Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page i — #

Fundamentals of Ship Hydrodynamics

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page i — #
j

Fundamentals of Ship Hydrodynamics

Fluid Mechanics, Ship Resistance and Propulsion

Lothar Birk
School of Naval Architecture and Marine Engineering
The University of New Orleans
New Orleans, LA
United States

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page ii — #
j

This edition first published 

©  John Wiley & Sons Ltd
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise,
except as permitted by law. Advice on how to obtain permission to reuse material from this title is available
at http://www.wiley.com/go/permissions.
The right of Lothar Birk to be identified as the author of this work has been asserted in accordance with law.

Registered Offices
John Wiley & Sons, Inc.,  River Street, Hoboken, NJ , USA
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO SQ, UK
Editorial Office
The Atrium, Southern Gate, Chichester, West Sussex, PO SQ, UK
For details of our global editorial offices, customer services, and more information about Wiley products
visit us at www.wiley.com.

Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that
appears in standard print versions of this book may not be available in other formats.

Limit of Liability/Disclaimer of Warranty

In view of ongoing research, equipment modifications, changes in governmental regulations, and the
constant flow of information relating to the use of experimental reagents, equipment, and devices, the reader
is urged to review and evaluate the information provided in the package insert or instructions for each
chemical, piece of equipment, reagent, or device for, among other things, any changes in the instructions or
indication of usage and for added warnings and precautions. While the publisher and authors have used
j their best efforts in preparing this work, they make no representations or warranties with respect to the j
accuracy or completeness of the contents of this work and specifically disclaim all warranties, including
without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty
may be created or extended by sales representatives, written sales materials or promotional statements for
this work. The fact that an organization, website, or product is referred to in this work as a citation and/or
potential source of further information does not mean that the publisher and authors endorse the
information or services the organization, website, or product may provide or recommendations it may make.
This work is sold with the understanding that the publisher is not engaged in rendering professional services.
The advice and strategies contained herein may not be suitable for your situation. You should consult with a
specialist where appropriate. Further, readers should be aware that websites listed in this work may have
changed or disappeared between when this work was written and when it is read. Neither the publisher nor
authors shall be liable for any loss of profit or any other commercial damages, including but not limited to
special, incidental, consequential, or other damages.

Library of Congress Cataloging-in-Publication Data

Names: Birk, Lothar, - author.
Title: Fundamentals of ship hydrodynamics : fluid mechanics, ship resistance and propulsion /
Lothar Birk, University of New Orleans.
Description: Hoboken, NJ : John Wiley & Sons, Ltd, [] | Includes bibliographical references
and index.
Identifiers: LCCN | ISBN  (hardcover) | ISBN  (epub)
Subjects: LCSH: Ships–Hydrodynamics.
Classification: LCC VM .B  | DDC ./–dc
LC record available at https://lccn.loc.gov/
Cover Design: Wiley
Cover Image: © zennie / Getty Images
Set in pt Warnock Pro Regular by Lothar Birk

Printed in Great Britain by TJ International Ltd, Padstow, Cornwall

         

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page iii — #
j

To My Family

They make everything worthwhile!

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page v — #
j

vii

Contents

List of Figures xvii

List of Tables xxvii
Preface xxxi
Acknowledgments xxxv
About the Companion Website xxxvii

1 Ship Hydrodynamics 
. Calm Water Hydrodynamics 
. Ship Hydrodynamics and Ship Design 
. Available Tools 

2 Ship Resistance 
. Total Resistance 
. Phenomenological Subdivision 
j . Practical Subdivision  j
.. Froude’s hypothesis 
.. ITTC’s method 
. Physical Subdivision 
.. Body forces 
.. Surface forces 
. Major Resistance Components 

3 Fluid and Flow Properties 

. A Word on Notation 
. Fluid Properties 
.. Properties of water 
.. Properties of air 
.. Acceleration of free fall 
. Modeling and Visualizing Flow 
. Pressure 

4 Fluid Mechanics and Calculus 

. Substantial Derivative 
. Nabla Operator and Its Applications 
.. Gradient 
.. Divergence 
.. Rotation 
.. Laplace operator 

5 Continuity Equation 
. Mathematical Models of Flow 
. Infinitesimal Fluid Element Fixed in Space 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page vi — #
j

viii Contents

. Finite Control Volume Fixed in Space 

. Infinitesimal Element Moving With the Fluid 
. Finite Control Volume Moving With the Fluid 
. Summary 

6 Navier-Stokes Equations 
. Momentum 
. Conservation of Momentum 
.. Time rate of change of momentum 
.. Momentum flux over boundary 
.. External forces 
.. Conservation of momentum equations 
. Stokes’ Hypothesis 
. Navier-Stokes Equations for a Newtonian Fluid 

7 Special Cases of the Navier-Stokes Equations 

. Incompressible Fluid of Constant Temperature 
. Dimensionless Navier-Stokes Equations 

8 Reynolds Averaged Navier-Stokes Equations (RANSE) 

. Mean and Turbulent Velocity 
. Time Averaged Continuity Equation 
. Time Averaged Navier-Stokes Equations 
j . Reynolds Stresses and Turbulence Modeling  j
9 Application of the Conservation Principles 
. Body in a Wind Tunnel 
. Submerged Vessel in an Unbounded Fluid 
.. Conservation of mass 
.. Conservation of momentum 

10 Boundary Layer Theory 

. Boundary Layer 
.. Boundary layer thickness 
.. Laminar and turbulent flow 
.. Flow separation 
. Simplifying Assumptions 
. Boundary Layer Equations 

11 Wall Shear Stress in the Boundary Layer 

. Control Volume Selection 
. Conservation of Mass in the Boundary Layer 
. Conservation of Momentum in the Boundary Layer 
.. Momentum flux over boundary of control volume 
.. Surface forces acting on control volume 
.. Displacement thickness 
.. Momentum thickness 
. Wall Shear Stress 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page vii — #
j

Contents ix

12 Boundary Layer of a Flat Plate 

. Boundary Layer Equations for a Flat Plate 
. Dimensionless Velocity Profiles 
. Boundary Layer Thickness 
. Wall Shear Stress 
. Displacement Thickness 
. Momentum Thickness 
. Friction Force and Coefficients 

13 Frictional Resistance 

. Turbulent Boundary Layers 
. Shear Stress in Turbulent Flow 
. Friction Coefficients for Turbulent Flow 
. Model–Ship Correlation Lines 
. Effect of Surface Roughness 
. Effect of Form 
. Estimating Frictional Resistance 

14 Inviscid Flow 

. Euler Equations for Incompressible Flow 
. Bernoulli Equation 
. Rotation, Vorticity, and Circulation 

j 15 Potential Flow  j

. Velocity Potential 
. Circulation and Velocity Potential 
. Laplace Equation 
. Bernoulli Equation for Potential Flow 

16 Basic Solutions of the Laplace Equation 

. Uniform Parallel Flow 
. Sources and Sinks 
. Vortex 
. Combinations of Singularities 
.. Rankine oval 
.. Dipole 
. Singularity Distributions 

17 Ideal Flow Around A Long Cylinder 

. Boundary Value Problem 
.. Moving cylinder in fluid at rest 
.. Cylinder at rest in parallel flow 
. Solution and Velocity Potential 
. Velocity and Pressure Field 
.. Velocity field 
.. Pressure field 
. D’Alembert’s Paradox 
. Added Mass 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page viii — #
j

x Contents

18 Viscous Pressure Resistance 

. Displacement Effect of Boundary Layer 
. Flow Separation 

19 Waves and Ship Wave Patterns 

. Wave Length, Period, and Height 
. Fundamental Observations 
. Kelvin Wave Pattern 

20 Wave Theory 

. Overview 
. Mathematical Model for Long-crested Waves 
.. Ocean bottom boundary condition 
.. Free surface boundary conditions 
.. Far field condition 
.. Nonlinear boundary value problem 
. Linearized Boundary Value Problem 

21 Linearization of Free Surface Boundary Conditions 

. Perturbation Approach 
. Kinematic Free Surface Condition 
. Dynamic Free Surface Condition 
. Linearized Free Surface Conditions for Waves 
j j
22 Linear Wave Theory 
. Solution of Linear Boundary Value Problem 
. Far Field Condition Revisited 
. Dispersion Relation 
. Deep Water Approximation 

23 Wave Properties 

. Linear Wave Theory Results 
. Wave Number 
. Water Particle Velocity and Acceleration 
. Dynamic Pressure 
. Water Particle Motions 

24 Wave Energy and Wave Propagation 

. Wave Propagation 
. Wave Energy 
.. Kinetic wave energy 
.. Potential wave energy 
.. Total wave energy density 
. Energy Transport and Group Velocity 

25 Ship Wave Resistance 

. Physics of Wave Resistance 
. Wave Superposition 
. Michell’s Integral 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page ix — #
j

Contents xi

. Panel Methods 

26 Ship Model Testing 

. Testing Facilities 
.. Towing tank 
.. Cavitation tunnel 
. Ship and Propeller Models 
.. Turbulence generation 
.. Loading condition 
.. Propeller models 
. Model Basins 

27 Dimensional Analysis 

. Purpose of Dimensional Analysis 
. Buckingham 𝜋-Theorem 
. Dimensional Analysis of Ship Resistance 

28 Laws of Similitude 

. Similarities 
.. Geometric similarity 
.. Kinematic similarity 
.. Dynamic similarity 
.. Summary 
j . Partial Dynamic Similarity  j
.. Hypothetical case: full dynamic similarity 
.. Real world: partial dynamic similarity 
.. Froude’s hypothesis revisited 

29 Resistance Test 

. Test Procedure 
. Reduction of Resistance Test Data 
. Form Factor 𝑘 
. Wave Resistance Coefficient 𝐶𝑊 
. Skin Friction Correction Force 𝐹𝐷 

30 Full Scale Resistance Prediction 

. Model Test Results 
. Corrections and Additional Resistance Components 
. Total Resistance and Effective Power 
. Example Resistance Prediction 

31 Resistance Estimates – Guldhammer and Harvald’s Method 

. Historical Development 
. Guldhammer and Harvald’s Method 
.. Applicability 
.. Required input 
.. Resistance estimate 
. Extended Resistance Estimate Example 
.. Completion of input parameters 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page x — #
j

xii Contents

.. Range of speeds 

.. Residuary resistance coefficient 
.. Frictional resistance coefficient 
.. Additional resistance coefficients 
.. Total resistance coefficient 
.. Total resistance and effective power 

32 Introduction to Ship Propulsion 

. Propulsion Task 
. Propulsion Systems 
.. Marine propeller 
.. Water jet propulsion 
.. Voith Schneider propeller (VSP) 
. Efficiencies in Ship Propulsion 

33 Momentum Theory of the Propeller 

. Thrust, Axial Momentum, and Mass Flow 
. Ideal Efficiency and Thrust Loading Coefficient 

34 Hull–Propeller Interaction 

. Wake Fraction 
. Thrust Deduction Fraction 
. Relative Rotative Efficiency 
j j
35 Propeller Geometry 
. Propeller Parts 
. Principal Propeller Characteristics 
. Other Geometric Propeller Characteristics 

36 Lifting Foils 

. Foil Geometry and Flow Patterns 
. Lift and Drag 
. Thin Foil Theory 
.. Thin foil boundary value problem 
.. Thin foil body boundary condition 
.. Decomposition of disturbance potential 

37 Thin Foil Theory – Displacement Flow 

. Boundary Value Problem 
. Pressure Distribution 
. Elliptical Thickness Distribution 

38 Thin Foil Theory – Lifting Flow 

. Lifting Foil Problem 
. Glauert’s Classical Solution 

39 Thin Foil Theory – Lifting Flow Properties 

. Lift Force and Lift Coefficient 
. Moment and Center of Effort 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xi — #
j

Contents xiii

. Ideal Angle of Attack 

. Parabolic Mean Line 

40 Lifting Wings 

. Effects of Limited Wingspan 
. Free and Bound Vorticity 
. Biot–Savart Law 
. Lifting Line Theory 

41 Open Water Test 

. Test Conditions 
. Propeller Models 
. Test Procedure 
. Data Reduction 

42 Full Scale Propeller Performance 

. Comparison of Model and Full Scale Propeller Forces 
. ITTC Full Scale Correction Procedure 

43 Propulsion Test 

. Testing Procedure 
. Data Reduction 
. Hull–Propeller Interaction Parameters 
.. Model wake fraction 
j j
.. Thrust deduction fraction 
.. Relative rotative efficiency 
.. Full scale hull–propeller interaction parameters 
. Load Variation Test 

44 ITTC 1978 Performance Prediction Method 

. Summary of Model Tests 
. Full Scale Power Prediction 
. Summary 
. Solving the Intersection Problem 
. Example 

45 Cavitation 
. Cavitation Phenomenon 
. Cavitation Inception 
. Locations and Types of Cavitation 
. Detrimental Effects of Cavitation 

46 Cavitation Prevention 

. Design Measures 
. Keller’s Formula 
. Burrill’s Cavitation Chart 
. Other Design Measures 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xii — #
j

xiv Contents

47 Propeller Series Data 

. Wageningen B-Series 
. Wageningen B-Series Polynomials 
. Other Propeller Series 

48 Propeller Design Process 

. Design Tasks and Input Preparation 
. Optimum Diameter Selection 
.. Propeller design task  
.. Propeller design task  
. Optimum Rate of Revolution Selection 
.. Propeller design task  
.. Propeller design task  
. Design Charts 
. Computational Tools 

49 Hull–Propeller Matching Examples 

. Optimum Rate of Revolution Problem 
.. Design constant 
.. Initial expanded area ratio 
.. First iteration 
.. Cavitation check for first iteration 
.. Second iteration 
j .. Final selection by interpolation  j
. Optimum Diameter Problem 
.. Design constant 
.. Initial expanded area ratio 
.. First iteration 
.. Cavitation check for first iteration 
.. Second iteration 
.. Final selection by interpolation 
.. Attainable speed check 

50 Holtrop and Mennen’s Method 

. Overview of the Method 
.. Applicability 
.. Required input 
. Procedure 
.. Resistance components 
.. Total resistance 
.. Hull–propeller interaction parameters 
. Example 
.. Completion of input parameters 
.. Resistance estimate 
.. Powering estimate 

51 Hollenbach’s Method 

. Overview of the method 
.. Applicability 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xiii — #
j

Contents xv

.. Required input 

. Resistance Estimate 
.. Frictional resistance coefficient 
.. Mean residuary resistance coefficient 
.. Minimum residuary resistance coefficient 
.. Residuary resistance coefficient 
.. Correlation allowance 
.. Appendage resistance 
.. Environmental resistance 
.. Total resistance 
. Hull–Propeller Interaction Parameters 
.. Relative rotative efficiency 
.. Thrust deduction fraction 
.. Wake fraction 
. Resistance and Propulsion Estimate Example 
.. Completion of input parameters 
.. Powering estimate 

Index 

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xv — #
j

xvii

List of Figures

. Ship sailing in its natural habitat 

. Self-propelled ship sailing in calm water with constant speed 
. Towed bare hull (no propeller or appendages) moving in calm water 
. Comparison of inflow conditions for a propeller operating in behind and
in open water condition 

. Comparison between Froude’s and ITTC’s current method of derivation

for the residuary resistance coefficient 𝐶𝑅 and wave resistance coefficient
𝐶𝑊 
. Viscosity of the fluid has significant effect on the flow within the boundary
layer around a ship hull 
. Results of a paint flow test. Photos courtesy of Dr. Alfred Kracht, Versuch-
sanstalt für Wasserbau und Schiffbau (VWS), Berlin, Germany 
. Resistance coefficients and resistance for a container ship as functions of
j the Froude number (velocity)  j
. Comparison of absolute and relative size of resistance components for
three different displacement type vessels at design speed. The data are
taken from Larsson and Raven (, pages ,) 

. Fresh and seawater properties as a function of temperature 

. The pressure force d𝐹 𝑝 acting on a small
surface element 
. Forces on a small cube in hydrostatic equilibrum 
. Hydrostatic pressure in a water column 
. Pressure distribution around a ship 

. Following a fluid particle and the flow properties it encounters along the
way 
. A moving, finite control volume 𝑉 which changes over time 
. The distance 𝑠𝑛 traveled by a surface element in normal direction 

. Four types of mathematical models for fluid flows and the resulting form
of the conservation law 
. Mass flux through the surface of a fluid element 
. Flux through the surface 𝑆 of a finite volume 𝑉 fixed in space 
. Flow through a contraction nozzle 

. Momentum flux in 𝑥-direction through the surface of an infinitesimal,

fixed fluid element d𝑉 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xvi — #
j

xviii List of Figures

. 𝑥-components of surface and body forces acting on the fixed, infinitesimal
fluid element d𝑉 
. Forces comprising the Navier-Stokes equations for an isotropic Newtonian
fluid 

. Mean and actual velocities in steady and unsteady turbulent flow 
. Velocity and turbulence distribution across an air duct 

. Body of revolution in a wind tunnel (simplified) 

. Ellipsoid moving in an unbounded fluid 

. Basic properties of the velocity distribution in the boundary layer 
. Transition from laminar to turbulent flow of the air rising from a burning
candle. Reproduced with kind permission by Dr. Gary S. Settles, Floviz,
Inc. 
. Flow characteristics of laminar and turbulent boundary layers 
. Development of the boundary layer along a flat surface. Note that the
outer limit of the boundary layer is not a streamline 
. Development of velocity profile in the boundary layer along a curved
surface with flow separation 

. Cross section through a finite, fixed control volume 𝑉 in the boundary
layer 
. Surface forces acting on the control volume 
j j
. Definition of displacement thickness 𝛿1 and displacement effect on exterior
flow 

. Laminar boundary layer along a flat plate 

. Boundary layer shear stress for laminar flow over a flat plate as dimension-
less position dependent skin friction coefficient 𝐶𝑓𝑥 
. Boundary layer thickness 𝛿, displacement thickness 𝛿1 , and momentum
thickness Θ for laminar flow over a flat plate 

. Features of a turbulent boundary layer over a flat plate (zero pressure
gradient) 
. A typical turbulent boundary layer velocity profile depicted in outer and
inner scaling 
. Comparing the modified log–wake law with experimental data from Öster-
lund () (Profile SWF) 
. Flat plate friction coefficients for smooth surfaces 
. Types of technical surface roughness and their effect on friction 
. Definition of equivalent sand roughness 𝑘𝑆 
. Flat plate friction coefficient for turbulent flow and its dependency on
Reynolds number and relative surface roughness 𝑘𝑆 ∕𝐿 

. A fluid element d𝑚 moves from point A to point B along a streamline 
. Determining the flow speed by measuring pressure difference in a contrac-
tion nozzle 
. Translation and linear deformation of a fluid element 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xvii — #
j

List of Figures xix

. Rotation and angular deformation of a fluid element 

. Definition of circulation Γ 
. Symmetric foil with lifting flow (Γ ≠ 0) and nonlifting flow (Γ = 0) 

. The work spent on moving an object from point A to point B 
. Definition of simply and multiply connected regions 
. Examples of basic potential flows 
. Flow field around a symmetric foil at angle of attack 𝛼 

. Planar uniform flow at angle 𝛼 

. Streamlines (𝜓 = const.) and isolines of velocity potential for a planar
source/sink flow. 𝑞 𝑇 = (𝜉, 𝜂) is the location of the source 
. Streamlines (𝜓 = const.) and equipotential lines for a planar source/sink
flow; the source/sink is located at (𝜉, 𝜂) = (2.5, 1.3)𝑇 
. Streamlines (𝜓 = const.) and equipotential lines for a planar vortex flow;
the vortex is located at 𝑞 = (2.5, 1.3)𝑇 
. Superposition of parallel flow and a source/sink pair 
. Flow field for a Rankine oval, a superposition of parallel flow, source, and
sink 
. Velocity and pressure distribution along the dividing streamline (Rankine
oval, 𝑈∞ = 1.0 m/s, 𝜎 = 2𝜋 m2 /s) 
. Creation of a dipole (doublet) by superposition of source and sink 
. Streamlines (𝜓 = const.) and isolines of velocity potential for planar dipole
j flows. Dipole is located at (𝜉, 𝜂) = (2.5, 1.3)𝑇  j

. An infinitely long cylinder moving with speed 𝑈∞ in positive 𝑥-direction

in a fluid at rest 
. An infinitely long cylinder at rest in parallel flow 
. Streamlines and velocity field for a cylinder in parallel flow 
. Contours of constant pressure coefficient 𝐶𝑝 for a cylinder in parallel
flow 
. Pressure coefficient 𝐶𝑝 distribution on the cylinder surface (𝑟 = 𝑅) for a
cylinder in parallel flow 

. The displacement effect of a boundary layer changes the effective hull
shape 
. The effect of viscous flow on the pressure distribution 
. Velocity profiles within the boundary layer near a separation point 
. Comparison of pressure and forces acting on a cylinder in inviscid and
viscous flow 
. Comparison of turbulent and laminar boundary layer flow around a cylin-
der 

. Definition of wave length 𝐿𝑤 and wave height 𝐻; the vertical scale is
exaggerated 
. Surface elevation of a harmonic, long-crested wave 
. Recording of surface elevation of a harmonic, long-crested wave at a fixed
position (𝑥 = 0) 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xviii — #
j

xx List of Figures

. Spatial extension of surface elevation of a linear, harmonic, long-crested

wave captured at time (𝑡 = 0) 
. A snapshot of the wave elevation in a wave group 
. Kelvin wave pattern in deep water 
. Change of Kelvin wave pattern with increasing velocity on deep water 
. Kelvin wave pattern like cloud formation in the slipstream of Amsterdam
Island in the southern Indian Ocean. Photo courtesy of NASA Earth
Observatory 
. Wave pattern of a ship at 𝐹𝑟 = 0.26 

. Definition of coordinate system and domain boundaries for wave theory
of long-crested waves 
. Simplified two-dimensional fluid domain for long-crested waves 
. The mathematical free surface model is valid for nonbreaking waves only 

. Simplified two-dimensional fluid domain for long-crested regular waves 
. The hyperbolic sine and cosine functions 
. Wave phase velocity as function of wave number and water depth based
on linear wave theory 
. The positive arm of the hyperbolic tangent function 

. Graphical verification of a solution of the nonlinear dispersion relation 

. Distribution of wave properties over . wavelength at the calm water level
j (𝑧 = 0) for a wave with wave period 𝑇 = 10 s  j
. Snapshot (𝑡 = 0) of the velocity field for a wave in restricted water depth 
. Amplitude of dynamic pressure over depth 𝑧 
. Photo of water particle trajectories. Photo courtesy of Dr. Walter L. Kuehn-
lein, seaice Ltd. & Co. KG, www.seaice.com 
. Water particle trajectories over one wave period 𝑇 for deep water (left)
and restricted water (right) 

. Propagation of wave profile and the movement of a water particle over one
wave period 
. Wave length 𝐿𝑤 as a function of water depth ℎ for constant wave period
𝑇 
. Effect of gradually decreasing water depth ℎ on wave propagation and
direction 
. Kinetic energy 𝐸kin in the control volume 𝑉 spanning one wave length 𝐿𝑤
in 𝑥-direction 
. Change in potential energy 𝐸pot when a fluid element is lifted above the
calm water level 
. Wave energy density distribution of a regular wave over one wave cycle
according to linear wave theory 
. Schematic propagation of wave energy for a deep water wave based on
linear wave theory 
. Wave elevation profiles after a few selected cycles of wave making (deep
water) 
. Distribution of energy density after a few selected cycles of wave making
(deep water) 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xix — #
j

List of Figures xxi

. Propagation of a group of regular waves over  wave periods 

. Wigley hull at Froude number 𝐹𝑟 = 0.26 showing the connection between
fluid and hull surface pressure and the resulting wave elevation. Light
colors indicate high pressure and high wave elevation. Dark colors indicate
low pressure and low values of wave elevation 
. Pronounced humps and hollows in a wave resistance curve. Data from
model tests with Wigley hulls (Bai and McCarthy, ) 
. Wave resistance coefficient of a single submerged sphere; see Equations (.)
and (.) 
. Wave pattern and wave profile created by a single submerged sphere at
position 𝑥∕𝐿 = +0.5 (forward). The sphere’s dimensionless speed is the
Froude number 𝐹𝑟 = 0.252, which is based on 𝐿 = 10𝐷 
. Wave pattern and wave profile created by a single submerged sphere at
position 𝑥∕𝐿 = −0.5 (aft). The sphere’s dimensionless speed is the Froude
number 𝐹𝑟 = 0.252 
. Combined wave pattern and profile of two submerged spheres. Froude
number 𝐹𝑟 = 0.252; favorable superposition of waves resulting in low
wave heights 
. Comparison of wave profiles created by submerged spheres at positions
𝑥∕𝐿 = ±0.5. The spheres’ dimensionless speed is the Froude number
𝐹𝑟 = 0.252 
. Wave pattern and wave profile of two submerged spheres. Froude number
j 𝐹𝑟 = 0.282; unfavorable superposition of waves, resulting in high wave j
heights 
. Comparison of wave profiles created by submerged spheres at positions
𝑥∕𝐿 = ±0.5. The spheres’ dimensionless speed is Froude number 𝐹𝑟 =
0.282 
. Comparison of wave patterns and wave profiles created by two submerged
spheres for Froude numbers 𝐹𝑟 = 0.252 (upper half, favorable superposi-
tion) and 𝐹𝑟 = 0.282 (lower half, unfavorable superposition) 
. Wave resistance coefficient for a system of two submerged spheres; dis-
tance between centers is 𝐿 = 10𝐷, submergence is 𝑠 = 𝐷 
. Wave resistance coefficient for a Wigley hull with length–beam ratio of
𝐿∕𝐵 = 10 and beam–draft ratio of 𝐵∕𝑇 = 1.6 
. Discretization of vessel bow into small panels for wave resistance compu-
tation 

. Towing tank of the Hamburg Ship Model Basin, Photo courtesy of Ham-
burgische Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.de 
. Towing tank at the School of Naval Architecture and Marine Engineering
of the University of New Orleans 
. Schematic view of a towing tank 
. Definition of rail sagitta 𝑠 
. Schematic of a cavitation tunnel without free surface 
. A model is prepared for cutting on the five axis mill. Photo courtesy
of Hamburgische Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.
de 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xx — #
j

xxii List of Figures

. The beginnings of a five bladed propeller model. Photo courtesy of Ham-
burgische Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.de 

. Simplified flow pattern at a propeller blade section. Kinematic similarity

requires that the angle of attack remains the same for full scale and model
propeller blade section 
. Relationship of total static pressures for model and full scale ship 

. Ship model set up for resistance test 

. Measured and derived data in the resistance test 
. Tank cross section area 𝐴 and blockage factor 𝑚 
. Measurements and length definitions for the computation of sinkage and
trim (sinkage and trim are exaggerated) 
. Method of Prohaska to determine the form factor 𝑘 

. Finding the form factor with Prohaska’s method; only data points with
0.1 ≤ 𝐹𝑟 ≤ 0.2 are used 
. Measured mean sinkage and running trim angle of model 
. Measured total resistance of model as function of model speed 
. Resistance coefficients of model 
. Resistance coefficients of full scale ship 
. Full scale total resistance prediction (calm water) 
. Full scale effective power 
j j
. Definition of the midship section and the computational length 𝐿 for Guld-
hammer and Harvald’s resistance estimate (Andersen and Guldhammer,
) 
. Guidance for the optimum location of 𝐿𝐶𝐵 as a function of Froude number
𝐹𝑟. Here, negative 𝐿𝐶𝐵 values indicate a location aft of midship 
. Resistance coefficients for the Guldhammer and Harvald method exam-
ple 
. Total resistance and effective power for the Guldhammer and Harvald
method example 
. Charts for standard residuary resistance coefficients 𝐶𝑅std after Guldham-
mer and Harvald () for vessels with length-speed ratio 𝐿∕𝑉 1∕3 = 6.0.
The values have been computed and redrawn based on the regression
formula provided by Andersen and Guldhammer () 
. Charts for standard residuary resistance coefficients 𝐶𝑅std after Guldham-
mer and Harvald () for vessels with length-speed ratio 𝐿∕𝑉 1∕3 = 6.5.
The values have been computed and redrawn based on the regression
formula provided by Andersen and Guldhammer () 

. Forces acting on ship without and with propulsion system 
. A five-bladed fixed pitch propeller with a Schneekluth nozzle to improve
propeller inflow 
. Schematic of a water jet 
. The propulsion system with transmission powers and efficiencies 

. Fixed control volume around an idealized propeller (actuator disk) 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxi — #
j

List of Figures xxiii

. Velocity and pressure distribution according to momentum theory 

. Ideal efficiency (jet efficiency) of propulsor momentum theory as a function
of thrust loading coefficient 

. Example of a nominal wake field of a single propeller ship with moderate
block coefficient 
. Example of a nominal wake field of a single propeller ship with high block
coefficient 
. Major contributions to the nominal wake fraction 
. Frictional wake for twin screw vessels 
. Effect of propeller on pressure and velocity distribution at the stern 

. Parts of a propeller. Shown is a right-handed, fixed pitch propeller with

four blades 
. Definition of propeller diameter 𝐷, blade radius 𝑅, hub radius 𝑟ℎ , and disk
area 𝐴0 
. Hydrofoil section within a propeller blade 
. Pitch angle variation for a propeller with constant pitch 𝑃 
. Helical paths for propeller with constant pitch 𝑃 
. Relationship between pitch 𝑃 and pitch angle 𝜙 
. Helical paths for propeller with variable pitch 𝑃 
. Basic geometric properties of a lifting hydrofoil 
. The expanded blade for a Wageningen B-Series propeller with four blades
j and 𝐴𝐸 ∕𝐴0 = 0.85. Drawing is based on data from Oosterveld and van j
Oossanen () 
. Definition of the expanded area 𝐴𝐸 ∕𝑍 of a propeller blade 
. Two examples of expanded area ratios. 𝐴𝐸 ∕𝐴0 = 0.55 on the left and
𝐴𝐸 ∕𝐴0 = 0.85 on the right 
. Side elevation of a propeller blade and definition of propeller rake 𝑖𝐺 (𝑥)
and rake angle 𝜃 
. Definition of propeller skew-back and skew angle 𝜃𝑠 
. Cupping of a propeller blade 

. Definition of foil geometry 

. Flow pattern and pressure distribution for a D foil section at angle of
attack 𝛼 
. Foil in stalled flow condition 
. Complete vortex system of the foil section 
. Forces acting on the foil section at angle of attack 𝛼 
. Typical lift–drag curves for a thin, cambered foil 
. Setup of boundary value problem for a thin foil operating at angle of
attack 𝛼 
. Definition of normal vectors for upper and lower foil surface 

. The boundary value problem of a symmetric thin foil with finite thickness
and zero angle of attack 
. The inverse tangent function 
. The source strength distribution 𝜎(𝜉) as a function of the slope d𝑡∕d𝑥 of
the foil surface 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxii — #
j

xxiv List of Figures

. Comparison of thin foil theory and conformal mapping (exact) pressure
coefficient for an elliptical foil with thickness to chord length ratio of
𝑡max ∕𝑐 = 0.10 
. Comparison of thin foil theory and conformal mapping (exact) pressure
coefficient for an elliptical foil with thickness to chord length ratio of
𝑡max ∕𝑐 = 0.05 

. Boundary value problem for an infinitely thin cambered plate at angle of
attack 𝛼 
. The first four elements of Glauert’s trigonometric series for the vortex
strength 𝛾 

. The effect of leading edge suction for a thin plate at angle of attack 𝛼 
. Section lift coefficient 𝐶𝑙 of thin, symmetric foil sections and a thin, cam-
bered foil section with zero lift angle 𝛼0 = −4 degree 
. Definition of the moment 𝑀𝑧 created by the pressure force acting on a
thin foil 
. Pressure distribution for a flat plate at  degrees angle of attack 
. A prescribed pressure distribution resulting in the NACA 𝑎 = 0.8 mean
line 

. Pressure distribution on a wing of finite span 𝑠 

. Simplest model of the vortex system of a wing 
j . Cross section through the velocity field of the wing tip vortices revealing j
downwash and updraft 
. A model of a wing with varying bound circulation Γ𝑏 (𝜂) and the resulting
free vortex sheet 
. Actual shape and roll up of trailing vortex sheet 
. Velocity vectors on upper and lower wing surfaces 
. Velocity vectors on upper and lower wing surface and mean velocity
𝑣𝑚 
. Vorticity vector 𝛾 and the resulting difference velocity 𝑣𝑑 
. Bound vorticity 𝛾 and its effect 
𝑏
. Free vorticity 𝛾 and its effect 
𝑓
. Establishing a connection between bound and free vorticity by integration
over the boundary of a simply connected region 
. The velocity induced by an element of a vortex filament 
. The velocity induced by a straight vortex filament 
. The effect of downwash on the angle of attack 
. Induced drag caused by the downwash 

. Simplified velocities triangle for an unrolled propeller blade section at

radius 𝑥 = 0.75 (induced velocities have been ignored) 
. Open water test of model propeller with propeller boat in towing tank 
. Open water test of model propeller in a circulating water tunnel / cavitation
tunnel 
. Typical propeller open water diagram 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxiii — #
j

List of Figures xxv

. Fluid forces acting on a propeller blade section at model scale 
. Comparison of model scale and full scale forces acting on a propeller blade
section 
. Comparison of measured open water characteristics and predicted full
scale propeller performance 

. Setup of model for propulsion test with single skin friction correction force
(continental method) 
. The relative difference of resistance for model and full scale vessel 
. Self propulsion point of model propeller under the assumption of thrust
identity 
. Self propulsion point of model propeller under the assumption of torque
identity 
. Setup of model for propulsion test with load variation (British method) 
. Typical results of a load variation test (British method) 

. Matching propeller thrust 𝐾𝑇𝑆 𝑂 with the thrust requirement of the ship
assuming thrust identity 
. Setup for solving the intersection problem with discrete open water data 

. Simplified phase diagram of fresh water 

. Flow around a cavitating foil section and the associated pressure distribu-
tion for back (upper) and face (lower) side 
j . Locations and common types of cavitation  j
. Open water test of five-bladed model propeller in a cavitation tunnel; tip
vortex and bubble cavitation 
. Open water test of five-bladed model propeller in a cavitation tunnel;
propeller blades completely covered in sheet cavitation 
. Loss of thrust and efficiency due to cavitation 
. Life of cavitation bubble 

. Limits for the propeller loading coefficient as a function of cavitation num-
ber and acceptable level of cavitation. After Burrill and Emerson (),
however, the curves represent the regression equations from Table .

. Usage of the Burrill chart 

. Open water chart for a Wageningen B-Series propeller with 𝑍 = 4 and
𝐴𝐸 ∕𝐴0 = 0.85 derived from 𝐾𝑇 and 𝐾𝑄 polynomials Equations (.) and
(.) 

. Design task  – Input: open water diagram for Wageningen B-series pro-
pellers with 𝑍 = 4 and 𝐴𝐸 ∕𝐴0 = 0.85 derived from 𝐾𝑇 and 𝐾𝑄 polynomi-
als (.) and (.). Torque coefficient curves 10𝐾𝑄 are emphasized 
. Design task  – Step : locate self propulsion points ◦ at which the pro-
pellers absorb the delivered power specified with the design constant
[𝐾𝑄 ∕𝐽 5 ] from Equation (.) 
. Design task  – Step : find open water efficiencies × for self propulsion
points ◦ 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxiv — #
j

xxvi List of Figures

. Design task  – Step : draw auxiliary curve through open water efficiency
values 
. Design task  – Result: optimum propeller is defined by maximum of
auxiliary curve 
. Design task  – Result: optimum propeller is defined by maximum of
auxiliary curve 
. Design task  – Result: optimum propeller is defined by maximum of
auxiliary curve 
. Design task  – Result: optimum propeller is defined by maximum of
auxiliary curve 
. Simplified task  propeller design 𝐵𝑃1 -chart for Wageningen B-Series pro-
peller with 𝑍 = 4 and 𝐴𝐸 ∕𝐴𝑜 = 0.85 
. Optimum diameter chart for design task  

. Propeller design 𝐵𝑢2 -chart for Wageningen B-series propeller B-. Cal-
culated and plotted based on the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and
van Oossanen () 
. Propeller design 𝐵𝑢2 -chart for Wageningen B-series propeller B-. Cal-
culated and plotted based on the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and
van Oossanen () 
. Auxiliary plot to determine the expanded area ratio of the final optimum
propeller 
. Auxiliary plots to determine final optimum propeller characteristics. The
j ⋆ mark results of first and second iterations  j
. Propeller design 𝐵𝑝1 -chart for Wageningen B-series propeller B-. Cal-
culated and plotted based on the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and
van Oossanen () 
. Propeller design 𝐵𝑝1 -chart for Wageningen B-series propeller B-. Cal-
culated and plotted based on the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and
van Oossanen () 
. Auxiliary plot to determine the expanded blade area ratio of the final
optimum propeller 
. Auxiliary plots to determine final optimum propeller characteristics. The
⋆ mark results of first and second iterations 
. Auxiliary plot to determine the attainable ship speed 

. Comparison of total resistance estimates for the methods by Hollenbach,

Guldhammer and Harvald, and Holtrop and Mennen 
. Comparison of predicted rate of revolution and delivered power for the
methods by Hollenbach, Guldhammer and Harvald, and Holtrop and
Mennen 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxv — #
j

xxvii

List of Tables

. Fresh water properties 

. A selection of model basins around the world, sorted alphabetically ac-
cording to their commonly used abbreviations 

. Particulars of full scale vessel and model used in the prediction example

. Testing and full scale environments for resistance prediction 
. Measured total resistance and sinkage of model; blockage correction
(Schuster, ), dynamic sinkage, and trim 
. Resistance coefficients for the model (𝑘 = 0.1566) 
. Predicted resistance coefficients for the full scale vessel 
. Full scale resistance 𝑅𝑇𝑆 and effective power 𝑃𝐸𝑆 

j . Range of parameters suitable for Guldhammer and Harvald’s method  j
. Required and optional input parameters for Guldhammer and Harvald’s
method 
. Bulbous bow corrections to the standard residuary resistance coefficient
(Andersen and Guldhammer, ) 
. Air resistance coefficients for different types of vessels (Kristensen and
Lützen, ) 
. Principal dimensions for resistance estimate example 
. Selected ship speeds and resulting Froude and Reynolds number 
. Residuary resistance value computation for the standard hull form 
. Computation of the 𝐿𝐶𝐵-correction for the residuary resistance coeffi-
cient 
. Comparison of old and updated bulbous bow correction to the residuary
resistance coefficient 
. Estimate of residual resistance coefficient 
. Frictional resistance estimate 
. Resistance coefficients computed with Guldhammer and Harvald’s method
using input from Table . 
. Total resistance and effective power computed with Guldhammer and
Harvald’s method using input from Table . 

. The three subtasks of thin foil theory 

. Example results of an open water test conducted in a towing tank 
. Open water characteristics of model propeller (see Table .) 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxvi — #
j

xxviii List of Tables

. Intermediate results for scaling open water characteristics of model pro-
peller (see Table .) 
. Predicted open water characteristics of full scale propeller (see Table .) 

. Propeller open water characteristics as a set of discrete data points 
. Example input data for the calculation of the self propulsion point for a
single ship speed 
. Data for required thrust parabola at 𝑣 = 11.472 m/s and 𝐶𝑆 = 0.43372
and propeller open water thrust coefficient 

. Regression equations for the limiting lines in the Burrill chart (Figure .) 

. Basic characteristics of the propellers in the Wageningen B-Series. For

each combination of 𝑍 and 𝐴𝐸 ∕𝐴0 propellers with 𝑃 ∕𝐷 = 0.5, 0.6, 0.8,
1.0, 1.2, and 1.4 have been tested 
. Factors and exponents for thrust coefficient polynomials of Wageningen
B-Series propellers (Oosterveld and van Oossanen, ) 
. Factors and exponents for torque coefficient polynomials of Wageningen
B-Series propellers (Oosterveld and van Oossanen, ) 
. Coefficients for the estimate of maximum thickness and chord length of
Wageningen B-Series propellers (Kuiper, ). For convenience values
have been added at radii 𝑥 = 0.15, 0.25, and 0.75 by interpolation 
. Factors and exponents for Reynolds number effects on thrust coefficient
j and torque coefficient of Wageningen B-Series propellers (Oosterveld and j
van Oossanen, ) 

. The four basic propeller design tasks 

. Input data to illustrate task : optimum propeller diameter selection based
on delivered power, speed of advance, and rate of revolution 
. Results for the propeller selection task  example 
. Results for the propeller selection task  example 

. Optimum rate of revolution problem – example input data for a container
ship 
. Optimum diameter problem – example input data for a bulk carrier 
. Resistance estimate for bulk carrier from Table . 

. Required and optional input parameters for Holtrop and Mennen’s method 
. Approximate values for appendage form factors 𝑘2𝑖 according to Holtrop
() 
. Coefficients for the wave resistance computation in Equation (.) if
𝐹𝑟 ≤ 0.4 (Holtrop, ) 
. Additional coefficients for the wave resistance computation in Equation (.)
if 𝐹𝑟 > 0.55 (Holtrop, ) 
. Coefficients for the full scale wake fraction of single screw vessels in Equa-
tion (.) (Holtrop, ) 
. Propeller data for powering estimate; see also Table . 
. Holtrop and Mennen resistance and powering estimate example; speed
independent procedural coefficients 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxvii — #
j

List of Tables xxix

. Holtrop and Mennen resistance and powering estimate example; speed
dependent procedural coefficients 
. Holtrop and Mennen resistance and powering estimate example; resistance
components and total resistance 
. Holtrop and Mennen resistance and powering estimate example; wake
fraction and self propulsion point analysis 
. Holtrop and Mennen resistance and powering estimate example; efficien-
cies, propeller rate of revolution, and delivered power 

. Recommended limits for principal dimensions and form parameters of

single screw vessels on design draft 
. Required and optional input parameters for Hollenbach’s method 
. Coefficients for an estimate of the wetted surface 
. Coefficients for computation of the standard residuary resistance coeffi-
cient in Hollenbach’s method 
. Coefficients for correction factors of the standard residuary resistance
coefficient in Hollenbach’s method 
. Factors for lower and upper limit formulas of the range of Froude numbers
in which the 𝐶𝑅 formulas are valid 
. Suggested values for the relative rotative efficiency 𝜂𝑅 , if the propulsion
estimate is based on Wageningen B-Series propeller data (Hollenbach,
) 
. Suggested values for the thrust deduction fraction 𝑡 (Hollenbach, ) 
j . Suggested values for constant 𝐶 for the hull efficiency of twin screw vessel j
models (Hollenbach, ) 
. Residuary resistance coefficients for minimum and mean resistance cases 
. Resistance coefficients for example vessel by Hollenbach’s method 
. Comparison of total calm water resistance estimates 
. Prediction of model and full scale wake fraction 
. Self propulsion point based on mean resistance curve 
. Prediction of rate of revolution and delivered power for trial condition
based on mean resistance curve 
. Predicted efficiencies based on mean resistance curve 
. Comparison of predicted rate of revolution and delivered power 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xv — #
j

xxxi

Preface

This book has been designed as a textbook to support ship resistance and propulsion Target
related courses at the undergraduate level. As such, its main audience is naval architec- readership
ture and marine engineering students and students in related fields. However, since
the book covers topics in fairly great detail, it is suited for self study for everybody
with a working knowledge of calculus, statics, and dynamics. Graduate students and
practicing engineers, who venture from other engineering disciplines into maritime
fields of study or work, might use this book as a preparation for new tasks.
Over the past  years, I have taught ship resistance and propulsion at three different Identified needs
universities and consistently made the following observations:

• The foundation laid by basic fluid mechanics courses in modern engineering

curricula is incomplete and does not cover everything essential to courses focused
on ship resistance and propulsion.

j • A wealth of excellent reference books exists covering all aspects of ship hydro- j
dynamics. However, no matter how strongly I recommend one of them, most
students find them too expensive or too intimidating and do not use them as
study aids.
• Arguably, most reference books are not organized in a way which lends itself to
support class work. The chapters are designed so broadly that it becomes difficult
to assign specific parts to individual class periods.

In many engineering curricula in the United States, basic fluid mechanics is covered in Interior vs.
a single course. This is just enough to cover hydrostatics and the basic equations of fluid exterior flow
dynamics but leaves hardly any room for boundary layer theory, potential flow, wave
theory, and foil and wing theory. In addition, teaching fluid mechanics courses is often
the responsibility of mechanical engineering departments, which naturally concentrate
on pipe flows and turbo machinery rather than exterior flows.
Authors of reference books assume more prerequisite knowledge than a typical under- Details vs.
graduate student of today actually has. After all, their target audience are practicing coverage
engineers. In addition, reference books attempt to be comprehensive and cover a broad
range of topics and tend to omit a lot of detail. The gaps may be easily filled by an
expert but often pose a seemingly insurmountable obstacle for students trying to un-
derstand the origins of a theory or fathom exactly how a certain method works. As a
consequence, I find myself compelled to explain to students what reference books cover
with statements like ‘as one can easily see’. Unfortunately, covering extensive details in
class distracts students from important assumptions and conclusions.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xvi — #
j

xxxii Preface

Objectives Based on these observations, I set out to write a textbook which meets the following
objectives:

• Whereas most reference books on ship hydrodynamics spend just a couple of

pages to summarize results of fluid mechanics, this textbook dedicates consider-
able space to it. However, it is not meant as a substitution, but rather a logical
continuation of a basic fluids mechanics course. The material is presented with
its application to ship resistance and propulsion in mind.
• Instead of covering all possible aspects of ship hydrodynamics, a selection of
topics is covered in greater detail. Sentences like ‘as one can easily see’ or ‘after
some manipulations’ are kept to a minimum. The detailed coverage allows the
teacher to concentrate on important assumptions and conclusions. Students will
find and study the details in the associated sections.
• Each chapter covers material for one, or sometimes two, class periods, which
should simplify reading assignments. As a consequence, the book is organized
into an unusually large number of chapters. Margin notes are used as an additional
organizational aid. There is a continuous thread throughout the book, but the
chapters are relatively independent from each other. This should make it easier
to skip some of them, assign them as extra reading, or rearrange their order
according to the needs of a specific course.

Content The junior level ship resistance and propulsion course serves a dual purpose in our naval
j j
overview architecture and marine engineering curriculum at the University of New Orleans. On
one hand, it identifies and explains basic flow patterns around a ship sailing at constant
speed. On the other hand, it prepares students to conduct basic ship design tasks like
resistance and powering estimates. Starting with basic fluid mechanics and ending with
powering estimates spans a wide arc. The only way to keep the page count in check was
to concentrate on the immediate topics at hand rather than venturing into all variations
and alternatives. The reader will notice that the book focuses on displacement type
monohulls driven by marine propellers. As a consequence, multihulls, planing boats,
and other propulsion systems are not covered. Fundamental analytical and experimental
methods are discussed but not computational fluid dynamics.
Organization The book is subdivided into  chapters organized into three parts: basic fluid mechanics,
ship resistance, and propulsion. However, the boundaries are blurred as I attempt to
connect basic theory with its application in ship hydrodynamics wherever possible. The
first chapter specifies the calm water resistance and propulsion problem. The second
chapter defines ship resistance and its major components. In Chapters  through  we
develop important equations describing viscous flow around submerged bodies and use
them to assess the frictional resistance of a ship. Chapters  through  analyze inviscid
flow and combine it with viscous flow theory to explain viscous pressure resistance.
Chapters  through  tackle wave theory and wave resistance.
Chapters  through  explain the concepts and theories which govern ship model
testing and the prediction of full scale resistance. Chapter  provides a first look at
resistance estimates for ship design purposes.
Chapter  marks the beginning of the ship propulsion part. Basic terminology, propul-
sor action, hull–propeller interaction, and propeller geometry are illustrated in Chap-

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xvii — #
j

Preface xxxiii

ters  through . Chapters  through  cover the basic flow theory for lifting
foils and wings. Chapters  through  deal again with model testing and discuss
experiments with model propellers and self-propelled ship models.
Chapters  through  address the problem of cavitation, cavitation avoidance, and
how to select a propeller for a specific ship. Finally, Chapters  and  describe in
detail two methods to estimate resistance and powering requirements in early design
stages.
Symbols are typically explained when they are introduced. A conscious effort has Nomenclature
been made to use the terminology and symbols according to the Dictionary of Hydro-
dynamics and the ITTC Symbol and Terminology List maintained and published by
the International Towing Tank Conference (ITTC). Both documents are part of the
quality systems manual and can be found on the ITTC’s website at www.ittc.info (ITTC,
a,b).
In most cases a Cartesian coordinate system < 𝑥, 𝑦, 𝑧 > is employed with its positive Cartesian
𝑥-axis pointing forward (in the direction of motion), its 𝑦-axis pointing to port, and its coordinate
system
𝑧-axis pointing upwards.
A textbook is always a conglomerate of the combined knowledge and wisdom of all Summary
who have worked in the specific field. All the presented work has originally been
developed by others and I have made every effort to point the reader to the correct
sources. My job has been to illustrate and explain everything, and as such the errors
are all mine. If you find any errors, please feel free to point them out to me via e-mail at
j lothar.birk@marine-hydrodynamics.com. j
Slidell, December 
Lothar Birk

References
ITTC (a). Dictionary of hydrodynamics – Alphabetic. International Towing Tank
Conference, Quality Systems Group.
ITTC (b). ITTC symbols and terminology list – Alphabetic. International Towing
Tank Conference, Quality Systems Group.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xix — #
j

xxxv

Acknowledgments

With the encouragement of my students, who told me on several occasions I should

convert my class notes into a book, I started writing this textbook roughly five years
ago. When Paul Petralia from Wiley & Sons, Ltd contacted me, I had the book proposal
ready to go and luckily the publishing house accepted my concept. I am grateful for the
opportunity and their support.
I would like to thank my colleagues Dr. Janou Hennig, Dr. Alfred Kracht, and Dr. Walter
Kühnlein for providing photos for the book. Many thanks also to Dr. Settles for his
photo of the laminar–turbulent transition in the airflow above a candle.
Writing a book is a milestone in a professional career. It provides time to pause and
reflect on how one got to this point. It is obvious to me that I had great teachers in
high school (which is called Gymnasium in Germany) and throughout my studies at
Technische Universität Berlin. Thank you, Mr. E. Jäckle, Mr. H. Riekert, Professor G.F.
Clauss, Professor H. Nowacki, and Professor E. Wolf, who all sparked my curiosity and
j inspired my desire to learn more. I hope all readers find great teachers and mentors j
like I did.
Last but not least, I want to thank my children Benjamin and Kathleen, who both
helped to correct their dad’s basic English, and of course my love and wife Carola, who
encouraged me all the way and patiently endured my prolonged occupation with this
book project.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page xxi — #
j

xxxvii

About the Companion Website

The companion website for this book is at

www.wiley.com/go/birk/hydrodynamics

The website includes:

• Python scripts
• Figures

Scan this QR code to visit the companion website.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

1
Ship Hydrodynamics

The field of ship hydrodynamics considers the interaction of vessels with surrounding
fluids. As the prefix ‘hydro’ suggests, we are most concerned with water; however, the
air flow around the super structure has to be dealt with as well. In this chapter, we
narrow down this broad field and define calm water hydrodynamics as the context of
this book. We will also discuss the role and responsibilities of the naval architect in
the analysis of ship hydrodynamics and – in broad terms – what tools we have at our
disposal to solve hydrodynamic tasks in ship design.

Learning Objectives
j j
At the end of this chapter students will be able to

• review the complexity of the ship propulsion problem

• explain the concepts of calm water, trial, service, and open water condition
• understand the role of ship resistance and propulsion in ship design
• distinguish basic tools to predict and investigate hydrodynamic performance

1.1 Calm Water Hydrodynamics

Boats and ships crossing any body of water have to negotiate the environment presented The real world
by wind, waves, currents, and the boundaries of their domain. Figure . shows the
major interactions between ship and environment.

• Wind blowing over water creates a seemingly chaotic pattern of waves through
which the ship sets its course. Wind and waves may come from different directions
relative to the ship’s path. A ship will change the shape of waves in its vicinity.
This is called wave diffraction. Wind and waves exert forces on the vessel which
vary with time.
• Wind and wave forces cause the ship to move. This movement creates additional
waves, similar to the waves created by a stone dropped into a pond. This is known
as wave radiation.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2 1 Ship Hydrodynamics

Figure 1.1 Ship sailing in its natural habitat

• Currents may cause the general flow direction to be oblique to the ship’s path.

j • A ship, even when moving in undisturbed water, creates a well organized wave j
pattern. It appears in a triangular region behind the ship and consists of divergent
and transverse waves.

• Some of the waves created by the vessel will break. A mix of water and air creates
a band of froth along the ship’s path.

• Small and large eddies appear next to and behind the ship. They are the result
of friction between hull and water. A boundary layer forms over the submerged
hull surface and merges into a disturbed flow region behind the ship generally
known as wake.

• A rotating propeller will generate its own twisted wake, further complicating the
flow patterns.

Unsteady flow A moving vessel constantly displaces water and air molecules from their original po-
sition. As stated in Newton’s laws of motion, forces (Latin: actio) must act on the
molecules to change magnitude or direction of their velocities. In turn, a reaction force
(Latin: reactio) is exerted by water and air on the ship. Vessels are usually self-propelled,
i.e. they have some means of propulsion. The most common propulsor today is the
marine propeller. Sails, oars, paddle wheels, and water jets may be applied depending
on purpose, size, and speed of the vessel. Propulsors create the force necessary to
overcome the reaction force by water and air. Although power settings of the engine
turning the propeller are kept constant, the speed of a vessel will still vary because
 Sir Isaac Newton (* – †), famous English mathematician, astronomer, and physicist. The

* marks the year of birth and the † marks the year of death

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

1.1 Calm Water Hydrodynamics 3

Figure 1.2 Self-propelled ship sailing in calm water with constant speed

waves and wind will alter the reaction force. In summary, the flow around a ship hull is
time dependent and, as a consequence, the flow will be unsteady.
In order to reduce the number of variables which influence the flow, it is often worth- Calm water
j while to study phenomena separately. In this book, we will eliminate the time varying condition j
components of the flow and ignore for the most part effects of wind, waves, and currents.

• We consider only the force necessary to move the ship through air and water
which are initially at rest.

• Only the waves generated by the ship itself are considered.

This is known as calm water condition (Figure .).

A ship trial is conducted before final delivery of a new vessel. Performance is measured Trial condition
and compared to the contracted requirements. Trials are conducted in deep, open
waters since the vicinity of the sea bottom or shore lines has a negative impact on ship
performance. At the time of a trial, the hull is freshly painted and free of any marine
growth. Together with the calm water condition this is referred to as trial condition.
The contract between owner and shipbuilder typically specifies a combination of engine Service condition
power and ship speed which has to be attained on the ship’s trial. However, the naval
architect must optimize the vessel for its intended service. A ship will encounter
currents, wind, and waves during its voyages. Marine fouling over time will increase
the roughness of any hull surface. These are the actual service conditions. As a result,
resistance will be higher in service than at the ship’s trial. Additional resistance has to
be considered during selection of engine and propeller, otherwise the ship will not be
able to perform as intended.
Application of calm water or trial conditions leaves us with a greatly simplified scenario
depicted in Figure .. The ship’s velocity is assumed to be constant in direction and

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

4 1 Ship Hydrodynamics

Figure 1.3 Towed bare hull (no propeller or appendages) moving in calm water

magnitude 𝑣𝑠 = const. Then all forces acting on the ship must be in equilibrium
according to Newton’s first law. The propulsor provides exactly the force 𝐹 𝑃 necessary
to compensate the force 𝐹 𝑅 exerted by water and air on the ship.

𝐹𝑃 + 𝐹𝑅 = 0 (.)
j In ship resistance and propulsion, we are concerned with the steady forward motion. For j
that reason the discussion may be restricted in many cases to just the force components
pointing in longitudinal direction.
Separation of Simulations of the flow around a ship–propulsor system are a challenge even for today’s
hull and multiprocessor computers. It is also quite difficult to make measurements in this closed
propeller
system which, according to Equation (.), has no resultant external force. Similar
problems exist in structural analysis. In order to reveal shear forces and bending
moments in a beam, one side of the beam is ‘removed’ to reveal the internal forces. To
that effect, the ship–propulsor system is split into two parts which are treated separately:

Total resistance • Bare hull: hull without propulsor and usually without any appendages like rudder,
struts, and bilge keels (Figure .). We remove the propulsor, which means the
bare hull has to be towed to achieve the desired speed. The required tow force is
equal to the ship hull’s total resistance 𝑅𝑇 .
Open water • Propulsor: the propulsor is removed from the ship and its properties are investi-
condition gated in undisturbed parallel flow instead of the disturbed flow field generated by
the hull (Figure .). This is known as open water condition.

Advantages Separation of the hull–propulsor system into its subsystems has advantages which are
exploited in experimental studies of ship hydrodynamics:

• forces of water and air on the hull and the force generated by the propulsor are
revealed,
 Underlined quantities, like velocity 𝑣, represent vectors. See the beginning of Chapter  for details.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

1.1 Calm Water Hydrodynamics 5

(a) Propeller working in a nonuniform flow field (b) Propeller working in a uniform flow (open water
behind the hull (behind condition) condition)

Figure 1.4 Comparison of inflow conditions for a propeller operating in behind and in open water condition

• hulls may be investigated without a specific propulsor, and

j j
• propulsor performance may be determined without the influence of a ship’s wake.

However, separating hull and propulsor considerably changes the hydrodynamic system. Disadvantages
As a consequence, we need to apply corrections to model test and simulation results
performed for hull or propulsor alone.

• The flow around the ship hull will be different for hulls with a propulsor attached
and for hulls without a propulsor. This is especially true for marine propellers.
Rotating propellers accelerate fluid already upstream of the propeller. Hence,
they have a direct impact on the flow around the stern of a vessel.

• Propellers, water jets, or paddle wheels do not operate without a ship attached
to them. The hull will create a nonuniform flow field called wake in which the
propulsor is working as indicated in Figure .(a). Therefore, the open water
condition with uniform inflow into the propulsor shown in Figure .(b) is an
unrealistic, hypothetical case.

• When performance is separately determined for hull and propulsor, the question
arises as to how the results are reconciled to make a prediction for the complete
hull–propeller system.

Although one may argue that it would be better to only investigate the complete hull– Hull-propulsor
propulsor system, it is today’s practice to perform model tests and even calculations for interaction
the separated components. Effects of the omitted part on the performance of the other
part are quantified by additional hydrodynamic characteristics. They will be discussed

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6 1 Ship Hydrodynamics

in Chapter  about hull–propeller interaction. This established practice goes back

many years, and naval architects have to remind themselves that careful interpretation
of these characteristics is required because the quantities they describe do not exist in
the original, combined ship–propulsor system.
Book topic The flow of water around the ship in calm water is the topic of this book. First, we will
discuss the resistance problem, i.e. the flow around a ship hull without its propulsor and
the forces associated with this flow. The resistance problem also provides the context
for review of the underlying concepts of fluid mechanics. In the second half, we will
study the propulsion problem. We examine marine propellers as the most widely used
propulsor and finally learn how to combine the results of hull and propeller analysis to
solve ship design related propulsion problems.

1.2 Ship Hydrodynamics and Ship Design

Your future job The definition of an effective hull shape is a naval architect’s responsibility. Hull geome-
try significantly impacts the performance of a vessel and must be assessed from earliest
design stages until completion of the design.
In a nutshell, the responsibilities of a naval architect charged with the design and
hydrodynamic assessment of a hull include:

• developing a form with minimum powering requirements

j j
• selection and design of an appropriate propulsor of high efficiency
• contributing information about necessary engine power and propulsor specifica-
tions to the overall ship design
• ensuring that the vessel behaves well in waves
• ensuring that the vessel is sufficiently course stable and/or maneuverable depend-
ing on its mission.

In this book, we consider the first three items related to resistance and propulsion in
calm water.
Objective The main objective in this element of ship design is the development of a ship–propulsor
combination which provides the most economic and ecological system to fulfill the
vessel’s mission. Overall size of the ship is usually defined by its purpose and target
route. Ports of call, seaways, and canals may impose further limits. Therefore, changes
to displacement and principal dimensions for hydrodynamic reasons will be small. The
same is usually true for the design speed of merchant vessels. A vessel will become part
of a transport chain sustaining a more or less well defined flow of goods. If the vessel is
slower than envisioned, it becomes a bottleneck in the transport chain. If it is too fast,
it will become idle awaiting the next batch of goods.
Optimization Within the design constraints a naval architect composes a hull shape with minimized
problem resistance and develops an optimum propulsor for it. Even better would be to formally
optimize the hull–propulsor system for overall high efficiency throughout the opera-
tional profile. A formal optimization is unfortunately often skipped for merchant vessels

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

1.3 Available Tools 7

since it requires sophisticated numerical flow simulations and supporting model tests.
Steadily increasing and more affordable computational capacities and more robust
simulation methods will eventually integrate formal optimization into day-to-day ship
design.
For a successful ship design, it is very important to accurately predict the power neces- Performance
sary to achieve the desired cruising or flank speed. The machinery for a main propulsor prediction
is usually the most expensive nonmilitary equipment item. An oversized engine wastes
space in the ship which could have been used for more (paying) cargo or a smaller
(cheaper) ship. Operating costs tend to increase with engine size as well. If the predicted
rating for an engine is larger than necessary, your design will be more expensive and
your customer might order your competitor’s design instead. The ship might not reach
its design speed if the naval architect happens to underpredict the power requirements.
In many cases there is no easy fix, and shipbuilding contracts commit the shipbuilder or
design agent to paying hefty fines if the contracted speed is not achieved. In the worst
case, the customer might refuse to take ownership of the ship, leaving a shipbuilder
with unpaid bills and an unwanted asset.
The essential challenge in ship hydrodynamics is to get it right the first time. Most Challenges
ships are one of a kind designs with investment cost so high that a prototype cannot
be built and tested before the actual vessel is constructed. Small series exist for naval
vessels and smaller ships like pleasure craft and work boats. Merchant vessel series
rarely show more than single digit repeats. Numbers that are typical for aircraft or car
models are never reached in shipbuilding. This does not imply that engineering of cars
j and aircraft is easier. It just means that the economic risk of engineering failures per j
unit is bigger for ships. In small series, results for the lead ship are exploited in the
construction of repeat designs. Again, only minor changes will occur. If the lead vessel
does not perform as desired, repeats will be unlikely.
The difficulty of our fluid dynamics problem is augmented by the fact that reliable
information about hull shape, resistance, and propulsion is needed early in the design.
Except for minor details, the hull shape needs to be settled as soon as possible since it
affects not only resistance and propulsion, but also stability, structure, and functionality
of the vessel. This leaves very little time for extensive computational analyses or model
tests for merchant vessels where the early design time is measured in weeks rather than
months. Naval vessels tend to have longer lead times, but in their case additional and
often changing mission constraints complicate the design problem.

1.3 Available Tools

A moving ship creates a complex flow field in its vicinity. As naval architects we Task
are interested in the physical properties of the flow field like fluid particle velocities,
pressure distributions, and their ultimate effects on the performance of a ship. Obtaining
quantitative descriptions of the flow processes around a ship hull poses a complex
exterior flow problem, characterized by:

• complicated three-dimensional geometries,

• two fluids (air and water), and

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8 1 Ship Hydrodynamics

• moving boundaries.

Moving Although the shape of hull and propeller are known in later design stages, their position
boundaries in the water and the boundary between air and water will change when the vessel is
underway. Changing water pressure along the hull causes it to deviate from its position
at rest. In computational fluid dynamics the boundary between air and water is often
called free surface. Its wavy shape has to be found as part of the solution.
About  years of research and advances in engineering sciences have developed a set
of tools which naval architects may employ to make accurate predictions of resistance
and propulsion properties. However, ship hydrodynamics is still an active field of
research, and improvements to existing methods and new tools for old and new design
problems will be developed for the foreseeable future.
Tools In order to obtain the desired data for our ship design, we can perform:

Model tests • model tests, since we have no prototype to test at full scale:
– Model tests are well established and arguably the most reliable performance
prediction method. The trial speed at a certain power output of the engine
has to be predicted within ±. kn. Model tests are performed in later stages
of the design.
– We cannot satisfy all required physical scaling laws at the same time. There-
fore, model test data have to be extrapolated to full scale. This can introduce
j significant errors if not done properly. Model testing will be discussed later j
in the book (Chapters – and –).
– Model tests are considered expensive. However, they are cheap compared
with the overall cost of a ship. Model tests are usually done just for the
final design to confirm earlier predictions. Modifications might be tested if
problems arise which invalidate earlier estimates, and the contracted trial
speed might not be reached.
Estimates • estimates:
– Estimates are only acceptable for early design phases.
– Available methods are the result of regression analyses of model test data
and ship trial data. The methods are quite simple but not always accurate
(about ±10%).
– Despite their limited accuracy, estimates are still the method of choice in
early design stages. We will discuss current methods in Chapters ,  and
.
Numerical • CFD (computational fluid dynamics) simulations:
analysis
– CFD simulations are time consuming, especially the grid generation, and
actual computation may take several hours or even days to complete.
– CFD is not very reliable yet and a lot of experience is needed to produce
a valid grid and to select appropriate boundary conditions and process
parameters. An expert user can produce results which are as good as a
model test, but the occasional user should apply CFD results with great
caution.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

1.3 Available Tools 9

– Reliable results can be obtained for partial problems, e.g. wave resistance
computations and related shape optimization of the forebody.
– CFD results are useful to compare and rank design alternatives.
– With the steady increase in computational speed and improvements in flow
modeling, CFD will play an increasingly important role in ship resistance
and propulsion predictions.

We will discuss marine model testing and practical estimates at length. CFD, although of
increasing accuracy and importance, is not presented here because it requires additional
background in numerical methods which does not fit the format of this book.

Self Study Problems

. Define the calm water condition.
. Discuss pros and cons of investigating hull and propulsor separately.
. Explain in your own words what total resistance means in ship hydrodynamics.
. Explain what open water condition means and where it is used.
. State the advantages and disadvantages of the separate analyses of hull and pro-
j peller. j

. Which tools are at the naval architect’s disposal for resistance estimates? Discuss
their advantages and disadvantages.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10

2
Ship Resistance

Resistance, or drag, is a force acting against the motion of a vessel. This chapter defines
resistance in the context of ship hydrodynamics and explores possible subdivisions to
better understand how characteristics of the ship and the fluid contribute to the resultant
force. All subdivisions of resistance provide useful insight but are also ambiguous
because clear boundaries do not exist between the components. The reader should
always keep in mind that the total resistance is the only force which can be directly
measured.
We will discuss three different subdivisions of the resistance: first, we observe flow
phenomena associated with the moving vessel and qualitatively explain how they con-
tribute to the resistance. Next, a more pragmatic subdivision is introduced which forms
j the basis for meaningful model tests. Finally, we will turn to the physical aspects of how j
fluids interact with the hull which are exploited in computational tools.

Learning Objectives
At the end of this chapter students will be able to

• discuss the concept of calm water total resistance

• distinguish flow phenomena around a ship hull

• define components of ship resistance

• summarize how water and air interact with a vessel

2.1 Total Resistance

Total resistance Naval architects link ship resistance or, better, total resistance 𝑅𝑇 with three distinct
conditions:

• Total resistance is the horizontal force acting on the hull. The effects of a propulsor
are deliberately excluded.

• The ship is moving with constant speed on a straight course.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.2 Phenomenological Subdivision 11

• The water is calm, i.e. without a current and without wind or wind generated
waves. In addition, we assume the water is deep enough so that the sea bottom
does not affect the waves generated by a moving vessel.

The total resistance 𝑅𝑇 of a hull may be measured as a towing force.

The total resistance for displacement type vessels is usually reported as a dimensionless Total resistance
total resistance coefficient 𝐶𝑇 . coefficient
𝑅
𝐶𝑇 = 𝜌 𝑇 (.)
𝑣2𝑆 𝑆
2
Resistance is normalized by the product of fluid density 𝜌, constant vessel speed 𝑣𝑆
squared, and wetted surface 𝑆. The wetted surface 𝑆 is taken for the ship at rest at
the selected displacement and trim. We ignore changes in wetted surface caused by
ship generated waves while the ship is underway. Typically, total resistance coefficient
values range from . to ..

2.2 Phenomenological Subdivision

Whether you watch a ship sail by or observe a duck paddling across a pond, you will
notice similar flow phenomena. Some are obvious, others less so.
j j
• All objects moving on or close to the water surface will create waves. Anyone Ship waves
who has sailed through a storm at sea or who has seen the destruction of coastal
installations caused by high surf can attest that waves contain a lot of energy.
Strong winds will build up high waves given the opportunity to act over a pro-
longed period. A ship sailing in calm water involuntarily spends some of its
kinetic energy to generate waves. Ship waves appear in a distinct pattern and
their creation is associated with the wave resistance. The higher the waves are,
the higher is the wave resistance.
• All images of a moving vessel show a disturbed flow region along the hull and in Eddies
its wake. Most visible is a band of white froth next to and behind the vessel which
mostly stems from the breaking of the bow wave. Underneath the froth, large
and small eddies form in the flow. Like the generation of ship waves, creation
of eddies siphons kinetic energy away from the ship. In other words, the hull is
performing additional work per unit time. For steady, straight motion the work
per unit time or power is equal to the product of speed and force. Since the speed
is constant and work increases due to the energy transferred into the eddies,
resistance will increase as well. The eddies are caused by friction in the fluid and
friction between hull and fluid. The eddy resistance is considered part of the
broader category of viscous resistance. The kinetic energy stored in the eddies is
eventually converted into other forms of energy and the eddies dissipate.
• Less obvious are the effects of friction between hull and water. The associated Boundary layer
wall shear stress forms the frictional resistance which is typically the biggest
contribution to the total resistance. Friction between hull surface and fluid and
within the fluid is significant only in the boundary layer. A boundary layer is a

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12 2 Ship Resistance

thin sheet of flow around material surfaces in which viscosity of the water plays a
dominant role. Frictional resistance is typically the largest part of the resistance.
• Like a submerged hull moves through water, everything above the waterline
moves through air. As a result, we have additional frictional and eddy resistance
components summarized as air resistance. It is, however, much smaller than the
resistance of the submerged part because the density of air is  times smaller
than that of water. Note that the air, like the water, is assumed to be initially at
rest.

Additional Additional resistance components may be associated with flow phenomena that occur
resistance only in special cases. Some examples are:
components

• Spray resistance: fast boats and ships create a spray pattern at the bow. Water is
forced up and to the sides, which results in an additional resistance component.
• Induced resistance: lifting surfaces like stabilizer fins and rudders generate flow
patterns with distinct vortices. Vortices create flow patterns with an asymmetric
pressure distribution. The main resultant force acts perpendicular to the onflow
(lift force). However, lift is always accompanied by a force component which acts
against the motion. We call this component induced resistance.

Some components of the resistance are associated with specific parts of the ship hull.
j Examples are j

• transom resistance,
• bow thruster tunnel resistance, and
• appendage resistance.

In subsequent chapters we will study the major flow phenomena more closely and
explore how to estimate their contribution to the total resistance.

2.3 Practical Subdivision

The problem of As early as the th century, scientists investigated the resistance of ships by towing
scale scale models across ponds (Calero, ). In , the Swedish admiral and shipbuilder
Fredrik Henrik af Chapman (* – †) even built a towing tank and tested system-
atically varied geometries and model ship hulls (Harris, ). However, the results of
all these efforts were inconclusive because two important questions had not yet been
answered.

(i) When a scale model is towed at a speed 𝑣𝑀 , what is the corresponding speed 𝑣𝑆
of the full scale vessel?
(ii) How do you convert the measured model resistance 𝑅𝑇𝑀 into the resistance 𝑅𝑇𝑆
of the full scale vessel?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.3 Practical Subdivision 13

William Froude (* – †), British engineer and naval architect, was the first to
postulate and successfully apply answers to these fundamental questions (Froude, ).
We will explain his ideas in due course.
Models of ship hulls are build at a fixed geometric scale Length scale

𝐿𝑆
𝜆 = (.)
𝐿𝑀
The subscript 𝑆 indicates a characteristic of the full scale vessel and the subscript 𝑀
stands for the model. All linear dimensions 𝐿𝑆 of the real vessel are shrunk by the
factor 𝜆 to construct the model. Consequently, surfaces 𝑆 and volumes 𝑉 of full scale
vessel and model are connected via the following relationships:
𝑆𝑆 𝑉𝑆
𝜆2 = and 𝜆3 = . (.)
𝑆𝑀 𝑉𝑀

A corresponding relationship for velocities requires an additional constant time scale 𝜏. Time scale

𝑇𝑆
𝜏 = (.)
𝑇𝑀
𝑇𝑆 is the duration of a process at full scale and 𝑇𝑀 the equivalent time at model scale.
The average velocity is defined by the ratio of the distance traveled 𝐿 and the time 𝑇
spent to cover the distance 𝐿.

j 𝐿𝑆 𝐿𝑆 j
𝑣𝑆 𝑇𝑆 𝐿𝑀 𝜆
= = = (.)
𝑣𝑀 𝐿𝑀 𝑇𝑆 𝜏
𝑇𝑀 𝑇𝑀
Given a geometric scale 𝜆, what is the appropriate time scale 𝜏? Can 𝜏 be chosen
independently of 𝜆? We will answer these questions shortly.
In essence, a scale 𝜅 is wanted for the forces 𝐹 acting on ship and model. Force scale

𝐹𝑆
𝜅 = (.)
𝐹𝑀
The weight force of a model of geometric scale 𝜆, for example, may be derived by
applying Archimedes’ principle: a floating body displaces as much water as it weighs.
Therefore, the scale 𝜅𝑤 of the weight forces is
𝑊𝑆 𝜌𝑆 𝑔𝑆 𝑉𝑆 𝜌 𝑔
𝜅𝑤 = = = 𝑆 𝑆 𝜆3 (.)
𝑊𝑀 𝜌𝑀 𝑔𝑀 𝑉𝑀 𝜌𝑀 𝑔𝑀
If we conduct model tests on Earth, the gravitational acceleration should be the same
for full scale vessel and model: 𝑔𝑆 = 𝑔𝑀 = 𝑔. Density of salt water and fresh water differ
by approximately 2.6%. Ignoring this for the moment, you can state that the weight
forces scale like volumes with the scale cubed.
𝜅𝑤 ≈ 𝜆3 (.)
Unfortunately, this simple force scale does not apply to a ship’s resistance.
 Named after the Greek scientist and engineer Archimedes (* BC – † BC), who formulated it first.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14 2 Ship Resistance

2.3.1 Froude’s hypothesis

Corresponding Froude answered the first question (i) with the proposal of ‘corresponding speeds:’
speeds
𝑣𝑆 √
= 𝜆 (.)
𝑣𝑀
Comparing
√ this with Equation (.), we see that this is equivalent to a time scale of
𝜏 = 𝜆. This relationship had previously been presented by the French naval engineer
Ferdinand Reech (* –†). However, it had never been applied to model tests.
Today, the relationship of corresponding speeds is called Froude’s law of similarity in
honor of Froude’s extraordinary contributions to ship model testing. In order to make
sense of model tests, we enforce that the selected model speed satisfies the requirement
of an equal Froude number.
𝑣
𝐹𝑟 = √ = 𝐹𝑟𝑆 = 𝐹𝑟𝑀 (.)
𝑔 𝐿𝑊𝐿
𝐿𝑊𝐿 is the length of vessel or model at the waterline. Assuming that the gravitational
acceleration 𝑔 is the same for model and ship, we may retrieve Froude’s expression for
the corresponding speeds.
𝐹𝑟𝑆 = 𝐹𝑟𝑀
𝑣𝑆 𝑣
√ = √ 𝑀
𝑔 𝐿𝑊𝐿𝑆 𝑔 𝐿𝑊𝐿𝑀
j (.) j
√ √
𝑣𝑆 𝑔 𝐿𝑊𝐿𝑆 𝐿𝑊𝐿𝑆 √
= √ = = 𝜆
𝑣𝑀 𝐿𝑊𝐿𝑀
𝑔 𝐿𝑊𝐿𝑀

A formal justification for this particular time scale is presented in Chapter .
With respect to the force scale 𝜅, Froude realized that a single scale is insufficient. The
wave resistance part requires a different scale factor than the viscous resistance. This is
due to the fact that we cannot attain full dynamic similarity of the fluid flow around
the model if we test the model in water. In order to achieve full dynamic similarity, the
water would have to be scaled somehow like the model itself. This is impossible and, in
fact, a fluid which has an appropriately scaled kinematic viscosity does not exist.
Froude’s In lieu of a fluid with the correct, scaled viscosity, Froude proposed to split the total
hypothesis resistance into two parts:

(i) frictional resistance 𝑅𝐹 of a flat plate with the same wetted surface 𝑆 as the ship
hull, and
(ii) residuary resistance 𝑅𝑅 , which mainly comprises the wave resistance.

The total resistance is the sum of these two parts:

𝑅𝑇 (𝑅𝑒, 𝐹𝑟) = 𝑅𝐹 (𝑅𝑒) + 𝑅𝑅 (𝐹𝑟) (.)
This is known as Froude’s hypothesis or Froude’s method. The hypothesis includes two
assumptions:

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.3 Practical Subdivision 15

• The frictional resistance 𝑅𝐹 is a function of the Reynolds number 𝑅𝑒. The

Reynolds number 𝑅𝑒 for a ship is defined as the dimensionless ratio of ship
speed 𝑣𝑆 times length over wetted surface 𝐿𝑂𝑆 divided by the kinematic viscosity
of the fluid 𝜈.
𝑣 𝐿
𝑅𝑒 = 𝑆 𝑂𝑆 (.)
𝜈
It is named in honor of British scientist Osborne Reynolds (* – †), who
investigated the phenomenon of laminar and turbulent flows.

• The residuary resistance 𝑅𝑅 is solely a function of the Froude number 𝐹𝑟. Froude
did not call it the Froude number, though. He called it corresponding speed.

Froude was well aware that his hypothesis is not entirely true. The residuary resistance
still includes a part of the viscous resistance which is dependent on the Reynolds number.
However, Froude’s hypothesis proved workable and laid the path to meaningful model
testing.
Froude () conducted model tests with flat plates of varying length and surface Froude’s method
roughness. From the results he developed formulas to compute the frictional resistance
component 𝑅𝐹 . The formulas could be cast into a form of

𝑅𝐹 = 𝑐 𝑣𝑚
𝑆𝑆 (no longer in use) (.)

j The exponent 𝑚 of the ship velocity 𝑣𝑆 changed with the length of the plates but seemed j
to level out at 𝑚 = 1.83 for long plates.
Knowing the frictional resistance component 𝑅𝐹 enabled William Froude to separate
the residuary resistance from the measured total resistance of the model.

𝑅𝑅 (𝐹𝑟) = 𝑅𝑇𝑀 (𝑅𝑒, 𝐹𝑟) − 𝑅𝐹𝑀 (𝑅𝑒) for the model (.)

The dimensionless residuary resistance coefficient

𝑅𝑅
𝐶𝑅 = 𝜌 = 𝐶𝑇𝑀 − 𝐶𝐹𝑀 (.)
2
𝑣2𝑀 𝑆𝑀

is applicable to both model and full scale vessel at the Froude number of the model test.
𝜌 2
𝑅𝑅 (𝐹𝑟) = 𝑣 𝑆 𝐶 for the full scale vessel (.)
2 𝑆 𝑆 𝑅
The total resistance of the full scale vessel is obtained by adding a computed frictional
resistance 𝑅𝐹 and small corrections which will be discussed in Chapter .

2.3.2 ITTC’s method

Froude’s initial procedure has been continuously refined and improved over the past ITTC 1957
 years. The International Towing Tank Conference (ITTC) publishes recommended model–ship
correlation line
procedures for model testing and CFD calculations on its website at http://www.ittc.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16 2 Ship Resistance

info. In its current set of recommended procedures, ITTC uses the  model–ship
correlation line as friction coefficient 𝐶𝐹 .
0.075
𝐶𝐹 = [ ]2 (.)
log10 (𝑅𝑒) − 2
Although not completely physical, this formula is widely used today. Chapter  dis-
cusses this topic in more detail. Equation (.) applies only to smooth surfaces, like
the surface of a ship model which has been sanded with very fine grit sandpaper. For
full scale ship hulls, a correction is added which depends on the roughness of the hull
surface. The dimensional frictional resistance is computed in the standard way:
𝜌
𝑅𝐹 = 𝑣2𝑆 𝑆 𝐶𝐹 (.)
2
Form factor Hughes () proposed using a form factor 𝑘 to augment the flat plate frictional resis-
tance. ITTC adopted this approach in an attempt to remove the viscous components
from Froude’s residuary resistance. The total resistance is now split into the viscous
resistance ( )
𝑅𝑉 = 1 + 𝑘 𝑅𝐹 (.)
and the remaining wave resistance 𝑅𝑊 .
( )
𝑅𝑇 = 1 + 𝑘 𝑅𝐹 + 𝑅𝑊 (.)
The form factor 𝑘 of the model is also applied to the full scale vessel. Determining the
form factor by experiments is quite challenging because it involves resistance mea-
j surements at low speeds for which the forces are small and are subject to considerable j
uncertainty.
ITTC’s residuary ITTC held on to the term residuary resistance until recently. Only in its  revision
resistance of the recommended procedures was the name changed from residuary resistance to
coefficient
wave resistance. Therefore, you will find many publications which state Equation (.)
with 𝑅𝑅 instead of 𝑅𝑊 .
Wave resistant In ITTC’s recommended procedure, the wave resistance coefficient 𝐶𝑊 is obtained
coefficient from model tests by subtracting the viscous resistance from the total resistance (ITTC,
b). ( )
𝐶𝑊 = 𝐶𝑇𝑀 − 1 + 𝑘 𝐶𝐹𝑀 (.)

Full scale total The full scale total resistance is obtained by adding an augmented frictional resistance
resistance and corrections to the wave resistance (ITTC, a). In dimensionless coefficients this
coefficient
reads as ( )
𝐶𝑇 𝑆 = 1 + 𝑘 𝐶𝐹𝑆 + 𝐶𝑊 + corrections (.)
Chapters  and  discuss the ITTC method in detail.
Although wave resistance is the major component of Froude’s residuary resistance,
it is important to realize that the wave resistance coefficient 𝐶𝑊 obtained by ITTC’s
method (.) is substantially different from the residuary resistance coefficient 𝐶𝑅
resulting from Froude’s method (.). Figure . illustrates the effect of the form factor
on the size of the remaining resistance components 𝐶𝑅 and 𝐶𝑊 , respectively. For the
same total resistance, the wave resistance coefficient is considerably smaller than the
residuary resistance coefficient. As a consequence, great care should be taken not to
mix the two methods.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.4 Physical Subdivision 17

Figure 2.1 Comparison between Froude’s and ITTC’s current method of derivation for the residuary resistance
coefficient 𝐶𝑅 and wave resistance coefficient 𝐶𝑊
j j

2.4 Physical Subdivision

Phenomenological and practical subdivision of resistance do not lend themselves to
numerical evaluations of the resistance. For this purpose we need a subdivision which
closely follows the physics of fluid mechanics and the mathematical models used to
capture the kinematic and dynamic relationships of flow properties.
In general, three generic types of forces act on any moving body: Forces acting on
a ship

• inertia forces,

• body forces, and

• surface forces.

Inertia is the property of all bodies to maintain their state of motion. The inertia
force acts against the motion and is present when the body accelerates or decelerates.
Newton’s second law states that the inertia force is in equilibrium with the resultant
of all external forces. External forces interact with the vessel as body forces or surface
forces. Body forces act on a vessel as a whole and are present in force fields like gravity
or magnetism. Surface forces act on the surface of the body, usually via pressure and
shear stress.
A body moving with constant speed and without change in direction experiences no
acceleration. Consequently, inertia forces will be zero and the resultant of all external

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

18 2 Ship Resistance

forces has to vanish as well. We consider our ship in this type of steady motion, i.e.
sailing with constant speed on a straight course.

2.4.1 Body forces

Body forces The only notable body force acting on our ship of mass Δ is the gravitational force.

⎛ 0 ⎞
𝑊 = ⎜ 0 ⎟ (.)
⎜ ⎟
⎝ −𝑔 Δ ⎠
Our Cartesian coordinate system < 𝑥, 𝑦, 𝑧 > has its positive 𝑥-axis pointing forward (in
the direction of motion), the 𝑦-axis pointing to port and the 𝑧-axis pointing upwards.
The gravitational force or weight (.) points in negative 𝑧-direction. Therefore, it acts
perpendicular to the resistance force (𝑥-direction).
According to Archimedes’ principle, the weight of a vessel is equal to the weight of the
displaced water at rest.
𝑔Δ = 𝜌𝑔𝑉 (.)
Δ is the mass of the vessel and 𝑉 is its volumetric displacement. For displacement
type vessels, weight is put into equilibrium by the resultant hydrostatic pressure force.
However, the total pressure distribution changes when a vessel is in motion, causing
small changes in vertical position, called dynamic sinkage, and trim.
j j

2.4.2 Surface forces

Surface forces Fluids interact with a ship hull in two ways:

• normal stress 𝜎, acting normal to the surface, and

• shear stress 𝜏, acting tangential to the surface.

The three-dimensional stress in a fluid is described by a stress tensor with three normal
stresses and six shear stresses.
⎛ 𝜎̂ 𝑥𝑥 𝜏𝑥𝑦 𝜏𝑥𝑧 ⎞
Stress tensor: 𝜎̂ = ⎜ 𝜏𝑦𝑥 𝜎̂ 𝑦𝑦 𝜏𝑦𝑧 ⎟ (.)
⎜ ⎟
⎝ 𝜏𝑧𝑥 𝜏𝑧𝑦 𝜎̂ 𝑧𝑧 ⎠
The tensor is symmetric, i.e. 𝜏𝑥𝑦 = 𝜏𝑦𝑥 , 𝜏𝑥𝑧 = 𝜏𝑧𝑥 , and 𝜏𝑦𝑧 = 𝜏𝑧𝑦 . In total, we have to
find six stresses to completely describe the stress at a point in the fluid.
In fluid mechanics, pressure 𝑝 represents the mean normal stress:
( )
1
𝑝 = 𝜎̂ 𝑥𝑥 + 𝜎̂ 𝑦𝑦 + 𝜎̂ 𝑧𝑧 (.)
3
Pressure acts equally in all directions. Separating pressure from the normal stresses
allows us to use the same basic equations for a fluid at rest and in motion.
𝜎̂ = −𝑝 + 𝜎

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.4 Physical Subdivision 19

The remaining stress tensor 𝜎 is dependent on the viscosity of the fluid.

In contrast to displacement type vessels, planing boat hulls develop an upwards oriented Planing boats
pressure force at higher speeds, which carries part of their weight. The corresponding
reduction in displacement and resistance allows them to reach much higher speeds
than displacement type vessels.
The total surface force 𝐹 𝑆 acting on a hull is obtained by the integration of pressure Ship resistance
and stresses over the actual wetted surface 𝑆actual of the ship when it is underway.

𝐹𝑆 = 𝑝 𝑛 d𝑆 + 𝜎 𝑛 d𝑆 (.)
∬ ∬
𝑆actual 𝑆actual

The 𝑥-component of 𝐹 𝑆 is the force commonly referred to as resistance.

The first integral captures pressure changes caused by the formation of a ship wave Wave resistance
pattern and is associated with the wave resistance.
Wave resistance may be eliminated by deeply submerging a body. However, a submerged D’Alembert’s
moving body will still experience a resistance force because water is a viscous fluid. paradox
The viscosity of water is, fortunately, very small. In fact, it is so small that scientists in
the th century widely assumed that frictional forces must be negligible. The French
scientist Jean-Baptiste le Rond d’Alembert (* – †) attempted to compute the
resistance of a submerged body assuming the fluid is frictionless (inviscid). As it turns
out, the resultant pressure force vanishes for inviscid fluids. This clearly does not match
j up to experimental observations and is therefore known as d’Alembert’s paradox. The j
question of how a fluid of such low viscosity can produce significant frictional forces is
partially explained by boundary layer theory. We will discuss viscous and inviscid flow
theories in greater detail in later chapters.
Pressure forces are also the main source of propeller thrust, which is necessary to Propulsion
overcome the resistance.
The second integral in Equation (.) considers the effects of the boundary layer and is Prandtl’s
associated with the viscous resistance. The boundary layer concept was introduced by boundary layer
theory
German engineer Ludwig Prandtl (* – †) in . For now it may suffice that
the boundary layer is a region of flow close to material surfaces in which shear stresses
within the fluid and between the fluid and the hull surface play an important role.
Figure . illustrates the flow regions. Boundary layers cover all surfaces and grow
in thickness from bow to stern. The thickness is exaggerated in Figure .. Typically,
boundary layers at the stern are roughly % of the ship length thick.
As the enlarged section in Figure . shows, fluid particles are actually sticking to Friction between
material surfaces like the hull. The velocity grows rapidly with distance from the hull fluid and hull
until it reaches the exterior flow speed. Boundary layer thickness is commonly defined
as the distance from the surface to where the local velocity has reached the exterior
flow speed. Figure . shows the result of a paint flow test. A custom made paint is put
in stripes on the dry hull of a ship model. The paint stripes roughly follow the ship’s
stations. While the paint is still wet, the model is put back into the water and towed at
the speed of interest. The water rushing over the hull surface smears the paint in the
direction of the flow, thus revealing the flow patterns across the hull.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20 2 Ship Resistance

Figure 2.2 Viscosity of the fluid has significant effect on the flow within the boundary layer around a ship hull

Frictional Paint flow tests are very useful for the proper placement of appendages like bilge
resistance keels, struts, and gratings covering thruster tunnels. In our discussion of resistance
components, the paint flow test is of interest because it confirms what we know from
experience: friction occurs between fluid and the hull surface. The associated shear
stress drags the paint over the surface. The shear stress results in frictional resistance,
j which is the largest resistance component for slow ships. j
Viscous pressure The boundary layer also has an effect on the pressure distribution around the hull.
resistance Compared with inviscid flow patterns, pressure remains lower at the downstream side
(stern) of vessels when a boundary layer is present. The combination of high pressure
at the bow and lower pressure at the stern creates the viscous pressure resistance.

2.5 Major Resistance Components

Major resistance Three resistance components stand out from our discussion of possible subdivisions of
components the ship resistance:

• frictional resistance (𝑅𝐹 , 𝐶𝐹 ) caused by the shear stress between fluid and hull,
• viscous pressure resistance (𝑅𝑉𝑃 , 𝐶𝑉𝑃 ), which reflects the changes in pressure
over a hull due to viscous flow effects, and
• wave resistance (𝑅𝑊 , 𝐶𝑊 ) associated with the generation of the ship wave
system.

Superposition The total resistance is obtained by summing up the components:

𝑅𝑇 = 𝑅𝐹 + 𝑅𝑉𝑃 + 𝑅𝑊 (.)
The sum of frictional and viscous pressure resistance is the viscous resistance.
𝑅𝑉 = 𝑅𝐹 + 𝑅𝑉𝑃 (viscous resistance, smooth) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.5 Major Resistance Components 21

(a) Entrance

(b) Midbody
j j

(c) Run

Figure 2.3 Results of a paint flow test. Photos courtesy of Dr. Alfred Kracht, Versuchsanstalt für
Wasserbau und Schiffbau (VWS), Berlin, Germany

Summations like Equations (.) and (.) imply that the summands are independent
of each other. For resistance components this is generally not the case. We can neither
measure the resistance components independently nor compute them all separately
from first principles. Equation (.) is not practically used because viscous pressure
resistance may not, if at all, be determined accurately enough.
Figure . shows a typical total resistance curve for a container ship with common Resistance curve
subdivisions in dimensionless and dimensional form as a function of Froude number
𝐹𝑟. For low speeds (𝐹𝑟 < 0.2), the total resistance coefficient 𝐶𝑇 𝑆 slightly declines with

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22 2 Ship Resistance

Figure 2.4 Resistance coefficients and resistance for a container ship as functions of the Froude number (velocity)

j j
increasing Froude number. The resistance 𝑅𝑇𝑆 is still increasing, however, at a rate
somewhat lower than velocity squared 𝑣2𝑆 . For higher Froude numbers, the slope of
the total resistance coefficient curve grows continuously, indicating that the resistance
increases at a rate higher than 𝑣2𝑆 .
Humps and The total resistance curves (𝐶𝑇 𝑆 , 𝑅𝑇𝑆 ) show minor humps and hollows in contrast to
hollows the frictional resistance curves (𝐶𝐹𝑆 , 𝑅𝐹𝑆 ) which are smooth. Hollows, with relatively
low resistance, are achieved when the waves generated by the ship hull superimpose into
a wave pattern with low wave heights due to wave cancellation. The humps represent
regions of unfavorable wave superposition with high wave heights. Humps tend to
appear around Froude numbers 0.23, 0.3, and 0.48, whereas favorable hollows appear
near 0.21, 0.25, and 0.34. The details will depend on the actual hull shape. We will
discuss this in greater detail in Chapter  on ship wave resistance.
The lines for frictional resistance (𝐶𝐹𝑆 , 𝑅𝐹𝑆 ), viscous resistance (𝐶𝑉𝑆 , 𝑅𝑉𝑆 ), and wave
resistance (𝐶𝑊 , 𝑅𝑊 ) show that at typical speeds frictional resistance is the largest
resistance component. For low Froude numbers (𝐹𝑟 < 0.12), total resistance is almost
equal to the viscous resistance. Wave resistance is negligible at low speeds, but, with
increasing velocity, wave resistance grows faster than viscous resistance. At Froude
numbers above 0.35, wave resistance will exceed frictional resistance for most vessels.
Additional Other resistance components indicated in Figure . include:
resistance
components
• Roughness allowance Δ𝐶𝐹 : the frictional resistance is increased if the surface is
not hydraulically smooth, i.e. the surface is rough. This applies to the full scale
ship only (see Chapter ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.5 Major Resistance Components 23

• Correlation allowance 𝐶𝐴 : for values derived from model tests, a correlation

allowance may be included, a catch-all for scaling effects not considered elsewhere
in the performance prediction procedure (see Chapter ).

• Appendage resistance 𝐶𝐴𝑃 𝑃 : additional viscous or induced resistance from ap-

pendages (rudder, struts, bossings, etc.) not yet considered in the viscous resis-
tance.

In Figure . the speed of the vessel is represented by the Froude number 𝐹𝑟. This Speed is relative
choice is deliberate. Although the curves will not change if the actual vessel speed
is substituted for the Froude number, it is important to note that vessel speed is a
relative quantity. For a short vessel like a trawler, a speed of 𝑣𝑆 = 10 kn is fast, whereas
for an 18 000 TEU container ship of 400 m length, 𝑣𝑆 = 10 kn is considered slow
steaming. The Froude number, rather than the absolute speed, has a major impact on
the subdivision of ship resistance.
Figure . shows the subdivisions of resistance for three different types of displacement Relative size of
vessels. The data have been taken from the excellent reference by Larsson and Raven resistance
components
(). The decompositions of the actual total resistance coefficients on the left (Fig-
ure .(a)) reflect the same trends shown in Figure .. Although the fishing vessel is
the slowest of the three in absolute velocity, it is sailing at the highest Froude number.
Because of the high Froude number, wave resistance has become the largest component
of its resistance. The high total resistance coefficient of 𝐶𝑇 = 8.1 ⋅ 10−3 for the fishing
vessel indicates that it will need a larger engine relative to vessel size than a container
j j
ship or a tanker. However, absolute value of resistance and installed engine power will
still be the smallest of the three.
The tanker sails at the lowest Froude number, and, as a consequence, its viscous resis-
tance makes up 92.5% of the total resistance (Figure .(b)). Tankers usually have high
block coefficients 𝐶𝐵 ≈ 0.8. Note that the wave breaking portion of the wave resistance
is twice as large as the wave pattern resistance. Blunt bows create a high bow wave
that dissipates a significant amount of energy by wave breaking. The fine lines of the
container vessel create a lower bow wave and most of the wave resistance is associated
with the energy contained in the wave pattern trailing the ship. Slender vessels also
tend to have less viscous pressure resistance.
As mentioned, subdivisions of the resistance provide insight into how vessel speed Only total
and form contribute to the overall resistance. Naval architects have to be mindful of resistance can be
measured
the contributions to the resistance so that design efforts can be properly prioritized. directly
However, one should also keep in mind that all subdivisions of resistance are based on
assumptions and that boundaries among components are blurry. The influence of a
resistance component on other components is usually ignored. Again, only the total
resistance can be measured directly.

References
Calero, J. (). The genesis of fluid mechanics –. Springer, Dordrecht, The
Netherlands.
 TEU stands for Twenty-foot Equivalent Unit, i.e. a  ft long container

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24 2 Ship Resistance

Tanker Container ship Fishing vessel

Fishing vessel 3
CT =2.210 CT =2.310 3 CT =8.110 3
9 vS =10kn 100 2.5%
Resistance components LWL =23m 5.0% 5.0%
wave breaking CWB CT =8.110 3 2.5%
8 wave pattern CWP 0.4050 15.0%

viscous pressure CVP 15.0%

roughness effect CF 80
7 10.0%
form effect
7.5%
flat plate friction CF
[10 3 ]

6 5.0% 10.0% 57.5%

[%]
4.6575 60 2.5%

Resistance coefficients C
Resistance coefficients C

Container ship 40
Tanker
vS =16kn vS =23kn 5.0%
3 LWL =248m 65.0%
LWL =316m 0.4050
CT =2.210 3 CT =2.310 3 60.0% 5.0%
0.057 0.4050 2.5%
0.110 0.2025
2 0.055 0.3450
0.3300 0.230 20
0.165
0.2300
0.110 0.057
1 2.0250 25.0%
1.4300 1.3800
j j
0 0
0.15 0.24 0.34 0.15 0.24 0.34
Froude number Fr [] Froude number Fr []

(a) Absolute values of resistance coefficients (b) Relative values of resistance coefficients

Figure 2.5 Comparison of absolute and relative size of resistance components for three different displacement
type vessels at design speed. The data are taken from Larsson and Raven (2010, pages 13,14)

Froude, W. (). Observations and suggestions on the subject of determining by

experiment the resistance of ships. Correspondence with the British Admiralty.
Published in The Papers of William Froude, A.D. Duckworth, The Institution of Naval
Architects, London, United Kingdom, pp. –, .
Froude, W. (). Experiments on surface-friction experienced by a plane moving
through water. Read before the British Association for the Advancement of Sci-
ence at Brighton. Published in The Papers of William Froude, A.D. Duckworth, The
Institution of Naval Architects, London, United Kingdom, pp. –, .
Harris, D. (). FH Chapman – The first naval architect and his work. Naval Institute
Press, Annapolis, Maryland.
Hughes, G. (). Friction and form resistance in turbulent flow, and a proposed
formulation for use in model and ship correlation. RINA Transactions and Annual
Report, ():–. doi:./rina.trans...
ITTC (a).  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

2.5 Major Resistance Components 25

ITTC (b). Resistance test. International Towing Tank Conference, Recommended

Procedures and Guidelines .---. Revision .
Larsson, L. and Raven, H. (). Ship resistance and flow. The Principles of Naval
Architecture Series. Society of Naval Architects and Marine Engineers (SNAME),
Jersey City, New Jersey.

Self Study Problems

. The next time you come near a pond, observe the wave systems created by water
fowl moving on the water surface.
. Throw two pebbles into a pond so that they hit the water a few feet apart. Observe
the radiating waves. Watch when the two wave systems intersect and what
happens to the wave amplitude as the waves spread.
. A ship model is built at scale 𝜆 = 30. If the full scale vessel has the following
principal dimensions, what are the corresponding values for the model?

full scale vessel model

dimension value value
length between perpendiculars 𝐿𝑃𝑃 179.3 m
j length in water line 𝐿𝑊𝐿 182.2 m j
length over wetted surface 𝐿𝑂𝑆 186.1 m
molded beam 𝐵 29.2 m
molded draft 𝑇 11.2 m
wetted surface 𝑆 6627.2 m2
displacement 𝑉 36355.7 m3
block coefficient 𝐶𝐵 0.62

. If the speed of the vessel in the previous problem is 𝑣𝑆 = 18 kn, what are its
Froude and Reynolds numbers? Assuming the same Froude number, what would
be the tow speed of the model at scale 𝜆 = 30?
. Use the ITTC  model–ship correlation line Equation (.) and the vessel
particulars from Problem  to compute the frictional resistance for the vessel.
Assume the vessel sails at Froude number 𝐹𝑟 = 0.231.
. Which component(s) of the resistance is (are) captured by the form factor 𝑘?
. What is the difference between Froude’s residuary resistance and ITTC’s wave
resistance?
. Explain to a friend the three major resistance components.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26

3
Fluid and Flow Properties

Ship hydrodynamics is concerned with the flow of water and air around hull and
superstructure. In this chapter we formulate basic concepts about fluids and summarize
those physical properties important to ship hydrodynamics. Fluid mechanics employs
different methods of observation and mathematical formulation to find the physical
properties in a fluid flow: velocity, pressure, density, and temperature. The difference
between Lagrange’s and Euler’s formulation of fluid motion is explained, and a brief
review of hydrostatics reinforces the concepts of pressure and pressure forces.

Learning Objectives
j j
At the end of this chapter students will be able to:

• explain the difference between solids and fluids

• know basic properties of water and air

• distinguish between Lagrangian and Eulerian mechanics

• understand and apply the concept of fluid pressure

3.1 A Word on Notation

Vectors In this book, equations are represented in vector form wherever possible. Whether
a quantity is a scalar or a vector should be obvious from the notation and not just
from the context. In many textbooks vectors are represented by symbols in bold print
or little arrows stacked on top of the symbol. I find both methods impractical. Bold
symbols cannot be reproduced on a blackboard and are sometimes hard to distinguish
from normal weighted letters. Arrows simply take too much time. For convenience, I
adopted the notation I use in my handwritten notes. Vectors are identified by a single
underscore, e.g. a velocity vector in space is

⎛ 𝑢 ⎞
𝑣 = ⎜ 𝑣 ⎟ (.)
⎜ ⎟
⎝ 𝑤 ⎠

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.1 A Word on Notation 27

Note that the vector 𝑣 stands for the velocity, which has a magnitude and a direction,
whereas the scalar 𝑣 denotes the transverse velocity component or in some cases the
magnitude of velocity. The vector components 𝑢, 𝑣, and 𝑤 indicate how much a basis
vector contributes to the total. For the Cartesian coordinate system we use the basis
vectors
⎛1⎞ ⎛0⎞ ⎛0⎞
𝑖 = ⎜0⎟ 𝑗 = ⎜1⎟ 𝑘 = ⎜0⎟ (.)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0⎠ ⎝0⎠ ⎝1⎠
The basis vectors are of unit length and are orthogonal to each other. Obviously, the
velocity vector is equal to 𝑣 = 𝑢𝑖 + 𝑣𝑗 + 𝑤𝑘. The magnitude or length of a vector is
defined as the square root of the sum of its squared components. For example, the
magnitude of a velocity vector is
√
|𝑣| = 𝑢2 + 𝑣2 + 𝑤2 (.)

Matrices are marked by double underscores as an extension of the vector notation. Matrices

⎛ 𝑎11 𝑎12 𝑎13 ⎞

𝐴 = ⎜ 𝑎21 𝑎22 𝑎23 ⎟ (.)
⎜ ⎟
⎝ 𝑎31 𝑎32 𝑎33 ⎠
Equation (.) shows a square matrix 𝐴 with dimension [3 × 3], i.e. with three rows and
j j
three columns. A vector is equivalent to a matrix with a single column.
The superscript operator 𝑇 transposes a column vector into a row vector. Appropriate Transposition
use of transposition 𝑇 allows us to employ the rules of matrix multiplication for the dot
product.
The dot product of two vectors is also called the inner product. It results in a scalar and Dot product
is one of the most important vector operations. You will encounter it frequently in this
book. A dot product is executed like a matrix multiplication:

. Two matrices of size [𝑚 × 𝑛] and [𝑛 × 𝑙] can be multiplied only if the number

of columns of the first matrix 𝑛 is equal to the number of rows 𝑛 of the second
matrix. The resulting matrix has as many rows as the first matrix and as many
columns as the second matrix, i.e. [𝑚 × 𝑛] ⋅ [𝑛 × 𝑙] → [𝑚 × 𝑙]. For the dot product
of two three-dimensional vectors, we multiply a row matrix of size [1 × 3] (hence
the transposition of the column vector) with a [3 × 1] matrix. The result is a trivial
matrix of size [1 × 1] which is equivalent to a scalar.
. A matrix multiplication is executed by multiplying a row of the first matrix with
a column of the second matrix element by element and then summing up the
products. In a dot product, each operand has only one row or column to multiply.
Therefore, the result is a scalar.

In contrast to the dot product, which results in a scalar, a cross product of two vectors Cross product
yields a vector. The cross product is properly defined only in three-dimensional space.

𝑎×𝑏 = 𝑐 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28 3 Fluid and Flow Properties

The resulting vector 𝑐 is perpendicular to the plane defined by the vectors 𝑎 and 𝑏. If 𝑎
and 𝑏 are parallel, their cross product will result in the null vector 𝑐 = (0, 0, 0)𝑇 , i.e. it
has neither a direction nor a length.
Dot product Given are two vectors 𝑝𝑇 = (𝑥, 𝑦, 𝑧) and 𝑞 𝑇 = (𝜉, 𝜂, 𝜁). Assuming the basis vectors form
examples an orthogonal set, the dot product is:

( )⎛ 𝜉 ⎞
𝑝𝑇 𝑞 = 𝑥, 𝑦, 𝑧 ⎜ 𝜂 ⎟ = 𝑥 𝜉 + 𝑦 𝜂 + 𝑧 𝜁 (.)
⎜ ⎟
⎝𝜁 ⎠

With actual numbers for the vector components, it looks like this

( )⎛ 4 ⎞
𝑝𝑇 𝑞 = 1, 2, 3 ⎜ 5 ⎟ = 1 ⋅ 4 + 2 ⋅ 5 + 3 ⋅ 6 = 32 (.)
⎜ ⎟
⎝6⎠

Geometrically the dot product represents the projected length of one vector onto the
direction of another.
Dot product The dot product is commutative, i.e. the order of operands may be reversed without
properties affecting the result.
𝑝𝑇 𝑞 = 𝑞 𝑇 𝑝 (.)

j The dot product is zero if 𝑝 and 𝑞 are perpendicular to each other. If you take the j
dot product of a vector 𝑣𝑇 = (𝑢, 𝑣, 𝑤) with itself, the result is equal to the squared
magnitude of 𝑣.
𝑣𝑇 𝑣 = |𝑣|2 = 𝑢2 + 𝑣2 + 𝑤2 (.)

Cross product Using the vectors 𝑝 and 𝑞 from above in a cross product yields
examples

⎛𝑥⎞ ⎛𝜉 ⎞ | 𝑖 𝑗 |
| 𝑘 |
| |
𝑝×𝑞 = ⎜ 𝑦 ⎟×⎜ 𝜂 ⎟ = | 𝑥 𝑦 𝑧 |
⎜ ⎟ ⎜ ⎟ | |
⎝𝑧⎠ ⎝𝜁 ⎠ | 𝜉 𝜂 𝜁 |
| |
( ) ( ) ( ) ⎛ 𝑦𝜁 − 𝑧𝜂 ⎞
= 𝑖 𝑦𝜁 − 𝑧𝜂 + 𝑗 𝑧𝜉 − 𝑥𝜁 + 𝑘 𝑥𝜂 − 𝑦𝜉 = ⎜ 𝑧𝜉 − 𝑥𝜁 ⎟ (.)
⎜ ⎟
⎝ 𝑥𝜂 − 𝑦𝜉 ⎠

With the numbers from above:

⎛1⎞ ⎛4 ⎞ | 𝑖 𝑗 𝑘 ||
|
𝑐 = 𝑝×𝑞 = ⎜ 2 ⎟×⎜ 5 ⎟ = || 1 2 3 |
|
⎜ ⎟ ⎜ ⎟ | |
⎝3⎠ ⎝6 ⎠ | 4 5 6 ||
|
( ) ( ) ( ) ⎛ −3 ⎞
= 𝑖 2⋅6−3⋅5 +𝑗 3⋅4−1⋅6 +𝑘 1⋅5−2⋅4 = ⎜ 6 ⎟ (.)
⎜ ⎟
⎝ −3 ⎠

Check the result by confirming that 𝑐 is perpendicular to 𝑎 and 𝑏 by showing that the
dot products are zero, i.e. 𝑐 𝑇 𝑎 = 0 and 𝑐 𝑇 𝑏 = 0.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.2 Fluid Properties 29

3.2 Fluid Properties

Ships travel in an environment formed by the fluids water and air. Although we all have What is a fluid?
a notion about fluids, it is not easily put into words. In contrast to solids, a fluid at rest
cannot bear external shear stresses. A solid loaded with external shear stresses will
deform until its internal stresses balance the external ones. The deformation of a solid
stops once equilibrium is reached. A fluid will continue to move until the external shear
stresses have vanished.
Fluids comprise liquids and gases. A mass of gas has neither a unique shape nor a Liquid versus gas
definite volume. It will completely fill any container, due to the small forces of attraction
between its molecules. If the container is not closed, gas will escape. Liquids do not
have a unique shape either, but they have a distinct volume. Due to moderate forces of
attraction between its molecules, a liquid will take the shape of its storage container,
but maintain its volume rather than fill the whole container. In the presence of gravity,
liquids typically form a clearly defined surface. Some materials, e.g. glass, have fluid
properties but flow so slowly that we rather treat them as solids.
The liquid considered in ship hydrodynamics is water. Chemically it is a combination Fresh water
of hydrogen and oxygen (H2 O). Depending on the combination of temperature and
pressure, water exists in three phases: solid as ice, liquid as water, and gaseous as
vapor. Varying levels of salt, air, and microscopic particles are dissolved in naturally
occurring water. To bring order into chaos, the international community organized the
International Association for the Properties of Water and Steam (IAPWS, www.iapws.
j j
org), which is tasked with the definition of the standard properties for pure water. The
ITTC has adopted the IAPWS guidelines and incorporated them into its own set of
recommended procedures (IAPWS, ; ITTC, ).
The essential difference between fresh water and seawater is the salt content or short Seawater
salinity. Salinity varies quite a bit across the Earth’s oceans. The Intergovernmental
Oceanographic Commission (IOC) has set a standard absolute salinity of 35.16504 g salt
per one kilogram of seawater. In ITTC’s recommended procedures, seawater properties
are taken from IOC et al. () and complemented with correlations from Sharqawy
et al. ().

3.2.1 Properties of water

Only a few of the water properties defined in the cited documents are relevant for our
discussion of ship hydrodynamics:

• density,

• viscosity, and

• vapor pressure.

All three quantities are a function of fluid temperature.

Density measures the compactness of matter and is given as mass unit per volume unit, Density
e.g. kg/m3 . As mentioned before, for any given temperature, density is treated as a
constant in most applications of ship hydrodynamics.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

30 3 Fluid and Flow Properties

Viscosity Viscosity is a measure for the internal friction and expresses how sticky a fluid is. The
viscosity of a fluid is either provided as:

• dynamic viscosity 𝜇, measured in units of force times time per area, i.e. Ns/m2 or
kg/(ms), or as
• kinematic viscosity 𝜈, which is defined as the ratio of dynamic viscosity and
density.
𝜇
𝜈 = (.)
𝜌
Kinematic viscosity 𝜈 is stated in units of m2 /s.

In ship hydrodynamics, kinematic viscosity 𝜈 is used more often because it is more

convenient for fluids with constant density.
Vapor pressure Sometimes molecules of a liquid escape and form a vapor (gas) above the liquid. Va-
por pressure is a measure for the tendency of molecules to transfer from liquid into
vapor stage. The higher the vapor pressure of a liquid, the more likely are transfers.
Vapor pressure is low for water at room temperature. Only when you heat water to
its boiling point (around 100 ◦ C) does its vapor pressure reach atmospheric pressure
levels. Transfers from liquid to vapor phase happen not only at the surface of a liquid
but also within a liquid. Just observe the bubbles that appear at the bottom of a pan
filled with boiling water next time you prepare pasta. High temperatures are unlikely in
ship hydrodynamics. However, sometimes the pressure in a flow falls below the vapor
j j
pressure and bubbles filled with vapor form in the fluid. This is called cavitation, which
has many detrimental effects. Details on this phenomenon and how to prevent it are
presented in Chapters  and .
Fresh water Most model tests are conducted in tanks filled with fresh water to avoid exposing
properties sensitive sensors and electronic equipment to the corrosive force of seawater. Table .
provides fresh water properties for a range of water temperatures typical for experi-
mental facilities. ITTC’s recommended procedure should be consulted for values at
other temperatures (ITTC, ).
Standard Water temperature varies considerably in the world’s oceans: from a balmy 30 ◦ C in
seawater the Caribbean to just below freezing in arctic waters. However, resistance predictions
properties
for ship designs are typically conducted for seawater of 15 ◦ C temperature. The rele-
vant quantities from the ITTC Recommended Procedure for fresh water and seawater
properties (ITTC, ) are:

• density: 𝜌𝑆 = 1026.021 kg/m3

• kinematic viscosity: 𝜈𝑆 = 1.1892 ⋅ 10−6 m2 /s seawater
• vapor pressure: 𝑝𝑣𝑠 = 1670.82 Pa

Figure . compares density, kinematic viscosity, and vapor pressure of fresh and seawa-
ter for the common range of water temperatures. The values in Table . and Figure .
have been computed for a standard atmospheric pressure of 101 325 Pa. The standards
should be consulted if the properties are needed for substantially different pressure
levels as may occur in deep sea diving.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.2 Fluid Properties 31

Table 3.1 Fresh water properties

1.05
of fresh water
𝑡𝑊 𝜌 𝜈 𝑝𝑣
−6 1.04 S of sea water
[10 m2∕s]

[103 kg/m3 ]
[◦ C] [kg∕m3 ] [Pa]
15.0 999.103 1.1386 1705.74 1.03
15.1 999.088 1.1356 1716.76
15.2 999.072 1.1326 1727.84 1.02 S =1 026.021kg/m
3
15.3 999.057 1.1296 1738.98
15.4 999.041 1.1267 1750.19

density
1.01
15.5 999.026 1.1237 1761.46 =999.103k g/m3
15.6 999.010 1.1208 1772.79
15.7 998.994 1.1179 1784.18 1.00
15.8 998.978 1.1150 1795.64
15.9 998.962 1.1121 1807.17 0.99
16.0 998.946 1.1092 1818.76 0 5 10 15 20 25 30
16.1 998.930 1.1064 1830.41
16.2 998.913 1.1035 1842.13 [10 6 m2 /s] 2.0
16.3 998.897 1.1007 1853.92 of fresh water
16.4 998.880 1.0978 1865.77 1.8 S of sea water
16.5 998.863 1.0950 1877.69
16.6 998.847 1.0922 1889.68 1.6
16.7 998.830 1.0894 1901.73
16.8 998.813 1.0866 1913.85 1.4 6
S =1.18921 0 m2 /s
kinematic viscosity

16.9 998.795 1.0839 1926.04

17.0 998.778 1.0811 1938.29 1.2
17.1 998.761 1.0784 1950.61
6
17.2 998.743 1.0756 1963.01 1.0 =1 .13861 0 m2 /s
17.3 998.725 1.0729 1975.47
17.4 998.708 1.0702 1988.00 0.8
17.5 998.690 1.0675 2000.60
17.6 998.672 1.0648 2013.27 0.6
17.7 998.654 1.0621 2026.01 0 5 10 15 20 25 30
17.8 998.635 1.0594 2038.82
j 17.9 998.617 1.0568 2051.70 4500 j
18.0 998.599 1.0541 2064.66 pv of fresh water
4000
18.1 998.580 1.0515 2077.68 pvS of sea water
vapor pressure pv [Pa]

18.2 998.561 1.0489 2090.78 3500

18.3 998.543 1.0462 2103.95 3000
18.4 998.524 1.0436 2117.19
18.5 998.505 1.0410 2130.50 2500
18.6 998.486 1.0385 2143.89 pv =1 705.74Pa
2000
18.7 998.467 1.0359 2157.35
18.8 998.447 1.0333 2170.89 1500
18.9 998.428 1.0308 2184.50 pvS =1 670.82Pa
1000
19.0 998.408 1.0282 2198.18
19.1 998.389 1.0257 2211.94 500
19.2 998.369 1.0232 2225.78
19.3 998.349 1.0207 2239.69 0
0 5 10 15 20 25 30
19.4 998.329 1.0182 2253.68
19.5 998.309 1.0157 2267.74 temperature T [C]
19.6 998.289 1.0132 2281.88
19.7 998.269 1.0107 2296.10
19.8 998.248 1.0083 2310.39 Figure 3.1 Fresh and seawater properties as a function of
19.9 998.228 1.0058 2324.76 temperature
20.0 998.207 1.0034 2339.21

3.2.2 Properties of air

The properties of air are somewhat neglected in ship hydrodynamics. The density of air Density of air
is about 800 times smaller than that of water. Hence, the forces on a ship’s hull are of
greater importance for ship resistance and propulsion in calm water conditions than
the forces on its superstructure. We only make use of the air density in the context of
resistance predictions for ships. The density of air depends on temperature, height above
sea level, and moisture content. For dry air of temperature 15 ◦ C, the International
Standard Atmosphere (ISA) model states an air density of 𝜌𝐴 = 1.225 kg/m3 (ISO,
).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

32 3 Fluid and Flow Properties

Kinematic Simulations of the flow around a ship superstructure also need a value for the kinematic
viscosity of air viscosity. For dry air of temperature 15 ◦ C at sea level, kinematic viscosity is 𝜈𝐴 =
1.4657 ⋅ 10−5 m2 /s. This is an order of magnitude higher than the kinematic viscosity of
water, which seems counterintuitive. Air is actually less viscous than water. However,
you have to compare the dynamic viscosities to make this obvious. The dynamic viscosity
for dry air is 𝜇𝑆 = 1.7955⋅10−5 Ns/m2 and for seawater it is 𝜇𝑆 = 122.0144⋅10−5 Ns/m2 .
Only because of its considerably lower density does the kinematic viscosity of air become
larger than that of seawater (see Equation (.)).

3.2.3 Acceleration of free fall

Standard The gravitational acceleration 𝑔 or acceleration of free fall is obviously not particular to
gravitational any fluid but is nonetheless important for fluid flows. The international community set
acceleration
a standard value for the gravitational acceleration (BIPM, ):
m
𝑔 = 9.80665 standard acceleration of gravity (.)
s2
For engineering applications we often employ a rounded value of 𝑔 = 9.807 m/s2 .
Effect of latitude However, gravitational acceleration varies with geographical latitude and height above
sea level. Therefore, the standard value may not be accurate enough for the calibration
of model testing equipment. The maximum gravitational acceleration occurs at the
j poles with 𝑔𝑝 = 9.8321849378 m/s2 . Due to the elliptical cross section of the globe, j
locations toward the Equator are farther away from the Earth’s center. As a result, they
experience lower 𝑔 values. The gravitational acceleration is 𝑔𝑒 = 9.7803253359 m/s2 at
the Equator. For a location at latitude 𝜃, the gravitational acceleration may be computed
from the following formula:
[ ]
1 + 𝑘 sin2 (𝜃)
𝑔(𝜃) = 𝑔𝑒 √ (.)
1 − 𝑒2 sin2 (𝜃)
with constants 𝑘 = 0.00193185265 and 𝑒2 = 0.00669437999

Data and formula have been stated according to the World Geodetic System  (WGS
, see e.g. NIMA, ).
Local Gravitational acceleration is also affected by various mineral deposits and changes
gravitational in thickness of the Earth’s crust. The Physikalisch-Technische Bundesanstalt in Ger-
acceleration
many maintains a web page at https://www.ptb.de/cartoweb/SISproject.php, which
allows the user to enter longitude, latitude, and height above sea level to retrieve the
gravitational acceleration for a specific location.

3.3 Modeling and Visualizing Flow

Kinematics and Basic courses in dynamics introduce the kinematics and kinetics of moving bodies.
kinetics Kinematic describes the movement of matter and kinetic connects the motion with its
causes, i.e. the forces acting on the objects. In particular, you studied Newton’s second

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.3 Modeling and Visualizing Flow 33

law stating that the inertia force 𝑚 𝑎 is equal to the resultant external force acting on
the body. ∑
𝑚 𝑎 = 𝐹 external = 𝐹𝑖
𝑖

Following individual bodies and describing their motions is the Lagrangian formulation Lagrangian
of mechanics. It is named after Joseph-Louis Lagrange (* – †), an Italian mechanics
mathematician who succeeded Leonard Euler as the director of mathematics at the
Prussian Academy of Sciences in Berlin. You can find details in your textbook on
dynamics or, for starters, read the article on Langrangian mechanics on Wikipedia
(http://en.wikipedia.org/wiki/Lagrangian_mechanics).
The Lagrangian description states physical quantities (velocity, momentum, etc.) as
properties of a piece of matter with mass 𝑚. This is necessary to formulate Newton’s
second law of motion. However, the Lagrangian approach does not lend itself well to
describe the motion of fluids because we would have to track an infinite number of
fluid particles to entirely describe a flow.
It is more practical to study the development of physical properties over time at various Eulerian
positions in the fluid domain. Jean-Baptiste le Rond d’Alembert, a French mathemati- mechanics
cian, physicist, philosopher, and music theorist, introduced this type of mathematical
modeling of continua. However, we now call it the Eulerian description. It is named
after Swiss mathematician and physicist Leonard Euler (pronounced ‘Oiler’, * –
†), who used this method to formulate the basic equations of motions for an invis-
j cid fluid. We will study these later in Chapters  and . In the Eulerian formulation, j
physical properties of the flow are described as functions of space 𝑥𝑇 = (𝑥, 𝑦, 𝑧) and
time 𝑡. For example, the pressure function is 𝑝 = 𝑝(𝑥, 𝑦, 𝑧, 𝑡). This is also known as field
theory. Chapter  reviews the substantial derivative, which converts flow properties
from Eulerian into Lagrangian coordinates.
The description of a flow field in Eulerian coordinates has two advantages:

(i) For many flow patterns, a coordinate system exists in which the Eulerian descrip-
tion becomes independent of time. For example, describing the flow around a
body moving forward with constant speed in a coordinate system which is fixed to
the body (moves with the same speed), makes the flow field steady. When you are
a passenger on a ship moving steadily in calm water, the wave pattern you observe
behind the vessel will not change over time.
(ii) The Eulerian description concurs with our instruments measuring flow properties.
We do not have pressure sensors which can be attached to a fluid particle and track
the changes in velocity along the particle’s path. However, we stick Pitot-static
tubes into the fluid to measure velocity over time at selected locations. Pitot-static
tubes are named after French hydraulic engineer Henri Pitot (*–†). They
actually measure dynamic pressure which is proportional to the squared flow
velocity.

 There are two old but very instructive videos which explain the differences between Lagrangian and

Eulerian description of fluid flows. Part : http://www.youtube.com/watch?v=XgL-dnUZc and Part :

http://www.youtube.com/watch?v=BRjBptzhnA

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34 3 Fluid and Flow Properties

Flow properties Since density and temperature are considered constants in ship hydrodynamics, the
objective is to find the spatial and time dependent distribution of velocity and pressure.

⎛ 𝑢(𝑥, 𝑦, 𝑧, 𝑡) ⎞
velocity vector 𝑣 = ⎜ 𝑣(𝑥, 𝑦, 𝑧, 𝑡) ⎟ (.)
⎜ ⎟
⎝ 𝑤(𝑥, 𝑦, 𝑧, 𝑡) ⎠
pressure 𝑝 = 𝑝(𝑥, 𝑦, 𝑧, 𝑡) (.)

Knowledge of these flow properties enables us to compute forces acting on a vessel.

Basic principles Three basic principles of continuum mechanics are available to formulate a mathemati-
cal model of the flow:

• conservation of mass,

• conservation of momentum, and

• conservation of energy.

We will concentrate on the first two of these principles. Conservation of mass leads to
the continuity equation. Conservation of momentum is equivalent to Newton’s second
law and results in the Navier-Stokes equations. For applications in ship hydrodynamics,
the equation for the conservation of energy is often not directly used because the
j j
remaining two are sufficient to describe flows with constant temperature, viscosity, and
density. However, conservation of energy is often employed to model turbulence.
Pathline Flow patterns can be visualized in different ways. If you follow an object and record
its path, you see its pathline. Just observe the flight of a bird or watch a leaf floating
down a stream. Of course, observing the pathline of a fluid particle in a flow is almost
impossible, because we cannot easily identify individual particles. For visualization
purposes, however, small, highly reflective particles may be used.
Streaklines Flow can also be visualized by inserting a constant stream of colored fluid or smoke
into the flow. The smoke enters the flow at a fixed point. The visible lines of smoke
emanating from that point are called streaklines. For instance, the smoke rising from a
candle which has just been extinguished forms a streakline.
Streamlines Most widely known are streamlines. These are lines which are tangent to the velocity
vector at all points they include. Since the flow is tangent to the velocity vector, fluid does
not cross streamlines. Therefore, streamlines divide the flow into regions. Streamlines
may be computed from the Eulerian formulation. Visualizing streamlines, however,
is not quite as simple. You may seed a flow with reflective particles and take a photo
with a suitable long exposure time. Particles will appear as short dashes on the picture
which may be interpreted as velocity vectors. Drawing curves tangent to the vectors
reconstructs the streamlines. In an unsteady flow, streamlines will constantly change.
For time independent flows (steady flows), pathlines, streaklines, and streamlines are
identical, which makes their visualization a lot easier.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.4 Pressure 35

Figure 3.2 The pressure force d𝐹 𝑝 acting on a small

surface element

3.4 Pressure
Pressure is the magnitude of a force per unit area exerted by one object onto another. It Pressure and
is comparable to the distributed load in beam theory. Pressure is a scalar. The direction pressure forces
of the resultant force is defined by the normal vector 𝑛 of the surface a pressure acts
upon. Thus, a pressure 𝑝 exerts the force d𝐹 𝑝 on an infinitesimally small surface area
d𝑆 with normal vector 𝑛 (Figure .).

d𝐹 𝑝 = 𝑝 𝑛 d𝑆 (.)

Depending on the orientation of the normal vector 𝑛, the equation may feature a
j j
preceding minus sign. In this text, the normal vector points into the body, i.e. out of
the fluid domain. Thus, a positive pressure is generating a force in the direction of the
normal vector.
Integration over the entire body surface 𝑆𝑏 yields the total pressure force 𝐹 𝑝 .

𝐹𝑝 = 𝑝 𝑛 d𝑆 (.)
∬
𝑆𝑏

For complex geometries like a ship hull, the integral has to be solved by numerical
methods. In ship hydrodynamics, finding the pressure 𝑝 is the more challenging problem
though.
A simpler case is a body in fluid at rest. Consider the small cube shown in Figure . Hydrostatics
submerged in water. Its sides are aligned with the coordinate system axes and the 𝑧-axis
points upward. The volume of the cube is Δ𝑉 = Δ𝑥Δ𝑦Δ𝑧. It is not moving, hence it is
in hydrostatic equilibrium. The external forces acting on the cube are its weight 𝐹 𝑔 and
pressure forces on its six faces. Archimedes’ principle dictates that the weight of the
body is equal to the weight of the displaced water.

⎛ 0 ⎞
𝐹 𝑔 = 𝜌 ⎜ 0 ⎟ Δ𝑉 (.)
⎜ ⎟
⎝ −𝑔 ⎠

For convenience, we assume that the cube is so small that the pressure may be considered
constant over each face of the cube. In that case, the pressure force integral (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36 3 Fluid and Flow Properties

Figure 3.3 Forces on a small cube in hydrostatic

equilibrum

simplifies to pressure times area, since the normal vectors are constant on each face as
well. The components of the pressure force acting on the cube are:
( )
𝑥-direction: 𝐹𝑝𝑥 = − 𝑝𝑥2 − 𝑝𝑥1 Δ𝑦Δ𝑧
j ( ) j
𝑦-direction: 𝐹𝑝𝑦 = − 𝑝𝑦2 − 𝑝𝑦1 Δ𝑥Δ𝑧 (.)
( )
𝑧-direction: 𝐹𝑝𝑧 = − 𝑝𝑧2 − 𝑝𝑧1 Δ𝑥Δ𝑦

Hydrostatic Since the cube is not moving, all forces must be in equilibrium:
equilibrium ( )
⎛ 𝑝𝑥2 − 𝑝𝑥1 Δ𝑦Δ𝑧 ⎞ ⎛ 0 ⎞
⎜( ) ⎟
− ⎜ 𝑝𝑦2 − 𝑝𝑦1 Δ𝑥Δ𝑧 ⎟ + 𝜌 ⎜ 0 ⎟ Δ𝑉 = 0 (.)
⎜ (𝑝 − 𝑝 ) Δ𝑥Δ𝑦 ⎟ ⎜ ⎟
⎝ −𝑔 ⎠
⎝ 𝑧2 𝑧1 ⎠
We divide the equation by the cube volume Δ𝑉 = Δ𝑥Δ𝑦Δ𝑧.
(𝑝𝑥2 −𝑝𝑥1 )
⎛ Δ𝑥
⎞ ⎛ 0 ⎞
⎜ (𝑝𝑦2 −𝑝𝑦1 ) ⎟ ⎜ ⎟
−⎜ ⎟+𝜌⎜ 0 ⎟ = 0 (.)
Δ𝑦
⎜ (𝑝𝑧2 −𝑝𝑧1 ) ⎟ ⎜ −𝑔 ⎟
⎝ Δ𝑧
⎠ ⎝ ⎠

Infinitesimally Letting the cube shrink to infinitesimal size, i.e. Δ𝑥 → d𝑥, Δ𝑦 → d𝑦, and Δ𝑧 → d𝑧,
small cube converts the first vector into the gradient of pressure grad 𝑝.
⎛ 𝜕𝑝 ⎞ ⎛ 0 ⎞
⎜ 𝜕𝑥 ⎟ ⎜ ⎟
− ⎜ 𝜕𝑦 ⎟ + 𝜌 ⎜ 0 ⎟ = 0
𝜕𝑝
(.)
⎜ ⎟ ⎜ −𝑔 ⎟
⎜ 𝜕𝑝 ⎟ ⎝ ⎠
⎝ 𝜕𝑧 ⎠

Several important conclusions follow from this analysis:

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.4 Pressure 37

(i) The 𝑥- and 𝑦-components of the equation state that the hydrostatic pressure is
constant in horizontal directions because the derivatives vanish.
𝜕𝑝 𝜕𝑝
= = 0 ⟶ 𝑝 is constant in 𝑥 and 𝑦 when fluid at rest (.)
𝜕𝑥 𝜕𝑦

(ii) From the 𝑧-component of the equation we conclude: Basic theorem of

hydrostatics
𝜕𝑝
− − 𝜌𝑔 = 0
𝜕𝑧

Since, according to conclusion (i), the pressure 𝑝 only depends on 𝑧, we may write:

d𝑝 + 𝜌𝑔 d𝑧 = 0 (.)

The latter is known as the basic theorem of hydrostatics. Formal integration with
respect to 𝑧 results in:
𝑝 + 𝜌𝑔𝑧 = const. (.)

(iii) Less obvious is a result for the pressure force on the infinitesimal cube: the first
vector in Equation (.) represents the resultant pressure force on the fluid
element d𝑉 per unit volume. Thus, it yields the same result as Equation (.):

d𝐹 𝑝 = 𝑝 𝑛 d𝑆 = −grad 𝑝 d𝑉 (.)

j The gradient of a scalar is a vector of its spatial derivatives j

𝜕𝑝
⎛ 𝜕𝑥 ⎞
⎜ 𝜕𝑝 ⎟
grad 𝑝 = ⎜ 𝜕𝑦 ⎟ (.)
⎜ 𝜕𝑝 ⎟
⎝ 𝜕𝑧 ⎠

Formal integration of Equation (.) yields Gauss’ integral theorem:

𝑝 𝑛 d𝑆 = − grad 𝑝 d𝑉 (.)
∬ ∭
𝑆 𝑉

Figure . depicts a water column in hydrostatic equilibrium. The column reaches Hydrostatic
from the water surface at 𝑧0 down to the level 𝑧 = −ℎ < 0. An atmospheric pressure pressure
𝑝𝐴 is recorded above the water surface. According to conclusion (i), no forces act in
the horizontal plane. Therefore, we may limit our static equilibrium condition to the
𝑧-direction.
0 = −𝑝𝐴 d𝑆 + 𝑝(𝑧)d𝑆 − 𝜌𝑔d𝑉 (.)
Division by the cross section area d𝑆 and reshuffling the terms yields the well known
equation for the distribution of static pressure in a fluid.

𝑝(𝑧) = 𝑝𝐴 + 𝜌𝑔 ℎ (.)

If we place the origin of the coordinate system at the water surface 𝑧0 = 0, the equation
simplifies to:
𝑝(𝑧) = 𝑝𝐴 − 𝜌𝑔 𝑧 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

38 3 Fluid and Flow Properties

Figure 3.4 Hydrostatic pressure in a water column

Figure 3.5 Pressure distribution around a ship

j j

Note that 𝑧 is negative for locations underwater.

Pressure force on The relationship (.) can be used to prove Archimedes’ principle. We compute
a floating ship the pressure force acting on a floating ship. As indicated in Figure ., a constant
atmospheric pressure 𝑝𝐴 is acting on all surfaces 𝑆𝑑 above the waterline, i.e. we ignore
the small changes in atmospheric pressure over the air draft of the vessel. The wetted
surface 𝑆 below the waterline is subject to the pressure 𝑝(𝑧) from Equation (.), which
varies over the draft.
Solving the integral (.) for these complex surfaces would be a formidable task. Luckily,
Equation (.) provides a much simpler solution. It allows us to convert the surface
integral into a volume integral.

𝐹𝑝 = 𝑝𝑛 d𝑆
∬
𝑆+𝑆𝑑

= − grad 𝑝 d𝑉 (.)
∭
𝑉 +𝑉𝑑

𝑉 and 𝑉𝑑 are the volumes associated with the underwater and above water surfaces 𝑆
and 𝑆𝑑 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

3.4 Pressure 39

The pressure 𝑝 is a function in two parts.

{
𝑝𝐴 for 𝑧 > 0
𝑝 = (.)
𝑝𝐴 − 𝜌𝑔𝑧 for 𝑧 ≤ 0

For the first case, 𝑧 > 0, above the water surface, the gradient is a zero vector, because
the derivatives of a constant vanish.
⎛0⎞
grad 𝑝 = grad 𝑝𝐴 = ⎜ 0 ⎟ for 𝑧 > 0 (.)
⎜ ⎟
⎝0⎠
Below the water line, the pressure gradient turns into

( ) ⎛ 0 ⎞
grad 𝑝 = grad 𝑝𝐴 − 𝜌𝑔𝑧 = ⎜ 0 ⎟ for 𝑧 ≤ 0 (.)
⎜ ⎟
⎝ −𝜌𝑔 ⎠
As a result, the constant part of the pressure will not yield a resultant force. Since we
did not restrict ourselves to a specific shape, this result is general: a constant pressure
acting on all sides of a body does not yield a resultant force.
As expected, the nonzero part of the pressure force points in vertical direction. Since Buoyancy force
the gradient of 𝑝 is zero above the waterline, the volume integral can be restricted to
the submerged volume 𝑉 .
j j
𝐹𝑝𝑧 = − (−𝜌𝑔) d𝑉
∭
𝑉

The integrand is constant across the volume and can be extracted from the integral.

𝐹𝑝𝑧 = −(−𝜌𝑔) d𝑉
∭
𝑉

The two minus signs cancel and the remaining integral yields the volumetric displace-
ment 𝑉 of the vessel.

𝐹𝑝𝑧 = 𝜌𝑔 𝑉 (.)

The result states that the buoyancy force points upwards and is equal to the weight of
the displaced water, which is Archimedes’ principle.

References
BIPM (). Resolution of the rd CGPM. https://www.bipm.org/en/CGPM/db///.
Online; accessed -July-.
IAPWS (). Revised supplementary release on properties of liquid water at . MPa.
IAPWS SR-(), The International Association for the Properties of Water and
Steam, Plzen, Czech Republic.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40 3 Fluid and Flow Properties

IOC, SCOR, and IAPSO (). The international thermodynamic equation of seawater
– : Calculation and use of thermodynamic properties. Manuals and Guides No.
, Intergovernmental Oceanographic Commission. UNESCO (English),  pp.
ISO (). Standard atmosphere. ISO :, International Organization for
Standardization, Geneva, Switzerland.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
NIMA (). Department of Defense World Geodetic System . Technical Report
NIMA TR., third edition, Amendment , National Imagery and Mapping
Agency.
Sharqawy, M., Lienhard V, J., and Zubair, S. (). Thermophysical properties of sea-
water: A review of existing correlations and data. Desalination and Water Treatment,
:–.

Self Study Problems

. Compute the dot products of the resulting vector 𝑐 of Equation (.) with the
arguments 𝑝 and 𝑞 of the cross product. Both should be zero as 𝑐 must be
perpendicular to 𝑝 and 𝑞.
j . Download the latest version of the fresh water and seawater properties document j
from ITTC’s website at www.ittc.info (ITTC, ).
. Read the Wikipedia pages about L. Euler and J.-L. Lagrange.
. A sunken ship is to be raised by a SSCV (semisubmersible crane vessel) with a
lifting capacity of 12 000 metric tons. The ship’s displacement afloat was 𝑉 =
6 000 m3 .
It is now completely filled with water
and sitting on the ground in water of
depth ℎ = 45 m. The contact area
with the ground is 𝐴bottom = 160 m2 .
There is no water between ship bottom
and sand. The density of seawater is
𝜌 = 1026.021 kg/m3 and the atmo-
spheric pressure is 𝑝𝐴 = 101 325 Pa.
Gravitational acceleration in SI units is
𝑔 = 9.807 m/s2 .

Is the crane capable of lifting the sunken

ship? Support your answer with a cal-
culation of the required lifting force.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

41

4
Fluid Mechanics and Calculus

Many of the fundamental pieces of calculus are related to fluid mechanics: total deriva-
tive, gradient, divergence, and rotation, among others. We will explore this connection
and the application of differential operators in fluid mechanics in this chapter.

Learning Objectives
At the end of this chapter students will be able to:

j • recapture selected concepts from calculus j

• understand the connection of differential operators with fluid mechanics

4.1 Substantial Derivative

Before we derive the important continuity and Navier-Stokes equations, we consider a
mathematical tool necessary in the Eulerian description of fluid flows. In the preceding
chapter, we noted that in fluid mechanics we are interested in the spatial distribution of
velocity and pressure over time. In ship hydrodynamics we assume that the density 𝜌 of
the fluid is constant. Such a fluid is called incompressible. For the limited space around
a ship, water temperature 𝑇 and dynamic viscosity 𝜇 are typically constant as well. Thus,
the remaining unknowns are the velocity field 𝑣 and the pressure distribution 𝑝.
Let us consider an infinitesimally small fluid element moving with the flow and track
the development of the pressure 𝑝 (Figure .). At 𝑡 = 𝑡1 the fluid element is at position
𝑥1 and will be subject to the pressure 𝑝 = 𝑝1 (𝑥1 , 𝑡1 ) at its center. Some time later, at
𝑡 = 𝑡2 , the fluid element will be at position 𝑥2 with pressure 𝑝 = 𝑝2 (𝑥2 , 𝑡2 ). We use
a Taylor series expansion to compute the value of 𝑝2 assuming we know 𝑝1 and its
derivatives.
A short reminder for those who have long forgotten their calculus lessons: Taylor series Taylor series
allow us to express an initially unknown value of a function 𝑓 (𝑥2 ) at 𝑥2 by the known

 Named after English mathematician Brook Taylor, * – †.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

42 4 Fluid Mechanics and Calculus

Figure 4.1 Following a fluid particle and the flow properties it encounters along the way

value and derivatives of the function 𝑓 at a nearby location 𝑥1 :

∑
∞
𝑓 (𝑛) (𝑥1 )
𝑓 (𝑥2 ) = (𝑥2 − 𝑥1 )𝑛
𝑛=0
𝑛!
𝑓 ′′ (𝑥1 )
= 𝑓 (0) (𝑥1 ) + 𝑓 ′ (𝑥1 )(𝑥2 − 𝑥1 ) + (𝑥2 − 𝑥1 )2 (.)
2
𝑓 ′′′ (𝑥1 )
+ (𝑥2 − 𝑥1 )3 + (4)
j 6 j
Of course, we have to make sure that the function is continuous and the derivatives
actually exist. Landau’s symbol  is used to represent terms of higher order. In the
equation above, (4) stands for the terms with derivatives of order 𝑛 ≥ 4.
If there is more than one free variable, things become a bit more complicated. With
two variables (𝑥, 𝑦), for instance, the Taylor series expansion of 𝑓 is
𝜕𝑓 (𝑥1 , 𝑦1 ) 𝜕𝑓 (𝑥1 , 𝑦1 )
𝑓 (𝑥2 , 𝑦2 ) = 𝑓 (0) (𝑥1 , 𝑦1 ) + (𝑥2 − 𝑥1 ) + (𝑦2 − 𝑦1 )
𝜕𝑥 𝜕𝑦
2 2
1 𝜕 𝑓 (𝑥1 , 𝑦1 ) 1 𝜕 𝑓 (𝑥1 , 𝑦1 )
+ (𝑥2 − 𝑥1 )2 + (𝑦2 − 𝑦1 )2
2 𝜕𝑥2 2 𝜕𝑦2 (.)
𝜕 2 𝑓 (𝑥1 , 𝑦1 )
+ (𝑥2 − 𝑥1 )(𝑦2 − 𝑦1 )
𝜕𝑥𝜕𝑦

+ (3)

Note that there are two first order derivatives and three second order derivatives.
Taylor series The pressure distribution 𝑝 = 𝑝(𝑥, 𝑦, 𝑧, 𝑡) has four variables. Writing only the zero and
example first order terms explicitly yields:
( ) ( ) ( )
𝜕𝑝 𝜕𝑝 𝜕𝑝
𝑝2 = 𝑝1 + (𝑥2 − 𝑥1 ) + (𝑦2 − 𝑦1 ) + (𝑧 − 𝑧1 )
𝜕𝑥 1 𝜕𝑦 1 𝜕𝑧 1 2
( ) (.)
𝜕𝑝
+ (𝑡 − 𝑡 ) + (2)
𝜕𝑡 1 2 1

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

4.1 Substantial Derivative 43

We subtract 𝑝1 on both sides and divide the equation by the time difference (𝑡2 − 𝑡1 ). Linear
Resorting zero and first order terms and omitting the higher order terms (2) yields approximation
the linear approximation of the pressure:
( ) ( ) ( ) ( )
(𝑝2 − 𝑝1 ) 𝜕𝑝 𝜕𝑝 (𝑥2 − 𝑥1 ) 𝜕𝑝 (𝑦2 − 𝑦1 ) 𝜕𝑝 (𝑧2 − 𝑧1 )
= + + +
(𝑡2 − 𝑡1 ) 𝜕𝑡 1 𝜕𝑥 1 (𝑡2 − 𝑡1 ) 𝜕𝑦 1 (𝑡2 − 𝑡1 ) 𝜕𝑧 1 (𝑡2 − 𝑡1 )
(.)

Of course, by omitting terms of second and higher order, the left and right sides of the
equation are no longer truly identical. The equal sign is justified, however, if we assume
that changes in position and time are small.
The left-hand side of Equation (.) can be interpreted as the mean rate of change of
pressure during the time interval 𝑡 ∈ [𝑡1 , 𝑡2 ]. We write:
(𝑝2 − 𝑝1 ) Δ𝑝
= (.)
(𝑡2 − 𝑡1 ) Δ𝑡
On the right-hand side of Equation (.), we obtain similar expressions for the change
in position for each of the three coordinate axes.
(𝑥2 − 𝑥1 ) Δ𝑥 (𝑦2 − 𝑦1 ) Δ𝑦 (𝑧2 − 𝑧1 ) Δ𝑧
= = = (.)
(𝑡2 − 𝑡1 ) Δ𝑡 (𝑡2 − 𝑡1 ) Δ𝑡 (𝑡2 − 𝑡1 ) Δ𝑡
Replacing the terms in Equation (.) with the abbreviations from Equations (.) and
j (.) yields: j
( ) ( ) ( ) ( )
Δ𝑝 𝜕𝑝 Δ𝑥 𝜕𝑝 Δ𝑦 𝜕𝑝 Δ𝑧 𝜕𝑝
= + + + (.)
Δ𝑡 𝜕𝑡 1 Δ𝑡 𝜕𝑥 1 Δ𝑡 𝜕𝑦 1 Δ𝑡 𝜕𝑧 1

Shrinking the length of the time interval Δ𝑡 to zero results in the total change of pressure
per unit time:
Δ𝑝 D𝑝
lim ≡ (.)
Δ𝑡→0 Δ𝑡 D𝑡
On the right-hand side, the fractions become the local components 𝑢, 𝑣, 𝑤 of the velocity
for Δ𝑡 → 0:
Δ𝑥 Δ𝑦 Δ𝑧
lim = 𝑢 lim = 𝑣 lim = 𝑤 (.)
Δ𝑡→0 Δ𝑡 Δ𝑡→0 Δ𝑡 Δ𝑡→0 Δ𝑡

Performing the transition Δ𝑡 → 0 in Equations (.) and (.) and substituting the Substantial
results into Equation (.) yields: derivative
( ) ( ) ( ) ( )
D𝑝 𝜕𝑝 𝜕𝑝 𝜕𝑝 𝜕𝑝
= +𝑢 +𝑣 +𝑤 (.)
D𝑡 𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
The term D𝑝∕D𝑡 is known as the substantial derivative (or material derivative) of the
pressure.
The substantial derivative consists of two parts:

𝜕𝑝
• the local derivative Local derivative
𝜕𝑡

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44 4 Fluid Mechanics and Calculus

( ) ( ) ( )
𝜕𝑝 𝜕𝑝 𝜕𝑝
Convective • and the convective derivative 𝑢 +𝑣 +𝑤
derivative 𝜕𝑥 𝜕𝑦 𝜕𝑧

In the Eulerian formulation, the total rate of change of fluid properties is composed of
a change over time at a selected location and a change caused by the movement of the
fluid element from one place to another. The former we call local derivative, and the
latter we call convective derivative.
For instance, if you start the air conditioning in a house, the temperature inside will
drop over a period of time – a local derivative. If you step outside, you might experience
a higher temperature, although the temperatures inside and outside have not changed
during the time span you needed to leave the house – a convective derivative.
Substantial The substantial differential D𝑝 is equivalent to the total differential d𝑝 from calculus.
derivative The capital D is retained in many references to indicate the physical meaning of the
operator
substantial derivative as the total rate of change. We introduce the substantial derivative
operator
( ) ( ) ( ) ( ) ( )
D 𝜕 𝜕 𝜕 𝜕 𝜕
= +𝑢 +𝑣 +𝑤 = + 𝑣𝑇 ∇ (.)
D𝑡 𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑡

It may be applied to all flow properties of interest. ∇ is the Nabla operator, which is
explained in the next section.

j j
4.2 Nabla Operator and Its Applications
Nabla operator ∇ The Nabla operator is a differential operator. It is also known as the del-operator. In
three-dimensional Cartesian space it is a vector composed of the three partial spatial
derivative operators
⎛ 𝜕 ⎞
⎜ 𝜕𝑥 ⎟
⎜ 𝜕 ⎟
∇ = ⎜ ⎟ (.)
⎜ 𝜕𝑦 ⎟
⎜ 𝜕 ⎟
⎝ 𝜕𝑧 ⎠

The Nabla operator is very useful in fluid mechanics because it significantly simplifies
notation and makes the meaning of individual terms more transparent; however, by
itself it has no physical meaning.
Throughout this book the Nabla operator will be used on numerous occasions. Hence,
it will be advantageous if you make yourself familiar with the following applications
of ∇.

4.2.1 Gradient

Gradient First, we apply the Nabla operator to a scalar physical property like the pressure 𝑝. In
this case, we get a vector with the spatial changes of pressure. This vector is known as

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

4.2 Nabla Operator and Its Applications 45

the pressure gradient. If 𝑠 represents a scalar property, the following expressions are
equivalent:
⎛ 𝜕𝑠 ⎞
⎜ 𝜕𝑥 ⎟
⎜ 𝜕𝑠 ⎟
∇𝑠 = grad 𝑠 = ⎜ ⎟ (.)
⎜ 𝜕𝑦 ⎟
⎜ 𝜕𝑠 ⎟
⎝ 𝜕𝑧 ⎠
The gradient of a scalar quantity 𝑠 has some important qualities:

• The gradient points into the direction of steepest ascent, i.e. if you always follow
this direction, you take the shortest path to the ‘summit’ or largest value.
• A direction normal to the gradient is part of an isoline, i.e. 𝑠 does not change.
• If the gradient vanishes, i.e. grad 𝑠 = 0, you have a possible extreme value of
the quantity 𝑠. It is a necessary but not a sufficient condition for a minimum or
maximum. The point could just be a saddle point.

4.2.2 Divergence

Nabla may also operate on vectorial quantities 𝑣. Since the Nabla operator itself is a
vector, three specific operations exist:
j j
• a dot product ∇ 𝑣 (see Section ..) which results in a scalar,
𝑇

• a cross product ∇ × 𝑣 (see Section ..) which results in a vector, or

• a dyad ∇ 𝑣𝑇 resulting in a matrix.

The dot and cross products with the Nabla operator are frequently used in fluid me-
chanics and warrant a closer look.
The divergence of the velocity vector 𝑣 has a special meaning as the rate of change of Moving control
volume of a fluid. For illustration, imagine a balloon filled with helium rising into the volume
air (Figure .). If the balloon is properly sealed, the mass contained in the volume
is constant as the balloon will always contain the same fluid particles. However, the
balloon may change its shape.
The divergence of a vector is defined as a dot product between the Nabla operator and
the vector 𝑣:

( )⎛ 𝑢 ⎞
𝜕 𝜕 𝜕⎜ ⎟
div 𝑣 = ∇ 𝑣 =
𝑇
, ,⎜ 𝑣 ⎟ (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
⎜ ⎟
⎝ 𝑤 ⎠
𝜕 𝜕 𝜕
= 𝑢+ 𝑣+ 𝑤
𝜕𝑥 𝜕𝑦 𝜕𝑧
𝜕𝑢 𝜕𝑣 𝜕𝑤
= + +
𝜕𝑥 𝜕𝑦 𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

46 4 Fluid Mechanics and Calculus

Figure 4.2 A moving, finite control

volume 𝑉 which changes over time

Figure 4.3 The distance 𝑠𝑛 traveled by a surface

element in normal direction

j j

Differential We observe a small area d𝑆 on the balloon’s surface. Points on the surface 𝑆 of the
change in control volume move with the flow velocity 𝑣. A surface element travels a distance
volume
Δ𝑠 = 𝑣 Δ𝑡 during a small time interval Δ𝑡. In the direction of the surface normal 𝑛, the
surface element will travel the distance 𝑠𝑛 (see Figure .).
( )𝑇
𝑠𝑛 = 𝑣 Δ𝑡 𝑛 (.)

The movement of the surface element d𝑆 causes a small change in the volume 𝑉 . The
volume of a prism is equal to the product of base surface d𝑆 and height 𝑠𝑛 normal to
the base surface. ( )𝑇
d𝑉 = 𝑠𝑛 d𝑆 = 𝑣 Δ𝑡 𝑛 d𝑆 (.)

Of course, if the integrity of the control volume is to be maintained, surface elements

cannot move independently. Over the time interval Δ𝑡, the control volume will change
by the integral of all differential volume changes d𝑉 .
( )𝑇
Δ𝑉 = 𝑣 Δ𝑡 𝑛 d𝑆 (.)
∬
𝑆

The time interval is independent of the integration variable d𝑆 which allows us to

extract it from the integral and to divide the equation by Δ𝑡.

Δ𝑉
= 𝑣𝑇 𝑛 d𝑆 (.)
Δ𝑡 ∬
𝑆

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

4.2 Nabla Operator and Its Applications 47

In the limit case Δ𝑡 ⟶ 0, we obtain the substantial derivative of the control volume 𝑉 .

D𝑉
= 𝑣𝑇 𝑛 d𝑆 (.)
D𝑡 ∬
𝑆

The divergence theorem from calculus states that for any continuous vector field 𝑣 the Divergence
following relationship exists: theorem

( 𝑇 )
𝑣𝑇 𝑛 d𝑆 = ∇ 𝑣 d𝑉 (.)
∬ ∭
𝑆 𝑉

Combining Equations (.) and (.), we can express the total derivative of the
volume 𝑉 as:
D𝑉 ( 𝑇 )
= ∇ 𝑣 d𝑉 (.)
D𝑡 ∭
𝑉

In a fluid we may subdivide a control volume until we ( end) up with a small volume
𝛿𝑉 . In fact, we make 𝛿𝑉 so small that the integrand ∇ 𝑇 𝑣 , in good approximation,
is constant across 𝛿𝑉 . Then, the volume integral on the right-hand side reduces to a
j product of the divergence of the velocity with the volume. j

D(𝛿𝑉 ) ( 𝑇 ) ( )
= ∇ 𝑣 d𝑉 ≈ ∇ 𝑇 𝑣 𝛿𝑉 (.)
D𝑡 ∭
(𝛿𝑉 )

Dividing this equation by 𝛿𝑉 yields an expression for the divergence of the velocity Divergence of
which bears a physical meaning. velocity

( )
D(𝛿𝑉 )
D𝑡 rate of change of volume
∇𝑇𝑣 = = (.)
𝛿𝑉 volume

The divergence of a velocity field is a measure for the rate of change in volume per unit
volume.

4.2.3 Rotation

The rotation or curl is defined as the cross product between the Nabla operator and Rotation = curl
a vector. It is therefore only properly defined in three dimensions. For instance, the

 You can also define it for two dimensions; however, the result needs special interpretation as it is not

part of the plane of the two original vectors.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48 4 Fluid Mechanics and Calculus

rotation of the velocity vector is rot 𝑣 with:

⎛ 𝜕 ⎞ | 𝑖 𝑗 𝑘 ||
⎜ 𝜕𝑥 ⎟ ⎛ 𝑢 ⎞ |
⎜ 𝜕 ⎟ ⎜ ⎟ | |
| 𝜕 𝜕 𝜕 ||
rot 𝑣 = ∇ × 𝑣 = ⎜ ⎟ × ⎜ 𝑣 ⎟ = ||
⎜ 𝜕𝑦 ⎟ ⎜ ⎟ | 𝜕𝑥 𝜕𝑦 𝜕𝑧 ||
⎜ 𝜕 ⎟ ⎝ 𝑤 ⎠ | 𝑢 𝑤 ||
⎝ 𝜕𝑧 ⎠ | 𝑣
( ) ( ) ( )
𝜕𝑤 𝜕𝑣 𝜕𝑢 𝜕𝑤 𝜕𝑣 𝜕𝑢
= − 𝑖+ − 𝑗 + − 𝑘 (.)
𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦

⎛ 𝜕𝑤
−
𝜕𝑣 ⎞
⎜ 𝜕𝑦 𝜕𝑧 ⎟
⎜ ⎟
𝜕𝑢 𝜕𝑤
= ⎜ − ⎟
⎜ 𝜕𝑧 𝜕𝑥 ⎟
⎜ 𝜕𝑣 𝜕𝑢 ⎟
⎜ − ⎟
⎝ 𝜕𝑥 𝜕𝑦 ⎠
The rotation of the velocity vector plays an important role in the definition of potential
flows and finds application in lifting flows.

4.2.4 Laplace operator

j Since the Nabla operator is a vector, we can form a dot product of the Nabla operator j
with itself. In Cartesian coordinates we get:

⎛ 𝜕 ⎞
( )⎜ 𝜕𝑥 ⎟
𝜕 𝜕 𝜕 ⎜ 𝜕 ⎟ 𝜕2 𝜕2 𝜕2
∇𝑇∇ = (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧 ⎜⎜ ⎟ =
, , + +
𝜕𝑦 ⎟ 𝜕𝑥2 𝜕𝑦2 𝜕𝑧2
⎜ 𝜕 ⎟
⎝ 𝜕𝑧 ⎠

As expected, the result of a dot product is a scalar. This scalar differential operator
computes the sum of the second order spatial derivatives. As we will see later, it occurs
quite frequently in the mathematical treatment of fluid mechanics and is known as the
Laplace operator 𝚫, so named after the man who defined it, the French mathematician
Pierre-Simon Laplace (*–†).
𝜕2 𝜕2 𝜕2
𝚫 = ∇𝑇∇ = + + (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2

The Laplace operator applied to a scalar quantity 𝜙 yields:

( 𝑇 ) 𝜕2𝜙 𝜕2𝜙 𝜕2𝜙
𝚫𝜙 = ∇ ∇ 𝜙 = + + (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2
The condition that the sum of second order spatial derivatives of 𝜙 vanishes is known
as the Laplace equation.
𝚫𝜙 = 0 Laplace equation (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

4.2 Nabla Operator and Its Applications 49

This linear, homogeneous, second order partial differential equation forms the foun-
dation of potential theory. The Laplace equation expresses conservation of mass for
potential flow, a specific class of inviscid flows. See Chapter  for an introduction and
Kellogg () for details. Besides fluid mechanics, potential theory is also used in the
field of magneto-electrodynamics.

References
Kellogg, O. (). Foundations of potential theory. Springer, Berlin, Germany.

Self Study Problems

. Compute the gradient ∇ 𝜙 for the function

−1
𝜙(𝑥) =
4𝜋𝑟
√
with the distance 𝑟 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2 between the field point
𝑥 = (𝑥, 𝑦, 𝑧)𝑇 and the fixed point 𝜉 = (𝜉, 𝜂, 𝜁)𝑇 (source point). You may assume
that 𝑥 ≠ 𝜉.
j j
. Show that 𝚫𝜙 = 0 for the function
−1
𝜙(𝑥) =
4𝜋𝑟
√
with the distance 𝑟 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2 between the field point
𝑥 = (𝑥, 𝑦, 𝑧)𝑇 and the fixed point 𝜉 = (𝜉, 𝜂, 𝜁)𝑇 (source point). You may assume
that 𝑥 ≠ 𝜉.

. Explain the physical meaning of the divergence of the velocity.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50

5
Continuity Equation

In Eulerian fluid mechanics, conservation laws may be set up for four different types
of control volumes. A control volume may be infinitesimally small or finite in size.
Independent of size, a control volume may be fixed in space or moving along with the
flow. In this chapter, we discuss the application of all four types of control volumes to
the conservation of mass principle. As a result, we obtain four equivalent mathematical
models for the conservation of mass principle known as the continuity equation.

Learning Objectives

j At the end of this chapter students will be able to: j

• define and apply different types of control volumes

• discuss the conservation of mass principle
• derive and apply the continuity equation

5.1 Mathematical Models of Flow

Finite, Different mathematical models may be derived for the same conservation principle.
infinitesimal, The models differ with respect to the size of the control volume (CV) and whether the
fixed or moving
control volume
control volume is fixed or moving. In total, we have four types of control volumes:

(i) infinitesimally small fluid element fixed in space

(ii) finite fluid element fixed in space
(iii) infinitesimally small fluid element moving with the fluid
(iv) finite fluid element moving with the fluid

Figure . shows which combination of fixed or moving and finite or infinitesimal
control volume yields which form of the conservation law. Finite control volumes lead
to integral forms, whereas infinitesimally small control volumes result in differential
forms of the conservation laws. Moving control volumes always contain the same fluid

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

5.2 Infinitesimal Fluid Element Fixed in Space 51

j j
Figure 5.1 Four types of mathematical models for fluid flows and the resulting form of the
conservation law

mass but may change their shape. They yield so-called nonconservative forms of the
equations. Fixed control volumes cannot change their shape or location. Fluid is passing
through the CV and may accumulate or recede in it. The resulting equations are in
conservative form.
The distinction conservative versus nonconservative is rather confusing. It refers to Conservative
properties of the discretized versions of the equations as they appear in numerical versus
nonconservative
solvers. Although the resulting equations may look different, all forms represent the
physical principle of conservation of mass, momentum, or energy. For details see
Anderson, Jr. (, Chapter ).
As an example, we will derive mathematical models for the conservation of mass
principle for all cases (i) through (iv).

5.2 Infinitesimal Fluid Element Fixed in Space

The sketch in Figure . shows a fixed, infinitesimally small control volume d𝑉 . The
cube is so small that flow velocities are assumed to be constant across each of its six
faces. An arrow on each face indicates the mass flux into and out of the fixed fluid
element. The arrows in Figure . point in a positive coordinate direction if they are

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

52 5 Continuity Equation

Figure 5.2 Mass flux through the

surface of a fluid element

on a positive face of the element and vice versa for negative faces of the element. As a
consequence, outflow is considered positive and inflow is considered a negative mass
flow rate. Conservation of mass requires that
j j
time rate decrease of mass inside d𝑉 = mass flow rate out of d𝑉 through d𝑆

or short −𝐶 = 𝐵 (.)

In other words, the negative rate of change of mass 𝐶 in d𝑉 is equal to the net mass
outflow per time unit 𝐵 over the boundary d𝑆 of the control volume d𝑉 .
Net mass flux A mass flow rate or mass flux is measured in kilograms per second. The mass flux across
through surface a surface element d𝑆 is equal to the product of density, velocity normal to d𝑆, and the
surface area. On the face ABCD of the cube, the normal vector is pointing in negative
𝑥-direction, i.e. 𝑛 = −𝑖. We get for the mass flux:

mass flux through face ABCD: − (𝜌 𝑣𝑇 𝑖) d𝑆 = −(𝜌𝑢) d𝑦d𝑧 (.)

On the opposite face EFGH with normal vector 𝑛 = +𝑖, the mass flow rate per area is
expressed as a truncated Taylor series of 𝜌𝑣. Then, the mass flux through face EFGH is
given by:

mass flux through face EFGH:

[ ]𝑇 [ ]
𝜕(𝜌𝑣) 𝜕(𝜌𝑢)
(𝜌𝑣) + d𝑥 𝑖 d𝑆 = (𝜌𝑢) + d𝑥 d𝑦d𝑧 (.)
𝜕𝑥 𝜕𝑥

Application of the Taylor series avoids introducing additional unknowns.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

5.3 Finite Control Volume Fixed in Space 53

Summing the mass flow rates in x-direction (.) and (.) yields:
[ ] ( )
𝜕(𝜌𝑢) 𝜕(𝜌𝑢)
−(𝜌𝑢) d𝑦d𝑧 + (𝜌𝑢) + d𝑥 d𝑦d𝑧 = −𝜌𝑢 + 𝜌𝑢 + d𝑥 d𝑦d𝑧
𝜕𝑥 𝜕𝑥
𝜕(𝜌𝑢)
= d𝑥d𝑦d𝑧
𝜕𝑥
𝜕(𝜌𝑢)
= d𝑉 (.)
𝜕𝑥

The resulting mass fluxes for the remaining four faces are:
𝜕(𝜌𝑣)
𝑦-direction, faces AEHB and CGFD: d𝑉 (.)
𝜕𝑦
𝜕(𝜌𝑤)
𝑧-direction, faces ADFE and BHGC: d𝑉 (.)
𝜕𝑧

The sum of Equations (.), (.), and (.) is equal to the total mass flow rate over the
surface d𝑆: [ ]
𝜕(𝜌𝑢) 𝜕(𝜌𝑣) 𝜕(𝜌𝑤)
𝐵 = + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

For the rate of change of mass in d𝑉 on the left-hand side of Equation (.), we first Rate of change of
determine the total mass contained in d𝑉 : mass

j j
d𝑚 = 𝜌 d𝑉 = 𝜌 d𝑥d𝑦d𝑧 (.)

We choose a fluid element d𝑉 so small that the density 𝜌 may be assumed constant
throughout d𝑉 .
The control volume d𝑉 is fixed in space. As a consequence, d𝑉 does not change over
time and the rate of change of mass will be
𝜕𝑚 𝜕𝜌
𝐶 = = d𝑉 (.)
𝜕𝑡 𝜕𝑡

We now state the conservation of mass by equating −𝐶 and 𝐵:

[ ]
𝜕𝜌 𝜕(𝜌𝑢) 𝜕(𝜌𝑣) 𝜕(𝜌𝑤)
− d𝑉 = + + d𝑉 (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
or after division by the volume d𝑉
𝜕𝜌 𝜕(𝜌𝑢) 𝜕(𝜌𝑣) 𝜕(𝜌𝑤)
+ + + = 0 (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧

Introducing the Nabla operator yields Continuity

equation
𝜕𝜌 (differential,
+ ∇𝑇 (𝜌𝑣) = 0 (.) conservative)
𝜕𝑡
Equation (.) represents the conservative, differential form of the continuity equation.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

54 5 Continuity Equation

Figure 5.3 Flux through the surface 𝑆 of a

finite volume 𝑉 fixed in space

5.3 Finite Control Volume Fixed in Space

Next, we investigate a finite control volume 𝑉 that is fixed in space (Figure .). 𝑆
denotes the surface enclosing the control volume 𝑉 . If mass is neither produced nor
destroyed within the control volume 𝑉 , conservation of mass requires that

time rate decrease of mass inside 𝑉 = net mass flow rate out of 𝑉 through 𝑆

or short −𝐶 = 𝐵 (.)

j Net mass flux We begin again with the mass flux 𝐵 across the control surface 𝑆. The orientation j
through surface of the surface is defined by its normal vector 𝑛. By definition, the normal vector is of
unit length |𝑛| = 1 and points out of the fluid domain. The mass flow rate through an
infinitesimally small part d𝑆 of the surface is a function of its size and the fluid velocity
normal to the surface element:

𝜌 𝑣𝑛 d𝑆 = 𝜌 𝑣𝑇 d𝑆 = 𝜌 𝑣𝑇 𝑛 d𝑆 (.)

The dot product 𝑣𝑛 = 𝑣𝑇 𝑛 yields the component of the velocity vector, which points in
the direction of the normal vector.
The net mass flow rate through all of 𝑆 is obtained by integrating Equation (.):

𝐵 = 𝜌 𝑣𝑇 𝑛 d𝑆 (.)
∬
𝑆

If the integrand is positive, mass leaves the control volume 𝑉 .

Rate of change of The mass contained in a fluid element is d𝑚 = 𝜌 d𝑉 . The total mass in 𝑉 will be
mass

d𝑚 = 𝜌 d𝑉 (.)
∫ ∭
𝑉

The rate of change with respect to time is:

𝜕 𝜕𝜌
𝐶 = 𝜌 d𝑉 = d𝑉 (.)
𝜕𝑡 ∭ ∭ 𝜕𝑡
𝑉 𝑉

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

5.4 Infinitesimal Element Moving With the Fluid 55

Since the volume 𝑉 is fixed in space, i.e. its limits do not change over time, interchanging
integration and time differentiation is allowed.
In summary, we have −𝐶 = 𝐵 or Continuity
equation
𝜕𝜌 (integral,
− d𝑉 = 𝜌 𝑣𝑇 𝑛 d𝑆 (.) conservative)
∭ 𝜕𝑡 ∬
𝑉 𝑆

This is the integral, conservative form of the continuity equation.

5.4 Infinitesimal Element Moving With the Fluid

Next, we consider an infinitesimally small fluid element d𝑉 moving with the flow. By
definition, it always contains the same amount of fluid. However, it may change its
shape over time. Since the mass in d𝑉 is constant, its substantial derivative must vanish.
D(d𝑚)
=0 (.)
D𝑡

We expand the substantial derivative of d𝑚 = 𝜌d𝑉 by applying the product rule:

D(𝜌 d𝑉 ) D𝜌 D(d𝑉 )
= d𝑉 +𝜌 = 0 (.)
j D𝑡 D𝑡 D𝑡 j
Division by d𝑉 yields:
[ ]
D𝜌 1 D(d𝑉 )
+𝜌 = 0 (.)
D𝑡 d𝑉 D𝑡

The term in brackets is the divergence of the velocity field (.): div 𝑣 = ∇𝑇 𝑣 Continuity
equation
D𝜌 (differential, non-
+ 𝜌∇𝑇 𝑣 = 0 (.) conservative)
D𝑡
We obtained the differential, nonconservative form of the continuity equation.

5.5 Finite Control Volume Moving With the Fluid

Finally, a very simple case! By definition, the mass in the moving finite control volume 𝑉 Continuity
does not change. Think of a weather balloon released into the atmosphere. It contains a equation
(integral, non-
fixed amount of helium, which expands as the outside pressure decreases with increasing conservative)
altitude. At take-off it will be slack and will burst eventually at high altitude when the
pressure difference becomes too great.
The total mass in the control volume 𝑉 is

𝑚 = 𝜌 d𝑉 (.)
∭
𝑉

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

56 5 Continuity Equation

If the mass does not change over time, its total derivative must vanish.

D𝑚 D
= 𝜌 d𝑉 = 0 (.)
D𝑡 D𝑡 ∭
𝑉

This is the continuity equation in its integral, nonconservative form.

5.6 Summary
Despite the different appearances, all four forms of the continuity equation express the
same physical principle: conservation of mass. With rigorous mathematical transfor-
mations, each form can be converted into the others.
Conversion For instance, the integral, conservative form of Equation (.) may be converted into
the differential, conservative form of the continuity Equation (.). As a first step, we
move the integrals in Equation (.) onto the same side.

𝜕𝜌
d𝑉 + 𝜌 𝑣𝑇 𝑛 d𝑆 = 0
∭ 𝜕𝑡 ∬
𝑉 𝑆

With the help of the divergence theorem (.), the surface integral may be converted
j into a volume integral. j
𝜕𝜌 ( )
d𝑉 + ∇ 𝑇 𝜌 𝑣 d𝑉 = 0
∭ 𝜕𝑡 ∭
𝑉 𝑉

Both volume integrals have the same limits. Therefore, they may be combined.
[ ]
𝜕𝜌 ( )
+ ∇ 𝑇 𝜌 𝑣 d𝑉 = 0 (.)
∭ 𝜕𝑡
𝑉

In general, a vanishing integral does not imply that the integrand must be zero as well.
Just check the sine function sin(𝑥) and its definite integral with limits  and 2𝜋. However,
we know that the conservation of mass principle must hold true in any control volume,
i.e. it must hold true not only in 𝑉 but also in any subdivision of 𝑉 . As a consequence,
the integrand must vanish in any infinitesimally small part of 𝑉 and the integrand must
be zero everywhere:
𝜕𝜌 ( )
+ ∇𝑇 𝜌𝑣 = 0 (.)
𝜕𝑡
This is, of course, the differential, conservative form of the continuity equation from
page . Anderson, Jr. () presents additional conversions.
Application The continuity equation has immediate engineering applications. Study the flow of
water through the contraction nozzle depicted in Figure ..
As a simplification, we assume that the flow velocity vectors are constant across inlet
1 and outlet 2 . The task is to find the outlet velocity as a function of the geometry

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

5.6 Summary 57

Figure 5.4 Flow through a contraction

nozzle

and inlet velocity. Obviously, the nozzle is a finite, fixed control volume. Hence, the
continuity equation (.) in its conservative integral form applies.
Water is considered incompressible, i.e. if the temperature does not change, the density
is constant and its time derivative vanishes.
𝜕𝜌 𝜕𝜌
because of = 0 the volume integral vanishes: d𝑉 = 0
𝜕𝑡 ∭ 𝜕𝑡
𝑉
The continuity Equation (.) converts into

0 = 𝜌 𝑣𝑇 𝑛 d𝑆
∬
j 𝑆 j
This means that the net flow rate across the boundary 𝑆 of the control volume is zero.
In other words, since no fluid may escape through the nozzle wall, the inflow must be
equal to the outflow.

0 = 𝜌 𝑣𝑇1 𝑛1 d𝑆 + 𝜌 𝑣𝑇2 𝑛2 d𝑆
∬ ∬
𝐴1 𝐴2

We substitute the vectors from Figure . into this equation and execute the dot prod-
ucts.
0 = −𝜌 𝑢1 d𝑆 + 𝜌 𝑢2 d𝑆
∬ ∬
𝐴1 𝐴2

As stated above, the velocities 𝑢1 and 𝑢2 are assumed constant across inlet and outlet
respectively. Consequently, the velocities may be extracted and the surface integrals
result in the inlet and outlet cross section areas.
0 = −𝜌 𝑢1 𝐴1 + 𝜌 𝑢2 𝐴2
Therefore,
𝐴1 𝑢1 = 𝐴2 𝑢2
and we find that the outlet velocity is proportional to the ratio of inlet to outlet cross
section areas.
𝐴
𝑢2 = 1 𝑢1
𝐴2
If 𝐴1 > 𝐴2 , the velocity will increase, i.e. 𝑢2 > 𝑢1 . For 𝐴1 ∕𝐴2 = 2.5 and 𝑢1 = 5.0 m/s,
we get 𝑢2 = 12.5 m/s.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

58 5 Continuity Equation

References
Anderson, Jr., J. (). Computational fluid dynamics – The basics with applications.
McGraw-Hill, Inc., New York, NY.

Self Study Problems

. A balloon is completely filled with water and thrown into a river. What type of
control volume does it represent?
. Convert the differential, nonconservative form of the continuity equation (.)
into its differential, conservative form (.). Justify the mathematical operation
in every step.
. Convert the integral, nonconservative form of the continuity equation (.) into
its integral, conservative form (.). Justify the mathematical operation in every
step. Hint: make use of the divergence theorem.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

59

6
Navier-Stokes Equations

The Navier-Stokes equations were independently derived in the first half of the th
century by the French engineer and physicist Claude Louis Marie Henri Navier (* –
†) and the Irish born scientist Sir George Gabriel Stokes (* – †). Although
we have known the equations for almost  years, we do not know much about them
except that they are incredibly hard to solve. Currently, we can solve only simplified ver-
sions of the equations by means of sophisticated numerical algorithms and substantial
computer power. These algorithms are discussed in books about CFD.
We will derive the equations here because they are so fundamental to fluid mechanics.
Afterwards, we will simplify the equations and study some cases that are relevant
to ship hydrodynamics. The Navier-Stokes equations express the conservation of
j momentum principle and represent Newton’s second law (simplified 𝐹 = 𝑚 𝑎) applied j
to a continuum. Newton’s second law states that the momentum of a body will change
only if an external force acts on it. If no resultant external force exists, a body will
continue to move with constant speed on a straight path.

Learning Objectives
At the end of this chapter students will be able to:

• discuss momentum in the context of fluid mechanics

• set up the equilibrium of forces for a control volume
• follow the derivation of the Navier-Stokes equations
• identify terms contained in the Navier-Stokes equations

6.1 Momentum
Momentum is the product of mass 𝑚 and velocity vector 𝑣. Consequently, momentum Momentum
𝑚 𝑣 is a vector quantity with magnitude and direction. It follows from Newton’s second
law that the time rate of change in momentum is equal to the vector sum of all external
forces acting on the mass.
d(𝑚𝑣) ∑
= 𝐹 external (.)
d𝑡

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

60 6 Navier-Stokes Equations

In solid mechanics we commonly assume that the mass is constant and independent of
d𝑣 ∑
time, which leads to the widely known form 𝑚 d𝑡 = 𝐹 external of Newton’s second law.
d𝑣
In addition, we concentrate the mass into a single point. Then d𝑡 is the acceleration
of a mass point caused by the external forces. A similar equation can be stated for the
rotational motion of a mass point.
In fluid mechanics we need a more general approach because we cannot track every
water molecule individually. In our discussion of the continuity equation, we employed
four mathematical models based on four different control volumes which resulted in
varying forms of the continuity equation. Derivation of the Navier-Stokes equations
may be done using either of these models. However, since it is a lengthier process, only
the derivation for an infinitesimally small control volume fixed in space is presented
here. It is arguably the easiest to follow. Other forms are presented by Anderson, Jr.
().

6.2 Conservation of Momentum

Like mass, momentum is a conserved quantity. Changes in momentum may be caused
by the flow of particles across the boundaries of a control volume (momentum transport)
or by external forces acting on the control volume, which cause the fluid particles to
accelerate, decelerate, or change direction. In a closed system, momentum will change
j only if there is a resultant external force. We will first define the rate of change of j
momentum before we consider the transport of momentum across the boundary of a
control volume and the external forces acting on a control volume.

6.2.1 Time rate of change of momentum

The total amount of mass in a fluid element d𝑉 is d𝑚 = 𝜌d𝑉 with momentum 𝜌d𝑉 𝑣.
Since the fluid element is fixed, the change of momentum per unit time is equal to the
local time derivative.
( )
𝜕
time rate of change of momentum: 𝜌d𝑉 𝑣 (.)
𝜕𝑡
Time rate of By definition, a fixed control volume d𝑉 may not change. Thus, d𝑉 represents a constant
change of factor for the time derivative.
momentum
( )
𝜕 𝜌𝑣
time rate of change of momentum: d𝑉 (.)
𝜕𝑡
Like the momentum itself, the time rate of change of momentum (.) is a vector too.

6.2.2 Momentum flux over boundary

We now compute the momentum flux across the boundaries of the fluid element. It
expresses the change of momentum a mass experiences because it moves in the flow
field. Figure . shows the fluid element with only the 𝑥-component of the momentum

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6.2 Conservation of Momentum 61

Figure 6.1 Momentum flux in 𝑥-direction through the surface of an infinitesimal, fixed fluid
element d𝑉

flux across all six faces. For clarity, we omitted the 𝑦-component and 𝑧-component.
Arrows indicate the sign associated with a specific momentum flux.
The mass flow rate (or mass flux) over an infinitesimal surface area d𝑆 is equal to
𝜌𝑣𝑇 𝑛d𝑆. The mass flux depends on density, the velocity 𝑣𝑇 𝑛 normal to the surface
j element, and the area d𝑆 of the surface element. Note that the mass flux through d𝑆 is j
a scalar. As before, we let the normal vector 𝑛 point out of the fluid domain (here the
control volume d𝑉 ). The velocity vector has the components 𝑣 = (𝑢, 𝑣, 𝑤)𝑇 .
Momentum is tied to mass. As mass moves, it carries momentum with it. The amount Transport of
of momentum transported per time unit through a surface element d𝑆 is momentum
( )
momentum flux across d𝑆 = 𝜌 𝑣𝑇 𝑛 𝑣 d𝑆 (.)

The momentum flux has the dimension of a force measured in Newtons:

[( ) ] [ ]
kg m m kg⋅m
⋅ ⋅1 ⋅ m2 = = [N]
m3 s s s2
The dot product (𝑣𝑇 𝑛) is a scalar. Therefore, the components of Equation (.) are given
by
( )
𝑥-component of momentum flux across d𝑆 = 𝜌𝑣𝑇 𝑛 𝑢 d𝑆 (.)
( 𝑇 )
𝑦-component of momentum flux across d𝑆 = 𝜌𝑣 𝑛 𝑣 d𝑆 (.)
( 𝑇 )
𝑧-component of momentum flux across d𝑆 = 𝜌𝑣 𝑛 𝑤 d𝑆 (.)

We now apply this to the infinitesimal control volume d𝑉 , which we consider fixed in
space as indicated by the supports (Figure .). This control volume has eight corners A,
B, C, D, E, F, G, and H, as well as six quadrilateral faces which we identify by their corner
points:

• with a normal vector parallel to the 𝑥-axis faces ABCD and EFGH

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

62 6 Navier-Stokes Equations

• with a normal vector parallel to the 𝑦-axis faces AEHB and CGFD

• with a normal vector parallel to the 𝑧-axis faces ADFE and BHGC

Flux of the We explicitly consider the flux of the 𝑥-component of momentum across all faces of
𝑥-component of the fluid element d𝑉 .
momentum
The normal vector of the face ABCD points in the direction of the negative 𝑥-axis,
Individual faces i.e. 𝑛ABCD = (−1, 0, 0)𝑇 = −𝑖 (Figure .). Substituting 𝑛ABCD and d𝑆 = d𝑦d𝑧 into
Equation (.) yields
( )
ABCD: 𝑥-component of momentum flux = 𝜌 𝑣𝑇 𝑛ABCD 𝑢 d𝑦d𝑧

( ) ⎛ −1 ⎞
= 𝜌 𝑢, 𝑣, 𝑤 ⎜ 0 ⎟ 𝑢 d𝑦d𝑧
⎜ ⎟
⎝ 0 ⎠
( )
= − 𝜌 𝑢 𝑢 d𝑦d𝑧 (.)

In Figure . the minus sign of the right-hand side term in Equation (.) is represented
by the arrow which points in negative 𝑥-direction.
Over the length d𝑥 of a fluid element, the momentum flux per unit surface area (𝜌𝑢𝑢)
may change. As for the conservation of mass, we express the change as a Taylor series
expansion that we truncate after the linear term. The normal vector for the face EFGH
points in positive 𝑥-direction 𝑛EFGH = (1, 0, 0)𝑇 = 𝑖.
j j
[ ( ) ]
𝜕 𝜌𝑢𝑢
EFGH: 𝑥-component of momentum flux = 𝜌𝑢𝑢 + d𝑥 d𝑦d𝑧 (.)
𝜕𝑥

The momentum flux contributions for the remaining four faces of the fluid element are
obtained in a similar manner by substituting the respective normal vectors and surface
edge lengths into Equation (.).

AEHB: 𝑥-component of momentum flux = − (𝜌 𝑣 𝑢) d𝑥d𝑧 (.)

[ ( ) ]
𝜕 𝜌𝑣𝑢
CGFD: 𝑥-component of momentum flux = 𝜌𝑣𝑢 + d𝑦 d𝑥d𝑧 (.)
𝜕𝑦

ADFE: 𝑥-component of momentum flux = − (𝜌 𝑤 𝑢) d𝑥d𝑦 (.)

[ ( ) ]
𝜕 𝜌𝑤𝑢
BHGC: 𝑥-component of momentum flux = 𝜌𝑤𝑢 + d𝑧 d𝑥d𝑦 (.)
𝜕𝑧

What may be confusing at first is the fact that we have a flow of 𝑥-momentum over
faces which are oriented in 𝑦- and 𝑧-directions. You have to remind yourself that the
mass flux across a surface is ultimately a scalar. Momentum, however, is a vectorial
quantity. Therefore, the mass flux over, for example, the top face BHGC, which is oriented
in positive 𝑧-direction, can carry momentum in 𝑥-direction.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6.2 Conservation of Momentum 63

Summing up all parts of the 𝑥-component of the momentum flux (.) through (.) Total
yields: 𝑥-component of
momentum flux
[ ( ) ]
( 𝑇 ) ( ) 𝜕 𝜌𝑢𝑢
𝜌 𝑣 𝑛 𝑢 d𝑆 = − 𝜌 𝑢 𝑢 d𝑦d𝑧 + 𝜌 𝑢 𝑢 + d𝑥 d𝑦d𝑧
𝜕𝑥
[ ( ) ]
𝜕 𝜌𝑣𝑢
− (𝜌 𝑣 𝑢) d𝑥d𝑧 + 𝜌 𝑣 𝑢 + d𝑦 d𝑥d𝑧 (.)
𝜕𝑦
[ ( ) ]
𝜕 𝜌𝑤𝑢
− (𝜌 𝑤 𝑢) d𝑥d𝑦 + 𝜌 𝑤 𝑢 + d𝑧 d𝑥d𝑦
𝜕𝑧

After simplifying Equation (.), only the differential terms remain on the right-hand
side. They represent the total flux of the 𝑥-component of momentum across the bound-
ary of d𝑉 . Each term is proportional to the volume d𝑉 = d𝑥d𝑦d𝑧 of the fluid element.
[ ( ) ( ) ( )]
( 𝑇 ) 𝜕 𝜌𝑢𝑢 𝜕 𝜌𝑣𝑢 𝜕 𝜌𝑤𝑢
𝑥-component: 𝜌 𝑣 𝑛 𝑢 d𝑆 = + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

The 𝑦- and 𝑧-components of the momentum flux follow in similar fashion using Equa- Fluxes of the 𝑦-
tions (.) and (.) instead of Equation (.): and
𝑧-component of
[ ( ) ( ) ( )] momentum
j ( 𝑇 ) 𝜕 𝜌𝑢𝑣 𝜕 𝜌𝑣𝑣 𝜕 𝜌𝑤𝑣
j
𝑦-component: 𝜌 𝑣 𝑛 𝑣 d𝑆 = + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
[ ( ) ( ) ( )]
( ) 𝜕 𝜌𝑢𝑤 𝜕 𝜌𝑣𝑤 𝜕 𝜌𝑤𝑤
𝑧-component: 𝜌 𝑣𝑇 𝑛 𝑤 d𝑆 = + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

6.2.3 External forces

Two types of external forces may act on a fluid element:

• body forces and

• surface forces

Figure . shows only the 𝑥-components of the external forces for clarity.
Body forces are also called volume forces. They act on the mass element as a whole. Body forces
Gravity and electromagnetic forces are of this type. The only body force we will consider
is the gravity force 𝐹 𝑔 , which we model as mass 𝜌d𝑉 times gravitational acceleration
vector 𝑓 = (𝑓𝑥 , 𝑓𝑦 , 𝑓𝑧 )𝑇 = (0, 0, −𝑔)𝑇 .

𝐹 𝑔 = 𝜌𝑓 d𝑉 (.)

The 𝑥- and 𝑦-components of the weight force (.), 𝜌𝑓𝑥 d𝑉 , actually vanish in Earth’s
gravity field. However, we keep them for the sake of completeness.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

64 6 Navier-Stokes Equations

Figure 6.2 𝑥-components of surface and body forces acting on the fixed, infinitesimal fluid
element d𝑉

Surface forces Surface forces are a result of normal and shear stresses acting on the fluid element.
Collecting the stress related forces in 𝑥-direction for all six faces yields:
[ ]
j 𝜕 𝜎̂ 𝑥𝑥 j
𝑥-comp. of surface force: − 𝜎̂ 𝑥𝑥 d𝑦d𝑧 + 𝜎̂ 𝑥𝑥 + d𝑥 d𝑦d𝑧
𝜕𝑥
[ ]
𝜕𝜏𝑦𝑥
− 𝜏𝑦𝑥 d𝑥d𝑧 + 𝜏𝑦𝑥 + d𝑦 d𝑥d𝑧 (.)
𝜕𝑦
[ ]
𝜕𝜏
− 𝜏𝑧𝑥 d𝑥d𝑦 + 𝜏𝑧𝑥 + 𝑧𝑥 d𝑧 d𝑥d𝑦
𝜕𝑧

Combining the terms and substituting d𝑉 = d𝑥d𝑦d𝑧 results in

[ ]
𝜕 𝜎̂ 𝑥𝑥 𝜕𝜏𝑦𝑥 𝜕𝜏𝑧𝑥
𝑥-component of surface force: + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

Repeating the surface force collection for the other two coordinate directions reveals:

[ ]
𝜕𝜏𝑥𝑦 𝜕 𝜎̂ 𝑦𝑦 𝜕𝜏𝑧𝑦
𝑦-component of surface force: + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
[ ]
𝜕𝜏𝑥𝑧 𝜕𝜏𝑦𝑧 𝜕 𝜎̂ 𝑧𝑧
𝑧-component of surface force: + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

There are nine stresses, but they are symmetric, i.e. 𝜏𝑥𝑦 = 𝜏𝑦𝑥 , 𝜏𝑥𝑧 = 𝜏𝑧𝑥 , and 𝜏𝑦𝑧 = 𝜏𝑧𝑦 .
This leaves us with six stresses which have to be determined in addition to the velocity
components and the pressure.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6.3 Stokes’ Hypothesis 65

6.2.4 Conservation of momentum equations

In the three previous subsections .., .., and .. we derived all parts needed to Conservation of
assemble the actual conservation of momentum equations. The 𝑥-component of the momentum in
𝑥-direction
momentum time rate of change (.) and the momentum flux across the fluid element
boundary (.) form the left-hand side of the conservation of momentum equation. It
represents the total rate of change in 𝑥-momentum for the fluid element d𝑉 . According
to Newton’s second law, this must be equal to the sum of external forces, which consists
of the 𝑥-components of body force (.) and surface force (.). Together, we get the
conservation equation for the 𝑥-component of the momentum:
( ) [ ( ) ( ) ( )]
𝜕 𝜌𝑢 𝜕 𝜌𝑢𝑢 𝜕 𝜌𝑣𝑢 𝜕 𝜌𝑤𝑢
d𝑉 + + + d𝑉
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
[ ]
𝜕 𝜎̂ 𝑥𝑥 𝜕𝜏𝑦𝑥 𝜕𝜏𝑧𝑥
= 𝜌𝑓𝑥 d𝑉 + + + d𝑉 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

Obviously, all terms are proportional to the small but finite volume d𝑉 , which allows
us to divide Equation (.) by d𝑉 :
( ) [ ( ) ( ) ( )] [ ]
𝜕 𝜌𝑢 𝜕 𝜌𝑢𝑢 𝜕 𝜌𝑣𝑢 𝜕 𝜌𝑤𝑢 𝜕 𝜎̂ 𝑥𝑥 𝜕𝜏𝑦𝑥 𝜕𝜏𝑧𝑥
+ + + = 𝜌𝑓𝑥 + + + (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧
j j
The corresponding 𝑦- and 𝑧-components of the conservation of momentum equations Conservation of
are: momentum in 𝑦-
and 𝑧- direction
( ) [ ( ) ( ) ( )] [ ]
𝜕 𝜌𝑣 𝜕 𝜌𝑢𝑣 𝜕 𝜌𝑣𝑣 𝜕 𝜌𝑤𝑣 𝜕𝜏𝑥𝑦 𝜕 𝜎̂ 𝑦𝑦 𝜕𝜏𝑧𝑦
+ + + = 𝜌𝑓𝑦 + + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧
(.)
( ) [ ( ) ( ) ( )] [ ]
𝜕 𝜌𝑤 𝜕 𝜌𝑢𝑤 𝜕 𝜌𝑣𝑤 𝜕 𝜌𝑤𝑤 𝜕𝜏𝑥𝑧 𝜕𝜏𝑦𝑧 𝜕 𝜎̂ 𝑧𝑧
+ + + = 𝜌𝑓𝑧 + + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧
(.)

The three components of the conservation of momentum equations describe the rela- Closure problem
tionships between ten yet unknown flow properties:

. density 𝜌

. three velocity components 𝑢, 𝑣, 𝑤

. six stresses 𝜎̂ 𝑥𝑥 𝜎̂ 𝑦𝑦 , 𝜎̂ 𝑧𝑧 , 𝜏𝑥𝑦 , 𝜏𝑥𝑧 , and 𝜏𝑦𝑧

Together with the continuity equation, we have four equations to solve for ten unknowns.
Obviously, we have to gather more equations to properly model the problem and to
ultimately obtain a solution.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

66 6 Navier-Stokes Equations

6.3 Stokes’ Hypothesis

Hooke’s Law The same problem occurs in mechanics of materials with the equilibrium of elastic
material: more unknowns than equations. The missing equations are derived from
observations and material properties. Robert Hooke (* – †), a th century
English natural philosopher and architect, stated that the normal stress 𝜎 in elastic
bodies is proportional to the change in length as a fraction of total length called strain
𝜀: 𝜎 = 𝐸𝜀, the material property being the modulus of elasticity 𝐸.
Newtonian fluids Newton was the first to show that the shear stress in some fluids is a function of the
rate of change of strain, i.e. the shear stress depends on how fast the fluid is deformed.
𝜕𝑢
𝜏=𝜇 (.)
𝜕𝑦
He derived this theoretically, based on flow observations and basic assumptions. How-
ever, experiments later confirmed this approach. In honor of his discovery, fluids are
called Newtonian fluids if they have a linear relationship between the stresses and the
rates of change of strain. The constant of proportionality is the dynamic viscosity 𝜇
which is measured in [Pa s] or [(N s)/m2 ]. For flow conditions typically encountered
in ship hydrodynamics, water and air are considered Newtonian fluids. We further
assume that the fluid is isotropic, i.e. the dynamic viscosity does not depend on the
direction in which it is measured.
In analogy to the general stress–strain relationship for elastic bodies, a relationship
j is derived for fluids which links stresses with the rate of change of strain. Note the j
difference!

• Elastic body: stress is proportional to the deformation (strain).

• Fluid: stress is a function of how fast (rate of change of strain) the fluid is deformed.

This exploits physical observations and mathematical properties of tensors.

The Irish mathematician and physicist Sir George Gabriel Stokes (* – †) intro-
duced the necessary equations for normal and shear stresses (.) during a lecture in
 (Stokes, ).
[ ]
𝑇 𝜕𝑢 𝜕𝑣 𝜕𝑢
𝜎̂ 𝑥𝑥 = −𝑝 + 𝜆∇ 𝑣 + 2𝜇 𝜏𝑥𝑦 = 𝜏𝑦𝑥 = 𝜇 +
𝜕𝑥 𝜕𝑥 𝜕𝑦
[ ]
𝜕𝑣 𝜕𝑢 𝜕𝑤
𝜎̂ 𝑦𝑦 = −𝑝 + 𝜆∇𝑇 𝑣 + 2𝜇 𝜏𝑥𝑧 = 𝜏𝑧𝑥 = 𝜇 + (.)
𝜕𝑦 𝜕𝑧 𝜕𝑥
[ ]
𝜕𝑤 𝜕𝑤 𝜕𝑣
𝜎̂ 𝑧𝑧 = −𝑝 + 𝜆∇𝑇 𝑣 + 2𝜇 𝜏𝑦𝑧 = 𝜏𝑧𝑦 = 𝜇 +
𝜕𝑧 𝜕𝑦 𝜕𝑧
For convenience, normal stresses are split into three parts:

. The pressure 𝑝, which is the negative mean of the three normal stresses:
1( )
𝑝 = − 𝜎̂ 𝑥𝑥 + 𝜎̂ 𝑦𝑦 + 𝜎̂ 𝑧𝑧 (.)
3
It is common to all coordinate directions.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6.4 Navier-Stokes Equations for a Newtonian Fluid 67

. A term which is proportional to the volume dilation ∇𝑇 𝑣, i.e. the divergence of

velocity.

. The part which is proportional to the rate of change of strain in the respective
coordinate direction, e.g. 𝜕𝑢∕𝜕𝑥.

By separating the pressure from the remaining normal stress, the limiting case for
vanishing velocity lets the equations collapse to the hydrostatic case.
A more elaborate discussion about the analogy of the stress–rate of change of strain
relationship for fluids and the stress–strain relationship of elastic bodies can be found
in Schlichting and Gersten (, Chapter ).
Two material properties appear in the Equations (.) for the stresses: Stokes’
hypothesis

• the dynamic viscosity 𝜇 and

• the second viscosity 𝜆.

Determining these properties is a task of the scientific field of rheology.

In principle, the viscosity can depend on time, position, as well as the state of a flow
field itself (temperature, stress, velocity, etc.). Luckily for us, air and water belong to
j the class of Newtonian fluids. For Newtonian fluids, the material properties 𝜇 and 𝜆 j
are independent of the actual stress and velocity values.
Dynamic viscosity differs from fluid to fluid. Viscosities of water and air are a function
of temperature (see Section ..). Very little is known about the second viscosity 𝜆.
Since it is associated with changes in volume, it cannot be observed with incompressible
fluids. Remember that the divergence of the velocity vanishes for fluids with constant
density: ∇𝑇 𝑣 = 0.
Stokes solved the difficulty posed by the unknown second viscosity by drawing an
analogy to elastic bodies. He postulated the relationship 3𝜆 + 2𝜇 = 0 between dynamic
and second viscosity, which has been named in his honor:

2
Stokes’ hypothesis for second viscosity 𝜆 = − 𝜇 (.)
3

Our experience shows that the resulting Navier-Stokes equations describe viscous
fluid flow correctly over a wide range of applications. However, the validity of Stokes’
hypothesis is still debated among experts. See for instance Gad-el Hak ().

6.4 Navier-Stokes Equations for a Newtonian Fluid

Finally, substituting the stress–rate of change of strain relationships (.) together Navier-Stokes
with Stokes’ hypothesis (.) into the conservation of momentum equations (.), equations
(.), and (.) yields the Navier-Stokes equations (NSE) for an isotropic Newtonian

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

68 6 Navier-Stokes Equations

fluid in their conservative, differential form, which describe viscous flows.

( ) [ ( ) ( ) ( )]
𝜕 𝜌𝑢 𝜕 𝜌𝑢𝑢 𝜕 𝜌𝑣𝑢 𝜕 𝜌𝑤𝑢 𝜕𝑝
+ + + = 𝜌𝑓𝑥 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥
[ ] [ ( )] [ ( )]
𝜕 2 𝑇 𝜕𝑢 𝜕 𝜕𝑣 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑤
− 𝜇∇ 𝑣 − 2𝜇 + 𝜇 + + 𝜇 + (.)
𝜕𝑥 3 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥

( ) [ ( ) ( ) ( )]
𝜕 𝜌𝑣 𝜕 𝜌𝑢𝑣 𝜕 𝜌𝑣𝑣 𝜕 𝜌𝑤𝑣 𝜕𝑝
+ + + = 𝜌𝑓𝑦 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑦
[ ( )] [ ] [ ( )]
𝜕 𝜕𝑣 𝜕𝑢 𝜕 2 𝑇 𝜕𝑣 𝜕 𝜕𝑤 𝜕𝑣
+ 𝜇 + − 𝜇∇ 𝑣 − 2𝜇 + 𝜇 + (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 3 𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕𝑧

( ) [ ( ) ( ) ( )]
𝜕 𝜌𝑤 𝜕 𝜌𝑢𝑤 𝜕 𝜌𝑣𝑤 𝜕 𝜌𝑤𝑤 𝜕𝑝
+ + + = 𝜌𝑓𝑧 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧
[ ( )] [ ( )] [ ]
𝜕 𝜕𝑢 𝜕𝑤 𝜕 𝜕𝑤 𝜕𝑣 𝜕 2 𝑇 𝜕𝑤
+ 𝜇 + + 𝜇 + − 𝜇∇ 𝑣 − 2𝜇 (.)
𝜕𝑥 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 3 𝜕𝑧

All terms in equations (.), (.), and (.) represent forces per unit fluid volume.
j Figure . identifies the force components in the NSE. On the left-hand side we have j
the inertia force and on the right-hand side we see the sum of external forces which
comprises body and surface forces.
C.L. Navier, G.G. The Navier-Stokes equations are named after Stokes and Navier, who published a form
Stokes of the NSE earlier than Stokes (Navier, ). However, Navier had no physical concept
for the material parameter we now know as viscosity (Navier, ).
Mathematical The Navier-Stokes equations are a set of coupled, nonlinear partial differential equations.
classification They are nonlinear because of the convection terms in brackets on the left side of (.),
(.), and (.), which contain products of the unknown velocity components. If you
also consider the density and viscosity distributions as unknown, then more terms are
nonlinear. Luckily, we may assume that kinematic viscosity and density are constant
for most problems in ship hydrodynamics. We will exploit this in the next chapter to
simplify the equations.
Even today we know very little about the existence and uniqueness of solutions to
the Navier-Stokes equations. The Clay Mathematical Institute (CMI) in Cambridge,
MA made the Navier-Stokes equations one of their famous Millenium Problems. A
contribution which represents substantial progress toward a mathematical theory which
will unlock the secrets hidden in the Navier-Stokes equations can earn you a $ million
award (see http://www.claymath.org/millennium-problems/navier-stokes-equation).
NSE solvers All modern approaches of CFD in ship hydrodynamics actually do not solve the equa-
tions above but a simplified version of them called the Reynolds averaged Navier-Stokes
equations or short RANSE (see Chapter ). The key is that the time dependent flow
velocity is subdivided into a time dependent average velocity and a highly fluctuating
turbulence. This leads to an additional term in the equations. It has the form of a stress

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

6.4 Navier-Stokes Equations for a Newtonian Fluid 69

j j

Figure 6.3 Forces comprising the Navier-Stokes equations for an isotropic Newtonian fluid

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

70 6 Navier-Stokes Equations

tensor and in honor of Osborne Reynolds is named the Reynolds stress tensor. For
the newly introduced unknowns additional equations are needed, which are known as
turbulence models.

References
Anderson, Jr., J. (). Solutions manual to accompany computational fluid dynamics
– The basics with applications. McGraw-Hill, Inc., New York, NY.
Gad-el Hak, M. (). Stokes’ hypothesis for a Newtonian, isotropic fluid. Journal of
Fluids Engineering, ():–.
Navier, C. (). Mémoire sur les lois du mouvement des fluides. Mém. Acad. Sci. Inst.
France, :–.
Schlichting, H. and Gersten, K. (). Boundary-layer theory. Springer-Verlag, Berlin,
Heidelberg, New York, eigth edition. Corrected printing.
Stokes, G. (). On the theories of the internal friction of fluids in motion, and of the
equilibrium and motion of elastic solids. Transactions of the Cambridge Philosophical
Society, :–.

Self Study Problems

j j
. Draw duplicates of the fluid element d𝑉 in Figure . and sketch in momentum
fluxes for 𝑦 and 𝑧 directions. Follow the steps of Equations (.) through (.)
and derive the momentum flux in 𝑦- and 𝑧-direction. You should arrive at the
result shown in Equations (.) and (.).
. Summarize in your own words Stokes’ hypothesis.
. Why are we not able to determine the second viscosity 𝜆 for water?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

71

7
Special Cases of the Navier-Stokes
Equations

The Navier-Stokes equations developed in the previous chapter contain the fluid prop-
erties density 𝜌 and dynamic viscosity 𝜇. The density is also part of the continuity
equation (.). We will simplify both the continuity equation and the NSE for the ship
resistance and propulsion problem by making reasonable assumptions about density
and viscosity. A dimensionless form of the NSE will provide insight into the different
scales of the forces driving the flow.

j
Learning Objectives j

At the end of this chapter students will be able to:

• find the effects of constant density on conservation of mass and momentum

• identify the terms in the NSE as forces per unit mass

• formulate a dimensionless version of the NSE

• introduce characteristic, dimensionless parameters

• explain the challenge of high Reynolds numbers

7.1 Incompressible Fluid of Constant Temperature

Ships sail at the phase boundary of water and air. We discussed important properties Constant
of these fluids in Chapter . Besides salinity, density and dynamic viscosity of water are temperature
mainly a function of fluid temperature. At the speeds achievable by ships, friction is
too low to cause a measurable change in water temperature. Therefore, it is opportune
to treat fluid temperature as constant.
Salinity tends to be uniform in a stretch of ocean or waterway. Thus, if the temperature Constant
does not change either, we can assume that the dynamic viscosity is constant for a viscosity and
density
specific flow problem. We also observe that water changes its volume only slightly even
under high pressure. With small dependency on pressure and constant temperature we

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

72 7 Special Cases of the Navier-Stokes Equations

treat the density of water as a constant as well. In particular, we assume that density is
independent of space and time and commonly say water is incompressible.
Air changes its volume a lot more easily under pressure. Just think of the hand pump
you use to inflate a ball. Even if you block the nozzle you can move the piston quite a
bit, thereby compressing the air within the pump. However, for the modest velocities
of ships, variations in pressure are not big enough to cause considerable changes in air
density. Therefore, air is treated as incompressible as well.
Continuity With density 𝜌 and dynamic viscosity 𝜇 fixed, continuity and Navier-Stokes equations
equation for may be simplified. We start with the conservative, differential form of the continuity
incompressible
fluid
equation (.):
𝜕𝜌
+ ∇ 𝑇 (𝜌𝑣) = 0 (.)
𝜕𝑡
Obviously, the time derivative of the density (first term) will vanish, if the density does
not change over time. From the second term we factor out the space independent
density 𝜌:

0
𝜕𝜌◁
◁ + 𝜌∇𝑇𝑣 = 0
◁𝜕𝑡

Finally, we divide by the density, which is always nonzero. Otherwise there would be a
vacuum and we would not be discussing fluid mechanics.
j j

∇𝑇 𝑣 = 0 (.)

The final expression (.) is the continuity equation for an incompressible fluid. The
left-hand side is the divergence of the velocity which we introduced previously. In
Section .., we discovered that the divergence of the velocity is related to the change
in volume. If the divergence of velocity vanishes, a given mass will always occupy the
same volume. Imagine a balloon filled with water. You can change its form but not its
volume, i.e. the fluid is incompressible.
NSE for constant Now let’s consider the components of the Navier-Stokes equations (.), (.), and
density and (.). Constant density 𝜌 and dynamic viscosity 𝜇 can be extracted from the derivatives
viscosity
on the left-hand and right-hand sides:
[ ]
𝜕𝑢 𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 𝜕𝑝
𝜌 +𝜌 + + = 𝜌𝑓𝑥 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥
[ ] [ ] [ ]
𝜕 2 𝑇 𝜕𝑢 𝜕 𝜕𝑣 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑤
−𝜇 ∇ 𝑣−2 +𝜇 + +𝜇 +
𝜕𝑥 3 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥

[ ]
𝜕𝑣 𝜕(𝑢 𝑣) 𝜕(𝑣 𝑣) 𝜕(𝑤 𝑣) 𝜕𝑝
𝜌 +𝜌 + + = 𝜌𝑓𝑦 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑦
[ ] [ ] [ ]
𝜕 𝜕𝑣 𝜕𝑢 𝜕 2 𝑇 𝜕𝑣 𝜕 𝜕𝑤 𝜕𝑣
+𝜇 + −𝜇 ∇ 𝑣−2 +𝜇 + (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 3 𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

7.1 Incompressible Fluid of Constant Temperature 73

[ ]
𝜕𝑤 𝜕(𝑢 𝑤) 𝜕(𝑣 𝑤) 𝜕(𝑤 𝑤) 𝜕𝑝
𝜌 +𝜌 + + = 𝜌𝑓𝑧 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧
[ ] [ ] [ ]
𝜕 𝜕𝑢 𝜕𝑤 𝜕 𝜕𝑤 𝜕𝑣 𝜕 2 𝑇 𝜕𝑤
+𝜇 + +𝜇 + −𝜇 ∇ 𝑣−2
𝜕𝑥 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 3 𝜕𝑧

The viscous forces on the right-hand side contain a term with the divergence of velocity.
According to the continuity equation for incompressible fluids, this term vanishes as
shown here for the 𝑥-component of Equation (.).
[ ]
𝜕𝑢 𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 𝜕𝑝
𝜌 +𝜌 + + = 𝜌𝑓𝑥 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥
[ ] [ ]
=0 [ ]
𝜕 2 𝑇 𝜕𝑢 𝜕 𝜕𝑣 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑤
(.)
>

−𝜇 ∇ 𝑣 − 2 +𝜇 + +𝜇 +
𝜕𝑥 3  𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥

All components are divided by the density 𝜌, and we introduce the kinematic viscosity
𝜈 = 𝜇∕𝜌.

𝜕𝑢 𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 1 𝜕𝑝

+ + + = 𝑓𝑥 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜌 𝜕𝑥
[ ] [ ] [ ]
𝜕 𝜕𝑢 𝜕 𝜕𝑣 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑤
−𝜈 −2 +𝜈 + +𝜈 +
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥
j j

𝜕𝑣 𝜕(𝑢 𝑣) 𝜕(𝑣 𝑣) 𝜕(𝑤 𝑣) 1 𝜕𝑝

+ + + = 𝑓𝑦 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜌 𝜕𝑦
[ ] [ ] [ ]
𝜕 𝜕𝑣 𝜕𝑢 𝜕 𝜕𝑣 𝜕 𝜕𝑤 𝜕𝑣
+𝜈 + −𝜈 −2 +𝜈 + (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑦 𝜕𝑧

𝜕𝑤 𝜕(𝑢 𝑤) 𝜕(𝑣 𝑤) 𝜕(𝑤 𝑤) 1 𝜕𝑝

+ + + = 𝑓𝑧 −
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜌 𝜕𝑧
[ ] [ ] [ ]
𝜕 𝜕𝑢 𝜕𝑤 𝜕 𝜕𝑤 𝜕𝑣 𝜕 𝜕𝑤
+𝜈 + +𝜈 + −𝜈 −2
𝜕𝑥 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑧

Next, we further differentiate the velocity derivatives on the right-hand side with respect
to the coordinate directions 𝑥, 𝑦, and 𝑧 and reorganize the resulting second order
derivatives. For brevity, we show this transformation only for the 𝑥-component of the
NSE.
𝜕𝑢 𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢)
+ + +
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
2
[ 2 ] [ 2 ]
1 𝜕𝑝 𝜕 𝑢 𝜕 𝑣 𝜕2𝑢 𝜕 𝑢 𝜕2𝑤
= 𝑓𝑥 − + 2𝜈 +𝜈 + +𝜈 +
𝜌 𝜕𝑥 𝜕𝑥2 𝜕𝑥𝜕𝑦 𝜕𝑦2 𝜕𝑧2 𝜕𝑥𝜕𝑧
( 2 2 2
) ( 2 )
1 𝜕𝑝 𝜕 𝑢 𝜕 𝑢 𝜕 𝑢 𝜕 𝑢 𝜕2𝑣 𝜕2𝑤
= 𝑓𝑥 − + 𝜈 + + +𝜈 + + (.)
𝜌 𝜕𝑥 𝜕𝑥2 𝜕𝑦2 𝜕𝑧2 𝜕𝑥2 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

74 7 Special Cases of the Navier-Stokes Equations

The first pair of parentheses on the right of (.) contains the sum of second order
derivatives of the horizontal velocity component 𝑢. We summarize this term using the
𝜕2 𝜕2 𝜕2
scalar Laplace operator 𝚫 = 𝜕𝑥2
+ 𝜕𝑦2
+ 𝜕𝑧2

𝜕2𝑢 𝜕2𝑢 𝜕2𝑢

𝚫𝑢 = + + (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2

The second pair of parentheses on the right of Equation (.) is the 𝑥-derivative of the
divergence of velocity which – you guessed correctly – vanishes for incompressible
fluids (.):
( 2 ) ( )
𝜕 𝑢 𝜕2𝑣 𝜕2𝑤 𝜕 𝜕𝑢 𝜕𝑣 𝜕𝑤
+ + = + +
𝜕𝑥2 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
= ∇𝑇𝑣
=0
𝜕 ( 𝑇
>)
(.)

= ∇ 𝑣
𝜕𝑥 
= 0

The continuity equation allows us to simplify the left-hand side as well. We first apply
the product rule of differentiation to the convective derivative:
j 𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑤 j
+ + = 𝑢 +𝑢 + 𝑣 +𝑢 + 𝑤 +𝑢
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

Reordering the summands results in

( )
𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑢 ⁓= 0
𝜕𝑣𝜕𝑤

+ + =𝑢 +𝑣 +𝑤 +𝑢 +  +
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

The last term in parenthesis is once more the divergence of the velocity and vanishes
for constant density 𝜌.

𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) 𝜕𝑢 𝜕𝑢 𝜕𝑢

+ + = 𝑢 +𝑣 +𝑤
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧
The right-hand side may now be rewritten in a convenient vector form:
𝜕(𝑢 𝑢) 𝜕(𝑣 𝑢) 𝜕(𝑤 𝑢) ( )
+ + = 𝑣𝑇 ∇ 𝑢 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
( )
Note that the expression 𝑣𝑇 ∇ is a scalar differential operator and is not the divergence
of velocity.
Substituting the results of Equations (.), (.), and (.) back into Equation (.) yields
the 𝑥-component of the NSE for incompressible fluids.
𝜕𝑢 ( ) 1 𝜕𝑝
+ 𝑣𝑇 ∇ 𝑢 = 𝑓𝑥 − + 𝜈𝚫𝑢 (.)
𝜕𝑡 𝜌 𝜕𝑥

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

7.2 Dimensionless Navier-Stokes Equations 75

We perform the same transformations for the 𝑦- and 𝑧-components of the NSE:

𝜕𝑣 ( ) 1 𝜕𝑝
+ 𝑣𝑇 ∇ 𝑣 = 𝑓𝑦 − + 𝜈𝚫𝑣 (.)
𝜕𝑡 𝜌 𝜕𝑦
𝜕𝑤 ( ) 1 𝜕𝑝
+ 𝑣𝑇 ∇ 𝑤 = 𝑓𝑧 − + 𝜈𝚫𝑤 (.)
𝜕𝑡 𝜌 𝜕𝑧

The three components of the NSE may now be assembled into a convenient vector Vector form of
equation: NSE
𝜕𝑣 ( ) 1
(𝜌, 𝜈 = const.)
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣 (.)
𝜕𝑡 ⏟⏟ ⏟ 𝜌 ⏟ ⏟ ⏟
⏟⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏟ ⏟⏟⏟
1 2 4
3
This looks a lot more compact than the Navier-Stokes equations we developed first in
Chapter . The individual terms still represent the same forces as in Figure .. This
time, however, all forces are given as accelerations, i.e. force per unit mass.

. Inertia force on the left-hand side.

. Vector of body forces 𝑓 .

1
. Pressure force 𝜌
∇ 𝑝.
j j
. Viscous force 𝜈𝚫𝑣.

Equation (.) represents the conservation of momentum for an incompressible New-

tonian fluid at constant temperature. Together with the continuity equation (.), it
describes the motion of water around ships. We now have four equations – the continu-
ity equation and the three components of the NSE – which in principle allow us to find
the four unknowns: pressure 𝑝 and the components 𝑢, 𝑣, and 𝑤 of the velocity vector.
In practice, most available CFD systems do not attempt to solve Equation (.) directly.
For high Reynolds numbers this direct numerical simulation (DNS) is still beyond our
computational capabilities. Instead, we solve the RANSE, which are derived by splitting
the instantaneous time dependent velocity into an average velocity (still time dependent)
and the turbulent velocity. This introduces the initially unknown Reynolds stress tensor.
The added unknowns require additional equations to represent the turbulence, so-
called turbulence models. The latter are still a focus of ongoing research and several
turbulence models of varying complexity are in use for ship resistance computations.
The interested reader can find a derivation of the RANSE form of the Navier-Stokes
equations for incompressible flow in Chapter .

7.2 Dimensionless Navier-Stokes Equations

Additional insight into the Navier-Stokes equations can be gained by rewriting them in a
dimensionless form. This is general practice for implementation of numerical solutions
methods. Therefore, this example might suffice to show this useful technique.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

76 7 Special Cases of the Navier-Stokes Equations

We revisit the NSE (.):

𝜕𝑣 ( ) 1
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣 (.)
𝜕𝑡 𝜌
As mentioned, both sides represent accelerations and are measured in the physical unit
m/s2 .
Dimensionless We make quantities dimensionless by dividing them by a suitable reference value. For
lengths and ∇ instance, all space variables are divided by a reference length 𝐿:
𝑥 𝑦 𝑧
𝑥∗ = 𝑦∗ = 𝑧∗ = (.)
𝐿 𝐿 𝐿
This seems fairly obvious. Less apparent is that the Nabla operator has a dimension as
well. The dimensionless Nabla operator ∇ ∗ is equal to

⎛ 𝜕 ⎞
⎜ ( ) ⎟
𝑥
⎛ 𝜕 ⎞ ⎜ 𝜕 𝐿 ⎟ ⎛𝐿𝜕 ⎞
⎜ 𝜕𝑥∗ ⎟ ⎜ ⎟ ⎜ 𝜕𝑥 ⎟
⎜ 𝜕 ⎟ ⎜ (𝜕 ) ⎟ ⎜ ⎟
∇∗ = ⎜ ∗ ⎟ = ⎜ 𝑦 ⎟ = ⎜𝐿𝜕 ⎟ = 𝐿∇ (.)
𝜕𝑦 𝜕
⎜ ⎟ ⎜ 𝐿 ⎟ ⎜ 𝜕𝑦 ⎟
⎜ 𝜕 ⎟ ⎜ 𝜕 ⎟ ⎜𝐿𝜕 ⎟
⎝ 𝜕𝑧∗ ⎠ ⎜ (𝑧) ⎟ ⎝ 𝜕𝑧 ⎠
⎜ 𝜕 ⎟
⎝ 𝐿 ⎠
j j
Dimensionless The dimensionless Laplace operator is derived from the identity 𝚫 = ∇ 𝑇 ∇ :
Laplace operator
𝚫 ( )𝑇 ( )𝑇
𝚫∗ = ∇∗ ∇∗ = 𝐿∇ 𝐿∇ = 𝐿2 𝚫 (.)

Any geometric characteristic of the flow problem may serve as reference length 𝐿, for
example the length or beam of a vessel, the diameter of a propeller, the waterdepth, and
others.
Dimensionless We also introduce a reference period 𝑇 . Possible choices are the duration of a process
time or the period of a recurring event. Our dimensionless time 𝑡∗ and partial derivative will
be
𝑡 𝜕 𝜕
𝑡∗ = = 𝑇 (.)
𝑇 𝜕𝑡∗ 𝜕𝑡

Dimensionless For convenience, we will select a reference velocity of magnitude 𝑈∞ , e.g. ship or
velocity flow speed. In practice, reference length and reference period can be used to define a
reference velocity or length and velocity constitute a reference time. The dimensionless
velocity vector becomes

⎛ 𝑢 ⎞
⎛ 𝑢∗ ⎞ ⎜ 𝑈∞ ⎟ ⎛ 𝑢 ⎞
⎜ ⎟ ⎜ 𝑣 ⎟ ⎜ ⎟ 𝑣
1
𝑣∗ = ⎜ 𝑣∗ ⎟ = ⎜ 𝑈 ⎟ = ⎜ 𝑣 ⎟ = (.)
⎜ ⎟ ⎜ ∞ ⎟ 𝑈∞ ⎜ ⎟ 𝑈∞
⎜ ∗⎟ ⎜ 𝑤 ⎟ ⎜ ⎟
⎝𝑤 ⎠ ⎜ ⎟ ⎝𝑤⎠
⎝ 𝑈∞ ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

7.2 Dimensionless Navier-Stokes Equations 77

Finally, we select a reference acceleration for the vector of body forces (per unit mass) Dimensionless
𝑓 . The gravity force is the only body force of note we will consider. Therefore, we use gravitation and
pressure
the gravitational acceleration 𝑔 = 9.807 m/s2 as a reference value. For the pressure we
select 𝑝∞ , which often represents the pressure far away from the body.
𝑓 𝑝
𝑓∗ = 𝑝∗ = (.)
𝑔 𝑝∞

The NSE (.) become dimensionless by replacing the dimensional quantities with Dimensionless
their dimensionless counterparts and reference quantities, i.e. 𝑥 = 𝐿𝑥∗ , 𝑣 = 𝑈∞ 𝑣∗ , NSE
𝑝 = 𝑝∞ 𝑝∗ , and so on.
( ∗
) [ ( ∗ )]
1 𝜕 𝑈∞ 𝑣 ( )
∗ 𝑇
∇ ( )
∗
+ 𝑈∞ 𝑣 𝑈∞ 𝑣∗
𝑇 𝜕𝑡 𝐿
( ∗)
∇∗( ) 𝚫 ( )
= 𝑔𝑓 ∗ − 𝑝∞ 𝑝∗ + 𝜈 𝑈∞ 𝑣∗ (.)
𝜌𝐿 𝐿2
We factor out the reference quantities in each term:
𝑈∞ 𝜕𝑣∗ 𝑈∞2
( ∗𝑇 ∗) ∗ 𝑝 ( ) 𝜈𝑈∞ ( ∗ ∗ )
+ 𝑣 ∇ 𝑣 = 𝑔𝑓 ∗ − ∞ ∇ ∗ 𝑝∗ + 𝚫 𝑣 (.)
𝑇 𝜕𝑡∗ 𝐿 𝜌𝐿 𝐿2
Finally, this equation is multiplied by 𝐿∕𝑈∞2 and we obtain:

[ ] ∗ [ ] [ ] [ ]
j 𝐿 𝜕𝑣 ( ∗𝑇 ∗) ∗ 𝑔𝐿 𝑝 ∞ 𝜈 j
+ 𝑣 ∇ 𝑣 = ∗
𝑓 − ∗ ∗
∇ 𝑝 + 𝚫∗ 𝑣∗ (.)
𝑇 𝑈∞ 𝜕𝑡∗ 𝑈∞2 𝜌 𝑈∞2 𝐿𝑈∞

All factors in (.) are now dimensionless, including the four terms in brackets which Characteristic
are important numbers characterizing the flow around a ship hull. numbers of the
flow
. On the left-hand side we have a form of the Strouhal number Strouhal number
𝐿
𝑆𝑡 = Strouhal number (.)
𝑇 𝑈∞
Vincenz Strouhal (*–†) was a Czech physicist who introduced this di-
mensionless number in  (Strouhal, ).
You probably have heard taught wires (rigging, guides of a cell tower) hum in
high winds. The sound is created when the periodic shedding of vortices in the
flow causes the wire to vibrate. The Strouhal number 𝑆𝑡 is used to describe the
frequency of unsteady processes like vortex shedding. Therefore, it precedes the
local time derivative of the velocity 𝜕𝑣∗ ∕𝜕𝑡∗ , which represents the acceleration in
the dimensionless Navier-Stokes equations (.).
. The first dimensionless number on the right-hand side is connected to the Froude Froude number
number 𝐹𝑟.
[ ]
𝑔𝐿 1
= with
𝑈∞2 𝐹𝑟2
𝑈
𝐹𝑟 = √ ∞ Froude number (.)
𝑔𝐿

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

78 7 Special Cases of the Navier-Stokes Equations

The Froude number 𝐹𝑟 is named after William Froude, the father of modern
ship model testing. Naval architects use the Froude number as a dimensionless
velocity. It is connected to the gravity forces and important for the similarity of
wave patterns.
Euler number . The next pair of brackets encloses a form of Euler number named after Leonard
Euler. [ ]
𝑝∞
𝐸𝑢 = Euler number of fluid mechanics (.)
𝜌 𝑈∞2

Euler made many important contributions to mathematics, structural mechanics,

and fluid mechanics, among others, and various theories, equations, as well as
numbers bear his name. In practice, we are more likely to encounter dimen-
sionless pressure coefficients 𝐶𝑝 which compare pressure differences with the
dynamic pressure (1∕2)𝜌 𝑈∞ 2:

𝑝 − 𝑝∞
𝐶𝑝 = pressure coefficient (.)
1 2
2
𝜌 𝑈∞

Reynolds . The last dimensionless number in brackets is the reciprocal of the Reynolds
number number 𝑅𝑒:
[ ]
𝜈 1
= with
𝐿𝑈∞ 𝑅𝑒
𝐿𝑈∞
j 𝑅𝑒 = Reynolds number (.) j
𝜈
The Reynolds number is named after Osborne Reynolds, who studied, among
many other things, the flow in pipes and under which conditions the flow tran-
sitioned from laminar to turbulent. The dimensionless Reynolds number is
obviously connected to the viscous forces caused by the viscosity of the water.

Froude number and Reynolds number are arguably the most important qualifiers of
flow conditions in ship hydrodynamics. We will encounter these numbers many times
in subsequent chapters.
Dimensionless We now substitute the dimensionless numbers 𝑆𝑡, 𝐹𝑟, 𝐸𝑢, and 𝑅𝑒 for the square brackets
NSE into the Navier-Stokes equations:
𝜕𝑣∗ ( )
1 ∗ 1 ∗ ∗
𝑆𝑡 + 𝑣∗ 𝑇 ∇ ∗ 𝑣∗ =
𝑓 − 𝐸𝑢 ∇ ∗ 𝑝∗ + 𝚫 𝑣 (.)
𝜕𝑡∗ 𝐹𝑟2 𝑅𝑒
Example An example will teach us something about the terms in the Navier-Stokes equations.
Consider the following data set:

length of vessel 𝐿𝑊𝐿 = 120.00 m

ship speed 𝑣𝑠 = 20.00 kn
period 𝑇 = 6.59 s
gravitational acceleration 𝑔 = 9.81 m/s2
density of salt water at  𝑜 C 𝜌𝑠 = 1026.021 kg/m3
standard atmospheric pressure 𝑝∞ = 101325.00 Pa
kinematic viscosity of seawater at  𝑜 C 𝜈 = .⋅10−6 m2 /s

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

7.2 Dimensionless Navier-Stokes Equations 79

We will employ the numbers to calculate the four dimensionless constants of the Navier-
Stokes equations.

. Strouhal number – Before we substitute the values above into Equation (.),
we must convert the ship speed into SI units. One knot is equal to one nautical
mile per hour, i.e. kn = M/h and
1852 m/M
1 kn = = 0.51444 m h/(s M) (.)
3600 s/h
Therefore, the ship speed we want to use as a reference velocity is

𝑣𝑠 = 20 kn = 10.289 m/s = 𝑈∞

As a reference period, we use the period of the transverse waves generated by the
ship. We will discuss this in depth later. With the reference length 𝐿 = 120 m,
reference period 𝑇 = 6.59 s, and reference velocity 𝑈∞ = 10.289 m/s we obtain a
Strouhal number of
𝐿 120 m
𝑆𝑡 = = = 1.76984
𝑇 𝑈∞ 6.59 s 10.289 m/s

It is good practice to check that the result is truly dimensionless.

. Froude number – With the preparation above, the Froude number poses no
j challenge: j
𝑈 10.289 m/s
𝐹𝑟 = √ ∞ = √ = 0.29988
𝑔𝐿 9.81 m/s2 ⋅ 120 m
Please note, that the Froude number is usually computed on the basis of length
in waterline 𝐿𝑊𝐿 .
The coefficient for the Navier-Stokes equations is
1
= 0.29988−2 = 11.12
𝐹𝑟2

. Euler number – In this example we get for the Euler number

𝑝∞ 101325 kg/(ms2 )
𝐸𝑢 = = = 0.93288
2
𝜌𝑈∞ 1026.021 kg/m3 ⋅ 10.2892 (m/s)2

. Reynolds number – The Reynolds number is:

𝐿𝑈∞ 120 m ⋅ 10.289 m/s
𝑅𝑒 = = = 1.03823 ⋅ 109
𝜈 1.1892 ⋅ 10−6 m2 /s
Note that the Reynolds number is usually based on the length over wetted surface
𝐿𝑂𝑆 .
The coefficient for the Navier-Stokes equations is the inverse of the Reynolds
number
1 1
= = 9.632 ⋅ 10−10
𝑅𝑒 1.03823 ⋅ 109

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

80 7 Special Cases of the Navier-Stokes Equations

We rewrite the Navier-Stokes equations using the coefficients we computed for this
example:

𝜕𝑣∗ ( )
1.76984 + 𝑣∗ 𝑇 ∇ ∗ 𝑣∗
𝜕𝑡∗
= 11.12 𝑓 ∗ − 0.93288 ∇ ∗ 𝑝∗ + 9.632 ⋅ 10−10 𝚫∗ 𝑣∗ (.)

Three of the coefficients are in the range of 1, but the coefficient for the body forces is
of magnitude 10, and the coefficient for the viscous forces is very small, roughly 10−9 .
Thus, the factor for the body forces is 10 000 000 000 times larger than the factor for
the viscous forces! This seems to indicate that viscous forces may be negligible.
Indeed, there are flow phenomena which support this notion. We know from detailed
observations that waves created by storms in the vicinity of the Antarctic Circle travel
halfway around the world to hit the shores of the North American Pacific coast. Their
height diminishes somewhat along the way but more due to adverse winds than friction.
On the other hand, it is a well known fact that if you give a boat a push, it will come to
rest again after a short while. Even if we avoid the generation of waves by considering a
well streamlined, submerged body, it will come to rest fairly soon without a continuous
propulsive force.
Boundary layer This apparent discrepancy between theory, in the form of the coefficients in the Navier-
j theory Stokes equations, and the reality of observations vexed scientists and engineers at the j
end of the th century. It was Ludwig Prandtl who proposed in  to divide the
flow around bodies into two regions: a thin sheet of fluid close to the body surface,
called the boundary layer, where viscous effects are present and an exterior flow outside
the boundary layer where viscous effects are mostly negligible. Prandtl introduced the
concept of fluid molecules sticking to the surface and also offered an explanation for
the phenomenon of flow separation. More on this later. Readers are encouraged to read
the paper by Anderson, Jr. () on Prandtl’s boundary layer theory and its impact on
aerodynamics and fluid mechanics.
Prandtl’s ideas allow a simplification of the Navier-Stokes equations into the boundary
layer equations. Although the boundary layer equations are a special case of the Navier-
Stokes equations, we will discuss them in subsequent chapters. They form the basis of
skin friction computations and provide important insights into the flow around ship
hulls.

References
Anderson, Jr., J. (). Ludwig Prandtl’s boundary layer. Physics Today, ():–.
Strouhal, V. (). Über eine besondere Art der Tonerregung. NF. Bd. V():–.
Deutsches Textarchiv http://www.deutschestextarchiv.de/strouhal_tonerregung_
, last visited July , .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

7.2 Dimensionless Navier-Stokes Equations 81

Self Study Problems

. Look at the following equation of fluid mechanics:

𝜕𝑣 ( ) 1
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣
𝜕𝑡 ⏟⏟ ⏟ 𝜌 ⏟⏟⏟
⏟⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏟ ⏟⏟⏟
1 2 4
3

Answer the following questions:

(a) What type of equation is this mathematically?
(b) What is the name of the equation?
(c) State what each of the terms 1 through 4 represents.
(d) What is the dimension of the terms in the equation?
. Provide a definition for each of the four dimensionless numbers: 𝑅𝑒, 𝐹𝑟, 𝐸𝑢, 𝑆𝑡.
. Consider the following form of the continuity equation.

𝜕𝜌
+ ∇ 𝑇 (𝜌𝑣) = 0
𝜕𝑡
(a) State whether the equation is in differential or integral form and whether it
j is the conservative or nonconservative form. j
(b) Convert the equation into a dimensionless form based on reference quanti-
ties 𝐿 for length, 𝑇 for time, 𝑈∞ for velocity, and 𝜌∞ for density.
(c) Which of the four dimensionless numbers appears in the resulting dimen-
sionless continuity equation?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

82

8
Reynolds Averaged Navier-Stokes
Equations (RANSE)

In practice, most available CFD systems do not attempt to solve Equation (.) di-
rectly. For high Reynolds numbers direct numerical simulation is still beyond our
computational capabilities. Instead, we solve the RANS equations, which are derived
by splitting the instantaneous time dependent velocity into an average velocity (still
time dependent) and the turbulent velocity. This introduces the initially unknown
Reynolds stress tensor. The added unknowns require additional equations to represent
the turbulence. Turbulence modeling is still a focus of ongoing research, and several
turbulence models of varying complexity are in use for ship resistance computations.
We derive the RANSE form of the Navier-Stokes equations for incompressible flow
j below to illustrate the difference. j

Learning Objectives
At the end of this chapter students will be able to:

• transform the NSE into RANSE

• formulate a dimensionless version of the NSE
• introduce characteristic dimensionless parameters
• explain the challenge of high Reynolds numbers

8.1 Mean and Turbulent Velocity

Turbulence Most instruments used to measure velocity and pressure report an average quantity.
This commonly encompasses averaging over a small volume and a short time period
as the instruments are not sensitive enough to follow extremely rapid fluctuations of
velocity and pressure like they occur in turbulence. By definition, we assume steady
flow to be time independent. Many of our instruments to measure flow velocity, e.g. a
Pitot-static tube, will register a constant velocity when they are immersed in a constant
stream. However, the flow may still experience rapid fluctuations 𝑢′ around a constant
mean velocity 𝑢, especially in a flow at high Reynolds number (see Figure .(a)).
Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.1 Mean and Turbulent Velocity 83

[−]
1.6 1.6
velocities u and u(t)= u + u ′ (t) [−]

1.4 1.4

velocities u(t) and u(t)= u(t)+ u ′ (t)

1.2 1.2

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4
u(t)= u + u ′ (t) u(t)= u(t)+ u ′ (t)
0.2 0.2
mean steady flow u mean unsteady flow u(t)
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
time t [s] time t [s]

(a) turbulent steady flow (b) turbulent unsteady flow

Figure 8.1 Mean and actual velocities in steady and unsteady turbulent flow

1.0
[m/s]

0.8
mean velocity u

0.6

0.4
j j
0.2

0.0
1.0 0.5 0.0 0.5 1.0
transverse position 2x/B []

(a) mean velocity (b) magnitude of longitudinal and transverse

turbulence

Figure 8.2 Velocity and turbulence distribution across an air duct

Hot wire anemometers (HWA) are sensitive enough to register rapidly varying turbulent Measuring
velocities 𝑢′ . An HWA consists primarily of a heated Tungsten wire exposed to the turbulence
flow. Passing fluid will cool the wire more or less depending on actual flow velocity. The
resistance of Tungsten is a function of temperature and even tiny temperature changes
cause measurable voltage changes in the attached circuitry.
Figure .(a) shows a typical mean velocity distribution found in the cross section
of an air duct. Figure .(b) presents the magnitude of longitudinal and transverse
turbulence. The longitudinal turbulence is parallel to the mean flow. Turbulence occurs
in all three coordinate directions and, as the example shows, is not necessarily equal in
all coordinate directions. Nevertheless, we often assume that turbulence is isotropic,
i.e. equal in all directions. In general, turbulence depends on several factors, including
disturbances in the on flow, mean flow velocity, geometry, pressure gradients, etc.
Our numerical tools and associated discretizations (grids) are not capable of resolving Reynolds’
the small scale of turbulence in space and time. Osborne Reynolds proposed to divide averaging

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

84 8 Reynolds Averaged Navier-Stokes Equations (RANSE)

all relevant flow properties into a mean value and a turbulent fluctuation.

𝑢(𝑡) = 𝑢(𝑡) + 𝑢′ (𝑡) 𝜌(𝑡) = 𝜌(𝑡) + 𝜌′ (𝑡)

𝑣(𝑡) = 𝑣(𝑡) + 𝑣′ (𝑡) 𝑝(𝑡) = 𝑝(𝑡) + 𝑝′ (𝑡) (.)
′ ′
𝑤(𝑡) = 𝑤(𝑡) + 𝑤 (𝑡) 𝑓 (𝑡) = 𝑓 (𝑡) + 𝑓 (𝑡)

Mean values of time series are calculated by integrating the series over time and dividing
by the length 𝑇 of the time series.
𝑇
1
𝑢 = 𝑢(𝑡) d𝑡 (.)
𝑇 ∫
0

For unsteady flows, 𝑇 must be small enough to follow the variation of the mean value
over time but large enough that the mean value of the turbulent velocity 𝑢′ vanishes.
𝑇
1
𝑢′ = 𝑢′ (𝑡) d𝑡 = 0 (.)
𝑇 ∫
0

Conversion of the NSE into the RANSE starts with introducing the approach Equa-
j tion (.) into the basic equations. Here we restrict ourselves to incompressible flows j
and use the continuity equation (.) and the NSE in the form of Equation (.).

8.2 Time Averaged Continuity Equation

Application of Starting with the continuity equation, we replace the time dependent velocity by the
Reynolds’ sum of mean velocity and turbulence.
averaging
⎛ 𝑢(𝑡) ⎞ ( ) ⎛ 𝑢(𝑡) + 𝑢′ (𝑡) ⎞
⎜ ⎟ 𝜕 𝜕 𝜕 ⎜
𝑇
∇ 𝑣(𝑡) = ∇ 𝑇
𝑣(𝑡) = , , 𝑣(𝑡) + 𝑣′ (𝑡) ⎟ = 0
⎜ ⎟ 𝜕𝑥 𝜕𝑦 𝜕𝑧 ⎜ ′ ⎟
⎝ 𝑤(𝑡) ⎠ ⎝ 𝑤(𝑡) + 𝑤 (𝑡) ⎠
( ) ( ) ( )
𝜕 𝜕 𝜕
𝑢(𝑡) + 𝑢′ (𝑡) + 𝑣(𝑡) + 𝑣′ (𝑡) + 𝑤(𝑡) + 𝑤′ (𝑡) = 0 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

Subsequently, we will omit the time argument, which shortens the writing quite a bit.
Repeating this substitution for the NSE (.) yields:

⎛ 𝑢 + 𝑢′ ⎞ [( ) ]⎛ 𝑢 + 𝑢
′ ⎞
𝜕 ⎜ ⎟ + 𝑢 + 𝑢′ , 𝑣 + 𝑣′ , 𝑤 + 𝑤′ ∇ ⎜ 𝑣 + 𝑣′ ⎟
𝑣 + 𝑣′
𝜕𝑡 ⎜ 𝑤 + 𝑤′ ⎟ ⎜ ′ ⎟
⎝ ⎠ ⎝ 𝑤+𝑤 ⎠
⎛ 𝑢 + 𝑢′ ⎞
1 ( )
= 𝑓 + 𝑓 − ∇ 𝑝 + 𝑝 + 𝜈𝚫 ⎜ 𝑣 + 𝑣′
′ ′ ⎟ (.)
𝜌 ⎜ ′ ⎟
⎝ 𝑤+𝑤 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.2 Time Averaged Continuity Equation 85

So far, we have not modified the continuity and Navier-Stokes equations, we have just
introduced the identities from Equation (.).
Next, we average Equations (.) and (.) over time. That means we integrate both
sides of the equations over time and divide by the time period 𝑇 .

𝑇 𝑇
1 1
… d𝑡 = … d𝑡 (.)
𝑇 ∫ 𝑇 ∫
0 0

This is a lot of cumbersome writing work. We can take a shortcut by exploiting the
rules for forming averages.
In a nutshell, the following computation rules are applied, which can all be derived Rules for
from basic rules of integration. averages

… equals …
(𝑎) average of a sum 𝑔+ℎ = 𝑔 + ℎ sum of averages

(𝑏) average of an average 𝑔 = 𝑔 average

𝜕𝑔 𝜕𝑔
(𝑐) average of a derivative = derivative of the average (.)
𝜕𝑠 𝜕𝑠
j j
(𝑑) average of an integral 𝑔 d𝑠 = 𝑔 d𝑠 integral of the average
∫ ∫

and the average of an average 𝑔 multiplied by another function ℎ

(𝑒) 𝑔⋅ℎ = 𝑔⋅ℎ

is the product of averages. However, the average of a product is not equal to the average
of products:

𝑔⋅ℎ ≠ 𝑔⋅ℎ (.)

Just consider the sine function 𝑔 = sin(𝑡). Its average is zero over one period 𝑇 = 2𝜋.

𝑇
1
𝑔 = sin(𝑡) = sin(𝑡) 𝑑𝑡
𝑇 ∫
0
[ ]2𝜋
1
= − cos(𝑡)
2𝜋 0
1 [ ]
= − 1 − (−1)
2𝜋
= 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

86 8 Reynolds Averaged Navier-Stokes Equations (RANSE)

However, the average of 𝑔 ⋅ ℎ with 𝑔 = sin(𝑡) and ℎ = sin(𝑡) is clearly not equal to the
product of the averages.

𝑇
1
𝑔 ⋅ ℎ = sin(𝑡) ⋅ sin(𝑡) = sin(𝑡) sin(𝑡) 𝑑𝑡
𝑇 ∫
0
[ ]2𝜋
1 𝑡 sin(2𝑡)
= −
2𝜋 2 4 0
1 [ ]
= 𝜋−0−0+0
2𝜋
1
= ≠ sin(𝑡) ⋅ sin(𝑡)
2
Equipped with the rules for averages, we return to the continuity equation (.) and
Navier-Stokes equations (.) for incompressible flow.
Reynolds’ Taking the average of the continuity equation (.) on both sides yields:
average of the
continuity ( ) ( ) ( )
𝜕 𝜕 𝜕
equation 𝑢 + 𝑢′ + 𝑣 + 𝑣′ + 𝑤 + 𝑤′ = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧

The average of a constant is equal to the constant. Applying rule (a) for averages of a
sum results in
j j
( ) ( ) ( )
𝜕 𝜕 𝜕
𝑢 + 𝑢′ + 𝑣 + 𝑣′ + 𝑤 + 𝑤′ = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧

and the average of a derivative is equal to the derivative of the average according to rule
(c):
( ) ( ) ( )
𝜕 𝜕 𝜕
𝑢 + 𝑢′ + 𝑣 + 𝑣′ + 𝑤 + 𝑤′ = 0 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

Again, we apply rule (a):

( ) ( ) ( )
𝜕 𝜕 𝜕
𝑢 + 𝑢′ + 𝑣 + 𝑣′ + 𝑤 + 𝑤′ = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧

Then, forming the derivatives of the sums, we apply rule (b) and resort the terms:

𝜕𝑢 𝜕𝑣 𝜕𝑤 𝜕𝑢′ 𝜕𝑣′ 𝜕𝑤′

+ + + + + = 0 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

The averages of the turbulent velocities vanish by definition, i.e.

0 0 0
𝜕𝑢 𝜕𝑣 𝜕𝑤 𝜕𝑢′
◁
 𝜕𝑣′

◁ 𝜕𝑤 ′

7
+ + + ◁ + ◁ +  = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧 ◁𝜕𝑥 𝜕𝑦
◁ 𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.3 Time Averaged Navier-Stokes Equations 87

What remains is the fact that the mean velocities satisfy the continuity equation (.).
𝜕𝑢 𝜕𝑣 𝜕𝑤
+ + = 0 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
Subtracting Equation (.) from Equation (.) reveals that the turbulent velocities
themselves also satisfy the continuity equation.
𝜕𝑢′ 𝜕𝑣′ 𝜕𝑤′
+ + = 0 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧
We will use this result to our advantage when we deal with the averaging of the Navier-
Stokes equations.

8.3 Time Averaged Navier-Stokes Equations

For the NSE, we illustrate the process using the 𝑥-component of Equation (.). Let us Reynolds’
start with the inertia forces (per unit mass) on the left-hand side. With the help of rules average of
inertia force
(a) and (c), the average of the local derivative turns into
0
𝜕 ( ) 𝜕𝑢 𝜕𝑢′

◁ 𝜕𝑢
𝑢 + 𝑢′ = + ◁ = (.)
𝜕𝑡 𝜕𝑡 ◁𝜕𝑡 𝜕𝑡
j The first summand of the convective derivative transforms into j
( )𝜕 ( ) 𝜕𝑢 𝜕𝑢′ 𝜕𝑢 𝜕𝑢′
𝑢 + 𝑢′ 𝑢 + 𝑢′ = 𝑢 +𝑢 + 𝑢′ + 𝑢′
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑢 𝜕𝑢′ 𝜕𝑢 𝜕𝑢′
+𝑢 = 𝑢 + 𝑢′ + 𝑢′
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑢 𝜕𝑢′ 𝜕𝑢 𝜕𝑢′
= 𝑢 +𝑢 + 𝑢′ + 𝑢′
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
The second and third terms on the right-hand side vanish because the mean value of
the turbulence is zero: 𝑢′ ≡ 0. However, the last term cannot be simplified:
( )𝜕 ( ) 𝜕𝑢 𝜕𝑢′
𝑢 + 𝑢′ 𝑢 + 𝑢′ = 𝑢 + 𝑢′ (.)
𝜕𝑥 𝜕𝑥 𝜕𝑥
Applying the same operations to the second and third terms of the convective derivative
in Equation (.) yields
( )𝜕 ( ) 𝜕𝑢 𝜕𝑢′
𝑣 + 𝑣′ 𝑢 + 𝑢′ = 𝑣 + 𝑣′ (.)
𝜕𝑦 𝜕𝑦 𝜕𝑦
and
( )𝜕 ( ) 𝜕𝑢 𝜕𝑢′
𝑤 + 𝑤′ 𝑢 + 𝑢′ = 𝑤 + 𝑤′ (.)
𝜕𝑧 𝜕𝑧 𝜕𝑧
Next are the external forces per unit mass on the right-hand side of Equation (.). Reynolds’
Again, the process is shown using the 𝑥-component as an example. average of
external forces

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

88 8 Reynolds Averaged Navier-Stokes Equations (RANSE)

. Body force
𝑓𝑥 + 𝑓𝑥′ = 𝑓𝑥 + 𝑓𝑥′ = 𝑓𝑥 (.)

. Pressure force
𝜕 ( ) 𝜕𝑝 𝜕𝑝′ 𝜕𝑝
𝑝 + 𝑝′ = + = (.)
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
. Viscous force

𝜕2 ( ) 𝜕2 ( ) 𝜕2 ( )
𝑢 + 𝑢′ + 𝑢 + 𝑢′ + 𝑢 + 𝑢′
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧2

⁓0
𝜕2𝑢 𝜕2𝑢 𝜕2𝑢 𝜕 2 𝑢′ 𝜕 2 𝜕 2
𝑢′ 𝑢′

= + + +  + +
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2  𝜕𝑥2 𝜕𝑦2 𝜕𝑧2
2 2 2
𝜕 𝑢 𝜕 𝑢 𝜕 𝑢
= + +
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2
= 𝚫𝑢 (.)

Reynolds’ We take stock and collect the time averaging results from Equations (.) through
average of NSE (.).
𝑥-component
𝜕𝑢 𝜕𝑢 𝜕𝑢′ 𝜕𝑢 𝜕𝑢′ 𝜕𝑢 𝜕𝑢′ 𝜕𝑝
+𝑢 + 𝑢′ + 𝑣 + 𝑣′ + 𝑤 + 𝑤′ = 𝑓𝑥 + + 𝚫𝑢 (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑥
j j
The equation looks similar to Equation (.) from which we started, except that there
are now three additional summands on the left-hand side. We subtract the additional
terms to move them over to the right-hand side.

𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑝 𝜕𝑢′ 𝜕𝑢′ 𝜕𝑢′

+𝑢 +𝑣 +𝑤 = 𝑓𝑥 + + 𝚫𝑢 − 𝑢′ − 𝑣′ − 𝑤′ (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑧

The physics of the new terms will be explained in due course, but first we will modify
them slightly. Consider the following expression, which we expand by application of
the product rule of differentiation.
𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢
𝜕𝑥 𝜕𝑦 𝜕𝑧
𝜕𝑢′ 𝜕𝑢′ ′ 𝜕𝑢′ 𝜕𝑣′ ′ 𝜕𝑢′ 𝜕𝑤′ ′
= 𝑢′ + 𝑢 + 𝑣′ + 𝑢 + 𝑤′ + 𝑢 (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧
( ′ ′ 
)
⁓0
′ 𝜕𝑢 𝜕𝑣 𝜕𝑤′ 

𝜕𝑢′ 𝜕𝑢′ 𝜕𝑢′
= 𝑢 +  +
 + 𝑢′ + 𝑣′ + 𝑤′
 𝜕𝑥
  𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

In the last line, the terms with factor 𝑢′ are summarized, and the continuity equation
for turbulent velocities (.) allows us to eliminate the term. The result of this exercise
is the following identity:

𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕𝑢′ 𝜕𝑢′ 𝜕𝑢′

𝑢𝑢 + 𝑣𝑢 + 𝑤 𝑢 = 𝑢′ + 𝑣′ + 𝑤′ (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.4 Reynolds Stresses and Turbulence Modeling 89

We average both sides of the equation over time.

𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕𝑢′ 𝜕𝑢′ 𝜕𝑢′

𝑢𝑢 + 𝑣𝑢 + 𝑤 𝑢 = 𝑢′ + 𝑣′ + 𝑤′ (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

Obviously, the right-hand side is identical to the three new terms in our time averaged
𝑥-component of the NSE (.). We will use the left-hand side of Equation (.) to
replace the three new terms.
It is left to the reader to repeat the time averaging process with the 𝑦- and 𝑧-components RANSE
of the NSE. As a result, we obtain the RANS equations.
[ ]
𝜕𝑢 ( 𝑇 ) 1 𝜕𝑝 𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
+ 𝑣 ∇ 𝑢 = 𝑓𝑥 − + 𝜈𝚫𝑢 − 𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢
𝜕𝑡 𝜌 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑧
[ ]
𝜕𝑣 ( 𝑇 ) 1 𝜕𝑝 𝜕 ′ ′ 𝜕 ′ ′ 𝜕
+ 𝑣 ∇ 𝑣 = 𝑓𝑦 − + 𝜈𝚫𝑣 − ′ ′
(𝑢 𝑣 ) + (𝑣 𝑣 ) + (𝑤 𝑣 ) (.)
𝜕𝑡 𝜌 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑧
[ ]
𝜕𝑤 ( 𝑇 ) 1 𝜕𝑝 𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
+ 𝑣 ∇ 𝑤 = 𝑓𝑧 − + 𝜈𝚫𝑤 − 𝑢𝑤 + 𝑣𝑤 + 𝑤𝑤
𝜕𝑡 𝜌 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

At this point, the bars on 𝑢, 𝑣, 𝑤, 𝑝 are omitted with the understanding that 𝑢, 𝑣, 𝑤, and
𝑝 now represent mean flow quantities without turbulence.
j j

8.4 Reynolds Stresses and Turbulence Modeling

If you study the RANSE (.), you will find that the new terms look similar to the Reynolds
stress terms in the conservation of momentum Equations (.), (.), and (.). The stresses
averaging process has introduced six, yet unknown, averages of the turbulent velocity
products (𝑢′ 𝑢′ ), (𝑢′ 𝑣′ ), (𝑢′ 𝑤′ ), (𝑣′ 𝑣′ ), (𝑣′ 𝑤′ ), and (𝑤′ 𝑤′ ). There is a total of nine terms,
but six appear in pairs.

(𝑢′ 𝑣′ ) = (𝑣′ 𝑢′ ) (𝑢′ 𝑤′ ) = (𝑤′ 𝑢′ ) (𝑣′ 𝑤′ ) = (𝑤′ 𝑣′ ) (.)

The new unknowns are often referred to as ‘Reynolds stresses’, which is not quite correct.
In order to obtain a stress unit [N/m2 ], the products of turbulent velocity [m2 /s2 ] have
to be multiplied by the density of the fluid first.
The unknown terms are collected into the Reynolds stress tensor 𝜏. Reynolds stress
tensor
′ ′ ′
⎛ 𝜏𝑥𝑥 𝜏𝑦𝑥 𝜏𝑧𝑥 ⎞ ⎛ 𝑢′ 𝑢′ 𝑣′ 𝑢′ 𝑤′ 𝑢′ ⎞
⎜ ′ ⎟ ⎜ ⎟
𝜏 = ⎜ 𝜏𝑥𝑦 ′
𝜏𝑦𝑦 ′
𝜏𝑧𝑦 ⎟ = −𝜌 ⎜ 𝑢′ 𝑣′ 𝑣′ 𝑣′ 𝑤′ 𝑣′ ⎟ (.)
⎜ 𝜏′ ′
𝜏𝑦𝑧 ′ ⎟
𝜏𝑧𝑧 ⎜ ⎟
⎝ 𝑥𝑧 ⎠ ⎝ 𝑢′ 𝑤′ 𝑣′ 𝑤′ 𝑤′ 𝑤′ ⎠

Note that the Reynolds stresses are only apparent stresses. Their appearance is a
consequence of separating mean and turbulent velocities and averaging the equation
over time.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

90 8 Reynolds Averaged Navier-Stokes Equations (RANSE)

Vector form of The RANSE (.) can be cast into a vector form by introducing the Reynolds stress
RANSE tensor 𝜏.
𝜕𝑣 ( ) 1 1 ( 𝑇 )𝑇
+ 𝑣∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣 + ∇ 𝜏 (.)
𝜕𝑡 𝜌 𝜌

Turbulence Similar to Stokes’ approach, which connected the unknown stresses in the conserva-
models tion of momentum equations to rates of change in the velocity field, we need to find
additional equations for the unknown Reynolds stresses. This is where turbulence
models come into play. Turbulence modeling is a very active field of research and we
will not dive into details here. The reader may start with the classic book on boundary
layer theory by Schlichting and Gersten () and then study turbulence models in
depth with Wilcox (). There are also excellent websites, like NASA’s Turbulence
Modeling Resource (Rumsey et al., ).
Boussinesq’s Some of the turbulence models use only one or two additional equations, which requires
eddy viscosity that the Reynolds stress tensor is expressed as a function of one or two parameters. In
hypothesis
the following, we outline the classic approach introduced by Joseph Valentin Boussinesq
(* – †), a French mathematician and physicist. Boussinesq () proposed
to link the apparent Reynolds stresses to the space derivatives of the mean velocities.
This is known as Boussinesq’s eddy viscosity hypothesis. Details can also be found in
Schmitt ().
The hypothesis is based on three basic assumptions.
j j
Apparent . Reynolds stresses are proportional to the eddy viscosity 𝜇𝑡 , also called apparent
viscosity viscosity. In contrast to the dynamic viscosity 𝜇 for Newtonian fluids, eddy
viscosity is not a material constant. 𝜇𝑡 still depends on the velocity distribution.
Isotropic . Turbulence is treated as isotropic, which means the turbulent velocity compo-
turbulence nents 𝑢′ , 𝑣′ , and 𝑤′ are all of identical magnitude. We require specifically that
𝑢′ 𝑢′ = 𝑣′ 𝑣′ = 𝑤′ 𝑤′ . This is definitely a stretch and cannot be true close to walls.
As we approach a wall, turbulence normal to the surface must diminish while
other components are not affected as much.
Mean kinetic . The mean kinetic turbulent energy is
energy of
( )
turbulence 1 ′ ′
𝑘 = 𝑢 𝑢 + 𝑣′ 𝑣′ + 𝑤′ 𝑤′ (.)
2

The assumptions lead to the following expression for the Reynolds stress tensor.
( ) ( )
⎛ 𝜕𝑢 2 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑤 ⎞
⎜ 2𝜇 𝑡 − 𝜌 𝑘 𝜇 𝑡 + 𝜇𝑡 + ⎟
⎜ (𝜕𝑥 3 ) 𝜕𝑦 𝜕𝑥
(
𝜕𝑧 𝜕𝑥 ⎟
)
⎜ 𝜕𝑢 𝜕𝑣 𝜕𝑣 2 𝜕𝑤 𝜕𝑣 ⎟
𝜏 = ⎜ 𝜇𝑡 (.)
𝜕𝑧 ⎟⎟
+ 2𝜇𝑡 − 𝜌𝑘 𝜇𝑡 +
⎜ ( 𝜕𝑦 𝜕𝑥 ) 𝜕𝑦 3 𝜕𝑦
( )
⎜ 𝜕𝑢 𝜕𝑤 𝜕𝑤 𝜕𝑣 𝜕𝑤 2 ⎟
⎜ 𝜇𝑡 𝜕𝑧 + 𝜕𝑥 𝜇𝑡 + 2𝜇𝑡 − 𝜌𝑘 ⎟
3
⎝ 𝜕𝑦 𝜕𝑧 𝜕𝑧 ⎠

This leaves us with the mean kinetic energy 𝑘 and the apparent viscosity 𝜇𝑡 as remaining
unknowns.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.4 Reynolds Stresses and Turbulence Modeling 91

We employ the first row of the new Reynolds stress tensor (.) and substitute it for Substituting
the Reynolds stresses in the 𝑥-component of the RANSE (.), i.e. Reynolds
stresses
[ ]
𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
− 𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢
𝜕𝑥 𝜕𝑦 𝜕𝑧
{ [ ] [ ( )] [ ( )]}
1 𝜕 𝜕𝑢 2 𝜕 𝜕𝑢 𝜕𝑣 𝜕 𝜕𝑢 𝜕𝑤
= 2𝜇𝑡 − 𝜌𝑘 + 𝜇 + + 𝜇 +
𝜌 𝜕𝑥 𝜕𝑥 3 𝜕𝑦 𝑡 𝜕𝑦 𝜕𝑥 𝜕𝑧 𝑡 𝜕𝑧 𝜕𝑥
[ ] [ ( )] [ ( )]
𝜕 𝜕𝑢 2 𝜕 𝜕𝑢 𝜕𝑣 𝜕 𝜕𝑢 𝜕𝑤
= 2𝜈𝑡 − 𝜌𝑘 + 𝜈𝑡 + + 𝜈𝑡 + (.)
𝜕𝑥 𝜕𝑥 3 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑧 𝜕𝑧 𝜕𝑥

For incompressible fluids, the apparent viscosity 𝜇𝑡 is replaced with the eddy kinematic
viscosity 𝜈𝑡 = 𝜇𝑡 ∕𝜌.
The derivatives on the right-hand side of Equation (.) must be taken with the product
rule because the kinematic eddy viscosity 𝜈𝑡 is a function of spatial variables 𝑥, 𝑦, and 𝑧.
We expand the right-hand side accordingly and rearrange its terms.
[ ]
𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
− 𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢
𝜕𝑥 𝜕𝑦 𝜕𝑧
2
𝜕 𝑢 𝜕𝜈 𝜕𝑢 2 𝜕𝑘
= 2𝜈𝑡 +2 𝑡 −
𝜕𝑥 2 𝜕𝑥 𝜕𝑥 3 𝜕𝑥
( ) ( 2 ) (.)
𝜕𝜈𝑡 𝜕𝑢 𝜕𝑣 𝜕 𝑢 𝜕2𝑣
j + + + 𝜈𝑡 + j
𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦2 𝜕𝑥𝜕𝑦
( ) ( 2 )
𝜕𝜈𝑡 𝜕𝑢 𝜕𝑤 𝜕 𝑢 𝜕𝑤
+ + + 𝜈𝑡 +
𝜕𝑧 𝜕𝑧 𝜕𝑥 𝜕𝑧2 𝜕𝑥𝜕𝑧

Rearranging the terms on the right-hand side results in

[ ]
𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′)
− 𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢
𝜕𝑥 𝜕𝑦 𝜕𝑧
( 2 ) ( 2 )
𝜕 𝑢 𝜕2𝑢 𝜕2𝑢 𝜕 𝑢 𝜕2𝑣 𝜕2𝑤
= 𝜈𝑡 + + + 𝜈𝑡 + + (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2 𝜕𝑥2 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑧
( ) ( ) ( )
2 𝜕𝑘 𝜕𝜈𝑡 𝜕𝑢 𝜕𝑢 𝜕𝜈𝑡 𝜕𝑢 𝜕𝑣 𝜕𝜈𝑡 𝜕𝑢 𝜕𝑤
− + + + + + +
3 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑧 𝜕𝑧 𝜕𝑥

The first pair of parentheses on the right involves the Laplace operator
( )
𝜕2𝑢 𝜕2𝑢 𝜕2𝑢
𝜈𝑡 + + = 𝚫𝑢 (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2

The second parenthesis vanishes because it contains the divergence of the velocity, i.e.
the continuity equation for incompressible fluids.
( ) ( )
𝜕2𝑢 𝜕2𝑣 𝜕2𝑤 𝜕 𝜕𝑢 𝜕𝑣 𝜕𝑤
𝜈𝑡 + + = 𝜈𝑡 + + = 0 (.)
𝜕𝑥 2 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
conti. eq. incompress. ∇ 𝑇 𝑣=0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

92 8 Reynolds Averaged Navier-Stokes Equations (RANSE)

The last three summands can be summarized in a convenient vector form.

( ) ( ) ( )
𝜕𝜈𝑡 𝜕𝑢 𝜕𝑢 𝜕𝜈𝑡 𝜕𝑢 𝜕𝑣 𝜕𝜈𝑡 𝜕𝑢 𝜕𝑤
+ + + + +
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑧 𝜕𝑧 𝜕𝑥
( )
( )𝑇 𝜕𝑣
= ∇ 𝜈𝑡 ∇𝑢 + (.)
𝜕𝑥

RANSE based on Finally, by substituting Equations (.), (.), and (.) into Equation (.), we
Boussinesq’s obtain for the three Reynolds stress terms on the 𝑥-component of RANSE the following
hypothesis
expressions:
[ ] ( )
𝜕 ( ′ ′) 𝜕 ( ′ ′) 𝜕 ( ′ ′) 2 𝜕𝑘 ( )𝑇 𝜕𝑣
− 𝑢𝑢 + 𝑣𝑢 + 𝑤𝑢 = 𝜈𝑡 𝚫𝑢 − + ∇ 𝜈𝑡 ∇𝑢 + (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧 3 𝜕𝑥 𝜕𝑥

Application of the same procedure to the Reynolds stress terms of the 𝑦- and 𝑧-components
yields a vector form of the RANS equations based on Boussinesq’s eddy viscosity hy-
pothesis.

𝜕𝑣 ( ) 1
+ 𝑣∇ 𝑣 = 𝑓 − ∇ 𝑝
𝜕𝑡 𝜌
( ) [( )𝑇 ] [ ]
2
+ 𝜈 + 𝜈𝑡 𝚫𝑣 − ∇ 𝑘 + ∇ 𝜈𝑡 ∇ 𝑣 + ∇ 𝑣𝑇 (∇ 𝜈𝑡 ) (.)
3
j j
Note that the dyadic product ∇ 𝑣𝑇 results in a [×] matrix. One can imagine how
much more effort turbulent flow computations require by comparing the RANSE (.)
with the equivalent incompressible NSE (.) applied to laminar flow, especially if we
consider that Equation (.) is already a simplified version of RANSE.
Need for Together with the continuity equation, we now have four equations for six unknowns.
turbulence
models 𝑢 = 𝑢(𝑥, 𝑦, 𝑧, 𝑡) 𝑣 = 𝑣(𝑥, 𝑦, 𝑧, 𝑡) 𝑤 = 𝑤(𝑥, 𝑦, 𝑧, 𝑡) (.)
𝑝 = 𝑝(𝑥, 𝑦, 𝑧, 𝑡) 𝜈𝑡 = 𝜈𝑡 (𝑥, 𝑦, 𝑧, 𝑡) 𝑘 = 𝑘(𝑥, 𝑦, 𝑧, 𝑡)

Consequently, two additional equations for the eddy viscosity 𝜈𝑡 and the mean turbulent
energy 𝑘 have to be found.
This is accomplished with two equations turbulence models like the 𝑘-𝜔 or the 𝑘-𝜀
models among others. In the latter, eddy viscosity is expressed as a function of the
mean turbulent kinetic energy 𝑘 and a dissipation rate 𝜀.

𝑘2
𝜈𝑡 = 𝐶𝜇
𝜀
𝐶𝜇 is an adjustable constant. For 𝑘 and 𝜀, two transport equations are provided which
are similar to the NSE themselves.
The resulting system of six coupled, nonlinear partial differential equations has to be
discretized and solved considering boundary conditions on the faces of the computa-
tional domain. All in all, a difficult and time consuming process. An introduction into
CFD can be found in Anderson, Jr. () and Versteeg and Malalasekera ().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

8.4 Reynolds Stresses and Turbulence Modeling 93

References
Anderson, Jr., J. (). Computational fluid dynamics – The basics with applications.
McGraw-Hill, Inc., New York, NY.
Boussinesq, J. (). Essai sur la théorie des eaux courantes. Mémoires présentés par
divers savants à l’Académie des Sciences, ():–.
Rumsey, C., Smith, B., and Huang, G. (). Turbulence modeling resource. http:
//turbmodels.larc.nasa.gov. Last visited December , .
Schlichting, H. and Gersten, K. (). Boundary-layer theory. Springer-Verlag, Berlin,
Heidelberg, New York, eigth edition. Corrected printing.
Schmitt, F. (). About Boussinesq’s turbulent viscosity hypothesis: historical re-
marks and a direct evaluation of its validity. Comptes Rendus Mécanique, (–
):–.
Versteeg, H. and Malalasekera, W. (). An introduction to computational fluid
dynamics: the finite volume method. Pearson, Harlow, England, second edition.
Wilcox, D. (). Turbulence modeling for CFD. DCW Industries, La Cañada, Califor-
nia, third edition.

Self Study Problems

j j
. Explain turbulence in your own words.
. What is the time average of the turbulent velocity components 𝑢′ , 𝑣′ , and 𝑤′ ?
. Derive the Reynolds averaged 𝑦- and 𝑧- component of the external forces in the
Navier-Stokes equation following the example in Equations (.) to (.).
. Why is turbulence model required to solve the Reynolds Averaged Navier Stokes
Equations (RANSE)?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

94

9
Application of the Conservation
Principles

The conservation of mass and momentum equations – also known as continuity equa-
tion and Navier-Stokes equations – form a system of coupled, nonlinear, partial dif-
ferential equations. They are without doubt complicated and no generally applicable
analytical solution is known. We will explain the basic concepts of using the conserva-
tion of mass and conservation of momentum principles with two examples:

• calculating the drag of an object in a wind tunnel, and

• computing the drag of a submerged vessel in an unbounded fluid.

j j
Admittedly, the examples are constructed and simplified to allow an analytical solution.
However, they show the principal procedures employed in a wake analysis, and the use
of such analysis in determining the resistance of a body in a flow. Along the way we will
use appropriate boundary conditions. This discussion will also provide an introduction
to an integral form of the conservation of momentum equations.

Learning Objectives
At the end of this chapter students will be able to:

• apply conservation of mass and momentum principles

• determine the drag of a body with known wake field

9.1 Body in a Wind Tunnel

Problem Consider a streamlined body of revolution in a wind tunnel with a circular cross section
(diameter ⌀𝐷 = 2𝑟0 = 2 m) (see Figure .). You are tasked with the computation of
the drag force 𝐹 𝐷 acting on the body. The installed measurement equipment allows
us to read pressure and velocity distribution at the inlet 1 upstream of the body and
outlet 2 downstream of the body.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.1 Body in a Wind Tunnel 95

Figure 9.1 Body of revolution in a wind tunnel (simplified)

Measured pressure at the inlet is 𝑝1 = 1005 hPa (= 100 500 N/m2 ) and at the outlet Input data
𝑝2 = 1000 hPa. The pressure is constant across the tunnel cross section. We neglect
the boundary layer at the tunnel wall and assume the inflow velocity is constant with
a steady 𝑥-component of 𝑣1 = 10 m/s. The measured local velocity at the outlet
j j
was simplified to a radial distribution (constant on circles concentric to the 𝑥-axis)
( )𝑇
𝑣2 (𝑟) = 𝑣max 𝑟∕𝑟0 , 0, 0 . The air density is 𝜌 = 1.225 kg/m3 .
The reaction of the drag force is an external force (surface force) acting on the fluid Solution
between inlet and outlet. We can employ the conservation of mass and momentum
principles to solve this problem. Since the equations are applicable to any flow problem,
we need additional equations to make the solution specific to the given problem. These
additional equations are derived from boundary conditions. You may remember from
mechanics of materials that the deflection of a beam under load depends on how the
beam is supported. We will specify appropriate boundary conditions below.
Conservation of momentum states that the rate of change of momentum is equal to the
sum of external forces. The integral form of the conservation of momentum principle
reads as
𝜕 ( ) ∑
𝜌 𝑣 d𝑉 + 𝜌 𝑣 𝑣𝑇 𝑛 d𝑆 = (external forces) (.)
𝜕𝑡 ∭ ∬
𝑉 𝑆
The fixed control volume 𝑉 is the wind tunnel content between inlet and outlet. Its
surface 𝑆 consists of inlet 𝑆1 , outlet 𝑆2 , the cylinder 𝑆𝑐 , and indeed the body surface
𝑆𝑏 .
The left-hand side consists of two terms. The volume integral describes the change of Rate of change of
momentum over time. Since the flow is steady and the control volume is fixed, the time momentum
derivative of the volume integral must vanish.
𝜕
𝜌𝑣 d𝑉 = 0 (.)
𝜕𝑡 ∭
𝑉

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

96 9 Application of the Conservation Principles

Momentum flux We split the surface integral for the momentum flux into four parts: inlet, outlet, cylinder
for wall and body and body.
( )
𝜌𝑣 𝑣𝑇 𝑛 d𝑆 = … d𝑆 + … d𝑆 + … d𝑆 + … d𝑆 (.)
∬ ∬ ∬ ∬ ∬
𝑆 𝑆1 𝑆2 𝑆𝑐 𝑆𝑏

The tunnel wall will not allow any air to escape, and the dot product of velocity and
normal vector vanishes everywhere on the cylindrical tunnel wall

𝑣𝑇 𝑛𝑐 = 0 on tunnel wall 𝑆𝑐

In addition, we consider the test body 𝑆𝑏 impermeable, i.e. no fluid flows through it, i.e.
𝑣𝑇 𝑛𝑏 = 0 on 𝑆𝑏 as well. Consequently, the integrals over 𝑆𝑐 and 𝑆𝑏 vanish.
Momentum flux Velocity distribution 𝑣 and normal vector 𝑛1 are constant at the inlet. Therefore, the
through inlet surface integral over the inlet reduces to

( 𝑇 ) ⎛ 𝑣21 ⎞
𝜌 𝑣1 𝑣1 𝑛1 d𝑆 = −𝜌𝜋𝑟0 0 ⎟
2⎜
(.)
∬ ⎜ ⎟
𝑆1 ⎝ 0 ⎠

Momentum flux For this exercise, we assume a heavily simplified, analytical radial velocity distribution.
j
through outlet In a real world application this would be replaced by a detailed wake measurement. j
Using polar coordinates for the cross section, the surface element becomes d𝑆 = 𝑟d𝜑d𝑟
(with 𝑟 ∈ [0, 𝑟0 ] and 𝜑 ∈ [0, 2𝜋]). We expand the integrand for the integral over 𝑆2 :

𝑟0 2𝜋 ⎛ 𝑣 𝑟 ⎞⎡
( ) ⎜ max
𝑟0 ⎟ ⎢( ) ⎛ 1 ⎞⎤
𝑟
∬
𝜌𝑣2 𝑣𝑇2 𝑛2 d𝑆 = 𝜌
∫ ∫ ⎜⎜ 0 ⎟ ⎢ 𝑣max 𝑟 , 0, 0 ⎜⎜ 0 ⎟⎟⎥⎥ 𝑟 d𝜑d𝑟 (.)
0 0 ⎝ 0 ⎟⎣ 0
⎝ 0 ⎠⎦
𝑆2 ⎠

The integrand on the right-hand side does not depend on the polar coordinate 𝜑, and
the inner integral simplifies to multiplication with the difference between its upper and
lower boundary.
𝑟 [ ]𝑟0
( 𝑇 ) ⎛ 𝑣2max ⎞ 0 3 ⎛ 𝑣2max ⎞
𝑟 𝑟4
𝜌𝑣2 𝑣2 𝑛2 d𝑆 = 2𝜋𝜌 ⎜ 0 ⎟ d𝑟 = 2𝜋𝜌 ⎜ 0 ⎟
∬ ⎜ ⎟ ∫ 𝑟2 ⎜ ⎟ 4𝑟20
𝑆2 ⎝ 0 ⎠ 0 0 ⎝ 0 ⎠ 0

⎛ 𝑣2max ⎞
1 2⎜ ⎟
= 𝜌𝜋𝑟0 0 (.)
2 ⎜ ⎟
⎝ 0 ⎠

In summary, the rate of change of momentum on the left-hand side of (.) is equal to

( 𝑇 ) ⎡ ⎛ 𝑣2max ⎞ ⎛ 𝑣21 ⎞⎤
𝜕 2 ⎢1 ⎜ ⎟−⎜ 0 ⎟⎥
𝜌𝑣 d𝑉 + 𝜌𝑣 𝑣 𝑛 d𝑆 = 𝜌𝜋𝑟0 0 (.)
𝜕𝑡 ∭ ∬ ⎢2 ⎜ ⎟ ⎜ ⎟⎥
𝑉 𝑆 ⎣ ⎝ 0 ⎠ ⎝ 0 ⎠⎦

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.1 Body in a Wind Tunnel 97

The only unknown term in this expression is the maximum velocity 𝑣max in the outlet. Conservation of
If it cannot be measured properly, we can compute it from the conservation of mass mass
principle. At a relatively low speed in the wind tunnel, we can assume that the flow is
incompressible (constant density).
For incompressible, steady flow the continuity equation for a fixed control volume states
( )
𝜌 𝑣𝑇 𝑛 d𝑆 = 0 (.)
∬
𝑆

The vanishing mass flux integral indicates that as much fluid is flowing out of the control
volume as is flowing into it. Again, we split the surface 𝑆 of the control volume into
four parts. Zero flow boundary conditions are imposed on the solid walls of 𝑆𝑐 and 𝑆𝑏 .
The mass flux through inlet and outlet must be equal but of opposite sign:
( ) ( )
0 = 𝜌 𝑣𝑇 𝑛 d𝑆 + 𝜌 𝑣𝑇 𝑛 d𝑆 (.)
∬ ∬
𝑆1 𝑆2
𝑟0 2𝜋 0 2𝜋 𝑟
( ) ⎛ −1 ⎞ ( )⎛ 1 ⎞
𝑟
= 𝜌 𝑣1 , 0, 0 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟 + 𝜌 𝑣max , 0, 0 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟 (.)
∫ ∫ ⎜ ⎟ ∫ ∫ 𝑟0 ⎜ ⎟
0 0 ⎝ 0 ⎠ 0 0 ⎝0⎠
[ ]
2
= 𝜌𝜋𝑟20 𝑣max − 𝑣1 (.)
3
j or j
3
𝑣max = 𝑣1 (.)
2
This simple relationship should not be mistaken for a general solution. It is a conse-
quence of our simplified velocity distribution at the outlet.
Substituting this result into the momentum flux (.) yields: Total momentum
flux
( 𝑇 ) ⎛ 𝑣21 ⎞
𝜕 1
𝜌 𝑣 d𝑉 + 𝜌 𝑣 𝑣 𝑛 d𝑆 = 𝜌𝜋𝑟0 0 ⎟
2⎜
(.)
𝜕𝑡 ∭ ∬ 8 ⎜ ⎟
𝑉 𝑆 ⎝ 0 ⎠

We now turn to the sum of external forces making up the right-hand side of our conser- Sum of external
vation of momentum equation (.). forces

There are body and surface forces acting on the fluid in our control volume. Gravity
will act as a body force on the fluid. However, the gravity force is acting vertically to the
flow and is balanced by the aero-static pressure distribution in the fluid. The influence
of gravity on our wind tunnel flow can safely be neglected.
Only the surface forces remain. There will be pressure forces on inlet 𝐹 𝑝 and outlet
1
𝐹 𝑝 . Of course, there is also the force exerted by the body onto the fluid −𝐹 𝐷 . The
2

 More exact: low Mach number 𝑀𝑎 = 𝑣 with 𝑎 being the speed of sound in the respective medium. For
𝑎
𝑀𝑎 < 0.2 air flows may be treated as incompressible.
 Our assumption that the pressure is constant across inlet and outlet is not % true. There is a small

variation of static pressure depending on the tunnel diameter (about  Pa or .% of 𝑝2 in the example).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

98 9 Application of the Conservation Principles

minus sign expresses the fact that the reaction of the drag force (fluid acting on body) is
acting against the motion of the fluid. It is consistent to ignore the friction force on the
wind tunnel wall 𝑆𝑐 , since we neglected the wall influence on the velocity distribution
near the wall as well. A more accurate assessment would measure complete velocity
profiles without and with body to distinguish between the change of momentum caused
by the tank wall and that caused by the body.
In our simplified case the sum of external forces reduces to
∑
(external forces) = 𝐹 𝑝 + 𝐹 𝑝 − 𝐹 𝐷 (.)
1 2

Inlet pressure Integration of the pressure over the surface yields the pressure forces. For the inlet this
force results in

𝐹𝑝 = − 𝑝1 𝑛1 d𝑆 (.)
1 ∬
𝑆1

The minus sign in front of the integral is a consequence of our choice for the normal
vector direction: it is pointing out of the control volume. Therefore, a positive pressure
will cause a force acting in the negative normal direction. Since the inlet is a flat disc
with constant normal vector and the pressure 𝑝1 is assumed constant, the integration
simplifies to a multiplication with the tunnel cross section area.
⎛ −1 ⎞ ⎛1⎞
j 𝐹 𝑝 = −𝜋𝑟20 𝑝1 ⎜ 0 ⎟ = 𝜋𝑟20 𝑝1 ⎜ 0 ⎟ (.) j
1 ⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝0⎠
As expected, the resulting force on 𝑆1 points in the positive 𝑥-direction.
Outlet pressure Analysis of the pressure force on the outlet yields
force
⎛1⎞
𝐹𝑝 = − 𝑝2 𝑛2 d𝑆 = −𝜋𝑟20 𝑝2 ⎜ 0 ⎟ (.)
2 ∬ ⎜ ⎟
𝑆2 ⎝0⎠

This force points in the negative 𝑥-direction.

Body drag force Our test object is a body of revolution, which allows the conclusion that only the 𝑥-
component of the drag force will be nonzero. Therefore, the sum of external forces is
equal to

∑ ⎛ 1 ⎞ ⎛ 1 ⎞ ⎛ 𝐹𝐷 ⎞
(external forces) = 𝜋𝑟20 𝑝1 ⎜ 0 ⎟ − 𝜋𝑟20 𝑝2 ⎜ 0 ⎟ − ⎜ 0 ⎟ (.)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝ 0 ⎠ ⎝ 0 ⎠

Drag force Equating the 𝑥-components of the rate of change in momentum (.) and the external
force (.) results in
1 ( )
𝜌𝜋𝑟20 𝑣21 = 𝜋𝑟20 𝑝1 − 𝑝2 − 𝐹𝐷 (.)
8
We solve for the unknown drag force 𝐹𝐷 :
[( ) ]
1
𝐹𝐷 = 𝜋𝑟20 𝑝1 − 𝑝2 − 𝜌 𝑣21 (.)
8

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.2 Submerged Vessel in an Unbounded Fluid 99

Finally, substituting values for the known quantities on the right-hand side results in:
[ ]
( ) 1 kg 2 m2
𝐹𝐷 = 𝜋1 m 2 2
100500 Pa − 100000 Pa − 1.22 10
8 m3 s2

= 1523 N (.)

Do not forget to check the units: pressure [Pa = N/m2 ] multiplied by surface area [m2 ]
results in a force [N]. The units in the last term reduce to
[ ( ) ( 2 )] [ ]
kg m kg m
m2 =
m3 s2 s2

which is also equal to Newton [N].

9.2 Submerged Vessel in an Unbounded Fluid

Our second example considers a body moving with constant velocity 𝑢0 in positive
𝑥-axis direction. The principal difference from the first example is the fact that there is
no tunnel wall. Instead, the body is moving in an unbounded fluid. The resistance of
the body is to be determined using conservation of mass and momentum principles:
j 𝜕 ( ) j
𝜌 d𝑉 + 𝜌 𝑣𝑇 𝑛 d𝑆 = 0 (.)
𝜕𝑡 ∭ ∬
𝑉 𝑆

𝜕 ( ) ∑
𝜌𝑣 d𝑉 + 𝜌𝑣 𝑣𝑇 𝑛 d𝑆 = (external forces) (.)
𝜕𝑡 ∭ ∬
𝑉 𝑆

We select a < 𝑥, 𝑦, 𝑧 >-coordinate system that is fixed to the moving body (Figure .). Fixed finite but
The control volume 𝑉 is a cylinder with mantle 𝑆𝑐 . Its length and its radius 𝑅𝑐 are open control
volume
chosen so large that the disturbance due to the body has declined. Specifically, we
require that the pressure has returned to its reference value 𝑝0 everywhere outside of
the control volume. The horizontal component of the velocity vector is again 𝑢0 . The
zero flow boundary condition of the wind tunnel example is replaced by a constant
pressure boundary condition. As a consequence, a small mass flow across the mantle
𝑆𝑐 must exist for the submerged body in an unbounded fluid.
The flow is steady and the control volume is fixed with respect to the chosen coordinate Steady flow
system. Consequently, the terms with time derivatives vanish in Equations (.) and assumption
(.), and they simplify to
( ) conservation of
𝜌 𝑣𝑇 𝑛 d𝑆 = 0 (.)
∬ mass, steady flow
𝑆

( ) ∑ conservation of
𝜌 𝑣 𝑣𝑇 𝑛 d𝑆 = (external forces) (.)
∬ momentum, steady flow
𝑆

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

100 9 Application of the Conservation Principles

Figure 9.2 Ellipsoid moving in an unbounded fluid

Velocity field At the inlet 𝑆𝑖 , fluid enters the control volume 𝑉 with constant velocity 𝑢0 . Again, for
j the outside observer the fluid is at rest, and the body moves with speed 𝑢0 . However, for j
an observer on the body, fluid is streaming toward the body. Due to friction between
water and body, a boundary layer develops and the momentum loss will cause a drag
force 𝐹𝐷 acting on the body. The momentum loss becomes apparent in the ‘wake dent’
behind the body (Figure .). At the outlet 𝑆𝑜 of the control volume fluid velocity is
reduced by Δ𝑢 from its initial value 𝑢0 .
In this example we assume the velocity loss Δ𝑢 is radially symmetric to the 𝑥-axis and
given by
⎧ 1 (𝑅 + 𝑟)2 (𝑅 − 𝑟)2 for |𝑟| < 𝑅
Δ𝑢(𝑟) ⎪ 4
= ⎨ 16𝑅 (.)
𝑢0 ⎪0
⎩ elsewhere

The velocity is reduced by a variable Δ𝑢 in a region of the outlet centered around the
𝑥-axis. Obviously, we must choose the control volume radius 𝑅𝑐 to be larger than
the radius 𝑅 of the region with reduced velocity. In an actual application, careful
measurement of the wake field behind the body would provide Δ𝑢 values, which then
would be integrated numerically.

9.2.1 Conservation of mass

Mass flux The boundary surface 𝑆 of the control volume consists of four parts: inlet 𝑆𝑖 , outlet 𝑆𝑜 ,
mantle 𝑆𝑐 , and body 𝑆𝑏 . The interior of the body is not part of the control volume. We

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.2 Submerged Vessel in an Unbounded Fluid 101

split the surface integral in Equation (.) into four parts.

=0
0 = … d𝑆 + … d𝑆 + … d𝑆 + 
>
… d𝑆 (.)
∬ ∬ ∬ ∬ 
𝑆𝑖 𝑆𝑜 𝑆𝑐 𝑆𝑏


Nothing flows through the body surface 𝑆𝑏 and the integrand 𝑣𝑇 𝑛 vanishes.
In contrast to the wind tunnel example, we may not claim that the mass flow integral Mass flux over
vanishes over the mantle surface 𝑆𝑐 . To the contrary, if the flow through the outlet mantle
.
𝑆𝑜 is smaller than the inflow through 𝑆𝑖 , a mass flow 𝑚𝑐 over the mantle exists. From
Equation (.) follows

. ( ) ( ) ( )
𝑚𝑐 = 𝜌 𝑣𝑇 𝑛 d𝑆 = − 𝜌 𝑣𝑇 𝑛 d𝑆 − 𝜌 𝑣𝑇 𝑛 d𝑆 (.)
∬ ∬ ∬
𝑆𝑐 𝑆𝑖 𝑆𝑜

At the inlet, velocity vector and normal vector are

⎛ −𝑢0 ⎞ ⎛1 ⎞
𝑣 = ⎜ 0 ⎟ 𝑛𝑖 = ⎜ 0 ⎟ (.)
⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝0 ⎠

The corresponding vectors for the outlet are

j j
⎛ −(𝑢0 − Δ𝑢) ⎞ ⎛ −1 ⎞
𝑣 = ⎜ 0 ⎟ 𝑛𝑜 = ⎜ 0 ⎟ (.)
⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝ 0 ⎠

Substituting the vectors into Equation . for the mass flow across the mantle yields:

. ( )⎛ 1 ⎞ [ ] ⎛ −1 ⎞
𝑚𝑐 = − 𝜌 − 𝑢0 , 0, 0 ⎜ 0 ⎟ d𝑆 − 𝜌 − (𝑢0 − Δ𝑢), 0, 0 ⎜ 0 ⎟ d𝑆
∬ ⎜ ⎟ ∬ ⎜ ⎟
𝑆𝑖 ⎝0⎠ 𝑆𝑜 ⎝ 0 ⎠

( )
= + 𝜌 𝑢0 d𝑆 − 𝜌 𝑢0 − Δ𝑢 d𝑆
∬ ∬
𝑆𝑖 𝑆𝑜

After extraction of all constant terms in the integrals over 𝑆𝑖 and 𝑆𝑜 , a single integral
remains to be solved:

.
𝑚𝑐 = 𝜌 𝜋𝑅2𝑐 𝑢0 − 𝜌 𝜋𝑅2𝑐 𝑢0 + 𝜌 Δ𝑢 d𝑆 = + 𝜌 Δ𝑢 d𝑆 (.)
∬ ∬
𝑆𝑜 𝑆𝑜

The integration is straightforward in polar coordinates < 𝑟, 𝜑 >. The radial distance 𝑟
from the 𝑥-axis takes values from  to 𝑅𝑐 , or more precisely 𝑟 ∈ [0, 𝑅 < 𝑅𝑐 ], because

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

102 9 Application of the Conservation Principles

Δ𝑢 = 0 for 𝑟 > 𝑅. The angle 𝜑 describes a full circle 𝜑 ∈ [0, 2𝜋]. Substituting
expression (.) for the velocity difference Δ𝑢 gives us

𝑅 2𝜋
.
𝑚𝑐 = 𝜌 Δ𝑢 d𝑆 = 𝜌 𝑢0
1
(𝑅 + 𝑟)2 (𝑅 − 𝑟)2 𝑟 d𝜑d𝑟
∬ ∫ ∫ 16𝑅4
𝑆𝑜 0 0
𝑅
1
= 2𝜋 𝜌 𝑢0 (𝑅 + 𝑟)2 (𝑅 − 𝑟)2 𝑟 d𝑟 (.)
16𝑅4 ∫
0

Expanding all terms leaves a simple polynomial integral to solve.

𝑅 [ ]
. 2𝜋 𝜌 𝑢0 ( ) 2𝜋 𝜌 𝑢0 𝑅4 𝑟2 𝑅2 𝑟4 𝑟6 𝑅
𝑚𝑐 = 𝑅4 − 2𝑅2 𝑟2 + 𝑟4 𝑟 d𝑟 = − +
16𝑅 ∫
4 16𝑅4 2 2 6 0
0
[ ]
2𝜋 𝜌 𝑢0 𝑅6 𝑅6 𝑅6 𝜌
= − + = 𝜋 𝑅2 𝑢0 (.)
16𝑅4 2 2 6 48
.
𝑚𝑐 represents how much mass is ‘escaping’ over the mantle surface 𝑆𝑐 per time unit.
We will employ this result shortly to estimate the amount of longitudinal momentum
that passes through 𝑆𝑐 .
j j
9.2.2 Conservation of momentum

Conservation of momentum states that the rate of change in momentum is equal to

the sum of external forces. We are most interested in the longitudinal component of
the equation. So, flow will be mostly axially symmetric if the body is axially symmetric.
Vortex shedding might cause a time varying transverse force, but we assume here that
the flow is steady.

Sum of external forces

Pressure force External forces consist of body and surface forces. The only body force is the gravity
vanishes force. It is acting perpendicular to the flow and will not affect the resistance of the
body. At the outer surfaces of the control volume 𝑆𝑖 , 𝑆𝑜 , and 𝑆𝑐 the pressure acts with a
constant value 𝑝0 and the resultant pressure force vanishes. The velocity has no gradient
perpendicular to the surfaces 𝑆𝑖 , 𝑆𝑜 , and 𝑆𝑐 . Therefore, no shear stresses occur.
Drag force The only boundary with a resultant pressure and shear stress related force is the body.
The resultant force in 𝑥-direction is the resistance 𝐹𝐷 we are looking for. The resistance
will act against the body’s motion. For the sum of external forces we need the resistance’s
reaction which acts on the fluid in positive 𝑥-direction.

∑ ⎛ 𝐹𝐷 ⎞
(external forces) = ⎜ 0 ⎟ (.)
⎜ ⎟
⎝ 0 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.2 Submerged Vessel in an Unbounded Fluid 103

Rate of change of momentum

The surface integral (.) for the momentum flux is broken into four parts. Momentum flux

=0
( ) 
>
𝜌 𝑣 𝑣𝑇 𝑛 d𝑆 = … d𝑆 + … d𝑆 + … d𝑆 + … d𝑆 (.)
∬ ∬ ∬ ∬ ∬ 
𝑆 𝑆𝑖 𝑆𝑜 𝑆𝑐 𝑆𝑏

The body 𝑆𝑏 is a stream surface and consequently no flow crosses into the body (𝑣𝑇 𝑛𝑐 =
0) and the corresponding integral vanishes.
In the case of the wind tunnel, the third integral vanishes as well. However, in the Momentum flux
current case 𝑆𝑐 is not a solid boundary, as we discovered applying the conservation of through mantle
mass principle. Instead, we made the radius 𝑅𝑐 of the control volume 𝑉 so big that the
velocity vector on the mantle 𝑆𝑐 is in good approximation constant and equal to

⎛ −𝑢0 ⎞
𝑣 = ⎜ 0 ⎟ for 𝑟 = 𝑅𝑐 (.)
⎜ ⎟
⎝ 0 ⎠

Since there is a small mass flow across surface 𝑆𝑐 , the velocity cannot be exactly parallel
to the 𝑥-axis everywhere. However, we may always increase the radius 𝑅 to justify
that 𝑣, 𝑤 ≪ 𝑢0 . As a consequence, the momentum flux across the mantle 𝑆𝑐 can be
.
expressed as a function of the mass flux 𝑚𝑐 (.).
j j
( ) ⎛ −𝑢0 ⎞ ( ) ⎛ −𝑢0 ⎞ ( )
𝜌 𝑣 𝑣𝑇 𝑛𝑐 d𝑆 ≈ 𝜌 ⎜ 0 ⎟ 𝑣𝑇 𝑛𝑐 d𝑆 = ⎜ 0 ⎟ 𝜌 𝑣𝑇 𝑛𝑐 d𝑆
∬ ∬ ⎜ ⎟ ⎜ ⎟∬
𝑆𝑐 𝑆𝑐 ⎝ 0 ⎠ ⎝ 0 ⎠ 𝑆𝑐
⎛ −𝑢0 ⎞
≈ ⎜ 0 ⎟ 𝑚𝑐
. (.)
⎜ ⎟
⎝ 0 ⎠

The momentum flux over the inlet 𝑆𝑖 is Momentum flux

through inlet

( ) ⎛ −𝑢0 ⎞ ⎡( ) ⎛ 1 ⎞⎤ ⎛ 𝑢20 ⎞
𝜌 𝑣 𝑣𝑇 𝑛𝑖 d𝑆 = ⎜
𝜌 0 ⎟ ⎢ − 𝑢0 , 0, 0 0⎜ ⎟⎥ d𝑆 = 𝜌 𝜋𝑅𝑐 0 ⎟
2⎜
(.)
∬ ∬ ⎜ ⎟⎢ ⎜ ⎟⎥ ⎜ ⎟
𝑆𝑖 𝑆𝑖 ⎝ 0 ⎠⎣ ⎝ 0 ⎠⎦ ⎝ 0 ⎠

For the momentum flux integral over the outlet 𝑆𝑜 , we employ polar coordinates. The Momentum flux
surface element becomes d𝑆 = 𝑟d𝜑d𝑟. through outlet

( ) ⎛ −(𝑢0 − Δ𝑢) ⎞ ⎡( ) ⎛ −1 ⎞⎤
𝜌 𝑣 𝑣𝑇 𝑛𝑜 d𝑆 = 𝜌⎜ 0 ⎟ ⎢ − (𝑢 − Δ𝑢), 0, 0 ⎜ 0 ⎟⎥ d𝑆
0
∬ ∬ ⎜ ⎟⎢ ⎜ ⎟⎥
𝑆𝑜 𝑆𝑜 ⎝ 0 ⎠⎣ ⎝ 0 ⎠⎦
𝑅𝑐 2𝜋
⎛ −𝑢20 + 2𝑢0 Δ𝑢 − (Δ𝑢)2 ⎞
= 𝜌 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟
∫ ∫ ⎜ ⎟
0 0 ⎝ 0 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

104 9 Application of the Conservation Principles

Splitting the integral into parts and solving the first integral yields:

𝑅 2𝜋
( 𝑇 ) ⎛ −𝑢20 ⎞ ⎛ 2𝑢0 Δ𝑢 ⎞
2⎜ ⎟ ⎜ 0 ⎟ 𝑟 d𝜑d𝑟
𝜌 𝑣 𝑣 𝑛𝑜 d𝑆 = 𝜌 𝜋𝑅𝑐 0 +𝜌 (.)
∬ ⎜ ⎟ ∫ ∫ ⎜ ⎟
𝑆𝑜 ⎝ 0 ⎠ 0 0 ⎝ 0 ⎠
𝑅 2𝜋
⎛ −(Δ𝑢)2 ⎞
+𝜌 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟
∫ ∫ ⎜ ⎟
0 0 ⎝ 0 ⎠

The first term in the momentum flux over the outlet (.) will cancel with the contri-
bution from the inlet (.).
The two remaining integrals involve the change in velocity Δ𝑢 and because Δ𝑢 = 0 for
𝑟 > 𝑅, the upper limit of the integrals has been adjusted to 𝑅 (which is smaller than 𝑅𝑐 ).
Substitution of Equation (.) into the first remaining integral results in an integral
.
similar to the one already solved for the mass flux 𝑚𝑐 .

𝑅 2𝜋 𝑅 2𝜋
⎛ 2𝑢0 Δ𝑢 ⎞ 2𝜌 𝑢20 ⎛ (𝑅 + 𝑟)2 (𝑅 − 𝑟)2 ⎞
𝜌 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟 = ⎜ 0 ⎟ 𝑟 d𝜑d𝑟
∫ ∫ ⎜ ⎟ 16𝑅4 ∫ ∫ ⎜ ⎟
0 0 ⎝ 0 ⎠ 0 0 ⎝ 0 ⎠
⎛ 𝑅6 ⎞
j 4𝜋 𝜌 𝑢20 ⎜ 6 ⎟ j
= ⎜ ⎟
16𝑅4 ⎜ 0 ⎟
⎝ 0 ⎠
⎛ 1 ⎞
⎜ 24 ⎟
= 𝜌 𝜋𝑅2 𝑢20 ⎜ ⎟ (.)
⎜ 0 ⎟
⎝ 0 ⎠

The second integral over (Δ𝑢)2 is often neglected if the changes in velocity at the outlet
are small. We have an analytical expression, therefore solving the integral poses no
further difficulty.

𝑅 2𝜋 [ ]
⎛ −(Δ𝑢)2 ⎞ 2𝜋 𝜌 𝑢20
𝑅
⎛ − (𝑅 + 𝑟)2 (𝑅 − 𝑟)2 2 ⎞
𝜌 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟 = ⎜ 0 ⎟ 𝑟 d𝑟
∫ ∫ ⎜ ⎟ 256 𝑅 8 ∫ ⎜ ⎟
0 0 ⎝ 0 ⎠ 0 ⎝ 0 ⎠
8 𝑅 6 2 4 4 2 6 8
⎛ −𝑅 + 4𝑅 𝑟 − 6𝑅 𝑟 + 4𝑅 𝑟 − 𝑟
2𝜋 𝜌 𝑢20 ⎞
= ⎜ 0 ⎟ 𝑟 d𝑟
256 𝑅8 ∫ ⎜ ⎟
0 ⎝ 0 ⎠
8 2 2 8 10 𝑅
⎡⎛ − 𝑅 𝑟 + 𝑅6 𝑟4 − 𝑅4 𝑟6 + 𝑅 𝑟 − 𝑟 ⎞⎤
2𝜋 𝜌 𝑢20 ⎢⎜ 2 2 10 ⎟⎥
= ⎢⎜ 0 ⎟⎥
256 𝑅8 ⎢⎜ ⎟⎥
⎣⎝ 0 ⎠⎦0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

9.2 Submerged Vessel in an Unbounded Fluid 105

10 10 10
𝑅 2𝜋
⎛ −(Δ𝑢)2 ⎞ ⎛ − 𝑅 + 𝑅10 − 𝑅10 + 𝑅 − 𝑅 ⎞
2𝜋 𝜌 𝑢20 ⎜ 2 2 10 ⎟
𝜌 ⎜ 0 ⎟ 𝑟 d𝜑d𝑟 = ⎜ ⎟
∫ ∫ ⎜ ⎟ 256 𝑅8 ⎜ 0 ⎟
0 0 ⎝ 0 ⎠ ⎝ ⎠
0
⎛− 1 ⎞
⎜ 1280 ⎟
= 𝜌 𝜋𝑅2 𝑢20 ⎜ ⎟ (.)
⎜ 0 ⎟
⎝ 0 ⎠

Finally, we can collect the results for the momentum flux over the boundary 𝑆 of the Momentum flux
control volume. With results from Equations (.), (.), (.), (.), (.), and result
(. we get
⎡⎛ 1 ⎞ ⎛ 1 ⎞ ⎛ 1 ⎞⎤
− −
( 𝑇 ) ⎢⎜ 24 ⎟ ⎜ 1280 ⎟ ⎜ 48 ⎟⎥
𝜌𝑣 𝑣 𝑛 d𝑆 = 𝜌 𝜋𝑅2 𝑢20 ⎢⎜
0 ⎟⎟ ⎜⎜ ⎟ + ⎜ 0 ⎟⎥
+
∬ ⎢⎜ 0 ⎟ ⎜ ⎟⎥
𝑆 ⎣⎝ 0 ⎠ ⎝ 0 ⎠ ⎝ 0 ⎠⎦
⎛ 77 ⎞
⎜ 3840 ⎟
𝜌 𝜋𝑅2 𝑢20 ⎜ (.)
0 ⎟⎟
=
⎜
⎝ 0 ⎠

Equating the result for the momentum flux with the sum of external forces provides us Drag force
j with an equation for the drag force 𝐹𝐷 acting on the body. j

⎛ 77 ⎞ ⎛ 𝐹𝐷 ⎞
2 2 ⎜ 3840 ⎟ ⎜ ⎟
𝜌 𝜋𝑅 𝑢0 ⎜ ⎟ = ⎜ 0 ⎟ (.)
⎜ 0 ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝ 0 ⎠

𝐹𝐷 ≈ 0.020052 𝜌 𝜋𝑅2 𝑢20 (.)

As it should be, the drag force will not depend on the cross section of the control
𝜋𝑅2𝑐
volume, if it is chosen large enough. It solely depends on magnitude and extent of the
wake dent.
If we assume a body velocity of 𝑢0 = 2.5 m/s and a radius of the wake 𝑅 = 1.0 m, the
resistance of the body in salt water of density 𝜌 = 1026.021 kg/m3 will be
𝐹𝐷 = 0.020052 ⋅ 1026.021 kg/m3 𝜋 1.02 m2 2.52 (m/s)2
kg m
= 403.96
s2
= 403.96 N
The units work out as expected and the result is a force in Newton.
As you hopefully noted, this type of analysis does not require any knowledge about the
geometry of the body. It could have any shape. The only requirement is the accurate
measurement of magnitude and extent of the wake field. If the body causes noticeable
asymmetries in the flow, careful assessment of the three-dimensional wake will also
deliver the force components in transverse and vertical direction.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

106

10
Boundary Layer Theory

This chapter introduces the basic equations of boundary layer theory, which was orig-
inally proposed by Ludwig Prandtl in  (see Prandtl et al., ). Together with
the following three chapters, this is necessarily a fairly short treatise on this complex
subject, which is still the topic of ongoing research. The most comprehensive text
on boundary layer theory is the book by Schlichting and Gersten (). Hermann
Schlichting (* – †) was a student of Ludwig Prandtl. Another good reference
is the book by Young ().
We will first discuss some phenomena associated with the boundary layer and then re-
turn to the Navier-Stokes equations. By making certain assumptions on flow conditions
in the boundary layer, we will derive the actual boundary layer equations.
j j

Learning Objectives
At the end of this chapter students will be able to:

• discuss the concept of boundary layers

• describe the differences between laminar and turbulent boundary layer flow
• explain flow separation
• work with dimensionless equations
• derive the boundary layer equations

10.1 Boundary Layer

Conundrum of In Section ., we derived the dimensionless factors in the NSE (.) and computed
viscous forces them for a  m long ship. The factor for the viscous forces equals the reciprocal value
1∕𝑅𝑒 of the Reynolds number. In the example, this viscous force factor is about 10−10
times smaller than the factors associated with other forces, which supports the notion
that viscous forces possibly could be ignored in ship hydrodynamics. However, simply
omitting the viscous forces leads to obviously erroneous resistance predictions.
Boundary layer Ludwig Prandtl solved this apparent dilemma by splitting the flow into two regions:

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.1 Boundary Layer 107

Figure 10.1 Basic properties of the velocity

distribution in the boundary layer

• A region close to the body which we call the boundary layer. Boundary layers
cover solid surfaces and boundaries between fluids that do not completely mix
(like air and water). Within this thin sheet of flow, viscous forces cannot be
j j
neglected, that is, (1∕𝑅𝑒) 𝚫∗ 𝑣∗ represents a noticeable force despite its small
leading factor.

• An exterior flow outside of the boundary layer where viscous forces can be
neglected in most practical cases.

Basic properties of the boundary layer are discussed below.

10.1.1 Boundary layer thickness

Figure . shows the velocity distribution near a body surface. For convenience, we Definition of
assume that the curvature of the surface is small and that the 𝑥-axis is parallel to the boundary layer
thickness
wall. With increasing distance 𝑦 from the wall, velocity grows continuously from zero
to the exterior flow velocity 𝑢𝑒 . Because there is no obvious break or knuckle in the
velocity distribution, it is hard to determine where the boundary layer ends and the
exterior flow begins. In experiments, we usually mark the outer edge of the boundary
layer at a distance from the wall where the velocity 𝑢 reaches % of the exterior flow
velocity (Figure .). The distance from the surface to the outer edge is called boundary
layer thickness 𝛿.
At the body surface, fluid molecules stick to the body. We call this the no slip condition of No slip condition
boundary layer theory. This assumption has been verified in experiments by observation
and measurements.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

108 10 Boundary Layer Theory

Figure 10.2 Transition from laminar to turbulent

flow of the air rising from a burning candle.
Reproduced with kind permission by Dr. Gary S.
Settles, Floviz, Inc.

j j

10.1.2 Laminar and turbulent flow

Laminar and You have probably watched the rising smoke of a cigarette or from a candle which just
turbulent flow got extinguished. The smoke forms a smooth band for some distance above the tip
but soon starts meandering and then seems to break up in chaotic swirls and eddies.
Figure . visualizes the air flow above a burning candle. We clearly see the straight
flow right above the flame. Apparently, the flow is stable and does not seem to be
time dependent. We call this flow laminar, derived from the Latin word ‘lamina’, which
means ‘thin sheet.’ After a short distance the flow starts swerving left and right before it
becomes fully turbulent. In turbulent flow, fluid particles deviate from the general flow
direction in seemingly random moves, sometimes even moving against the general flow
for a short while. Therefore, turbulent flow is constantly changing over time. However,
more often than not, we treat turbulent flow as steady by splitting the time dependent
velocity 𝑣(𝑡) into a mean velocity 𝑣 and a turbulent fluctuation 𝑣′ (𝑡): 𝑣(𝑡) = 𝑣 + 𝑣′ (𝑡).
For details see Chapter .
Laminar flow Boundary layer flows are laminar only for small local Reynolds numbers 𝑅𝑒𝑥 < 500 000
(Fox and McDonald, ). The local Reynolds number 𝑅𝑒𝑥 is based on the current
position 𝑥 rather than on a fixed reference length 𝐿.
𝑢𝑒 𝑥
𝑅𝑒𝑥 = local Reynolds number (.)
𝜈
𝑅𝑒𝑥 changes along the body surface!

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.1 Boundary Layer 109

(a) Flow in laminar boundary layer (b) Flow in turbulent boundary layer

Figure 10.3 Flow characteristics of laminar and turbulent boundary layers

In laminar boundary layers, the flow appears to be neatly stacked in tiers (see Fig-
ure .(a)). There is neither momentum exchange with the exterior flow nor between
the tiers.
j j
Under ideal conditions, a boundary layer can stay laminar beyond 𝑅𝑒𝑥 > 500 000. Laminar-
However, every flow experiences a varying level of small disturbances, called preturbu- turbulent
transition
lence. Preturbulence, surface roughness, or obstacles on the surface may trigger the
boundary layer to transition from laminar to turbulent flow at lower Reynolds numbers.
With increasing local Reynolds number, the flow transitions from laminar to turbulent
somewhere in the range 200 000 < 𝑅𝑒𝑥 < 1 000 000. The exact position depends on the
amount of preturbulence, surface roughness, and other factors.
For Reynolds numbers 𝑅𝑒𝑥 > 500 000 the boundary layer is usually treated as turbulent Turbulent flow
(see Figure .(b)). With turbulence, fluctuations in the transverse velocity 𝑣 – though
small overall – become large enough to cause momentum exchange within the boundary
layer and with the exterior flow. Transverse velocity must diminish close to the wall.
Therefore, turbulent boundary layers usually have a laminar sublayer. In modern
literature it is called viscous sublayer because it is governed by the viscosity of the fluid.
Figure . illustrates the development of a boundary layer over a flat plate. Its thickness Boundary layer
is zero at the beginning of a plate and continuously grows toward the end of the plate. thickness
Note that we exaggerate the thickness in essentially all schematic drawings of the
boundary layer in order to show what is happening in it. The boundary layer thickness
at the stern of a ship is in the order of % of the vessel’s length.
Initially, the flow is laminar in the boundary layer. The formula given in Figure . Laminar versus
for the laminar boundary layer thickness 𝛿𝐿 reflects Blasius’ solution of the boundary turbulent
layer equations (Blasius, ). 𝛿𝐿 is a function of distance from the leading edge 𝑥
and the local Reynolds number 𝑅𝑒𝑥 . However, with increasing local Reynolds number
𝑅𝑒𝑥 , the boundary layer becomes unstable and transitions into a turbulent flow. The
boundary layer thickens more quickly in a turbulent boundary layer. The equation

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

110 10 Boundary Layer Theory

Figure 10.4 Development of the boundary layer along a flat surface. Note that the outer limit of
the boundary layer is not a streamline

stated in Figure . for the turbulent boundary layer thickness 𝛿𝑇 is an approximate
solution for moderate Reynolds numbers (≈ 107 ). It ignores the initial laminar part of
the boundary layer.
The exact line of transition is hard to predict because it depends not only on the local
Reynolds number but also on preturbulence, surface roughness, and even the shape of
j the leading edge of the plate. Low preturbulence, smooth surface, and round leading j
edges tend to delay the transition.
The distinction between laminar and turbulent boundary layers is essential in ship
model testing. Boundary layers of full scale ships are turbulent due to the high Reynolds
numbers (𝑅𝑒 > 108 ) for ship flows. Naval architects must ensure that the boundary
layer on a ship model is also turbulent. Otherwise, flow patterns observed on a model
may not be comparable to those of the full scale vessel, rendering the model test results
pretty much useless.

10.1.3 Flow separation

Separation On curved surfaces we may encounter an additional flow feature called separation. It
commonly occurs with blunt bodies like spheres. A well-designed ship hull should not
suffer measurably from flow separation, especially with a working propeller at the stern.
Water sticks to the hull surface. Therefore, the velocity tangential to the wall vanishes
there: 𝑢(𝑦 = 0) = 0. Due to the viscosity, fluid sticking to the wall slows down fluid
particles close to the wall. Those, in turn, will slow down neighboring particles. With
increasing position 𝑥, the zone of slowed down particles stretches farther out into
the flow. Consequently, the boundary layer thickens. The initially positive transverse
derivative 𝜕𝑢∕𝜕𝑦 diminishes (Figure .). Especially in regions where the exterior
flow slows down (𝜕𝑢𝑒 ∕𝜕𝑥 < 0), the flow near but not quite at the wall stops as well.
Eventually, the flow reaches a separation point with 𝜕𝑢∕𝜕𝑦 = 0. Further downstream a
region of separated flow develops in which the direction of flow is reversed.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.2 Simplifying Assumptions 111

Figure 10.5 Development of velocity profile in the boundary layer along a curved surface with
flow separation

Shear stress in laminar boundary layers is smaller than shear stress in turbulent bound- Effect on
ary layers. As a consequence, frictional resistance is lower for laminar boundary layers. resistance
However, the latter are more prone to flow separation. Within a region of separated
flow, pressure is not rebuilt to the levels at the bow of a vessel, which causes a significant
increase in viscous pressure resistance. Turbulent boundary layers withstand separation
better because the momentum exchange with the exterior flow replenishes energy lost
to friction. With delayed separation, pressure is rebuilt over a larger portion of the hull
surface. As a consequence, the total viscous resistance for turbulent boundary layers
with less or no separation in many cases is smaller than the total viscous resistance of
laminar boundary layers.
j j

10.2 Simplifying Assumptions

Exterior flow and boundary layer flow must satisfy the principles of conservation of Two-dimensional
mass and momentum. Therefore, continuity and Navier-Stokes equations apply to both flow
boundary layer and exterior flow. For our discussion of the boundary layer, we restrict
ourselves to two-dimensional, incompressible flow in the 𝑥-𝑦–plane. Two dimensional
flow simply means that the same flow field is observed in any plane with constant 𝑧
coordinate. Consequently, derivatives in 𝑧-direction vanish and the velocity vector
becomes 𝑣 = (𝑢, 𝑣)𝑇 .
We assume here that the fluid is incompressible and hence the density 𝜌 is constant. Con- Incompressible,
sequently, the continuity equation (.) reduces to the requirement that the divergence steady 2D-flow
of the velocity vanishes:
∇𝑇 𝑣 = 0 (.)
The Navier-Stokes equations for incompressible flow of a Newtonian fluid read as

𝜕𝑣 ( ) 1
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣 (.)
𝜕𝑡 𝜌

We further assume that the flow is steady, which eliminates the first term in Equa-
tion (.):
𝜕𝑣
= 0 for steady flow (.)
𝜕𝑡

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

112 10 Boundary Layer Theory

Restating continuity and Navier-Stokes equations in components for two-dimensional,

incompressible, steady flow yields:

𝜕𝑢 𝜕𝑣
+ = 0 D continuity equation
𝜕𝑥 𝜕𝑦
𝜕𝑢 𝜕𝑢 1 𝜕𝑝
𝑢 +𝑣 = 𝑓𝑥 − + 𝜈𝚫𝑢 𝑥-component of D NSE (.)
𝜕𝑥 𝜕𝑦 𝜌 𝜕𝑥
𝜕𝑣 𝜕𝑣 1 𝜕𝑝
𝑢 +𝑣 = 𝑓𝑦 − + 𝜈𝚫𝑣 𝑦-component of D NSE
𝜕𝑥 𝜕𝑦 𝜌 𝜕𝑦

Unfortunately, this system of nonlinear, partial differential equations is very difficult to

solve.
2D conservation Following Prandtl’s ideas, we will eliminate some terms in Equations (.). For details
of mass and see Schlichting and Gersten (, Chapter ). In order to decide which terms are of
momentum
equations
lesser importance, we have to determine their relative magnitude. This is best done
by studying the dimensionless equivalent of Equations (.). External flow speed 𝑢𝑒 ,
length of body 𝐿, gravitational acceleration 𝑔, and twice the dynamic pressure 𝜌𝑢2𝑒 serve
as reference quantities. Following the process illustrated in Section ., we get the two-
dimensional conservation of mass and momentum equations of steady, incompressible
flow in dimensionless form:
𝜕𝑢∗ 𝜕𝑣∗
+ = 0
𝜕𝑥∗ 𝜕𝑦∗
j ( 2 ∗ ) j
𝜕𝑢∗ 𝜕𝑢∗ 1 ∗ 𝜕𝑝∗ 1 𝜕 𝑢 𝜕 2 𝑢∗
𝑢∗ ∗ + 𝑣∗ ∗ = 𝑓𝑥 − 𝐸𝑢 ∗ + + (.)
𝜕𝑥 𝜕𝑦 𝐹𝑟2 𝜕𝑥 𝑅𝑒 𝜕𝑥∗ 2 𝜕𝑦∗ 2
∗ ∗ ∗
( )
∗ 𝜕𝑣 ∗ 𝜕𝑣 1 ∗ 𝜕𝑝 1 𝜕 2 𝑣∗ 𝜕 2 𝑣∗
𝑢 +𝑣 = 𝑓 − 𝐸𝑢 ∗ + +
𝜕𝑥∗ 𝜕𝑦∗ 𝐹𝑟2 𝑦 𝜕𝑦 𝑅𝑒 𝜕𝑥∗ 2 𝜕𝑦∗ 2

Geometric The longitudinal coordinate 𝑥∗ = 𝑥∕𝐿 is pointing along the body contour, whereas the
relationships transverse coordinate 𝑦∗ = 𝑦∕𝐿 is pointing in normal direction to the body surface.
Although we use Cartesian coordinates in the equation, 𝑥∗ and 𝑦∗ may also be read
as more general curvilinear coordinates. The longitudinal coordinate 𝑥 takes values
between zero and the reference length 𝐿. Consequently, the dimensionless coordinate
𝑥∗ ranges from zero to one. We say the magnitude of a dimensionless coordinate 𝑥∗ is
of order one.
𝑥∗ ∼ 1 (.)
The transverse coordinate 𝑦∗ = 𝑦∕𝐿 grows from zero up to the boundary layer thickness
𝛿 ∗ = 𝛿∕𝐿. We conclude that 𝑦∗ is of the same order of magnitude as the boundary layer
thickness 𝛿 ∗ .
𝑦∗ ∼ 𝛿 ∗ (.)

Longitudinal The longitudinal velocity 𝑢 reaches the value 𝑢𝑒 at the outer edge of the boundary layer.
velocity and Consequently, the dimensionless longitudinal velocity 𝑢∗ = 𝑢∕𝑢𝑒 is also of order one.
derivatives
𝑢∗ ∼ 1 (.)

We further assume that the magnitude of 𝑢∗ remains of the same order along the body.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.2 Simplifying Assumptions 113

This implies that the derivative 𝜕𝑢∗ ∕𝜕𝑥∗ must be of order one as well:

𝜕𝑢∗
𝑢∗ ≈ 𝑥∗ (.)
𝜕𝑥∗
order of magnitude ∼1 ≈∼1⋅1

Otherwise the velocity would grow beyond magnitude one. In practice this means that
the body surface changes only gradually and has no sharp corners. With both 𝜕𝑢∗ ∕𝜕𝑥∗
and 𝑥∗ of magnitude one, the second order derivative 𝜕 2 𝑢∗ ∕𝜕𝑥∗ 2 will be of magnitude
one as well.
The transverse derivative of the horizontal velocity 𝜕𝑢∗ ∕𝜕𝑦∗ must be of magnitude 1∕𝛿 ∗
in order for the velocity 𝑢 to grow from zero at the wall (no-slip condition) to its full
value at the outer edge of the boundary layer:

𝜕𝑢∗
𝑢∗ ≈ 𝑦∗ (.)
𝜕𝑦∗
1
order of magnitude ∼ 1 ≈ ∼ 𝛿∗ ∗ = 1
𝛿
Then, since the transverse distance 𝑦∗ is of magnitude 𝛿 ∗ , the second order transverse
derivative 𝜕 2 𝑢∗ ∕𝜕𝑦∗ 2 must be of magnitude 1∕𝛿 ∗ 2 :

𝜕𝑢∗ 𝜕 2 𝑢∗
≈ 𝑦∗ (.)
j 𝜕𝑦 ∗
𝜕𝑦∗ 2 j
1 1 1
order of magnitude ∼ ∗ ≈ ∼ 𝛿∗ = ∗
𝛿 𝛿∗2 𝛿

The transverse velocity 𝑣∗ vanishes at the wall (no flow through wall). From the dimen- Transverse
sionless continuity equation follows that 𝜕𝑢∗ ∕𝜕𝑥∗ and 𝜕𝑣∗ ∕𝜕𝑦∗ must be of the same velocity and
derivatives
order of magnitude (one).

𝜕𝑢∗ 𝜕𝑣∗
∗
+ = 0 (.)
𝜕𝑥 𝜕𝑦∗
∼1− ∼1 = 0

With the derivative in 𝑦∗ -direction being of order one, the transverse velocity 𝑣∗ will be
of the same order as dimensionless boundary layer thickness 𝛿 ∗ = 𝛿∕𝐿.

𝜕𝑣∗
𝑣∗ ≈ 𝑦∗ (.)
𝜕𝑦∗
order of magnitude ∼ 𝛿∗ ≈ ∼ 𝛿 1 = 𝛿∗
∗

With 𝑣∗ of magnitude 𝛿 ∗ and 𝑥∗ of magnitude one, its derivative 𝜕𝑣∗ ∕𝜕𝑥∗ will be of
magnitude 𝛿 ∗ .

𝜕𝑣∗
𝑣∗ ≈ 𝑥∗ (.)
𝜕𝑥∗
order of magnitude ∼ 𝛿∗ ≈ ∼ 1 𝛿∗ = 𝛿∗

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

114 10 Boundary Layer Theory

Its second order derivative 𝜕 2 𝑣∗ ∕𝜕𝑥∗ 2 will also be of magnitude 𝛿 ∗ :

𝜕𝑣∗ 2 ∗
∗ 𝜕 𝑣
≈ 𝑥 (.)
𝜕𝑥∗ 𝜕𝑥∗ 2
order of magnitude ∼ 𝛿 ≈ ∼ 1 𝛿∗ = 𝛿∗
∗

Relative Now that you are thoroughly confused, let’s rewrite first the longitudinal conservation
magnitude of of momentum Equation (.) and note underneath each term the order of magnitude
terms
we found in the paragraphs above.
( 2 ∗ )
𝜕𝑢∗ 𝜕𝑢∗ 1 ∗ 𝜕𝑝∗ 1 𝜕 𝑢 𝜕 2 𝑢∗
𝑢∗ ∗ + 𝑣∗ ∗ = 𝑓𝑥 − 𝐸𝑢 ∗ + + (.)
𝜕𝑥 𝜕𝑦 𝐹𝑟2 𝜕𝑥 𝑅𝑒 𝜕𝑥∗ 2 𝜕𝑦∗ 2
( )
1 1
∼ 1 ⋅ 1 + 𝛿∗ ∗ ≈ ∼ 0 − 1 + 𝛿∗2 1 +
𝛿 𝛿∗2
Obviously, the left-hand side representing the 𝑥-component of the inertia force is of
magnitude one. The right-hand side of Equation (.) is equivalent to the sum of
external forces in 𝑥-direction.
We know from observation that the boundary layer thickness is small compared with
the length of a body, i.e. 𝛿 ≪ 𝐿 or 𝛿 ∗ ≪ 1. Accordingly, the second term dominates the
first term in the parentheses on the right-hand side:
1
≫ 1
𝛿∗2
j j
However, the components of an external force may not be larger than one in magnitude.
Otherwise, the equality cannot be maintained in Equation (.). As a consequence,
the inverse of the Reynolds number must be of small magnitude to reduce the viscous
forces to magnitude one.
1 1
∼ 𝛿∗2 or 𝛿∗ ∼ √ (.)
𝑅𝑒 𝑅𝑒
In turn, the dimensionless boundary layer thickness is proportional to the square root
of the inverse Reynolds number. We will come back to this important conclusion. Note
that for the limit case 𝑅𝑒 → ∞, the boundary layer vanishes as lim𝑅𝑒→∞ 𝛿 = 0. This
applies to a hypothetical, inviscid fluid.
Before we simplify this equation, we repeat the process for the 𝑦-component of the con-
servation of momentum equation, assuming the same relationship between Reynolds
number and boundary layer thickness.
( )
𝜕𝑣∗ 𝜕𝑣∗ 1 ∗ 𝜕𝑝∗ 1 𝜕 2 𝑣∗ 𝜕 2 𝑣∗
𝑢∗ ∗ + 𝑣∗ ∗ = 𝑓𝑦 − 𝐸𝑢 ∗ + + (.)
𝜕𝑥 𝜕𝑦 𝐹𝑟2 𝜕𝑦 𝑅𝑒 𝜕𝑥∗ 2 𝜕𝑦∗ 2
( )
1
∼ 1 𝛿∗ + 𝛿∗ 1 ≈ ∼ 0 − 1 + 𝛿∗2 𝛿∗ + ∗
𝛿
Again, inertia force and viscous force are of the same magnitude. However, the 𝑦-
components are of magnitude 𝛿 ∗ ≪ 1, which renders them considerably smaller than
their 𝑥-direction counterparts.
Assumptions of Based on these observations, Prandtl formulated his basic assumptions of boundary
boundary layer layer theory:
theory

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.3 Boundary Layer Equations 115

. The boundary layer is thin, i.e. Boundary layer

must be thin
𝛿 ≪ 𝑥 ≤ 𝐿 (.)

. Fluid sticks to the wall, i.e. No-slip condition

𝑢 = 0 for 𝑦 = 0, no-slip condition (.)

. Flow speed changes rapidly across the boundary layer from  at the wall (no-slip
condition (.)) to the exterior flow speed at the boundary layer thickness 𝛿.
This means the derivative of the flow speed in direction normal to the wall is
large.
𝜕𝑢 𝑢
∼ 𝑒 is large (.)
𝜕𝑦 𝛿
. Flow speed changes slowly along 𝑥 so that Slope and
curvature of
𝜕𝑢 𝑢 body are small
∼ 𝑒 is moderate (.)
𝜕𝑥 𝐿

. The second order derivative of 𝑢 with respect to 𝑥 is negligible compared with its
second order derivative with respect to 𝑦:

𝜕2𝑢 𝜕2𝑢
≪ (.)
2 𝜕𝑦2
j 𝜕𝑥 j

. Transverse velocity 𝑣 and its 𝑥-derivatives are small within the boundary layer

𝛿 𝜕𝑣 𝛿 𝜕2𝑣 𝛿
𝑣 ∼ 𝑢𝑒 ∼ 𝑢𝑒 ∼ 𝑢𝑒 (.)
𝐿 𝜕𝑥 𝐿2 𝜕𝑥 2 𝐿3

. The 𝑦-derivatives of the transverse velocity are moderate to large:

𝜕𝑣 𝑢 𝜕2𝑣 𝑢
∼ 𝑒 and ∼ 𝑒 (.)
𝜕𝑦 𝐿 𝜕𝑦2 𝐿𝛿

. In addition, we will ignore the effect of gravity on the boundary layer flow. In Body forces will
a control volume completely filled with incompressible fluid, gravity forces and be ignored
hydrostatic forces are in equilibrium. When we omit the gravity forces in Equa-
tion (.), the pressure 𝑝 is understood to be the dynamic pressure only.

In the next section, the assumptions of boundary layer theory are employed to derive
the simplified boundary layer equations.

10.3 Boundary Layer Equations

We start again with the two-dimensional continuity equation and Navier-Stokes equa- Two-dimensional
tions for steady, incompressible flow (.). The body forces 𝑓𝑥 and 𝑓𝑦 are omitted flow
based on the facts that the gravity force is the only body force of interest and that it

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

116 10 Boundary Layer Theory

may be neglected according to assumption No. . Some terms are annotated to indicate
whether they are small or very small based on the assumptions of boundary layer theory
above.
𝜕𝑢 𝜕𝑣
+ = 0 D continuity equation
𝜕𝑥 𝜕𝑦
very small
𝜕𝑢 𝜕𝑢 1 𝜕𝑝 2
𝜕 𝑢 𝜕2𝑢
𝑢 +𝑣 = − + 𝜈 + 𝜈 (.)
𝜕𝑥 𝜕𝑦 𝜌 𝜕𝑥 𝜕𝑥 2 𝜕𝑦2
small small very small small
𝜕𝑣 𝜕𝑣 1 𝜕𝑝 2
𝜕 𝑣 𝜕 2 𝑣
𝑢 +𝑣 = − + 𝜈 + 𝜈
𝜕𝑥 𝜕𝑦 𝜌 𝜕𝑦 𝜕𝑥2 𝜕𝑦2

2D boundary Finally, omitting all terms that are labeled ‘small’ or ‘very small,’ we obtain the two-
layer equations dimensional boundary layer equations
𝜕𝑢 𝜕𝑣
+ = 0 D continuity equation (.)
𝜕𝑥 𝜕𝑦
𝜕𝑢 𝜕𝑢 1 𝜕𝑝 𝜕2𝑢
𝑢 +𝑣 = − + 𝜈 (.)
𝜕𝑥 𝜕𝑦 𝜌 𝜕𝑥 𝜕𝑦2
1 𝜕𝑝
0 = − (.)
𝜌 𝜕𝑦
j This set of coupled partial differential equations is much less complex than the original j
Navier-Stokes equations (.). However, the set is still nonlinear due to the convection
terms on the left-hand side of the conservation of momentum equation (.). An
analytic solution is not known for general cases, but this set of equations can be inte-
grated numerically. Schlichting and Gersten (, Chapter ) provides an overview
and further details may be found in Krause ().
Blasius () solved the equations by transformation and series expansion for the case
of a flat plate. An English version of his text can be found in Blasius (). We will
study the case of a flat plate in Chapter .
Pressure is However, without actually solving Equations (.) through (.), we may already
constant across draw an important conclusion. Take a closer look at the 𝑦-component of the conserva-
thin boundary
layers
tion of momentum equation (.)! It states that the derivative of pressure vanishes
across the boundary layer:
𝜕𝑝
0 = or 𝑝(𝑦) = const. across boundary layer (.)
𝜕𝑦
In other words, the pressure does not change across the boundary layer. As a conse-
quence, the pressure of the exterior flow will also act on the body surface.
This conclusion is important because it allows us to derive pressure forces from exterior
flow solutions, which we may compute by neglecting viscous effects. This is still far
from trivial and we will discuss some basics later on. Of course, you have to remember
that this will yield useful results only as long as the boundary layer is thin.
Next, we will apply the mass and momentum conservation principles to a piece of the
boundary layer to study the relationship between flow velocities governed by Equa-

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

10.3 Boundary Layer Equations 117

tions (.)–(.) and the shear stress acting on the wall. Knowing the shear stress
will enable us to compute the frictional resistance.

References
Blasius, H. (). Grenzschichten in Flüssigkeiten mit kleiner Reibung. PhD thesis,
Universität Göttingen.
Blasius, H. (). The boundary layers in fluids with little friction. Technical Report
TM-, National Advisory Committee for Aeronautics (NACA). Translation of
Blasius (). Grenzschichten in Flüssigkeiten mit kleiner Reibung. Zeitschrift für
Mathematik und Physik, ():–.
Fox, R. and McDonald, A. (). Introduction to Fluid Mechanics. John Wiley & Sons,
New York, NY.
Krause, E. (). Numerical solution of the boundary-layer equations. AIAA Journal,
():–. doi: ./..
Prandtl, L., Oswatitsch, K., and Wieghardt, K. (). Führer durch die Strömungslehre.
Friedrich Vieweg & Sohn, Braunschweig/Wiesbaden, ninth edition. Verbesserte und
erweiterte Auflage.
Schlichting, H. and Gersten, K. (). Boundary-layer theory. Springer-Verlag, Berlin,
Heidelberg, New York, eigth edition. Corrected printing.
j Young, A. (). Boundary layers. AIAA Education Series. American Institute of j
Aeronautics and Astronautics (AIAA), Washington, DC.

Self Study Problems

. How is the outer limit of the boundary layer determined in experiments?
. Discuss the differences of laminar and turbulent boundary layers and their effect
on resistance.
. What characterizes the separation point?
. Summarize the assumptions of boundary layer theory.
. Explain under which conditions and why we may assume that the pressure is
constant across the boundary layer.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

118

11
Wall Shear Stress in the Boundary Layer

In this chapter we apply our knowledge of boundary layers to discover a relationship

between flow properties and the resultant wall shear stress which is at the root of fric-
tional resistance. The discussion makes use of the conservation of mass and momentum
principles, and you will see that it is very similar to the examples from Chapter .

Learning Objectives
At the end of this chapter students will be able to:

j • apply conservation principles for mass and momentum to boundary layer flow j
problems

• further understand the implications of the assumptions made in boundary layer

theory

• define boundary layer thickness, displacement thickness, and momentum thick-

ness

11.1 Control Volume Selection

As a control volume 𝑉 , we cut a thin sliver with length d𝑥 out of the boundary layer
(see Figure .). For convenience, we assume that the 𝑥-axis points along the wall and
the 𝑦-axis normal to the wall. In addition, we restrict ourselves to two-dimensional,
incompressible, and steady flow. In order to check the physical units in the equations,
we assume that the control volume is of width 𝑏 in the direction of the third, invisible
𝑧-axis. Note that the normal vector 𝑛 points out of the control volume everywhere on
𝑆 and that the control volume is fixed in space.
The boundary 𝑆 of the control volume is split into four parts. Starting on the left and
going clockwise we have:

. An inlet surface 𝑆AB , marked by the letters AB. Fluid is flowing into the control
volume through the inlet.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.2 Conservation of Mass in the Boundary Layer 119

Figure 11.1 Cross section through a finite, fixed control volume 𝑉 in the boundary layer

. The outer edge of the boundary layer 𝑆BC stretches between B and C. Note that
this boundary does not represent a streamline. There will be a mass flow across
this boundary (see Equation (.)).

j . On the right side, we have the outlet 𝑆CD marked by letters CD. j
. Finally, a piece of the wall forms the lower boundary 𝑆DA marked with letters DA.

11.2 Conservation of Mass in the Boundary Layer

Conservation of mass states that in a control volume 𝑉 the rate of change of mass must Integral form of
be equal to the resultant mass flow across the boundary 𝑆 of the control volume. We continuity
equation
express this physical principle with the continuity equation. Its integral, conservative
form was derived in Chapter .
𝜕𝜌
d𝑉 = − 𝜌 𝑣𝑇 𝑛 d𝑆 (.)
∭ 𝜕𝑡 ∬
𝑉 𝑆
⏟⏞⏞⏞⏟⏞⏞⏞⏟ ⏟⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏟
rate of change resultant mass flow
of mass in 𝑉 across boundary 𝑆

Since the density 𝜌 is constant and the control volume is fixed in space, the volume Steady flow in a
integral on the left-hand side vanishes for incompressible, steady flow. fixed control
volume
𝜕𝜌
d𝑉 = 0 (.)
∭ 𝜕𝑡
𝑉

Consequently, at any given moment, as much fluid is flowing out as is flowing into the
control volume.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

120 11 Wall Shear Stress in the Boundary Layer

Mass flux across We split the surface integral into four integrals, one for each part of the boundary. Each
boundaries .
integral represents the mass flux 𝑚𝑖𝑗 across the boundary 𝑆𝑖𝑗 .

0 =
.
𝜌 𝑣𝑇 𝑛 d𝑆 = 𝑚AB + 𝑚BC + 𝑚CD + 𝑚DA
. . . (.)
∬
𝑆

Let us discuss the individual contributions to the overall mass flux using the definitions
from Figure .:

Mass flux . The mass flux over the inlet boundary 𝑆AB is
through inlet
.
𝑚AB = 𝜌 𝑣𝑇 𝑛 d𝑆 with (.)
∬
𝑆AB
( ) ( )
𝑢(𝑥, 𝑦) −1
velocity 𝑣(𝑥, 𝑦) = and normal 𝑛AB (𝑥, 𝑦) = (.)
𝑣(𝑥, 𝑦) 0

The surface element d𝑆 is simply the product of depth 𝑏 (not visible in Figure .)
and a differential step d𝑦 in 𝑦-direction:

d𝑆 = 𝑏 d𝑦 (.)

j The double integral reduces to a single integral over the boundary layer thickness j
𝛿(𝑥). We get for the mass flux

𝛿(𝑥) ( ) 𝛿(𝑥)
. ( )𝑇 −1
𝑚 AB = 𝜌 𝑢, 𝑣 𝑏 d𝑦 = −𝑏 𝜌 𝑢(𝑥, 𝑦) d𝑦 (.)
∫ 0 ∫
0 0

Without any knowledge about the velocity distribution, we cannot proceed any
further at the moment.

Mass flux . We keep the boundary layer limit 𝑆BC for last and tackle the outlet 𝑆CD next. Its
through outlet normal vector points in the positive 𝑥-axis direction.
( )
1
𝑛CD (𝑥 + 𝑑𝑥, 𝑦) = = −𝑛AB
0

Over the short distance d𝑥, boundary layer thickness 𝛿 and horizontal velocity
𝑢 have changed in a yet unknown manner. To avoid introducing even more
unknowns, we develop the mass outflow as a Taylor series expansion of the inflow.
Since the normal vector of 𝑆CD points in the opposite direction, we take the
negative of the Taylor series. It may be truncated after its linear term because d𝑥
is small.
( . )
.
𝑚CD = − 𝑚AB +
. 𝜕 𝑚AB
d𝑥 (.)
𝜕𝑥

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.3 Conservation of Momentum in the Boundary Layer 121

.
Substituting the result for 𝑚AB from Equation (.) yields:

⎡ 𝛿(𝑥) ⎛ 𝛿(𝑥) ⎞ ⎤
.
𝑚 ⎢
= −𝑏 − 𝜌 𝑢(𝑥, 𝑦) d𝑦 −
𝜕 ⎜
𝜌 𝑢(𝑥, 𝑦) d𝑦⎟ d𝑥⎥
CD
⎢ ∫ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
⎣ 0 ⎝0 ⎠ ⎦
⎡ 𝛿(𝑥) ⎛ 𝛿(𝑥) ⎞ ⎤
⎢ 𝜕 ⎜
= 𝑏 𝜌 𝑢(𝑥, 𝑦) d𝑦 + 𝜌 𝑢(𝑥, 𝑦) d𝑦⎟ d𝑥⎥ (.)
⎢∫ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
⎣0 ⎝0 ⎠ ⎦

. Determining the mass flow across the wall 𝑆DA is easy for a change: it is a wall, No mass flux
so nothing can flow through it. through wall

.
𝑚DA = 0 (.)

. Finally, we have to address the mass flow across the boundary layer edge 𝑆BC . Mass flux
Considering the facts that nothing flows through the wall and that the mass fluxes through
boundary layer
for inlet and outlet are different, something must flow across the outer boundary limit
to balance inflow and outflow.
We employ Equation (.) to determine the respective mass flow.
. .
𝑚BC = −𝑚AB − 𝑚CD − 𝑚DA
. .
.
With 𝑚DA = 0 we have
j j
.
𝑚
. .
= −𝑚AB − 𝑚CD
BC

⎛ 𝛿(𝑥) ⎞
𝜕 ⎜
= −𝑏 𝜌 𝑢(𝑥, 𝑦) d𝑦⎟ d𝑥 (.)
𝜕𝑥 ⎜ ∫ ⎟
⎝0 ⎠
.
The mass flow 𝑚BC is the difference between inflow and outflow which makes
sense if nothing flows through the wall part of the boundary. Let us keep this
result (.) in mind, since we will need it shortly.

11.3 Conservation of Momentum in the Boundary Layer

Conservation of momentum states that the rate of change of momentum in 𝑉 is equal
to the sum of external forces acting on the control volume according to Newton’s second
law. We are interested in the force and shear stress acting parallel to the wall, i.e. in this
case the 𝑥-direction.
For our finite, fixed control volume, we apply an integral form of the conservation Integral form of
of momentum equations (see Chapter ), but use only its 𝑥-component. The 𝑦- and conservation of
momentum
𝑧-components are obtained by replacing 𝜌 𝑢 with 𝜌 𝑣 or 𝜌 𝑤 and the corresponding principle
external force component.
𝜕 ( ) ∑
𝜌𝑢 d𝑉 + 𝜌 𝑢 𝑣𝑇 𝑛 d𝑆 = (external forces in 𝑥-direction) (.)
𝜕𝑡 ∭ ∬
𝑉 𝑆

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

122 11 Wall Shear Stress in the Boundary Layer

The volume integral vanishes for steady, incompressible flow in a fixed control volume.
Body and surface forces contribute to the sum of external forces. Recalling assumption 
of boundary layer theory from page , we neglect the body forces. Only the sum of
surface forces 𝐹𝑆𝑥 remains.

( )
𝜌 𝑢 𝑣𝑇 𝑛 d𝑆 = 𝐹𝑆𝑥 (.)
∬
𝑆

11.3.1 Momentum flux over boundary of control volume

As for the conservation of mass, we split the surface integral on the left-hand side
representing the 𝑥-momentum flux through 𝑆 into four parts.

( ) . . . .
𝜌 𝑢 𝑣𝑇 𝑛 d𝑆 = 𝑀AB + 𝑀BC + 𝑀CD + 𝑀DA (.)
∬
𝑆

Momentum flux . We start with the inlet boundary 𝑆AB . Velocity vector 𝑣, normal vector 𝑛, and
through inlet surface element d𝑆 remain the same as in Equations (.) and (.). Thus, we
get for the momentum flux in 𝑥

j
𝛿(𝑥) ( ) j
. ( ) ( )𝑇 −1
𝑀AB = 𝜌 𝑢 𝑣𝑇 𝑛 d𝑆 = (𝜌 𝑢) 𝑢, 𝑣 𝑏 d𝑦
∬ ∫ 0
𝑆AB 𝑜
𝛿(𝑥)

= −𝑏 (𝜌 𝑢) 𝑢 d𝑦 (.)
∫
𝑜

Momentum flux . The strategy used for the mass transport through the outlet is also employed for
through outlet the momentum flux: we develop the flux through the outlet as a Taylor series
expansion of the flux through the inlet. Again, the change in orientation of the
normal vector from inlet to outlet triggers a change in sign for the Taylor series:
( . )
. . 𝜕 𝑀AB
𝑀CD = − 𝑀AB + d𝑥 (.)
𝜕𝑥

.
Substituting the result for 𝑀AB from Equation (.) yields:

. ⎡ 𝛿(𝑥) ⎛ 𝛿(𝑥) ⎞ ⎤
⎢ 𝜕 ⎜
𝑀CD = −𝑏 − (𝜌 𝑢) 𝑢 d𝑦 − (𝜌 𝑢) 𝑢 d𝑦⎟ d𝑥⎥
⎢ ∫ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
⎣ 0 ⎝0 ⎠ ⎦
⎡ 𝛿(𝑥) ⎛ 𝛿(𝑥) ⎞ ⎤
⎢ 𝜕 ⎜
= 𝑏 (𝜌 𝑢) 𝑢 d𝑦 + (𝜌 𝑢) 𝑢 d𝑦⎟ d𝑥⎥ (.)
⎢∫ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
⎣0 ⎝0 ⎠ ⎦

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.3 Conservation of Momentum in the Boundary Layer 123

. Like the mass transport, momentum flux through the wall 𝑆DA vanishes because No momentum
the velocity and normal vectors are
( perpendicular
) to each other and, consequently, flux through wall
their dot product vanishes, i.e. 𝑣𝑇 𝑛 = 0.
.
𝑀DA = 0 (.)
.
. Finally, the momentum flux 𝑀BC through the outer boundary 𝑆BC needs some Momentum flux
careful consideration. through
boundary layer
At the edge of the boundary layer, the velocity is equal to the exterior flow speed limit
𝑢𝑒 . In the general case of a curved body, the exterior flow velocity 𝑢𝑒 will change
over the length d𝑥.
𝜕𝑢𝑒 (𝑥)
𝑢𝑒 (𝑥 + d𝑥) = 𝑢𝑒 (𝑥) + d𝑥 (.)
𝜕𝑥
We cannot say how much specifically without any knowledge about the body
shape. However, the derivative 𝜕𝑢𝑒 ∕𝜕𝑥 in Equation (.) is considered constant if
d𝑥 is sufficiently small. That allows us to accurately estimate the mean horizontal Mean velocity
velocity 𝑢̄ 𝑒 for the boundary BC.
( )
1 𝜕𝑢 1 𝜕𝑢𝑒
𝑢̄ 𝑒 = 𝑢𝑒 + 𝑢𝑒 + 𝑒 d𝑥 = 𝑢𝑒 + d𝑥 (.)
2 𝜕𝑥 2 𝜕𝑥

The momentum (in 𝑥-direction) flowing through the boundary BC is approxi-

mately equal to the product of mass flow times the mean velocity 𝑢̄ 𝑒 .
j j
. ( )
𝑀BC = 𝜌 𝑢 𝑣𝑇 𝑛 d𝑆
∬
𝑆BC
( )
≈ 𝑢̄ 𝑒 𝜌 𝑣𝑇 𝑛 d𝑆
∬
𝑆BC

= 𝑢̄ 𝑒 𝑚BC
. (.)

Now it is time to substitute the result for the mass flux over BC (.) into the
Equation (.) for the momentum flux. Using Equation (.) to replace the
mean velocity 𝑢̄ 𝑒 yields:
( )⎡ ⎛ 𝛿(𝑥) ⎞ ⎤
. 1 𝜕𝑢𝑒 𝜕
𝑀BC = 𝑢𝑒 + d𝑥 ⎢−𝑏 ⎜ 𝜌 𝑢 d𝑦⎟ d𝑥⎥
2 𝜕𝑥 ⎢ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
⎣ ⎝0 ⎠ ⎦
⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥) ⎞  ⁓≈ 0

𝜕 ⎜ 1 𝜕𝑢 𝜕
𝜌 𝑢 d𝑦⎟ d𝑥 − 𝑏 ⎜ 𝜌 𝑢 d𝑦⎟ (d𝑥)2
𝑒

(.)

= −𝑏 𝑢𝑒 
𝜕𝑥 ⎜ ∫ ⎟ 2 
𝜕𝑥𝜕𝑥 ⎜∫ ⎟
⎝0 ⎠  ⎝0 ⎠
The last term is proportional to (d𝑥)2 and therefore of second order small. Like
before in the Taylor series, we neglect second order and higher order terms.

. ⎛ 𝛿(𝑥) ⎞
𝜕 ⎜
𝑀BC = −𝑏 𝑢𝑒 𝜌 𝑢 d𝑦⎟ d𝑥 (.)
𝜕𝑥 ⎜∫ ⎟
⎝0 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

124 11 Wall Shear Stress in the Boundary Layer

Resultant All parts of the momentum flux have been computed and are ready to be assembled
.
momentum flux into the resultant momentum flux equation. We substitute the results for 𝑀AB (.),
. . .
𝑀CD (.), 𝑀DA (.), and 𝑀BC (.) back into Equation (.).
( ) . . . .
𝜌 𝑢 𝑣𝑇 𝑛 d𝑆 = 𝑀AB + 𝑀CD + 𝑀DA + 𝑀BC
∬
𝑆
𝛿(𝑥)
⎡ 𝛿(𝑥) ⎛ 𝛿(𝑥) ⎞ ⎤
𝜕
= −𝑏 (𝜌 𝑢) 𝑢 d𝑦 + 𝑏 ⎢ (𝜌 𝑢) 𝑢 d𝑦 + ⎜ (𝜌 𝑢) 𝑢 d𝑦⎟ d𝑥⎥
∫ ⎢∫ 𝜕𝑥 ⎜ ∫ ⎟ ⎥
𝑜 ⎣0 ⎝0 ⎠ ⎦
⎛ 𝛿(𝑥) ⎞
𝜕 ⎜
+ 0 − 𝑏 𝑢𝑒 𝜌 𝑢 d𝑦⎟ d𝑥 (.)
𝜕𝑥 ⎜ ∫ ⎟
⎝0 ⎠
Obviously, the first two integrals in Equation (.) cancel each other out. We obtain
the following expression for the total 𝑥-momentum flux across the boundary 𝑆 of our
fluid element 𝑉 .

( 𝑇 ) ⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥) ⎞
𝜕 ⎜ ⎟ 𝜕 ⎜
𝜌 𝑢 𝑣 𝑛 d𝑆 = 𝑏 (𝜌 𝑢) 𝑢 d𝑦 d𝑥 − 𝑏 𝑢𝑒 𝜌 𝑢 d𝑦⎟ d𝑥 (.)
∬ 𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟
𝑆 ⎝0 ⎠ ⎝0 ⎠

j 11.3.2 Surface forces acting on control volume j

We now turn to the right-hand side of the conservation of momentum Equation (.),
the sum of external forces. As mentioned above, only surface forces 𝐹𝑆𝑥 in the 𝑥-
direction need to be considered.
Surface forces to Before we dive into the mathematics of it, let us take a look at physics, geometry, and our
consider assumptions. Surface forces result from pressure acting normal to the surface or from
shear stress acting tangential to the surface. Since the normal vectors of inlet and outlet
boundaries AB and CD are parallel to the 𝑥-axis, only pressure forces will have an effect.
Shear forces on AB and CD act perpendicular to the 𝑥-axis. On the outer boundary
BC, we have reached the exterior flow and, as per boundary layer theory assumptions,
we may neglect viscous forces there. Consequently, only a pressure force is acting on
boundary BC. On the wall DA, shear forces and pressure forces will be present. However,
the wall is parallel to the 𝑥-axis and its normal vector 𝑛DA points in 𝑦-direction. The
resulting pressure force will have no component in 𝑥-direction. Finally, remember
that, according to our conclusions from the boundary layer equations, pressure 𝑝(𝑥) is
constant across the boundary layer as long as the boundary layer is sufficiently thin.
We move on to the forces acting on the individual boundaries.

Pressure force on . The exterior flow pressure 𝑝(𝑥) acts on the inlet surface 𝑆AB . As usual, a pressure
inlet force is computed by integration of pressure over the surface. Directions of
differential forces are determined by the direction of the normal vector.

𝐹𝑆 = − 𝑝 𝑛AB d𝑆 (.)
AB ∬
𝑆AB

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.3 Conservation of Momentum in the Boundary Layer 125

For the inlet surface, this simplifies considerably as both pressure and normal
vector are constant across the inlet surface. As a consequence, the resultant
pressure force in 𝑥-direction reduces to the product of pressure and area. The
rectangular inlet area is equal to the product of boundary layer thickness 𝛿(𝑥)
(height) and depth 𝑏.
𝐹𝑆AB𝑥 = 𝑝(𝑥) 𝛿(𝑥) 𝑏 (.)

. The same is true for the outlet surface 𝑆CD . Again, pressure and normal vector Pressure force on
are constant. However, they both have changed by a differential amount. In outlet
addition, we have to consider that the pressure force will act against the positive
𝑥-direction.
( )( )
𝜕𝑝(𝑥) 𝜕𝛿(𝑥)
𝐹𝑆CD𝑥 = − 𝑝(𝑥) + d𝑥 𝛿(𝑥) + d𝑥 𝑏
𝜕𝑥 𝜕𝑥

Expanding the products results in:

𝜕𝑝(𝑥)
𝐹𝑆CD𝑥 = −𝑝(𝑥) 𝛿(𝑥) 𝑏 − 𝛿(𝑥) d𝑥 𝑏
𝜕𝑥
𝜕𝛿(𝑥) ⁓≈ 0
𝜕𝑝(𝑥) 𝜕𝛿(𝑥) 
− 𝑝(𝑥) d𝑥 𝑏 −  (d𝑥)2 𝑏
𝜕𝑥 𝜕𝑥 𝜕𝑥

The last term is proportional to (d𝑥)2 and thus of second order small. Neglecting
j the second order term, we obtain for the pressure force on the outlet: j

𝜕𝑝(𝑥) 𝜕𝛿(𝑥)
𝐹𝑆CD𝑥 = −𝑝(𝑥) 𝛿(𝑥) 𝑏 − 𝛿(𝑥) d𝑥 𝑏 − 𝑝(𝑥) d𝑥 𝑏 (.)
𝜕𝑥 𝜕𝑥

. As discussed above, the pressure force has no component in 𝑥-direction. The Friction force on
force acting on the fluid in 𝑦-direction is the reaction force which keeps the wall in wall
place. The only 𝑥-component for the exterior force results from the yet unknown
shear stress 𝜏𝑤 .
𝐹𝑆 = − 𝜏 𝑖 d𝑆 (.)
DA ∬ 𝑤
𝑆DA

The vector 𝑖 = (1, 0)𝑇 is the unit vector in positive 𝑥-direction and represents the
tangent vector to the surface 𝑆DA .
As with all other flow quantities, we develop the value at point D as a Taylor series Mean wall shear
based on the value 𝜏𝑤 (𝑥) at point A (Figure .(a)). Like the mean exterior flow stress
velocity (.), we estimate the mean shear stress 𝜏̄𝑤 (𝑥) acting over the short
distance d𝑥 to be:
( )
1 𝜕𝜏 1 𝜕𝜏𝑤
𝜏̄𝑤 = 𝜏𝑤 + 𝜏𝑤 + 𝑤 d𝑥 = 𝜏𝑤 + d𝑥 (.)
2 𝜕𝑥 2 𝜕𝑥

Note that although the force on the wall points in positive 𝑥-direction, i.e. the
fluid attempts to drag the plate with it, we have to consider the forces acting on
the fluid. The wall exerts a force on the control volume against the motion of the
fluid.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

126 11 Wall Shear Stress in the Boundary Layer

(a) Mean shear stress acting on the wall 𝑆DA (b) Mean pressure acting on the outer limit of the
boundary layer 𝑆BC

Figure 11.2 Surface forces acting on the control volume

With the mean shear stress being constant (𝜏̄𝑤 = const.) over the small area
𝑆DA = d𝑥 𝑏, the integral in Equation (.) reduces to the product of mean shear
stress and area.
𝐹𝑆DA𝑥 = −𝜏̄𝑤 d𝑥 𝑏 (.)

Pressure force on . Finally, a pressure force is acting on the outer edge of the boundary layer 𝑆BC .
j boundary layer The transverse velocity derivative vanishes, 𝜕𝑢∕𝜕𝑦 = 0 for 𝑦 = 𝛿(𝑥). Therefore, j
limit
no shear stress acts on the outer limit of the boundary layer. This is in accord
with our assumption that the exterior flow may be treated as inviscid flow.
In analogy to the mean shear stress on the wall, we compute a mean pressure 𝑝̄
acting on the surface 𝑆BC = d𝑠 𝑏 (see Figure .(b)).
( )
1 𝜕𝑝 1 𝜕𝑝
𝑝̄ = 𝑝 + 𝑝+ d𝑥 = 𝑝 + d𝑥 (.)
2 𝜕𝑥 2 𝜕𝑥

The resultant pressure force will be:

𝐹𝑆 = − 𝑝 𝑛BC d𝑆 (.)
BC ∬
𝑆BC

Assuming that pressure 𝑝 and normal vector 𝑛BC are constant with
( )
− sin(𝛼)
𝑝 = 𝑝̄ 𝑛BC =
cos(𝛼)

the vectorial pressure force is

𝐹𝑆 = −𝑝̄ 𝑛BC d𝑠 𝑏 (.)

BC

The length d𝑠 of the boundary layer edge will be slightly larger than the length
of the wall d𝑥 and the angle 𝛼 is derived from the geometry of the cross section

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.3 Conservation of Momentum in the Boundary Layer 127

through the control volume 𝑉 . The sine of 𝛼 is equal to the ratio of d𝛿 and d𝑠
d𝛿
sin(𝛼) =
d𝑠
Using this geometric relationship, the 𝑥-component of the pressure force becomes
𝐹𝑆BC𝑥 = 𝑝̄ sin(𝛼) d𝑠 𝑏
d𝛿
= 𝑝̄ d𝑠 𝑏
d𝑠
= 𝑝̄ d𝛿 𝑏 (.)
The total differential d𝛿 of the boundary layer thickness is equal to
𝜕𝛿
d𝛿 = d𝑥 (.)
𝜕𝑥
Therefore, using Equation (.) and the equation for the mean pressure (.),
the pressure force (.) becomes:
( )( )
1 𝜕𝑝 𝜕𝛿
𝐹𝑆BC𝑥 = 𝑝 + d𝑥 d𝑥 𝑏
2 𝜕𝑥 𝜕𝑥
⌃≈ 0

𝜕𝛿 1 𝜕𝑝 𝜕𝛿 2
= 𝑝 d𝑥 𝑏 +  (d𝑥) 𝑏 (.)
𝜕𝑥 2𝜕𝑥𝜕𝑥
Yet again, we will ignore the second order term proportional to the small quantity
j (d𝑥)2 . j

Summarizing the results (.), (.), (.), and (.), we compute the total Sum of external
force 𝐹𝑆𝑥 on the surface 𝑆 acting in 𝑥-direction. For convenience we omit the position forces
argument, knowing that pressure 𝑝 = 𝑝(𝑥), boundary layer thickness 𝛿 = 𝛿(𝑥), and wall
shear stress 𝜏𝑤 = 𝜏𝑤 (𝑥) are functions of 𝑥.
𝜕𝑝 𝜕𝛿  𝜕𝛿 
𝐹𝑆𝑥 =  𝑏 −
𝑝 𝛿 𝑏−𝛿
𝑝 𝛿 d𝑥 𝑏 − 𝑝 d𝑥 𝑏 + 𝑝 d𝑥 𝑏 − 𝜏𝑤 d𝑥 𝑏
𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝑝
= − 𝜏𝑤 d𝑥 𝑏 − 𝛿 d𝑥 𝑏 (.)
𝜕𝑥

Equipped with the resultant momentum flux (.) and the resultant exterior force (.), Conservation of
the conservation of momentum equation is assembled. momentum
equation
𝛿(𝑥) 𝛿(𝑥)
⎛ ⎞ ⎛ ⎞
𝜕 ⎜ 𝜕 ⎜
◁𝑏 (𝜌 𝑢) 𝑢 d𝑦⎟ 
d𝑥
 − ◁𝑏 𝑢𝑒 𝜌 𝑢(𝑥, 𝑦) d𝑦⎟ 
d𝑥
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟

⎝0 ⎠ ⎝0 ⎠
𝜕𝑝
= − 𝜏𝑤 d𝑥𝑏− 𝑏 (.)
d𝑥
𝛿(𝑥) 
𝜕𝑥
Dividing by the wall surface (d𝑥𝑏) and solving for the unknown wall shear stress 𝜏𝑤
yields:

⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥) ⎞
𝜕 ⎜ ⎟ 𝜕 ⎜ 𝜕𝑝
𝜏𝑤 (𝑥) = − (𝜌 𝑢) 𝑢 d𝑦 + 𝑢𝑒 𝜌 𝑢 d𝑦⎟ − 𝛿(𝑥) (.)
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥
⎝0 ⎠ ⎝0 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

128 11 Wall Shear Stress in the Boundary Layer

Equation (.) relates the wall shear stress to the flow properties in the control volume.
In order to compute the shear stress and the resultant resistance, we would have to
know the exterior flow velocity 𝑢𝑒 (𝑥), the boundary layer thickness 𝛿(𝑥), the velocity
distribution within the boundary layer 𝑢(𝑥, 𝑦), and the pressure distribution 𝑝(𝑥). Right
now, we do not have enough equations to resolve this.
Express pressure We can eliminate the pressure from this equation by employing the Bernoulli equa-
as function of tion (.) which is derived from the Euler equations (see later in Chapter ). The
velocity
Bernoulli equation relates pressure and flow velocity. As stated before, we consider the
flow to be steady and neglect the influence of body forces. The exterior flow is inviscid
as per boundary layer assumptions. If we add the condition that the flow is irrotational,
i.e. rot 𝑣 = 0, the exterior flow may be treated as potential flow (see Chapter ). Under
these conditions the Bernoulli equation states that the sum of pressure 𝑝 and dynamic
pressure 1∕2 𝜌 𝑢2𝑒 remains constant at all points in the flow.

1 2
𝑝+ 𝜌 𝑢 = const. (.)
2 𝑒
At point B we have pressure 𝑝 and exterior flow velocity 𝑢𝑒 . At point C both may have
changed by small amounts, which we approximate by Taylor series expansion. Thus,
setting up Bernoulli’s equation between points B and C yields:
( )2
1 2 𝜕𝑝 1 𝜕𝑢
𝑝+ 𝜌 𝑢𝑒 = 𝑝 + d𝑥 + 𝜌 𝑢𝑒 + 𝑒 d𝑥 = const. (.)
2 𝜕𝑥 2 𝜕𝑥
j j
We expand the terms and ignore the second order term at the end:
( ⌃≈ 0
)2 
1  2 𝜕𝑝 1  2 𝜕𝑢𝑒 1 𝜕𝑢𝑒 
𝑝+ 𝜌𝑢 = 
𝑝 + d𝑥 + 𝜌 𝑢 + 𝜌 𝑢𝑒 d𝑥 + 𝜌  (d𝑥)2
 2 𝑒
 𝜕𝑥 2 𝑒
 𝜕𝑥 2  𝜕𝑥

The result is an expression for the pressure gradient as a function of the exterior flow
velocity and its derivative.
𝜕𝑝 𝜕𝑢
= −𝜌 𝑢𝑒 𝑒 (.)
𝜕𝑥 𝜕𝑥
We use Equation (.) to replace the pressure term in our Equation (.) for the
shear stress:

⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥) ⎞
𝜏𝑤 (𝑥) = −
𝜕 ⎜
(𝜌 𝑢) 𝑢 d𝑦⎟ + 𝑢𝑒
𝜕 ⎜ 𝜌 𝑢 d𝑦⎟ + 𝛿(𝑥) 𝜌 𝑢 𝜕𝑢𝑒 (.)
𝑒
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥
⎝0 ⎠ ⎝0 ⎠

This is still not the final form of the equation we need.

Simplifying the Study the following expression where the right-hand side is derived by applying the prod-
equation for wall uct rule: 𝑢𝑒 (𝑥) is representing the first factor and the integral the second 𝑥-dependent
shear stress
factor.

⎛ 𝛿(𝑥)( ) ⎞ 𝜕𝑢
𝛿(𝑥)
( ) ⎛ 𝛿(𝑥) ⎞
𝜕 ⎜ 𝜕
𝑢𝑒 𝜌 𝑢 d𝑦⎟ = 𝑒
𝜌 𝑢 d𝑦 + 𝑢𝑒 ⎜ 𝜌 𝑢 d𝑦⎟ (.)
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ∫ 𝜕𝑥 ⎜ ∫ ⎟
⎝ 0 ⎠ 0 ⎝0 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.3 Conservation of Momentum in the Boundary Layer 129

The last integral also appears in Equation (.) for the shear stress. We rearrange
Equation (.) to solve for this integral:

⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥)( ) ⎞ 𝜕𝑢
𝛿(𝑥)
( )
𝜕 ⎜ 𝜕
𝑢𝑒 𝜌 𝑢 d𝑦⎟ = ⎜𝑢𝑒 𝜌 𝑢 d𝑦⎟ − 𝑒
𝜌 𝑢 d𝑦 (.)
𝜕𝑥 ⎜∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ∫
⎝0 ⎠ ⎝ 0 ⎠ 0

We substitute the right-hand side for the integral in Equation (.) which – at first
glance – seems to complicate the equation. We are almost there!

⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥)( ) ⎞
𝜕 ⎜ 𝜕
𝜏𝑤 (𝑥) = − (𝜌 𝑢)𝑢 d𝑦⎟ + ⎜𝑢 𝜌 𝑢 d𝑦⎟
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ 𝑒 ∫ ⎟
⎝0 ⎠ ⎝ 0 ⎠
𝛿(𝑥)
𝜕𝑢 ( ) 𝜕𝑢
− 𝑒 𝜌 𝑢 d𝑦 + 𝛿(𝑥) 𝜌 𝑢𝑒 𝑒 (.)
𝜕𝑥 ∫ 𝜕𝑥
0

Remember that the exterior flow velocity 𝑢𝑒 (𝑥) is a function of 𝑥 only. Consequently,
we can move the factor 𝑢𝑒 in front of the second integral underneath the integral sign.
From the last two terms, we factor out the derivative 𝜕𝑢𝑒 ∕𝜕𝑥:

⎛ 𝛿(𝑥) ⎞ ⎛ 𝛿(𝑥)( ) ⎞
j 𝜕 ⎜ 𝜕 j
𝜏𝑤 (𝑥) = − (𝜌 𝑢) 𝑢 d𝑦⎟ + ⎜ 𝜌 𝑢 𝑢𝑒 d𝑦⎟
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎜ ∫ ⎟
⎝0 ⎠ ⎝0 ⎠
𝛿(𝑥)
𝜕𝑢𝑒 ⎡⎢ ( ) ⎤
+ − 𝜌 𝑢 d𝑦 + 𝛿(𝑥) 𝜌 𝑢𝑒 ⎥ (.)
𝜕𝑥 ⎢ ∫ ⎥
⎣ 0 ⎦
The first two integrals can be combined because a sum of derivatives is equal to the
derivative of a sum.

⎛ 𝛿(𝑥) ( ) ⎞ ⎡ 𝛿(𝑥)( ) ⎤
𝜕 ⎜ ( ) 𝜕𝑢
𝜏𝑤 (𝑥) = 𝜌 𝑢 𝑢𝑒 − 𝑢 d𝑦⎟ + 𝑒 ⎢− 𝜌 𝑢 d𝑦 + 𝛿(𝑥) 𝜌 𝑢𝑒 ⎥ (.)
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ⎢ ∫ ⎥
⎝0 ⎠ ⎣ 0 ⎦

In order to combine the terms in square brackets, we rewrite the last term:
𝛿(𝑥) 𝛿(𝑥)

𝛿(𝑥) 𝜌 𝑢𝑒 = 1 d𝑦 𝜌 𝑢𝑒 = 𝜌 𝑢𝑒 d𝑦 (.)
∫ ∫
0 0

Again, we utilized that density 𝜌 and exterior flow velocity 𝑢𝑒 are not functions of 𝑦
within the boundary layer.

⎛ 𝛿(𝑥) ( ) ⎞ 𝜕𝑢
𝛿(𝑥)
( )
𝜕 ⎜
𝜏𝑤 (𝑥) = 𝜌 𝑢 𝑢𝑒 − 𝑢 d𝑦⎟ + 𝑒
𝜌 𝑢𝑒 − 𝑢 d𝑦 (.)
𝜕𝑥 ⎜ ∫ ⎟ 𝜕𝑥 ∫
⎝0 ⎠ 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

130 11 Wall Shear Stress in the Boundary Layer

(a) Definition of 𝛿1 (b) Displacement effect

Figure 11.3 Definition of displacement thickness 𝛿1 and displacement effect on exterior flow

Finally, we normalize the velocities in the integrands by multiplying them with 𝑢2𝑒 ∕𝑢2𝑒
and 𝑢𝑒 ∕𝑢𝑒 respectively and extract from the integrals all factors which are independent
of the integration variable 𝑦.
j j
𝛿(𝑥) ( ) ⎤ 𝛿(𝑥)( )
⎡ 𝜕𝑢 ( )
𝜕 ⎢ 2 𝑢 𝑢 𝑢
𝜏𝑤 (𝑥) = 𝜌 𝑢𝑒 1− d𝑦⎥ + 𝑒
𝜌 𝑢𝑒 1− d𝑦 (.)
𝜕𝑥 ⎢ ∫ 𝑢𝑒 𝑢𝑒 ⎥ 𝜕𝑥 ∫ 𝑢𝑒
⎣ 0 ⎦ 0

The remaining integrals in Equation (.) have special meaning in boundary layer
theory and will be explained in more detail in the following two subsections.

11.3.3 Displacement thickness

The last integral in Equation (.) is the displacement thickness.

𝛿(𝑥)( )
𝑢
𝛿1 = 1− d𝑦 (.)
∫ 𝑢𝑒
0

The displacement thickness 𝛿1 can be interpreted as a ‘thickening’ of the body. Mathe-

matically, the integral in Equation . is equal to the gray shaded area in Figure .(a).
It is a measure of how much less viscous fluid flows through the boundary layer com-
pared with an inviscid fluid. The displacement thickness can be used to include the
displacement effects of the boundary layer in inviscid flow calculations. The body is
enlarged by 𝛿1 which usually improves the prediction of wave patterns and streamlines
(see Figure .(b)).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

11.4 Wall Shear Stress 131

11.3.4 Momentum thickness

The first integral in Equation (.) is the momentum thickness.

𝛿(𝑥) ( )
𝑢 𝑢
Θ = 1− d𝑦 (.)
∫ 𝑢𝑒 𝑢𝑒
0
Momentum thickness Θ is a measure of how much momentum is lost in the boundary
layer. As we will see, Θ plays an important role in the computation of the wall shear
stress and the frictional resistance.

11.4 Wall Shear Stress

Introducing displacement thickness 𝛿1 (.) and momentum thickness Θ (.) into
Equation (.) for the wall shear stress yields:
( ) 𝜕𝑢
𝜕
𝜏𝑤 (𝑥) = 𝜌 𝑢2𝑒 Θ + 𝜌 𝑢𝑒 𝑒 𝛿1 (.)
𝜕𝑥 𝜕𝑥
This is a general result of boundary layer theory. The partial differential equation (.)
for the shear stress involves three unknown functions and their derivatives:

. exterior flow velocity 𝑢𝑒 ,

j j
. displacement thickness 𝛿1 , and
. momentum thickness Θ.

Displacement and momentum thickness, in turn, depend on the boundary layer thick-
ness 𝛿, which is used as the upper limit for the defining integrals. We will discuss an
approximate solution of the shear stress equation (.) in the next chapter. It will get
us a step closer to an answer on how to determine the frictional resistance of a ship.

Self Study Problems

. Annotate the text in this chapter by indicating which assumption of boundary
layer theory is used in each step.
. Why is the outer limit of the boundary layer not a streamline?
. Consider the following velocity profile for a laminar boundary layer:
(𝜋 𝑦)
𝑢
= sin
𝑢𝑒 2𝛿
Compute displacement and momentum thickness (as functions of boundary layer
thickness 𝛿).
. How does Equation (.) change when a flow with vanishing pressure gradient
𝜕𝑝∕𝜕𝑥 = 0 is considered? Hint: Go back to Equation (.) and perform the steps
used in simplifying the equation for shear stress.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

132

12
Boundary Layer of a Flat Plate

In previous chapters we derived the simplified conservation of mass and momentum

equations which describe the flow in a thin boundary layer. Here, these equations are
solved for the special case of a flat plate. The approximate solution provides estimates
for boundary layer thickness, displacement thickness, momentum thickness, and the
important wall shear stress. We derive flat plate frictional coefficients for laminar flow.
A solution for turbulent flow will be presented in the following chapter.

Learning Objectives

j At the end of this chapter students will be able to: j

• solve boundary layer flow problems
• interpret results from boundary layer theory
• understand the origin of flat plate friction coefficients

12.1 Boundary Layer Equations for a Flat Plate

Flat plate Consider the flow over a flat plate of length 𝐿 and beam 𝑏 (Figure .), with the latter
measured in the invisible 𝑧-direction. The onflow is uniform in positive 𝑥-direction
with constant velocity 𝑢0 (parallel flow). As usual, thickness of the boundary layer is
greatly exaggerated in Figure ..
Constant exterior A flat plate does not force the fluid to change course. Therefore, the exterior flow speed
velocity 𝑢𝑒 = 𝑢0 will be maintained everywhere outside the boundary layer.
𝜕𝑢𝑒
𝑢𝑒 = 𝑢0 = const. or = 0 for 𝑦 > 𝛿(𝑥) (.)
𝜕𝑥
This assumption is not completely true. In the previous chapter we learned that the
boundary layer has a displacement effect. This will also occur on the plate and, conse-
quently, the exterior flow experiences a ‘body’ with finite thickness which will increase
exterior flow speed. However, since the boundary layer is thin and much smaller than
the length of the plate, this increase in exterior velocity will be very small and the
assumption of 𝑢0 = const. is justified.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.1 Boundary Layer Equations for a Flat Plate 133

Figure 12.1 Laminar boundary layer along a flat plate

With the vanishing 𝑥-derivative of the external flow velocity, the pressure derivative Constant
must vanish as a consequence of the Bernoulli equation (.). pressure along
plate

𝜕𝑝 𝜕𝑢
= −𝜌 𝑢𝑒 𝑒 = 0 for a flat plate
𝜕𝑥 𝜕𝑥

Thus, for a flat plate, pressure is constant along and across the boundary layer (𝜕𝑝∕𝜕𝑥 =
0 ∧ 𝜕𝑝∕𝜕𝑦 = 0), the latter being one of the conclusions from the boundary layer
equations (.).
j Utilizing this simplification enables us to eliminate the last term in Equation (.) for Wall shear stress j
the wall shear stress. With 𝑢𝑒 (𝑥) = 𝑢0 = const. and 𝜕𝑢0 ∕𝜕𝑥 = 0 we get: equation

=0
( ) 𝜕𝑢 
𝜕
𝜏𝑤 (𝑥) = 𝜌 𝑢20 Θ + 𝜌 𝑢0 0 𝛿1
𝜕𝑥 𝜕𝑥

Extracting the constant density 𝜌 and the squared flow velocity 𝑢0 from the first term on
the right-hand side yields a relationship between wall shear stress 𝜏𝑤 and the momentum
thickness Θ.

𝜕Θ
𝜏𝑤 (𝑥) = 𝜌 𝑢20 (.)
𝜕𝑥

We replace Θ with its defining integral.

𝛿(𝑥) ( )
𝜕 𝑢 𝑢
𝜏𝑤 (𝑥) = 𝜌 𝑢20 1− d𝑦 (.)
𝜕𝑥 ∫ 𝑢0 𝑢0
0

The solution of Equation (.) is defined by a dimensionless velocity distribution 𝑢∕𝑢0

over the thickness of the boundary layer. In fact, we may define the velocity distribution
first and employ it to find an expression for the unknown boundary layer thickness
𝛿, which marks the upper limit of the integral. Suitable velocity distributions may be
found from measurements.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

134 12 Boundary Layer of a Flat Plate

12.2 Dimensionless Velocity Profiles

Shape of velocity Based on his observations, Prandtl suggested that the basic shape of the velocity distri-
profile assumed bution remains approximately the same for increasing 𝑥-position along the boundary
to stay similar
layer. It is simply stretched in 𝑦-direction as indicated in Figure .. If a simple equation
for the velocity distribution can be derived from measurements, Equation (.) can be
solved for the wall shear stress. An example will demonstrate this.
Parabolic Let us assume the velocity distribution is reasonably well described by a parabola. It is
velocity profile one of several usable shapes.
𝑢
= 𝑎̄ + 𝑏̄ 𝑦 + 𝑐̄ 𝑦2 (.)
𝑢0
Please note that simple formulas like Equation (.) apply only to boundary layers with
laminar flow. The velocity distribution in turbulent boundary layers is more complicated
(see next chapter).
Next, proper values need to be found for the coefficients 𝑎, ̄ and 𝑐̄ in Equation (.).
̄ 𝑏,
Dimensionless To simplify this task, we introduce the dimensionless 𝑦-coordinate 𝜂.
coordinates
𝑦
𝜂 = or 𝑦 = 𝜂 𝛿(𝑥) (.)
𝛿(𝑥)
𝜂 grows from zero at the wall to the value one at the edge of the boundary layer (Fig-
ure .). We replace the 𝑦-coordinate in Equation (.) with 𝜂.
j 𝑢 j
= 𝑎̄ + 𝑏̄ 𝑦 + 𝑐̄ 𝑦2
𝑢0
= 𝑎 + 𝑏̄ 𝜂 𝛿 + 𝑐̄ 𝜂 2 𝛿 2
= 𝑎 + 𝑏 𝜂 + 𝑐 𝜂2 (.)
The new coefficients 𝑎, 𝑏, and 𝑐 are
𝑎 = 𝑎̄ 𝑏 = 𝑏̄ 𝛿 𝑐 = 𝑐̄ 𝛿 2

Matching profile With the coefficients 𝑎, 𝑏, and 𝑐, the assumed velocity profile may be adjusted to the
to boundary specific boundary conditions of the flow problem at hand. Three conditions are required
conditions
to determine the coefficients:

(i) Based on the basic boundary layer assumptions (see page ), fluid sticks to the
plate surface.
(ii) Also, the flow velocity 𝑢 increases from zero to the external flow speed across the
boundary layer.
(iii) At the boundary layer edge, the transverse derivative of 𝑢 vanishes, i.e. the velocity
remains constant outside of the boundary layer.

Mathematically, this results in the set of equations below. The no-slip boundary condi-
tion at the wall requires:
𝑢
for 𝑦 = 0 ∶ 𝑢 = 0 or for 𝜂 = 0 ∶ = 0 (.)
𝑢0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.3 Boundary Layer Thickness 135

At the outer edge of the boundary layer we impose:

𝑢
for 𝑦 = 𝛿 ∶ 𝑢 = 𝑢0 or for 𝜂 = 1 ∶ = 1 (.)
𝑢0
( )
𝜕𝑢 𝜕 𝑢
also for 𝑦 = 𝛿 ∶ = 0 or for 𝜂 = 1 ∶ = 0 (.)
𝜕𝑦 𝜕𝜂 𝑢0

From Equation (.) and the first boundary condition (.) follows the value for Coefficient 𝑎
coefficient 𝑎:
𝑢 ||
from = 0 follows 𝑎 + 𝑏 ⋅ 0 + 𝑐 ⋅ 02 = 0 and 𝑎 = 0 (.)
𝑢0 ||𝜂=0

Obviously, the 𝑦-axis part of the velocity distribution must vanish or otherwise the
no-slip condition would be violated.
The second boundary condition Equation (.), provides a relationship between coeffi- Coefficients 𝑏
cients 𝑏 and 𝑐. and 𝑐

𝑢 ||
from = 1 follows 𝑎
◁ + 𝑏 ⋅ 1 + 𝑐 ⋅ 1 = 1 and 𝑏 + 𝑐 = 1
2
(.)
𝑢0 ||𝜂=1

Of course, we exploit that the coefficient 𝑎 is equal to zero.

From the third boundary condition, Equation (.), we derive a second expression
j connecting 𝑏 and 𝑐. The derivative of the velocity profile with respect to 𝜂 is j
( ) ( )
𝜕 𝑢 𝜕
= 𝑎 + 𝑏𝜂 + 𝑐𝜂 2 = 𝑏 + 2𝑐 𝜂 (.)
𝜕𝜂 𝑢0 𝜕𝜂

Equating it to zero at the edge of the boundary layer yields:

( )
𝜕 𝑢 ||
| = 0 or 𝑏 + 2𝑐 ⋅ 1 = 0 (.)
𝜕𝜂 𝑢0 ||𝜂=1

We find that
𝑏 = −2𝑐 (.)
Substituting the result (.) into condition (.) results in

𝑐 = −1 and 𝑏 = 2

Thus, the dimensionless velocity profile within the boundary layer is defined as Dimensionless
velocity profile
𝑢
= 2𝜂 − 𝜂 2 (.)
𝑢0

It is important to note that although Equation (.) satisfies the boundary conditions
of boundary layer theory, we only assume that it represents a realistic approximation
of the actual velocity distribution. Other useful approximations are left as self study
problems. Having an expression for the velocity distribution across the boundary layer
enables us to compute the boundary layer thickness.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

136 12 Boundary Layer of a Flat Plate

12.3 Boundary Layer Thickness

Momentum First, we apply the velocity profile to the definition of the momentum thickness (.).
thickness
integral 𝛿(𝑥) ( )
𝑢 𝑢
Θ = 1− d𝑦 (.)
∫ 𝑢0 𝑢0
0

The integration variable and limits of integration are replaced with their dimensionless
counterparts.

𝑦 = 𝜂𝛿 d𝑦 = 𝛿 d𝜂 𝑦=0 ⟶ 𝜂=0 𝑦=𝛿 ⟶ 𝜂=1

Considering that the boundary layer thickness 𝛿 can be extracted from the integral
because it is only a function of 𝑥 and not of 𝑦, we obtain:

𝛿(𝑥) ( ) 1 ( )
𝑢 𝑢 𝑢 𝑢
1− d𝑦 = 𝛿 1− d𝜂 (.)
∫ 𝑢0 𝑢0 ∫ 𝑢0 𝑢0
0 0

Next, the assumed velocity profile 2𝜂 − 𝜂 2 is substituted for 𝑢∕𝑢0 .

j 1 1 j
( )(
2
) ( )
Θ(𝑥) = 𝛿 2𝜂 − 𝜂 1 − 2𝜂 + 𝜂 2 d𝜂 = 𝛿 2𝜂 − 5𝜂 2 + 4𝜂 3 − 𝜂 4 d𝜂
∫ ∫
0 0
[ ]1
5 1 2
= 𝛿 𝜂2 − 𝜂3 + 𝜂4 − 𝜂5 = 𝛿(𝑥) (.)
3 5 0 15

For our assumed velocity profile 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 , the momentum thickness is / of
the geometric boundary layer thickness. Note that both boundary layer thickness 𝛿(𝑥)
and momentum thickness Θ(𝑥) are functions of position 𝑥 along the plate. Only the
ratio Θ(𝑥)∕𝛿(𝑥) is constant, which is a consequence of assuming that the shape of the
velocity profile 𝑢∕𝑢0 is independent of 𝑥.
Wall shear stress We dump the result for the momentum thickness (.) into the equation for the wall
(1st equation) shear stress (.)
( )
𝜕 2 2 𝜕𝛿
𝜏𝑤 = 𝜌 𝑢20 𝛿 = 𝜌 𝑢2 for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝜕𝑥 15 15 0 𝜕𝑥
This equation has two unknowns: the boundary layer thickness 𝛿(𝑥) and the wall shear
stress 𝜏𝑤 (𝑥). Obviously, we need to consult a second equation to solve this.
Netwon’s shear The observant reader may remember that we came across a similar problem when we
stress derived the Navier-Stokes equations. There the unknown stresses in the fluid were
(2nd equation)
related to the spatial rate of change of velocities. In that context, we stated the simplest
form of this relationship as it was derived by Newton (see Equation (.) on page ).
Newton derived this relationship for simple shear flow between two plates. In our case,
the second plate is very far away but the flow conditions are similar (D over a flat

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.3 Boundary Layer Thickness 137

surface). If we employ Newton’s relationship for shear stress and use the derivative of
the velocity at the wall (𝑦 = 0), we get

𝜕𝑢 ||
𝜏𝑤 = 𝜇 (.)
𝜕𝑦 ||𝑦=0

Again, we introduce the dimensionless velocity profile 𝑢∕𝑢0 and the coordinate 𝜂 = 𝑦∕𝛿
by multiplying the numerator with 1 = 𝑢0 ∕𝑢0 and the denominator with 1 = 𝛿∕𝛿,
respectively.
( ) ( )
| 𝑢0 𝑢 | [ ( )]
𝜕𝑢| 𝜕 |
| 𝑢0 𝑢0 𝑢0 | 𝑢0 𝜕 𝑢
𝜏𝑤 = 𝜇 ( )| = 𝜇 ( )| = 𝜇 (.)
𝛿 | 𝛿 𝑦 | 𝛿 𝜕𝜂 𝑢
𝜕𝑦 𝛿 | 𝜕 𝛿 | 0 𝜂=0
|𝑦=0 |𝑦=0

Extracting external flow speed 𝑢0 and boundary layer thickness 𝛿 from the differential
operators is possible because they do not depend on the variables 𝑦 or 𝜂.
The additional, required equation for wall shear stress and boundary layer thickness
is obtained by introducing the derivative of the assumed velocity profile (.) into
Equation (.).
𝑢0 [ ] 𝑢
𝜏𝑤 = 𝜇 2 − 2𝜂 = 2𝜇 0 for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝛿 𝜂=0 𝛿
j j
Next, we equate the right-hand sides of Equations (.) and (.) for the wall shear ODE for
stress. boundary layer
thickness
2 𝜕𝛿 𝑢
𝜌 𝑢20 = 2𝜇 0
15 𝜕𝑥 𝛿

Division by 2∕15 and 𝜌𝑢20 and reintroducing the kinematic viscosity 𝜈 = 𝜇∕𝜌 results in

𝜕𝛿 15𝜈 1
− = 0 (.)
𝜕𝑥 𝑢0 𝛿

Finally, we have derived an equation which only bears the boundary layer thickness as
an unknown quantity.
In fact, Equation (.) is an ordinary differential equation (ODE) for the boundary
layer thickness. 𝛿 is only a function of the 𝑥-coordinate. The equation is of type
𝛿 ′ − 𝛼𝛿 −1 = 0. This means we are looking for a function whose derivative is equal to the
inverse of the function itself. You know a function with this property! It is the square
root function. We state the boundary layer thickness as a generic square root function:
√ 𝜕𝛿 𝐶 1
𝛿 = 𝐶1 𝑥 + 𝐶2 and its derivative: = 1 𝑥− 2 (.)
𝜕𝑥 2
The constant 𝐶2 = 0 in Equation (.) must vanish because the boundary layer Integration
thickness is zero at the leading edge of the plate 𝛿(𝑥 = 0) = 0, as stated in one of our constants
basic assumptions of boundary layer theory.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

138 12 Boundary Layer of a Flat Plate

Substituting the function and its derivative from Equation (.) into the ordinary
differential equation (.) yields:
𝐶1 − 1 15𝜈 1
𝑥 2 − 1
= 0
2 𝑢0
𝐶1 𝑥 2

or
( )
1 𝐶1 15𝜈 1
𝑥− 2 − = 0
2 𝑢0 𝐶1

This must hold true for all 𝑥, therefore the term in parentheses must vanish:

30𝜈 1
𝐶1 − = 0
𝑢0 𝐶1
The last equation yields for the constant 𝐶1 :
√
30𝜈
𝐶1 = for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝑢0

Laminar With the constant 𝐶1 our expression is complete for the boundary layer thickness of a
boundary layer flat plate. √
j thickness j
approximation 30𝜈 √
𝛿(𝑥) = 𝑥 (.)
𝑢0
The boundary layer grows with the square root of the position 𝑥 along the plate. If you
study sketches and pictures of boundary layers, you will recognize the basic shape of
the square root function.
Finally, we introduce the local Reynolds number 𝑅𝑒𝑥 with
𝑢0 𝑥
𝑅𝑒𝑥 = (.)
𝜈
Note that 𝑅𝑒𝑥 differs from our usual definition of the Reynolds number. 𝑅𝑒𝑥 uses a
variable position 𝑥 as reference length. As a consequence, the local Reynolds number
changes from 𝑅𝑒𝑥 = 0 at the beginning of the plate (𝑥 = 0) to 𝑅𝑒𝑥 = 𝑢0 𝐿∕𝜈 at the
end of the plate (𝑥 = 𝐿). Introducing 𝑅𝑒𝑥 into Equation (.) yields the following
equation for the boundary layer thickness based on the parabolic velocity profile:
√ 𝑥 𝑥
𝛿(𝑥) = 30 √ ≈ 5.48 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝑅𝑒𝑥 𝑅𝑒𝑥

As expected, the boundary layer thickness is proportional to the inverse square root of
the local Reynolds number as we discussed in Section ..
It is important to remember that we achieved this result under the assumption that the
velocity profile in the boundary layer is well represented by the parabolic function of
Equation (.): 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 . It turns out that this is a reasonable, but inaccurate,
assumption for laminar boundary layers. Boundary layers on smooth, flat plates without

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.3 Boundary Layer Thickness 139

excessive disturbances in the onflow are laminar up to a position 𝑥 for which 𝑅𝑒𝑥 =
200 000 (Fox and McDonald, ).
Blasius used a mathematically more rigorous approach. He first converted the boundary Blasius’ laminar
layer Equations (.)–(.) into an ODE of third order by introducing a stream boundary layer
thickness
function. Then he solved the third order, nonlinear ODE for the stream function by a solution
series expansion which allows control of the desired accuracy (Blasius, ).
Blasius’ result for the boundary layer thickness for laminar flow over a flat plate is
(Schlichting and Gersten, )
𝑥
𝛿(𝑥) ≈ 5.0 √ Blasius’ result (.)
𝑅𝑒𝑥

According to Blasius’ more accurate result, our approximation (.) overpredicts the
thickness by about %.
For a flat plate of length 𝐿 = 0.2 m in a parallel flow of velocity 𝑢0 = 1.0 m/s in fresh Boundary layer
water at  ◦ C, we can expect the following boundary layer thickness: thickness
example

. The kinematic viscosity of fresh water ( ◦ C) is 𝜈 = 1.1386⋅10−6 m2 ∕s, according

to the ITTC water properties tables (ITTC, ).
. The resulting local Reynolds number at the end of the plate will be
𝑢0 𝑥 || 1.0 m/s ⋅ 0.2 m
j 𝑅𝑒𝑥 = = = 175654.3 < 200000 j
𝜈 ||𝑥=𝐿 1.1386 10−6 m2 ∕s
The result is indeed dimensionless as it should be and the value of 𝑅𝑒𝑥 = 175654.3
indicates that the assumption of laminar flow is acceptable.
. Finally, we compute the maximum boundary layer thickness for the assumed
parabolic velocity profile (.):
|
𝑥 || 0.2 m
𝛿(𝑥 = 𝐿) = 5.48 √ | = 5.48 √ = 0.0026 m
|
𝑅𝑒𝑥 |𝑥=0.2 m 175654.3

With . mm the boundary layer thickness has reached .% of the plate length.
Although this is fairly thin, we must expect that any predictions based on the
boundary layer equations will deviate more and more from the actual flow as the
boundary layer becomes thicker.

Luckily, boundary layers of ships are thin as well. Reynolds numbers for ships are in Laminar
the range of 109 which results√in low 𝛿∕𝐿 ratios as the boundary layer thickness is
boundary layer
results do not
proportional to the inverse of 𝑅𝑒𝑥 . However, ship boundary layers are turbulent and apply to ships
the formulas presented in this chapter do not apply. Thickness of turbulent boundary
layers grows faster along the plate than the thickness of laminar boundary layers. We
will discuss more relevant solutions for turbulent boundary layers in Chapter .
Having finally found an expression for the thickness of a laminar boundary layer, we
can compute the remaining boundary layer characteristics and compare them with
Blasius’ more accurate results.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

140 12 Boundary Layer of a Flat Plate

12.4 Wall Shear Stress

Shear stress We have two equations to solve for the wall shear stress: Equations (.) and (.).
approximation Of course, both yield the same result. For convenience, we use Equation (.) and
substitute the result for the boundary layer thickness (.) into it:
𝑢 𝑢
𝜏𝑤 = 2𝜇 0 = 2𝜇 √ 0 for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝛿 30 √ 𝑥
𝑅𝑒𝑥

Rearranging the last term and expanding it by the factor 1 = 𝜌𝑢0 ∕𝜌𝑢0 yields:
√ √ ( )
𝑅𝑒𝑥 𝜇 𝑢0 𝜌 𝑢0 𝑅𝑒𝑥 2 𝜇
𝜏𝑤 = 2 √ ⋅ = 2 √ 𝜌 𝑢0 (.)
30 𝑥 𝜌 𝑢0 30 𝜌 𝑢0 𝑥

The last factor in parentheses is the inverse of the local Reynolds number 𝑅𝑒𝑥 .
2 1
𝜏𝑤 = √ 𝜌 𝑢20 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
30 𝑅𝑒𝑥

Skin friction In order to compare this result with measurements and other approximation, it is
coefficient common practice to present the wall shear stress as a dimensionless skin friction
approximation
coefficient 𝐶𝑓𝑥 .
𝜏𝑤
𝐶𝑓𝑥 (𝑥) = (.)
1
j 2
𝜌 𝑢20 j
For the case of the quadratic velocity profile with wall shear stress from Equation (.),
the local skin friction coefficient is equal to
4 1 1
𝐶𝑓𝑥 (𝑥) = √ √ ≈ 0.730 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
30 𝑅𝑒𝑥 𝑅𝑒𝑥
Blasius’ more With Blasius’ solution for the boundary layer thickness, a local skin friction coefficient
accurate solution of
1
𝐶𝑓𝑥 (𝑥) = 0.665 √ Blasius’ result (.)
𝑅𝑒𝑥
is obtained.
Figure . compares our approximation with Blasius’ result. It is interesting to note
that both solutions predict an infinitely high shear stress at the beginning of the plate
at 𝑥 = 0. This is a singularity of the mathematical boundary layer theory model rather
than a real phenomenon. Studying the ODE (.), we see that we cannot get a valid
solution for 𝛿(𝑥 = 0) = 0 because of the 1∕𝛿-term.
We noted in Figure . that the velocity profiles are geometrically similar along the
plate but stretched in 𝑦-direction. The stretching results in a visible reduction of the
transverse derivative of the velocity
𝜕𝑢 || 𝜕𝑢 || 𝜕𝑢 ||
| > | > for 𝑥1 > 𝑥2 > 𝑥3 (.)
𝜕𝑦 |𝑥=𝑥1 𝜕𝑦 |𝑥=𝑥2 𝜕𝑦 ||𝑥=𝑥3
This decline in the slope of the velocity profile is also reflected in a decreasing shear
stress along the plate.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.5 Displacement Thickness 141

Figure 12.2 Boundary layer shear stress for laminar flow over a flat plate as dimensionless
position dependent skin friction coefficient 𝐶𝑓𝑥

j j
12.5 Displacement Thickness

The displacement thickness is defined as the integral Displacement

thickness
approximation
𝛿(𝑥)( )
𝑢
𝛿1 = 1− d𝑦 (.)
∫ 𝑢0
0

Its computation is straight forward for our assumed velocity profile from Equation (.)
𝑢∕𝑢0 = 2𝜂 − 𝜂 2 :

1
( ) [ ]1
1 1
𝛿1 = 𝛿 1 − 2𝜂 + 𝜂 2 d𝜂 = 𝛿 𝜂 − 𝜂 2 + 𝜂 3 = 𝛿
∫ 3 0 3
0
√
30 𝑥 𝑥
= √ ≈ 1.826 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
3 𝑅𝑒𝑥 𝑅𝑒𝑥

Based on Blasius’ results for a laminar boundary layer, the displacement thickness Blasius’ more
should be accurate result

𝑥
𝛿1 (𝑥) = 1.72 √ Blasius’ result (.)
𝑅𝑒𝑥

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

142 12 Boundary Layer of a Flat Plate

0.014
/L (Blasius)
1 /L (Blasius)
[]

0.012 /L (Blasius)
2
/L for u/u0 =2
/L

2
0.010 1 /L for u/u0 =2
2
/L for u/u0 =2
1 /L,

0.008
dimensionless thickness /L,

u0 =1.000 m/s
0.006
L =0.200 m

0.004

0.002

0.000
0.0 0.2 0.4 0.6 0.8 1.0
position along plate x/L []

Figure 12.3 Boundary layer thickness 𝛿, displacement thickness 𝛿1 , and momentum thickness
Θ for laminar flow over a flat plate

j j
12.6 Momentum Thickness
Momentum The integral for the momentum thickness of a flat plate boundary layer has been solved
thickness as part of our quest to find the boundary layer thickness (see page ).
approximation
√
2 2 30 𝑥 𝑥
Θ = 𝛿 = √ ≈ 0.730 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
15 15 𝑅𝑒𝑥 𝑅𝑒𝑥

Blasius’ more Blasius’ solution for the momentum thickness of a laminar boundary layer is
accurate result
𝑥
Θ(𝑥) = 0.665 √ Blasius’ result (.)
𝑅𝑒𝑥

Figure . compares the results for boundary layer thickness, displacement thickness,
and momentum thickness of Blasius’ with our approximation based on the velocity
profile 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 . The approximate solution somewhat overpredicts the more
accurate values of Blasius. In the self study problems readers can explore other ap-
proximations. In any case, the following relationship holds true between the three
characteristic boundary layer thicknesses:

Θ < 𝛿1 < 𝛿 (.)

With Blasius’ results for the laminar boundary layer, the following ratios are found:
𝛿1 1.72 Θ 0.665
= ≈ 0.344 = ≈ 0.133 (.)
𝛿 5.0 𝛿 5.0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.7 Friction Force and Coefficients 143

12.7 Friction Force and Coefficients

Integration of the shear stress distribution (.) across the plate results in the total Friction force
friction force 𝐹𝐹𝑥 on one side of the plate.

𝐿 𝐿 𝐿
𝜕Θ
𝐹𝐹𝑥 = 𝜏𝑤 d𝑆 = 𝑏 𝜏𝑤 d𝑥 = 𝜌 𝑢20 𝑏 d𝑥 = 𝜌 𝑢20 𝑏 dΘ
∬ ∫ ∫ 𝜕𝑥 ∫
0 0 0
[ ]𝐿
= 𝜌 𝑢20 𝑏 Θ = 𝜌 𝑢20 𝑏 Θ𝐿 (.)
0

Θ𝐿 represents the value of the momentum thickness at the end of the plate at 𝑥 = 𝐿.
As expected, the total friction force is equal to the momentum lost over the length of
the plate.
For the √
parabolic velocity profile (.), momentum thickness (.) is Θ(𝑥) =
0.730𝑥∕ 𝑅𝑒𝑥 . The momentum thickness at the end of the plate is

𝐿
Θ(𝑥 = 𝐿) = 0.730 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝑅𝑒𝐿

Substituting this result into the formula for the friction force and using the standard
Reynolds number 𝑅𝑒 = 𝑅𝑒𝐿 = 𝑢0 𝐿∕𝜈 yields:
j j
1
𝐹𝐹𝑥 = 0.730 𝜌 𝑢20 𝑏 𝐿 √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
𝑅𝑒
The product of beam 𝑏 and plate length 𝐿 represents the wetted surface of the plate.
Based on the data from the example on page  (𝐿 = 0.2 m, 𝑢0 = 1.1 m/s, 𝜈 =
1.1386 ⋅ 10−6 m2 /s, and density 𝜌 = 999.1026 kg/m3 ), the friction force per unit plate
width will be
𝐹𝐹𝑥 ( )2 1
= 0.730 ⋅ 999.1026 kg/m3 1.1 m/s 0.2 m √
𝑏 193168.8
= 0.402 N/m

This is a fairly small force and hints at how difficult it is to conduct measurements which
are accurate enough to validate assumptions and theories.
Analog to the coefficients introduced for the components of ship resistance, we intro- Laminar flow
duce a dimensionless friction coefficient for the flat plate in laminar flow. friction
coefficient
𝐹𝐹𝑥 1.46
𝐶𝐹 = = √ for 𝑢∕𝑢0 = 2𝜂 − 𝜂 2 (.)
1 2 𝑅𝑒
𝜌𝑢 𝑏𝐿
2 0

With Blasius’ result for the momentum thickness the friction coefficient becomes
1.33
𝐶𝐹 = √ Blasius’ result, valid for 𝑅𝑒 < 200 000 (.)
𝑅𝑒

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

144 12 Boundary Layer of a Flat Plate

Unlike the local skin friction coefficient 𝐶𝑓𝑥 , flat plate friction coefficients are indepen-
dent of the position. 𝐶𝐹 is valid for plates as a whole.
Unfortunately, the friction coefficient for laminar boundary layers (.) is not appli-
cable to ships. Reynolds numbers for ships are considerably higher than the upper limit
for laminar flow. Appropriate friction coefficients for turbulent flows will be discussed
in the following chapter.

References
Blasius, H. (). Grenzschichten in Flüssigkeiten mit kleiner Reibung. PhD thesis,
Universität Göttingen.
Fox, R. and McDonald, A. (). Introduction to Fluid Mechanics. John Wiley & Sons,
New York, NY.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
Schlichting, H. and Gersten, K. (). Boundary-layer theory. Springer-Verlag, Berlin,
Heidelberg, New York, eigth edition. Corrected printing.

j Self Study Problems j

. Order by increasing magnitude: boundary layer thickness 𝛿, displacement thick-

ness 𝛿1 , and momentum thickness Θ.

. Show that the integral for the ratio of momentum and boundary layer thickness
1 ( )
Θ 𝑢 𝑢
= 1− d𝜂
𝛿 ∫ 𝑢0 𝑢0
0

is equal to
1 ( )
𝑢 𝑢 4−𝜋
1− d𝜂 =
∫ 𝑢0 𝑢0 2𝜋
0

for the following assumed velocity profile

( )
𝑢 𝜋
= sin 𝜂 for 0 ≤ 𝜂 ≤ 1
𝑢0 2

. Following the procedures explained in Section ., compute boundary layer

thickness 𝛿, displacement thickness 𝛿1 , momentum thickness Θ, and wall shear
stress 𝜏𝑤 assuming the velocity profile is given by
( 𝑦)
𝑢 𝜋
= sin for 0 ≤ 𝑦 ≤ 𝛿
𝑢0 2 𝛿

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

12.7 Friction Force and Coefficients 145

. Following the procedures explained in Section . compute boundary layer

thickness 𝛿, displacement thickness 𝛿1 , momentum thickness Θ, and wall shear
stress 𝜏𝑤 assuming the velocity profile is given by

𝑢 3 1
= 𝜂 − 𝜂3 for 0 ≤ 𝜂 ≤ 1
𝑢0 2 2

Remember that 𝜂 = 𝑦∕𝛿.

. Why is the following equation not a suitable velocity profile for the boundary
layer flow?
𝑢 2 1
= 𝜂 − 𝜂3 for 0 ≤ 𝜂 ≤ 1
𝑢0 3 2

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

146

13
Frictional Resistance

Major resistance For most ships, frictional resistance is the major component of their calm water re-
component sistance. Wave resistance exceeds the frictional component only at Froude numbers
above 0.3. Obviously, the accurate prediction of frictional resistance 𝑅𝐹 is essential in a
ship’s performance assessment. Notwithstanding more than a century of research and
great advances in fluid mechanics, we still lack a unified theory of turbulent flow and
the resulting surface friction and momentum loss. As a consequence, our frictional re-
sistance predictions rely on approximations extracted from experimental observations
combined with physical insight drawn from conservation of mass, momentum, and
energy principles.
Ongoing The frictional resistance of ships results from their turbulent boundary layers, which are
j
research the topic of ongoing research efforts all over the world. Progress is regularly published j
through a number of dedicated conferences and journals. This chapter provides only a
short summary of turbulent boundary layer features. For further study, the textbooks
by Young () and White () may be of help. An in-depth discussion of the topic
is given by Schlichting and Gersten ().

Learning Objectives
At the end of this chapter students will be able to:

• discuss fundamental properties of turbulent boundary layers

• select flat plate friction coefficients
• apply the ITTC  model–ship correlation coefficient
• estimate effects of surface roughness

13.1 Turbulent Boundary Layers

Ships have In preceding chapters, we discussed the basics of boundary layer theory and derived
turbulent flat plate friction coefficients applicable to laminar flow. Reynolds numbers for boats
boundary layers
and ships are well beyond the limit for laminar flow. Consequently, ship boundary
layers will be turbulent except for a negligible fraction of their length at the bow. So

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.1 Turbulent Boundary Layers 147

why did we discuss laminar boundary layers at all? Simply put, we derived important
relationships which also apply to turbulent boundary layers.
The frictional resistance or drag of a flat plate of length 𝑥 follows from the total mo- Drag force
mentum loss along the plate as expressed by Equation (.). The momentum lost up
to the position 𝑥 is quantified by the momentum thickness Θ(𝑥).
𝑥

𝐹𝐹𝑥 = 𝜏𝑤 d𝑆 = 𝑏 𝜏𝑤 (𝜉) d𝜉 = 𝜌 𝑢20 𝑏 Θ(𝑥) (.)

∬ ∫
0

Accordingly, the frictional resistance coefficient 𝐶𝐹 for a flat plate of width 𝑏 and length
𝑥 is equal to
𝐹𝐹𝑥 Θ(𝑥)
𝐶𝐹 = = 2 (.)
1 2 𝑥
𝜌𝑢 𝑏 𝑥
2 0

Flat plate boundary layers are characterized by a vanishing pressure gradient in the exte- Shear stress
rior flow. With constant pressure along the boundary layer, shear stress is proportional
to the lengthwise change in momentum thickness. See Equation (.). Therefore, the
local skin friction coefficient is
𝜏𝑤 dΘ(𝑥)
𝐶𝑓𝑥 (𝑥) = = 2 (.)
1 d𝑥
2
𝜌 𝑢20
j Equations (.) and (.) are derived with flat plate boundary layer theory assump- j
tions, i.e. the boundary layer is assumed to be thin and the streamwise pressure gradient
𝜕𝑝∕𝜕𝑥 vanishes. However, they are valid for laminar as well as turbulent flow. Critical in
both expressions is the momentum thickness Θ as a function of the lengthwise position
𝑥. The momentum thickness, in turn, is a function of the velocity profile 𝑢∕𝑢0 in the
boundary layer (Equation (.)). A substantial amount of research is dedicated to the
development of suitable mathematical models for the velocity distribution in turbulent
boundary layers.
The basic nature of turbulence has been explained in Section .. Essentially, turbulence Turbulence
results in a rapid, irregular fluctuation of velocity which may be registered with fast
reacting instruments like hot wire anemometers. As introduced by Reynolds (),
it is convenient to split the velocity within a turbulent boundary layer into a time
independent mean velocity 𝑢 and a time dependent turbulence 𝑢′ .

𝑢(𝑡) = 𝑢 + 𝑢′ (𝑡) Reynolds decomposition (.)

Similar to laminar flat plate boundary layers, the thickness of turbulent boundary layers Boundary layer
is defined by the point where the flow velocity 𝑢 approaches the exterior flow speed 𝑢0 . thickness
Due to the unsteady turbulence, the distance of this point from the wall is fluctuating
over time at every position 𝑥 along the plate. A snapshot for a selected time 𝑡 = 𝑡0 is
shown on the upper right in Figure .. Most researchers define the boundary layer
thickness 𝛿 as the distance where the mean velocity 𝑢 reaches the exterior flow speed.
This is indicated by the dashed line in the snapshot and used as the definition of 𝛿 in this
chapter. However, some authors define the boundary layer thickness as the envelope of
all turbulent fluctuations. For example, see Grigson ().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

148 13 Frictional Resistance

Figure 13.1 Features of a turbulent boundary layer over a flat plate (zero pressure gradient)

Intermittency The flow outside of the boundary layer is free of turbulence and usually treated as
j inviscid. Approaching the boundary layer limit from the exterior flow, short bursts of j
turbulence occur where the boundary layer temporarily overshoots the limit 𝑦 = 𝛿(𝑥).
Klebanoff () found that these bursts may be detected at a distance of up to %
of 𝛿 from the wall. The duration of turbulence bursts rapidly increases within the
boundary layer. The ratio of periods with turbulence to total time observed is expressed
in the intermittency factor 𝛾 (Klebanoff, ). An intermittency factor of zero (𝛾 = 0)
indicates that the flow is free of turbulence. As shown in Figure ., the intermittency
factor grows rapidly toward the wall and reaches % at approximately % of the
boundary layer thickness.
Outer scaling When velocity profiles for turbulent boundary layers are plotted in the usual manner as
dimensionless ratio 𝑢∕𝑢0 over a dimensionless wall distance 𝜂 = 𝑦∕𝛿, very little can be
deduced about the flow close to the wall. This is known as outer scaling because it uses
the exterior flow speed 𝑢0 and the boundary layer thickness 𝛿 as reference scales. An
example is shown on the left side of Figure ..
Wall friction Prandtl (b) suggested that it is beneficial to use the wall friction velocity 𝑢𝜏 and
velocity and wall wall units 𝑦+ as reference quantities instead. Both parameters are extensively used in
units
the theory of turbulent boundary layers.
√
𝜏𝑤 𝑦𝑢𝜏
𝑢𝜏 = 𝑦+ = (.)
𝜌 𝜈
The wall friction velocity 𝑢𝜏 scales the velocity profile by the amount of shear stress
it creates. Observant readers may notice that the wall units 𝑦+ are a form of local
Reynolds number, which uses the distance from the wall 𝑦 as reference length and the
friction velocity 𝑢𝜏 as reference speed.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.1 Turbulent Boundary Layers 149

Figure 13.2 A typical turbulent boundary layer velocity profile depicted in outer and inner scaling

j j
Velocity profiles for turbulent boundary layers are usually presented in inner scaling. Inner scaling
The dimensionless velocity 𝑢+ = 𝑢∕𝑢𝜏 is plotted with a logarithmic axis of wall units 𝑦+
as shown on the right-hand side of Figure .. The logarithmic scaling of the wall unit
axis stretches the distances close to the wall so that the inner region covers most of the
graph.
Turbulent boundary layers have a more complex structure than laminar boundary layers. Turbulent
Broadly speaking, the turbulent boundary layer is subdivided into an inner and an outer boundary layer
structure
region. The inner region makes up about % of the boundary layer thickness in the
example shown in Figure .. Inner and outer region share an overlap layer.
For the outer region, von Kármán () stated that the dimensionless velocity profile Outer region
should only be a function of the exterior flow velocity 𝑢0 and the distance from the wall
𝜂. This is formulated as the velocity defect law.

𝑢0 − 𝑢
= 𝑓 (𝜂) velocity defect law (.)
𝑢𝜏

The difference 𝑢0 −𝑢 expresses how much the fluid has slowed down within the boundary
layer due to friction. In flows with nonzero pressure gradient (𝜕𝑝∕𝜕𝑥 ≠ 0), the velocity
distribution in the outer layer will also depend on 𝜕𝑝∕𝜕𝑥.
Prandtl (a) argued that the mean velocity distribution near the wall should be Inner region
governed by the wall friction, fluid properties, and the distance from the wall

𝑢 = 𝐺(𝜏𝑤 , 𝜌, 𝜇, 𝑦) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

150 13 Frictional Resistance

Dimensional analysis, which we will discuss in a later chapter, reveals that the dimen-
sionless velocity profile in the inner layer must be of the form

𝑢
𝑢+ = = 𝑔(𝑦+ ) inner law (.)
𝑢𝜏

The inner region is subdivided into three parts: a viscous sublayer, the buffer layer, and
the overlap layer, which may also be considered a part of the outer region (Figure .).
Viscous sublayer Directly adjacent to the wall lies the viscous sublayer. The presence of the wall dampens
the transverse turbulence and the flow attains laminar characteristics. Hence, some
older publications name this layer the laminar sublayer. The viscous sublayer stretches
only up to about 𝑦+ = 5. Schlichting and Gersten (, Chapter , p. ) report a
universal law of the wall which applies to flat plates. As shown in Figure ., 𝑢+ = 𝑦+
may be used as a first approximation.
Buffer layer The connection between viscous sublayer and the overlap layer is formed by the buffer
layer. In Figure ., the velocity profile is completed with Spalding’s approximation of
the velocity profile for the inner region (White, , Chapter ). The approximation
is given in an implicit form.

⎡ ( + )2 ( + )3 ⎤
𝜅𝑢 𝜅𝑢
𝑦 + +
= 𝑢 +e −𝜅𝐵 ⎢e𝜅𝑢+ − 1 − 𝜅𝑢+ − − ⎥ inner region (.)
⎢ 2 6 ⎥
⎣ ⎦
j j
Spalding () used a value of . for the von Kármán constant 𝜅 and the value 𝐵 = 4.1.
Spalding’s approximation fits velocity profiles measured at high Reynolds numbers
reasonably well in the inner region (see Figure .).
Wall layer The combination of viscous sublayer and buffer layer is sometimes referred to as wall
layer. The wall layer is very thin and takes up roughly % of the boundary layer thickness,
as shown in the velocity profile in outer scaling of Figure ..
Overlap layer The velocity profile must be continuous throughout the boundary layer. Therefore,
inner and outer region must share an overlap layer in which both velocity defect law
and inner law must be valid. Setting Equations (.) and (.) equal yields:

𝑢 𝑢
= 𝑔(𝑦+ ) = 0 − 𝑓 (𝜂) (.)
𝑢𝜏 𝑢𝜏

The general form of the velocity profile in the overlap layer is given by the logarithmic
overlap law.
1
𝑢+ = ln(𝑦+ ) + 𝐵 logarithmic overlap law (.)
𝜅
Some publications call this the ‘logarithmic law of the wall,’ which is somewhat mis-
leading. It is not valid at the wall. The logarithmic overlap law fits experimental data
fairly well in the range of 70 < 𝑦+ < 1000 for high Reynolds number flows.
The constants 𝜅 and 𝐵 have to be determined from experiments. 𝜅 is known as the von
Kármán constant and usually assigned the value 𝜅 = 0.41. However, reported values
range from . to .. There is indication from recent experiments that the von
Kármán constant has some dependency on the Reynolds number (Österlund et al.,

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.1 Turbulent Boundary Layers 151

data of Österlund (1999) data of Österlund (1999), =0.38

(1/7)
u/u0 = power law approximation 4 modified log−wake law Eq. (13.13) b.l. edge
10
inner region, Spalding (1961)
viscous sublayer u + = y +
1.5
Rex =2 0156864 log law 1 ln(y + )+ B
x =5 .50m

[−]
[−]

u0 =54.0360m/s
=0.07864m y + =1000
103

wall distance y + = yu /
= y/

boundary layer edge

1.0
wall distance

102
y + =70
u =1.7647m/s
−5
0.5 =1.4744⋅10 m2 /s
=0 .40
B =4.73
Π=0.7577
101
y + =1000
y + =70 y + =5
y + =5
0.0

0.0 0.2 0.4 0.6 0.8 1.0 0 5 10 15 20 25 30

velocity u/u0 [−] velocity u + = u/u [−]

j Figure 13.3 Comparing the modified log–wake law with experimental data from Österlund (1999) (Profile j
SW981113F)

). Reported values for the constant 𝐵 vary from . to ., but most frequently a
value of 𝐵 = 5.0 is used (Grigson, ).
Close to the boundary layer limit of flat plates, measured velocities exceed the values Log–wake law for
predicted by the logarithmic overlap law. Coles () added the wake function 𝑊 (𝜂) the outer region
to the logarithmic law to extend its use to the outer edge of the boundary layer. So
far, the wake function lacks a concise physical theory. Expressions are derived by data
fitting. A commonly used sine-squared form of the wake function is attributed to Hinze
(). Together with the logarithmic law, we get the log–wake law.
( )
1 2Π 2 𝜋𝜂
𝑢+ = ln(𝑦+ ) + 𝐵 + sin log–wake law (.)
𝜅 𝜅 2
The constant Π is Coles’ wake strength parameter. Its value depends on the pressure
gradient along the body. For high Reynolds number flows on flat plates, a value of
Π = 0.7577 is used (Guo et al., ). A drawback of this equation is that it violates
the boundary condition of vanishing slope 𝜕𝑢∕𝜕𝑦 = 0 at the boundary layer limit 𝑦 = 𝛿.
Guo () added a correction term and introduced the modified log–wake law which
has a vanishing derivative 𝜕𝑢∕𝜕𝑦 = 0 at 𝑦 = 𝛿.
( ) 𝜂3
1 2Π 2 𝜋𝜂
𝑢+ = ln(𝑦+ ) + 𝐵 + sin − modified log–wake law (.)
𝜅 𝜅 2 3𝜅

The modified log–wake law fits experimental data very well over the complete outer
region of the boundary layer. Figure . compares the modified log–wake law (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

152 13 Frictional Resistance

with the experimental data provided by Österlund (). As of May , Österlund’s
data sets are available at http://www.mech.kth.se/~jens/zpg/.
Power law The left-hand side plot in Figure . with the velocity profile in outer scaling also
contains an approximation of the velocity profile which was introduced by Prandtl
(a).
𝑢
= 𝜂 ∕7 power law (.)
1

𝑢0
It fits the data astonishingly well, but it neither satisfies the zero slope condition at
the outer edge of the boundary layer nor provides a finite slope at the wall. For higher
Reynolds number flows, the exponent may be lowered to 1∕9.
Ongoing and There are still many unanswered questions which hamper our ability to predict the
future work behavior of turbulent boundary layers. Unfortunately, we have not yet been able to
solve the Navier-Stokes equations for high Reynolds number flows even for the simple
looking case of a flat plate. Available direct numerical solutions are so far restricted to
low Reynolds numbers. Experimental investigations are also scarce for high Reynolds
numbers. There are several research groups actively working on the subject, though
(see for instance Monkewitz et al., ; Marusic et al., ). Accurate measurements
of wall shear stress and fluid velocity are very difficult in unsteady turbulent boundary
layers. Even the smallest hot wire anemometers have a finite diameter and locally
disturb the flow. Pressure taps have to be accurately drilled and the edges carefully
deburred to avoid affecting the flow. Any wind tunnel or circulating water tunnel is
j limited in the achievable Reynolds number by its dimensions and attainable flow speeds. j
An outdoor laboratory on a flat salt lake near Salt Lake City, Utah, is an attempt to
remove the dimensional restrictions. However, weather and geographical influences
render control of test conditions more difficult (Metzger et al., ).

13.2 Shear Stress in Turbulent Flow

Wall shear stress In contrast to viscous laminar flow, turbulent boundary layers feature two mechanisms
for shear stress generation. They share the viscosity driven shear stress with laminar
boundary layers, but also generate shear stress by dissipation of energy through turbu-
lent eddies. Using Reynolds’ decomposition of the flow speed (.), wall shear stress
may be expressed as a sum of viscous and turbulent shear stress.

𝜏𝑤 = 𝜏𝑣 + 𝜏𝑡 (.)

Viscous shear Close to the wall, viscosity of the fluid and its associated viscous stress will dominate.
stress Viscous stress is computed as in the case of a laminar boundary layer:

𝜕𝑢 𝜕𝑢
𝜏𝑣 = 𝜇 = 𝜌𝜈 viscous shear stress (.)
𝜕𝑦 𝜕𝑦

Turbulent shear In the outer region of the boundary layer flow, turbulent stress will be the dominant
stress component. Energy is dissipated in small and large eddies which form in the boundary
layer. Turbulent stress can be related to averages of products of turbulence velocities.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.3 Friction Coefficients for Turbulent Flow 153

For a flat plate the dominant component is

𝜕𝑢
𝜏𝑡 = 𝜌𝑢′ 𝑣′ = 𝜌𝜈𝑡 turbulent shear stress (.)
𝜕𝑦

The eddy viscosity 𝜈𝑡 is introduced to unify the notation. Turbulence models are used
to calculate the eddy viscosity (see Section .).
Combining Equations (.) and (.) for viscous and turbulent shear stresses and
dividing Equation (.) by the density yields

𝜏𝑤 ( 𝜕𝑢
= 𝜈 + 𝜈𝑡 ) (.)
𝜌 𝜕𝑦

This equation is not very helpful in determining shear stress because neither the eddy
viscosity nor the derivative of the mean velocity is easily determined. It is more practical
to rely on the relationship between momentum thickness and shear stress expressed in
Equations (.) and (.) to derive friction coefficient formulas.

13.3 Friction Coefficients for Turbulent Flow

Some of the first theoretical formulas for friction coefficients were developed by Ludwig Prandtl’s
j Prandtl and his research group in Göttingen, Germany. A number of these formulas semi-empirical
j
formula
are based on the power law velocity profile (.), and we outline the procedure below.
Initially, we follow the procedure from Section . and establish a relationship between Momentum
boundary layer thickness 𝛿 and momentum thickness Θ. To that effect, the power law thickness
𝑢∕𝑢0 = 𝜂 1∕7 is substituted into the definition of the momentum thickness (.)

1 ( ) 1
[ ]
𝑢 𝑢
d𝜂 = 𝛿 𝜂 ∕7 1 − 𝜂 ∕7 d𝜂
1 1
Θ = 𝛿 1−
∫ 𝑢0 𝑢0 ∫
0 0
1
[ ] [ ]
7 8 7 9 1 7
𝜂 ∕7 − 𝜂 ∕7 d𝜂 = 𝛿 𝜂 ∕7 − 𝜂 ∕7 = (.)
1 2
= 𝛿 𝛿
∫ 8 9 0 72
0

Thus, the local skin friction coefficient from Equation (.) becomes:

7 𝜕𝛿
𝐶𝑓𝑥 = (.)
36 𝜕𝑥

With the chosen velocity profile, we cannot exploit Newton’s shear stress formula,
Equation (.), because the 𝑦-derivative of 𝜂 1∕7 does not exist at the wall (𝑦 = 𝜂 = 0).
Prandtl (a) circumvented this problem by utilizing a shear stress formula that
Blasius () had previously derived for turbulent flow in pipes.

0.045
𝐶𝑓𝑥 = 1∕4
(.)
𝑅𝑒𝛿

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

154 13 Frictional Resistance

Combining Equations (.) and (.) and expanding the Reynolds number based on
boundary layer thickness 𝑅𝑒𝛿 = 𝑢0 𝛿∕𝜈 results again in an ordinary differential equation
for the boundary layer thickness:
( )1∕4
𝜈 7 𝜕𝛿
0.045 − = 0 (.)
𝑢0 𝛿 36 𝜕𝑥

With the approach 𝛿 = 𝐶𝑥4∕5 , a constant 𝐶 = 0.3707 is derived, and turbulent boundary
layer thickness and momentum thickness are given by
𝑥 𝑥
𝛿 = 0.3707 1∕5
and Θ = 0.036 1∕5
(.)
𝑅𝑒𝑥 𝑅𝑒𝑥

Note that turbulent boundary layers grow much faster than laminar boundary layers:
𝑥0.8 compared with 𝑥0.5 .
The desired flat plate friction coefficient is derived from Equation (.) with 𝑥 = 𝐿:

0.0721
𝐶𝐹 = (do not use for ships) (.)
𝑅𝑒 ∕5
1

This formula is not applicable to ships because the underlying Equation (.) for shear
stress in pipes is not accurate enough for high Reynolds numbers.
Prandtl– In a later paper, Prandtl (b) reported a flat plate friction formula which is based
j Schlichting on the logarithmic overlap law velocity profile (.). The momentum thickness was j
formula
computed numerically for a set of Reynolds numbers and converted into a convenient
regression formula.

0.455
𝐶𝐹 = [ ]2.58 Prandtl–Schlichting, 𝑅𝑒 ≤ 109 (.)
log10 (𝑅𝑒)

Equation (.) is known as the Prandtl–Schlichting formula and widely used in fluid
mechanics for cases with 𝑅𝑒 ≤ 109 .
Friction White () recommends the following flat plate friction formula for general use:
coefficient based
on inner 0.523
variables 𝐶𝐹 = [ ]2 flat plate, turbulent flow (.)
ln(0.06𝑅𝑒)

It is based on an analysis of Spalding’s formula for the inner region (.) by Kestin and
Persen ().
Summary Dozens of these friction coefficient formulas have been developed over the past 
years. A selection of friction coefficient curves is compared in Figure .. They all lie
within the band of available experimental data, and none of them may really claim it is
better than the others. It is the engineer’s responsibility to select a formula that fits the
posed problem and range of Reynolds numbers.
None of the formulas stated in this section is used for the evaluation of model test
results. Figure . also includes model–ship correlation lines, which are discussed in
the following section.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.4 Model–Ship Correlation Lines 155

j j

Figure 13.4 Flat plate friction coefficients for smooth surfaces

13.4 Model–Ship Correlation Lines

William Froude’s successful method of predicting full scale resistance from model tests Importance of
requires an accurate method to calculate the frictional resistance for model and full the friction
coefficient
scale vessel (see Section ..). The frictional resistance is subtracted from the measured
total resistance to determine the residual resistance coefficient 𝐶𝑅 .

𝐶𝑅 = 𝐶𝑇𝑀 − 𝐶𝐹𝑀 (Froude) (.)

In the current recommended procedures of the ITTC, frictional resistance is augmented

by a form factor 𝑘 to capture effects of the three-dimensional hull form. This method
was originally proposed by Hughes (). Subtracting the augmented flat plate friction
from the total resistance results in the wave resistance 𝐶𝑊 .

𝐶𝑊 = 𝐶𝑇𝑀 − (1 + 𝑘) 𝐶𝐹𝑀 (ITTC) (.)

ITTC used to call the result residuary resistance, until its most recent  edition of
the recommended procedures. This latest convention mirrors physics more closely,
because an accurate form factor will capture all viscous effects beyond the flat plate
resistance.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

156 13 Frictional Resistance

ITTC 1978 The wave resistance coefficient 𝐶𝑊 is considered a function of the Froude number alone.
performance Consequently, it is identical for model and full scale vessel. The full scale resistance is
prediction
method
then reassembled using the wave resistance coefficient 𝐶𝑊 , an augmented frictional
resistance coefficient (1 + 𝑘)𝐶𝐹𝑆 for the full scale vessel, plus some smaller corrections.
This is part of the ITTC  performance prediction method (ITTC, ). Details
are presented in Chapters  and .
Model basins In order to make full scale resistance predictions from model tests, Froude obtained a
grant from the British Royal Navy to build a towing tank, which is a long water basin
for testing ship models (Brown, ). He towed flat planks of up to . m length
and plotted their resistance as a function of speed (Froude, ). Since flat plates
have essentially no displacement, the measured resistance was taken as the frictional
resistance. Froude’s measurements were later cast into formula with the basic form of
𝑅𝐹 = 𝑐𝑣𝑚 𝑆
𝑆. 𝑆 is the wetted surface and 𝑐 is a length and surface roughness dependent
constant. The exponent 𝑚 approached a value of . for higher Reynolds numbers
(Schoenherr, ).
Schoenherr, Froude’s work initiated the advent of model basins all over the world, and numerous
ATTC line researchers conducted similar tests in the late th and early th centuries. Schoenherr
() collected the results and added his own experiments. Subsequently, Schoenherr
created a curve fit following an implicit function model suggested originally by von
Kárman.
0.242 ( )
0 = √ − log10 𝑅𝑒 𝐶𝐹𝐴𝑇 𝑇 𝐶 Schoenherr,  (.)
𝑅𝑒
j j
In , this friction line was adopted by the American Towing Tank Conference (ATTC)
for use with Froude’s method in the evaluation of ship model tests and henceforth called
the ATTC line. Although the underlying data covers only Reynolds numbers up to
about 𝑅𝑒 < 5 ⋅ 108 , it was applied to model and full scale vessels alike.
ITTC 1957 Equation (.) is awkward to use because it is implicit and nonlinear. Comparisons
model–ship of results with geometrically similar models of different scales indicated that a steeper
correlation line
slope in the range of model Reynolds numbers would be desirable. After a long and
sometimes contentious discussion of various proposed friction lines, the ITTC adopted
the following model–ship correlation line (ITTC, ).
0.075
𝐶𝐹 = 𝐶𝐹𝐼𝑇 𝑇 𝐶 = [ ]2 ITTC,  (.)
log10 (𝑅𝑒) − 2

The ITTC  line is a compromise for the specific purpose of making full scale
predictions from model test results, and its physical foundation is weak. To cite from
the ITTC’s decision (Saunders et al., ):

In view of the above, the Conference decides that the line given by the
formula [Equation (.)] is adopted as the ‘ITTC  model–ship cor-
relation line,’ it being clearly understood that this is regarded only as an
interim solution to the problem for practical engineering purposes.

Use in model Chapters  and  explain in detail how the ITTC  line is used in the evaluation
tests of model tests and the prediction of full scale resistance.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.5 Effect of Surface Roughness 157

Equation (.) is an astonishingly simple solution to a very complicated problem. The Possible
‘interim solution’ has now been in use for more than  years. Research on turbulent improvements
boundary layers has progressed considerably, as summarized in Section .. Several
proposals for an improved model–ship correlation line have been made but none has
been universally accepted yet. Grigson () developed an algorithm for the flat
plate friction coefficient considering a wake–log form of the velocity distribution in
the boundary layer. The algorithm is fairly complex, but Grigson provides a simple
modification to the ITTC  line which represents the results of his algorithm quite
nicely.
{
𝐺1 (𝑥1 ) 𝐶𝐹𝐼𝑇 𝑇 𝐶 for 1.5 ⋅ 106 < 𝑅𝑒 < 2 ⋅ 107
𝐶𝐹 = (.)
𝐺2 (𝑥2 ) 𝐶𝐹𝐼𝑇 𝑇 𝐶 for 2 ⋅ 107 ≤ 𝑅𝑒 < 6 ⋅ 109
The factors 𝐺1 (𝑥1 ) and 𝐺2 (𝑥2 ) are given as

𝐺1 (𝑥1 ) = 0.9335 + 0.147 𝑥21 − 0.071 𝑥31 (.)

with 𝑥1 = log10 (𝑅𝑒) − 6.3

𝐺2 (𝑥2 ) = 1.0096 + 0.0456 𝑥2 − 0.013944 𝑥22 + 0.0019444 𝑥32 (.)

with 𝑥2 = log10 (𝑅𝑒) − 7.3

Grigson’s friction line is also shown in Figure .. The line is very similar to Equa-
tion (.) by White (), which is reassuring, since both are based on formulations
j of the boundary layer velocity profile in inner variables. So far however, the ITTC  j
model–ship correlation line Equation (.) is still the recommended formula for the
evaluation of model test results.

13.5 Effect of Surface Roughness

The ITTC  model–ship correlation line and the other 𝐶𝐹 -formulas presented above Effect of surface
are designed for hydraulically smooth plates. Hydraulically smooth means that surface roughness
roughness has no influence on the flow in the boundary layer. This is the case as long as
surface imperfections do not penetrate through the viscous sublayer as illustrated in
Figure .(a). The dimensionless height of the bumps is smaller than 𝑦+ = 𝑦𝑢𝜏 ∕𝜈 = 5.
Surface imperfections which penetrate through the viscous sublayer into the buffer layer,
i.e. bumps higher than 𝑦+ > 5, the friction coefficient increases beyond the value for
the smooth surface. This is known as the transitional regime depicted in Figure .(b).
In cases where the surface imperfections reach across the wall layer into the overlap
layer, the frictional coefficient becomes independent of the Reynolds number 𝑅𝑒. In
this fully rough regime, the frictional coefficient is only a function of the relative height
of the surface roughness (Figure .(c)).
Defining the roughness of a surface is actually quite complicated. Surface imperfections Equivalent sand
may vary in size, form, and distribution. Different statistical parameters can be used roughness
to define the technical roughness 𝑘tech . Often the measured surface profile is divided
into smaller samples and a mean value is taken of the maximum surface imperfection
heights in each sample. The concept of equivalent sand roughness 𝑘𝑆 is introduced
in order to measure the effects of surface roughness independently of its shape and

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

158 13 Frictional Resistance

(a) Hydraulically smooth regime; no (b) Transitional regime: (c) Fully rough regime: vanishing
influence of surface roughness on noticeable effect of surface influence of viscous shear
shear stress and 𝐶𝐹 roughness on shear stress and stress. 𝐶𝐹 becomes
𝐶𝐹 independent of 𝑅𝑒

Figure 13.5 Types of technical surface roughness and their effect on friction

Figure 13.6 Definition of equivalent

sand roughness 𝑘𝑆

j j

distribution. Imagine a surface completely covered with spheres of diameter 𝑘𝑆 , as

shown on the right in Figure .. This can be achieved by covering a surface with a
uniform grain, like you find on sandpaper.
For each technical roughness an equivalent sand roughness is found by comparing the
frictional coefficients at high Reynolds numbers. To be more precise, the value of 𝑘+
𝑆
must be larger than  so that we are in the fully rough regime.

𝑘𝑆 𝑢𝜏
𝑘+
𝑆
= > 70 fully rough regime (.)
𝜈
Again, the friction velocity 𝑢𝜏 of Equation (.) is employed to define the Reynolds
number 𝑘+𝑆
based on the equivalent sand roughness as the reference length.
The influence of surface roughness was extensively studied by Prandtl’s research group
in Göttingen. Nikuradse () studied the flow in pipes with varying sand roughness.
Based on his results, Prandtl and Schlichting () developed a theoretical model for
the flow over smooth and rough plates. Its details are beyond the scope of this book, but
the interested reader may study Chapters  and  of Schlichting and Gersten ()
for an up-to-date version. The theoretical model results in a set of implicit equations
for the frictional coefficient 𝐶𝐹 of a plate as a function of the Reynolds number 𝑅𝑒 and
the relative surface roughness 𝑘𝑆 ∕𝐿.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.5 Effect of Surface Roughness 159

−4
7 kS /L =5⋅10
transitional
regime
fully rough limit kS+ =70
[10−3 ]

6
−4
kS /L =2⋅10
turbulent flat plate friction coefficient CF

fully rough regime

5 −4
kS /L =1⋅10

−5
kS /L =5⋅10
4
−5
kS /L =2⋅10
−5
3 kS /L =1⋅10
−6
kS /L =5⋅10
−6
ITTC 1957 line kS /L =2⋅10
−6
kS /L =1⋅10
−7
2 kS /L =5⋅10
−7
kS /L =2⋅10
CF (Re, ks /L) based on equiv. sand roughness kS /L =5⋅10 −8

1 CF (Re, ktech /L) based on technical roughness

ITTC 1957 model−ship correlation line smooth plate
approx. of fully rough CF by Equation (13.34)
j 0
105 106 107 108 109 1010
j
u0 L
Reynolds number Re = [−]

Figure 13.7 Flat plate friction coefficient for turbulent flow and its dependency on Reynolds number and relative
surface roughness 𝑘𝑆 ∕𝐿

Figure . depicts the resulting friction coefficient curves. Some may recognize the Friction
similarity with the Moody chart which is used in the design of piping systems. For coefficient for
rough plates
resistance predictions, ITTC recommends the use of an equivalent sand roughness
for welded steel ship hulls of 𝑘𝑆 = 150 ⋅ 10−6 m. Thus, for a ship of length 𝐿 = 300 m,
relative surface roughness would be 𝑘𝑆 ∕𝐿 = 2 ⋅ 10−6 . If we pick up the 𝐶𝐹 -curve for
𝑘𝑆 ∕𝐿 = 2⋅10−6 at the right side of the chart and follow it to lower Reynolds numbers (to
the left), we see that the frictional coefficient is initially constant with 𝐶𝐹 = 2.3858⋅10−3 .
Around 𝑅𝑒 = 2 ⋅ 109 the 𝐶𝐹 -curve starts to dip, and we enter the transitional regime
where the surface roughness affects the viscous flow in the wall layer. Close to 𝑅𝑒 = 108
it merges with the curve for smooth plates.
Two important conclusions can be drawn from the 𝐶𝐹 -curves in Figure .:

• In the fully rough regime, the frictional resistance coefficient becomes indepen-
dent of the Reynolds number. As a consequence, frictional resistance grows
quadratic with speed for high Reynolds numbers and the higher the relative
surface roughness 𝑘𝑆 ∕𝐿 the higher the final value of 𝐶𝐹 .
• A surface that is hydraulically smooth at low Reynolds numbers may show effects
of surface roughness at high Reynolds numbers. A plate with relative roughness

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

160 13 Frictional Resistance

𝑘𝑆 ∕𝐿 = 2 ⋅ 10−5 shows almost no roughness effect at Reynolds number 𝑅𝑒 = 107

(model scale) whereas at 𝑅𝑒 = 109 (full scale) its friction coefficient is more than
twice that of a smooth plate.

Technical versus Figure . also features a set of dashed curves which deviate from the solid curves
equivalent sand in the transitional regime. The dashed curves represent the friction coefficients for
roughness
plates with technical roughness, i.e. plates roughened by randomly distributed surface
imperfections. In contrast to the plates with uniform roughness, the dip in 𝐶𝐹 -curves
through the transitional regime is noticeably absent in experiments with plates with
technical roughness 𝑘tech . Therefore, more conservative resistance estimates are ob-
tained for technical systems if the 𝐶𝐹 -vales are taken from the dashed curves in the
transitional regime.
Approximation In older editions of Schlichting’s book on boundary layer theory, an approximation is
of fully rough 𝐶𝐹 provided for the flat plate friction coefficient in the fully rough regime. It seems to have
been omitted in Schlichting and Gersten () but is restated in White ().
[ ( )]−2.5
𝐿
𝐶𝐹 rough = 1.89 + 1.62 log10 (.)
𝑘𝑆
The values are shown as squares on the right-hand side of Figure .. The estimates
are fairly accurate for the relative roughness of ship hulls 𝑘𝑆 ∕𝐿 ≈ 10−6 .
For comparison, the ITTC  model–ship correlation line (.) has been added to
j Figure .. It emphasizes that the ITTC  line predicts lower frictional coefficients j
at full scale than modern theory and experiments suggest.
Roughness Towing tank models are built to be hydraulically smooth, but actual ship hulls do not
allowance satisfy this criterion. Therefore, we will augment the resistance curve for the full scale
vessel by a roughness allowance Δ𝐶𝐹 (see Chapter , page ).
Keep the Revisit the distribution of wall shear stress over the length of a plate in Figure ..
bow/leading Although the specific graph is for laminar flow, the fact that wall shear stress is highest
edge smooth
at the beginning of a plate or body and then declines also applies to turbulent boundary
√
layers. High wall shear stress 𝜏𝑤 , in turn, implies high friction velocity 𝑢𝜏 = 𝜏𝑤 ∕𝜌.
With higher friction velocity the Reynolds number 𝑘+ 𝑆
= 𝑘𝑆 𝑢𝜏 ∕𝑛𝑢 is more likely to
exceed the limit for fully rough regime (𝑘+ 𝑆
= 70) at the beginning than at the end of
a plate, i.e. the initial part of a plate may fall into the fully rough regime although the
plate overall may not. Therefore, it is more important to smooth and clean the bow of a
vessel or the leading edge of foil than their downstream parts.

13.6 Effect of Form

Thin plate versus Evidently, ship hulls are not flat plates. An infinitely thin plate would not produce waves
ship and would not have a wave resistance. However, an infinitely flat plate has no volume
and, as a result, no buoyancy. Consequently, our imaginary thin ship would have a low
resistance but could carry neither cargo nor its own weight.
Varying external For flat plates we assumed that the exterior flow pressure is constant along the plate.
pressure Then, the exterior flow velocity 𝑢𝑒 is constant according to the Bernoulli equation. In

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.7 Estimating Frictional Resistance 161

the literature this is called a zero pressure gradient flow. The assumption of 𝑝 = const.
allows us to eliminate the second term in Equation (.) for the wall shear stress.
Pressure varies over a three-dimensional, voluminous body, and the second term in
Equation (.) has to be retained. As a consequence, computation of shear stress
and boundary layer thickness becomes significantly more complicated. The changes in
exterior flow pressure affect the growth of the boundary layer thickness and the actual
shear stress distribution. In addition, the boundary layer pushes the exterior flow away
from the body: the displacement effect. We will revisit this problem in Chapter , after
we have learned more about the pressure distribution in exterior flows.

13.7 Estimating Frictional Resistance

Frictional resistance 𝑅𝐹 is the integral of the shear stress vectors over the wetted surface Frictional
of the ship hull. Especially for ships at lower speeds, 𝑅𝐹 is by far the largest resistance resistance
component. Although the ITTC  model–ship correlation line (.) is technically
not an accurate friction line, it is widely used in estimates of the frictional resistance of
ships.
First, the Reynolds number 𝑅𝑒 needs to be computed: Reynolds
number
𝑣𝑆 𝐿𝑂𝑆
𝑅𝑒 = (.)
𝜈
j j
As input, we require ship speed 𝑣𝑆 , length over wetted surface 𝐿𝑂𝑆 , and the kinematic
viscosity 𝜈. The kinematic viscosity 𝜈 is the ratio of dynamic viscosity and density. For
example, let us compute the frictional resistance for a ship of 212 m length over wetted
surface sailing at 20 knots. It is standard practice to make resistance predictions for
salt water at temperature 15 ◦ C.
Obviously, the speed has to be converted into suitable SI-units first. 1 knot is equivalent Unit conversion
to the speed of 1 international nautical mile per hour. A ‘knot’ is not an official SI-unit
but its usage is allowed (BIPM, ). Conversion into meter per second is obtained
with the following factor:

1 international nautical mile 1852.0 m m

1 kn = = = 0.51444̄ (.)
1 hour 3600.0 s s
Using an abbreviated value of 0.51444 is sufficiently accurate. Therefore, our ship speed
of 20 kn is equivalent to

20 kn = 20 ⋅ 0.51444 m/s = 10.289 m/s

Using the kinematic viscosity value of 1.1892 ⋅ 10−6 m2 /s for seawater at temperature
 degree Celsius (ITTC, ), the Reynolds number in our example is
10.289 m/s ⋅ 212 m
𝑅𝑒 = = 1 834 231 416.1
1.1892 ⋅ 10−6 m2 ∕s

Reynolds numbers for full scale ships are very high, typically in the range of 108 to
1010 . When you compute the frictional resistance coefficient according to the 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

162 13 Frictional Resistance

model–ship correlation line (.), make sure you use the common logarithm function
with base  and not the natural logarithm.

0.075 −3
𝐶𝐹 = ( )2 = 1.4216 ⋅ 10
log10 (1834231416.1) − 2

This value is valid only for a hydraulically smooth surface.

The frictional resistance itself will be
𝜌 2
𝑅𝐹 = 𝑣 𝑆 𝐶𝐹
2 𝑆
1026.021 kg m−3 ( )2
= 10.289 m s−1 5000 m2 ⋅ 1.4216 ⋅ 10−3
2
kg m
= 386 026 = 386 026 N = 386.03 kN
s2
The frictional resistance component is the only part of the resistance for which we have
such a simple procedure. Other components, like the viscous pressure resistance and the
wave resistance, require model test results or the numerical solution of the fundamental
equations. We will come back to these resistance components in Chapters  and ,
respectively.

j References j

BIPM (). The international system of units (SI). The International Bureau of
Weights and Measures (BIPM), Sèvres, France, eigth edition.
Blasius, H. (). Das Ähnlichkeitsgesetz bei Reibungsvorgängen in Flüssigkeiten: über
den Gültigkeitsbereich der beiden Ähnlichkeitsgesetze in der Hydraulik, volume 
of Mitteilungen über Forschungsarbeiten auf dem Gebiete des Ingenieurwesens, ins-
besondere der Technischen Hochschulen. VDI-Verlag Berlin.
Brown, D. (). The way of a ship in the midst of the sea – The life and work of William
Froude. Periscope Publishing Ltd.
Coles, D. (). The law of the wake in turbulent boundary layer. Journal of Fluid
Mechanics, ():–.
Froude, W. (). Experiments on surface-friction experienced by a plane moving
through water. Read before the British Association for the Advancement of Sci-
ence at Brighton. Published in The Papers of William Froude, A.D. Duckworth, The
Institution of Naval Architects, London, United Kingdom, pp. –, .
Grigson, C. (). An accurate smooth friction line for use in performance prediction.
In Transactions of The Royal Institution of Naval Architects (RINA), volume ,
pages –. The Royal Institution of Naval Architects.
Grigson, C. (). A planar friction algorithm and its use in analysing hull resistance.
In Transactions of The Royal Institution of Naval Architects (RINA), volume ,
pages –. The Royal Institution of Naval Architects.
 Unfortunately, the natural logarithm is named log() in many programming languages. Logarithm

functions with base  are often accessible as log().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

13.7 Estimating Frictional Resistance 163

Guo, J. (). Turbulent velocity profiles in clear water and sediment-laden flows. PhD
thesis, Colorado State University, Department of Civil Engineering, Fort Collins, CO.
Guo, J., Julien, P., and Meroney, R. (). Modified log⣓wake law for zero-pressure-
gradient turbulent boundary layers. Journal of Hydraulic Research, ():–.
Hinze, J. (). Turbulence. McGraw-Hill, New York, NY.
Hughes, G. (). Friction and form resistance in turbulent flow, and a proposed
formulation for use in model and ship correlation. RINA Transactions and Annual
Report, ():–. doi:./rina.trans...
ITTC (). th International Towing Tank Conference, Madrid Spain. ITTC. Available
at www.ittc.info.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
ITTC ().  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
Kestin, J. and Persen, L. (). Application of Schmidt’s, method to the calculation of
Spalding’s function and of the skin-friction coefficient in turbulent flow. Int. Journal
of Heat and Mass Transfer, (–):–.
Klebanoff, P. (). Characteristics of turbulence in a boundary layer with zero pres-
sure gradient. NACA Technical Report , National Advisory Committee for
Aeronautics.
Marusic, I., Hutchins, N., and Mathis, R. (). High Reynolds number effects in
j wall turbulence. In Sixth International Symposium on Turbulence and Shear Flow j
Phenomena, pages –, Seoul, Korea.
Metzger, M., McKeon, J., and Holmes, H. (). The near-neutral atmospheric surface
layer: turbulence and non-stationarity. Philosophical Transactions of the Royal Society
of London A, :–⣓.
Monkewitz, P., Chauhan, K., and Nagib, H. (). Self-consistent high-Reynolds-
number asymptotics for zero-pressure-gradient turbulent boundary layers. Physics
of Fluids, ().
Nikuradse, J. (). Laws of flow in rough pipes. Technical Memorandum TM ,
National Advisory Committee for Aeronautics (NACA). Translation of Strömungs-
gesetze in rauhen Rohren, published in  as VDI Forschungsheft , Beilage zu
Forschung auf dem Gebiet des Ingenieurwesens, Ausgabe B, Band , Juli/August.
Österlund, J. (). Experimental studies of zero pressure-gradient turbulent boundary
layer flow. PhD thesis, Royal Institute of Technology, Department of Mechanics,
Stockholm, Sweden. ISRN KTH/MEK/TR–/–SE.
Österlund, J., Johansson, A., Nagib, H., and Hites, M. (). Wall shear stress mea-
surements in high Reynolds number boundary layers from two facilities. In th
AIAA Fluid Dynamics Conference, Norfolk, VA. American Institute of Aeronautics
and Astronautics. AIAA -.
Prandtl, L. (a). Über den Reibungswiderstand strömender Luft. In Tollmien, W.,
Schlichting, H., and Görtler, H., editors, Gesammelte Abhandlungen zur angewandten
Hydro- und Aerodynamik, volume , pages –. Springer Verlag, Berlin, Heidel-
berg. Original published in  as Ergebnisse der Aerodynamischen Versuchsanstalt
zu Göttingen, Oldenbourg München-Berlin, III. Lief., S.–.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

164 13 Frictional Resistance

Prandtl, L. (b). Zur turbulenten Strömung in Rohren und längs Platten. In Tollmien,
W., Schlichting, H., and Görtler, H., editors, Gesammelte Abhandlungen zur ange-
wandten Hydro- und Aerodynamik, volume , pages –. Springer Verlag,
Berlin, Heidelberg. Original published in  as Ergebnisse der Aerodynamischen
Versuchsanstalt zu Göttingen, Oldenbourg München-Berlin, IV. Lief., S.–.
Prandtl, L. and Schlichting, H. (). Das Widerstandsgesetz rauher Platten. In
Tollmien, W., Schlichting, H., and Görtler, H., editors, Gesammelte Abhandlungen
zur angewandten Hydro- und Aerodynamik, volume , pages –. Springer
Verlag, Berlin, Heidelberg. Original published  in Werft, Reederei, Hafen . Jg.,
S.–.
Reynolds, O. (). On the dynamical theory of incompressible, viscous fluids and
the determination of the criterion. Philosophical Transactions of the Royal Society
London. A, :–.
Saunders, H., Walker, W., and Mazarredo, I. (). Concluding technical session. In
th International Towing Tank Conference, pages –, Madrid Spain. ITTC.
Schlichting, H. and Gersten, K. (). Boundary-layer theory. Springer-Verlag, Berlin,
Heidelberg, New York, eigth edition. Corrected printing.
Schoenherr, K. (). Resistance of flat surfaces moving through a fluid. SNAME
Transactions, :–.
Spalding, D. (). A single formula for the law of the wall. Journal of Applied
Mechanics, ():–. doi:./..
j von Kármán, T. (). Mechanical similitude and turbulence. Technical Memorandum j
, National Advisory Committee for Aeronautics.
White, F. (). Viscous fluid flow. McGraw-Hill, New York, NY, third edition.
Young, A. (). Boundary layers. AIAA Education Series. American Institute of
Aeronautics and Astronautics (AIAA), Washington, DC.

Self Study Problems

. State and describe the regions and layers in a turbulent boundary layer.

. Download a velocity profile from Österlund’s data sets, available at http://www.

mech.kth.se/~jens/zpg/. Plot the data in outer and inner scaling (see Figure .).
Add the power law profile Equation (.) and compare!
. Explain why surface roughness at the bow is more detrimental to its drag than
roughness at the stern.
. Repeat the estimate of frictional resistance from Section . for the speeds 10 kn
and 15 kn.
. Repeat the estimate of the frictional resistance from Section . for the speeds
10 kn, 15 kn, and 20 kn with the friction formula (.). Compare the results
with those from the text and previous problem.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

165

14
Inviscid Flow

As stated in Section ., the coefficient for the shear stress forces is very small for water.
We will discuss a number of flow conditions, where we treat the fluid as inviscid, i.e. we
set its kinematic viscosity to zero:

𝜈 = 0 for inviscid fluid (.)

This assumption leads to many practical solutions in fluid mechanics. However, we

have to be careful in interpreting the results because an actual inviscid fluid does not
exist. The hypothetical inviscid fluid is also known as ideal fluid.
After defining the Euler equations, we derive the important Bernoulli equation for a
streamline. It connects pressure and flow velocity and is widely used in fluid mechanics.
j Later, we will derive more specific forms of the Bernoulli equation for specific flow j
types. This chapter also introduces the concepts of rotation, vorticity, and circulation.

Learning Objectives
At the end of this chapter students will be able to:

• identify the Euler equations

• discuss and apply the Bernoulli equation
• define vorticity and circulation

14.1 Euler Equations for Incompressible Flow

The conservation of momentum equations for inviscid fluids were formulated by Euler Euler equations
() roughly  years before Navier included viscous forces in his version of the
conservation of momentum equations. We can obtain the Euler equations for the
conservation of momentum for an incompressible (𝜌 = const.) and inviscid fluid by
setting the kinematic viscosity to zero (𝜈 = 0) and, consequently, omitting the last term
in the Navier-Stokes equations (.).
𝜕𝑣 ( ) 1
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 Euler equations (.)
𝜕𝑡 𝜌

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

166 14 Inviscid Flow

This is the differential, conservative form of the Euler equations in Cartesian coordinates.
Careful Solving the Euler equations will lead to incorrect results for the total resistance of a
application body. They ignore the viscous effects which occur in the boundary layer. Although less
required
demanding than the Navier-Stokes equations, the Euler equations (.) still represent
a formidable challenge for numerical solutions. The fundamental tasks of discretiz-
ing a three-dimensional domain and solving large systems of linear equations remain.
Nonetheless, the Euler equations still provide the basis for practical solutions of impor-
tant problems in ship hydrodynamics like wave theory, wave resistance, lifting surfaces,
and propeller theory to name a few. This requires some extra restrictions in addition
to the underlying assumptions in Equations (.) that the fluid is incompressible and
inviscid.
Alternative form A slightly different form of the Euler equations (.) is obtained by modifying the
of Euler left-hand side with the following vector identity.
equations
( 𝑇 ) 1
𝑣 ∇ 𝑣 = ∇ (𝑣𝑇 𝑣) − 𝑣 × (∇ × 𝑣) (.)
2
This vector identity is confirmed by expanding the terms on both sides.
Substituting the identity (.) into the Euler equations (.) above yields
𝜕𝑣 1 1
+ ∇ (𝑣𝑇 𝑣) − 𝑣 × (∇ × 𝑣) = 𝑓 − ∇ 𝑝 (.)
𝜕𝑡 2 𝜌
j We will later use this second form of the Euler equations to derive the important j
Bernoulli equation for potential flow. In potential flows, the last term on the left-hand
side vanishes because the curl of the fluid velocity vanishes, i.e. ∇ × 𝑣 = 0. We call these
flows irrotational. More on this in Chapter .

14.2 Bernoulli Equation

The terms in the Euler equations (.) represent forces per unit mass. An integration
of the forces along a streamline yields the work done by the forces. Of course, for
unsteady flow this is just a snapshot as the streamlines may change over time. Before we
perform this integration, we take a look at the geometry and kinematics of the problem
(Figure .).
Tangent vector The tangent 𝑠 to the streamline is a vector of unit length |𝑠| = 1.

𝑠 = (𝑠𝑥 , 𝑠𝑦 , 𝑠𝑧 )𝑇

Its coordinates 𝑠𝑥 , 𝑠𝑦 , and 𝑠𝑧 represent the cosines of the angles a streamline forms with
the coordinate axes at that point.
A line element d𝑠 of the streamline is defined as a vector stretching an infinitesimally
small distance d𝑠 in tangent direction.

⎛ d𝑥 ⎞ √
d𝑠 = ⎜ d𝑦 ⎟ with length d𝑠 = d𝑥2 + d𝑦2 + d𝑧2 (.)
⎜ ⎟
⎝ d𝑧 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14.2 Bernoulli Equation 167

Figure 14.1 A fluid element d𝑚 moves from point A to point B along a streamline

Per definition, the velocity vector 𝑣 is tangent to the streamline in every instant. Velocity vector

⎛𝑢⎞ √
𝑣=⎜𝑣⎟ with speed |𝑣| = 𝑢2 + 𝑣2 + 𝑤2
⎜ ⎟
⎝𝑤⎠
Consequently, both line element d𝑠 and velocity 𝑣 can be rewritten as a product of the
tangent vector 𝑠 and their respective length.
d𝑠 = 𝑠 d𝑠 𝑣 = 𝑠 |𝑣| (.)
j Work is often defined as force times distance traveled. In D space, however, force Work in a vector
j
vector and motion are not always parallel. Only the portion of the forces pointing field
in the direction of motion will contribute to the work done on a fluid element. The
dot product allows us to compute the component of a force vector 𝐹 pointing in the
direction of the line element d𝑠. The work done by the force 𝐹 along the distance d𝑠 is
equal to
d𝑊 = d𝑠𝑇 𝐹 = 𝑠𝑇 𝐹 d𝑠 (.)
Of course, we assumed that d𝑠 is so small that the force is considered constant over the
distance d𝑠.
Integration of the differential work element d𝑊 along the streamline results in the total Line integral
work done between the points A and B. We dot-multiply Equation (.) with the unit
length tangent vector 𝑠 and obtain the forces per unit mass that point in the direction
of motion. Subsequently, we integrate the forces per unit mass along the streamline
from point A to point B.
𝑠B 𝑠B 𝑠B 𝑠B
𝜕𝑣 ( ) 1
𝑠 𝑇
d𝑠 + 𝑇
𝑠 𝑇
𝑣 ∇ 𝑣 d𝑠 = 𝑇
𝑠 𝑓 d𝑠 − 𝑠𝑇 ∇ 𝑝 d𝑠 (.)
∫ 𝜕𝑡 ∫ ∫ ∫ 𝜌
𝑠A 𝑠A 𝑠A 𝑠A

We will study the integrals individually and start with the second integral on the left- Study of second
hand side of Equation (.). Its integrand contains the dot product of velocity vector term
and Nabla operator. We substitute the velocity vector with the product of tangent
vector times speed, i.e. 𝑣 = 𝑠|𝑣|.
( 𝑇 ) ( )𝑇
𝑣 ∇ = 𝑠|𝑣| ∇ = |𝑣|𝑠𝑇 ∇ (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

168 14 Inviscid Flow

The speed |𝑣| is a scalar and can be extracted from the dot product. The dot product of
a unit vector and the Nabla operator is called a directional derivative.
𝜕
𝑠𝑇 ∇ = (.)
𝜕𝑠
It specifies the component of the gradient that points in direction of the unit vector
(here 𝑠). Together we have:
( 𝑇 ) 𝜕
𝑣 ∇ = |𝑣| (.)
𝜕𝑠
We substitute this into the second integral of Equation (.).
𝑠B 𝑠B 𝑠 𝑠
( ) ( ) B
( )𝑇 𝜕𝑣
B
𝜕𝑣
𝜕
𝑠 𝑣𝑇 ∇ 𝑣 d𝑠 =
𝑇
𝑠 |𝑣|
𝑇
𝑣 d𝑠 = 𝑠 |𝑣| d𝑠 = 𝑣𝑇 d𝑠
∫ ∫ 𝜕𝑠 ∫ 𝜕𝑠 ∫ 𝜕𝑠
𝑠A 𝑠A 𝑠A 𝑠A
(.)
Again, we utilized the facts that a scalar factor in a dot product can be moved from one
operand to the other and that 𝑣 = 𝑠 |𝑣|.
We apply the product rule to the following expression:
( )𝑇
𝜕( 𝑇 ) 𝜕𝑣 𝜕𝑣 𝜕𝑣
𝑣 𝑣 = 𝑣𝑇 + 𝑣 = 2 𝑣𝑇 (.)
𝜕𝑠 𝜕𝑠 𝜕𝑠 𝜕𝑠
j j
Multiplication with 1∕2 yields an alternative expression for the integrand of the second
integral.
𝜕𝑣 1 𝜕( 𝑇 )
𝑣𝑇 = 𝑣 𝑣 (.)
𝜕𝑠 2 𝜕𝑠
𝑠B 𝑠B 𝑠B
( ) 𝜕𝑣
1 𝜕( 𝑇 )
𝑠𝑇 𝑣𝑇 ∇ 𝑣 d𝑠 = 𝑣𝑇 d𝑠 = 𝑣 𝑣 d𝑠 (.)
∫ ∫ 𝜕𝑠 2 ∫ 𝜕𝑠
𝑠A 𝑠A 𝑠A

The integral of a derivative of a function is the function itself. Therefore, the second
integral in Equation (.) results in
𝑠B 𝑠B
( ) [ ]
1 𝜕( 𝑇 ) 1 ( 𝑇 ) 𝑠B
𝑠 𝑣𝑇 ∇ 𝑣 d𝑠 =
𝑇
𝑣 𝑣 d𝑠 = 𝑣 𝑣 (.)
∫ 2 ∫ 𝜕𝑠 2 𝑠A
𝑠A 𝑠A

Since the dot product of a vector with itself is equal to its length squared, the work per
unit mass associated with the convective acceleration is
𝑠B
( ) [ ]𝑠B
1 1 1
𝑠𝑇 𝑣𝑇 ∇ 𝑣 d𝑠 = |𝑣|2 = |𝑣B |2 − |𝑣A |2 (.)
∫ 2 𝑠A 2 2
𝑠A

𝑣A being the velocity vector at point A and 𝑣B being the velocity vector at point B,
respectively. The expression on the right-hand side is equivalent to the increase in
kinetic energy per unit mass.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14.2 Bernoulli Equation 169

Next, we take a closer look at the integral of the body forces per unit mass. It is the first Study of third
integral on the right-hand side of Equation (.). The gravity force is the only body term
force of note that we consider here. Thus, the vector of body forces 𝑓 per unit mass is
pointing in negative 𝑧-direction (𝑧 is positive upwards) with strength 𝑔 (gravitational
acceleration).
𝑠B 𝑠B 𝑠
)⎛ 0 ⎞
B
( ( )
𝑇
𝑠 𝑓 d𝑠 = 𝑠𝑥 , 𝑠𝑦 , 𝑠𝑧 ⎜ 0 ⎟ d𝑠 = − 𝑔 𝑠𝑧 d𝑠 (.)
∫ ∫ ⎜ ⎟ ∫
𝑠A 𝑠A ⎝ −𝑔 ⎠ 𝑠A

The product of the cosine 𝑠𝑧 of the angle between path element and vertical axis and
the length d𝑠 of the path elements is equal to the differential d𝑧 = 𝑠𝑧 d𝑠. The work per
unit mass due to the gravity force then depends only on the vertical distance (𝑧B − 𝑧A )
between points A and B, a fact you hopefully remember from your physics classes.
𝑠B 𝑧B
( )
𝑇
𝑠 𝑓 d𝑠 = − 𝑔 d𝑧 = −𝑔 𝑧B − 𝑧A (.)
∫ ∫
𝑠A 𝑧A

For the integral of the work done by pressure forces (surface force), we employ findings
from previous integrals.
𝑠B 𝑠B 𝑠B
𝜕𝑝 [ ]𝑠B
1 1
𝑇 1 1
− 𝑠 ∇ 𝑝 d𝑠 = − 𝑠𝑇 ∇ 𝑝 d𝑠 = − d𝑠 = − 𝑝
j ∫ 𝜌 𝜌∫ 𝜌 ∫ 𝜕𝑠 𝜌 𝑠A j
𝑠A 𝑠A 𝑠A
1( )
= − 𝑝B − 𝑝A (.)
𝜌

Finally, we substitute the results for the integrals (.), (.), and (.) into Equa- Bernoulli
tion (.). equation for
unsteady flow
𝑠B
𝜕𝑣 1 1 ( ) 1( )
𝑠𝑇 d𝑠 + |𝑣B |2 − |𝑣A |2 = −𝑔 𝑧B − 𝑧A − 𝑝B − 𝑝A (.)
∫ 𝜕𝑡 2 2 𝜌
𝑠A

We multiply the equation by the density, move terms depending on the initial position
A to the right-hand side, and move terms depending on the final point B to the left. The
result is known as the Bernoulli equation for unsteady flow.
𝑠B
𝜕𝑣 1 1
𝜌 𝑠𝑇 d𝑠 + 𝜌 |𝑣B |2 + 𝜌 𝑔 𝑧B + 𝑝B = 𝜌 |𝑣A |2 + 𝜌 𝑔 𝑧A + 𝑝A (.)
∫ 𝜕𝑡 2 2
𝑠A

This equation may be used, for instance, to compute the water hammer pressure which
occurs in a flow line when a valve is rapidly closed.
Note that points A and B lie on the same streamline. The right-hand side of the unsteady
Bernoulli equation (.) may conveniently be expressed as a constant.
𝑠B
𝜕𝑣 1
𝜌 𝑠𝑇 d𝑠 + 𝜌 |𝑣B |2 + 𝜌 𝑔 𝑧B + 𝑝B = 𝐶A (.)
∫ 𝜕𝑡 2
𝑠A

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

170 14 Inviscid Flow

Figure 14.2 Determining the flow

speed by measuring pressure difference
in a contraction nozzle

The constant 𝐶A usually changes from streamline to streamline.

j j
Bernoulli The remaining path integral on the left-hand side vanishes for steady flow because
equation for 𝜕𝑣∕𝜕𝑡 = 0.
steady flow
1 1
𝜌 |𝑣B |2 + 𝜌 𝑔 𝑧B + 𝑝B = 𝜌 |𝑣A |2 + 𝜌 𝑔 𝑧A + 𝑝A (.)
2 2
Equation (.) emphasizes the important result that the sum of dynamic, hydrostatic,
and ambient pressure is constant along a streamline (inviscid flow).
1
𝜌 |𝑣|2 + 𝜌 𝑔 𝑧 + 𝑝 = const. (.)
2
We will derive more specialized versions of the Bernoulli equation later.
Application of The Bernoulli equation is an important tool in the evaluation of flow properties. Con-
Bernoulli sider the flow through the contraction nozzle of a cavitation tunnel (Figure .). We
equation
intend to determine the flow velocity from the difference in pressure between inlet and
outlet. The circulation pump is moving water through the nozzle from left to right. We
assume the flow is steady and incompressible. The tunnel is equipped with pressure taps
before and after the nozzle. The taps are perpendicular to the flow direction. A u-tube
manometer, filled with mercury or more environmentally friendly fluids, measures the
pressure difference between inlet and outlet.
Conservation of The conservation of mass principle dictates that mass fluxes are equal at inlet 1 and
mass 2 . For steady, incompressible flow, the continuity equation for a fixed control volume
requires that the net flow across the surface of the control volume vanishes.
( ) 𝐴2
𝜌 𝑣𝑇 𝑛 d𝑆 = −𝑣1 𝐴1 + 𝑣2 𝐴2 = 0 or 𝑣1 = 𝑣2 (.)
∬ 𝐴1
𝑆

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14.3 Rotation, Vorticity, and Circulation 171

The Bernoulli equation is our representation of the conservation of momentum princi- Conservation of
ple. We restate Equation (.) for the steady flow along a streamline between inlet momentum
and outlet.
1 1
𝜌 |𝑣1 |2 + 𝜌 𝑔 𝑧1 + 𝑝1 = 𝜌 |𝑣2 |2 + 𝜌 𝑔 𝑧2 + 𝑝2 (.)
2 2
Along the center line of the tunnel hydrostatic pressure remains unchanged (𝑧1 = 𝑧2 ).
This eliminates the hydrostatic pressure term 𝜌𝑔𝑧 on both sides. We solve the remaining
equation for the velocity 𝑣2 .
√
√
√2 𝑝1 − 𝑝2
𝑣2 = √ √𝜌 [ ( )2 ] (.)
√ 𝐴
√ 1− 2
𝐴1

The u-tube manometer measures the pressure difference 𝑝1 − 𝑝2 . Hydrostatic equilib- Pressure
rium between left and right arm of the manometer requires that difference

𝑝1 = 𝑝2 + 𝜌𝐻𝑔 𝑔 Δℎ or 𝑝1 − 𝑝2 = 𝜌𝐻𝑔 𝑔 Δℎ (.)

We substitute this result into Equation (.):

√
√𝜌
√ 𝐻𝑔 2 𝑔 Δℎ
𝑣2 = √√ 𝜌 [ ( )2 ] (.)
√ 𝐴2
j √ 1− j
𝐴1

The density of the fresh water in the tank is 𝜌 = 998.2072 kg/m3 (for a temperature of
 ◦ C) and the density of mercury is 𝜌𝐻𝑔 = 13579.04 kg/m3 . Given a cavitation tunnel
with a contraction ratio of 𝐴2 ∕𝐴1 = 0.4 and a reading of Δℎ = 85.2 mm mercury on
the manometer, the velocity at the outlet 2 will be
√
13579.04 2 ⋅ 9.807 m/s2 0.0852 m
𝑣2 = ⋅ [ ] ≈ 5.202 m/s (.)
998.2072 1 − 0.42

High density ratios of gauge fluid and water (here 𝜌𝐻𝑔 ∕𝜌) enable the measurement of
large pressure differences. Mercury is about . times heavier than water. However,
mercury is a potent poison and an environmental hazard. Therefore, manometers filled
with mercury have to be handled with great care. Colored oil might be used as an
alternative gauge fluid.

14.3 Rotation, Vorticity, and Circulation

We study a differential fluid element moving with the flow. In general, a fluid element Fluid element
will change position and shape in four different ways: movement

. translation

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

172 14 Inviscid Flow

Figure 14.3 Translation and linear deformation of a fluid element

. stretching (linear deformation)

. rotation
. shearing (angular deformation)

Translation and For better visualization, we momentarily restrict ourselves to planar (two-dimensional)
j stretching flow. As the fluid element changes position in the 𝑥-𝑦-plane (translation) it may also j
lengthen and shorten (stretching). Assume that the fluid element has the side lengths
d𝑥 and d𝑦 and an area d𝑆 = d𝑥d𝑦 at time 𝑡 (Figure .). Horizontal velocity at corner
A will be 𝑢 and 𝑢 + (𝜕𝑢∕𝜕𝑥) d𝑥 at corner B. The latter is again obtained by a Taylor series
expansion which is truncated after the linear term.
The distances 𝑢d𝑡 and 𝑣d𝑡 express the translation of the fluid element as a whole. At time
𝑡+d𝑡, the length of side AB will have changed from d𝑥 to the new length d𝑥+(𝜕𝑢∕𝜕𝑥) d𝑥d𝑡
due to the difference in velocities at points A and B. In the vertical direction, the length
of side AD will change from d𝑦 to d𝑦 + (𝜕𝑣∕𝜕𝑦) d𝑦d𝑡 (Figure .). For a moving fluid
element filled with incompressible fluid (𝜌 = const.), conservation of mass requires
that the area (planar flow, or volume for D-flow) of the element remains constant. If
one side lengthens, the other side must shrink.
area at time 𝑡 = area at time 𝑡 + d𝑡
d𝑥d𝑦 = AB ⋅ AD
( )( )
𝜕𝑢 𝜕𝑣
= d𝑥 + d𝑥d𝑡 d𝑦 + d𝑦d𝑡
𝜕𝑥 𝜕𝑦
We expand the product on the right-hand side and divide by the area d𝑆 = d𝑥d𝑦 on
both sides. The remaining change of the surface area must vanish:
𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣
0 = d𝑥d𝑦d𝑡 + d𝑥d𝑦d𝑡 + d𝑥d𝑦d𝑡2
𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦
The last term is proportional to d𝑡2 , i.e. it is of second order small. Since we already cut
off higher order terms in the Taylor expansions for the velocities, it is only consequent

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14.3 Rotation, Vorticity, and Circulation 173

Figure 14.4 Rotation and angular deformation of a fluid element

to ignore this second order term as well. The remaining equation is divided by the area
d𝑆 = d𝑥d𝑦, and we obtain the two-dimensional continuity equation for incompressible
fluids.
𝜕𝑢 𝜕𝑣
+ = 0 (.)
𝜕𝑥 𝜕𝑦
j This form of the continuity equation enforces that elongation in one direction is accom- j
panied by shortening in another direction. As a result, area or volume stay constant.
This is not necessarily the case for compressible fluids like air of course.
Figure . explains the effects of changing velocities on rotation and angular deforma- Rotation and
tion of the fluid element. Like the horizontal velocity 𝑢, the transverse velocity 𝑣 may shearing

also change along the side AB. Over a short time period d𝑡 point B will rise above point
A by the distance
( )
𝜕𝑣 𝜕𝑣
𝑣+ d𝑥 d𝑡 − 𝑣 d𝑡 = d𝑥 d𝑡
𝜕𝑥 𝜕𝑥
As a result, side AB rotates by the angle d𝛼1 around point A with

𝜕𝑣
d𝑥 d𝑡
𝜕𝑥 𝜕𝑣
tan(d𝛼1 ) = = d𝑡 ≈ d𝛼1
d𝑥 𝜕𝑥
For small angles sin(d𝛼1 ) ≈ d𝛼1 and cos(d𝛼1 ) ≈ 1. Hence, tan(d𝛼1 ) ≈ d𝛼1 . The angular
velocity 𝜔1 of side AB becomes

d𝛼1 𝜕𝑣
𝜔1 = = (.)
d𝑡 𝜕𝑥

Side AD of the fluid element is moving in 𝑥 direction. If point D is moving faster than
point A, side AD will turn clockwise around point A (Figure .). Since we used the
positive sign for the anti-clockwise rotation of side AB, clockwise rotation is negative.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

174 14 Inviscid Flow

Side AD rotates by the angle d𝛼2 with

𝜕𝑢
d𝑦 d𝑡
𝜕𝑦 𝜕𝑢
tan(d𝛼2 ) = − = − d𝑡
d𝑦 𝜕𝑦

We assume d𝛼2 is small and obtain the angular velocity of side AD

d𝛼2 𝜕𝑢
𝜔2 = = − (.)
d𝑡 𝜕𝑦

Rotation We define the mean value of the angular velocities 𝜔1 and 𝜔2 as the angular velocity of
the element. The axis of rotation is perpendicular to the 𝑥-𝑦-plane and parallel to the
𝑧-axis. ( )
1( ) 1 𝜕𝑣 𝜕𝑢
𝜔𝑧 = 𝜔1 + 𝜔2 = − (.)
2 2 𝜕𝑥 𝜕𝑦
Considering the rotation of fluid elements in the 𝑦-𝑧-plane and 𝑧-𝑥-plane respectively,
leads to angular velocities about the 𝑥- and 𝑦-axes.
( )
1 𝜕𝑤 𝜕𝑣
𝜔𝑥 = − (.)
2 𝜕𝑦 𝜕𝑧
( )
1 𝜕𝑢 𝜕𝑤
𝜔𝑦 = − (.)
2 𝜕𝑧 𝜕𝑥
j j
Summarizing Equations (.) through (.) yields the general vector of angular
velocity 𝜔 for a three-dimensional fluid element d𝑉 .

⎛ 𝜕𝑤 𝜕𝑣
− ⎞
⎜ 𝜕𝑦 𝜕𝑧 ⎟
⎛ 𝜔𝑥 ⎞ ⎜ ⎟ ( )
1⎜ 1 𝜕𝑢 𝜕𝑤 ⎟ = 1 ∇ × 𝑣 = 1 rot 𝑣
𝜔 = 𝜔𝑦 ⎟ = ⎜ − (.)
2⎜ ⎟ 2⎜ 𝜕𝑧 𝜕𝑥 ⎟ 2 2
⎝ 𝜔𝑧 ⎠ ⎜ 𝜕𝑣 𝜕𝑢 ⎟
⎜ − ⎟
⎝ 𝜕𝑥 𝜕𝑦 ⎠
The rotation of the velocity rot 𝑣 is also known as curl 𝑣.
Vorticity The vector of vorticity 𝛾 is defined as twice the angular velocity.

𝛾 = 2 𝜔 = ∇ × 𝑣 = rot 𝑣 (.)

According to Helmholtz’s theorems of fluid mechanics, vorticity cannot change in

an inviscid fluid (Helmholtz, ). More information on Helmholtz’s theorems is
provided in Section .. Generation and dissipation of vorticity may only be explained
by viscous flow theory.
Circulation This leads us to the concept of circulation. Circulation Γ is defined as the line integral
of the projection of the velocity 𝑣 onto a closed path 𝐶 (Figure .). The projection is
computed as a dot product and, consequently, a scalar quantity.

Γ = 𝑣𝑇 𝑠 d𝑠 (.)
∮
𝐶

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

14.3 Rotation, Vorticity, and Circulation 175

Figure 14.5 Definition of circulation Γ

Figure 14.6 Symmetric foil with lifting flow (Γ ≠ 0) and nonlifting flow (Γ = 0)

j j
The integral is similar to the work integral which we discussed at the beginning of the
chapter.
Stokes’ integral theorem reveals a connection between circulation and vorticity:

Γ = 𝑣𝑇 𝑠 d𝑠 = (∇ × 𝑣)𝑇 𝑛 d𝑆 = 𝛾 𝑇 𝑛 d𝑆 (.)
∮ ∬ ∬
𝐶 𝑆 𝑆

Circulation is equal to the vorticity integrated over the surface 𝑆 enclosed by the path
𝐶.
At first glance, circulation Γ appears to be a rather abstract, mathematical concept.
However, circulation is important in potential theory and for the description of lifting
flows as they occur on foils, e.g. rudders and propeller blades (Figure .). An asym-
metric foil with camber, or a symmetric foil at a nonzero angle of attack, produces a
lifting force perpendicular to the onflow. According to Kutta-Joukowsky’s lift theorem,
the lift force 𝐿 is proportional to density 𝜌 of the fluid, onflow velocity 𝑈∞ , and the
circulation Γ. A symmetric foil at zero angle of attack will not develop a lift force, and
the circulation is zero. We will come back to this when we discuss foil theory and
propellers.

References
Euler, L. (). Principes généraux du mouvement des fluides. In Mémoires de
l’Académie des Sciences de Berlin, volume , pages –.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

176 14 Inviscid Flow

Helmholtz, H. (). On the integrals of the hydrodynamic equations that correspond

to vortex motions. Translation by Uwe Parpart. Int. Journal of Fusion Energy, (–
):–. Original published in German in Journal für die reine und angewandte
Mathematik, :–, .

Self Study Problems

. Show that the vector identity of Equation (.) is true.
( 𝑇 ) 1
𝑣 ∇ 𝑣 = ∇ (𝑣𝑇 𝑣) − 𝑣 × (∇ × 𝑣) (.)
2

. Explain the difference between the Euler and the Navier-Stokes equations.
. Which important function does the Bernoulli equation serve?
. State the conditions under which the Bernoulli equation (.) may be applied.
. Define and explain circulation.
. Consider the following velocity field:
−𝑦
j ⎛ [ ] ⎞ j
⎜ 𝑥2 + 𝑦2 ⎟
𝛾 ⎜ 𝑥 ⎟
2𝜋 ⎜⎜ [𝑥2 + 𝑦2 ] ⎟
𝑣 =
⎟
⎜ ⎟
⎝ 0 ⎠

Compute the vector of vorticity 𝛾.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

177

15
Potential Flow

In this section we discuss potential flow. Although viscous effects are ignored, it has
many practical applications. Potential flow is used to determine basic properties of
hydrofoils and lifting wings. That includes keels, rudders, and propellers. It is also
employed in the prediction of cavitation inception and in seakeeping. We start this
section by introducing the concept of the velocity potential. The velocity potential is
then used to derive the important Laplace and Bernoulli equations from the principles
of conservation of mass and momentum.

Learning Objectives
j j
At the end of this chapter students will be able to:

• explain the concept of a velocity potential

• apply the principles of conservation of mass and conservation of momentum to

potential flow

• predict velocities and pressure for potential flow

• identify limitations of potential flow

15.1 Velocity Potential

In Chapter , we introduced the circulation as the integral of the velocity vector along Circulation
a closed path through the flow field.

Γ = 𝑣𝑇 𝑠 d𝑠 (.)
∮
𝐶

The dot product between velocity vector and path element 𝑣𝑇 𝑠d𝑠 looks similar to the
dot product of force vector and path element when work is computed.
The work to move an object from point A to point B is determined by integration of the Work

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

178 15 Potential Flow

Figure 15.1 The work spent on moving an

object from point A to point B

external force 𝐹 acting on the object along the path from A to B.

B

𝑊 = 𝐹 𝑇 𝑠 d𝑠 (.)
∫
A

Again, vector 𝑠 represents the unit length tangent vector to the path. A major difference
to the definition of circulation is that the path for the work does not have to be closed.
As the dot product between force 𝐹 and tangent vector indicates, only the component
of the force vector pointing in the tangent direction is contributing to the work. No
j work is done when the force is perpendicular to the path. j
Usually, we prefer to move objects along the shortest path from A to B to conserve
energy. Sometimes a longer path is used because the required force might be too large
otherwise. Mountain roads have switchbacks because your car engine would not be
strong enough to push the car directly up the hill. Does that mean you spend more
Conservative energy following the switchbacks? Not really, because the car works in a conservative
force field force field. In conservative force fields, like Earth’s gravity, the work done in moving
an object from A to B becomes independent of the path. Assuming the force field 𝐹
depicted in Figure . is conservative, the work integrals for paths 𝐴, 𝐵, and 𝐶 will be
equal.
B

𝑊 = 𝐹 𝑇 𝑠 d𝑠 = 𝑊𝐴 = 𝑊𝐵 = 𝑊𝐶 (.)
∫
A
As a consequence, the work 𝑊 is only a function of the location of the end points A and
B. To be fair, in our mountain road analogy we probably spend some more energy on the
switchback road due to the longer exposure to roll and wind resistance. However, the
work related to just lifting the car is the same for the steep hill climb and the switchbacks.
Potential The capability of a conservative force field to perform work is expressed by its potential.
Like work, potential is a scalar function. Work is equal to the difference in potential
between point B and point A.
𝑊 = 𝐸B − 𝐸A (.)

In Earth’s gravity field, the potential for an object of mass 𝑚 is

𝐸 = −𝑔 𝑧 𝑚 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.1 Velocity Potential 179

The minus sign stems from the chosen positive 𝑧-direction (positive upwards). The
work related to lifting the car will be
( )
𝑊 = 𝐸B − 𝐸A = −𝑔 𝑧B 𝑚 − (−𝑔 𝑧A 𝑚) = −𝑔 𝑚 𝑧B − 𝑧A (.)

The value is negative if 𝑧B > 𝑧A because we have to work against the force field to lift
the car higher. If it runs downhill, the work is done by the force field and is positive.
Note that the amount of work follows from the change in potential energy between A
and B. The absolute size of the potential is irrelevant, there is no difference in work if
you lift the car from  m to  m or from  m to  m.
The rate of change in potential is equal to the vector field. If we take the gradient of the
potential (.), we obtain the gravity force acting on mass 𝑚.

⎛ 𝜕𝐸 ⎞ ⎛ 𝜕(−𝑔 𝑚 𝑧) ⎞
⎜ 𝜕𝑥 ⎟ ⎜ 𝜕𝑥 ⎟ ⎛ 0 ⎞
⎜ ⎟ ⎜ ⎟
∇𝐸 = ⎜
𝜕𝐸
⎟ = ⎜⎜
𝜕(−𝑔 𝑚 𝑧) ⎟ = ⎜⎜ 0 ⎟
⎟ (.)
⎜ 𝜕𝑦 ⎟ 𝜕𝑦 ⎟ ⎜ ⎟
⎜ 𝜕𝐸 ⎟ ⎜ 𝜕(−𝑔 𝑚 𝑧) ⎟ ⎝ −𝑔 𝑚 ⎠
⎝ ⎠ ⎜ ⎟
𝜕𝑧 ⎝ 𝜕𝑧 ⎠

We will come back to this important relationship between a vector field and its potential.
Not all force fields are conservative. So you may ask, when is a force field 𝐹 conservative Sufficient
and when does a potential exist? Mathematically speaking, a force field is conservative condition
j j
and a potential exists, if the rotation of the vector field vanishes in a simply connected
region (Kellogg, ).
rot (𝐹 ) = ∇ × 𝐹 = 0 (.)
Such a vector field is called irrotational. Equation (.) is a sufficient condition, i.e. if it
holds true, the vector field 𝐹 is conservative. If ∇ × 𝐹 ≠ 0, we cannot draw a conclusion.
Additional checks will be necessary to determine whether a potential exists or not.
A simply connected region is a space where all closed curves can be contracted to a Simply and
point without leaving the region. Study Figure .: In a simply connected region a multiply
connected
sling can be pulled tight to any point within the region without any part of the sling regions
leaving the gray shaded area. In Figure .(a) no part of the loop leaves the gray shaded
area during the tightening of the sling. If the region contains an object like a foil, any
sling which encompasses the object can no longer be reduced to a point. This situation
is depicted in Figure .(b). The closest you get is the dotted line. Such a region is
called multiply connected. Any region that does not contain the object is still simply
connected.
Let us switch back to fluid mechanics. We started this section by pointing out that the
definition of the circulation (.) looks similar to the work integral (.). However,
the path 𝐶 is always closed and the force field 𝐹 is replaced with the velocity field 𝑣.
We apply the concept of a potential to the velocity field. A flow field is called irrotational Velocity
if it satisfies equation (.) in all simply connected regions. potential

rot (𝑣) = ∇ × 𝑣 = 0 (.)

 See Section .. for an introduction to rotation, also called curl

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

180 15 Potential Flow

(a) Simply connected regions can be (b) A multiply connected region cannot be
reduced to a point reduced to a point

Figure 15.2 Definition of simply and multiply connected regions

A velocity potential 𝜙 will exist under this condition.

Potential flow A flow field represented by a velocity potential is a potential flow. The only formal
condition is that the flow is irrotational. In practice, this usually implies that the fluid is
j also inviscid. Shear stresses in the fluid will cause rotation. j
Analog to the relationship (.) between weight force and the potential of the gravity
field, the velocity vector is equal to the gradient of the velocity potential

𝑣 = ∇𝜙 (.)

This relationship is extremely useful. For a potential flow, we may replace the velocity
vector which has three unknown components with the yet unknown scalar potential
leading to a reduction in equations and unknowns to solve for.
Two basic examples may illustrate this reduction.

Parallel flow (i) Consider the following velocity potential:

𝜙(𝑥, 𝑦, 𝑧) = −𝑈 𝑥 (.)

The velocity potential has the physical units m2 /s. Taking the gradient yields the
velocity vector.
( )
⎛ 𝜕 − 𝑈𝑥 ⎞
⎜ 𝜕𝑥 ⎟ ⎛ −𝑈 ⎞
⎜ 𝜕 ( )⎟ ⎜ ⎟
𝑣(𝑥) = 𝑣(𝑥, 𝑦, 𝑧) = ∇ 𝜙 = ⎜ − 𝑈𝑥 ⎟ = ⎜ 0 ⎟ (.)
⎜ 𝜕𝑦 ⎟ ⎜ ⎟
( ) ⎝ 0 ⎠
⎜ 𝜕 − 𝑈𝑥 ⎟
⎝ 𝜕𝑧 ⎠

At every point 𝑥 = (𝑥, 𝑦, 𝑧)𝑇 in the flow field, the velocity in 𝑥-direction is equal
to −𝑈 and zero in all other directions. This represents parallel flow in negative

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.1 Velocity Potential 181

2.0 2.0
lines of constant potential

1.5 1.5

1.0 1.0

0.5 0.5
y [m]

y [m]
0.0 0.0

0.5 0.5

1.0 1.0

1.5 1.5
lines of constant potential
2.0 2.0
2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0
x [m] x [m]

(a) Parallel flow 𝜙 = −𝑈 𝑥 in the 𝑥-𝑦-plane (b) Vortex flow 𝜙 = 𝛾∕(2𝜋) arctan(𝑦∕𝑥) in the 𝑥-𝑦-plane

Figure 15.3 Examples of basic potential flows

j 𝑥-direction. Figure .(a) shows the velocity field and selected isolines of the j
potential (𝜙 = const.).

(ii) The next example is a bit more complex. Vortex flow

𝛾 (𝑦)
𝜙(𝑥, 𝑦, 𝑧) = arctan (.)
2𝜋 𝑥
The flow field described by this potential has this velocity distribution:

𝑣(𝑥, 𝑦, 𝑧) = ∇ 𝜙
[ ( 𝑦 )] −𝑦
⎛ 𝜕 ⎞ ⎛ ⎞
arctan
⎜ 𝜕𝑥 𝑥 ⎟ ⎜ 𝑥2 + 𝑦2 ⎟
𝛾 ⎜ [ ( 𝑦 )] ⎟ 𝛾 ⎜ ⎟
𝜕 𝑥 (.)
2𝜋 ⎜⎜ ⎟ = 2𝜋 ⎜ ⎟
= arctan
𝜕𝑦 𝑥 ⎟ 2
⎜ 𝑥 +𝑦 ⎟2
[ ( 𝑦 )]
⎜ 𝜕
arctan ⎟ ⎜ ⎟
⎝ 𝜕𝑧 𝑥 ⎠ ⎝ 0 ⎠

The strength of the vortex 𝛾 is given in units of m2 /s.

( ) ( ) ( )
At points 𝑥1 = 1, 0, 0 m, 𝑥2 = 0, 1, 0 m, and 𝑥3 = 1, 1, 0 m the velocity
vectors are
1 1
⎛0 m ⎞ ⎛ −1 m ⎞
𝛾 ⎜ 1 ⎟ 𝛾 ⎜ 1 ⎟
2𝜋 ⎜⎜ ⎟ 2𝜋 ⎜⎜ ⎟
𝑣1 (𝑥1 ) = 1 m
𝑣2 (𝑥2 ) = 0 m
1 ⎟ 1 ⎟
⎝0 m
⎠ ⎝ 0 m
⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

182 15 Potential Flow

1 1
⎛ −2 m ⎞
𝛾 ⎜ 1 1 ⎟
2𝜋 ⎜⎜ 2 ⎟
𝑣3 (𝑥3 ) = m
1 ⎟
⎝ 0 m
⎠
When we evaluate the velocity vectors at many points, we see that the fluid moves
in concentric circles around the 𝑧-axis (Figure .(b)). The magnitude of the
velocity declines rapidly as we move further away from the origin.
At the origin (𝑥, 𝑦)𝑇 = (0, 0) m, we cannot evaluate the velocity. The derivatives
𝜕𝜙∕𝜕𝑥 and 𝜕𝜙∕𝜕𝑦 of the potential do not exist, and the velocity components
become infinite:
𝑢, 𝑣 ⟶ (−∞, ∞) for (𝑥, 𝑦)𝑇 ⟶ (0, 0)
A point at which a function approaches infinity is called a singular point or, suc-
cinctly, a singularity. Therefore, we call the vortex potential a singularity. We will
discuss more of these special velocity potentials in the following chapter.

Velocity is The gradient of a function always points in the direction of its steepest ascent and is
perpendicular to always perpendicular to isolines of the function. Therefore, the velocity – the gradient of
isolines of
potential
a potential – is always perpendicular to the lines of constant potential (see Figure .).

15.2 Circulation and Velocity Potential

j j
Before we study the equations that allow us to actually compute a potential, we explore
the somewhat confusing connection between circulation Γ and velocity potential 𝜙.
No circulation – We started this chapter recapping the definition of circulation Γ:
no lift force
Γ = 𝑣𝑇 𝑠 d𝑠 = dΓ (.)
∮ ∮
𝐶 𝐶

We also mentioned previously that the lift force 𝐿 generated by hydrofoils, sails, etc.
is proportional to the circulation Γ. This is known as Kutta-Joukowsky’s lift theorem
(Moran, ).
𝐿 = 𝜌𝑈 Γ (D, force per unit length) (.)
Thus, vanishing circulation Γ = 0 means there is no lift force 𝐿 either.
With Stokes’ integral theorem (see Equation (.) on page ), we also found a
relationship between circulation and rotation (curl) of the velocity vector.

Γ = 𝑣𝑇 𝑠 d𝑠 = (∇ × 𝑣)𝑇 𝑛 d𝑆 = (rot 𝑣)𝑇 𝑛 d𝑆 (.)

∮ ∬ ∬
𝐶 𝑆 𝑆

The last integral of Equation (.) above will vanish in potential flow fields because
rot 𝑣 = 0 in simply connected regions. Therefore, the circulation is zero for all paths 𝐶
enclosing a simply connected region 𝑆.

Γ = dΓ = 0 for potential flow (simply connected) (.)

∮
𝐶

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.2 Circulation and Velocity Potential 183

So, how can we use potential flow to describe lifting flows around hydrofoils?
The answer to that question lies in the difference between simply and multiply connected
regions. Consider the differential line element in the path integral of Equation (.):

𝑠 d𝑠 = (d𝑥, d𝑦, d𝑧)𝑇 (.)

A comparison of the integrands in (.) above enables us to state the total differential
dΓ of the circulation:

dΓ = 𝑣𝑇 𝑠 d𝑠 = 𝑢d𝑥 + 𝑣d𝑦 + 𝑤d𝑧 (.)

For potential flow, the components 𝑢, 𝑣, and 𝑤 of the velocity vector may be replaced Total differentials
with the spatial derivatives of the velocity potential. dΓ and d𝜙

𝜕𝜙 𝜕𝜙 𝜕𝜙
dΓ = d𝑥 + d𝑦 + d𝑧 (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

The term on the right-hand side also represents the definition of the total differential
for the potential 𝜙.

𝜕𝜙 𝜕𝜙 𝜕𝜙
d𝜙 = d𝑥 + d𝑦 + d𝑧 = dΓ (potential flow) (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

Thus, the total differentials of circulation and potential are identical for potential flow:
j d𝜙 = dΓ. j
Combining the findings of Equations (.) and (.) with the definition of circula-
tion (.) yields:

Γ = dΓ = d𝜙 = 0 for potential flow (simply connected) (.)

∮ ∮
𝐶 𝐶

A definite integral is equal to the difference of antiderivative values at the limits.

B

d𝜙 = 𝜙B − 𝜙A (.)
∫
A

Start and end point will be equal for a closed path A=B. Employing this in Equa-
tion (.) yields
B=A

Γ = d𝜙 = 𝜙B=A − 𝜙A = 0 (.)
∮
A

Therefore, the potential must be single valued in a simply connected region:

𝜙A = 𝜙B=A for a simply connected region (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

184 15 Potential Flow

(a) Path 𝐶 encloses the foil. The region is multiply (b) Path 𝐶 ′ excludes the foil. The region is simply
connected and the circulation does not vanish connected and the circulation vanishes

Figure 15.4 Flow field around a symmetric foil at angle of attack 𝛼

Lifting foil Let’s apply this to the lifting flow around a foil. Figure .(a) shows a path 𝐶 enclosing
a foil. The symmetric foil is operating at a small angle of attack 𝛼. We know from
experience and observation that the foil will generate a lift force 𝐿 perpendicular to
the onflow. The presence of a lift force 𝐿 ≠ 0 requires that Γ ≠ 0. Consequently, the
potential at points A=B cannot be single valued.
Γ = 𝜙B − 𝜙A ≠ 0 for a multiply connected region (.)
j j
This is not a contradiction of potential flow theory because the region enclosed by the
path 𝐶 is multiply connected with the foil in it.
A simply connected region is constructed by extending the path to exclude the foil
(Figure .(b)). The extended path 𝐶 ′ follows the wake streamline from point B to
the trailing edge of the foil, then wraps around the foil and returns along the wake
streamline to point A. Since fluid does not cross the wake streamline this artificial
barrier will not disrupt the flow. In contrast to the figure, there is no distance between
the path elements above and below the wake barrier. Integration of dΓ along path 𝐶 ′
will find Γ′ = 0 because the region enclosed by 𝐶 ′ is simply connected.

15.3 Laplace Equation

Velocity potentials are a powerful tool for the description of a large class of flows, so-
called potential flows. The application of the conservation of mass and conservation
of momentum principles for Newtonian fluids led us to the continuity equation and
Navier-Stokes equations (see Chapters  and ). We restricted ourselves to fluids with
Viscous, constant density 𝜌 = const., which we call incompressible. The differential conservative
incompressible forms of continuity equation and Navier-Stokes equations are (see Chapter ).
flow
∇𝑇𝑣 = 0 continuity equation, 𝜌 = const. (.)
𝜕𝑣 ( ) 1
+ 𝑣𝑇 ∇ 𝑣 = 𝑓 − ∇ 𝑝 + 𝜈𝚫𝑣 NSE, 𝜌 = const. (.)
𝜕𝑡 𝜌

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.3 Laplace Equation 185

Outside the boundary layer, viscous effects are small and may often be neglected. This Inviscid,
simplification leads to the Euler equation (see Chapter ). The continuity equation incompressible
flow
does not change of course.
∇𝑇𝑣 = 0 continuity equation, 𝜌 = const. (.)
𝜕𝑣 ( 𝑇 ) 1
+ 𝑣 ∇ 𝑣 = 𝑓 − ∇𝑝 Euler eq., 𝜈 = 0, 𝜌 = const. (.)
𝜕𝑡 𝜌

For potential flow, we add the restriction that the flow must be irrotational in sim- Inviscid,
ply connected regions: rot 𝑣 = 0. In that case, the velocity vector may be replaced incompressible,
irrotational flow
by the gradient of a velocity potential. We substitute 𝑣 = ∇ 𝜙 into the continuity
equation (.).
∇𝑇𝑣 = 0
∇𝑇∇𝜙 = 0
𝚫𝜙 = 0 (.)
The dot product of the Nabla operator with itself results in the Laplace operator, which
has been introduced in Section ... Rewriting the last equation in Cartesian coordi-
nates yields the important Laplace equation.

𝜕2𝜙 𝜕2𝜙 𝜕2𝜙

+ + = 0 Laplace equation (.)
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑧2
j j
The Laplace equation forms the basis of potential theory. Besides in fluid mechanics, Laplace equation
potential theory is used in other fields of physics, like gravity, electrodynamics, and
magnetism. In mathematics, potential theory and the Laplace equation are closely
linked to harmonic functions. We will get back to this when we discuss wave theories.
Its favorable properties make the Laplace equation so important. The Laplace equation Properties of
is Laplace equation

(i) a partial differential equation (PDE)

(ii) of second order (derivatives of second order),
(iii) homogeneous (because the right-hand side is zero), and
(iv) linear.

The linearity (iv) of the Laplace equation is its most consequential property. Do not get
distracted by the ‘exponents’ of 2. They indicate second order derivatives and not that
the terms are multiplied by themselves. Note that
( )2
𝜕2𝜙 𝜕𝜙
≠
𝜕𝑥2 𝜕𝑥
Any linear combination of two or more solutions of a linear equation is again a solution Laplace equation
of the equation. Specifically, if 𝜙1 and 𝜙2 are solutions of the Laplace equation (.), is linear
i.e.
𝚫𝜙1 = 0 and 𝚫𝜙2 = 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

186 15 Potential Flow

then the function 𝜙 formed by a linear combination of 𝜙1 and 𝜙2

𝜙 = 𝑎1 𝜙1 + 𝑎2 𝜙2

will also satisfy the Laplace equation. The finite valued constants 𝑎1 and 𝑎2 have to be
independent of the spatial coordinates 𝑥, 𝑦, and 𝑧.
( )
𝚫𝜙 = 𝚫 𝑎1 𝜙1 + 𝑎2 𝜙2 = 𝑎1 𝚫𝜙1 + 𝑎2 𝚫𝜙2 = 0 (.)

Laplace equation The Laplace equation is also homogeneous (iii). A homogeneous equation has an
is homogeneous infinite number of solutions. In order to find the velocity potential for a particular
flow problem, we will need boundary conditions in addition to the Laplace equation.
Examples will be presented in Chapters  and .
Polar In many practical applications the Laplace equation is stated in polar, cylindrical, or
coordinates spherical coordinates. Planar flows make use of polar coordinates with the transforma-
tions
√
(𝑥, 𝑦) ⟶ (𝑟, 𝜑) with 𝑥 = 𝑟 cos(𝜑), 𝑦 = 𝑟 sin(𝜑) and 𝑥2 + 𝑦2 = 𝑟 (.)

Laplace equation The variable 𝑟 measures distance from the origin, and 𝜑 measures the angle between
in polar the vector (𝑥, 𝑦)𝑇 and the positive 𝑥-axis. Performing the variable substitution in the
coordinates
Laplace equation results in
j j
𝜕2𝜙 1 𝜕𝜙 1 𝜕2𝜙
+ + = 0 (.)
𝜕𝑟2 𝑟 𝜕𝑟 𝑟2 𝜕𝜑2

The velocity vector in polar coordinates is derived from the gradient in polar coordi-
nates.
⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑟 ⎟
𝑣(𝑟, 𝜑) = ⎜ (.)
𝜕𝜙 ⎟
⎜ 1 ⎟
⎝ 𝑟 𝜕𝜑 ⎠
Note that the components of 𝑣(𝑟, 𝜑) are not pointing in 𝑥- and 𝑦-direction. The first
component points radially away from the origin, and the second component is normal
to the radial.
For axially symmetric three-dimensional flows, or flows that rotate around a common
axis, cylindrical coordinates might be helpful. The transformation of Cartesian coordi-
nates into cylindrical coordinates is accomplished by adding the longitudinal 𝑧-axis to
the planar polar coordinates:

(𝑥, 𝑦, 𝑧) ⟶ (𝑟, 𝜑, 𝑧)
(.)
with 𝑥 = 𝑟 cos(𝜑), 𝑦 = 𝑟 sin(𝜑) and 𝑧 = 𝑧

Laplace equation The Laplace equation becomes

in cylindrical
coordinates
𝜕2𝜙 1 𝜕𝜙 1 𝜕2𝜙 𝜕2𝜙
+ + + = 0 (.)
𝜕𝑟2 𝑟 𝜕𝑟 𝑟2 𝜕𝜑2 𝜕𝑧2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.4 Bernoulli Equation for Potential Flow 187

The corresponding velocity vector in cylindrical coordinates is

⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑟 ⎟
⎜ ⎟
1 𝜕𝜙
𝑣(𝑟, 𝜑, 𝑧) = ⎜ ⎟ (.)
⎜ 𝑟 𝜕𝜑 ⎟
⎜ 𝜕𝜙 ⎟
⎜ ⎟
⎝ 𝜕𝑧 ⎠

For point symmetric spatial flows, spherical coordinates may be used.

(𝑥, 𝑦, 𝑧) ⟶ (𝑟, 𝜑, 𝜃)
(.)
with 𝑥 = 𝑟 sin(𝜃) cos(𝜑), 𝑦 = 𝑟 sin(𝜃) cos(𝜑) and 𝑧 = 𝑟 cos(𝜃)

The first angle 𝜑 (azimuth) describes the longitude with 𝜑 ∈ [0, 2𝜋] and is measured
with respect to the positive 𝑥-axis. The second angle defines the latitude and is measured
from the positive 𝑧-axis: 𝜃 ∈ [0, 𝜋] with 𝜃 = 0 at the North pole, 𝜃 = 𝜋∕2 at the equator,
and 𝜃 = 𝜋 at the South pole of the sphere.
The Laplace equation now reads Laplace equation
in spherical
coordinates
𝜕2𝜙 2 𝜕𝜙 1 𝜕2𝜙 1 𝜕2𝜙 1 cos(𝜃) 𝜕𝜙
+ + + + = 0 (.)
𝜕𝑟 2 𝑟 𝜕𝑟 2 2 2 2 𝑟2 sin(𝜃) 𝜕𝜃
𝑟 sin (𝜃) 𝜕𝜑
2 𝑟 𝜕𝜃
j j
The corresponding velocity vector in spherical coordinates is

⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑟 ⎟
⎜ ⎟
1 𝜕𝜙
𝑣(𝑟, 𝜑, 𝜃) = ⎜ ⎟ (.)
⎜ 𝑟 sin(𝜃) 𝜕𝜑 ⎟
⎜ 1 𝜕𝜙 ⎟
⎜ ⎟
⎝ 𝑟 𝜕𝜃 ⎠

The Laplace equation is derived from the continuity equation for incompressible fluids. Laplace eq.
For irrotational flow, the velocity vector is replaced by the gradient of the velocity represents
conservation of
potential. In conclusion, the Laplace equation represents the conservation of mass mass
principle for incompressible, inviscid, and irrotational flow. We did not specifically use
the assumption of inviscid fluid (𝜈 = 0) in its derivation. However, a viscous fluid is
usually never really irrotational.
If the Laplace equation replaces the continuity equation, what happens to the conserva-
tion of momentum equations? We will discuss this next.

15.4 Bernoulli Equation for Potential Flow

In Section ., we integrated the Euler equation along a streamline and obtained the
important Bernoulli equation (.).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

188 15 Potential Flow

We revisit the alternative form of the Euler equation for incompressible flow.
𝜕𝑣 1 1
+ ∇ (𝑣𝑇 𝑣) − 𝑣 × (∇ × 𝑣) = 𝑓 − ∇ 𝑝 (.)
𝜕𝑡 2 𝜌
For potential flow the last term on the left-hand side of (.) will vanish:

with rot 𝑣 = ∇ × 𝑣 = 0 follows 𝑣 × (∇ × 𝑣) = 0

Only In addition, we will allow only conservative volume forces. The only volume force
conservative we usually consider is the gravity force. We already stated that the weight force is
force fields
allowed
conservative and its potential is given on page .: 𝐸 = −𝑔 𝑧 𝑚. The Euler equation is
expressed as force per unit mass. Therefore, the vector of weight force per unit mass 𝑓
may be replaced with the gradient of its potential per unit mass.

𝑓 = ∇ (−𝑔𝑧) (.)

We substitute the new expression for the body forces 𝑓 and replace the velocity vector
with the gradient of the velocity potential.
𝜕∇ 𝜙 [( )𝑇 ]
1 1
+ ∇ ∇ 𝜙 ∇ 𝜙 = ∇ (−𝑔𝑧) − ∇ 𝑝 (.)
𝜕𝑡 2 𝜌
Assuming that the velocity potential has continuous second order derivatives, we may
j interchange the time derivative with the Nabla operator (taking spatial derivatives). j
This is also known as Schwartz’s or Clairaut’s theorem (Rogawski, ).
𝜕𝜙 [( )𝑇 ]
1 1
∇ + ∇ ∇ 𝜙 ∇ 𝜙 = ∇ (−𝑔𝑧) − ∇ 𝑝 (.)
𝜕𝑡 2 𝜌

We notice that every term in Equation (.) involves a gradient. Since a sum of
derivatives is equal to the derivative of the sum, we shuffle all terms to the left-hand
side and extract the Nabla operator.
[ ]
𝜕𝜙 1( )𝑇 𝑝
∇ + ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 0 (.)
𝜕𝑡 2 𝜌
The term in brackets has to be constant and independent from 𝑥, 𝑦, and 𝑧 for the spatial
Bernoulli derivatives to be zero everywhere. The constant may be a function of time, though.
equation for Finally, we have found the Bernoulli equation for potential flow (incompressible, inviscid,
potential flow
and irrotational fluid).
𝜕𝜙 1( )𝑇 𝑝
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 𝐶(𝑡) (.)
𝜕𝑡 2 𝜌
The major difference to the Bernoulli equation for general incompressible, inviscid
flow (.) is that the constant 𝐶 is now valid for the whole domain and not just for
a single streamline. This gives us great freedom to adjust the constant according to
practical considerations.
The main application of the Bernoulli equation is to compute pressure after the velocity
potential, and hence the fluid velocity vector, are known. If we choose a reference point

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

15.4 Bernoulli Equation for Potential Flow 189

far away from the body where the potential is steady 𝜕𝜙∞ ∕𝜕𝑡 = 0, velocity id equal to
𝑣)∞ , height is 𝑧∞ , and the pressure is 𝑝∞ , the constant becomes
1 𝑇 𝑝
𝐶 = 𝑣∞ 𝑣∞ + 𝑔𝑧∞ + ∞
2 𝜌
Remember that the dot product of a vector with itself yields its length (magnitude)
squared. Substituting 𝐶 into Bernoulli’s equation (.) yields
𝜕𝜙 1 𝑝 1 𝑝
+ |∇ 𝜙|2 + 𝑔𝑧 + = |𝑣∞ |2 + 𝑔𝑧∞ + ∞ (.)
𝜕𝑡 2 𝜌 2 𝜌
We can solve this for the change in pressure Pressure in
[ ] potential flow
𝜕𝜙 1
𝑝 − 𝑝∞ = −𝜌 + 𝜌 |𝑣∞ |2 − |∇ 𝜙|2 + 𝜌 𝑔(𝑧∞ − 𝑧) (.)
𝜕𝑡 2
Pressure forces are obtained by integration of the pressure difference (.) over the
body surface.
In steady flow all local time derivatives vanish and Equation (.) reduces to Pressure in
[ ] steady potential
1 flow
𝑝 − 𝑝∞ = 𝜌 |𝑣∞ |2 − |∇ 𝜙|2 + 𝜌 𝑔(𝑧∞ − 𝑧) (.)
2

We often assume in wave theory and seakeeping that the local velocity ∇ Φ does not Linearized
deviate much from the far field velocity 𝑣∞ . This allows us to neglect the nonlinear Bernoulli
j [ ] equation for j
term 12 𝜌 |𝑣∞ |2 − |∇ 𝜙|2 in Equation (.). For convenience, if the reference height unsteady flow
is put at the calm water level 𝑧∞ = 0, we obtain the linearized Bernoulli equation:
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
The 𝑧-axis is pointing upwards and all points in the fluid have negative 𝑧 values.
In summary of this section on potential flow, we see that the physical principles of Conservation of
conservation of mass and momentum are cast into the Laplace equation and Bernoulli mass and
momentum
equation for potential flow.
𝚫𝜙 = 0 Laplace eq. (.)
𝜕𝜙 1
𝜌 + 𝜌|∇ 𝜙|2 + 𝜌 𝑔𝑧 + 𝑝 = 𝐶(𝑡) Bernoulli eq. (.)
𝜕𝑡 2
The constant 𝐶(𝑡) is valid for the whole fluid domain.

References
Kellogg, O. (). Foundations of potential theory. Springer, Berlin, Germany.
Moran, J. (). An introduction to theoretical and computational aerodynamics. John
Wiley & Sons, New York, NY.
Rogawski, J. (). Multivariable calculus – early transcendentals. W.H. Freeman and
Company, New York, NY.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

190 15 Potential Flow

Self Study Problems

. Which principle of fluid mechanics is expressed by the Laplace equation, and
what are the necessary assumptions for applying the Laplace equation?
. State the general assumptions necessary to apply potential theory.
. How do you find the velocity vector once a scalar velocity potential is known?
. What are advantages of potential flow theory compared with solving the Euler
equations directly?
. Explain which assumptions are made to get from the Navier-Stokes equations to
the Bernoulli equation for potential flow.
. What is the key difference of the two forms of the Bernoulli equation stated in
Equations (.) and (.)?

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

191

16
Basic Solutions of the Laplace Equation

The Laplace equation forms the basis of potential theory.

𝚫𝜙 = 0 (.)

It is a linear, homogeneous, partial differential equation of second order. Because it is

homogeneous, it has an infinite number of solutions. Its linearity implies that linear
combinations of solutions form another solution to the Laplace equation (.). In fluid
mechanics, solutions to the Laplace equation represent velocity potentials. This chapter
introduces basic potentials which are often used as building blocks for potentials which
describe more complicated flow patterns. Besides the parallel flow potential, we discuss
singularities like source, vortex, and dipole, as well as combinations of these and their
resulting flow fields.
j j

Learning Objectives
At the end of this chapter students will be able to:

• discuss basic building blocks of potential flows

• derive velocity fields from potentials
• superimpose singularities to represent more complex flows

16.1 Uniform Parallel Flow

The most basic flow pattern is a uniform flow field, which features the same velocity Uniform parallel
vector at every point of the domain. This flow is known as parallel flow. For example, flow
the velocity vector of a flow of constant magnitude 𝑈∞ in negative 𝑥-direction is given
as
⎛ 𝑈∞ ⎞
𝑣(𝑥, 𝑦, 𝑧) = ∇ 𝜙∞ = −∇ (𝑈∞ 𝑥) = − ⎜ 0 ⎟ (.)
⎜ ⎟
⎝ 0 ⎠
The corresponding velocity potential is the scalar function

𝜙∞ (𝑥, 𝑦, 𝑧) = −𝑈∞ 𝑥 (.)

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

192 16 Basic Solutions of the Laplace Equation

Figure 16.1 Planar uniform flow at angle 𝛼

Sometimes we desire uniform flow at a given angle 𝛼 with respect to the coordinate
system. The velocity potential
[ ]
𝜙∞ (𝑥, 𝑦, 𝑧) = −𝑈∞ 𝑥 cos(𝛼) − 𝑦 sin(𝛼) (.)

describes a planar flow rotated clockwise by 𝛼 (Figure .). Its gradient yields the
following velocity vector:

⎛ 𝑢 ⎞ ⎛ −𝑈∞ cos(𝛼) ⎞
⎜ ⎟ ⎜ ⎟
𝑣(𝑥, 𝑦, 𝑧) = ∇ 𝜙∞ = ⎜ 𝑣 ⎟ = ⎜ 𝑈∞ sin(𝛼) ⎟ (.)
j ⎜ ⎟ ⎜ ⎟ j
⎝𝑤⎠ ⎝ 0 ⎠
The speed is as before |𝑣| = 𝑈∞ .
Obviously, all second order space derivatives of the parallel flow potential 𝜙∞ vanish.
Therefore, it satisfies the Laplace equation.

𝜕 2 𝜙∞ 𝜕 2 𝜙∞ 𝜕 2 𝜙∞
+ + = 0 (.)
𝜕𝑥2 𝜕𝑦2 𝜕𝑧2

16.2 Sources and Sinks

Source A source is a point in space, or on a plane, from which a volume flow emanates. Picture
the water flowing out from a natural spring or a garden stone with a hole drilled through
fed by a pond pump. Fluid will flow equally in every direction if the surrounding
landscape is radially symmetric. The thick, solid lines in Figure . represent such a
flow pattern.
2D source The corresponding velocity potential in two dimensions is
potential
𝜎
𝜙𝑠 (𝑝, 𝑞) = ln(|𝑝 − 𝑞|) (.)
2𝜋
The source strength 𝜎 controls how much volume is discharged by the source per unit
time. It has the unit cubic meter per second in space and meter squared per second in

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.2 Sources and Sinks 193

y
4

2
p
0.10

05
-0. q
0
4 2 -0. 0 2 4 x
0.0

10
0

0.05
0.
15

0.2
0

5
velocity potential 4 0.2
stream function

j Figure 16.2 Streamlines (𝜓 = const.) and isolines of velocity potential for a planar source/sink j
flow. 𝑞 𝑇 = (𝜉, 𝜂) is the location of the source

a planar flow. The source turns into a sink if the strength 𝜎 becomes negative which
means that a sink removes fluid from the domain. Lines of constant velocity potential
form concentric circles around the source (see lines in Figure .). A proof that
the velocity potential in Equation (.) satisfies the Laplace equation is left as a self
study problem.
The point 𝑞 in Equation (.) fixes the location of the source. We call it a source point. Source and field
point
( ) ⎛𝜉 ⎞
𝜉
D flows 𝑞 = D flows 𝑞 = ⎜ 𝜂 ⎟ (.)
𝜂 ⎜ ⎟
⎝𝜁 ⎠

In Figure ., the source point is identical to the origin 𝑞 𝑇 = (0, 0). The vector 𝑝 marks
the point where we inquire about flow properties. We call it field or collocation point.
( ) ⎛𝑥⎞
𝑥
D flows 𝑝 = D flows 𝑝 = ⎜ 𝑦 ⎟ (.)
𝑦 ⎜ ⎟
⎝𝑧⎠

We will make use of the distance 𝑟𝑝𝑞 between source point and field point in a number
of cases. For two-dimensional problems 𝑟𝑝𝑞 is
√
𝑟𝑝𝑞 = |𝑝 − 𝑞| = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 in D (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

194 16 Basic Solutions of the Laplace Equation

y6
er
e p

pq
r 0.1
0

2 05
-0. q

0.0
-0.1
0
0

0 0.05
x
0.

2 0 2 4 6
15

0.2
0

2
5
velocity potential 0.2
stream function

j Figure 16.3 Streamlines (𝜓 = const.) and equipotential lines for a planar source/sink flow; the j
source/sink is located at (𝜉, 𝜂) = (2.5, 1.3)𝑇

In three-dimensional problems use

√
𝑟𝑝𝑞 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2 in D (.)

Velocity field The velocity vector of the two-dimensional source flow in Cartesian coordinates is

⎛ (𝑥 − 𝜉) ⎞
)] ⎜ [ ( ⎟
𝜎 𝜎 ⎜ 𝑟2𝑝𝑞 ⎟
𝑣𝑠 (𝑝) = ∇ 𝜙𝑠 = ∇ ln |𝑝 − 𝑞| = (.)
2𝜋 2𝜋 ⎜ (𝑦 − 𝜂) ⎟
⎜ 𝑟2 ⎟
⎝ 𝑝𝑞 ⎠
The velocity vector is tangent to the streamlines. As shown in Figures . and .,
streamlines for source flows are radials emanating from the source in a regular pattern.
Fluid is flowing away from the source point if 𝜎 is positive and it is flowing toward the
source point if 𝜎 < 0. In Figure ., the source point 𝑞 has been moved off the origin
of the coordinate system. The resulting formulas for source potential (.) and velocity
field (.) are useful for the assembly of more complicated flow patterns.
Singularity Note that the velocity vector does not exist if field point and source point are identical,
i.e. 𝑝 = 𝑞. The flow velocity becomes infinite as we approach the source:

|𝑣𝑠 | ⟶ ∞ for 𝑝 ⟶ 𝑞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.2 Sources and Sinks 195

Because the Function (.) is singular for 𝑝 = 𝑞, we call the source/sink a singularity.
Later we will discuss other basic solutions of the Laplace equation which classify as
singularities. Strictly speaking, a singularity does not satisfy the Laplace equation at
the source point. Therefore, equations involving singularities always have to be treated
with care during integration and differentiation.
Sometimes it is more convenient to describe a two-dimensional flow in polar coordi- 2D source in
nates. The source potential and its gradient are given by polar
coordinates
(√ )
𝜎 ( )2 ( )2
𝜙𝑠 (𝑟, 𝜑) = ln 𝑟 cos(𝜑) − 𝜉 + 𝑟 sin(𝜑) − 𝜂 (.)
2𝜋
Here, the polar coordinate 𝑟 is measured from the origin and not from the position of
the source (Figure .). It is only equal to the distance between source and field point
if the source point is at the origin.

𝑣𝑠 (𝑟, 𝜑) = ∇ 𝜙𝑠 (𝑟, 𝜑)
[ (√ )]
⎛ 𝜕 𝜎 ( )2 ( )2 ⎞
⎜ 𝜕𝑟 2𝜋 ln 𝑟 cos(𝜑) − 𝜉 + 𝑟 sin(𝜑) − 𝜂 ⎟
= ⎜ ⎟
⎜ 1 𝜕 [𝜎 (√
( )2 ( )2
)] ⎟
⎜ ln 𝑟 cos(𝜑) − 𝜉 + 𝑟 sin(𝜑) − 𝜂 ⎟
⎝ 𝑟 𝜕𝜑 2𝜋 ⎠
[ ]
⎛ 𝑟 − 𝜉 cos(𝜑) − 𝜂 sin(𝜑) ⎞
⎜ ( )2 ( )2 ⎟
j 𝜎 ⎜ 𝑟 cos(𝜑) − 𝜉 + 𝑟 sin(𝜑) − 𝜂 ⎟ j
[ ] (.)
2𝜋 ⎜⎜ ⎟
=
𝜉 sin(𝜑) − 𝜂 cos(𝜑) ⎟
⎜ ( )2 ( )2 ⎟
⎝ 𝑟 cos(𝜑) − 𝜉 + 𝑟 sin(𝜑) − 𝜂 ⎠
Always remember that the components of the velocity vector in polar coordinates
generally do not point in 𝑥- and 𝑦-direction. They point in radial 𝑒𝑟 and tangential 𝑒𝜑
direction (see point 𝑝 in Figure .).
The velocity potential for a source in three dimensions is slightly different. It is also 3D source
known as a Rankine source, named after the Scottish engineer William J.M. Rankine potential
(*–†).
−𝜎
𝜙𝑠 (𝑝, 𝑞) = with 𝑟𝑝𝑞 = |𝑝 − 𝑞| (.)
4 𝜋 𝑟𝑝𝑞
The factor −1∕(4𝜋) ensures that the flow is directed away from the source point 𝑞 for
positive source strength 𝜎 and the volume flow is equal to the value of 𝜎 in cubic meter
per second.
Its velocity field consists of streamlines that emanate from 𝑞 equally in all directions. 3D source
velocity field
⎛ 𝜎 (𝑥 − 𝜉) ⎞
⎜ 4𝜋 𝑟3𝑝𝑞 ⎟
( ) ⎜ ⎟
−𝜎 ⎜ 𝜎 (𝑦 − 𝜂) ⎟
𝑣𝑠 (𝑝) = ∇ = ⎜ 4𝜋 𝑟3𝑝𝑞 ⎟ (.)
4 𝜋 𝑟𝑝𝑞 ⎜ ⎟
⎜ 𝜎 (𝑧 − 𝜁) ⎟
⎜ 4𝜋 𝑟3𝑝𝑞 ⎟
⎝ ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

196 16 Basic Solutions of the Laplace Equation

y6

/2
2/3

/3
4
5/6
/6

p
2
q
0

0
2 0 2 4 6 x
11
/6
7/6

2
5/3
4/3

velocity potential
3/2

stream function

j Figure 16.4 Streamlines (𝜓 = const.) and equipotential lines for a planar vortex flow; the vortex j
is located at 𝑞 = (2.5, 1.3)𝑇

Distributions of Rankine sources over body surfaces play an important role in mod-
ern numerical methods for displacement flow and wave resistance calculations (see
Section .).

16.3 Vortex
2D vortex Two-dimensional vortex flow was introduced as an example of potential flow in Fig-
potential ure .(b). At first glance, streamlines and equipotential lines of the vortex in Fig-
ure . look like the corresponding illustration of the source flow in Figure .. How-
ever, streamlines and equipotential lines are interchanged. For the vortex, streamlines
form concentric circles whereas for the source, streamlines are radials. Fluid is rotating
around the vortex center. The narrowing of streamlines indicates that the flow velocity
is increasing rapidly with decreasing distance from the center 𝑞.
Cartesian In Cartesian coordinates, the potential of a D vortex located at the source point
coordinates 𝑞 𝑇 = (𝜉, 𝜂) is given by
( )
𝛾 𝑦−𝜂
𝜙𝑣 (𝑥, 𝑦) = arctan (.)
2𝜋 𝑥−𝜉
 It is common practice to call the location of any singularity ‘source point.’

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.3 Vortex 197

Its velocity field is provided by the gradient of 𝜙𝑣 . 2D vortex

velocity field
⎛ [ −(𝑦 − 𝜂) ⎞
⎜ ] ⎟
𝛾 ⎜ (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 ⎟
𝑣𝑣 = ∇ 𝜙𝑣 = (.)
2𝜋 ⎜ (𝑥 − 𝜉) ⎟
⎜ [ ] ⎟
⎝ (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 ⎠
Again, we have a singular point at 𝑞 = 𝑝 at which the velocity would be infinite.
The velocity potential of a vortex at the origin is given by the function Polar
coordinates
𝛾
𝜙𝑣 (𝑟, 𝜑) = 𝜑 for 𝑞 = (0, 0)𝑇 (.)
2𝜋
As expected, the corresponding velocity vector has no component in radial direction
and the tangential velocity declines with increasing distance 𝑟 from the vortex.

⎛ 0 ⎞
𝛾 ⎜ ⎟
𝑣𝑣 (𝑟, 𝜑) = ∇ 𝜙𝑣 (𝑟, 𝜑) = for 𝑞 = (0, 0)𝑇 (.)
2𝜋 ⎜ 1 ⎟
⎝ 𝑟 ⎠

We compute the circulation Γ𝑣 for the vortex flow. We select a streamline located an 2D vortex
arbitrary distance 𝑟𝑝𝑞 = |𝑝 − 𝑞| away from the vortex as the integration path 𝐶. Since circulation
the velocity vector is always tangent to a streamline, the tangent vector of the circular
j integration path is equal to j
𝑣𝑣
𝑠 = (.)
|𝑣𝑣 |
Equation (.) allows us to compute the circulation of the vortex flow.
𝑣𝑣
Γ𝑣 = 𝑣𝑇𝑣 𝑠 d𝑠 = 𝑣𝑇𝑣 d𝑠 = |𝑣𝑣 | d𝑠 (.)
∮ ∮ |𝑣𝑣 | ∮
𝐶 𝐶 𝐶

The magnitude of velocity (.) is constant along each streamline and inversely pro-
portional to the distance 𝑟𝑝𝑞 .
√
√[ ]2
√
𝛾 √ − (𝑥 − 𝜉) + (𝑦 − 𝜂)2 𝛾
|𝑣𝑣 | = √ [ ] = (.)
2𝜋 2 2𝜋 𝑟𝑝𝑞
(𝑥 − 𝜉)2 + (𝑦 − 𝜂)2

This result is confirmed by Equation (.), which shows the velocity vector in polar
coordinates for a vortex at the origin. The magnitude of the tangential velocity is equal
to (.).
Substitution of this result into the integral of Equation (.) yields Circulation and
vortex strength
𝛾 𝛾
Γ𝑣 = d𝑠 = 2𝜋 𝑟𝑝𝑞 = 𝛾 (.)
2𝜋 𝑟𝑝𝑞 ∮ 2𝜋 𝑟𝑝𝑞
𝐶

The remaining path integral is equal to the circumference of a circle with radius 𝑟𝑝𝑞 .
Thus, the vortex strength 𝛾 is equal to the circulation of the vortex. In addition, the
circulation is constant and independent of the radius 𝑟𝑝𝑞 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

198 16 Basic Solutions of the Laplace Equation

Figure 16.5 Superposition of parallel flow and a

source/sink pair

At this point it is worthwhile to point out that the vortex strength and the circulation
are nonzero: 𝛾 = Γ𝑣 ≠ 0. This seems to violate the assumption of irrotational flow which
forms the basis of potential flow theory. As explained in Section ., it is sufficient
if the curl of the velocity vector vanishes in all simply connected regions. However,
any path that includes the vortex is multiply connected. The velocity field is singular
for 𝑝 = 𝑞. As a consequence, the vortex location 𝑞 cannot be a part of the domain. In
self-study problem #, you may show that the circulation vanishes for any path that
does not include the vortex itself.

16.4 Combinations of Singularities

j j
Before we discuss additional singularities, an example may show how they are used
in the description of more complex and more practical flow patterns. Exploiting the
linearity of the Laplace equation, new velocity potentials may be formed by linear
superposition of singularities.

16.4.1 Rankine oval

Superposition of A classic example is the Rankine oval (Rankine, ). Its flow pattern is described by a
parallel flow, velocity potential which consists of parallel flow, a source of strength +𝜎, and a sink of
source, and sink
strength −𝜎.
𝜎 √ 𝜎 √
𝜙 = −𝑈∞ 𝑥 + ln (𝑥 − 𝑎)2 + 𝑦2 − ln (𝑥 + 𝑎)2 + 𝑦2 (.)
2𝜋 2𝜋
As illustrated in Figure ., the source at 𝑥 = +𝑎 is facing the parallel flow which points
in negative 𝑥-axis direction. The sink is located at 𝑥 = −𝑎.
Differentiation of the potential with respect to 𝑥 (position of the field point) produces
the horizontal component of the velocity vector. This requires multiple applications of
the chain rule.
𝜕𝜙 ( ) 2(𝑥 − 𝑎)
𝜎 1 1
𝑢 = = −𝑈∞ + √ √
𝜕𝑥 2𝜋 (𝑥 − 𝑎)2 + 𝑦2 2 (𝑥 − 𝑎)2 + 𝑦2
( ) 2(𝑥 + 𝑎)
𝜎 1 1
− √ √
2𝜋 (𝑥 + 𝑎)2 + 𝑦2 2 (𝑥 + 𝑎)2 + 𝑦2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.4 Combinations of Singularities 199

𝜎 (𝑥 − 𝑎) 𝜎 (𝑥 + 𝑎)
𝑢 = −𝑈∞ + [ ] − [ ] (.)
2𝜋 (𝑥 − 𝑎)2 + 𝑦2 2𝜋 (𝑥 + 𝑎)2 + 𝑦2

The 𝑦-derivative of the potential yields the transverse velocity component:

𝜕𝜙 ( ) (2𝑦)
𝜎 1 1
𝑣 = = √ √
𝜕𝑦 2𝜋 (𝑥 − 𝑎)2 + 𝑦2 2 (𝑥 − 𝑎)2 + 𝑦2
( ) (2𝑦)
𝜎 1 1
− √ √
2𝜋 (𝑥 + 𝑎)2 + 𝑦2 2 (𝑥 + 𝑎)2 + 𝑦2
𝜎 (𝑦) 𝜎 (𝑦)
= [ ] − [ ] (.)
2𝜋 (𝑥 − 𝑎)2 + 𝑦2 2𝜋 (𝑥 + 𝑎)2 + 𝑦2

In summary, the velocity field is given by the vector

⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑥 ⎟
𝑣 = ∇𝜙 = ⎜ ⎟
𝜕𝜙
⎜ ⎟
⎝ 𝜕𝑦 ⎠
⎛ 𝜎 (𝑥 − 𝑎) 𝜎 (𝑥 + 𝑎) ⎞
⎜ −𝑈∞ + 2𝜋 [ 2 2
] − [
2𝜋 (𝑥 + 𝑎)2 + 𝑦2
] ⎟
⎜ (𝑥 − 𝑎) + 𝑦 ⎟
= (𝑦) (𝑦) (.)
⎜ 𝜎 𝜎 ⎟
⎜ [ ] − [ ] ⎟
⎝ 2𝜋 2
(𝑥 − 𝑎) + 𝑦 2 2𝜋 (𝑥 + 𝑎)2 + 𝑦2 ⎠
j j
Figure . shows an example of the flow field generated by the superposition of parallel Rankine oval
flow in negative 𝑥-axis direction, a source of strength 𝜎, and a sink of strength −𝜎.
Parallel, source, and sink flow are symmetric with respect to the 𝑥-axis. It is fair to
assume that the flow resulting from superposition is symmetric to the 𝑥-axis as well,
which is confirmed by the streamlines in Figure .. A striking feature of the flow is
the dividing streamline which splits the flow into two independent regions.

• interior flow – Fluid originating from the source is diverted toward the sink which
swallows fluid at the same rate as it emanates from the source.
• exterior flow – The parallel flow (from right to left in Figure . is diverted
around the interior flow between source and sink.

The dividing streamline encloses the interior flow. Since fluid particles never cross
streamlines, no fluid exchange happens between interior and exterior flow. As a conse-
quence, we may interpret the dividing streamline as a material body contour. The body
is known as a Rankine oval or Rankine ovoid.
We investigate the flow field and search for stagnation points, i.e. points at which
the fluid velocity vanishes |𝑣| = 0. In order to find the coordinates (𝑥𝑠 , 𝑦𝑠 )𝑇 of the
stagnation points we use the conditions 𝑢 = 0 and 𝑣 = 0. Starting with the latter, the
second component of Equation (.) yields the statement

𝜎 (𝑦𝑠 ) 𝜎 (𝑦𝑠 )
𝑣 = [ ] − [ ] = 0 (.)
2𝜋 (𝑥𝑠 − 𝑎) + 𝑦
2 2 2𝜋 (𝑥𝑠 + 𝑎)2 + 𝑦2
𝑠 𝑠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

200 16 Basic Solutions of the Laplace Equation

4
parallel flow velocity |U |=1.00m/s
source strength =6.28m 2 /s

2 dividing streamline

1
aft stagnation point
interior flow
[m]

0
+
y

forward stagnation point

1

exterior flow equipotential lines

j j
4
4 3 2 xs 1 0 1 +xs 2 3 4
x [m]

Figure 16.6 Flow field for a Rankine oval, a superposition of parallel flow, source, and sink

The equality is satisfied for 𝑎 = 0 or 𝑦𝑠 = 0. 𝑎 = 0 would put source and sink together
at the origin and they would cancel each other. Only the parallel flow would remain.
Therefore, stagnation points must be located on the 𝑥-axis with 𝑦𝑠 = 0. This corresponds
to the points where the dividing streamline splits and reunites.
We exploit 𝑦𝑠 = 0 and study the second condition:
𝜎 (𝑥𝑠 − 𝑎) 𝜎 (𝑥𝑠 + 𝑎)
𝑢 = 0 = −𝑈∞ + [ ] − [ ] (.)
2𝜋 (𝑥𝑠 − 𝑎)2 + 𝑦2 2𝜋 (𝑥𝑠 + 𝑎)2 + 𝑦2
𝑠 𝑠

With 𝑦𝑠 = 0 this simplifies to

𝜎 (𝑥𝑠 − 𝑎) 𝜎 (𝑥𝑠 + 𝑎)
0 = −𝑈∞ + [ ] − [ ]
2𝜋 (𝑥𝑠 − 𝑎) 2 2𝜋 (𝑥𝑠 + 𝑎)2
[ ]
𝜎 (𝑥𝑠 + 𝑎) − (𝑥𝑠 − 𝑎)
= −𝑈∞ +
2𝜋 (𝑥 − 𝑎)(𝑥𝑠 + 𝑎)
[ 𝑠 ]
𝜎 2𝑎
= −𝑈∞ + ( ) (.)
2𝜋 𝑥2𝑠 − 𝑎2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.4 Combinations of Singularities 201

Figure 16.7 Velocity and pressure distribution along the dividing streamline (Rankine oval,
𝑈∞ = 1.0 m/s, 𝜎 = 2𝜋 m2 /s)
j j

We solve this for the 𝑥-coordinates of the stagnation points:

( ) 𝜎𝑎
𝑈∞ 𝑥2𝑠 − 𝑎2 =
𝜋
2 𝜎𝑎
𝑥𝑠 = + 𝑎2
𝜋 𝑈∞
√
𝜎𝑎
𝑥𝑠 = ± + 𝑎2 (.)
𝜋 𝑈∞

For the values 𝑈∞ = 1.0 m/s, 𝜎 = 2𝜋 m2 /s, and 𝑎 = 1.0 m used in Figure ., the half
length of the Rankine oval is 𝑥𝑠 = 1.732 m.
Figure . illustrates velocity and pressure distribution along the dividing streamline. Velocity
Far in front (𝑥 ⟶ ∞) and far behind (𝑥 ⟶ −∞) the Rankine oval, the fluid flows in distribution
negative 𝑥-axis direction with velocity 𝑣𝑇 = (−𝑈∞ , 0). Obviously, the velocity vanishes
at the forward stagnation point (+𝑥𝑠 , 0). The onflow splits at the stagnation point.
While fluid is diverted up or down, it accelerates and reaches its maximum velocity
at the largest width of the body at 𝑥 = 0. For −𝑥𝑠 < 𝑥 < 0, fluid decelerates until it
reaches the aft stagnation point at −𝑥𝑠 . Behind the body fluid accelerates again until
the disturbance of the source and sink pair has declined.
The dimensionless pressure distribution follows from the Bernoulli equation for steady Pressure
flow, Equation (.). distribution
𝑝 − 𝑝∞ |𝑣|2
𝐶𝑝 = = 1− (.)
1 2 𝑈∞ 2
2
𝜌𝑈∞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

202 16 Basic Solutions of the Laplace Equation

Figure 16.8 Creation of a dipole (doublet) by

superposition of source and sink

Obviously, we ignored the hydrostatic pressure in this by assuming it is constant in the

plane of the flow.
The actual computation of 𝐶𝑝 is quite tedious and a computer program has been
employed to create the data in Figure .. First, the dividing streamline was located by
tracking a fluid particle over time through integration of the velocity vector. Next, the
pressure coefficient has been computed using the stored locations and velocity values.

16.4.2 Dipole
j j
Combination of Imagine a source of strength 𝜎 > 0 is placed at the origin 𝑞 (+) = (0, 0)𝑇 and a sink
source and sink
of strength −𝜎 is placed further to the left on the negative 𝑥-axis at 𝑞 (−) = (−ℎ, 0)𝑇 .
Figure . illustrates the setup.
The combined velocity potential of the source/sink arrangement is

−𝜎 √ 𝜎 √
𝜙𝑠𝑠 = ln (𝑥 + ℎ)2 + 𝑦2 + ln 𝑥2 + 𝑦2 (.)
2𝜋 2𝜋
If the distance ℎ is reduced to zero, source +𝜎 and sink −𝜎 will cancel each other, unless
we keep the product of source strength and distance constant. We define the new
parameter 𝜇 = 𝜎 ℎ > 0 and investigate the limit of the potential 𝜙𝑠𝑠 (.) for ℎ ⟶ 0.
[ ( √ √ )]
−𝜎 ℎ
lim 𝜙𝑠𝑠 = lim ln (𝑥 + ℎ)2 + 𝑦2 − ln 𝑥2 + 𝑦2
ℎ→0 ℎ→0 2𝜋 ℎ
[ √ √ ]
−𝜎 ℎ
= lim ln (𝑥 + ℎ)2 + 𝑦2 − ln 𝑥2 + 𝑦2
2𝜋 ℎ ℎ→0
[ √ √ ]
−𝜇 ln (𝑥 + ℎ)2 + 𝑦2 − ln 𝑥2 + 𝑦2
= lim (.)
2𝜋 ℎ→0 ℎ

The limit value of the term in square brackets looks like the definition of a first order
partial derivative with respect to 𝑥.

𝜕𝑓 𝑓 (𝑥 + ℎ) − 𝑓 (𝑥)
= lim
𝜕𝑥 ℎ→0 ℎ

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.4 Combinations of Singularities 203

√
In our case, the function is 𝑓 = ln 𝑥2 + 𝑦2 . Replacing the limit value in Equa-
tion (.) with the first order 𝑥-derivative results in a new potential which is known
as dipole or doublet potential.
−𝜇 𝜕 ( √ 2 )
𝜙𝑑 = ln 𝑥 + 𝑦2 (.)
2𝜋 𝜕𝑥
−𝜇 (𝑥)
= ( ) (.)
2𝜋 𝑥2 + 𝑦2

Essentially, the dipole potential is the negative derivative of the source potential. Only,
the source strength 𝜎 is exchanged for the dipole strength 𝜇.
The dipole potential above may also be expressed in polar coordinates. 2D dipole in
polar
−𝜇 cos(𝜑) coordinates
𝜙𝑑 (𝑟, 𝜑) = (.)
2𝜋 𝑟
This follows from Equation (.) with the substitutions from Equation (.): 𝑥 =
𝑟 cos(𝜑) and 𝑥2 + 𝑦2 = 𝑟2 .
The velocity field is defined in radial and tangential components as

⎛ cos(𝜑) ⎞
𝜇 ⎜⎜ 𝑟2 ⎟
⎟
𝑣𝑑 (𝑟, 𝜑) = ∇ 𝜙𝑑 (𝑟, 𝜑) = for 𝑞 = (0, 0)𝑇 (.)
2𝜋 ⎜ sin(𝜑) ⎟
j ⎜ ⎟ j
⎝ 𝑟2 ⎠

With the specific dipole strength of Cylinder flow

𝜇 = 2𝜋 𝑈∞ 𝑅2 (.)

the velocity potential (.) becomes identical to the disturbance potential (.),
which we will also find as part of the solution of a boundary value problem (see Chap-
ter ). The dipole potential, with axis in positive 𝑥-direction, combined with a parallel
flow in negative 𝑥-axis direction represents the flow around a long cylinder of radius 𝑅.
The velocity potential for the cylinder flow is equal to

𝑈∞ 𝑅2
𝜙cyl (𝑟, 𝜑) = − 𝑈∞ 𝑟 cos(𝜑) − cos(𝜑) (.)
𝑟

in polar coordinates. In Cartesian coordinates we may state it as

𝑥
𝜙cyl (𝑥, 𝑦) = − 𝑈∞ 𝑥 − 𝑈∞ 𝑅2 ( ) (.)
𝑥2 + 𝑦2

If the source point 𝑞 = (𝜉, 𝜂)𝑇 is moved away from the origin, the potential of a dipole 2D dipole
in Cartesian coordinates is equal to potential

𝜕𝜙𝑠 −𝜇 (𝑥 − 𝜉)
𝜙𝑑 = − = [ ] (.)
𝜕𝑥 2𝜋 (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

204 16 Basic Solutions of the Laplace Equation

Figure .(a) shows the corresponding pattern of streamlines and equipotential lines.
The curves form an orthogonal mesh, with sets of circles that are all tangent to two
perpendicular lines through the source point.
2D dipole As before, the components of the velocity vector follow from the gradient of the dipole
velocity field potential.
⎛ (𝑥 − 𝜉)2 − (𝑦 − 𝜂)2 ⎞
⎜ [ ]2 ⎟
𝜇 ⎜ (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 ⎟
(.)
2𝜋 ⎜⎜ ⎟
𝑣𝑑 = ∇ 𝜙𝑑 =
2(𝑥 − 𝜉)(𝑦 − 𝜂) ⎟
⎜ [(𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 ]2 ⎟
⎝ ⎠

Dipole axis Dipoles have a distinct axis. The dipole described by the potential in Equation (.) is
moving fluid through the source point in positive 𝑥-axis direction. The axis of the dipole
may be oriented in any direction 𝑠. The directional derivative of the source potential in
direction −𝑠 defines the corresponding dipole potential.

𝜕𝜙𝑠
𝜙𝑑 = − = −𝑠𝑇 ∇ 𝜙𝑠 (.)
𝜕𝑠

This equation is valid in two-dimensional and three-dimensional flow. For a directional

vector
( )
cos(𝛼)
j 𝑠 = j
sin(𝛼)

with |𝑠| = 1 we get

[ ]
−𝜇 (𝑥 − 𝜉) cos(𝛼) (𝑦 − 𝜂) sin(𝛼)
𝜙𝑑 = [ ] + [ ] (.)
2𝜋 (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2

The axis of the dipole depicted in Figure .(b) is rotated by 𝛼 = 65 degrees with
respect to the positive 𝑥-axis.

16.5 Singularity Distributions

Superposition of Cylinder and Rankine oval flow serve as examples of flows around simple geometries
discrete modeled by superposition of velocity potentials. More complex flow patterns may be
singularities
modeled by adding more singularities of varying strength. A closed body contour is
usually obtained if the total strength of sources and sinks adds up to zero.
Line and surface Rather than using discrete singularities, it is more efficient to distribute them along
distributions contours or boundary surfaces. Later we will use line distributions of sources and
vortices to model the flow past thin lifting foils. Sources and dipoles play an important
role in the numerical solution of potential flow problems. It can be shown that any
potential flow may be represented by distributions of sources and dipoles over the
boundaries of the fluid domain. For details see Katz and Plotkin ().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

16.5 Singularity Distributions 205

y6 y6

0.00
0.0
2
2 04
-0.
0 -0.
4 0.0 4
0

-0.0
4

-0
-0

.0
.02

8
p p

.12
-0
2 2
q -0 q
.08
0.12
0.08

-0.16
0.1 -0 2

2
.16 0.0

0.1
6
2

0.08
-0.1 0.
16

0 0
2 0
0.0
42 4 6 x 2 0 2 4 6 x

2 2
0.04
velocity potential velocity potential
stream function stream function

(a) Dipole axis parallel to 𝑥-axis (b) Dipole axis rotated by 65 degrees relative to +𝑥-axis

Figure 16.9 Streamlines (𝜓 = const.) and isolines of velocity potential for planar dipole flows. Dipole is located at
(𝜉, 𝜂) = (2.5, 1.3)𝑇

j j
References
Katz, J. and Plotkin, A. (). Low-speed aerodynamics. Cambridge Aerospace Series.
Cambridge University Press, New York, NY, second edition.
Rankine, W. (). Elementary demonstrations of principles relating to stream-lines.
In Millar, W., editor, Miscellaneous scientific papers, chapter XXXI., pages –.
Charles Griffin and Company, London. Originally published in The Engineer, Otc.
, .

Self Study Problems

. Show that the two-dimensional source potential
𝜎 √
𝜙𝑠 (𝑝, 𝑞) = ln(𝑟) with 𝑟 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2
2𝜋
satisfies the Laplace equation for 𝑝 ≠ 𝑞.

. Demonstrate that the three-dimensional source potential

−𝜎 √
𝜙𝑠 (𝑝, 𝑞) = with 𝑟 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2
4𝜋 𝑟
satisfies the Laplace equation for 𝑝 ≠ 𝑞.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

206 16 Basic Solutions of the Laplace Equation

. Employ Equations (.) and (.) to derive the velocity potential of a three-
dimensional dipole whose axis points in positive 𝑥-direction.
. A vortex flow is irrotational in any area which excludes the vortex itself. Show
that the circulation vanishes along an arbitrary path 𝐶 as shown in the figure
below.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

207

17
Ideal Flow Around A Long Cylinder

As an example of potential flow, we solve the basic equations that describe the fluid
flow around an infinitely long cylinder oriented with its axis perpendicular to the flow.
Since the flow properties will not change in the direction of the cylinder axis, we may
pick any plane transverse to the cylinder axis to set up the necessary equations.
Potential flow assumes that the flow is incompressible, inviscid, and irrotational. With
these restrictions, the equations to be solved are the Laplace equation (.) and the
Bernoulli equation (.). From the former follows the velocity potential Φ. Its gradient
provides the velocity vector 𝑣. The Bernoulli equation yields the pressure distribution.

𝚫Φ(𝑥, 𝑡) = 0 for all 𝑥 ∈ 𝑉 (.)

𝜕Φ 1
j 𝜌 + 𝜌|𝑣|2 + 𝑝 + 𝜌𝑔𝑧 = 𝐶(𝑡) for all 𝑥 ∈ 𝑉 (.) j
𝜕𝑡 2
We will see that the solution provides useful insight and forms the basis for the solution
of other flow problems, despite omitting viscous effects from the get go.

Learning Objectives
At the end of this chapter students will be able to:

• set up and solve a boundary value problem

• compute velocities, pressure, and forces for potential flow around a cylinder
• understand d’Alembert’s paradox
• understand the concept of added mass

17.1 Boundary Value Problem

As stated before, the Laplace equation (.) is a linear partial differential equation of Problem
second order. Since it is also homogeneous, it will have an infinite number of solutions. formulation
Similar to problems in elasticity theory where the deflection of a beam loaded with a
force or distributed load depends on the types of supports (clamped, pinned, slide, etc.),
we need boundary conditions to make the solution unique to the given problem.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

208 17 Ideal Flow Around A Long Cylinder

Figure 17.1 An infinitely long cylinder moving with speed 𝑈∞ in positive 𝑥-direction in a fluid at
rest

j Point of view The problem can be formulated in two different ways which ultimately are completely j
equivalent.

• We can build a mathematical model that reflects the view of an outside observer,
who will see an object moving through a fluid. The fluid is initially at rest and
moves only to make way for the cylinder.
• Or, we build a model representing the view of an observer moving with the object.
For the observer on the cylinder, the object is at rest and only the fluid moves.

We will formulate both below, but show the detailed solution only for the latter case. It
is much more convenient because the flow will be steady (invariant with time) in the
moving coordinate system. That simplifies the calculation of the velocity and pressure
distributions.

17.1.1 Moving cylinder in fluid at rest

Cylinder is Figure . shows a cylinder moving with velocity 𝑈∞ in positive 𝑥-direction. The fluid
moving is at rest except for the particles that have to move to make room for the cylinder. The
fixed finite control volume 𝑉 has two boundary surfaces: the body surface 𝑆𝑏 and a
surface 𝑆∞ in the far field. We assume the surface 𝑆∞ is far enough away from the body
that any disturbance due to the moving cylinder has subsided. Note that the normal
vector 𝑛 to the surfaces is pointing out of the fluid domain 𝑉 . As a consequence, the
normal vector of the body surface is pointing into the cylinder. The interior of the
cylinder is not part of the fluid domain.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.1 Boundary Value Problem 209

In order to properly pose the problem, we must define boundary conditions on all Types of
surfaces of the control volume. For boundary value problems three types of boundary boundary
conditions
conditions are considered:

. Dirichlet boundary conditions impose values on the unknown function itself. In

our case the unknown function is the velocity potential.

. Neumann boundary conditions prescribe the values of derivatives of the unknown

function, for instance the velocity components which are the spatial derivatives
of the velocity potential.

. Mixed boundary conditions which involve the unknown functions and their
derivatives simultaneously.

In order to obtain a unique solution, we must define boundary conditions on all surfaces Physical
of the control volume. There are two boundary surfaces, and we have to specify a boundary
conditions
boundary condition for each of them.

• As stated above, the disturbance due to the moving cylinder has subsided in the
far field. Therefore, fluid velocity (induced by the moving cylinder) has to vanish
in the far field.
j ∇Φ ⟶ 0 for 𝑟 ⟶ ∞ j

This is a Neumann boundary condition, since we impose a requirement for the

derivative of the unknown potential Φ.

• We want the fluid to stay out of the cylinder, i.e. no fluid particle is allowed to flow
across the body surface 𝑆𝑏 . In a viscous fluid, we expect the molecules to stick to
the surface. However, we ignore friction here, which allows molecules to move
tangentially along the surface. To prevent fluid from crossing the body surface,
we must require that the normal component of the relative velocity between fluid
and body vanishes on 𝑆𝑏 .

𝑛𝑇 𝑣relative = 0 for points on 𝑆𝑏

The relative velocity is the difference between fluid flow velocity 𝑣 = ∇ Φ and the
local velocity 𝑣𝑏 of the body surface.

In summary, at any given time 𝑡 we have to satisfy the following set of equations which Boundary value
make up a boundary value problem for unsteady flow. problem for
unsteady flow

𝚫Φ(𝑥, 𝑡) = 0 for 𝑥 ∈ 𝑉
𝜕Φ
− 𝑛𝑇 𝑣𝑏 = 0 for 𝑥 ∈ 𝑆𝑏 , for all 𝑡 (.)
𝜕𝑛
lim ∇ Φ(𝑥, 𝑡) = 0 for 𝑥 ∈ 𝑆∞
|𝑥|→∞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

210 17 Ideal Flow Around A Long Cylinder

Figure 17.2 An infinitely long cylinder at rest in parallel flow

17.1.2 Cylinder at rest in parallel flow

j Fluid is moving The boundary value problem (.) formulated above is useful if the cylinder is moving j
with varying forward speed. However, for a cylinder moving with a constant velocity it
is better to formulate the problem in a coordinate system which is moving along with
the cylinder. The flow becomes steady, we have to satisfy the following set of equations
which make up a boundary value problem.
Figure . shows the cylinder in a parallel flow which is directed in negative 𝑥-axis
direction with constant speed 𝑈∞ . The parallel flow speed corresponds to the forward
speed of the cylinder as seen by an outside observer.
Parallel flow The linearity of the Laplace equation allows us to combine solutions of it into new
potential solutions. The velocity potential of parallel flow in negative 𝑥-direction is

𝜙∞ = −𝑈∞ 𝑥 (.)

Its gradient is

⎛ 𝜕𝜙∞ ⎞
( )
⎜ 𝜕𝑥 ⎟ −𝑈∞
∇ 𝜙∞ = ⎜ ⎟ = (.)
⎜ 𝜕𝜙∞ ⎟ 0
⎜ ⎟
⎝ 𝜕𝑦 ⎠
Obviously, all second order derivatives of 𝜙∞ vanish, and the parallel flow velocity
potential satisfies the Laplace equation.
In polar coordinates the parallel stream velocity potential is

𝜙∞ (𝑟, 𝜑) = −𝑈∞ 𝑟 cos 𝜑 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.2 Solution and Velocity Potential 211

Introducing the parallel flow potential splits the yet unknown velocity potential 𝜙 of Superposition
the flow around a cylinder into two parts: the known parallel flow potential 𝜙∞ and the principle
yet unknown disturbance potential 𝜙𝑑 . The latter is so named because it represents the
disturbance of parallel flow caused by the cylinder.

𝜙 = 𝜙∞ + 𝜙𝑑 (.)

We now reformulate our boundary conditions. There are two boundaries: the far field Boundary
𝑆∞ and the body surface 𝑆𝑏 . conditions

• In the far field, we expect the disturbance caused by the cylinder to subside. A
submarine passing by far out at sea will not change flow patterns in the harbor.
Mathematically speaking, we request that ∇ 𝜙 → ∇ 𝜙∞ for 𝑟 → ∞, i.e. we have
parallel flow conditions in the far field. This is equivalent to
( )
0
∇ 𝜙𝑑 → for 𝑟 → ∞
0

• On the body surface 𝑆𝑏 , we now request that the component of the fluid particle
velocity normal to the body surface vanishes.

𝑛𝑇 ∇ 𝜙 = 0 for 𝑥 ∈ 𝑆𝑏

Remember that the body does not move relative to the chosen coordinate system.
j j
Again, this is a Neumann boundary condition. For the dot product of normal vec-
tor 𝑛 and gradient of the potential ∇ 𝜙, we usually write 𝜕𝜙∕𝜕𝑛 (normal derivative).

In summary, the boundary value problem for the cylinder in parallel flow is defined by Boundary value
problem for
𝚫𝜙(𝑥) = 0 for 𝑥 ∈ 𝑉 steady cylinder
flow
𝜕𝜙
= 0 for 𝑥 ∈ 𝑆𝑏 (.)
𝜕𝑛
lim ∇ 𝜙(𝑥) = ∇ 𝜙∞ (𝑥) for 𝑥 ∈ 𝑆∞
|𝑥|→∞

17.2 Solution and Velocity Potential

Solution of the boundary value problem (.) is easier if it is done in polar coordinates Solution by
< 𝑟, 𝜑 > rather than Cartesian coordinates < 𝑥, 𝑦 >. The Laplace equation in polar separation of
variables
coordinates is provided by Equation (.) and repeated here for convenience.

𝜕2𝜙 1 𝜕𝜙 1 𝜕2𝜙
+ + = 0 (.)
𝜕𝑟2 𝑟 𝜕𝑟 𝑟2 𝜕𝜑2
The following equations transform coordinates from a polar into a Cartesian coordinate
system:

𝑥 = 𝑟 cos(𝜑) 𝑦 = 𝑟 sin(𝜑) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

212 17 Ideal Flow Around A Long Cylinder

The reverse transformation is given by

√ (𝑦)
𝑟 = 𝑥2 + 𝑦2 𝜑 = arctan (.)
𝑥

The unknown disturbance potential will be a function of (𝑟, 𝜑). We seek a solution by
separation of variables:
𝜙𝑑 (𝑟, 𝜑) = 𝑃 (𝑟) 𝑄(𝜑) (.)
Instead of 𝜙𝑑 (𝑟, 𝜑), we have to find two functions 𝑃 (𝑟) and 𝑄(𝜑). 𝑃 and 𝑄 are univariate
functions.
We substitute 𝜙 = 𝜙∞ + 𝑃 𝑄 into the Laplace equation (.). Because of the super-
position principle and the fact that 𝜙∞ satisfies the Laplace equation, we need only to
investigate the product 𝑃 𝑄. Any derivative with respect to 𝑟 affects function 𝑃 only
while 𝑄 is a constant factor. Derivatives with respect to 𝜑 affect𝑄 only, with 𝑃 being
treated as a constant factor.

𝜕2𝑃 1 𝜕𝑃 1 𝜕2𝑄
𝑄+ 𝑄+ 𝑃 = 0 (.)
𝜕𝑟2 𝑟 𝜕𝑟 𝑟2 𝜕𝜑2

Dividing the equation by the product 𝑃 𝑄 yields

𝜕2𝑃 1 1 𝜕𝑃 1 1 𝜕2𝑄 1
+ + = 0 (.)
j 𝜕𝑟2 𝑃 𝑟 𝜕𝑟 𝑃 𝑟2 𝜕𝜑2 𝑄 j
Of course, we are excluding the trivial solution 𝑃 𝑄 ≡ 0.
Multiplication with 𝑟2 separates the equation into terms which depend on either 𝑟 or 𝜑
but not both.
𝜕2𝑃 1 𝜕𝑃 1 𝜕2𝑄 1
𝑟2 +𝑟 + = 0 (.)
𝜕𝑟2 𝑃 𝜕𝑟 𝑃 𝜕𝜑2 𝑄

We move the last term to the right-hand side:

𝜕2𝑃 1 𝜕𝑃 1 𝜕2𝑄 1
𝑟2 +𝑟 = − = 𝐶 (.)
2
𝜕𝑟 𝑃 𝜕𝑟 𝑃 𝜕𝜑2 𝑄
For this equation to be true for all possible combinations of 𝑟 and 𝜑, both sides have to
be equal to a constant 𝐶, otherwise there would be points where the difference between
left and right side would not result in zero. 𝐶 must be independent of the variables
(𝑟, 𝜑).
PDE converted to The partial differential equation (.) is now split into two ordinary differential equa-
system of ODEs tions (ODEs). Using subscripts to indicate differentiation we get:

𝑟2 𝑃𝑟𝑟 + 𝑟𝑃𝑟 − 𝑃 𝐶 = 0 (.)

𝑄𝜑𝜑 + 𝑄 𝐶 = 0 (.)

Note that the constant 𝐶 must be the same in both equations. It represents the coupling
between the ODEs. The differential equations (.) and (.) are of standard shape.
Solutions can be found in any book on differential equations.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.2 Solution and Velocity Potential 213

We solve for the function 𝑃 first and assume that it takes the form of a reciprocal Basic solution for
function. 𝑃 (𝑟)

1
𝑃 = 𝐶1
𝑟
𝜕𝑃 1
𝑃𝑟 = 𝐶1 = −𝐶1 (.)
𝜕𝑟 𝑟2
𝜕2𝑃 1
𝑃𝑟𝑟 = 𝐶1 = 2𝐶1
𝜕𝑟2 𝑟3

Substituting this into equation (.) yields the desired result.

1 1 1
0 = 𝑟2 2𝐶1 − 𝑟 𝐶1 − 𝐶1 𝐶
𝑟3 𝑟2 𝑟
𝐶1 ( )
0 = 1−𝐶 (.)
𝑟
The last equation is satisfied for 𝐶 = 1. The constant 𝐶1 remains to be determined.
With 𝐶 = 1 substituted into (.), we need a function 𝑄 that is the negative of its Basic solution for
second order derivative. The sine and cosine functions are of this type. Because our 𝑄(𝜑)
flow is expected to be symmetric to the 𝑦-axis, we select the cosine function.

𝑄 = 𝐶2 cos 𝜑
j 𝑄𝜑𝜑 = −𝐶2 cos 𝜑 (.) j

Combining the basic solutions 𝑃 = 𝐶1 ∕𝑟 and 𝑄 = 𝐶2 cos 𝜑, creates a new constant

𝐶3 = 𝐶1 𝐶2 . We obtain for the disturbance potential 𝜙𝑑 :

1
𝜙𝑑 (𝑟, 𝜑) = 𝐶1 𝐶2 cos 𝜑
𝑟 (.)
1
= 𝐶3 cos 𝜑
𝑟

The boundary conditions are exploited to find the constant 𝐶3 . Let us start with the Identification of
body boundary condition. On the body surface 𝑆𝑏 we have to enforce that constant

𝜕𝜙
= 0 for 𝑟 = 𝑅 (.)
𝜕𝑛
The normal vector on the cylinder surface points always toward the center or, more
importantly, in the opposite direction of 𝑒𝑟 . Thus, the normal derivative is equal to the
negative radial derivative.

𝜕 𝜕
=− on 𝑆𝑏 with 𝑟 = 𝑅 (.)
𝜕𝑛 𝜕𝑟

Replacing the normal derivative with −𝜕∕𝜕𝑟

𝜕𝜙 |
− | = 0
𝜕𝑟 |𝑟=𝑅

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

214 17 Ideal Flow Around A Long Cylinder

and splitting the potential in parallel stream and disturbance part yields:

𝜕𝜙∞ | 𝜕𝜙 |
− | − 𝑑| = 0
𝜕𝑟 |𝑟=𝑅 𝜕𝑟 |𝑟=𝑅
𝜕𝜙 | 𝜕𝜙∞ |
− 𝑑| = |
|
𝜕𝑟 𝑟=𝑅 𝜕𝑟 |𝑟=𝑅
Finally, utilizing Equation (.) on the right-hand side and Equation (.) on the
left-hand side yields for the constant 𝐶3 :

𝜕𝜙 |
− | = −𝑈∞ cos 𝜑
𝜕𝑟 |𝑟=𝑅
cos 𝜑 |
𝐶3 | = −𝑈∞ cos 𝜑
𝑟2 |𝑟=𝑅
cos 𝜑
𝐶3 = −𝑈∞ cos 𝜑
𝑅2
𝐶3 = −𝑈∞ 𝑅2 (.)

The constant depends on the basic parameters defining the flow problem: the velocity
of the parallel stream and the radius 𝑅 of the cylinder, which makes sense. The unit of
the resulting potential is m2 /s, which is also correct.
Potential of Substituting the constant 𝐶3 into Equation (.) for the disturbance potential gives
disturbance
−𝑈∞ 𝑅2
j 𝜙𝑑 (𝑟, 𝜑) = cos 𝜑 (.) j
𝑟
This is a dipole or doublet potential, which we already encountered in the previous
chapter (see Equation (.)).
Velocity Finally, the velocity potential for the flow around a cylinder in an unbounded fluid is
potential for the given by the sum of parallel flow and a dipole with its axis pointing against the parallel
cylinder flow
flow.
𝑈 𝑅2
𝜙(𝑟, 𝜑) = −𝑈∞ 𝑟 cos 𝜑 − ∞ cos 𝜑 (.)
𝑟
In Cartesian coordinates the velocity potential for the cylinder flow is given by
𝑥
𝜙(𝑥, 𝑦) = −𝑈∞ 𝑥 − 𝑈∞ 𝑅2 ( ) (.)
𝑥2 + 𝑦2

Far field Hopefully, you have not forgotten about the boundary condition for the far field surface
boundary 𝑆∞ . We have not used it yet explicitly but, if you check, you can see that the gradient
condition
of the second term of the potential vanishes for 𝑟 ⟶ ∞ and the parallel flow part
remains. See also the velocity field discussed in the next section.

17.3 Velocity and Pressure Field

Properties of In the previous section we solved the boundary value problem for the cylinder in parallel
cylinder flow flow, and derived the velocity potential (.). Now we will study the resulting velocity
and pressure distributions.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.3 Velocity and Pressure Field 215

4
parallel flow velocity U =1.00m/s

stagnation points
y [m]

j j
4
4 3 2 1 0 1 2 3 4
x [m]

Figure 17.3 Streamlines and velocity field for a cylinder in parallel flow

17.3.1 Velocity field

The gradient of the potential represents the velocity vector for points (𝑟, 𝜑) in the fluid Velocity vector
domain. The gradient is stated with respect to the vector basis formed by the unit
vectors 𝑒𝑟 and 𝑒𝜑 .

⎛ 𝜕𝜙 ⎞ ⎛ 𝑈 𝑅2 ⎞
⎜ −𝑈∞ cos 𝜑 + ∞ cos 𝜑 ⎟
⎜ 𝜕𝑟 ⎟ 𝑟2
∇ 𝜙<𝑟,𝜑> = ⎜ ⎟ = ⎜ ⎟ (.)
⎜ 1 𝜕𝜙 ⎟ ⎜ 𝑈 𝑅2 ⎟
⎝ 𝑟 𝜕𝜑 ⎠<𝑟,𝜑> ⎜ 𝑈∞ sin 𝜑 + ∞ sin 𝜑 ⎟
⎝ 𝑟2 ⎠<𝑟,𝜑>

Figure . shows the streamlines and velocity arrows for the cylinder in parallel flow.
Since fluid cannot move across streamlines, the distance between streamlines may
serve as a gage for the magnitude of the velocity. Where streamlines widen, the flow
will slow down, and where they narrow, fluid will move faster. The length of the arrows
is proportional to the flow speed. Obviously, we have areas of slow flow speeds in front
of (𝑥 = +1 m, 𝑦 = 0) and behind the point (𝑥 = −1 m, 𝑦 = 0) on the cylinder. High
flow speeds occur at the upper and lower side of the cylinder. As expected, the flow is

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

216 17 Ideal Flow Around A Long Cylinder

completely symmetric to the 𝑥-axis. The minimum velocity is zero, and the maximum
velocity is apparently two times the onflow velocity.
Velocity vector Let us investigate the velocity distribution on the body surface itself. On 𝑆𝑏 the radius
is 𝑟 = 𝑅, and the angle takes values from the interval 𝜑 ∈ [0, 2𝜋]. Substituting 𝑟 = 𝑅
into Equation (.) yields:
( ) ( )
𝑣𝑟 0
𝑣(𝑟 = 𝑅) = = (.)
𝑣𝜑 <𝑟,𝜑> 2𝑈∞ sin 𝜑 <𝑟,𝜑>

This seems surprising at first but remember that the velocity vector is to the base
< 𝑒𝑟 , 𝑒𝜑 >. The vanishing radial velocity 𝑣𝑟 = 0 simply reflects the body boundary
condition from Equation (.).
The component 𝑣𝜑 = 2𝑈∞ sin 𝜑 is always tangent to the cylinder surface. Since the sine
is limited to the range −1 to +1, the maximum velocity of 2𝑈∞ occurs for 𝜑 = 90 degree
and 𝜑 = 270 degree which represent the points where the cylinder is widest with
respect to the onflow. Stagnation points with 𝑣 = 0 are found for 𝜑 = 0 degree and
𝜑 = 180 degree, i.e. at the front and back of the cylinder.

17.3.2 Pressure field

Pressure The pressure field is derived from the Bernoulli equation. Again, for inviscid, incom-
j j
distribution pressible, and irrotational flow it is
𝜕𝜙 1 ( )𝑇
𝜌 + 𝜌 ∇ 𝜙 ∇ 𝜙 + 𝑝 + 𝜌𝑔𝑧 = 𝐶(𝑡) (.)
𝜕𝑡 2
The equation has been derived from the conservation of momentum per unit volume
equation, essentially by integration in space. Consequently, the terms in the equation
represent the energy per volume unit of fluid. We consider the cylinder flow for a steady
flow speed. Therefore, the unsteady pressure term 𝜌𝜕𝜙∕𝜕𝑡 vanishes. Now the constant
𝐶 will be independent of location and also be independent of time. We are free to
choose a suitable value for 𝐶 because the potential is only unique up to a constant.
For the cylinder flow a suitable value is
1 2
𝐶 = 𝜌𝑈 + 𝑝∞ (.)
2 ∞
Since forces result from pressure differences, we form:
( ( )𝑇 )
1
𝑝 − 𝑝∞ = 𝜌 𝑈∞ 2
− ∇𝜙 ∇𝜙 − 𝜌𝑔𝑧 (.)
2 ⏟⏟⏟
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
hydrostatic pressure
hydrodynamic pressure

Hydrostatic We assume that the cylinder is neutrally buoyant. In that case, the resultant force of the
pressure hydrostatic pressure component 𝜌𝑔𝑧0 compensates the weight of the cylinder.
 For a vertical cylinder, we could claim that the static pressure is equal all around the circumference.

Hence, it has no resultant force.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.3 Velocity and Pressure Field 217

4
0.0

0.0
3

2 -0.5
-0.
8
0
-1.

1
-0.
0.1
-0.
-2.5

-1
1
1

.5

-0.3
0.1

0.5
0.8
y [m]

0
0.8
0.5
0.3 0.3
-0.3
-0.5

1 -1 0.1
.5 -2.5
0.1
1

0
-0.

-0. -1.
8
-0.1

3 0.0

j 0.0 j
4
4 3 2 1 0 1 2 3 4
x [m]

Figure 17.4 Contours of constant pressure coefficient 𝐶𝑝 for a cylinder in parallel flow

We divide Equation (.) by the dynamic pressure, which yields a dimensionless

pressure coefficient 𝐶𝑝 .

𝑝 − 𝑝∞ |𝑣|2
𝐶𝑝 = = 1− (.)
1 2 2
𝑈∞
2
𝜌𝑈∞

The pressure distribution clearly depends on the magnitude of the velocity |𝑣|. Since
the velocity magnitude is symmetric to both axes, the pressure distribution will be
symmetric as well (Figure .).
On the cylinder surface, we find with |𝑣|2 = 𝑣2𝜑 = 4𝑈∞
2 sin2 𝜑 that the pressure coeffi- Pressure on
cient is equal to cylinder surface

𝐶𝑝 = 1 − 4 sin2 𝜑 (.)

It ranges from +1 at the stagnation points (𝜑 = 0 and  degrees) to −3 at the shoulders
where the velocity is highest (𝜑 = 90 and  degrees).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

218 17 Ideal Flow Around A Long Cylinder

forward stagnation point aft stagnation point

1

[]
0
pressure coefficient Cp

0 /2 3/2 2
position angle [rad]

Figure 17.5 Pressure coefficient 𝐶𝑝 distribution on the cylinder surface (𝑟 = 𝑅) for a cylinder in
parallel flow

17.4 D’Alembert’s Paradox

j Pressure force We will now use the pressure distribution on the cylinder surface to compute the j
resultant pressure force. This problem is also known as d’Alembert’s paradox because,
in contrast to all practical experience, our computation will find a net zero force in all
directions.
A pressure force on a body is obtained by integrating the pressure over the body surface:
( )
𝐹𝑝𝑥
𝐹𝑝 = = 𝑝𝑛 d𝑆 (.)
𝐹𝑝𝑦 ∬
𝑆𝑏

This equation will show a minus sign in front of the integral when the normal vector
points into the fluid. In our case, a positive pressure (higher than some reference pres-
sure) will cause a force that points in positive normal direction. As mentioned before,
we are not interested in the hydrostatic force. We will only integrate the hydrodynamic
pressure component of Equation (.).
Surface element We assume our cylinder has the length 𝓁 in 𝑧-direction. A surface element of the
cylinder surface is then given as

d𝑆 = 𝓁𝑅 d𝜑 (.)

𝑅d𝜑 represents a small piece of the circular arc of the cylinder contour. Our double
integral over 𝑆𝑏 reduces to a single integral over the angle 𝜑 ∈ [0, 2𝜋] because the flow
is not changing in 𝑧-direction.
Pressure Substituting the hydrodynamic pressure yields for the pressure force
integration

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.5 Added Mass 219

( ) 2𝜋 ( )
𝐹𝑝𝑥 ( ) − cos(𝜑)
1 2
𝐹𝑝 = = 𝜌𝑈∞ 1 − 4 sin2 (𝜑) 𝓁𝑅d𝜑 (.)
𝐹𝑝𝑦 ∫ 2 − sin(𝜑)
0

Extracting all terms independent of the integration variable 𝜑 results in the following
expression:
2𝜋 ( )
1 2 − cos(𝜑) + 4 sin2 (𝜑) cos(𝜑)
𝐹𝑝 = 𝜌𝑈∞ 𝓁𝑅 d𝜑
2 ∫ − sin(𝜑) + 4 sin3 (𝜑)
0

After splitting the integral at the plus sign, we have

( ) 2𝜋 ( )
⎡ 2𝜋 − cos(𝜑) sin2 (𝜑) cos(𝜑) ⎤
1 2
𝐹𝑝 = 𝜌𝑈∞ 𝓁𝑅 ⎢ d𝜑 + 4 d𝜑⎥
2 ⎢∫ − sin(𝜑) ∫ sin3 (𝜑) ⎥
⎣0 0 ⎦

and with the antiderivatives from an integral table follows

[( )]2𝜋 [( )]2𝜋
⎛ − sin(𝜑)
1
sin3 (𝜑) ⎞
1 2
𝐹𝑝 = 𝜌𝑈∞ 𝓁𝑅 ⎜ +4 3 ⎟ (.)
2 ⎜ + cos(𝜑) cos(𝜑) − 1
cos3 (𝜑) ⎟
⎝ 0 3 0 ⎠
j j
Substituting the integral limits finally gives us d’Alembert’s
( ) ( ) paradox
𝐹𝑝𝑥 0
𝐹𝑝 = = (.)
𝐹𝑝𝑦 0

Both 𝑥- and 𝑦-components of the force vanish. Since the body is symmetric to the 𝑥-axis,
we expected no transverse force 𝐹𝑝𝑦 . However, 𝐹𝑝𝑥 = 0 clearly violates any practical
observation and experiment. If we place a cylinder in a stream, it will experience a drag
force, and we have to secure the cylinder or it will be washed downstream. The theory
leads to a contradiction with the experiment, a paradox. The conclusion that something
is wrong with the theory was drawn by Jean-Baptiste le Rond d’Alembert in  and
is therefore named d’Alembert’s paradox. Today we know that neglecting the viscous
forces (solving Euler equations instead of Navier-Stokes equations) is the cause for the
unrealistic pressure force prediction.

17.5 Added Mass

Of course, pressure forces do not vanish for all potential flows. The picture changes
already when we consider, for example, the unsteady potential flow around a cylinder
moving through the fluid at rest. Although this thematically belongs to seakeeping
and maneuvering, it is shown below that the resultant pressure force gives rise to the
so called added mass. It is a hydrodynamic pressure force proportional to the relative
acceleration between body and fluid.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

220 17 Ideal Flow Around A Long Cylinder

Potential of Solving the boundary value problem (.) of the cylinder moving through a fluid at
unsteady rest yields the velocity potential of a dipole.
cylinder flow
−𝑈 (𝑡)𝑅2
𝜙(𝑟, 𝜑, 𝑡) = cos(𝜑) (.)
𝑟
However, here the velocity 𝑈 (𝑡) of the cylinder (in 𝑥-direction) is considered a function
of time. The gradient of a potential provides the velocity vector for the fluid. With
respect to a polar coordinate vector base we have
2
⎛ 𝑈 (𝑡)𝑅 cos(𝜑) ⎞
⎜ 𝑟 2 ⎟
𝑣(𝑟, 𝜑, 𝑡) = ∇𝜙(𝑟, 𝜑, 𝑡) = ⎜ ⎟ (.)
𝑈 (𝑡)𝑅 2
⎜ sin(𝜑) ⎟
⎝ 𝑟2 ⎠

The magnitude of the velocity vector is

| 𝑈 (𝑡)𝑅2 |
| |
|𝑣| = | | (.)
| 𝑟2 |
| |

Unsteady The corresponding unsteady pressure distribution is obtained from the unsteady Bernoulli
pressure field equation (.).
𝜕𝜙 1 ( )𝑇
𝜌 + 𝜌 ∇ 𝜙 ∇ 𝜙 + 𝜌𝑔𝑧 + 𝑝 = 𝐶(𝑡) (.)
j 𝜕𝑡 2 j
Like before, we assume that resultant hydrostatic pressure force and weight are in
equilibrium. As a constant 𝐶(𝑡), we choose the constant pressure 𝑝∞ in the far field.
The remaining dynamic pressure acting on the cylinder surface will be
𝜕𝜙 1 ( )𝑇
𝑝dyn (𝑟, 𝜑, 𝑡) = 𝑝 − 𝑝∞ = −𝜌 − 𝜌 ∇𝜙 ∇𝜙 (.)
𝜕𝑡 2

Only the body velocity is time dependent in the velocity potential (.). Therefore,
the time derivative of the potential is
( )
𝜕𝜙 𝜕 −𝑈 (𝑡)𝑅2 𝜕𝑈 (𝑡) 𝑅2
= cos(𝜑) = − cos(𝜑) (.)
𝜕𝑡 𝜕𝑡 𝑟 𝜕𝑡 𝑟
Substituting Equations (.) and (.) into (.) and evaluating the expression
for 𝑟 = 𝑅 results in the pressure on the cylinder surface.
𝜕𝑈 (𝑡) 1
𝑝dyn (𝑟 = 𝑅, 𝜑, 𝑡) = 𝜌 𝑅 cos(𝜑) − 𝜌 |𝑈 (𝑡)|2 (.)
𝜕𝑡 2

Unsteady For the resulting pressure force, we use the definition (.) and the surface element
pressure force d𝑆 from Equation (.).
( ) 2𝜋 ( )
𝐹𝑝𝑥 (𝑡) ( 𝜕𝑈 (𝑡) ) − cos(𝜑)
1
𝐹 𝑝 (𝑡) = = 𝜌𝑅 cos(𝜑) − 𝜌 |𝑈 (𝑡)|2 𝓁𝑅d𝜑
𝐹𝑝𝑦 (𝑡) ∫ 𝜕𝑡 2 − sin(𝜑)
0
(.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

17.5 Added Mass 221

As before, we split the integral into terms and extract all factors which do not depend
on the integration variable 𝜑.

2𝜋 ( )
2 𝜕𝑈 (𝑡)
− cos2 (𝜑)
𝐹 𝑝 (𝑡) = 𝜌 𝓁𝑅 d𝜑
𝜕𝑡 ∫ − sin(𝜑) cos(𝜑)
0
2𝜋 ( )
1 − cos(𝜑)
− 𝜌𝓁𝑅 |𝑈 (𝑡)|2 d𝜑 (.)
2 ∫ − sin(𝜑)
0

The second integral in Equation (.) already appeared in the steady parallel flow
case (.) and does not produce a force.
The remaining integral is easily solved.

2𝜋 ( )
𝜕𝑈 (𝑡) − cos2 (𝜑)
𝐹 𝑝 (𝑡) = 𝜌 𝓁𝑅2 d𝜑
𝜕𝑡 ∫ − sin(𝜑) cos(𝜑)
0
[( 1 )]2𝜋
2 𝜕𝑈 (𝑡) 2
𝜑 + 14 sin(2𝜑)
= −𝜌 𝓁𝑅 1
𝜕𝑡 sin(2𝜑)
2 0
[( )]2𝜋
j 𝜕𝑈 (𝑡) 𝜋+0−0−0 j
= −𝜌 𝓁𝑅2 (.)
𝜕𝑡 0−0
0

Again, due to the symmetry with respect to the 𝑥-axis the force in transverse direction
vanishes as expected.
However, in contrast to the steady flow case, we now have a resultant force in the Added mass
𝑥-direction: the added mass force force

𝜕𝑈 (𝑡)
𝐹𝑝𝑥 = −𝜌 𝓁𝜋𝑅2 (.)
𝜕𝑡
The minus sign indicates that it is acting against the direction of motion. The force
is proportional to the acceleration 𝜕𝑈 ∕𝜕𝑡 of the cylinder. Obviously, if the cylinder
moves at a constant speed and 𝜕𝑈 ∕𝜕𝑡 = 0, we return to the d’Alembert paradox case.
In general, the added mass force is proportional to the relative acceleration between
fluid and body.
The factor in front of the acceleration must have the dimension of mass. Indeed, 𝜌 𝓁𝜋𝑅2 Added mass
is equivalent to the mass of the water displaced by the cylinder and is called added
mass. For noncylindrical shapes, the added mass is generally different from the mass of
the displaced water. The term ‘added mass’ is somewhat misleading. Unlike the ballast
water pumped into a ship, there is no actual mass added to the cylinder. However,
the unsteady pressure force creates an apparent increase in the inertia of the cylinder
because it is proportional to the relative acceleration between body and fluid. Added
mass is important for the correct assessment of the dynamics of ships in waves and
during maneuvering in calm water.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

222 17 Ideal Flow Around A Long Cylinder

Self Study Problems

. Derive the velocity vector in Cartesian coordinates for the flow potential (.)
for a cylinder in parallel flow.
. The velocity potential for the flow around a rotating cylinder is given by
(𝑦)
𝑥 Γ
𝜙(𝑥, 𝑦) = −𝑈∞ 𝑥 − 𝑈∞ 𝑅2 + arctan Cartesian coordinates
𝑥2 + 𝑦2 2𝜋 𝑥

𝑈∞ 𝑅2 Γ
𝜙(𝑟, 𝜑) = −𝑈∞ 𝑟 cos(𝜑) − cos(𝜑) + 𝜑 polar coordinates
𝑟 2𝜋
Γ is the circulation strength, 𝑅 the radius of the cylinder, and 𝑈∞ the velocity of
the parallel flow.
(a) Derive the equations for the velocity vector in Cartesian and polar coordi-
nates.
(b) Compute the location of the stagnation points as a function of the position
angle 𝜑 for circulation strength values Γ ≤ 4𝜋𝑈∞ 𝑅.
(c) What happens to the stagnation point if Γ > 4𝜋𝑈∞ 𝑅?
(d) Compute the resultant pressure force acting on the cylinder as a function of
𝑈∞ , 𝑅, and Γ. This is best done in polar coordinates.
j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

223

18
Viscous Pressure Resistance

The ideal flow assumption used in potential theory is useful in a number of cases.
However, as d’Alembert’s paradox (Section .) expresses, potential theory is incapable
of predicting the total resistance of a body. Even if we include the frictional resistance
caused by the shear stresses in the boundary layer (Chapter ), the result will not match
experimental drag data. What is missing is the displacement effect of the boundary
layer onto the exterior flow. The displacement effect changes the pressure distribution
around a hull, which results in the viscous pressure resistance. In some cases, especially
for blunt bodies, flow separation may amplify the effect.

j
Learning Objectives j

At the end of this chapter students will be able to:

• illustrate the effects of a boundary layer on the external flow

• discuss viscous pressure resistance

• recognize flow separation

18.1 Displacement Effect of Boundary Layer

The viscosity of the fluid results in the frictional resistance. In addition, the resulting Displacement
boundary layer also affects the exterior flow. Since the fluid in the boundary layer effect
is slowed down by friction, less mass is flowing through the boundary layer per unit
time compared to a hypothetical inviscid fluid flow. The displacement thickness 𝛿1 ,
introduced in Chapter , quantifies this effect.
The flow outside of the boundary layer is pushed away from the hull (Figure .). Due
to this displacement effect, the body appears to be thicker for the exterior flow than its
actual structural dimensions. Since the boundary layer thickness grows from bow to
stern, the aft ship is more affected than the bow.
The outer limit of the boundary layer is defined as the position off the hull surface,
where local velocity reaches the undisturbed exterior (inviscid) flow speed. As noted
before, the curve marking the boundary layer thickness (dashed line in Figure .)

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

224 18 Viscous Pressure Resistance

Figure 18.1 The displacement effect of a boundary layer changes the effective hull shape

j j

Figure 18.2 The effect of viscous flow on the pressure distribution

is not a streamline of the flow, and neither is the solid line marking the displacement
effect. However, the displacement effect boundary may be interpreted as a streamline
for a hypothetical, inviscid flow which would have the same mass flow properties as the
actual viscous flow. This imaginary streamline extends the hull aft beyond the actual
physical surface with significant effects on the pressure distribution around the hull.
Change in In our studies of potential (inviscid) flow around a cylinder, we found stagnation points
pressure at the front and back side of the body. This applies to other geometries submerged in
distribution
an ideal fluid. The pressure coefficient takes the value 𝐶𝑝 = 1 at stagnation points. The
pressure forces on front and back side are equal but act in opposite directions. As a

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

18.1 Displacement Effect of Boundary Layer 225

consequence, the resulting pressure force vanishes in the flow direction if the fluid is
considered inviscid (d’Alembert’s paradox). This hypothetical case is depicted at the
top of Figure ..
In a real fluid, only the stagnation point at the bow remains. The aft stagnation point Viscous pressure
has vanished due to the displacement effect of the boundary layer. Therefore, the resistance
pressure remains lower there compared to the ideal flow case. High pressure at the
bow (stagnation point) and lower pressure at the stern create a net force which acts
against the body’s forward motion. We call this force the viscous pressure resistance.
Like other resistance components, the viscous pressure resistance cannot be measured
directly. In theory, the pressure distribution over the hull may be measured, but we still
would be unable to separate the contribution of the displacement effect from other flow
features like, for instance, the ship generated waves.
As the name suggests, viscous pressure resistance 𝑅𝑉𝑃 is a consequence of the viscosity Function of
of fluid. Therefore, it must be dependent on the Reynolds number 𝑅𝑒. Despite this Reynolds
number
fact, William Froude knowingly included 𝑅𝑉𝑃 in the residuary resistance which we
assume to be independent of the Reynolds number and to be solely a function of the
Froude number 𝐶𝑅 (𝐹𝑟). Without a practical means of determining 𝑅𝑉𝑃 , Froude had
little choice.
The recommended procedures of the International Towing Tank Conference (ITTC, Form factor
http://www.ittc.info) encapsulate the viscous pressure resistance in the form factor
𝑘 (Gross, ). The major component of viscous resistance for model and full scale
vessel is the flat plate equivalent frictional resistance, represented by the ITTC 
j j
model–ship correlation coefficient 𝐶𝐹 .
( )
𝐶𝑉𝑀 = 1 + 𝑘 𝐶𝐹𝑀 model (.)
( )
𝐶𝑉𝑆 = 1 + 𝑘 𝐶𝐹𝑆 + Δ𝐶𝐹 ship (.)

The portion 𝑘𝐶𝐹 captures the viscous pressure resistance and other three-dimensi-
onal effects. The same form factor is used for model and ship. Δ𝐶𝐹 is the roughness
correction which is applied only to the full scale vessel because models are considered
hydrodynamically smooth.
Prohaska () proposed a method to determine the form factor from model tests
(see Section . for details). It is considered the most reliable procedure available.
However, the uncertainty of results is still high. Several formulas have been proposed
to estimate the form factor 𝑘. Unfortunately, neither of them is universal nor does their
use significantly improve the resistance prediction from model tests (Gross, ). One
of the simpler formula has been presented by Watanabe ():

25.6 𝐶𝐵
𝑘 = −0.095 + ( ) √ (.)
𝐿𝑊𝐿 2 𝐵
𝐵 𝑇

𝐶𝐵 is the block coefficient, 𝐵 the molded beam, 𝑇 the molded draft, and 𝐿𝑊𝐿 the length
in the waterline. Granville () derived a formula based on Schoenherr’s friction line.
The following version of Granville’s formula has been corrected for use with the ITTC

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

226 18 Viscous Pressure Resistance

Figure 18.3 Velocity profiles within the boundary layer near a separation point

 model–ship correlation line (Gross, ).

𝐶𝐵2
𝑘 = −0.03 + 32.8 ( ) (.)
𝐿𝑊𝐿 2 ( 𝐵 )
𝐵 𝑇

j Unfortunately, the correlation of form factors derived from Equations (.) and (.) j
with results from model tests is poor (Gross, ).
Grigson () defines a regression formula specifically for usage with his friction line
Equation (.).
[ √ ]
𝑆 𝐵
𝑘 = 0.028 + 3.30 𝐶𝐵 (.)
(𝐿𝑊𝐿 )2 𝐿𝑊𝐿
use only in conjunction with Equation (.)

Additional formulas are available in more recent methods for resistance estimates which
use a form factor (see e.g. Chapter ).

18.2 Flow Separation

Separation point Friction within the fluid, especially at the wall, causes a continuous loss of momentum.
The fluid particles slow down in the boundary layer and the boundary layer grows in
thickness. Except for turbulent fluctuations, the flow maintains the direction of the
external flow and follows the body contour. The velocity profile across the boundary
layer has a positive gradient 𝜕𝑢∕𝜕𝑦 > 0 at the wall (Figure .). However, the gradient
diminishes due to the continuous loss of momentum. Eventually, the velocity profile
has a zero gradient perpendicular to the wall: 𝜕𝑢∕𝜕𝑦 = 0. This condition marks the
separation point. Just downstream of the separation point the fluid flows in the opposite
direction (back flow). Further downstream we encounter a confused flow field with
large and small eddies.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

18.2 Flow Separation 227

1.0
laminar viscous flow Re =10 5
turbulent viscous flow Re =1 07
inviscid flow Cp =14s in2 ()
0.5

0.0
[]

0.5
pressure coefficient Cp

1.0

1.5

2.0

2.5
j j

3.0
7/8 3/4 5/8 /2 3/8 /4 /8 0
position angle [rad]

(a) Typical pressure distributions over one side of a cylinder for (b) Resulting pressure forces for inviscid flow
inviscid flow as well as laminar and turbulent viscous flow and turbulent viscous flow

Figure 18.4 Comparison of pressure and forces acting on a cylinder in inviscid and viscous flow

Ludwig Prandtl’s boundary layer theory provided the first theory capable of explaining
the phenomenon of separation and a method to estimate the location of the separation
point. However, the estimates are not necessarily accurate because the boundary layer is
no longer thin at the separation point, which violates the basic assumption of Prandtl’s
boundary layer theory.
It should be emphasized that flow separation does not occur on a well designed ship hull.
If at all, ships with high block coefficients, like tankers and bulk carriers, may encounter
small areas of flow separation at the stern. Blunt objects like spheres, cylinders, or foils
at high angles of attack will suffer from flow separation.
We study the effect of the boundary layer and flow separation on the pressure distri- Pressure
bution around a cylinder. Figure .(a) compares the pressure coefficient derived for distribution in
ideal and viscous
inviscid flow flow
𝐶𝑝 = 1 − 4 sin2 (𝜑) (𝑟 = 𝑅) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

228 18 Viscous Pressure Resistance

(a) Turbulent boundary layer flow around a (b) Laminar boundary layer flow around a
cylinder with delayed separation cylinder with early separation

Figure 18.5 Comparison of turbulent and laminar boundary layer flow around a cylinder

with typical experimental results for laminar viscous flow and turbulent viscous flow.
𝜑 is the position angle with respect to the positive 𝑥-axis. The results for turbulent
flow (𝑅𝑒 = 107 , marked with ■) match the ideal flow results pretty well, except for the
backside of the cylinder where the pressure coefficient does not rise back to the level +1
of a stagnation point. Instead, it levels off at a roughly constant value. The root cause is
j the displacement effect of the boundary layer, augmented by flow separation. j
Viscous pressure Figure .(b) compares the pressure force distributions on the cylinder surface for
resistance of ideal flow and turbulent viscous flow. For the latter, the pressure force distribution is no
cylinder
longer symmetric with respect to the 𝑦-axis. Both, front and back, feature differential
pressure forces d𝐹 𝑝 which point in flow direction (against the movement of the cylinder)
and result in the viscous pressure resistance 𝑅𝑉𝑃 .
Turbulent flow The pressure distribution for laminar flow in Figure .(a) (𝑅𝑒 = 105 , marked with
separation ⋆) shows little resemblance to the potential flow pressure coefficient (solid line). This
is the result of early flow separation and typically results in a higher drag coefficient
compared with the turbulent flow case. At first glance, this is surprising because wall
friction is higher for turbulent flow. However, momentum in the turbulent boundary
layer is replenished through energy exchange with the exterior flow, allowing it to stay
attached longer and recover more of the pressure compared with the laminar boundary
layer. This is depicted in Figure .(a). The separation points are on the back side of the
cylinder and the region of separated flow is comparatively small. The exact locations of
the separation points depend on a number of factors: Reynolds number, preturbulence
of the flow, surface roughness, and others.
Laminar flow The laminar flow case is shown in Figure .(b). It lacks energy exchange with the
separation exterior flow. Thus, laminar flow reaches the condition 𝜕𝑢∕𝜕𝑦 = 0 for separation earlier.
In Figure .(b) separation occurs even before the maximum width of the body. A
large area of separated flow is created and, as a consequence, even less of the pressure
is recovered, causing a high viscous pressure resistance (Figure .(a)).
As discussed in Chapter , boundary layers of ships are turbulent due to the high
Reynolds numbers. Ship models use a trip wire, studs, or a strip of sand to trigger

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

18.2 Flow Separation 229

the laminar–turbulent transition of the boundary layer. This not only ensures that the
boundary layer of the model is mostly turbulent, it also prevents the unwanted effects
of possible laminar flow separation.

References
Granville, P. (). Progress in the analysis of the viscous resistance of surface
ships. Technical Report SPD -, Naval Ship Research and Development Center
(NSRDC).
Grigson, C. (). A planar friction algorithm and its use in analysing hull resistance.
In Transactions of The Royal Institution of Naval Architects (RINA), volume ,
pages –. The Royal Institution of Naval Architects.
Gross, A. (). Form factor. In Proceedings of the th ITTC, pages –, Ot-
tawa, Canada. International Towing Tank Conference. Report of the Performance
Committee, Appendix .
Prohaska, C. (). A simple method for the evaluation of the form factor and the low
speed wave resistance. In Proceedings of the th ITTC, pages –, Tokyo, Japan.
International Towing Tank Conference. Resistance Session, Written Contributions.
Watanabe, K. (). Note to the Performance Committee. International Towing Tank
Conference.

j j

Self Study Problems

. What is the main cause of the viscous pressure resistance of a ship?

. Sketch how the pressure distribution around a cylinder in a viscous fluid deviates
from the ideal flow case.
. Golf balls with dimples fly about % farther than experimental smooth golf balls.
Based on your knowledge of laminar and turbulent flow separation, discuss why
a golf ball with dimples flies farther than a perfectly smooth golf ball.

. How does flow separation increase the resistance of a blunt body?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

230

19
Waves and Ship Wave Patterns

Ships generate a very distinctive wave pattern, also known as the Kelvin wave pattern
in honor of William Thomson’s (Lord Kelvin, * – †) efforts in developing a
suitable theory (Thomson , Lord Kelvin). This chapter discusses basic properties of
waves and the Kelvin wave pattern utilizing results of linear wave theory. This may serve
as a foundation for our more detailed discussion of linear wave theory in subsequent
chapters.

Learning Objectives

j At the end of this chapter students will be able to: j

• discuss basic wave descriptors like wave height, wave length, and wave period.

• understand the phenomenon of wave dispersion

• distinguish phase velocity and group velocity

• describe ship wave patterns

19.1 Wave Length, Period, and Height

Free surface Before discussing the ship wave pattern itself, we will review some basic properties
waves of water waves. The surface separating the liquid water from the gaseous air is free of
parallel shear stresses when the fluids are in equilibrium. In physics, this is called a free
surface. Changes in pressure at the free surface, e.g. due to wind or a moving object,
create waves, which subsequently propagate across the water surface.
Wave profile Cutting through the water surface parallel to the direction of propagation reveals the
wave elevation 𝜁, which measures the deviation from the calm water level 𝑧 = 0. The
wave profile will resemble the solid curve shown in Figure ..
Wave length and The horizontal distance between two wave crests marks the wave length 𝐿𝑤 . The
height maximum vertical distance between wave trough and wave crest defines the wave
height 𝐻. Note that the vertical scale is enlarged in Figure .. Typically waves have
steepness ratios of 𝐻∕𝐿𝑤 less than .. In theory, waves will break once they reach a

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.1 Wave Length, Period, and Height 231

Figure 19.1 Definition of wave length 𝐿𝑤 and wave height 𝐻; the vertical scale is exaggerated

Figure 19.2 Surface elevation of a harmonic, long-crested wave

wave steepness of 𝐻∕𝐿𝑤 = 1∕7, but usually break before they reach this limit (Michell,
).
In simulations, we often repeat the same wave profile over and over again: a regular Regular,
j j
wave. Natural waves are irregular and have to be described by probabilistic methods. long-crested
waves
Furthermore, the wave profile is assumed to be the same in every plane parallel to the
direction of propagation. The crests and troughs of this long-crested wave stretch to
infinity in the transverse direction. In linear wave theory, the actual wave profile is
approximated by a cosine function.

𝜁(𝑧, 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)

This simplified, harmonic wave is represented in Figure . by the dashed curve. Note
that the wave crest of the real wave profile is higher than the wave trough is deep. The
wave trough is wider than the wave crest.
In Equation (.), 𝜃 = 𝑘𝑥−𝜔𝑡 defines the phase of a wave. Figure . shows the wave Wave amplitude
elevation 𝜁(𝑥, 𝑡) for one cycle 𝜃 ∈ [0, 2𝜋] from wave crest to wave crest. Since the cosine
function is limited to values between ±1, the wave elevation takes values between ±𝜁𝑎 .
Thus, the total range of values for a harmonic wave is two times the wave amplitude 𝜁𝑎 .
The wave amplitude is always positive and defines the maximum deviation of the wave
elevation from its mean value 𝑧 = 0.
Wave crests (maxima of 𝜁) occur at phases equal to even multiples of 𝜋, i.e. 0, 2𝜋, 4𝜋, Wave crests and
etc.: troughs
max(𝜁(𝜃)) = +𝜁𝑎 for 𝜃 = 2𝑘𝜋 with k=,,,…
Wave troughs (minima of 𝜁) occur when the phase 𝜃 is equal to odd multiples of 𝜋. For
harmonic waves, the wave height 𝐻 is equal to twice the wave amplitude.

𝐻 = 2𝜁𝑎 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

232 19 Waves and Ship Wave Patterns

Figure 19.3 Recording of surface elevation of a harmonic, long-crested wave at a fixed position
(𝑥 = 0)

Wave period Let us consider a wave probe which records the free surface elevation 𝜁(𝑡) as a function
of time at the fixed position 𝑥 = 0 (Figure .). The time that passes between the
recording of one wave crest and the next subsequent wave crest is the wave period 𝑇 .
Wave frequency The completion of a single cycle during one period 𝑇 requires a change in phase of Δ𝜃
equal to two 𝜋.
Δ𝜃 = 2𝜋 = (𝑘𝑥 − 𝜔𝑡)(𝑥=0,𝑡=𝑇 ) = − 𝜔𝑇
In other words, the wave frequency 𝜔 is reciprocal to the wave period 𝑇 .

2𝜋 2𝜋
𝜔 = 𝑇 = (.)
𝑇 𝜔
j As an angular frequency, the wave frequency measures how much of a single wave cycle j
is happening in 2𝜋 seconds. Its physical unit is [/s], however, we often denote it as
[rad/s] to distinguish it from a frequency measured in Hz = /s. A wave frequency of
𝜔 = 1 rad/s means a cycle is completed in 𝑇 = 2𝜋 s or approximately . seconds.
With a wave frequency of 𝜔 = 0.5 rad/s, only half of the cycle is completed within
𝜋 ≈ 6.283 seconds. The period of this wave would be 𝑇 = 4𝜋 s = 12.566 seconds.
Typical wave frequencies for ocean and ship generated waves are in the range of 0.1 rad/s
to 2.0 rad/s.
Wave length If we take a photograph with a very short exposure time, the wave is captured as a
function of space 𝑥. For convenience, we choose the time 𝑡 = 0. Cutting through the
wave in direction of wave progression 𝑥 reveals the wave length 𝐿𝑤 (see Figure .). If
the wave length is long, the wave is called long. As we will show later, long waves also
have a high wave period (small frequency). Short waves have small wave periods and
high wave frequencies.

Figure 19.4 Spatial extension of surface elevation of a linear, harmonic, long-crested wave
captured at time (𝑡 = 0)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.2 Fundamental Observations 233

Figure 19.5 A snapshot of the wave elevation in a wave group

At any selected point in time 𝑡, the change in phase of Δ𝜃 from one wave crest to the Wave number
other is still equal to 2𝜋.

Δ𝜃 = 2𝜋 = (𝑘𝑥 − 𝜔𝑡)(𝑥=𝐿𝑤 ,𝑡=0) = 𝑘 𝐿𝑤

The wave number 𝑘 is the spatial equivalent to the wave frequency. Wave number and
wave length are reciprocal.
2𝜋 2𝜋
𝑘 = 𝐿𝑤 = (.)
𝐿𝑤 𝑘
The wave number expresses how much of the wave length fits into a distance of 2𝜋 meter.
Wave numbers are noted in the unit /m. Some prefer the unit rad/m to indicate the
j factor 2𝜋. A wave frequency of 𝑘 = 1 rad/m is equivalent to a wave length of 𝐿𝑤 = j
2𝜋 m ≈ 6.283 meter. With a wave number of 𝑘 = 0.1 rad/m, a wave is approximately
. meters long and a tenth of it fits into 2𝜋 meters. Since a wave length is always a
positive distance, wave numbers are positive as well.
A wave travels one wave length during one wave period. Therefore, the speed of Phase velocity
propagation of a water wave is
𝐿𝑤 𝜔
𝑐 = = (.)
𝑇 𝑘
If you move in lockstep with a wave crest or any other point along the wave, you have
to move with velocity 𝑐. Since you will have a constant phase relationship to the wave,
the velocity of wave propagation is known as phase velocity.

19.2 Fundamental Observations

When you throw a pebble into a calm pond, a group of waves is created. Figure . Wave groups
shows the vertical displacement of the water surface 𝜁 along the wave group. As the
waves propagate and expand in the form of concentric circles, three observations can
be made:

(i) Long waves propagate faster than short waves, i.e. if 𝐿𝑤1 > 𝐿𝑤2 then 𝑐1 > 𝑐2 .
(ii) At the front of the wave group, wave amplitude diminishes until the wave vanishes,
whereas at the end of the wave group, waves seem to emerge from nothing.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

234 19 Waves and Ship Wave Patterns

(iii) The waves seem to travel from the back to the front of the group, while the envelope
of the wave group moves slower than the individual waves.

Wave dispersion The first observation (i) is known as wave dispersion. Sound waves of varying frequency
all travel with a constant speed, i.e. the speed of sound. Electromagnetic waves of
different frequency propagate with the speed of light. In contrast, water waves with small
wave frequency (long waves) travel faster than short waves with high wave frequency.
Very long waves, like the ones created by tsunamis, may cross whole oceans in a few
hours. The phase velocity is also influenced by the water depth ℎ. We will derive
the exact relationship between wave frequency, water depth, and phase velocity in
Section ..
Wave energy Observation (ii) is related to the energy transport in waves. Chapter  presents a
detailed discussion of wave energy and wave propagation. For now, it may suffice to
state that waves contain two types of energy:

• kinetic energy – water particles underneath a wave move in elliptical paths. Since
water particles have mass and velocity, they possess kinetic energy.
• potential energy – the up and down of the free surface contains potential en-
ergy comparable to the energy of a pendulum. However, potential wave energy
propagates with the wave.

The water is initially at rest in front of the wave group. For the water particles to assume
j j
the motion characteristic to waves, potential energy is converted into kinetic energy at
the front of the wave. As a consequence, the wave amplitude diminishes. At the back of
the wave group, kinetic energy is converted into potential energy and waves emerge.
Once all kinetic energy has been converted into potential energy and transported away,
the water is at rest again.
Viscous effects Viscosity of the water will dissipate some wave energy during this process. This is
especially true for waves of very short length (< 0.1 m), so-called capillary waves. They
are also affected by surface tension. Ship and ocean waves have longer wave lengths,
and the effects of surface tension and viscosity may be neglected for many engineering
applications.
Group velocity The observations (ii) and (iii) are connected. Individual waves travel with their respective
phase velocity 𝑐. As the wave group approaches, water particles are set in motion, but
more or less come to rest in their initial position once the wave group has passed. This
means that only the potential energy propagates. As we will discuss in Chapter , wave
energy travels at the group velocity 𝑐𝐺 . On deep water it is just half of the phase velocity.
The envelope of a wave group reflects the reduced speed of the energy transport. For
that reason, you will see waves emerge from the back of the wave group, travel through
the group to the front, and finally diminish because their energy is used to set water
particles in motion. The relationship between group and phase velocity affects the
formation of the Kelvin wave pattern, which we will discuss in the following section.
Wave Next time you pass a quiet pond, throw two pebbles into the water a few yards apart.
superposition Watch how the two systems of radiated circular waves spread and interact. For the most
part, they will just pass each other and continue on undisturbed. A notable change will
occur only if the superposition of waves causes wave breaking. You may notice that

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.3 Kelvin Wave Pattern 235

Figure 19.6 Kelvin wave pattern in deep water

the waves appear to be circular after a short distance, no matter what the shapes of the
pebbles were.

j j
19.3 Kelvin Wave Pattern
Anything moving at or close enough to the water surface creates waves: You may drag
a stick through the water, watch a swan glide across a pond, or observe a koi swimming
just underneath the water surface. In all cases, you will observe the wave pattern that is
characteristic to all ships traveling the world’s oceans, lakes, and rivers.
Thomson (Lord Kelvin) studied the wave pattern created by a traveling point distur-
bance (Figure .). The solid curves represent lines of constant phase, like the position
of wave crests. The parametric equations in Figure . have been used to create the plot.
They represent an approximation of the rather complicated mathematics behind this
problem. From Figure . and observation, the following features may be discerned
for wave patterns:

• Waves appear only in a triangular region behind the vessel. Wave elevations Kelvin angle
rapidly diminish outside of the triangle. In deep water, the interior angle at the
tip is . degrees and is known as the Kelvin angle.
• The Kelvin wave pattern consists of two types of waves: divergent waves and Kelvin wave
transverse waves. We also call the divergent waves ‘diagonal waves’ and the pattern
transverse waves ‘following waves.’ Divergent waves appear at the sides of the
wave pattern and travel away from the ship’s path. Transverse waves follow the
ship with their crests and troughs aligned perpendicular to the ship’s course.
• Observed from a moving boat, the wave pattern appears to be stationary, which Waves follow the
means that the waves themselves propagate with ship speed in the direction of ship
the ship’s forward motion.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

236 19 Waves and Ship Wave Patterns

Figure 19.7 Change of Kelvin wave pattern with increasing velocity on deep water

Ship speed • The wave length 𝐿𝑤 depends on the ship’s speed and, in some cases, also on the
determines wave water depth (Figure .). The faster a vessel sails, the longer its waves become.
length
Wave breaking • Especially at higher speeds or with blunt bows, waves near the bow break, like
the waves rolling onto a beach.

Linear wave theory provides a relationship for wave length 𝐿𝑤 and wave phase velocity 𝑐.

𝑐2
𝐿𝑤 = 2𝜋 (.)
𝑔 tanh(𝑘ℎ)
On deep water, the product of wave number 𝑘 and water depth ℎ is larger than 𝜋, i.e.
j 𝑘ℎ > 𝜋 and tanh(𝑘 ℎ) ≈ 1. The transverse waves progress with the same speed as the j
disturbance, i.e. 𝑐 = 𝑣𝑆 or

𝑐 2 = 𝑣2𝑆 = 𝐹𝑟2 𝑔 𝐿𝑊𝐿 (.)

We substitute Equation (.) into (.) and obtain a formula which provides the length
of transverse waves as a function of Froude number for deep water.

𝐿𝑤 = 2𝜋 𝐹𝑟2 𝐿𝑊𝐿 for 𝑘ℎ > 𝜋 (.)

The photo in Figure . may exemplify that Lord Kelvin’s theoretical model is a pretty
good match to nature. The satellite photo shows a Kelvin wave pattern visualized by
clouds in the wake of Amsterdam Island in the Indian Ocean. The volcano on the island
creates a Kelvin pattern at the boundary of two layers of air with different temperatures.
As warm, moist air is pushed up into cooler air, water condenses and clouds form (wave
crests). When the air sinks, it warms up and the clouds dissolve again (wave troughs).
Ship wave A ship disturbs the water surface over its whole length rather than just at a single point.
pattern Figure . shows the simulation of the wave pattern behind a ship hull. The specific
shape of the vessel influences the height of the waves and the details of the wave pattern.
The basic pattern shape, however, does not change as long as the vessel is sailing on
deep water and is carried by its buoyancy. The nature of the wave pattern changes if
the vessel maneuvers from displacement to planing mode. A planing vessel no longer
produces a Kelvin wave pattern.
Wave Once waves have been generated by a moving vessel, they become independent of the
propagation vessel. When the vessel changes its course, newly generated waves will be oriented

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

19.3 Kelvin Wave Pattern 237

Figure 19.8 Kelvin wave pattern like cloud formation in the slipstream of Amsterdam Island in
j the southern Indian Ocean. Photo courtesy of NASA Earth Observatory j

according to the new course. The waves generated before the course change will
continue on their original path. If the vessel stops, wave generation ceases, but the
already existing waves continue to spread.
Waves contain and transport energy. Therefore, the creation of waves requires energy. Wave resistance
Wave energy grows quadratic with wave amplitude. Provided two waves have the same
wave length, a wave with twice the wave height contains four times the energy. In calm
water, the moving ship is the only possible source of energy for wave making. Kinetic
energy of the ship is constantly converted into wave energy. The loss of energy for the
ship results in its wave resistance. As naval architects we aim to minimize the wave
height in the wave pattern to minimize the loss of energy (see Chapter ).
If the water depth is less than half the wave length, the ocean bottom or river bed will Shallow water
influence the behavior of the waves. A first visible sign is a widening of the Kelvin effect
angle. Unfortunately, shallow water effects also result in increased wave resistance
(Schlichting, ).

References
Michell, J. (). The hightest wave in water. Philosophical Magazine Series , :–
.
Schlichting, O. (). Schiffswiderstand auf beschränkter Wassertiefe; Widerstand von

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

238 19 Waves and Ship Wave Patterns

Figure 19.9 Wave pattern of a ship at 𝐹𝑟 = 0.26

j j
Seeschiffen auf flachem Wasser. In Jahrbuch der Schiffbautechnischen Gesellschaft
(STG), volume , pages –. Springer Verlag, Berlin.
Thomson (Lord Kelvin), W. (). On ship waves, volume  of Cambridge Library Col-
lection – Physical Sciences, pages –. Cambridge University Press, Cambridge,
United Kingdom. First presented in .

Self Study Problems

. Define wave height and wave length in your own words.
. How does a real wave profile deviate from a harmonic signal (cosine function)?

. Which geometric properties of a Kelvin wave pattern change when the ship moves
into less deep water while maintaining its speed?
. Compute the wave length for a ship sailing with velocity 𝑣𝑆 = 24 kn?
. In deep water, a ship creates a wave pattern with a basic wave length of 𝐿𝑤 =
50.0 m. How fast is the ship? Compute the wave length and phase velocity of the
divergent waves!

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

239

20
Wave Theory

Waves are fascinating objects. At times, they demurely lap on a beach and at other
times they cause havoc and destruction. In this chapter we develop a mathematical
model for waves that will enable us to study many of their properties. Because waves
independently propagate from their source, we deliberately ignore how they are created.
It is helpful to first study the behavior of waves alone.

Learning Objectives
At the end of this chapter students will be able to:
j j
• formulate wave flow as a boundary value problem

• derive free surface boundary conditions

• interpret the physical meaning of boundary conditions in wave theory

20.1 Overview
Attempts to capture the physics of waves in mathematical models go way back in time. History
Craik () provides a discussion of the development of water wave theory over the
past three centuries. Many famous scientists and mathematicians contributed to the
field: Newton, Lagrange, Laplace, Green, Cauchy, and Poisson to name a few.
The mathematics of the problem are quite difficult because we deal with nonlinear Stokes and Airy
partial differential equations. Different theories have been developed for general, as wave theories
well as specialized, applications. The most commonly used were developed by Stokes
() who provided approximations of first, second, and higher order for the original
nonlinear problem. The approximations are known as Stokes’ wave theories of first,
second, third, and fifth order. We encountered Sir George Gabriel Stokes already in
the context of the Navier-Stokes equations. The first order, or linear solution, is also
known as Airy wave theory because Airy published his work a few years before Stokes
did (Airy, ). Sir George Biddell Airy (*–†), an English mathematician
and astronomer, became Astronomer Royal and established the prime meridian in
Greenwich, UK, which is still in use as the reference line for longitude.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

240 20 Wave Theory

j Figure 20.1 Definition of coordinate system and domain boundaries for wave theory of j
long-crested waves

Linear wave Linear wave theory is one of the most important tools in seakeeping analyis and prob-
theory abilistic prediction of the short and long term behavior of marine vessels in waves.
Although it is limited to waves with a small wave height to wave length ratio 𝐻∕𝐿𝑤 ≪ 1,
linear wave theory provides important insight into the behavior of waves and the wave
making of a ship. Therefore, we will discuss linear wave theory in some detail.
Wave resistance The Australian mathematician Michell () developed a model for wave resistance
based on linear theory. It can be used to estimate wave resistance of fast, slender vessels,
especially multihull vessels. More on this in Section ..

20.2 Mathematical Model for Long-crested Waves

Long-crested In order to develop a two-dimensional mathematical model for waves, we assume that
waves their crests and troughs stretch in 𝑦-direction from negative infinity to positive infinity
(Figure .). These are commonly known as long-crested waves. This simplification
renders all flow properties independent of the 𝑦-coordinate and restricts wave flow
modeling to the 𝑥-𝑧-plane. The 𝑥-axis points in the direction of wave propagation and
the 𝑧-axis is pointing upward. We place the origin at the calm water level (𝑧 = 0). As
a consequence, points below the calm water level, i.e. in the water, will have negative
𝑧-coordinates (𝑧 < 0).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 241

Figure 20.2 Simplified two-dimensional fluid domain for long-crested waves

Figure . shows the simplified two-dimensional flow domain. The fluid domain 𝑉
is bordered at the top by the free surface 𝑆𝐹 . The ocean bottom 𝑆𝐵 forms the lower
boundary of the domain. We assume that the ocean bottom is parallel to the calm water
surface, i.e. the water depth ℎ is constant throughout the domain. Note that the water
depth ℎ is defined as a positive length.
As noted before, viscous effects are small for water waves of lengths typically created by Ignoring viscous
a moving ship. However, a notable amount of energy is dissipated in the case of breaking effects
j waves. If we exclude the phenomenon of wave breaking, the fluid flow in waves may be j
treated as inviscid for practical applications. Again, water is considered incompressible
(𝜌 = const.). In addition, we assume that the flow is irrotational (rot 𝑣 = 0), which
enables us to employ potential theory.
The Laplace equation and the Bernoulli equation represent the conservation of mass Conservation of
and conservation of momentum principles of fluid mechanics for potential flow. mass and
momentum
𝚫𝜙 = 0 Laplace eq. (.)
𝜕𝜙 𝜌
𝜌 + |∇ 𝜙|2 + 𝜌 𝑔 𝑧 + 𝑝 = 𝐶(𝑡) Bernoulli eq. for pot. flow (.)
𝜕𝑡 2
The wave flow is represented by its velocity potential 𝜙, which in general is a function
of space 𝑥 and time 𝑡. In order to define a unique solution of the Laplace equation,
boundary conditions are imposed at the borders of the flow region 𝑉 .
The domain 𝑉 is bordered by the ocean bottom 𝑆𝐵 and the free surface 𝑆𝐹 (see Fig- Domain
ure .). From our discussion of the two-dimensional flow around a cylinder in boundaries
Chapter , we know that an additional condition for the far field boundary 𝑆∞ at the
inlet and outlet of the flow is also required.

20.2.1 Ocean bottom boundary condition

Let us start with a simple boundary condition. Like the cylinder surface, the ocean No flow through
bottom 𝑆𝐵 is considered a fixed, impenetrable surface. Since we ignore friction, fluid ocean bottom

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

242 20 Wave Theory

motion parallel to the bottom is allowed. However, the normal velocity has to vanish
over 𝑆𝐵 .
𝜕𝜙
= 0 on 𝑆𝐵 (.)
𝜕𝑛
Since the ocean floor is parallel to the water surface, its outward oriented normal
vector points in the negative 𝑧-direction, i.e. 𝑛𝑇 = (0, 0, −1). Consequently, the normal
derivative is equal to the derivative in the negative 𝑧-direction.

𝜕𝜙 ( ) ⎛ 𝜙𝑥 ⎞ 𝜕𝜙
= 𝑛𝑇 ∇ 𝜙 = 0, 0, −1 ⎜ 𝜙𝑦 ⎟ = −
𝜕𝑛 ⎜ ⎟ 𝜕𝑧
⎝ 𝜙𝑧 ⎠
As a means to shorten the notation, we represent the partial derivatives with subscripts
𝜕𝜙 𝜕𝜙 𝜕𝜙
𝜙𝑥 = 𝜙𝑦 = 𝜙𝑧 = (.)
𝜕𝑥 𝜕𝑦 𝜕𝑧

The minus sign for the 𝑧 derivative is of no consequence, since we require the vertical
velocity component to vanish. Therefore, our final ocean bottom boundary condition
reads
𝜕𝜙
𝜙𝑧 = = 0 for 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
𝜕𝑧
Physically, it is a kinematic boundary condition because it restricts how the water may
j move. Mathematically it is classified as a Neumann boundary condition because it j
applies to a derivative of the unknown potential 𝜙.

20.2.2 Free surface boundary conditions

The ocean bottom 𝑆𝐵 is easily described mathematically: 𝑆𝐵 consists of all points with
𝑧 = −ℎ. Similarly, the calm water surface is formed by points 𝑥 = (𝑥, 𝑦, 0). However,
the position of the water surface is a priori unknown when waves are present. The shape
of the free surface 𝑆𝐹 has to be computed as part of the solution of the flow problem.
As a consequence the boundary conditions for the free surface have to be satisfied at a
yet unknown surface 𝑆𝐹 .
Mathematical As a first step, a mathematical model is introduced for the shape of the free surface.
model for free We employ an implicit function as you may have seen in the definition of a circle of
surface
radius 𝑅: 𝑥2 + 𝑦2 − 𝑅2 = 0. The contour is formed by all points whose distance from
the origin is equal to 𝑅. The implicit function for the free surface captures all points
whose vertical distance from the calm water level (𝑧 = 0) is equal to the free surface
elevation 𝜁.
𝐹 (𝑥, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑡) = 0 (.)
We call 𝜁 the wave elevation. Because we restricted the problem to long-crested waves,
the implicit function 𝐹 and the wave elevation 𝜁 do not depend on the transverse
coordinate 𝑦.
Limitations of The formulation (.) implies that 𝐹 and 𝜁 are analytic functions. In that case, a unique
free surface 𝜁 value must exist for all combinations (𝑥, 𝑡) which automatically excludes overturning
model
or breaking waves. As indicated in Figure ., in cases of wave breaking more than one

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 243

Figure 20.3 The mathematical free surface model is valid for nonbreaking waves only

𝜁 value might occur for a pair (𝑥, 𝑡). As a consequence the function 𝑧 = 𝜁(𝑥, 𝑡) is not
unique and does not exist. For the discussion of breaking waves, other mathematical
models have to be employed.
At the ocean bottom, we specify a single boundary condition because its geometry is Two boundary
j j
known. The only unknown is the velocity potential. For the free surface we need two conditions
needed
boundary conditions because we have two unknowns: the potential 𝜙 and the wave
elevation 𝜁. The two free surface boundary conditions will be discussed in the following
subsections.

Kinematic free surface boundary condition

The free surface must be a stream surface at all times 𝑡. Consequently, no fluid particle
may cross the stream surface 𝑆𝐹 . Like in the case of the body boundary condition
for the cylinder flow, we restrict the movement of the fluid, therefore this will be a
kinematic free surface boundary condition. The explicit expression for the kinematic
boundary condition at the free surface may be derived in two different ways: (a) using a
physical argument or (b) a more mathematical argument.
Let us start with the longer but probably more comprehensible physical argument. If Physical
𝑆𝐹 is a stream surface, fluid particles at the surface may move tangential to the surface argument
but not normal to it. Therefore, the relative normal velocity between surface and water
particles must vanish. The velocity of a water particle is 𝑣 = ∇ 𝜙. The surface moves up
and down with local velocity
𝜕𝜁
𝑣𝑆 = 𝑘
𝐹 𝜕𝑡
where 𝑘 = (0, 0, 1)𝑇 is the unit vector in vertical direction. Thus, the relative velocity is

𝜕𝜁
𝑣 − 𝑣𝑆 = ∇𝜙 − 𝑘 (.)
𝐹 𝜕𝑡

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

244 20 Wave Theory

According to our stream surface requirement, the component of the relative velocity
in normal direction must vanish. The latter is obtained by forming the dot product of
normal vector 𝑛 and relative velocity:
( )
𝜕𝜁
𝑇
𝑛 ∇𝜙 − 𝑘 = 0 (.)
𝜕𝑡

Normal vector For implicit surfaces, the normal vector 𝑛 is identical to the gradient normalized to unit
length.
∇𝐹
𝑛 = (.)
|∇ 𝐹 |
This relationship follows from the characteristic of gradients to always point into the
direction of steepest ascent, which in turn is perpendicular to the lines of constant
function value.
We substitute the expression (.) into Equation (.) and multiply it with the length
|∇ 𝐹 | of the gradient:
( )
( )𝑇 𝜕𝜁
∇𝐹 ∇𝜙 − 𝑘 = 0
𝜕𝑡
Normal relative Expanding the dot products yields
velocity must
vanish ⎛ 𝜕𝜙 ⎞
⎜ 𝜕𝑥 ⎟
j ( )⎜ ⎟ j
𝜕𝐹 𝜕𝐹 𝜕𝐹 ⎜
, ,
𝜕𝜙 ⎟ − 𝜕𝜁 𝜕𝐹 = 0
𝜕𝑥 𝜕𝑦 𝜕𝑧 ⎜ 𝜕𝑦 ⎟ 𝜕𝑡 𝜕𝑧
⎜ 𝜕𝜙 ⎟
⎜ ⎟
⎝ 𝜕𝑧 ⎠

and we get

𝜕𝐹 𝜕𝜙 𝜕𝐹 𝜕𝜙 𝜕𝐹 𝜕𝜙 𝜕𝜁 𝜕𝐹
+ + − = 0 (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 𝜕𝑡 𝜕𝑧

The partial derivatives of the implicit surface function 𝐹 = 𝑧 − 𝜁(𝑥, 𝑡) are

𝜕𝐹 𝜕𝜁 𝜕𝐹 𝜕𝜁
= − = −
𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦
(.)
𝜕𝐹 𝜕𝐹 𝜕𝜁
= +1 = −
𝜕𝑧 𝜕𝑡 𝜕𝑡
Finally, we substitute the derivatives (.) into the condition (.) that the normal
component of the relative velocity between water particles and free surface vanishes.
3D kinematic free For a three-dimensional wave field 𝐹 (𝑥, 𝑦, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑦, 𝑡) = 0, the following is
surface condition obtained for the kinematic free surface condition:
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧

2D kinematic free In the case of long-crested waves, the transverse derivatives 𝜕∕𝜕𝑦 vanish, and the kine-
surface condition

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 245

matic boundary condition for long-crested waves requires that the following condition
is satisfied at the free surface.
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧

The equivalent mathematical argument will of course lead to the same kinematic bound- Mathematical
ary condition: The condition 𝐹 = 𝑧 − 𝜁(𝑥, 𝑡) = 0 must be valid at all times in order for argument
𝐹 to accurately describe the free surface 𝑆𝐹 . Consequently, its value 𝐹 = 0 cannot
change over time, and its total derivative with respect to time must vanish. Applying
the substantial derivative, fluid mechanics’ equivalent of the total derivative, to 𝐹 yields
in vector form:
D𝐹 𝜕𝐹
= + 𝑣𝑇 ∇ 𝐹 = 0
D𝑡 𝜕𝑡
and in components
𝜕𝐹 𝜕𝐹 𝜕𝐹 𝜕𝐹
+𝑢 +𝑣 +𝑤 = 0 (.)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧
We rewrite the latter equation by replacing the partial derivatives of 𝐹 with the results
from Equation (.) and by substituting partial derivatives of the potential 𝜙 for the
components 𝑢, 𝑣, and 𝑤 of the velocity vector. As expected, the result is the same as
Equation (.)
𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙
− − − + 1 = 0 (.)
j 𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 j
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms
In long-crested waves, the transverse velocity 𝑣 = 0 vanishes and we obtain again
Equation (.) as the kinematic boundary condition for the free surface.
Unfortunately, the resulting kinematic free surface conditions (.) and (.) are Nonlinear and
nonlinear because they contain products of the derivatives of the unknown surface implicit
elevation 𝜁 and the velocity potential 𝜙. As an added difficulty, the condition has to be
satisfied at a yet unknown location 𝑧 = 𝜁(𝑥, 𝑡).

Dynamic free surface boundary condition

As mentioned above, a second boundary condition is needed because we do not know

the actual location of the free surface. This second condition will be employed to
eliminate the unknown surface elevation 𝜁 from the kinematic boundary condition.
Besides kinematic boundary conditions that impose restrictions on how the fluid can Physical
move, we can also prescribe dynamic conditions which define how forces interact or do argument
not interact with the fluid. We are interested in the behavior of free waves, i.e. waves
that have been created by wind or a ship but are now progressing on their own. That
means, we will explicitly ignore the interaction of waves with the atmosphere. Pressure
forces on the surface vanish if pressure differences at the water surface vanish.
To cast this into a mathematical model we recall Bernoulli’s equation for potential flow: Mathematical
model
 Admittedly, this is a simplification which works for the linearized equations only. The exact boundary

conditions form a system of coupled equations that has to be solved simultaneously.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

246 20 Wave Theory

𝜕𝜙 1( )𝑇 𝑝
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 𝐶(𝑡) (.)
𝜕𝑡 2 𝜌
The fluid is at rest far away from the progressing waves, and the free surface is exposed
to atmospheric pressure 𝑝0 . This forms our constant for the right-hand side 𝐶(𝑡) = 𝑝0 ∕𝜌.
𝜕𝜙 1( )𝑇 𝑝 𝑝
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 + = 0 at 𝑆𝐹 (.)
𝜕𝑡 2 𝜌 𝜌
The dynamic boundary condition requests that the pressure difference 𝑝0 − 𝑝 vanishes
at the free surface:
𝜕𝜙 1( )𝑇
+ ∇ 𝜙 ∇ 𝜙 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2
Note that all terms in Equation (.) have to be evaluated at 𝑧 = 𝜁(𝑥, 𝑦, 𝑡). Thus,
solving Equation (.) for the unknown wave elevation does not provide a means to
immediately calculate 𝜁:

1 𝜕𝜙 1 ( )𝑇
𝜁(𝑥, 𝑦, 𝑡) = − − ∇ 𝜙 ∇ 𝜙 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝑔 𝜕𝑡 2𝑔
The wave elevation is also part of the right-hand side in (.) as argument of potential
𝜙.
Nonlinear Expansion of the dot product of the gradients of the potential reveals that the dynamic
j dynamic free free surface boundary condition is nonlinear, like the kinematic free surface boundary j
surface condition
condition.
[( ) ( ) ( ) ]
2 2 2
𝜕𝜙 1 𝜕𝜙 𝜕𝜙 𝜕𝜙
+ + + + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2 𝜕𝑥 𝜕𝑦 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms

The second nonlinear term vanishes for the two-dimensional problem of long-crested
waves and the dynamic free surface boundary condition reads:
[( ) ( ) ]
2 2
𝜕𝜙 1 𝜕𝜙 𝜕𝜙
+ + + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡) (.)
𝜕𝑡 2 𝜕𝑥 𝜕𝑧
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
nonlinear terms

20.2.3 Far field condition

Radiation We have considered boundary conditions for the ocean bottom 𝑆𝐵 and the free surface
condition 𝑆𝐹 . What remains to be specified are conditions for the inlet and outlet surfaces 𝑆∞
in the far field. In the case of waves, the far field condition is often referred to as the
radiation condition.
Throw a stone A radiation condition commonly involves statements about the direction in which
into a pond waves are traveling and how their amplitude develops. Observe the waves created
after you throw a stone into a quiet pond: a set of circular waves emerges which travel

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.2 Mathematical Model for Long-crested Waves 247

away from the spot where the stone hit the water surface. The amplitudes of the waves
diminish as the rings expand until they finally fade away. In modeling this wave system,
the radiation condition would have to enforce the direction of wave propagation (away
from the stone) and that the amplitudes of waves vanish far away from their origin.
Friction forces do not exist in an inviscid fluid. Therefore, energy contained in the wave Radiation
flow cannot be dissipated into heat or eddies. In long-crested waves, no forces act in condition for
long-crested
the 𝑦-direction and thus energy cannot be distributed along the wave crests (parallel waves
to the 𝑦-axis). As a consequence, the wave height cannot vanish in the far field of our
two-dimensional domain. The only far field condition we impose is the requirement
that the waves progress with constant velocity 𝑐 in positive 𝑥-direction.

𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)

20.2.4 Nonlinear boundary value problem

Analogous to the cylinder flow problem, the Laplace equation together with the bound- 3D boundary
ary conditions form a boundary value problem whose solution contains the velocity value problem
potential 𝜙 and the wave elevation 𝜁.

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑦, 𝑧)𝑇 ∈ 𝑉
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑦, 𝑡)
j 𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 j
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧 = 0 (.)
2
𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑦, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
propagation, energy conservation on 𝑆∞

This reduces to the following set of equations for long-crested waves in the two- 2D boundary
dimensional case. value problem

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉
𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧 = 0 (.)
2
𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞

A general analytic solution is not known for the exact two- and three-dimensional
boundary value problems because of the nonlinear free surface boundary conditions.
However, approximate solutions can be derived by using a Fourier series expansion
or similar methods. If required, applications use the Stokes’ wave theories with ap-
proximations from first to fifth order. Most important for practical applications is the
first order wave theory, which solves the linearized problem as outlined below and in
subsequent chapters.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

248 20 Wave Theory

20.3 Linearized Boundary Value Problem

Linearized The nonlinear free surface boundary condition can be linearized if the analysis is
boundary value restricted to waves with very small steepness 𝐻∕𝐿 ≪ 1. The linearization process is
problem
based on a perturbation series expansion of the exact solution. A detailed description
of the process is provided in the following chapter. For now we simply study the result
of the linearization and point out the major differences to the exact boundary value
problem.
Linearized Obviously, the Laplace equation (.) and boundary conditions at the sea bottom
boundary value 𝑆𝐵 (.) and the far field 𝑆∞ (.) are unchanged. As expected, the nonlinear
problem
terms have vanished on the left-hand side of the kinematic and dynamic free surface
conditions (.) and (.). In addition, however, the two free surface conditions
are now imposed on the calm water surface 𝑧 = 0 instead of on the actual free surface
𝑆𝐹 . This is a major simplification enabling an analytic solution which will be derived in
Chapter  on linear wave theory.

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉 (.)

𝜕𝜁 𝜕𝜙
− + = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧
𝜕𝜙
+ 𝑔𝜁 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
j 𝜕𝑧 j
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)

A less obvious change in the linearized boundary value problem is that potential 𝜙 and
wave elevation 𝜁 are now interpreted as generally small disturbances of the fluid at rest.
Combined linear Now that the free surface conditions are enforced at the calm water level, they are no
free surface longer implicit. This allows us to solve the dynamic boundary condition (.) for the
condition
wave elevation 𝜁.
1 𝜕𝜙
𝜁(𝑥, 𝑡) = − 𝑥 with 𝑧 = 0 (.)
𝑔 𝜕𝑡
We take the time derivative of Equation (.) and substitute the result for the
time derivative of the wave elevation into the linearized kinematic boundary con-
dition (.):
𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡 2 𝜕𝑧

With the combined free surface condition for the free surface, the linearized boundary
value problem can be stated with the velocity potential 𝜙 as the only unknown:

𝚫𝜙 = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉
𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0
𝜕𝑡2 𝜕𝑧 (.)
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ
𝜕𝑧
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

20.3 Linearized Boundary Value Problem 249

Once we know the potential, the flow field in the fluid domain is given by the gra-
dient of the potential 𝑣 = ∇ 𝜙. The pressure follows from the linearized Bernoulli
equation (.):
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
and the wave elevation from Equation (.).

References
Airy, G. (). Tides and waves. In Smedley, E., Rose, H. J., and Rose, H. J., editors,
Encyclopaedia Metropolitana, volume V of Mixed Sciences, Vol. , pages –.
B. Fellowes et al., London.
Craik, A. (). The origins of water wave theory. Annual Review of Fluid Mechanics,
:–.
Michell, J. (). The wave resistance of a ship. Philosophical Magazine Series ,
():–.
Stokes, G. (). On the theory of oscillatory waves. Transactions of the Cambridge
Philosophical Society, :–.

j Self Study Problems j

. Why may we ignore viscosity for typical wind or ship generated free surface
waves?
. Why do we need two boundary conditions for the free surface but only one for
the ocean bottom?
. Derive the exact kinematic and dynamic free surface conditions for the three-
dimensional case, i.e. with 𝐹 (𝑥, 𝑦, 𝑧, 𝑡) = 𝑧 − 𝜁(𝑥, 𝑦, 𝑡) = 0.
. Name and explain in words the four boundary conditions of the exact boundary
value problem for long-crested waves:

𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − + = 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑧
𝜕𝜙 1
+ ||∇ 𝜙|| + 𝑔𝑧
2
= 0 𝑥 on 𝑆𝐹 , 𝑧 = 𝜁(𝑥, 𝑡)
𝜕𝑡 2
𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −𝑑
𝜕𝑧
𝑐 = const. > 0 at 𝑆∞ for |𝑥| → ∞

. Why is no analytic solution of the exact two-dimensional boundary value problem

(.) known?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

250

21
Linearization of Free Surface Boundary
Conditions

In this chapter, we formally derive the linearized versions of the kinematic and dynamic
free surface conditions used in wave theory. The procedure is known as a perturbation
approach and is applicable to the linearization of nonlinear equations in general. As
results of the linearization process we obtain the kinematic and dynamic free surface
boundary conditions which are employed in linear wave theory.

Learning Objectives
j j
At the end of this chapter students will be able to:

• apply the perturbation approach

• perform linearizations of nonlinear equations
• understand the origin of linearized free surface conditions
• know the limitations of linear wave theory

21.1 Perturbation Approach

Nonlinear, For our potential wave flow problem, we derived two free surface boundary conditions.
implicit free Two boundary conditions are required because we have two unknown functions: the
surface
conditions
velocity potential 𝜙(𝑥, 𝑦, 𝑧, 𝑡) and the actual position 𝜁(𝑥, 𝑦, 𝑡) of the free surface relative
to the calm water level 𝑧 = 0. Unfortunately these boundary conditions are:

• nonlinear because they contain products of the unknown functions 𝜙 and 𝜁 or

their derivatives. As a consequence the equations cannot be resolved to eliminate
one unknown function from either of the two equations,
• implicit because all terms have to be evaluated at the unknown free surface 𝑆𝐹
with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡).

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.1 Perturbation Approach 251

A general analytical solution of the nonlinear problem does not exist. Therefore, we
will linearize the free surface boundary conditions based on a perturbation approach.
Of course, this comes at a price: the solution of the linearized problem will be valid
only within certain limits, which we will discuss.
The perturbation approach follows the general concept of mathematical series like the Series expansion
Taylor series or the Fourier series. The idea is that an approximate, basic solution for the
unknown function is improved with additional terms that become smaller and smaller.
The basic solution is also called the ‘zero order’ solution and is assumed to be known.
The additional terms are of first order, second order, third order, and so on.
The unknown functions, in our case the velocity potential 𝜙 and the wave elevation
𝜁, are expressed as a power series expansion with respect to a small parameter 𝜀. The
perturbation series for our two unknowns are:

𝜙 = 𝜙(0) + 𝜀𝜙(1) + 𝜀2 𝜙(2) + 𝜀3 𝜙(3) + … (.)

𝜁 = 𝜁 (0)
+ 𝜀𝜁 (1) 2 (2)
+𝜀 𝜁 +𝜀 𝜁3 (3)
+… (.)

The superscripts (0), (1), (2), etc. express the order. 𝜙(𝓁) denotes a velocity potential
of 𝓁 𝑡ℎ -order and 𝜁 (𝓁) an 𝓁 𝑡ℎ -order wave elevation. The perturbation series may more
concisely be written as the sums:

∑
∞
𝜙(𝑥, 𝑦, 𝑧, 𝑡, 𝜀) = 𝜀𝓁 𝜙(𝓁) (𝑥, 𝑦, 𝑧, 𝑡),
j 𝓁=0 j
(.)
∑∞
𝜁(𝑥, 𝑦, 𝑡, 𝜀) = 𝜀𝓁 𝜁 (𝓁) (𝑥, 𝑦, 𝑡)
𝓁=0

The small parameter 𝜀 ≪ 1 (also called the perturbation parameter) can be interpreted
as a dimensionless wave steepness 𝜀 = 𝑘𝜁𝑎 = 𝐿2𝜋 𝐻2 ∼ 𝐿𝐻 . If this is too mathematical
𝑊 𝑊
for you, just think of replacing the exact solutions by a sum of potentials or wave
elevations that have fast decreasing contributions

𝜙(𝑥, 𝑦, 𝑧, 𝑡) = 𝜙(0) + 𝜙(1) + 𝜙(2) + … with 𝜙(0) > 𝜙(1) > 𝜙(2) > …
(.)
𝜁(𝑥, 𝑦, 𝑡) = 𝜁 (0) + 𝜁 (1) + 𝜁 (2) + … with 𝜁 (0) > 𝜁 (1) > 𝜁 (2) > …

Different levels of approximation (of the exact solution) are derived by truncating the
series expansion after a selected order. Series expansions with respect to 𝜀 lead to
solutions known as Stokes’ wave theories. Stokes’ wave theories up to fifth order are
used in offshore hydrodynamics.
In order to linearize the free surface boundary conditions we truncate the perturbation Linearization
series after the linear term

𝜙 = 𝜙(0) + 𝜀𝜙(1) + (2) (.)

𝜁 = 𝜁 (0)
+ 𝜀𝜁 (1)
+ (2) (.)

Furthermore, we assume the terms of second and higher order represented by (2) are
negligibly small, and we will omit them from here on.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

252 21 Linearization of Free Surface Boundary Conditions

21.2 Kinematic Free Surface Condition

Kinematic free Let us apply the perturbation approach to the kinematic free surface boundary condition
surface (.), which expresses the requirement that no particle leave the free surface.
boundary
condition 𝜕𝜁 𝜕𝜁 𝜕𝜙 𝜕𝜁 𝜕𝜙 𝜕𝜙
− − − + = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧
Another interpretation is that the velocity of a particle normal to the surface has to be
equal to the velocity of the surface in that direction.
Simplify notation Before we proceed, we rewrite the boundary condition using subscripts to indicate
partial differentiation.

− 𝜁𝑡 − 𝜁𝑥 𝜙𝑥 − 𝜁𝑦 𝜙𝑦 + 𝜙𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)

Substitute Now we replace the unknown functions 𝜙 and 𝜁 with their linear perturbation approxi-
perturbations mations.
into boundary
condition ( ) ( ) ( )
− 𝜁 (0) + 𝜀𝜁 (1) 𝑡 − 𝜁 (0) + 𝜀𝜁 (1) 𝑥 𝜙(0) + 𝜀𝜙(1) 𝑥
( ) ( ) ( )
− 𝜁 (0) + 𝜀𝜁 (1) 𝑦 𝜙(0) + 𝜀𝜙(1) 𝑦 + 𝜙(0) + 𝜀𝜙(1) 𝑧 = 0
on 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

j Note that we apply the perturbation also to the position where we satisfy the boundary j
condition.
The differential operators are applicable to the individual terms, for instance:
( (0) )
𝜁 + 𝜀𝜁 (1) 𝑡 = 𝜁𝑡(0) + 𝜀𝜁𝑡(1)

Likewise for the other derivatives. The perturbation parameter 𝜀 is treated as a constant
factor. We first perform the differentiations within the parentheses and then expand
the products. Equation (.) transforms into

⁓≈ 0
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0) (0) (1) (1) (0)
𝜀2
𝜁𝑥(1)𝜙(1)

𝑥 − 𝜁𝑥 𝜀𝜙𝑥 − 𝜀𝜁𝑥 𝜙𝑥 −  𝑥
⁓≈ 0
− 𝜁𝑦(0) 𝜙(0) (0) (1) (1) (0)
𝜀2 𝜁𝑦(1)𝜙
(1)

𝑦 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦 −  𝑦
+ 𝜙(0) (1)
𝑧 + 𝜀𝜙𝑧 = 0
on 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

Neglect higher The products of derivatives of the first order potential 𝜙(1) and the first order wave
order elevation 𝜁 (1) result in second order terms as indicated by the factor 𝜀2 . Since we
contributions
neglected second order contribution already in our perturbation approximation, it is
only logical to neglect the newly created second order terms as well.
Equation (.) is now linear if we consider the zero order functions and their derivatives
known. There are no longer products of the unknown first order functions 𝜙(1) and 𝜁 (1)
or their derivatives with each other. This is also reflected in the fact that there are only
terms left that have a factor 𝜀 of one.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.2 Kinematic Free Surface Condition 253

We are not done yet, however. The boundary condition still has to be satisfied at the Not done yet
unknown position 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) . This affects the derivatives of both zero and first
order potentials. In order to remedy this problem, the potentials are developed in a
Taylor series with respect to the zero order wave elevation. The 𝑥-derivative of the zero
order potential may serve as an example.

𝜕𝜙(0) || 1( ) 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝑧, 𝑡) = | + 𝑧 − 𝜁 (0) |
𝜕𝑥 ||𝑧=𝜁 (0) 1! 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

1( )2 𝜕 3 𝜙(0) ||
+ 𝑧 − 𝜁 (0) | + (3) (.)
2! 𝜕𝑥𝜕𝑧2 ||𝑧=𝜁 (0)

Of course, all terms of order two and higher will be neglected.

𝜕𝜙(0) || ( ) 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝑧, 𝑡) = | + 𝑧 − 𝜁 (0) | (.)
𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

At the free surface 𝑆𝐹 we have 𝑧 = 𝜁. Together with the linear perturbation we obtain
( )
𝑧 = 𝜁 = 𝜁 (0) + 𝜀𝜁 (1) or 𝑧 − 𝜁 (0) = 𝜀𝜁 (1) (.)

Applying this result to the Taylor series expansion of the zero order potential yields Taylor series
approximation
j j
𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑥 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)

The same Taylor series expansion is applied to all other potentials and derivatives in
Equation (.):

𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑦 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑦 ||𝑧=𝜁 (0) 𝜕𝑦𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(0) || 𝜕 2 𝜙(0) ||
𝜙(0)
𝑧 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑧 ||𝑧=𝜁 (0) 𝜕𝑧2 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 2 (1) |
(1) 𝜕 𝜙 |
𝜙(1) (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 | (.)
𝑥 𝜕𝑥 ||𝑧=𝜁 (0) 𝜕𝑥𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 𝜕 2 𝜙(1) ||
𝜙(1)
𝑦 (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 (1) | (.)
𝜕𝑦 ||𝑧=𝜁 (0) 𝜕𝑦𝜕𝑧 ||𝑧=𝜁 (0)
𝜕𝜙(1) || 2 (1) |
(1) 𝜕 𝜙 |
𝜙(1) (𝑥, 𝑦, 𝜁 , 𝑡) = | + 𝜀𝜁 | (.)
𝑧 𝜕𝑧 ||𝑧=𝜁 (0) 𝜕𝑧2 ||𝑧=𝜁 (0)

The first and second order derivatives of the potentials on the right-hand side are now
evaluated at the known zero order wave elevation 𝜁 (0) .
The right-hand sides of Equations (.) through (.) replace the corresponding
terms in Equation (.). We again revert to subscripts for derivatives and omit the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

254 21 Linearization of Free Surface Boundary Conditions

function arguments, but note that all potential derivatives are evaluated at 𝑧 = 𝜁 (0)
from here on.
( )
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0)
𝑥 + 𝜀𝜁 𝜙𝑥𝑧
(1) (0)

( ) ( )
− 𝜁𝑥(0) 𝜀 𝜙(1)
𝑥 + 𝜀𝜁 (1) (1)
𝜙 𝑥𝑧 − 𝜀𝜁 (1)
𝑥 𝜙 (0)
𝑥 + 𝜀𝜁 (1) (0)
𝜙 𝑥𝑧
( ) ( )
− 𝜁𝑦(0) 𝜙(0) (1) (0) (0)
𝑦 + 𝜀𝜁 𝜙𝑦𝑧 − 𝜁𝑦 𝜀 𝜙𝑦 + 𝜀𝜁 𝜙𝑦𝑧
(1) (1) (1)

( ) ( )
− 𝜀𝜁𝑦(1) 𝜙(0)𝑦 + 𝜀𝜁 (1) (0)
𝜙 𝑦𝑧 + 𝜙 (0)
𝑧 + 𝜀𝜁 (1) (0)
𝜙 𝑧𝑧
( )
+ 𝜀 𝜙(1)𝑧 + 𝜀𝜁 𝜙𝑧𝑧
(1) (1)
= 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)

As a result, the kinematic boundary condition is now enforced at the known zero order
wave elevation instead of the unknown exact wave elevation.
Linearized Expanding the products results in five additional second order terms which we conse-
kinematic free quently neglect. Finally, we have morphed the nonlinear kinematic boundary condition
surface
boundary
into its linearized form.
condition
− 𝜁𝑡(0) − 𝜀𝜁𝑡(1) − 𝜁𝑥(0) 𝜙(0) (0) (1) (0) (0) (1) (1) (0)
𝑥 + 𝜁𝑥 𝜀𝜁 𝜙𝑥𝑧 − 𝜁𝑥 𝜀𝜙𝑥 − 𝜀𝜁𝑥 𝜙𝑥
− 𝜁𝑦(0) 𝜙(0) (0) (1) (0) (0) (1) (1) (0)
𝑦 + 𝜁𝑦 𝜀𝜁 𝜙𝑦𝑧 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦
+ 𝜙(0) (1) (0) (1)
𝑧 + 𝜀𝜁 𝜙𝑧𝑧 + 𝜀𝜙𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)
j j
Only the first order wave elevation 𝜁 (1) and the first order velocity potential 𝜙(1) are
unknown in Equation (.). The equation is linear because it no longer contains
products or powers of the unknown functions or their derivatives. At most a first order
function is multiplied with known zero order functions.
Zero order What is left is the selection of an appropriate zero order solution! However, before we
solution do so, the same linearization process is applied to the dynamic free surface boundary
condition. The zero order solution is introduced in Section ..

21.3 Dynamic Free Surface Condition

The dynamic free surface condition followed from the unsteady Bernoulli equation and
expresses the requirement that the pressure is to be constant over the free surface 𝑆𝐹 .
𝜕𝜙 1
+ |∇ 𝜙|2 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
𝜕𝑡 2
The same equation reads as follows if again subscripts are used to represent partial
differentiation.
[ ]
1 2
𝜙𝑡 + 𝜙𝑥 + 𝜙2𝑦 + 𝜙2𝑧 + 𝑔𝑧 = 0 at 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑦, 𝑡) (.)
2

Subsequently, the same process is followed as was used for the kinematic free surface
boundary condition. Only major way points are provided here and the step by step
execution is left as a self study problem (see page ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.3 Dynamic Free Surface Condition 255

. Introduce the linear perturbations Linear

perturbation
𝜙 = 𝜙(0) + 𝜀𝜙(1) (.)
𝜁 = 𝜁 (0) + 𝜀𝜁 (1) (.)

into Equation (.).

. Take derivatives and expand all products.

. Delete all terms proportional to 𝜀2 (second order, negligibly small)! As a result, Neglect higher
we obtain for the dynamic free surface condition the following expression: order terms

[ ]
1 ( (0) )2 ( (0) )2 ( (0) )2
𝜙(0) (1)
𝑡 + 𝜀𝜙𝑡 + 𝜙𝑥 + 𝜙𝑦 + 𝜙𝑧
2
+ 𝜀𝜙𝑥 𝜙𝑥 + 𝜀𝜙(0)
(0) (1) (1) (0) (1)
𝑦 𝜙𝑦 + 𝜀𝜙𝑧 𝜙𝑧 + 𝑔𝜁
(0)
+ 𝑔𝜀𝜁 (1) = 0
at 𝑆𝐹 with 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) (.)

. The derivatives of the potential still have to be computed at the unknown free Taylor series
surface location 𝑧 = 𝜁 (0) + 𝜀𝜁 (1) . The derivatives are, once again, approximated expansion
by the truncated Taylor series approximations (.) through (.) and corre-
sponding Taylor series for the time derivatives 𝜙(0) (1)
𝑡 and 𝜙𝑡 of the zero and first
order potential respectively. The dynamic free surface condition is now satisfied
j at the zero order wave elevation 𝑧 = 𝜁 (0) . j

. After renewed expansion of products and the omission of second order terms we Linearized
have as linearized dynamic free surface boundary condition: dynamic free
surface condition
[ ]
1 ( (0) )2 ( (0) )2 ( (0) )2
𝜙(0) (1) (0) (1)
𝑡 + 𝜀𝜁 𝜙𝑡𝑧 + 𝜀𝜙𝑡 + 𝜙𝑥 + 𝜙𝑦 + 𝜙𝑧
2
+ 𝜙(0)
𝑥 𝜀𝜁 (1) (0)
𝜙 (0) (1) (0) (0) (1) (0)
𝑥𝑧 + 𝜙𝑦 𝜀𝜁 𝜙𝑦𝑧 + 𝜙𝑧 𝜀𝜁 𝜙𝑧𝑧
+ 𝜀𝜙(0) (1) (0) (1) (0) (1)
𝑥 𝜙𝑥 + 𝜀𝜙𝑥 𝜙𝑦 + 𝜀𝜙𝑥 𝜙𝑥 + 𝑔𝜁
(0)
+ 𝑔𝜀𝜁 (1) = 0
at 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)

Admittedly, the linearized three-dimensional kinematic and dynamic boundary condi-

tions for the free surface, (.) and (.) respectively, look even more complicated
than their nonlinear equivalents. However, the linearized forms have two distinct
advantages:

• The unknown first order wave elevation and potential are multiplied with known
constants or zero order functions only.

• The conditions are applied at the known location of the zero order wave elevation
instead of the exact position 𝜁(𝑥, 𝑦, 𝑡).

Finally, a suitable pair of zero order wave elevation and velocity has to be selected to
formulate the linearized boundary value problem already introduced in Section ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

256 21 Linearization of Free Surface Boundary Conditions

21.4 Linearized Free Surface Conditions for Waves

Zero order The zero order potential 𝜙(0) and zero order wave elevation 𝜁 (0) were introduced as
solution must be basic solutions to the problem which we assume to know. In fact, we will select a basic
physical
solution. A valid basic solution has to be physical, i.e. it does not violate the underlying
conditions of the problem.
Simplest zero It was mentioned earlier that linear wave theory is only valid for waves with small
order solution steepness
wave height 𝐻
= ≪ 1 (.)
wave length 𝐿𝑤
For diminishing wave steepness the disturbance of the free surface eventually becomes
invisible. If the surface is not moving, the fluid underneath will not move either (we
ignore currents). Thus, a suitable set of zero order solutions is the trivial case of no
wave elevation and no fluid velocity, or in other words calm water. We select for the
zero order potential:

𝜙(0) (𝑥, 𝑦, 𝑧, 𝑡) ≡ 0 (.)

and for the zero order wave elevation:

𝜁 (0) (𝑥, 𝑦, 𝑡) ≡ 0 (.)

j Obviously, derivatives of the zero order solutions vanish as well. j

Simplifying the First, the zero order solutions are substituted into the linearized kinematic free surface
kinematic condition.
boundary
condition
=0
⌃ = 0(0) (1) ⁓ = 0(0) (1⁓ = 0 (1)(0)
⁓= 0
− 𝜁𝑡(0)
 − 𝜀𝜁 (1) − 𝜁 (0) (0)
 
(0)
𝑡 𝜙 +
𝜁𝑥𝜀𝜁𝜙𝑥𝑧 − 

𝜁𝑥𝜀𝜙𝑥 ) − 
𝜀𝜁
𝑥 𝜙𝑥


𝑥 𝑥

⌃ = 0(0) (1) ⌃ = 0 (1) (0)
⁓ = 0(0) (1)
 ⌃= 0

− 𝜁𝑦(0) (0) (0)
 
𝜙 𝑦 + 𝜙
𝜁𝑦 𝜀𝜁 𝑦𝑧 − 𝜁𝑦 𝜀𝜙𝑦 − 𝜀𝜁𝑦 𝜙𝑦

  
=0
⁓= 0
+ 𝜙(0) (1)(0)
𝜙𝑧𝑧 + 𝜀𝜙(1) on 𝑆𝐹 with 𝑧 = 𝜁 (0) (.)


𝑧 + 𝑧 = 0
 𝜀𝜁

Astonishingly, only two terms remain:

− 𝜀𝜁𝑡(1) + 𝜀𝜙(1)
𝑧 = 0 on 𝑆𝐹 with 𝑧 = 𝜁 (0) = 0 (.)

Kinematic free At this point we may drop the perturbation parameter and omit the superscript for
surface condition the first order classification. The first order wave elevation and the first order velocity
for linear wave
theory
potential are the only nonvanishing components of the solution left.
𝜕𝜁 𝜕𝜙
− + = 0 on 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧

Physical Let us take a step back after this long and tedious mathematical derivation. The deriva-
interpretation tive of the potential with respect to 𝑧 is the vertical component 𝑤 of the fluid velocity
𝑣. The time derivative of the wave elevation 𝜕𝜁∕𝜕𝑡 is the velocity of the free surface in

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

21.4 Linearized Free Surface Conditions for Waves 257

𝑧-direction. As originally claimed, the kinematic free surface condition enforces that
the relative velocity between free surface and fluid particles vanishes. In the case of
the linear kinematic free surface condition, the slope of the free surface is ignored. In
fact, it is treated as if it is flat. Therefore, the linear kinematic condition is valid only for
waves with small steepness 𝐻∕𝐿𝑤 .
Applying the selected zero order solution and its derivatives to the dynamic free surface Simplifying the
condition (.) yields: dynamic
boundary
=0 =0 [ ] condition
=(0 ) =(0 ) =0
1 (  )
2 2 2
𝜙(0) (1) (0) (1) (0)
⌃ (0)⌃ (0)⌃
𝑡 + 𝜀𝜁 𝜙𝑡𝑧 + 𝜀𝜙𝑡 + 𝜙 + 𝜙 + 𝜙
   
2 𝑥 𝑦 𝑧

⁓ = 0 (0) (1)
 ⁓ = 0 (0) (1) ⁓= 0

𝜙(0) (1) (0) 𝜙(0) (0)
  𝜙
+𝑥𝜀𝜁 𝜙𝑥𝑧 +  𝜙𝑦𝜀𝜁 𝑦𝑧 + 
𝜙𝑧𝜀𝜁 𝑧𝑧

⁓ = 0 (0)(1)
(0)(1) ⁓= 0  ⁓ = 0 (1)
𝜀𝜙(0)
(1)

  

+ 𝑥 𝜙𝑥 + 
𝜀𝜙 𝑥 𝜙𝑦 + 
𝜀𝜙 
𝑥 𝜙𝑥 + 𝑔𝜀𝜁 = 0

at 𝑆𝐹 with 𝑧 = 𝜁 (0) = 0 (.)

What remains is
𝜀𝜙(1)
𝑡 + 𝑔𝜀𝜁
(1)
= 0 at 𝑆𝐹 with 𝑧 = 0 (.)

As for the linear kinematic free surface condition, the perturbation parameter and the Dynamic free
superscript for first order may be dropped knowing that the first order wave elevation surface condition
j for linear wave j
and potential are the only substantial parts of the solution. theory
𝜕𝜙
+ 𝑔𝜁 = 0 at 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡
Note that in linear wave theory the free surface conditions are satisfied at the calm
water level 𝑧 = 0.
In Section (.), we introduced the linearized Bernoulli equation (.) Physical
interpretation
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
The dynamic boundary condition requires that the pressure is equal everywhere at the
free surface with 𝑧 = 𝜁. Therefore, the pressure difference must vanish 𝑝 − 𝑝∞ = 0 for
𝑧 = 𝜁.
𝜕𝜙
0 = −𝜌 − 𝜌 𝑔𝜁 (.)
𝜕𝑡
Dividing the linear Bernoulli equation applied to the free surface by the negative fluid
density −𝜌 results in the linear dynamic free surface condition (.). Here we neglect
the influence of the nonlinear dynamic pressure term 1∕2𝜌|𝑣|2 , which we assume is
small.
The first order wave elevation is derived from the dynamic free surface boundary First order wave
condition (.). elevation
𝜕𝜙 ||
𝜁(𝑥, 𝑦, 𝑡) = −𝑔 (.)
𝜕𝑡 ||𝑧=0
In long-crested waves, the wave elevation is only a function of 𝑥 and time 𝑡.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

258 21 Linearization of Free Surface Boundary Conditions

Self Study Problems

. Linearize the following differential equation, assuming the zero order basis solu-
tion is linear 𝑓 (0) = 𝑎 ⋅ 𝑧. with 𝑎 being a constant factor.

𝜕𝑓 𝜕𝑓 𝜕2𝑓
+ 𝜙 = 0
𝜕𝑥 𝜕𝑧 𝜕𝑧2
Hint: in this case the Taylor series expansion around a specific 𝑧 value is not
needed.
. Linearize the dynamic free surface condition by executing in full detail the steps
 through  on page .

. Assume steady flow (all time derivatives vanish) and simplify the linearized
kinematic and dynamic free surface boundary conditions (.) and (.) with
respect to the following zero order solutions (Neumann-Kelvin linearization):

𝜙(0) = −𝑈 𝑥 parallel flow in negative 𝑥-direction

𝜁 (0) = 0 calm water

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

259

22
Linear Wave Theory

After thoroughly explaining the mathematical model for long-crested waves, it is time
to actually solve the boundary value problem. In this chapter the linearized boundary
value problem is solved for the two-dimensional case of long-crested waves progressing
in positive 𝑥-direction. Again, the separation of variables is applied to derive the now
time dependent velocity potential. Equipped with the potential, we will study the
behavior and properties of waves.

Learning Objectives

j At the end of this chapter students will be able to: j

• apply potential theory to water waves
• solve potential flow problems
• compute the wave number
• distinguish exact formulation from deep water approximation

22.1 Solution of Linear Boundary Value Problem

The linearized boundary value problem for long-crested waves has been introduced in Mathematical
Section .. Chapter  provides a detailed account of the linearization process. A model
recaptured
solution to the boundary value problem (.) (repeated below) will yield a velocity
potential that describes the unsteady flow in waves with small wave steepness.

Laplace equation (conservation of mass)

𝜕2𝜙 𝜕2𝜙
𝚫𝜙(𝑥, 𝑡) = + = 0 𝑥 = (𝑥, 𝑧)𝑇 ∈ 𝑉 (.)
𝜕𝑥 2 𝜕𝑧2
generalized linear free surface boundary condition

𝜕2𝜙 𝜕𝜙
+𝑔 = 0 𝑥 with 𝑧 = 0 (.)
𝜕𝑡2 𝜕𝑧

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

260 22 Linear Wave Theory

Figure 22.1 Simplified two-dimensional fluid domain for long-crested regular waves

ocean bottom boundary condition

𝜕𝜙
= 0 𝑥 on 𝑆𝐵 , 𝑧 = −ℎ (.)
𝜕𝑧
far field boundary condition

j 𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.) j

The coordinate system, fluid domain, and its boundaries are depicted in Figure ..
The origin is at the calm water level 𝑧 = 0. With the 𝑧-axis pointing upwards, points in
the fluid domain 𝑉 have negative 𝑧-coordinates 𝑧 ≤ 0. Note that the linear free surface
boundary condition (.) is satisfied at the calm water level 𝑧 = 0 and not at the yet
unknown displaced free surface 𝑆𝐹 with 𝑧 = 𝜁(𝑥, 𝑡).
Separation of The velocity potential is a function of 𝑥, 𝑧, and time 𝑡. Therefore, we seek a solution
variables which represents the potential as a product of three functions 𝑋(𝑥), 𝑍(𝑧), and 𝑇 (𝑡).
Each of these is a function of just one independent variable.

𝜙(𝑥, 𝑧, 𝑡) = 𝑋(𝑥) 𝑍(𝑧) 𝑇 (𝑡) (.)

This approach is called separation of variables because it will convert the partial differ-
ential Laplace equation into a set of three simultaneous, ordinary differential equations.
The second order spatial derivatives of the potential (.) are needed for the Laplace
equation.
𝜕2𝜙 ( )
𝜕2 𝜕2𝑋
= 𝑋𝑍𝑇 = 𝑍 𝑇 = 𝑋𝑥𝑥 𝑍 𝑇
𝜕𝑥 𝜕𝑥2 𝜕𝑥2 (.)
𝜕2𝜙 ( )
𝜕2 𝜕2𝑍
= 𝑋𝑍𝑇 = 𝑋 𝑇 = 𝑋 𝑍𝑧𝑧 𝑇
𝜕𝑧 𝜕𝑧2 𝜕𝑧2
With substitution of the derivatives (.) into (.) and using subscripts to denote
differentiation, the Laplace equation becomes:

𝑋𝑥𝑥 𝑍 𝑇 + 𝑋 𝑍𝑧𝑧 𝑇 = 0 for 𝑥 = (𝑥, 𝑧)𝑇 in 𝑉 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.1 Solution of Linear Boundary Value Problem 261

Remember that the zero order potential 𝜙(0) ≡ 0 describes the fluid at rest. Therefore,
it is safe to assume that 𝑋 𝑍 ≠ 0, and we may divide Equation (.) by 𝑋𝑍:
[ ]
𝑋𝑥𝑥 𝑍𝑧𝑧
+ 𝑇 = 0 (.)
𝑋 𝑍

For this to be true at all times, the term in brackets must vanish, i.e.
𝑋𝑥𝑥 𝑍
+ 𝑧𝑧 = 0 (.)
𝑋 𝑍
The first fraction is a function of 𝑥 alone. The second fraction only depends on 𝑧. As a
consequence, the fractions will be equal for all (𝑥, 𝑧) only if they are constant in (𝑥, 𝑧)
and one is the negative of the other.

𝑋𝑥𝑥 𝑍
= − 𝑧𝑧 = const. = 𝐶 (.)
𝑋 𝑍
Equation (.) may be split into two ordinary differential equations that are connected Ordinary
by the same constant 𝐶 differential
𝑋𝑥𝑥 equations
for 𝑋(𝑥): = 𝐶
𝑋 (.)
𝑍
for 𝑍(𝑧): − 𝑧𝑧 = 𝐶
𝑍
j We select the constant to be 𝐶 = −𝑘2 with 𝑘 being the wave number. Equations (.) j
are multiplied by 𝑋 and 𝑍 respectively, and all terms are moved to the left-hand side.

𝑋𝑥𝑥 + 𝑘2 𝑋 = 0 (.)
2
𝑍𝑧𝑧 − 𝑘 𝑍 = 0 (.)

The complimentary solution of the ordinary differential equation in 𝑥 (.) is a Solution for 𝑋
function whose second order derivative is equal to the negative of the function itself,
except for a constant factor. The sine and cosine functions are of this type. However, we
do not know which of the two. Since combinations of sine and cosine are also plausible,
we define
𝑋 = 𝐶1 ei𝑘𝑥 (.)
The exponential function combines sine and cosine with Euler’s formula for complex
numbers ei𝑥 = cos 𝑥 + i sin 𝑥. The complex notation is a useful mathematical tool which
simplifies the treatment of our equations significantly. However, the velocity field is
real and only the real part of the resulting velocity potential will bear physical meaning.
Differentiating Equation (.) twice and substituting it back shows that this is indeed
a solution of ODE (.). Boundary conditions will be used later to solve for the new
constant 𝐶1 which generally differs from constant 𝐶 above.
The second ODE (.) is satisfied by functions whose second order derivative is Solution for 𝑍
equal to the function itself. This time hyperbolic sine and cosine functions will suffice.
Selecting the hyperbolic cosine, a general solution to (.) is given by

𝑍 = 𝐶2 cosh(𝑘𝑧 + 𝛼2 ) (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

262 22 Linear Wave Theory

Solution for 𝑇 Finally, a solution is needed for the time dependent function 𝑇 . Equation (.) is of no
use because it would only yield the trivial solution 𝑇 ≡ 0. Instead, the product (.) 𝜙 =
𝑋 𝑌 𝑍 is introduced into the generalized linear free surface boundary condition (.).

𝜙𝑡𝑡 + 𝑔 𝜙𝑧 = 𝑋 𝑍 𝑇𝑡𝑡 + 𝑔 𝑋 𝑍𝑧 𝑇 = 0 for 𝑧 = 0 (.)

Division by the product 𝑋 𝑍 yields

𝑍𝑧 ||
𝑇𝑡𝑡 + 𝑔 𝑇 = 0 (.)
𝑍 ||𝑧=0
The expression 𝑍𝑍 ∕𝑍 has to be evaluated at 𝑧 = 0 because Equation (.) is derived
from the free surface boundary condition. 𝑧 is the only variable in the expression 𝑍𝑍 ∕𝑍.
Consequently, the term 𝑔𝑍𝑧 ∕𝑍|𝑧=0 is constant. We select
𝑍𝑧 ||
𝑔 = 𝜔2 (.)
𝑍 ||𝑧=0
𝜔 is the wave frequency.
The ordinary differential equation for 𝑇 is of the same type as the ODE (.) for 𝑋.

𝑇𝑡𝑡 + 𝜔2 𝑇 = 0 (.)

An appropriate complimentary solution is

𝑇 = 𝐶3 e−i𝜔𝑡 (.)
j j
The minus sign is deliberate in the argument of the exponential function. We come
back to this when we discuss the far field condition.
So far, the solution for the velocity potential is using the results from Equations (.),
(.), and (.):

𝜙(𝑥, 𝑧, 𝑡) = 𝑋(𝑥) 𝑍(𝑧) 𝑇 (𝑡) = 𝐶1 𝐶2 𝐶3 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 (.)

The product of constants is yet another constant. We substitute 𝐶4 = 𝐶1 𝐶2 𝐶3 and get

𝜙(𝑥, 𝑧, 𝑡) = 𝐶4 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 (.)

Constants The remaining constants 𝐶4 and 𝛼2 in (.) are selected so that the solution 𝜙 =
𝑋 𝑌 𝑍 satisfies the remaining boundary conditions of the problem.
The ocean bottom boundary condition requires that the normal velocity vanishes at
the ocean floor.
𝜕𝜙
= 0 for 𝑧 = −ℎ (.)
𝜕𝑧
Introducing the partial solution (.) yields:
[ ]
𝜕
𝐶4 cosh(𝑘𝑧 + 𝛼2 ) ei𝑘𝑥 e−i𝜔𝑡 = 0 for 𝑧 = −ℎ (.)
𝜕𝑧

Since only the hyperbolic cosine is a function of 𝑧 and 𝐶4 ei𝑘𝑥 e−i𝜔𝑡 ≠ 0 this simplifies to
[ ]
𝜕
cosh(𝑘𝑧 + 𝛼2 ) = 0 for 𝑧 = −ℎ (.)
𝜕𝑧

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.1 Solution of Linear Boundary Value Problem 263

4 Figure 22.2 The hyperbolic sine and cosine functions

1
sinh(x),cosh( x)

3
sinh(x)
cosh(x)
4
3 2 1 0 1 2 3
x

Differentiation results in

j 𝑘 sinh(𝑘𝑧 + 𝛼2 ) = 0 for 𝑧 = −ℎ (.) j

The hyperbolic sine only vanishes if its argument is zero (see Figure .). Substituting
the negative water depth for 𝑧, we get
𝑘(−ℎ) + 𝛼2 = 0 (.)
Therefore, the constant 𝛼2 is equal to
𝛼2 = 𝑘ℎ (.)

This leaves the constant 𝐶4 to be determined. Except for the far field boundary condition,
we have explicitly used all boundary conditions listed in the linearized boundary value
problem (.) through (.). The far field condition is of no use to determine 𝐶4 . One
has to remember that the generalized free surface condition (.) actually consists of
two boundary conditions, i.e. the kinematic and the dynamic free surface condition.
The linearized dynamic free surface condition (.) requires that the pressure at the
free surface remains constant.
𝜕𝜙
+ 𝑔𝜁 = 0 at 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡
Solving (.) for the unknown wave elevation 𝜁 and substituting the partial solution
for the potential (.) and the new found constant 𝛼2 from Equation (.) yields:
1 𝜕𝜙 || 1
𝜁(𝑥, 𝑡) = − = − 𝐶4 cosh(𝑘 ⋅ 0 + 𝑘ℎ) ei𝑘𝑥 (−i𝜔)e−i𝜔𝑡
𝑔 𝜕𝑡 ||𝑧=0 𝑔
i𝜔
= 𝐶 cosh(𝑘ℎ) ei(𝑘𝑥−𝜔𝑡) (.)
𝑔 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

264 22 Linear Wave Theory

Wave amplitude The exponential function represents a general harmonic oscillation of unit amplitude.
ei(𝑘𝑥−𝜔𝑡) = cos(𝑘𝑥 − 𝜔𝑡) + i sin(𝑘𝑥 − 𝜔𝑡)
Consequently, the factor preceding the harmonic function may be interpreted as an
amplitude 𝜁𝑎 , which must be real valued since the surface elevation is real.
i𝜔
𝜁𝑎 = 𝐶 cosh(𝑘ℎ) (.)
𝑔 4
Accordingly, the constant 𝐶4 must be complex and is given by
i𝑔
𝐶4 = − 𝜁 (.)
𝜔 cosh(𝑘ℎ) 𝑎

Solution for This completes the solution of the linearized boundary value problem. We found that
wave elevation the first order wave elevation is a cosine function with amplitude 𝜁𝑎 . In complex notation
the wave profile is
̄ 𝑡) = 𝜁𝑎 ei(𝑘𝑥−𝜔𝑡)
𝜁(𝑥, (.)
However, only its real part ℜ(𝜁) ̄ is of interest to us:
( )
𝜁(𝑥, 𝑡) = ℜ 𝜁(𝑥,̄ 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)

Solution for The second part of the solution is the first order velocity potential for a long-crested
velocity wave in two dimensions. Equation (.) with the results for constants 𝛼2 (.) and
potential
𝐶4 (.) becomes
j j
̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ei𝑘𝑥 e−i𝜔𝑡
𝜙(𝑥, (.)
𝜔 cosh(𝑘ℎ)
In order to compute the velocity field, we need the real valued potential 𝜙.
( )
̄ 𝑧, 𝑡)
𝜙(𝑥, 𝑧, 𝑡) = ℜ 𝜙(𝑥,
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ( i𝑘𝑥 −i𝜔𝑡 )
= − 𝑎 ℜ ie e
𝜔 cosh(𝑘ℎ)
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) ( )
= − 𝑎 ℜ i cos(𝑘𝑥 − 𝜔𝑡) + i2 sin(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
which results in
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)
Note that the velocity potential (.) is valid for all water depths ℎ as long as the
general assumptions for linear wave theory are satisfied.
Wave parameters A wave on constant water depth ℎ is defined by two parameters:

• wave frequency 𝜔, and

• wave amplitude 𝜁𝑎

Wave frequency and wave number are connected via the dispersion relation (see below).
Therefore, the wave number 𝑘 may be used instead of the wave frequency 𝜔 to define
the wave.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.2 Far Field Condition Revisited 265

22.2 Far Field Condition Revisited

In the solution of the boundary value problem we hinted that the choice of e−i𝜔𝑡 (see Direction of
page ) as a basis for the time dependent part of the potential is deliberate to satisfy propagation
the far field condition.
𝑐 = const. ≥ 0 at 𝑆∞ for |𝑥| → ∞ (.)
The far field condition enforces that the wave is moving with constant velocity in positive
𝑥-direction.
The velocity of a wave is determined by how fast the crest-trough-crest contour travels. Constant phase
An observer moving as fast as the wave crest experiences a stationary wave field as we
have observed for our ship wave system. Traveling with a specific point of the wave
contour, like for instance a crest, requires a constant phase 𝜃 = 𝑘𝑥 − 𝜔𝑡 = const.
Therefore, the velocity with which a wave progresses is called phase velocity. If the
phase is to be constant in the far field, it cannot change over time and its derivative
must vanish.
d𝜃 d( )
= 𝑘𝑥 − 𝜔𝑡 = 0 (.)
d𝑡 d𝑡
Wave frequency 𝜔 and wave number 𝑘 are constants for the wave. Therefore, the phase Phase velocity
velocity of a regular wave is
d𝑥 𝜔 𝐿
𝑐 = = = 𝑤 ≥ 0 (.)
d𝑡 𝑘 𝑇
j Since both wave length 𝐿𝑤 and wave period 𝑇 are positive and constant, the phase j
velocity 𝑐 is positive and constant as well. The far field condition is satisfied.
If we had opted for e+i𝜔𝑡 instead of e−i𝜔𝑡 as the basic time function (see Equation (.)),
the condition of constant phase (.) would have resulted in 𝑐 = −𝐿𝑤 ∕𝑇 and the
wave would move in negative 𝑥-direction.

22.3 Dispersion Relation

Back on page , we stated that wave number and wave frequency are connected. The Link between
linearized kinematic free surface condition (.) provides this connection. frequency and
wave number
𝜕𝜁 𝜕𝜙
−+ = 0 on 𝑆𝐹 with 𝑧 = 0 (.)
𝜕𝑡 𝜕𝑧
Based on our linear solution for wave elevation (.) and potential (.), we form
the derivatives:
𝜕𝜁
= 𝜔𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑡

𝜕𝜙 𝜁 𝑔 sinh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 𝑎 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑧 𝜔 cosh(𝑘ℎ)
Both derivatives are evaluated at 𝑧 = 0 and substituted into the kinematic free surface
condition above.
𝜁 𝑔 sinh(𝑘ℎ)
−𝜔𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡) + 𝑘 𝑎 sin(𝑘𝑥 − 𝜔𝑡) = 0
𝜔 cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

266 22 Linear Wave Theory

The ratio of hyperbolic sine and cosine is equal to the hyperbolic tangent.
[ 𝑔 ]
𝜁𝑎 sin(𝑘𝑥 − 𝜔𝑡) −𝜔 + 𝑘 tanh(𝑘ℎ) = 0 (.)
𝜔
The latter equation is satisfied when the sine vanishes for 𝜃 = 𝑛𝜋 with 𝑛 = 0, 1, 2, … ,
or, much more generally, if the term in brackets vanishes.
[ 𝑔 ]
−𝜔 + 𝑘 tanh(𝑘ℎ) = 0 (.)
𝜔
Dispersion This may be transformed into
relation
𝜔2 = 𝑘 𝑔 tanh(𝑘ℎ) (.)
which is known as the dispersion relation. Since wave frequencies are considered
positive, one may also write
√
𝜔 = 𝑘 𝑔 tanh(𝑘ℎ) (.)
The relationship between wave frequency and wave number is a nonlinear, transcenden-
tal equation. It is influenced by the gravitational acceleration 𝑔 and the water depth ℎ.
Thus, waves will behave differently depending on the water depth. If one would consider
wave experiments on the Moon, with one sixth of Earth’s gravity, quite different wave
behavior could be expected. Obviously, the wave frequency can be readily determined
for any combination of water depth and wave number. In the next chapter, you will
j learn how to compute the wave number 𝑘 for a given water depth and wave frequency j
using the dispersion relation.
Phase velocity As stated above, the surface elevation moves with phase velocity 𝑐 = 𝜔∕𝑘 = 𝐿𝑤 ∕𝑇 .
Replacing the wave frequency 𝜔 with the dispersion relation (.) yields the phase
velocity as a function of wave length. The water depth ℎ serves as a parameter.
√ √
𝜔 𝑘 𝑔 tanh(𝑘ℎ) 𝑔 tanh(𝑘ℎ)
𝑐 = = = (.)
𝑘 𝑘 𝑘
Figure . visualizes this relationship. The primary horizontal axis is the wave number
𝑘. The lower axis shows the corresponding wave length 𝐿𝑤 . Both axes use a logarithmic
scale. Wave number is increasing from left to right. The reciprocal wave length grows
from right to left, and the vertical axis represents the phase velocity 𝑐. Note that the
vertical axis is logarithmic in the left part of Figure . but linear in the right subfigure
which shows values appropriate for model scale.
Dispersion Several important observations can be made:

• The phase velocity of waves changes with wave length (or wave frequency). As
mentioned, this effect is called dispersion. Water waves are therefore quite
different from sound and electromagnetic waves, which have constant phase
velocities that are independent of wave length (speed of sound, speed of light).
• Waves become faster with increasing wave length (decreasing wave number).
Long waves progress astonishingly fast. A wave of  m length reaches a
velocity of almost  m/s. A wave of  m length travels in deep water as fast
as a commercial air liner.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.4 Deep Water Approximation 267

j j

Figure 22.3 Wave phase velocity as function of wave number and water depth based on linear wave theory

• Depending on the water depth ℎ, waves reach a limiting phase velocity. The
lower the water depth, the lower the maximum attainable phase velocity. In very
shallow water, dispersion vanishes and waves of different wave length have the
same phase velocity.

22.4 Deep Water Approximation

Ship trial conditions usually include the requirement of ‘deep water.’ In this context,
water is deep if it does not affect wave motion at the free surface and other parts of the
flow around a ship hull. We still have to clarify which water depth is considered deep
enough.
Figure . shows the hyperbolic tangent function as it appears in the dispersion relation.
For large arguments 𝑘ℎ ⟶ ∞, the hyperbolic tangent asymptotically approaches the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

268 22 Linear Wave Theory

1.2
lim [tanh(kh)]=1
kh
1.0

kh tanh(kh)
0.8
3.000000.99505
3.141590.99627
tanh(kh)

0.6 4.000000.99933
5.000000.99991
10.000001. 00000
0.4

0.2

0.0
0 1 2 3 4 5
kh

Figure 22.4 The positive arm of the hyperbolic tangent function

value one. For a value of 𝑘ℎ = 𝜋, the deviation is only .% from the asymptotic
limit . For 𝑘ℎ = 5, the difference is less than .% and for 𝑘ℎ = 10, the tangent is
smaller than one by less than 1.0 ⋅ 10−8 .
j j
Deep water For practical purposes, water is considered deep if:
condition
𝑘ℎ ≥ 𝜋

Under this premise, we obtain the following relationship between wave length and
water depth:
2𝜋
𝑘ℎ = ℎ ≥ 𝜋
𝐿𝑤
which is equivalent to
𝐿𝑤
ℎ ≥ or 𝐿𝑤 ≤ 2 ℎ (.)
2
Thus, if the water depth is larger than half the wave length, the influence of the ocean
floor on a wave becomes negligible.
Deep water wave If 𝑘ℎ is large enough, the hyperbolic tangent will be close to one (tanh(𝑘ℎ) ≈ 1) and the
number dispersion relation (.) becomes
√ 𝜔2
𝜔 = 𝑘0 𝑔 or 𝑘0 = (.)
𝑔

𝑘0 is the deep water wave number.

Deep water With the deep water wave number, phase velocity (.) simplifies to
phase velocity
√
𝑔
𝑐0 = (.)
𝑘0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

22.4 Deep Water Approximation 269

Therefore, the relationships between wave length and phase velocity are
√
𝑔 𝐿𝑤 2𝜋 2
𝑐0 = or 𝐿𝑤 = 𝑐 on deep water (𝑘ℎ ≥ 𝜋) (.)
2𝜋 𝑔 0

The assumption of deep water simplifies the velocity potential as well. The only term Deep water
in the velocity potential (.) (or (.) for the complex form) that depends on the velocity
potential
water depth is the ratio of hyperbolic cosines.
( )
cosh 𝑘(𝑧 + ℎ)
cosh(𝑘ℎ)
The hyperbolic functions may be replaced by a special combination of exponential
functions (Abramowitz and Stegun, , p. ).
1 𝑥
sinh(𝑥) = (e − e−𝑥 ) (.)
2
1
cosh(𝑥) = (e𝑥 + e−𝑥 ) (.)
2
Using the latter for our ratio of hyperbolic cosines yields:
( )
cosh 𝑘(𝑧 + ℎ) e𝑘𝑧+𝑘ℎ + e−(𝑘𝑧+𝑘ℎ) e𝑘𝑧 e𝑘ℎ + e−𝑘𝑧 e−𝑘ℎ
= = (.)
cosh(𝑘ℎ) e𝑘ℎ + e−𝑘ℎ e𝑘ℎ + e−𝑘ℎ

j If the product of wave number and water depth becomes very large, i.e. 𝑘ℎ → ∞, then j
the exponential of (−𝑘ℎ) will vanish e−𝑘ℎ → 0. Therefore, the limit of the ratio of
hyperbolic cosines for deep water is equal to the exponential function with argument
𝑘𝑧.
( ) ⌃0
cosh 𝑘(𝑧 + ℎ) e𝑘𝑧 e𝑘ℎ + e−𝑘𝑧e−𝑘ℎ
 
e𝑘𝑧 e𝑘ℎ
lim = lim = lim = e𝑘0 𝑧 (.)
𝑘ℎ→∞ cosh(𝑘ℎ) 𝑘ℎ→∞ 0 𝑘ℎ→∞ e𝑘ℎ
e−𝑘ℎ
e𝑘ℎ +  ⌃


Finally, the linear velocity potential for waves in deep water becomes

̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 e𝑘0 𝑧 ei𝑘0 𝑥 e−i𝜔𝑡

𝜙(𝑥, (complex) (.)
𝜔
𝜁𝑎 𝑔 𝑘 𝑧
𝜙(𝑥, 𝑧, 𝑡) = e 0 sin(𝑘0 𝑥 − 𝜔𝑡) (real) (.)
𝜔
This simplification was very important in the pre-computer age. With modern desktop
computers, it is better to work with the exact solution for the velocity potential (.),
which is valid for all water depths.

References
Abramowitz, M. and Stegun, I., editors (). Handbook of mathematical functions
with formulas, graphs, and mathematical tables. Number  in Applied Mathemat-
ics Series. National Bureau of Standards, Washington, DC. Tenth printing, with
corrections.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

270 22 Linear Wave Theory

Self Study Problems

. Summarize the assumptions that have been made in deriving the linearized
boundary value problem (.) through (.).
. Show that the first order velocity potential for a harmonic wave

𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = sin(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
satisfies the Laplace equation (.) and boundary conditions (.), (.), and
(.).

. At which water depth ℎ is a wave of length 𝐿𝑤 = 80 m considered a deep water

wave?
. Compute wave number and wave frequency for the wave of length 𝐿𝑤 = 80 m in
deep water.
. You have two waves of different wave length: 𝐿𝑤1 = 100 m and 𝐿𝑤2 = 120 m.
Which wave has the greater phase velocity (assume deep water)? Is this still true
for a water depth of 20 m? Use Figure . to justify your answer.
. A ship is producing a transverse wave of wave length 𝐿𝑤 = 120 m. Assuming
deep water, what is the ship’s speed? Use Figure . to check if your answer is
j reasonable. j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

271

23
Wave Properties

Based on the linear wave theory solution, basic kinematics and dynamics of water
waves are investigated in this and the following chapter. First, a method is introduced
to solve the nonlinear dispersion relation. With the wave number known, wave par-
ticle velocities, accelerations, motions, and unsteady pressure are discussed. A more
comprehensive analysis of water wave properties may be found in offshore engineering
books by Dean and Dalrymple () and by Clauss et al. ().

Learning Objectives

j At the end of this chapter students will be able to: j

• understand wave kinematics and dynamics

• solve the dispersion relation

• apply potential flow theory results

• compute wave properties

23.1 Linear Wave Theory Results

This section briefly summarizes the most important results of Chapter  for the reader
who skipped the finer details of linear wave theory. The results form a basis for the
discussion of wave properties.
The solution, which satisfies the Laplace equation and the linearized boundary con- Regular waves
ditions for long-crested progressing waves, predicts a sinusoidal wave elevation with
amplitude 𝜁𝑎 . Waves with a sinusoidal profile are commonly referred to as regular waves
of wave frequency 𝜔.

̄ 𝑡) = 𝜁𝑎 ei(𝑘𝑥−𝜔𝑡)
𝜁(𝑥, (complex)
(.)
𝜁(𝑥, 𝑡) = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (real)

Although the discussion here uses almost exclusively the real valued equations, it should
be pointed out that the complex notation has many advantages in computational tools.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

272 23 Wave Properties

For example, the superposition of regular waves of different frequencies, amplitudes,

and phases into more realistic looking wave patterns is much easier in the complex
than the real domain.
Velocity The basis for all analysis of kinematics and kinetics of regular waves is the velocity
potential potential:
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ) i𝑘𝑥 −i𝜔𝑡
̄ 𝑧, 𝑡) = −i
𝜙(𝑥, e e (complex)
𝜔 cosh(𝑘ℎ)
(.)
𝜁 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
𝜙(𝑥, 𝑧, 𝑡) = 𝑎 sin(𝑘𝑥 − 𝜔𝑡) (real)
𝜔 cosh(𝑘ℎ)
Equation (.) represents the exact solution of the linearized boundary value problem
and is valid for all water depths ℎ.
Dispersion From the kinematic boundary condition follows the dispersion relation, which expresses
relation the relationship between wave frequency 𝜔, wave number 𝑘, and water depth ℎ.
√
𝜔 = 𝑘 𝑔 tanh(𝑘ℎ) (.)

Deep water Influence of the sea bottom on the wave motion becomes negligibly small for water
approximation depths larger than half the wave length, i.e. ℎ > 𝐿𝑤 ∕2. As shown in Section ., the
velocity potential simplifies to the following expression under the assumption that
𝑘ℎ > 𝜋.
j ̄ 𝑧, 𝑡) = − 𝜁𝑎 i 𝑔 e𝑘0 𝑧 ei𝑘0 𝑥 e−i𝜔𝑡
𝜙(𝑥, (complex) j
𝜔 (.)
𝜁 𝑔
𝜙(𝑥, 𝑧, 𝑡) = 𝑎 e𝑘0 𝑧 sin(𝑘0 𝑥 − 𝜔𝑡) (real)
𝜔
The corresponding deep water dispersion relation states

𝜔2 = 𝑘0 𝑔 (.)

Based on the assumptions made during the solution process, the equations above
are valid for fluids with negligible viscosity and constant density as long as the wave
steepness 𝐻∕𝐿𝑤 ≪ 1 is small.

23.2 Wave Number

Waves are often specified by providing values for wave period 𝑇 and wave amplitude 𝜁𝑎 .
From the wave period follows the wave frequency 𝜔 = 2𝜋∕𝑇 . For deep water (𝑘ℎ > 𝜋),
the wave number 𝑘0 follows immediately from Equation (.).

𝜔2
𝑘0 = (.)
𝑔
A wave of period 𝑇 = 8 s has a frequency of

2𝜋 6.2832
𝜔 = ≈ = 0.7854 s−1
𝑇 8s

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.2 Wave Number 273

Therefore, its deep water wave number is

𝜔2 0.78542 s−2
𝑘0 = ≈ = 0.06288 m−1
𝑔 9.81 m s−2
Its associated wave length is 𝐿𝑤 = 2𝜋∕𝑘0 = 99.923 m.
In the case of restricted water depth ℎ, the nonlinear dispersion relation (.) must
be solved. Prior to the age of programmable calculators and desktop computers this
was a cumbersome task, but with today’s computer support it is no longer a hurdle.
The method of choice to solve Equation (.) for the wave number is the Newton- Newton-
Raphson method (see for example Akai, ). Like all methods for nonlinear problems, Raphson
method
it is iterative, i.e. it starts with an initial, approximate solution and improves the solution
up to the desired accuracy. The Newton-Raphson method follows from a Taylor series
expansion around a known solution 𝑘𝑖 . The series is truncated after the linear term and
higher order terms are ignored.
( ) d𝑓 |
𝑓 (𝑘𝑖+1 ) = 𝑓 (𝑘𝑖 ) + 𝑘𝑖+1 − 𝑘𝑖 | (.)
d𝑘 ||𝑘=𝑘𝑖

The function 𝑓 in Equation (.) represents the nonlinear equation to be solved written
in an homogeneous form. We square Equation (.) and shift all terms to the left
side.
j 𝜔2 j
0 = − 𝑘 tanh(𝑘ℎ)
𝑔

Thus, the function 𝑓 (𝑘) and its derivative are

𝜔2
𝑓 (𝑘) = − 𝑘 tanh(𝑘ℎ) (.)
𝑔
d𝑓 ℎ
= −𝑘 − tanh(𝑘ℎ) (.)
d𝑘 cosh2 (𝑘ℎ)

The solution 𝑘 we seek satisfies the condition 𝑓 (𝑘) = 0. Consequently, the left-hand
side of the truncated Taylor series expansion vanishes if we assume that 𝑘 = 𝑘𝑖+1 is the
desired solution.
( ) d𝑓 |
𝑓 (𝑘𝑖+1 ) = 0 = 𝑓 (𝑘𝑖 ) + 𝑘𝑖+1 − 𝑘𝑖 | (.)
d𝑘 ||𝑘=𝑘𝑖

The remainder of Equation (.) is solved for 𝑘𝑖+1 . Iterative

procedure
𝑓 (𝑘𝑖 )
𝑘𝑖+1 = 𝑘𝑖 − (.)
d𝑓 |
|
d𝑘 |𝑘=𝑘𝑖

Equation (.) represents an iterative procedure. Repeated application will converge

toward the real solution in a few steps as long as the derivative in the denominator does
not vanish, i.e. d𝑓 ∕d𝑘 ≠ 0. The deep water wave number 𝑘0 is readily available as a

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

274 23 Wave Properties

starting value. The iterations are stopped when the difference between two subsequent
approximations is smaller than a tiny threshold value 𝜀.

|𝑘𝑖+1 − 𝑘𝑖 | ≤ 𝜀 (.)

Application The following example may serve as an illustration of the procedure. We want to know
example the wave number 𝑘 and wave length 𝐿𝑤 of a wave with period 𝑇 = 8 s when it travels in
water of depth ℎ = 20 m. Consequently, the wave frequency is 𝜔 = 0.7854 s−1 and the
deep water wave number is 𝑘0 = 0.0628797 m−1 . Equipped with 𝑘0 as the starting value,
we compute 𝑓 (𝑘0 ) = 0.0094068 m−1 and the derivative at 𝑘0 as d𝑓 ∕d𝑘 = −1.1985276.
These values are substituted into Equation (.) to derive 𝑘1 = 0.0707284 m−1 . For
the next iteration 𝑖 = 1, the values of 𝑓 (𝑘1 ) and its derivative are computed. Application
of Equation (.) results in 𝑘2 and so on. After 𝑖 = 2 no changes in the first seven
digits can be seen, and 𝑘 = 0.0707624 m−1 is considered the result for the wave number
with sufficient precision for all practical purposes. Values used in the iteration are
summarized in the table below:

𝑖 𝑘𝑖 𝑓 (𝑘𝑖 ) d𝑓 ∕d𝑘 |𝑘𝑖

[−] [m−1 ] [m−1 ] [−]
 0.0628797 0.0094068 −1.1985276
 0.0707284 0.0000404 −1.1864226
 0.0707624 0.0000000 −1.1863489
j j
 0.0707624 −0.0000000 −1.1863489
 0.0707624 0.0000000 −1.1863489
… … … …

As one can see, the iteration converges very fast. Usually less than a handful of iterations
will yield a wave number with high accuracy. The wave length 𝐿𝑤 derived from the
new wave number is 𝐿𝑤 = 2𝜋∕𝑘 = 88.793 m which is considerably less than the wave
length on deep water (𝐿𝑤 = 99.923 m). Note that the length of a wave decreases with
decreasing water depth for constant wave period.
Solutions of nonlinear equations should be thoroughly checked.

(i) Does the solution satisfy the equation? In our case the solution clearly satisfies
𝑓 (𝑘) = 𝑓 (0.0707624) = 0 within numerical precision.
(ii) You may also perform a graphical check by plotting the function 𝑓 (𝑘) in the
vicinity of a solution. Figure . shows the function 𝑓 (𝑘) using the input data
provided in the example. The intersection of 𝑓 (𝑘) with the 𝑘-axis falls on the value
𝑘 = 0.0707624 m−1 as expected.

Therefore, the solution 𝑘 = 0.0707624 m−1 makes sense.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.3 Water Particle Velocity and Acceleration 275

0.06
water depth h =20.00m
[1/m]

wave period T =8.000s

0.04 wave frequency =0.7854 0s 1
2
grav. acceleration g =9.807ms
implicit dispersion relation f(k)

0.02

0.00
1
solution f(k)=0 for k =0.0707786m
0.02 2
f(k)= g k tanh(kh)

0.04
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
wave number k [1/m]

Figure 23.1 Graphical verification of a solution of the nonlinear dispersion relation

23.3 Water Particle Velocity and Acceleration

j j
Equipped with

• water depth ℎ,

• wave frequency 𝜔,

• wave number 𝑘,

• and wave amplitude 𝜁𝑎

the real-valued velocity potential 𝜙 from Equation (.) provides the tool to investi-
gate water wave kinematics.
As we explored in our discussion of potential flow and specifically the flow around a Water particle
cylinder, flow velocities follow from the spatial derivatives of the velocity potential. The velocity
velocity vector 𝑣 = (𝑢, 𝑤)𝑇 in the two-dimensional wave flow problem has components
in 𝑥- and 𝑧-directions.
𝜕𝜙
𝑢 = horizontal particle velocity
𝜕𝑥
[ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑥 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 cos(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

276 23 Wave Properties

𝜕𝜙
𝑤 = vertical particle velocity
𝜕𝑧
[ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑧 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 sinh(𝑘𝑧 + 𝑘ℎ)
= 𝑘 sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜔 cosh(𝑘ℎ)
If the dispersion relation (.) is brought into the form (.)

𝜔2
𝑘𝑔 = (.)
tanh(𝑘ℎ)
and substituted into the equation for the horizontal velocity 𝑢, one gets:

𝜔2 𝜁𝑎 cosh(𝑘𝑧 + 𝑘ℎ)
𝑢 = cos(𝑘𝑥 − 𝜔𝑡)
tanh(𝑘ℎ) 𝜔 cosh(𝑘ℎ)
cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜁𝑎 𝜔 cos(𝑘𝑥 − 𝜔𝑡)
tanh(𝑘ℎ) cosh(𝑘ℎ)
Exploiting that tanh(𝑘ℎ) = sinh(𝑘ℎ)∕ cosh(𝑘ℎ) yields:

cosh(𝑘𝑧 + 𝑘ℎ)
𝑢 = 𝜁𝑎 𝜔 cos(𝑘𝑥 − 𝜔𝑡) (.)
sinh(𝑘ℎ)
j j

The analog result for the vertical particle velocity 𝑤 is

sinh(𝑘𝑧 + 𝑘ℎ)
𝑤 = 𝜁𝑎 𝜔 sin(𝑘𝑥 − 𝜔𝑡) (.)
sinh(𝑘ℎ)
Both velocity components 𝑢 and 𝑤 are harmonic functions of time and 𝑥-coordinate.
The factor preceding the sine and cosine functions may be interpreted as velocity
amplitudes 𝑢𝑎 and 𝑤𝑎 which are a function of submergence 𝑧.
Water particle The water particle acceleration is of great importance in the computation of wave forces
acceleration and vessel motions. Equations for the horizontal and vertical accelerations are obtained
by taking the time derivatives of Equations (.) and (.). Observe that the time
derivative of 𝜃 = 𝑘𝑥 − 𝜔𝑡 introduces an additional factor −𝜔.
.
𝑢 =
𝜕𝑢
= 𝜁𝑎 𝜔2
cosh(𝑘𝑧 + 𝑘ℎ)
sin(𝑘𝑥 − 𝜔𝑡) (.)
𝜕𝑡 sinh(𝑘ℎ)

. 𝜕𝑤 sinh(𝑘𝑧 + 𝑘ℎ)
𝑤 = = −𝜁𝑎 𝜔2 cos(𝑘𝑥 − 𝜔𝑡) (.)
𝜕𝑡 sinh(𝑘ℎ)

Phase Figure . shows a snapshot of several wave properties at time (𝑡 = 0). From top to
relationships bottom, we see wave elevation 𝜁, particle velocities 𝑢 and 𝑤, and particle accelerations
. .
𝑢 and 𝑤. The graph at the bottom of Figure . shows the dynamic pressure, which we
will discuss in the next section. Each curve represents the spatial distribution of the
respective property over . wave lengths, calculated at the calm water level (𝑧 = 0).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.3 Water Particle Velocity and Acceleration 277

j j

Figure 23.2 Distribution of wave properties over 1.5 wavelength at the calm water level (𝑧 = 0)
for a wave with wave period 𝑇 = 10 s

Wave elevation 𝜁 and horizontal particle velocity 𝑢 are in phase, i.e. wave elevation has Velocity
its positive maximum at the same time as the horizontal particle velocity. Under a wave amplitudes
crest, particles move with maximum velocity in direction of wave propagation, whereas
under a wave trough, the particles move with maximum velocity against the direction
of wave propagation. The amplitude of the wave elevation is obviously equal to 𝜁𝑎 . At

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

278 23 Wave Properties

𝑧 = 0 we obtain an amplitude 𝑢𝑎 of the velocity of

cosh(𝑘𝑧 + 𝑘ℎ) ||
𝑢 𝑎 = 𝜁𝑎 𝜔
sinh(𝑘ℎ) ||𝑧=0
𝜁𝑎 𝜔
= for 𝑧 = 0
tanh(𝑘ℎ)
For a deep water wave, the numerator becomes equal to one and the maximum hori-
zontal particle velocity is 𝑢𝑎 = 𝜁𝑎 𝜔.
The vertical particle velocity 𝑤 is phase shifted by a quarter cycle (Δ𝜃 = −𝜋∕2). It
is zero when the wave elevation is at its maximum or minimum (Figure .). The
maximum value 𝑤𝑎 at 𝑧 = 0 is equal to
2𝜋𝜁𝑎
𝑤𝑎 = 𝜁𝑎 𝜔 = for 𝑧 = 0 (.)
𝑇
There is no difference in vertical velocity amplitude 𝑤𝑎 whether the wave is in deep
water or restricted water depths. An expansion of wave amplitude multiplied by wave
frequency on the right-hand side reveals that the maximum vertical particle velocity is
equal to the ratio of the circumference of a circle with radius 𝜁𝑎 and the wave period 𝑇 .
Acceleration
.
The horizontal particle acceleration 𝑢 is phase shifted by Δ𝜃 = −𝜋∕2 like the vertical
amplitudes particle velocity (Figure ., fourth graph). Its maximum value is for 𝑧 = 0:

j . 𝜁𝑎 𝜔2 j
𝑢𝑎 = for 𝑧 = 0 (.)
tanh(𝑘ℎ)
For the amplitude of the vertical particle acceleration we get at the calm water level
(𝑧 = 0):
.
𝑤𝑎 = 𝜁𝑎 𝜔2 for 𝑧 = 0 (.)
.
𝑤 is in opposite phase to the wave elevation. Particles experience the largest accelera-
tions downward under a crest and upward under a wave trough.
Vertical Figure . emphasizes the change of particle velocities with increasing submergence.
distribution of We learned from Figure . that the horizontal particle velocity has its positive maxi-
velocities
mum (in the direction of wave propagation) underneath the wave crest. The vertical
particle velocity has its positive (upward) maximum where the wave elevation vanishes
in front of the approaching wave crest.
The horizontal velocity vanishes where the vertical velocity has an extreme value and vice
versa. Velocities are largest at the calm water level and rapidly decline with increasing
depth. The decline is governed by the expressions
cosh(𝑘𝑧 + 𝑘ℎ) sinh(𝑘𝑧 + 𝑘ℎ)
and
sinh(𝑘ℎ) sinh(𝑘ℎ)
for horizontal and vertical particle velocities respectively. On deep water both velocities
decline with e𝑘𝑧 .
The vertical particle velocity vanishes at the sea bottom (𝑧 = −ℎ) as requested by the
bottom boundary condition (.). However, in restricted water depth the horizontal
particle velocity does not vanish at the sea bottom. Every passing wave moves water

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.4 Dynamic Pressure 279

Figure 23.3 Snapshot (𝑡 = 0) of the velocity field for a wave in restricted water depth

back and forth parallel to the bottom. Just watch the sand rolling back and forth on the
sea bottom next time you visit a beach.
Observe that the origin of the 𝑧-axis has been laid into the calm water level. Since Mind the vertical
j the 𝑧-axis is positive upward, values of 𝑧 are negative for all points in the water. The axis
j
attentive reader will notice that the velocity distributions 𝑢(𝑧) and 𝑤(𝑧) in Figure .
have been drawn up to the calm water level irrespective of whether the position is under
a wave crest or a wave trough. It shall serve as a reminder that in linear wave theory
wave amplitudes are assumed to be very small compared with the wave length.

23.4 Dynamic Pressure

The pressure distribution in a wave is obtained from the Bernoulli equation (see Sec-
tions .). Since we linearized the boundary value problem, it is opportune to use the
linearized form of the Bernoulli equation.
𝜕𝜙
𝑝 − 𝑝∞ = −𝜌 − 𝜌𝑔𝑧 (.)
𝜕𝑡
The dynamic pressure in the wave field is equal to the unsteady portion of the pressure:
𝜕𝜙
𝑝dyn = −𝜌
𝜕𝑡 [ ]
𝜕 𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= −𝜌 sin(𝑘𝑥 − 𝜔𝑡)
𝜕𝑡 𝜔 cosh(𝑘ℎ)
𝜁𝑎 𝑔 cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜌𝜔 cos(𝑘𝑥 − 𝜔𝑡)
𝜔 cosh(𝑘ℎ)
cosh(𝑘𝑧 + 𝑘ℎ)
= 𝜌 𝑔 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) (.)
cosh(𝑘ℎ)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

280 23 Wave Properties

Figure 23.4 Amplitude of dynamic pressure over

0
depth 𝑧

10

20

[m]
depth z 30

40
water depth h =50.00m
wave period T =10.00s
1
wave number k =0.04154m
2
grav. acceleration g =9.807ms
50
0.0 0.2 0.4 0.6 0.8 1.0
pdyna
dynamic pressure amplitude []
g a
j j

As the bottom graph in Figure . shows, dynamic pressure is in phase with the wave
elevation. The maximum value is equivalent to the hydrostatic pressure underneath
a water column of height 𝜁𝑎 . The positive maximum is found at the calm water level
(𝑧 = 0) under wave crests (𝜃 = 𝑛𝜋, with 𝑛 = 0, 2, 4, … ).

𝑝dyn𝑎 = 𝜌 𝑔 𝜁𝑎 for 𝑧 = 0 (.)

The dynamic pressure 𝑝dyn at the sea bottom 𝑧 = −ℎ fluctuates between the values

𝜌 𝑔 𝜁𝑎 𝜌 𝑔 𝜁𝑎
− < 𝑝dyn < + for 𝑧 = −ℎ.
cosh(𝑘ℎ) cosh(𝑘ℎ)

Like wave particle velocities and accelerations, dynamic pressure amplitude declines
rapidly with depth 𝑧 (see Figure .). Dynamic pressure is positive underneath wave
crests and negative underneath wave troughs.

23.5 Water Particle Motions

Particle paths Integration of the wave particle velocities Equations (.) and (.) (or their deep
water equivalents) over time 𝑡 results in the paths of a water particle with initial position
(𝑥, 𝑧). The photo in Figure . shows the unsteady movement of small, neutrally
buoyant particles underneath a wave. The picture was taken in a small wave flume at the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.5 Water Particle Motions 281

Figure 23.5 Photo of water particle trajectories. Photo courtesy of Dr. Walter L. Kuehnlein,
sea2ice Ltd. & Co. KG, www.seaice.com

Technical University of Berlin. Its walls are made out of acrylic. The photo’s exposure
j j
time was chosen to be equal to one wave period. Wave steepness 𝐻∕𝐿𝑤 is small and, as
linear wave theory predicts, paths of particles are closed. Therefore, water particles are
not transported along with the wave elevation. On average, they stay at their original
location (𝑥, 𝑧). The vertical axis of the elliptical paths declines visibly with depths. The
horizontal extensions decline very little because the wave length is about  times the
water depth. As a consequence, shallow water effects are strong.
In steep waves with high 𝐻∕𝐿𝑤 ratio, particle paths are no longer closed. They look
like a spiral and during each wave cycle particles are moved a small distance further
in the direction of wave propagation. This mass transport contributes considerably to
coastal storm surge in heavy storms like hurricanes.
Figure . depicts water particle trajectories for deep water (left side) and restricted Deep water
water depth (on the right). The × mark the position of water particles when they are trajectories
at rest. For deep water the trajectories are circles. The diameter is largest at the calm
water level and equal to twice the wave amplitude. With increasing submergence the
diameters of trajectories decline rapidly and vanish almost completely at a depth of half
the wave length.
On the right side of Figure ., trajectories are plotted for the same wave period and Trajectories on
amplitude but consider a finite water depth of ℎ = 50 m. Due to the influence of the restricted water
depth
ocean floor, wave length has shrunk by %. The ratio of water depth to wave length is
about ℎ∕𝐿𝑤 ≈ 0.2. The trajectories have become ellipses. The eccentricity is still small
close to the water surface but increases with increasing submergence. At the ocean
floor, the vertical axis of all trajectories vanishes and particles simply move back and
forth parallel to the bottom (see also Figure .). This is of course a consequence of
the bottom boundary condition which prohibits flow through the ocean floor.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

282 23 Wave Properties

horizontal part. displacement x [m] horizontal part. displacement x [m]

10 5 0 5 10 10 5 0 5 10

0 0

10 10
vertical particle displacement z [m]

restricted water depth h

20 20
deep water

30 30
j j
depth z,

40 40
a = 2.500 m a =2 .500m

T = 12.00 s T =1 2.00s
Lw = 224.760 m Lw =204.784m
h h =5 0.000m
50 50

Figure 23.6 Water particle trajectories over one wave period 𝑇 for deep water (left) and
restricted water (right)

References
Akai, T. (). Applied numerical methods for engineers. John Wiley & Sons, Inc.,
New York, NY.
Clauss, G., Lehmann, E., and Östergaard, C. (). Offshore structures, volume :
Conceptual design and hydrodynamics. Springer Verlag, London.
Dean, R. and Dalrymple, R. (). Water wave mechanics for engineers and scientists,
volume  of Advanced Series on Ocean Engineering. World Scientific, Hackensack,
NJ.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

23.5 Water Particle Motions 283

Self Study Problems

. Compute the wave number 𝑘 and wave length 𝐿𝑤 for a wave with period 𝑇 = 6 s
in ℎ = 12 m water depth.
. Will the wave length become shorter or longer when the wave propagates from
deep into shallow water?
. Compute the wave length for the wave with frequency 𝜔 = 0.95 s−1 in water of
depth ℎ = 25 m.

. Find the wave length 𝐿𝑤 of a wave of frequency 𝜔 = 0.95 s−1 in deep water.
. Use the real-valued deep water potential
𝜁𝑎 𝑔 𝑘 𝑧
𝜙(𝑥, 𝑧, 𝑡) = e 0 sin(𝑘0 𝑥 − 𝜔𝑡) (.)
𝜔
to derive equations for the water particle velocities, accelerations, and dynamic
pressure for the case 𝑘ℎ > 𝜋.
. Use the equations from Problem  to plot maxima of velocities 𝑢𝑎 (𝑧), 𝑤𝑎 (𝑧),
. .
accelerations 𝑢𝑎 (𝑧), 𝑤𝑎 (𝑧), and dynamic pressure 𝑝dyn (𝑧) from 𝑧 = −𝐿𝑤 ∕2 to
𝑧 = 0 for a wave with period 𝑇 = 5 s and amplitude 𝜁𝑎 = 1 m.

j . Derive the equations for the complex water particle velocities and accelerations j
from the complex linear wave theory velocity potential (.).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

284

24
Wave Energy and Wave Propagation

The preceding chapter discusses wave kinematics and the distribution of wave particle
velocities, accelerations, and dynamic pressure throughout the water column under-
neath a long crested wave. In this chapter, we will use these results to define the energy
contained in a regular wave and how the wave and its energy propagate. The results will
form the basis for a brief discussion of ship wave resistance in the following chapter.

Learning Objectives
At the end of this chapter students will be able to:
j j
• distinguish between the kinematic energy and potential energy of waves

• describe and discuss how wave energy propagates

• compute group velocity and energy density of waves

24.1 Wave Propagation

Wave elevation Over the time of a single wave period 𝑇 a wave travels the distance of one wave length
𝐿𝑤 . We used this to define the phase velocity 𝑐 = 𝜔∕𝑘 in Equation (.). See also
Section .. Figure . depicts a wave in  time instances during one wave period
(𝜔𝑡 ∈ [0, 2𝜋]). The wave crest, marked with a small triangle ▿, travels exactly one wave
length from 𝑘𝑥 = 0 to 𝑘𝑥 = 2𝜋.
Water particles It is important to understand that the wave elevation is propagating but not the water
particles. In the center of Figure . at 𝑘𝑥 = 2𝜋, a particle (small dot) is tracked
over one wave period. On average, it stays in the same position. Its horizontal and
vertical excursions are equal to the wave amplitude ±𝜁𝑎 . If you look closely, you will
notice that the dot marking the water particle is not always exactly on the plotted wave
contour. This is most evident at the locations of maximum horizontal excursion. The
apparent discrepancy is a consequence of the linearization of the free surface boundary
conditions. The harmonic wave elevation 𝜁 = 𝜁𝑎 cos(𝑘𝑥 − 𝜔𝑡) is only truly valid for
vanishing wave amplitude or more precisely if 2𝜁𝑎 ∕𝐿𝑤 ≪ 1.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.1 Wave Propagation 285

wave crest water particle

0

2
6

4
6

5
6
[]
ttime

7
6

8
6

9
6

j j
10
6

11
6

0 3 2 5 3
2 2 2
position kx []

Figure 24.1 Propagation of wave profile and the movement of a water particle over one wave period

The larger the wave steepness 𝐻∕𝐿𝑤 , the less accurate is the sinusoidal wave profile. Mass transport in
In fact, as mentioned previously, in steep waves the particle paths are no longer closed steep waves
loops, and a slow but steady mass transport occurs in the direction of wave propagation.
Once the water depth ℎ is less than half the wave length, the sea bottom begins to affect Effects of finite
the way a wave propagates. Did you ever wonder why waves always roll onto a beach water depths
with their crests and troughs parallel to the shoreline? As a wave moves from deeper
into shallower water, its wave length becomes shorter. Figure . shows the wave
length 𝐿𝑤 as a function of water depth ℎ for a selection of wave periods 𝑇 . The wave
length stays constant in deep water where 2ℎ > 𝐿𝑤 .
A decreasing wave length causes the wave number 𝑘 to increase. As a consequence, the
wave phase velocity 𝑐 = 𝜔∕𝑘 must decrease, i.e. a wave slows down as it moves into

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

286 24 Wave Energy and Wave Propagation

j j
Figure 24.2 Wave length 𝐿𝑤 as a function of water depth ℎ for constant wave period 𝑇

Figure 24.3 Effect of gradually decreasing water depth ℎ on wave propagation and direction

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 287

shallower water. Figure . shows waves approaching a beach. Further offshore, in
deeper water, waves move in the direction of prevailing winds. In the figure, waves
travel at an oblique angle with respect to the beach. The parts of waves closer to shore
will be affected first by the decreasing water depth. Those parts will slow down while
the parts further offshore are still traveling at the original phase velocity. The difference
in phase velocity causes the wave to turn and eventually its crest becomes parallel to
the shoreline. As it further approaches the beach, the front of a wave continues to
slow down. The wave becomes higher and steeper, and it will eventually break if the
steepness exceeds 𝐻∕𝐿𝑤 > 1∕7 (Michell, ).

24.2 Wave Energy

Whoever has sailed through a storm at sea or watched waves crashing into a sea wall
has experienced the destructive force of waves first hand. In fact, waves carry so much
energy that it is a potential source of renewable power. The Bureau of Ocean Energy
Management states on its web pages (https://www.boem.gov/Ocean-Wave-Energy/)
that in the United States alone the recoverable wave power along the U.S. shelf edge
would amount to more than  TWh/yr (i.e. tera-watt hours per year), which is
enough to power millions of homes.
Waves contain two forms of energy: Contributions to
wave energy
j j
(i) Kinetic energy: In the previous chapter we discussed the motion of water particles
in waves. Obviously, if a mass of fluid moves, it possesses kinetic energy.

(ii) Potential energy: The water surface at a selected point moves up and down in a
way similar to the movement of a mass hanging from a spring. In fact, when a wave
crest passes, fluid elements have to be lifted above their equilibrium condition.
The necessary work done by the flow is converted into potential energy in the
gravity field.

Below, both forms of energy are computed and, as before, it is assumed that the wave is
regular and its wave steepness is small enough to apply linear wave theory.

24.2.1 Kinetic wave energy

Figure . shows a slice of ocean 𝑉 which is one wave length 𝐿𝑤 long, has a width of Kinetic energy
𝑏 in 𝑦-direction, and stretches from the ocean floor at 𝑧 = −ℎ to the calm water level
𝑧 = 0. The wave propagates in 𝑥-direction and the wave crests are parallel to the 𝑦-axis
(long crested wave). The volume is cut into small volumes d𝑉 of size

d𝑉 = d𝑥 d𝑧 𝑏 (.)

The kinetic energy contained in the volume d𝑉 is

1
d𝐸kin = 𝜌 |𝑣|2 d𝑉 (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

288 24 Wave Energy and Wave Propagation

Figure 24.4 Kinetic energy 𝐸kin in the control volume 𝑉 spanning one wave length 𝐿𝑤 in
𝑥-direction

Obviously, the total kinetic energy in the fluid space 𝑉 is found by integrating d𝐸kin
across the volume.
1
𝐸kin = 𝜌 |𝑣|2 d𝑉 (.)
∫ 2
𝑉
j j
Since the flow is planar in the 𝑥-𝑧-plane, the volume integral reduces to a D integral
over wave length and water depth. The components of the velocity vector 𝑣𝑇 = (𝑢, 𝑤)
are given by Equations (.) and (.).

0 𝐿𝑤
1 ( 2 )
𝐸kin = 𝜌𝑏 𝑢 + 𝑤2 d𝑥d𝑧
2 ∫ ∫
−ℎ 0
( )2 0 𝐿𝑤
1 𝜁𝑎 𝜔 [
= 𝜌𝑏 cosh2 (𝑘𝑧 + 𝑘ℎ) cos2 (𝜃)
2 sinh(𝑘ℎ) ∫ ∫
−ℎ 0
]
+ sinh2 (𝑘𝑧 + 𝑘ℎ) sin2 (𝜃) d𝑥d𝑧 (.)

Simplifying the As a first step, the integrand is simplified using the following identities for the hyperbolic
integrand cosine and sine functions.
( ) ( )
1 𝛼 1 𝛼
cosh(𝛼) = e + e−𝛼 and sinh(𝛼) = e − e−𝛼
2 2

from which follow the relationships

( ) ( )
1 2𝛼 1 2𝛼
cosh2 (𝛼) = e + e−2𝛼 + 2 and sinh2 (𝛼) = e + e−2𝛼 − 2
4 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 289

Setting 𝛼 = 𝑘𝑧 + 𝑘ℎ, the integrand in Equation (.) becomes

cosh2 (𝛼) cos2 (𝜃) + sinh2 (𝛼) sin2 (𝜃)

1 [( 2𝛼 ) ( ) ]
= e + e−2𝛼 + 2 cos2 (𝜃) + e2𝛼 + e−2𝛼 − 2 sin2 (𝜃)
4
1 [( 2𝛼 )( ) ( )]
= e + e−2𝛼 cos2 (𝜃) + sin2 (𝜃) + 2 cos2 (𝜃) − sin2 (𝜃)
4
1 [( 2𝛼 ) ( )]
= e + e−2𝛼 + 2 cos2 (𝜃) − sin2 (𝜃)
4

For the second part, we used the identity 1 = cos2 (𝜃) + sin2 (𝜃).
Further, with cos(2𝜃) = cos2 (𝜃) − sin2 (𝜃) we finally get a much simpler form of the
integrand:

cosh2 (𝛼) cos2 (𝜃) + sinh2 (𝛼) sin2 (𝜃)

[ ]
1
= cosh(2𝑘𝑧 + 2𝑘ℎ) + cos(2𝑘𝑥 − 2𝜔𝑡) (.)
2

This expression captures the magnitude of velocity |𝑣| as a function of 𝑥 and 𝑧.

Equation (.) is rewritten based on the result (.): Integration

j ( )2 ⎧ 0 𝐿𝑤 j
𝜁𝑎 𝜔 ⎪
𝐸kin = 𝜌 𝑏
sinh(𝑘ℎ) ⎨ ∫ ∫ cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑥d𝑧
⎪−ℎ 0
⎩
0 𝐿𝑤 ⎫
⎪
+ cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥d𝑧⎬ (.)
∫ ∫ ⎪
−ℎ 0 ⎭

The remaining two integrals are of standard form.

The integrand of the first integral in Equation (.) is independent of 𝑥. The 𝑥-
integration simply yields a factor of one wave length and the integral becomes:

0 𝐿𝑤 0

cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑥d𝑧 = 𝐿𝑤 cosh(2𝑘𝑧 + 2𝑘ℎ) d𝑧

∫ ∫ ∫
−ℎ 0 −ℎ
[ ]0
1
= 𝐿𝑤 sinh(2𝑘𝑧 + 2𝑘ℎ) (.)
2𝑘 −ℎ

𝐿𝑤
= sinh(2𝑘ℎ)
2𝑘

The second integrand in Equation (.) is only a function of 𝑥 and time 𝑡. Integration
over the water depth yields the factor ℎ. However, the remaining integral over the wave
length vanishes.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

290 24 Wave Energy and Wave Propagation

0 𝐿𝑤 𝐿𝑤 [ ]𝐿𝑤
sin(2𝑘𝑥 − 2𝜔𝑡)
cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥d𝑧 = ℎ cos(2𝑘𝑥 − 2𝜔𝑡) d𝑥 = ℎ
∫ ∫ ∫ 2𝑘 0
−ℎ 0 0
[ ]𝐿𝑤
sin(2𝑘𝑥) cos(2𝜔𝑡) − cos(2𝑘𝑥) sin(2𝜔𝑡)
= ℎ (.)
2𝑘 0

= 0

Kinetic energy in With the results from Equations (.) and (.), we obtain from Equation (.) the
control volume kinetic wave energy in the control volume 𝑉 :
𝑉
( )2
𝜁𝑎 𝜔 1
𝐸kin = 𝜌 𝑏 𝐿𝑤 sinh(2𝑘ℎ)
2 sinh(𝑘ℎ) 2𝑘

Utilizing that sinh(2𝑘ℎ) = 2 sinh(𝑘ℎ) cosh(𝑘ℎ) results in

( 2 )
1 𝜔 cosh(𝑘ℎ)
𝐸kin = 𝜌 𝑏 𝐿𝑤 𝜁𝑎2 (.)
4 𝑘 sinh(𝑘ℎ)

The ratio of hyperbolic cosine and sine is equivalent to the reciprocal hyperbolic tangent:

j cosh(𝑘ℎ) 1 j
= (.)
sinh(𝑘ℎ) tanh(𝑘ℎ)
The resulting expression may be simplified based on the dispersion relation (.).

𝜔2
= 𝑔 (.)
𝑘 tanh(𝑘ℎ)
Thus, the kinetic energy contained in our control volume 𝑉 is
1
𝐸kin = 𝜌 𝑔 𝜁𝑎2 𝑏 𝐿𝑤 (.)
4

Kinetic wave The kinetic energy per unit ocean surface is known as kinetic wave energy density. We
energy density divide Equation (.) by the area of the free surface 𝑏 𝐿𝑤 .
𝐸kin 1
𝐸 kin = = 𝜌 𝑔 𝜁𝑎2 (.)
𝑏 𝐿𝑤 4
Note that the energy density is independent of water depth and wave frequency. It is
solely a function of the squared wave amplitude.

24.2.2 Potential wave energy

Potential energy Potential energy in a gravity field is determined by the gravitational acceleration, eleva-
tion above a datum level, and mass of the object. When calculating the potential wave
energy, we consider the work done when a volume of water is lifted above the calm

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.2 Wave Energy 291

Figure 24.5 Change in potential energy 𝐸pot when a fluid element is lifted above the calm
water level

water level. Figure . shows a piece of the free surface. We consider a small volume
d𝑉 of length d𝑥, width 𝑏, and height equal to the local wave elevation 𝜁. Its mass will be
d𝑚 = 𝜌 d𝑉 = 𝜌 𝑏 𝜁 d𝑥 (.)
The center of a mass element d𝑚 is lifted above the calm water surface by half of its
height. The work necessary for the lift is equal to the change in potential energy.
𝜁 1
d𝑚 = 𝜌 𝑔 𝑏 𝜁 2 d𝑥
d𝐸pot = 𝑔 (.)
2 2
The same work has to be done to depress the free surface. Outside the range of a
progressing wave elevation, overall mass distribution remains unchanged and water
j j
fills the control volume. Therefore, changes in potential energy occur at the free surface
only.
We integrate Equation (.) over one wave length 𝑥 ∈ [0, 𝐿𝑤 ]. Potential wave
energy
2𝜋
1
𝐸pot = 𝜌𝑔𝑏 𝜁 2 d𝑥
2 ∫
0
2𝜋
1
= 𝜌𝑔𝑏 𝜁 2 cos2 (𝑘𝑥 − 𝜔𝑡) d𝑥 (.)
2 ∫ 𝑎
0
[ ]𝐿𝑤
1 𝑥 1
= 𝜌 𝑔 𝑏 𝜁𝑎2 − sin(2𝑘𝑥 − 2𝜔𝑡)
2 2 4𝑘 0
( )
1 2 𝐿 𝑤
= 𝜌 𝑔 𝑏 𝜁𝑎 −0
2 2

Thus, the potential energy distributed over one wave length is equal to the kinetic
energy.
1
𝐸pot = 𝜌 𝑔 𝜁𝑎2 𝑏 𝐿𝑤 (.)
4

The potential wave energy density is Potential wave

energy density
𝐸pot 1
𝐸 pot = = 𝜌 𝑔 𝜁𝑎2 (.)
𝑏 𝐿𝑤 4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

292 24 Wave Energy and Wave Propagation

1.5

Ekin + Epot
mean total wave
energy density Etot
[−]

1.0 potential wave

energy density
2
a

Epot
E
g 1
2
energy density

0.5

kinetic wave energy density Ekin

0.0
0 /2 3/2 2
phase [rad]
j j
Figure 24.6 Wave energy density distribution of a regular wave over one wave cycle according
to linear wave theory

24.2.3 Total wave energy density

The total energy density is equal to the sum of kinetic and potential wave energy density.
According to linear wave theory, the energy density of a regular wave amounts to

1
𝐸 = 𝐸 pot + 𝐸 kin = 𝜌 𝑔 𝜁𝑎2 (.)
2
This result is significant as it relates to the statistical energy density of regular waves
𝑆(𝜔)d𝜔 = 𝜁𝑎2 ∕2. 𝑆(𝜔) is the value of the spectral energy density spectrum also known
as wave spectrum (Dean and Dalrymple, ). The latter plays an important role in
the statistical treatment of irregular sea states which are modeled as a superposition
of many regular components with varying frequency and phase. More on this may be
found in offshore engineering references like Barltrop and Adams ().
For our discussion of wave resistance, we may draw the conclusion that doubling the
wave amplitude in the ship’s wave pattern will quadruple its energy density. Low wave
resistance is achieved if wave amplitudes in the wave pattern stay as small as possible.
Distribution of Figure . shows the distribution of wave energy density for a wave over one period.
wave energy Kinetic energy density is constant across a wave cycle because the amplitude of the wave
density
particle velocity is only varying with submergence. Equation (.) reveals that the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 293

potential wave energy density follows a cos2 distribution over one cycle. The potential
wave energy density is at a maximum for wave crest and trough. It vanishes at the zero
crossings of the wave elevation. However, as we calculated in (.), its mean value is
equal to the kinetic energy density distribution.

24.3 Energy Transport and Group Velocity

Waves transport energy in the direction of wave propagation. However, since mass
transport does not exist in linear waves due to the closed particle paths, only the
potential wave energy portion is transported by the wave. In this section we determine
how fast wave energy propagates.
First, we study the idealized simulation depicted in Figure .. It follows an idea taken
from Clauss et al. (). Imagine a wave maker which transfers  energy density
units into the water during each wave cycle. The horizontal axis in Figure . marks
the distance from the wave maker which acts at location 𝑥 = 0. The vertical scale
depicts time measured from top to bottom in wave periods.
Beginning at the top of Figure ., we track the energy density distribution over  Simulation from
wave maker cycles. The wave maker begins by introducing  energy density units, cycle 0 to 1
one half of which is converted into kinetic energy; the other half becomes potential
energy. During the first cycle, the potential energy propagates one wave length forward
j along with the wave elevation. The kinetic wave energy remains at the wave maker j
which adds another  energy density units. At the end of cycle  a total of 
energy density units are at the wave maker while  units have been propagated to
location 𝑥 = 𝐿𝑤 .
At station 𝑥 = 𝐿𝑤 , half of the  units must be used to get the water particles moving Simulation from
(kinetic energy) and subsequently remain at 𝑥 = 𝐿𝑤 . The other half ( units) is cycle 1 to 2
propagated to station 𝑥 = 2𝐿𝑤 . At the same time, half of the  energy density units
from 𝑥 = 0 have moved to station 𝑥 = 𝐿𝑤 . The remaining  units at 𝑥 = 0 are
augmented by another  units from the wave maker. Thus, the energy distribution
after two wave maker cycles is

position 𝑥∕𝐿𝑤   
energy density units   

The lowest row in Figure . shows the energy density distribution after  wave maker
cycles. Note that we are slowly approaching equilibrium at the wave maker, in which as
much energy is propagated as is added. With increasing distance from the wave maker,
energy density declines rapidly. Since the energy density is tied to the wave amplitude,
the latter must diminish as well with increasing distance from the wave maker.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

294 24 Wave Energy and Wave Propagation

Birk — “fshy” — // — : — page  — #
j

j
Trim Size: mm × mm Single Column Tight

Figure 24.7 Schematic propagation of wave energy for a deep water wave based on linear wave theory

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 295

j j

Figure 24.8 Wave elevation profiles after a few selected cycles of wave making (deep water)

Although the wave maker adds an equal amount of energy in each cycle, it takes six Wave elevation
cycles to raise the energy density to within % of its asymptotic value. Figure . depicts
the resulting wave elevations for wave maker cycles , , , and . The unavoidable
build-up of wave amplitude in the tank makes model testing in regular waves difficult
as the model will experience the change in amplitude as a wave of different frequency.
Figure . shows the wave energy density distributions corresponding to the wave Wave energy
elevations in Figure .. A small amount of the wave energy has progressed as many density
distribution
wave lengths as there are wave cycles. However, the mean value of the wave energy
density has traveled only half that distance. This implies that the energy propagates
with half the phase velocity.
𝑐
𝑐𝐺 = for deep water waves (.)
2

The propagation speed of wave energy was first derived by Sir Gabriel Stokes and based
on a physical argument (Darrigol, ). Later Strutt, Baron Rayleigh () provided

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

296 24 Wave Energy and Wave Propagation

Figure 24.9 Distribution of energy density after a few selected cycles of wave making (deep
water)

a formal proof by computing the work done by the dynamic pressure in a cross section
j of the fluid. In general, the energy propagation velocity of water waves depends also on j
the water depth.
[ ]
𝑐 2𝑘ℎ
𝑐𝐺 = 1+ (.)
2 sinh(2𝑘ℎ)

Group velocity The velocity at which wave energy is transported corresponds to the propagation speed
of wave groups. In continuation of our wave making simulation from Figure ., a
wave maker is operated for  cycles and shut down. Figure . depicts the resulting
group of regular waves after , , and  wave periods. After  cycles, the actual
front of the wave group has traveled  wave lengths, but the amplitudes at the front are
very small and cut off on the plots. The dashed lines at the top and bottom of the groups
represent group envelopes, which interpolate wave crests and troughs, respectively,
A gray triangle marks the maximum peak of the group envelop. During the  wave
periods between individual plots, the peak of the wave group has traveled exactly  wave
lengths because the energy transport velocity (.) is just half the phase velocity on
deep water. Therefore, the energy transport velocity is also known as group velocity 𝑐𝐺 .
The plots also track a wave labeled #. As expected, it progresses with phase velocity
 wave lengths during  wave periods.

References
Barltrop, N. and Adams, A. (). Dynamics of fixed marine structures. Butterworth-
Heinemann Ltd, Oxford, United Kingdom, third edition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

24.3 Energy Transport and Group Velocity 297

j j

Figure 24.10 Propagation of a group of regular waves over 20 wave periods

Clauss, G., Lehmann, E., and Östergaard, C. (). Offshore structures, volume :
Conceptual design and hydrodynamics. Springer Verlag, London.
Darrigol, O. (). The spirited horse, the engineer, and the mathematician: water
waves in nineteenth-century hydrodynamics. Archive for History of Exact Sciences,
():–. doi ./s---.
Dean, R. and Dalrymple, R. (). Water wave mechanics for engineers and scientists,
volume  of Advanced Series on Ocean Engineering. World Scientific, Hackensack,
NJ.
Michell, J. (). The hightest wave in water. Philosophical Magazine Series , :–
.
Strutt, Baron Rayleigh, J. (). On progressive waves. In Proceedings of the London
Mathematical Society, volume IX, pages –.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

298 24 Wave Energy and Wave Propagation

Self Study Problems

. Compute the wave energy density for a wave of wave height 𝐻 = 2 m and wave
length 𝐿𝑤 = 40 m in ℎ = 50 m water depth.
. Explain to a friend why wave crests always align themselves parallel to a beach
with a gentle slope.
. Throw a stone into a calm pond. Watch the developing ripples closely. Look out
for the waves that emerge at the trailing edge of the wave group and observe how
they move to the front, decline, and finally vanish at the leading edge of the wave
group.
. Complete  cycles of a table like the one shown is Figure . for a wave maker
which adds  energy units during each cycle. Round numbers to one decimal.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

299

25
Ship Wave Resistance

Every object traveling at or close to a free fluid interface will create a wave pattern:
a ship sailing across the ocean, a koi close to the water surface, or a sea plane about
to land. In this chapter we apply our knowledge about waves to the problem of wave
resistance. The actual computation of wave resistance is beyond the scope of this book,
but we will discuss basic concepts and introduce Michell’s integral and panel methods,
which may be employed to assess wave pattern resistance.

Learning Objectives

j At the end of this chapter students will be able to: j

• understand the origin and mechanism of wave resistance
• discuss favorable and unfavorable wave superposition
• know about methods for wave resistance computation

25.1 Physics of Wave Resistance

A deeply submerged body in a frictionless fluid will not experience a drag force. D’Alem- Effect of free
bert investigated this hypothetical case and found that the resulting pressure force surface
vanishes for a body deeply submerged in an inviscid fluid regardless of the body shape.
This result is known as d’Alembert’s paradox (see Section .). However, vessels moving
at or close to the free surface create the typical ship wave pattern and experience an
additional resistance component that we call wave resistance 𝑅𝑊 . It exists even in a
hypothetical inviscid fluid. Thus, wave resistance is linked to the presence of a free
surface. The dimensionless wave resistance coefficient is defined in the usual manner as
2𝑅𝑊
𝐶𝑊 = (.)
𝜌 𝑣2 𝑆

Two models may be used to explain how the generation of a wave system results in
wave resistance. We may argue either that

(i) continuously generating waves extracts kinetic energy from the ship, or that

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

300 25 Ship Wave Resistance

j j

Figure 25.1 Wigley hull at Froude number 𝐹𝑟 = 0.26 showing the connection between fluid
and hull surface pressure and the resulting wave elevation. Light colors indicate high pressure
and high wave elevation. Dark colors indicate low pressure and low values of wave elevation

(ii) the generated waves change the pressure distribution around the hull, which
results in an additional pressure force.

Both are true, of course, and explain the same phenomenon. Let us discuss the models
in more detail.
Loss of energy In the preceding chapter we learned that waves contain and transport energy. As a
consequence, creating and sustaining a wave pattern requires energy. The higher the
waves are, the more energy is used. In a fluid otherwise at rest, only the moving ship
can provide the necessary energy. Apparently, a portion of the ship’s kinetic energy is
continuously converted into wave energy. The extracted kinetic energy must be replaced
with work done by the propulsion system, otherwise the vessel would slow down. The
work done per unit time is equivalent to the wave resistance multiplied by ship speed
(in steady conditions).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.2 Wave Superposition 301

Wind creates waves by locally changing pressure above the water surface. Just blow Pressure force
into your coffee mug and you will create small ripples. A body traveling at or just
beneath the free surface also changes the pressure distribution. Figure . shows the
pressure distribution over a Wigley hull and the associated water surface elevations.
Low pressure (dark gray to black) on the hull pulls the water surface down and creates
a wave trough. High pressure (light gray to white) pushes the water surface upwards
and generates wave crests. Any deformation of the free surface, in turn, changes the
pressure distribution over the hull. Wave resistance is the 𝑥-component of the net
pressure force resulting from the changes in pressure.
Wave resistance 𝑅𝑊 is commonly subdivided into two parts: Wave pattern
and wave
𝑅𝑊 = 𝑅𝑊𝑃 + 𝑅𝑊𝐵 (.) breaking
resistance

• wave pattern resistance 𝑅𝑊𝑃 , associated with the work done to generate the ship
wave system, and

• wave breaking resistance 𝑅𝑊𝐵 , associated with energy dissipation that occurs
when waves break. Ships with high block coefficients 𝐶𝐵 sail slowly. Wave heights
are generally small and wave pattern resistance makes up only a small portion of
the overall resistance. Their blunt bows, however, create steep bow waves which
immediately break. The swirling motion caused by wave breaking dissipates
energy. Again, kinetic energy is lost to the ship, which results in an additional
resistance component.
j j

We are unable to measure wave resistance directly. It may, however, be computed from
measured wave profiles or deduced from total resistance by subtracting estimates or
predictions of other resistance components. Wave resistance is typically the largest
portion of the residual resistance.

25.2 Wave Superposition

According to linear wave theory, energy contained in a wave is proportional to half its Humps and
squared amplitude 𝐸̄ ∼ 𝜁𝑎2∕2. Thus, wave resistance is high if the waves in a ship wave hollows
pattern are high, which means a lot of energy is spent in creating and sustaining the
waves. Ship waves tend to become higher with increasing ship speed. However, wave
amplitudes do not grow monotonically with velocity. At certain speeds, waves in the
wave pattern are less high than at other speeds. As a result, the wave resistance curve
shows distinct humps and hollows. Figure . shows the wave resistance coefficient
𝐶𝑊 of a Wigley hull form as a function of Froude number 𝐹𝑟. At humps, the wave
resistance is high compared to the values at adjacent speeds. In contrast, hollows in the
𝐶𝑊 -curve seem to indicate favorable conditions with diminished wave resistance.
Humps and hollows are the effect of favorable and unfavorable wave superposition. Wave
Waves can be superimposed without significant effects onto each other. Just throw two superposition

 In case you wonder why the waves in your mug quickly dissipate once you stop blowing, whereas ship

and ocean waves persist for fairly long time periods, here is an explanation: The short waves in your mug are
heavily affected by the surface tension of the fluid. In ocean waves with much greater wave lengths, effects of
surface tension are negligible.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

302 25 Ship Wave Resistance

5
range of experimental values

[10 3 ]
4
wave resistance coefficient CW

3 humps

hollows
0
0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Froude number Fr []

Figure 25.2 Pronounced humps and hollows in a wave resistance curve. Data from model tests
with Wigley hulls (Bai and McCarthy, 1979)

stones into a pond some distance apart! The circular waves will cross but continue on
their individual paths essentially unimpeded. Significant changes will occur only if the
j superimposed waves become too steep and break. j
Wave resistance In general, analytical solutions of the wave resistance problem are not known for ship
of a submerged shaped bodies. Therefore, we study the phenomenon of wave superposition and its effect
sphere
on wave resistance with the wave systems created by submerged spheres. Linear wave
theory was successfully employed by Havelock () to compute the wave resistance
𝑅𝑊 of a sphere traveling at constant speed 𝑣 with its center submerged at a depth 𝑠
below the calm water level. Using a Froude depth number
𝑣 1
𝐹𝑟𝑠 = √ and the factor 𝛼 = (.)
𝑔𝑠 𝐹𝑟2𝑠

the wave resistance of a submerged sphere with diameter 𝐷 is given by

𝜋𝜌 𝑔 4 𝐷6 −𝛼 [ (
1
) ]
𝑅𝑊sphere = e 𝐾0 (𝛼) + 1 + 𝐾1 (𝛼) (.)
64 𝑣6 2𝛼
𝐾0 and 𝐾1 are modified Bessel functions of the second kind (Abramowitz and Stegun,
).
Figure . shows the resultant wave resistance coefficient as a function of the Froude
number based on a reference length 𝐿 = 10𝐷. 𝑆sphere = 𝜋𝐷2 is the wetted surface of the
sphere. Wave resistance coefficient curves are shown for three submergence–diameter
ratios 𝑠∕𝐷 = 1.0, ., and .. The absolute values of the wave resistance coefficient are
of no further use to us. However, the curves allow three general conclusions regarding
wave resistance:

• The wave resistance is negligible for small Froude numbers below 𝐹𝑟 < 0.1.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.2 Wave Superposition 303

Figure 25.3 Wave resistance coefficient of a single submerged sphere; see Equations (25.4) and
j j
(25.1)

• Submergence has a significant influence. Moving the sphere just one radius
further away from the free surface reduces the maximum wave resistance of the
sphere by %. This concurs with our experience that slender ships with deep
draft tend to have less wave resistance than beamy vessels with shallow draft
where more displacement is concentrated close to the water surface.

• The wave resistance coefficient curve is not monotonic. In the case of the sphere,
it has a distinct maximum. Its speed and value strongly depend on submergence.
A declining coefficient means that the wave resistance force will grow slower and
slower with increasing speed once the maximum has been passed.

Besides the analytical solutions, Figure . also shows some numerical results for the
submergence–diameter ratio 𝑠∕𝐷 = 1. They have been obtained with a panel method
(see later). Although it overpredicts the maximum by about %, the results are generally
in good agreement with the exact potential theory result from Equation (.).
Figure . shows the wave pattern created by a sphere of diameter 𝐷∕𝐿 = 0.1 posi- Wave pattern of
tioned forward at 𝑥∕𝐿 = +0.5. The sphere center is located one diameter below the a submerged
sphere
calm water surface, i.e. its submergence is 𝑠 = 𝐷. A moving
√ reference frame is traveling
with the sphere at constant Froude number 𝐹𝑟 = 𝑣∕ 𝑔 𝐿 = 0.252. Imagine that the
position 𝑥∕𝐿 = +0.5 marks the bow if the reference length 𝐿 is interpreted as the ship’s
length. The line plot at the bottom of Figure . depicts the wave profile through the
sphere’s center plane 𝑦∕𝐿 = 0. The larger plot above it shows the port side half of its

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

304 25 Ship Wave Resistance

j j

Figure 25.4 Wave pattern and wave profile created by a single submerged sphere at position
𝑥∕𝐿 = +0.5 (forward). The sphere’s dimensionless speed is the Froude number 𝐹𝑟 = 0.252,
which is based on 𝐿 = 10𝐷

symmetric wave pattern. A D view of sphere and wave pattern is presented in the
small insert at the top right corner.
According to Equation (.), the wave length of the transverse waves is on deep water
equal to
𝐿𝑤 = 2𝜋 𝐹𝑟2 𝐿 for 𝑘 ℎ > 𝜋 (.)
Then the ratio of wave length and reference length becomes

𝐿𝑤
= 2𝜋 𝐹𝑟2 (.)
𝐿
With Froude number 𝐹𝑟 = 0.252 in Figure ., we get for the sphere:

𝐿𝑤 𝐿
= 0.4 or = 2.5 (.)
𝐿 𝐿𝑤

Wave pattern of After this preparatory work, we study the wave pattern of a second sphere. It has
second the same speed 𝐹𝑟 = 0.252, diameter 𝐷∕𝐿 = 0.1, and submergence but is located
submerged
sphere
aft at 𝑥∕𝐿 = −0.5 (Figure .). Comparing the wave profiles of forward sphere
(Figure .) and aft sphere (Figure .), we notice that both produce the same wave

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.2 Wave Superposition 305

j j

Figure 25.5 Wave pattern and wave profile created by a single submerged sphere at position
𝑥∕𝐿 = −0.5 (aft). The sphere’s dimensionless speed is the Froude number 𝐹𝑟 = 0.252

length 𝐿𝑤 = 0.4𝐿. This is of course expected for identical Froude numbers. The waves
following immediately behind the aft sphere also have the same heights as the waves
following the forward sphere.
Figure . shows the wave pattern created by both spheres simultaneously. The Superposition of
combined waves trailing the aft sphere are now much lower. The transverse waves in wave patterns
the pattern have almost vanished.
Figure . explains this effect. The wave profiles of forward and aft sphere are exactly Favorable wave
out of phase, i.e. the bow sphere produces a wave crest where the stern sphere creates a superposition
wave trough. As a consequence, waves created by the stern sphere are almost canceled
by the waves of the bow sphere. The low wave heights in the wave pattern are an
indication for small wave resistance and a hollow in the wave resistance curve.
Next, we study the superposition of waves for the same pair of spheres at the slightly Unfavorable
higher Froude number 𝐹𝑟 = 0.282. The ratio of wave length to reference length is wave
superposition
according to Equation (.).
𝐿𝑤 𝐿
= 0.5 or = 2.0 (.)
𝐿 𝐿𝑤
The resultant wave pattern is shown in Figure .. It is apparent that the transverse
waves are much higher compared with the waves at the smaller Froude number in
Figure ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

306 25 Ship Wave Resistance

j j

Figure 25.6 Combined wave pattern and profile of two submerged spheres. Froude number
𝐹𝑟 = 0.252; favorable superposition of waves resulting in low wave heights

Figure 25.7 Comparison of wave profiles created by submerged spheres at positions

𝑥∕𝐿 = ±0.5. The spheres’ dimensionless speed is the Froude number 𝐹𝑟 = 0.252

In contrast to the wave profiles at 𝐹𝑟 = 0.252, the waves of bow and stern sphere now
superimpose crest on crest (in phase, see Figure .). This results in wave amplification
rather than cancellation. This unfavorable superposition of waves will result in high
wave resistance and a hump in the wave resistance curve.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.2 Wave Superposition 307

j j
Figure 25.8 Wave pattern and wave profile of two submerged spheres. Froude number
𝐹𝑟 = 0.282; unfavorable superposition of waves, resulting in high wave heights

Figure 25.9 Comparison of wave profiles created by submerged spheres at positions

𝑥∕𝐿 = ±0.5. The spheres’ dimensionless speed is Froude number 𝐹𝑟 = 0.282

In our study case, bow and stern spheres are separated by ten sphere diameters 𝐿 = 10𝐷. Effect on wave
The flow around the bow sphere should have only a minimal impact on the flow around resistance
the stern sphere. One might initially guess that the wave resistance of the two sphere
system is approximately twice the wave resistance of a single sphere. However, favorable
and unfavorable wave superpositions create a much more complicated picture.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

308 25 Ship Wave Resistance

j j

Figure 25.10 Comparison of wave patterns and wave profiles created by two submerged
spheres for Froude numbers 𝐹𝑟 = 0.252 (upper half, favorable superposition) and 𝐹𝑟 = 0.282
(lower half, unfavorable superposition)

Figure . compares the wave patterns and profiles created by the spheres moving
with different velocities. It is obvious that the waves generated by the moving spheres
are much higher for the Froude number with unfavorable wave superposition and,
consequently, this results in a much higher wave resistance. 𝐹𝑟 = 0.282 is not the only
case of unfavorable wave superposition. Figure . compares the wave resistance
curves of a single sphere with the curve for the two sphere system. The latter shows
large fluctuations around the single sphere wave resistance. In the case of 𝐹𝑟 = 0.252,
the wave resistance coefficient of two spheres is much smaller than the 𝐶𝑊 -value of

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.2 Wave Superposition 309

j j
Figure 25.11 Wave resistance coefficient for a system of two submerged spheres; distance
between centers is 𝐿 = 10𝐷, submergence is 𝑠 = 𝐷

a single sphere. As expected, for 𝐹𝑟 = 0.282, we see a much higher wave resistance
coefficient.
Ships represent continuous pressure disturbances. Therefore, contributions to the ship Back to ships
wave system come from every part of the hull and not just bow and stern. The humps
and hollows in the wave resistance curves of ships are much less pronounced than
for the two sphere system. In fact, sometimes they are almost invisible. Details of the
pressure distribution along the vessel depend on hull shape and speed. Only a detailed
assessment with model tests or numerical analysis will provide a clear picture of the
wave generation for a specific hull.
The example of the two spheres makes clear that wave resistance may not be effectively
minimized for all speeds simultaneously. If the hull is optimized to produce a minimum
wave resistance at a specific speed, it may increase wave resistance at other speeds.
Therefore, optimization should be applied with care to avoid extreme measures which
decrease the overall performance of the ship.
One of the most common design measures to lower wave resistance is the bulbous Bulbous bow
bow. It lengthens and smooths the pressure disturbance created by the bow. The bow
wave system becomes lower, which decreases the wave pattern resistance. Even for
slow ships with already small wave pattern resistance, a bulbous bow might lower wave
breaking resistance by reducing the height of the bow wave.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

310 25 Ship Wave Resistance

25.3 Michell’s Integral

John Henry The first reasonably successful attempt to compute the wave resistance of a ship is
Michell attributed to the Australian mathematician John H. Michell (*–†). Michell
() derived an equation for the wave resistance that resulted in a triple integral. The
derivation of the equation is not for the mathematically fainthearted and is beyond the
scope of this book. In his derivation, Michell made the following basic assumptions:

Assumptions (i) potential flow, i.e. the flow is inviscid, incompressible, and irrotational
(ii) linearized free surface boundary condition, i.e. wave amplitudes are small com-
pared to wave lengths
(iii) the body boundary condition is linearized and satisfied at the center plane, i.e. the
beam is small compared with the length of the vessel, and the angle of hull surface
elements to the center plane is small.

The limitations of the third and second assumptions render the wave resistance values
from Michell’s theory close to useless for modern merchant vessel hull forms and speeds.
Practical applications are limited to extreme narrow hulls as they are used in catamarans
and other multihull vessels and to high Froude numbers. That having been said, one
has to acknowledge that Michell’s solution of the problem is mathematically sound and
proofed to be a big step forward in our understanding of ship waves and the resultant
resistance. Michell’s solution resulted in a multidimensional integral which can only
j j
be solved numerically for realistic hull forms. Michell () included a numerical
solution of his wave resistance integral for a ship and he reportedly has solved it for
many more cases. At a time where even slide rules were scarce and computers did not
exist even in science fiction, this was a formidable accomplishment by itself. Even with
today’s computer power special algorithms are necessary to accurately solve Michell’s
wave resistance integral. Readers are encouraged to read the summary on Michell’s
accomplishment by Tuck ().
Michell’s integral The following discussion of Michell’s integral uses the notation by Tuck () and
Tuck et al. (). A coordinate system is fixed to the bow and its 𝑥-axis points aft. The
vertical 𝑧-axis is positive upward and starts at the waterline. In Michell’s integral the
wave resistance is obtained from an analysis of the wave energy contained in the ship’s
wave pattern.
+𝜋∕2
𝜋
𝑅𝑤 = 𝜌 𝑣2 |𝐴(𝜃)|2 cos3 (𝜃) d𝜃 (.)
2 ∫
−𝜋∕2
The complex function 𝐴(𝜃) is known as a free wave spectrum or Kochin function, and
is a measure for the energy of ship waves traveling in direction 𝜃 (Tuck et al., ). 𝜃
measures the angle between the ship’s path and the direction of wave propagation. For
𝜃 = 0, waves follow the ship (transverse waves) and for 𝜃 = ±𝜋∕2 waves have crests
parallel to the ship’s path.
Free wave The free wave spectrum is a function of ship geometry and the propagation angle 𝜃.
spectrum
2
2i 𝑘0 [ ]
𝐴(𝜃) = 𝑃 (𝜃) + i𝑄(𝜃) (.)
𝜋 cos4 (𝜃)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.3 Michell’s Integral 311

The fundamental (transverse) wave number 𝑘0 on deep water is equal to

𝑔 1
𝑘0 = = (.)
𝑣2 𝐹𝑟2 𝐿

The functions 𝑃 (𝜃) and 𝑄(𝜃) incorporate the geometry of the hull. They are computed
as integrals along the length of a vessel from bow to stern.

𝐿 ( )
𝑘0 𝑥
𝑃 (𝜃) = 𝐹 (𝑥, 𝜃) cos d𝑥 (.)
∫ cos(𝜃)
0
𝐿 ( )
𝑘0 𝑥
𝑄(𝜃) = 𝐹 (𝑥, 𝜃) sin d𝑥 (.)
∫ cos(𝜃)
0

The function 𝐹 (𝑥, 𝜃) is an integral over the submerged center plane of the hull from 𝑧-integral
𝑧 = −𝑇 to the waterline at 𝑧 = 0.
0 ( 𝑘 𝑧 )
0
𝐹 (𝑥, 𝜃) = 𝑌 (𝑥, 𝑧) e cos2 (𝜃) d𝑧 (.)
∫
−𝑇

j Michell’s original integral employed the longitudinal derivatives of the hull offsets j
𝜕𝑌 (𝑥, 𝑧)∕𝜕𝑥 whereas Tuck’s solution works directly with the half breadths 𝑌 (𝑥, 𝑧) (Tuck,
). The equations above assume that the hull offsets at bow and stern vanish, i.e.
𝑌 (0, 𝑧) = 𝑌 (𝐿, 𝑧) = 0. Practical algorithms for numerical integration of the integrals in
Equations (.), (.), (.), and (.) are described in Tuck () and Tuck
et al. (). Read () also contains a description and provides some specifics on
the treatment of transoms where 𝑌 (𝐿, 𝑧) ≠ 0.
Figure . shows the wave resistance coefficient for a standard Wigley hull with offsets Wigley hull
[ ( ) ][ ( )2 ]
𝐵 2𝑥 − 𝐿 2 𝑧
𝑌 (𝑥, 𝑧) = 1− 1− (.)
2 𝐿 𝑇
with 𝑥 ∈ [0, 𝐿] and 𝑧 = [−𝑇 , 0]. 𝐿, 𝐵, and 𝑇 are length, beam, and draft of the vessel,
respectively.
As for the case of two spheres, the theoretical wave resistance curve for the Wigley
hull computed with Michell’s integral (.) features numerous humps and hollows.
With decreasing Froude number, fluctuations between humps and hollows become
more frequent but also decline in amplitude. Figure . also contains the range of
experimental results for the wave resistance already featured in Figure . (Bai and
McCarthy, ). It is pretty obvious that discrepancies between experimental and
numerical results are significant in the relevant range of Froude numbers for merchant
vessels 𝐹𝑟 ∈ [0.15, 0.35]. Even the humps and hollows are not completely aligned.
However, one has to bear in mind, that wave resistance cannot be measured directly. Its
derivation from experimental results includes estimates of other resistance components
which have their own limitations.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

312 25 Ship Wave Resistance

Wigley hull wave resistance

5
Cw from Michell's integral
range of experimental values
[10−3 ]

4
wave resistance coefficient Cw

0
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Froude number Fr [−]

Figure 25.12 Wave resistance coefficient for a Wigley hull with length–beam ratio of 𝐿∕𝐵 = 10
and beam–draft ratio of 𝐵∕𝑇 = 1.6

j Michell’s integral predicts wave resistance quite accurately for high Froude numbers j
𝐹𝑟 > 0.4 and for very slender vessels with large 𝐿∕𝐵 and small 𝐵∕𝑇 ratios. However,
linear wave theory is of limited use in predicting wave resistance for common merchant
and naval vessel hull forms at the low and moderate Froude numbers they usually sail at.
Reliable prediction of wave resistance requires the solution of a nonlinear problem using
advanced numerical tools which go beyond the scope of this book. A brief introduction
and further references are provided in the following section.

25.4 Panel Methods

Panel methods Today, panel methods are the most reliable way of computing the wave resistance of
ships. These numerical methods derive their name from the small elements which
are used to discretize the underlying mathematical equations. Figure . shows
the bow of a container ship discretized into small quadrilateral and triangular panels.
Quadrilateral panels are usually preferred because they reduce the total number of
panels. However, quadrilaterals may not be flat. If quadrilateral panels become too
twisted, like at the intersection of bulbous bow and stem in Figure ., it is better to
split quadrilaterals into triangles.
Panel methods belong to the larger class of boundary element methods (BEM), which
are also employed in structural analysis. Boundary element methods solve three-
dimensional problems by reducing them to equations which have to be satisfied on the
domain boundaries. This effectively reduces the D problem to a two-dimensional
 Hence the name boundary element method.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.4 Panel Methods 313

Figure 25.13 Discretization of vessel bow into small panels for wave resistance computation

j problem. Instead of subdividing the domain into several hundred thousand or millions j
of small finite volumes, the boundaries of the domain are subdivided into a couple of
hundred or thousand panels 𝑆0 , 𝑆1 , … , 𝑆𝑁 thus substantially reducing the computa-
tional effort.
D panel methods have been successfully used first in the design of aircraft wings and
fuselages (Hess and Smith, ). A general derivation of the underlying theory and
equations can be found in Katz and Plotkin (). Two principal methods are in use:

(i) The direct method: Green’s integral theorem is used to derive a Fredholm integral
equation of the second kind for the velocity potential. It utilizes the Laplace
equation and the boundary conditions of the problem to reduce the portion of
the boundary which affects the solution. Typically, only the ship hull itself has to
be discretized. If a nonlinear free surface boundary condition is employed, the
free surface has to be discretized as well (Newman, ).

(ii) The indirect method: the potential is assumed to be the sum of singularity dis-
tributions over all panels. Different types of singularities or combinations of
singularities are in use. Distribution of sources are most common. Typically, the
source strength 𝜎𝑗 is assumed to be constant for each panel. The body boundary
condition and, if necessary, the free surface boundary condition are employed to
determine the source strength values. The boundary conditions will only be truly
satisfied at the panel centers, the so-called collocation points.

Important developments and modern panel methods for ship wave resistance com-
putations are described by Jensen (), Nakos (), Raven (), and Janson
().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

314 25 Ship Wave Resistance

Although viscous effects are ignored in panel methods, they are well suited for wave
resistance computations and the shape optimization of bulbous bows and the entrance
of ship hulls. They may also serve as a preprocessor for more complex D boundary
layer and general viscous flow computations, especially if the latter involve a free surface.

References
Abramowitz, M. and Stegun, I., editors (). Handbook of mathematical functions
with formulas, graphs, and mathematical tables. Number  in Applied Mathemat-
ics Series. National Bureau of Standards, Washington, DC. Tenth printing, with
corrections.
Bai, K. and McCarthy, J. (). Proceedings of the workshop on ship wave resistance
computations. Technical report, David W. Taylor Ship Research and Development
Center, Bethesda, Maryland .
Havelock, T. (). Wave resistance: some cases of three-dimensional fluid motion.
Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathe-
matical and Physical Character, ():–.
Hess, J. and Smith, A. (). Calculation of potential flow about arbitrary bodies.
Progress in Aeronautical Sciences, :–.
Janson, C.-E. (). Potential flow panel methods for the calculation of free surface
j j
flows with lift. PhD thesis, Department of Naval Architecture and Ocean Engineering,
Chalmers University of Technology, Gothenburg, Sweden.
Jensen, G. (). Berechung der stationären Potentialströmung um ein Schiff unter
Berücksichtigung der nichtlinearen Randbedingung an der Wasseroberfläche. Tech-
nical Report , Technische Universität Hamburg-Harburg.
Katz, J. and Plotkin, A. (). Low-speed aerodynamics. Cambridge Aerospace Series.
Cambridge University Press, New York, NY, second edition.
Michell, J. (). The wave resistance of a ship. Philosophical Magazine Series ,
():–.
Nakos, D. (). Ship wave patterns and motions by a three-dimensional Rankine
panel method. PhD thesis, Massachusetts Institute of Technology, Cambridge, MA,
USA.
Newman, J. (). Marine hydrodynamics. The MIT Press, Cambridge, MA.
Raven, H. (). A solution method for the nonlinear ship wave resistance problem.
PhD thesis, Technical University Delft, Delft, The Netherlands.
Read, D. (). A drag estimate for concept-stage ship design optimization. PhD thesis,
The Graduate School, The University of Maine, Orono, ME.
Tuck, E. (). Wave resistance of thin ships and catamarans. Technical Report T,
Applied Mathematics Department, The University of Adelaide, Australia. Unaltered
October  reprint of internal report T.
Tuck, E. (). The wave resistance formula of J.H. Michell () and its significance
to recent research in ship hydrodynamics. Journal of Australian Math. Soc. Series B,
():–.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

25.4 Panel Methods 315

Tuck, E., Lazauskas, L., and Scullen, D. (). Sea wave pattern evaluation – Part 
report: Primary code and test results (surface vessels). Technical report, Applied
Mathematics Department, The University of Adelaide, Australia.

Self Study Problems

. A ship of length 𝐿𝑊𝐿 = 172 m creates pronounced bow and stern wave systems.
Similar to the system of two spheres you may assume that other parts of the wave
making are negligible for this exercise. Compute the length of the transverse
waves and reason whether or not the design speed of 𝐹𝑟 = 0.23 is a good choice
regarding low wave resistance.
. If you want to lower the wave resistance of a ship, would it be better to make it
wider and shallower or narrower (smaller beam) and deeper (higher draft)? Give
reasons.
. Read the paper by Tuck () on Michell’s wave resistance theory.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

316

26
Ship Model Testing

Resistance and powering requirements have to be determined as early as possible in

the design process because size and layout of engine and propulsor heavily influence
general arrangements of the ship. Despite enormous advances made in computational
methods, model testing is still the most reliable way of predicting resistance and power-
ing requirements for a future ship. In this chapter we discuss the physical assets used
in model tests: ship model, propeller model, and the towing tank or cavitation tunnel
which represent the waterway. The chapter closes with a selection of websites for model
basins around the world.

Learning Objectives
j j

At the end of this chapter students will be able to:

• understand layout and function of scale models and testing facilities

• discuss advantages and disadvantages of different testing facilities
• know about modern model testing facilities

26.1 Testing Facilities

Model basins Facilities for ship model testing are found in research institutions and universities. Ship
models are tested either in long water basins called towing tanks or in circulating water
tunnels. Basins and circulating water tunnels stand in for oceans, lakes, and waterways
and provide a controlled testing environment. Research institutions with a towing tank
are often referred to as model basins due to importance and size of the towing tank.
Figure . shows a view of the towing tank of the Hamburg Ship Model Basin (HSVA)
in Hamburg, Germany. It is  m long,  m wide, and  m deep and of fairly typical
size for a model basin that offers testing and consulting services for marine industry
and government clients.
Most universities have smaller tanks for teaching and research purposes. Figure .
shows the towing tank at the University of New Orleans as an example (length . m,
width . m, water depth . m). Towing tank buildings often block sun light to slow
algae growth which reduces the need for chemical additives and filtering of the water.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26.1 Testing Facilities 317

j j
Figure 26.1 Towing tank of the Hamburg Ship Model Basin, Photo courtesy of Hamburgische
Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.de

26.1.1 Towing tank

Towing tanks are long basins filled with fresh water. They range in length from  m to Towing tank
more than  m, with widths between  m and  m, and with water depths between
. m and  m. Figure . shows the principal setup of a towing tank. One end typically
features a narrow appendix called the trim tank, which provides easy access to a floating
model. Before the actual testing, a model is lowered into the trim tank for ballasting.
Afterwards the model is hooked up to sensors and the data acquisition system on the
towing carriage. Smaller tanks are equipped with underwater viewing windows which
are often considered impractical for basins with great water depth.
A pair of rails for the towing carriage is mounted on top of the side walls. The rails Rails
have to be precisely parallel to each other and parallel to the water surface. The latter
requires that the rails follow the Earth’s curvature, i.e. the rails form an arc when viewed
from the side. This is especially critical for tests where the model is rigidly attached to a
force balance to measure maneuvering or wave forces. In the case of a typical resistance
test, sinkage would be underestimated in the middle of a towing tank if the rails did not
follow the Earth’s curvature.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

318 26 Ship Model Testing

j j
Figure 26.2 Towing tank at the School of Naval Architecture and Marine Engineering of the
University of New Orleans

Figure 26.3 Schematic view of a towing tank

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26.1 Testing Facilities 319

Figure 26.4 Definition of rail sagitta 𝑠

The height or sagitta 𝑠 of a circular arc is a function of its radius 𝑅 and the length 𝐿 of
the chord.
√
𝐿2
𝑠 = 𝑅 − 𝑅2 − (.)
4
According to the World Geodetic System  (WGS ), the mean radius of the Earth
is 𝑅 =  . m (NIMA, ). For the towing tank at the Hamburg Ship
Model Basin (HSVA, Hamburg, Germany) with a length of 𝐿 = m, the rails are
𝑠 = 1.8 mm higher in the middle than at the ends. The 𝐿 = 1324 m long tank at the
Krylov State Research Center (St. Petersburg, Russia) needs a sagitta of 𝑠 =. mm
( / in).
A towing carriage is a moving laboratory in itself. It carries power supplies and elec- Towing carriage
tronics to process data from the sensors. In large basins the carriage is manned and has
its own drive motors. In smaller model basins, the towing carriage is often pulled by a
j cable system and operated from a fixed outside station. Towing carriages must be free j
of vibrations that could be transferred into the resistance measurement system. This
places great demands on the speed control of the carriage motors, roundness of wheels,
alignment of rails, and stiffness of the carriage structure.
Most basins feature a wave maker at the end opposite to the trim tank. Wave makers Wave maker
consist of a hydraulically actuated flap or board. A computer controls the board move-
ment to create waves with a wide range of frequencies and wave amplitudes. Typically,
wave forces and motions of stationary or moving models can be measured only in
head seas. However, model basins with wide tanks might also feature segmented wave
makers on the long side of the tank capable of creating waves that propagate at an angle
to the model’s course. A beach at the trim tank end of the basin dampens the waves
generated by models and wave maker. An effective beach will reflect very little wave
energy back into the tank.
Towing a ship model through the basin has the advantage that it reflects the real ship– Advantages and
waterway system: the water is at rest and the vessel moves. A model creates a wave disadvantages
pattern dynamically similar to the wave system of the full scale vessel, if the tests are
conducted at the same Froude number. Moving the model at a constant speed is much
easier and cheaper than creating a uniform flow over the entire cross section of a tank
by pumping water around. Limited length of a basin, however, restricts the usable
observation time, especially for high speed vessels. Another difficulty for towing tanks
is the simulation of cavitation. Very few model basins have the necessary capability to
lower the ambient air pressure above the water surface. Such a tank must be enclosed
in an air tight bunker and carriage operation has to be fully automated. The interested
reader might look for the depressurized wave tank of the Dutch model basin MARIN
in Wageningen as an example.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

320 26 Ship Model Testing

Figure 26.5 Schematic of a cavitation tunnel without free surface

j 26.1.2 Cavitation tunnel j

Circulating water In a circulating water tunnel the model is stationary, and a pump moves the water past
tunnel the model. Most circulating water tunnels can reduce the static pressure in the fluid
by reducing the effect of atmospheric pressure. Pressure distributions and not just
pressure forces become dynamically similar, which is necessary to simulate propellers
and foils at model scale that are subject to cavitation (see Chapter ). For this reason,
circulating water tunnels are more commonly known as cavitation tunnels.
Cavitation tunnel Figure . shows a schematic view of a cavitation tunnel. In contrast to wind tunnels,
circulating water tunnels are built upright: in small tunnels the test section might
be as little as  m height above the ciculation pump. Large cavitation tunnels easily
span several stories of a building. Although construction is more expensive, the height
difference between measuring section and circulation pump increases the static pressure
at the latter, thus forestalling cavitation of the pump itself. The fluid accelerated by the
circulation pump is slowed down in a diffusor to prevent cavitation at the vanes. Foil
shaped vanes help to redirect the flow around corners in the tunnel.
Flow equalizer Before the nozzle accelerates the fluid into the test section, the flow passes through
and test section a grid which acts as a flow equalizer. Together, grid and nozzle aim to make the flow
as uniform as possible across the test section. Test sections vary in shape and size
considerably. Some are circular but most are rectangular, with widths from a few inches
to several meters.
Propeller models are mounted to a dynamometer which rotates the propeller and
records thrust, torque, and rate of revolution (see Chapter ). In tunnels with large

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26.2 Ship and Propeller Models 321

Figure 26.6 A model is prepared for cutting on the five axis mill. Photo courtesy of
Hamburgische Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.de

j j
cross section, models of the submerged hull of a ship may be mounted in front of the
propeller model to simulate the wake field.
In a cavitation tunnel, the flow into the test section is never truly uniform. However, Advantages and
deviations from an average flow velocity will be small for a carefully designed tunnel. disadvantages
Circulating water tunnels allow measurements for prolonged periods of time. It is
relatively easy to adjust the ambient pressure to the correct scale of the model which
enables the detailed investigation of cavitation phenomena. Most tunnels are completely
filled with water. Therefore, effects of waves cannot be studied due to the lack of a
free surface. Some of the larger tunnels can be operated with a free surface so that the
total resistance of a ship model may be measured. Since the cross sections of cavitation
tunnels are usually much smaller than the cross section of towing tanks, effects due to
blockage and walls have to be carefully studied.

26.2 Ship and Propeller Models

Ship models for resistance and propulsion tests are scale replicas of the actual vessels’ Scale
outer hull shape. All linear dimensions are reduced by the same scale factor 𝜆.

𝐿𝑆 𝐵𝑆 𝑇
𝜆 = = = 𝑆 (.)
𝐿𝑀 𝐵𝑀 𝑇𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

322 26 Ship Model Testing

Consequently, surfaces and volumes scale with 𝜆2 and 𝜆3 , respectively.

𝑆𝑆 𝑉𝑆
𝑆𝑀 = 𝑉𝑀 = (.)
𝜆2 𝜆3

The subscript 𝑆 marks properties of the full scale vessel and the subscript 𝑀 stands
for model properties. In general, models do not feature deck houses, equipment, or
interior structure.
Construction Models are made as long as possible given the restrictions of towing tank or circulating
water tunnel. Some of the larger model basins test models up to  m in length. Most
models are made of wood. Wood is relatively cheap and easily machined or shaped with
hand tools. The resulting models are dimensionally stable if the model is properly treated
and sealed. First, thick boards are roughly sawed according to water line contours and
glued together into a block (see Figure .). The actual shaping is done with industrial
five axis mills. The massive computer controlled router can be seen hanging from its
gantry in the right half of Figure .. As a final step, a model is sanded with  or
 grit sand paper and painted to achieve a hydraulically smooth surface. Cheaper
than wooden models are models made of paraffin wax. The wax can be remelted and
reused. However, the models are more sensitive to temperature changes and will not
keep their shape as long as wooden models.
Accuracy Models are manufactured according to specifications in the ITTC recommended pro-
cedures (ITTC, b). The length of a model should be accurate within ±. % of
j length between perpendiculars or at least within ±1 mm of its target value. The latter j
limit also applies to all other linear dimensions. The locations of knuckle lines and
transoms in particular have to be modeled accurately because they may significantly
influence the resistance.
Appendages If possible, the ITTC recommends conducting tests with and without appendages to
gauge their effects on the resistance. Movable appendages, like rudders and stabilizer
fins, should not be included. However, fixed appendages like bilge keels, bossings, and
struts are quite often omitted in resistance tests as well. The larger the scale factor 𝜆,
the smaller a model will be and the more difficult it becomes to accurately represent the
appendages. They might get too flimsy and would easily be bent out of shape during
testing. In addition, appendages mostly affect the viscous resistance which is not scaled
correctly in model tests due to the lower Reynolds numbers (see Chapter ).

26.2.1 Turbulence generation

Sand strips or Lower Reynolds numbers at model scale are also responsible for the requirement to have
trip wires turbulence generators. On a full scale vessel, the boundary layer will transition from
laminar to turbulent flow a couple of centimeters behind the bow. As a consequence,
the boundary layer will be turbulent over essentially the entire hull. On a model this
transition would happen at a position considerably further aft relative to its length. To
ensure that most of the model boundary layer is turbulent, the flow is tripped by a wire
(.–. mm diameter) or a strip of rough sand at a station % of the length behind the
forward perpendicular. A bulbous bow receives its own turbulence generator about
one third of its length from the tip.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26.2 Ship and Propeller Models 323

Some model basins prefer a series of studs parallel to the bow contour. Cylindrical studs
stick out into the flow and create a turbulent wake behind them. Diameter, height, and
separation should follow the guidance given in the ITTC recommended procedures
(ITTC, b).
Enforcing a turbulent boundary layer allows us to use the same model–ship correlation
line in the computation of the frictional resistance of model and ship (see Chapter ).
It also prevents problems with separation, which is more likely in a laminar boundary
layer and may lead to an exaggerated viscous pressure resistance for the model. As a
consequence, the performance prediction would fail for the full scale vessel.

26.2.2 Loading condition

Typically, a model is tested over a range of speeds at different loading conditions (design Ballast based on
draft, fully loaded, ballast, etc.). The model is ballasted according to its volumetric volumetric
displacement
displacement for the loading condition in question. Waterline markings on the model
are solely for checking heel and trim angles. The distribution of ballast in the model is not
important for resistance tests as long as displacement and attitude are correct. However,
dynamically similar moments of inertia and centers of gravity must be maintained for
maneuvering and seakeeping tests. Ballast must be stored accordingly on the model.
Calculation of the necessary ballast is straightforward: Ballast
calculation
j j
. Determine volumetric displacement 𝑉 for the targeted loading condition, e.g.
from the ship’s curves of form. Equation (.) provides the relationship between
model volume 𝑉𝑀 and ship volume 𝑉𝑆 .

. Weigh the model as built and record its mass 𝑚𝑀 . This mass should include all
the equipment that will be placed on the model during testing. The equipment
includes half of the mass of the yoke which connects model and towing carriage
(see Figure .).

. Measure the temperature 𝑡𝑊 of the tank water at a depth of half the model draft.

. Measure the tank water density 𝜌𝑀 with a calibrated hydrometer or, if this is not
available, use the value from the water properties tables in ITTC ().

. Determine the total model mass for the selected loading condition

Δ𝑀 = 𝜌𝑀 𝑉𝑀 (.)

. Compute the ballast mass 𝑚ballast by subtracting the model mass 𝑚𝑀 from the
total mass 𝜌𝑀 𝑉𝑀 of the model for the target loading condition

𝑚ballast = Δ𝑀 − 𝑚𝑀 (.)

. Adjust the ballast mass if tests are conducted over a longer time period where
water temperature and density may change.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

324 26 Ship Model Testing

Figure 26.7 The beginnings of a five bladed propeller model. Photo courtesy of Hamburgische
Schiffbau-Versuchsanstalt GmbH (HSVA), www.hsva.de

j j

26.2.3 Propeller models

Fabrication of propeller models requires high precision (Figure .). In most cases
computer controlled milling machines are used; however, D-printing may also be
effective. The diameter of model propellers manufactured according to ITTC (a)
may deviate by at most ±0.1 mm from its nominal value. Other important propeller
parameters like pitch and shape of blade sections also have tight tolerances. The blade
surfaces are polished to less than . 𝜇m mean roughness. Turbulence generation is
not used on propeller models because it would negatively affect their lift to drag ratio.
Relevant Reynolds numbers for the propeller at model scale must be high enough to
ensure that the boundary layer flow is mostly turbulent.

26.3 Model Basins

Table . lists model basins and their websites. This list is definitely incomplete and
my apologies if your favorite institution is missing. Emphasis has been placed on model
basins whose websites provide information and photos of their facilities. All reported
web links were active in September . The ITTC offers an extensive list of facilities
on its website at www.ittc.info. Some of the institutions, like the David Taylor Model
Basin in the United States, do not provide a lot of information because they mostly
perform classified research for their respective naval forces.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

26.3 Model Basins 325

Table 26.1 A selection of model basins around the world, sorted alphabetically according to their commonly used
abbreviations

Name Location Website

China Ship Scientific Research Center Wuxi, Jiangsu, P.R. www.cssrc.com/
(CSSRC) China
David Taylor Model Basin (DTMB), Naval West Bethesda, MD, www.navsea.navy.mil/Home/
Surface Warfare Center at Carderock USA WarfareCenters/NSWCCarderock.aspx
Canal de Experiencias Hidrodinámicas de Madrid, Spain www.cehipar.es/cehiparweb/index.php?lang=
El Pardo (El Pardo Model Basin) english
Marine Technology Research Institute Rome, Italy www.insean.cnr.it/
(INSEAN)
Hamburg Ship Model Basin (HSVA) Hamburg, Germany www.hsva.de
Korea Research Institute of Ships and Daejeon, South www.kriso.re.kr/eng/
Ocean Engineering (KRISO) Korea
Krylov State Research Center St. Petersburg, krylov-center.ru/eng
Russia
Maritime Research Institute Netherlands Wageningen, The www.marin.nl
(MARIN) Netherlands
SINTEF Ocean (formerly Marintek) Trondheim, www.sintef.no/en/ocean/laboratories/
Norway
National Maritime Research Institute Tokyo, Japan www.nmri.go.jp/index_e.html
j (NMRI) j
National Research Council Canada St. John’s, www.nrc-cnrc.gc.ca/eng/solutions/facilities/
Newfoundland, marine_performance/towing_tank.html
Canada
QinetiQ Maritime Test Facilities Haslar, United www.qinetiq.com/services-products/
Kingdom maritime/platform-readiness/Pages/
maritime-test-facilities.aspx
SSPA Sweden AB Gothenburg, http://www.sspa.se
Sweden
Schiffbau-Versuchsanstalt Potsdam Potsdam, Germany www.sva-potsdam.de/home.html
(SVA)
Vienna Model Basin (SVA) Vienna, Austria www.sva.at

References
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
ITTC (a). Propeller model accuracy. International Towing Tank Conference,
Recommended Procedures and Guidelines .---. Revision .
ITTC (b). Ship models. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .
NIMA (). Department of Defense World Geodetic System . Technical Report
NIMA TR., third edition, Amendment , National Imagery and Mapping
Agency.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

326 26 Ship Model Testing

Self Study Problems

. What are the advantages of conducting a resistance test in a towing tank compared
with a resistance test in a cavitation tunnel?
. Why must the rails follow the Earth’s curvature?
. What is necessary to conduct model tests with cavitation conditions similar to
the full scale vessel?

. Which geometric property of a hull determines how much ballast must be added
to a ship model before testing?
. Research the dimensions of the towing tank of the Vienna Model Basin.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

327

27
Dimensional Analysis

Dimensional analysis is an important tool in the search for functional relationships

in experimental as well as theoretical models. Typically, this topic is part of courses
in fluid mechanics. However, for the sake of completeness this chapter provides a
short summary, explaining the purpose and application of dimensional analysis. The
Buckingham 𝜋-theorem is recaptured and applied to the ship resistance problem. We
rediscover the characteristic numbers already found by making the Navier-Stokes
equations dimensionless.

Learning Objectives
j j
At the end of this chapter students will be able to:

• apply dimensional analysis

• define dimensionless characteristic numbers

27.1 Purpose of Dimensional Analysis

Experiments represent an obvious means to obtain more insight into the flow processes
around a ship hull and the resultant resistance. However, before we conduct any
experiments, two important questions need to be answered:

(i) which parameters influence the resistance, and

(ii) how can we obtain the most information from the smallest set of experiments?

Let us assume that the flow around a submerged sphere is governed by four variables: Need for
dimensional
analysis (DA)
(a) water density 𝜌 (type of fluid),
(b) speed 𝑣 (operation),
(c) diameter 𝐷 (geometry), and
(d) dynamic viscosity 𝜇 (another fluid property).

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

328 27 Dimensional Analysis

Then the sphere’s resistance 𝑅 will be a yet unknown function  of 𝜌, 𝑣, 𝐷, and 𝜇.

𝑅 =  (𝜌, 𝑣, 𝐷, 𝜇) (.)
To determine the function  , we may select  different values for each variable. The
total number of experiments to cover all combinations of variable values would be
N = 104 = 10 000 experiments. Even if we manage to perform  experiments per day,
it would take  days, or more than half a year, to assemble the results. We rarely have
this much time.

27.2 Buckingham 𝜋-Theorem

Buckingham Dimensional analysis helps to reduce the number of relevant variables by combin-
𝜋-theorem ing them into dimensionless parameters. The theory is based on the Buckingham
𝜋-theorem, named after the American physicist Edgar Buckingham (* – †),
although it was first proved by the French mathematician Joseph Bertrand (* –
†) (Bertrand, ). Assume we have a physically correct formula that involves 𝑛
physical variables. The variables may be expressed in terms of 𝑘 independent fundamen-
tal physical quantities, e.g. time, length, and force. Then the Buckingham 𝜋-theorem
postulates that the physically correct formula is equivalent to an equation involving
a set of 𝑚 = 𝑛 − 𝑘 dimensionless parameters, so-called 𝜋-terms. The dimensionless
parameters are constructed from the original variables.
j j
Considering the example in the previous section, the formula for the resistance of a
sphere has 𝑛 = 4 variables. Independent physical quantities are force measured in
Newtons [N], length [m], and time [s] or simply 𝑘 = 3. Therefore, a single dimensionless
parameter 𝑚 = 𝑛−𝑘 = 4−3 = 1 can be used to describe the relationship  resulting in a
significant reduction in the number of necessary experiments. We go from  (𝜌, 𝑣, 𝐷, 𝜇)
to  (𝑅𝑒). Thus,  instead of   experiments would yield the same information
about the drag of spheres.

27.3 Dimensional Analysis of Ship Resistance

The resistance 𝑅𝑇 of a ship may serve as an illustration of the steps involved in a
dimensional analysis (DA).

Collection of . First, we collect all variables known (or believed) to affect ship resistance. This is
relevant the most difficult task of DA, often supported by observation, experience, and
variables
trial and error.
Let us assume 𝑅𝑇 depends on density of water 𝜌, velocity 𝑣 of the ship, dimen-
sion of the ship represented by its length 𝐿, dynamic viscosity 𝜇 of the water,
acceleration of gravity 𝑔, and pressure 𝑝:
𝑅𝑇 =  (𝜌, 𝑣, 𝐿, 𝜇, 𝑔, 𝑝) (.)

There is a total of six variables (𝑛 = 6) which involve the following fundamental

units:

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

27.3 Dimensional Analysis of Ship Resistance 329

[ ] [ ]
kg Ns2 kg m
(a) 𝜌 water density or using N =
m3 m 4 s2
[ ]
m
(b) 𝑣 ship speed
s
(c) 𝐿 ship length [m]
] [
Ns
(d) 𝜇 dynamic viscosity relevant to viscous forces
m2
[ ]
m
(e) 𝑔 acceleration of gravity relevant to gravity forces
s2
[ ]
N
(f ) 𝑝 pressure relevant to pressure forces
m2
Therefore, the fundamental physical quantities involved are force, length, and
time with the independent units Newton [N], meter [m], and seconds [s], i.e.
𝑘 = 3. Mass [kg] could be used instead of force, but the former is more con-
venient in this case. The Buckingham 𝜋-Theorem tells us that the number of
dimensionless parameters 𝑚 is equal to the number of variables minus the number
of independent units
𝑚 = 𝑛−𝑘 = 6−3 = 3

. The unknown resistance function 𝑅𝑇 =  (𝜌, 𝑣, 𝐿, 𝜇, 𝑔, 𝑝) will contain products of Product of

the variables, possibly raised to yet unknown powers 𝑎 through 𝑓 , and a constant variables

j factor 𝐶. j
𝑅𝑇 = 𝐶 𝜌𝑎 𝑣𝑏 𝐿𝑐 𝜇𝑑 𝑔 𝑒 𝑝𝑓 (.)

In physical units the equation requests [N] = [N]. We will derive the values of
the powers in subsequent steps.
. The equation above can only be satisfied if the product on the right-hand side Comparison of
yields the independent unit of a force [N]. We substitute the variables with their coefficients
corresponding independent units in Equation (.).
[ ]1 [ 2 ]𝑎 [ ] [ ] [ ]𝑑 [ ]𝑒 [ ]𝑓
Ns m 𝑏 𝑐 Ns m N
N = ⋅ ⋅ m ⋅ ⋅ ⋅ (.)
m4 s m2 s2 m2
From Equation (.) follow three conditions, one for each physical quantity. The
sum of powers on the left- and right-hand side have to be equal for each unit
separately, because they are independent of each other.
[ ]
for force N : 1 = 𝑎+𝑑 +𝑓 (.)
[ ]
for length m : 0 = −4𝑎 + 𝑏 + 𝑐 − 2𝑑 + 𝑒 − 2𝑓 (.)
[]
for time s : 0 = 2𝑎 − 𝑏 + 𝑑 − 2𝑒 (.)

The equations are derived by collecting corresponding terms from Equation (.).
On the left-hand side of Equation (.) only N1 appears. The power is equal to 
for the force unit [N] (.) and zero for unit length [m] (.) and time [s] (.).
On the right-hand side we collect powers of all appearances of the independent
unit. Powers which are part of a denominator are preceded with minus signs.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

330 27 Dimensional Analysis

Reduction . In total, we derive three equations for six unknown powers 𝑎, 𝑏, 𝑐, 𝑑, 𝑒, and 𝑓 .
Although we cannot solve for all of them, we are able to reduce the number of
unknown powers to three by expressing powers 𝑎, 𝑏, and 𝑐 in terms of powers 𝑑,
𝑒, and 𝑓 . From Equation (.) follows

𝑎 = 1−𝑑 −𝑓 (.)

and from Equation (.) we get

𝑏 = 2𝑎 + 𝑑 − 2𝑒

Substituting the result for 𝑎 into the latter yields

𝑏 = +2 − 2𝑑 − 2𝑓 + 𝑑 − 2𝑒
= 2 − 𝑑 − 2𝑒 − 2𝑓 (.)

Using Equations (.) and (.) allows us to derive 𝑐 from Equation (.):

0 = −4 + 4𝑑 + 4𝑓 + 2 − 𝑑 − 2𝑒 − 2𝑓 + 𝑐 − 2𝑑 + 𝑒 − 2𝑓
−𝑐 = −2 + 𝑑 − 𝑒
𝑐 = 2−𝑑 +𝑒 (.)

Dimensionless . We replace 𝑎, 𝑏, and 𝑐 in (.) with the results from Equations (.), (.), and
j parameters (.). Subsequently, we sort the terms with respect to equal power coefficients. j

𝑅𝑇 ∼ 𝜌1−𝑑−𝑓 ⋅ 𝑣2−𝑑−2𝑒−2𝑓 ⋅ 𝐿2−𝑑+𝑒 ⋅ 𝜇𝑑 ⋅ 𝑔 𝑒 ⋅ 𝑝𝑓

[ ] [ 𝑒 𝑒] [ 𝑓 ]
2 2 𝜇𝑑 𝑔 𝐿 𝑝
∼ 𝜌𝑣 𝐿 ⋅ 𝑑 𝑑 𝑑 ⋅ ⋅
2𝑒 𝑓 2𝑓
𝜌 𝑣 𝐿 𝑣 𝜌 𝑣
𝑓
[ ]𝑑 [ ]𝑒 ⎡ ⎤
1 2 2 𝜇 𝑔𝐿 ⎢ 𝑝 ⎥
∼ 𝜌𝑣 𝐿 ⋅ ⋅ ⋅ (.)
2 𝜌𝑣𝐿 𝑣2 ⎢ 1 𝜌𝑣2 ⎥
⎣ ⎦ 2

The terms in brackets are indeed dimensionless. Introducing the kinematic

viscosity 𝜈 = 𝜇∕𝜌, the first pair of square brackets (power 𝑑) is the reciprocal
value of the Reynolds number 𝑅𝑒.
𝑣𝐿
Reynolds number 𝑅𝑒 = (.)
𝜈

The second pair of brackets (power 𝑒) contains the reciprocal value of the squared
Froude number 𝐹𝑟.
𝑣
Froude number 𝐹𝑟 = √ (.)
𝑔𝐿

The third set of brackets yields a form of the Euler number or pressure coefficient.

𝑝
Euler number 𝐸𝑢 = (.)
1
2
𝜌𝑣2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

27.3 Dimensional Analysis of Ship Resistance 331

Dimensional analysis reveals that the calm water resistance of a ship depends on the
type of fluid (density 𝜌), ship speed 𝑣, ship geometry (represented by length 𝐿 or wetted
surface 𝑆), and the three dimensionless parameters: Reynolds number 𝑅𝑒, Froude
number 𝐹𝑟, and Euler number 𝐸𝑢.
1 2
𝑅𝑇 = 𝜌 𝑣 𝑆 ⋅  (𝑅𝑒, 𝐹𝑟, 𝐸𝑢) (.)
2

We often express this in the form of a dimensionless resistance coefficient 𝐶𝑇

2𝑅𝑇
𝐶𝑇 = =  (𝑅𝑒, 𝐹𝑟, 𝐸𝑢) (.)
𝜌 𝑣2 𝑆
The actual function  will have to be determined by experiments.

References
Bertrand, J. (). Sur l’homogénéité dans les formules de physique. Comptes rendus,
():–.

j Self Study Problems j

. Assume that thrust 𝑇 of a propeller depends on the seven physical variables

density 𝜌, kinematic viscosity 𝜈, gravitational acceleration 𝑔, diameter 𝐷, speed
of advance 𝑣𝐴 , pressure 𝑝, and rate of revolution 𝑛. How many dimensionless
characteristic numbers will govern the similarity of model tests measuring thrust?
. Thrust 𝑇 of a propeller is assumed to depend on water density 𝜌, dynamic viscosity
of the water 𝜇, acceleration of gravity 𝑔, diameter 𝐷, speed of advance 𝑣𝐴 , pressure
in the fluid 𝑝, and rate of revolution 𝑛. Perform a dimensional analysis to derive
the dimensionless coefficients governing propeller thrust. Look-up and name the
dimensionless coefficients.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

332

28
Laws of Similitude

Need for model Even with cutting-edge CFD tools naval architects cannot reliably compute the resis-
testing tance of a ship for daily design purposes. Although accuracy and reliability of CFD
results have steadily improved, model tests will be used to confirm results for the fore-
seeable future. Since most ships are one-of-a-kind designs, prototypes cannot be used
to improve the design which further increases the importance of model tests.
However, model tests are useful only if they are accompanied by methods which transfer
physical quantities measured at model scale to the full scale design. This chapter explains
the fundamental concepts of similarity and points out the difficulties that arise if only
partial dynamic similarity is possible.

j j
Learning Objectives
At the end of this chapter students will be able to:

• understand the concepts of geometric, kinematic, and dynamic similarity

• interpret Froude number, Reynolds number, and Euler number in the context of
dynamic similarity

• identify the limitations of similarity in ship model tests

• discuss Froude’s hypothesis and the ITTC form factor method

28.1 Similarities
Required Three basic conditions regulate the conversion of model test results into full scale
similarities results:

. geometric similarity,

. kinematic similarity, and

. dynamic similarity.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.1 Similarities 333

The individual similarities will be discussed below. Unfortunately, we often cannot

maintain all three similarities. Reliable approximations have to be introduced which
provide corrections for the condition of partial similarity in the model test.

28.1.1 Geometric similarity

The necessity for geometric similarity is the most obvious. If you want to know the Geometric
resistance of a sphere, you most likely will not get usable results by testing a box. We similarity
require that our ship is converted into a model by applying a fixed scale factor 𝜆 to all
linear dimensions. Any length 𝐿𝑆 at the ship will be 𝜆 times the corresponding length Scale of length,
on the model 𝐿𝑀 or area and volume

𝐿𝑆
𝜆= scaling of lengths (.)
𝐿𝑀

Areas are of dimension (length unit) × (length unit) and volumes are of dimension
(length unit)3 . Consequently, areas and volumes will scale according to

𝑆𝑆
𝜆2 = scaling of areas (.)
𝑆𝑀
𝑉
𝜆3 = 𝑆 scaling of volumes (.)
j 𝑉𝑀 j
As a result of the constant scale in length, angles are the same for full scale ship and
model.
The recommended ITTC procedure on model ships (ITTC, ) states that the model
should be accurate within ±1 mm in linear dimensions for all surfaces that get in contact
with the water. The surfaces should be smoothly sanded with  to  grit sand
paper. Small appendages, like struts or bilge keels, may not be reproduced in many
cases, because members may become too thin at model scale. Details of the surface
like weld seems or rivets are not reproduced at model scale.
In the context of geometric similarity, it is easily forgotten that we should scale the Tank dimensions
ocean environment as well. Due to space limitations, however, depth and especially
width of our model oceans are limited. A careful assessment has to be performed to
determine whether restricted water depth and tank width influence the model test
results or not. If you are a stickler for detail, water particles should be scaled too, but
let us file this under impossible.

28.1.2 Kinematic similarity

Kinematic similarity in fluid mechanics is often referred to as the similarity of streamline Time scale
patterns. Fluid particles should not only take geometric similar paths, but, in addition,
should maintain a constant time scale 𝜏 at full scale and model scale.
Δ𝑇𝑆
𝜏 = (.)
Δ𝑇𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

334 28 Laws of Similitude

Figure 28.1 Simplified flow pattern at a propeller

blade section. Kinematic similarity requires that
the angle of attack remains the same for full scale
and model propeller blade section

As a consequence, ratios of velocities are constant between full and model scale.
Similarity of A section of a rotating propeller blade may serve as an example (Figure .). In order
velocities to have the same flow pattern, the angle of attack 𝛼 between blade section and inflow
must be equal in full scale and model scale. This is achieved if the ratio of axial flow
velocity (speed of advance 𝑣𝐴 ) and tangential velocity (transverse speed due to propeller
rotation, 1∕2𝜋 𝑛 𝐷 for blade tip) is constant. For convenience we omit the factor 𝜋.
( ) ( )
𝑣𝐴𝑀 𝑣𝐴𝑆
= = 𝐽 (.)
𝑛𝑀 𝐷𝑀 𝑛𝑆 𝐷𝑆

j The fraction 𝑣𝐴 ∕(𝑛 𝐷) is called advance coefficient 𝐽 . Obviously, kinematic similarity j

cannot be achieved without geometric similarity.
Scale of Together the scales for geometry 𝜆 and time 𝜏 define the scales of velocities and accel-
velocities and erations.
acceleration
𝐿𝑆 𝐿𝑆
𝑣𝑆 Δ𝑇𝑆 𝐿𝑀 𝜆
velocity scale = = = (.)
𝑣𝑀 𝐿𝑀 Δ𝑇𝑆 𝜏
Δ𝑇𝑀 Δ𝑇𝑀
𝐿𝑆 𝐿𝑆
𝑎𝑆 Δ𝑇𝑆2 𝐿𝑀 𝜆
acceleration scale = = = (.)
𝑎𝑀 𝐿𝑀 Δ𝑇𝑆2 𝜏2
2
Δ𝑇𝑀 Δ𝑇 2 𝑀

28.1.3 Dynamic similarity

Types of forces Dynamic similarity requires that forces in full and model scale have the same direction
and the ratio 𝜅 of magnitudes full scale to model scale is constant. According to the
Navier-Stokes equations for incompressible flow, four classes of forces act on our ship:

. inertia force

. gravity force (volume force)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.1 Similarities 335

. friction force (surface force, shear stress)

. pressure force (surface force, normal stress)

Below we determine the force ratios and the requirements for dynamic similarity.

Inertia forces

Inertia forces may be expressed as mass times acceleration: 𝐹𝑖 = 𝑚 𝑎. Mass is usually Inertia forces
replaced by volume and density 𝑚 = 𝜌 𝑉 . As a first part of dynamic similarity, the ratio
of inertia forces must be constant.
𝐹𝑖𝑆 𝜌𝑆 𝑉𝑆 𝑎𝑆 𝜌 𝜆 𝜌 𝜆4
𝜅𝑖 = = = 𝑆 𝜆3 = 𝑆 (.)
𝐹𝑖𝑀 𝜌𝑀 𝑉𝑀 𝑎𝑀 𝜌𝑀 𝜏2 𝜌𝑀 𝜏 2
Whatever geometric scale and time scale have been chosen, if only inertia forces would
act on a system, they would scale by the ratio of the fluid densities times 𝜆4 ∕𝜏 2 . Of
course, whenever there are inertia forces, at least one other class of force will act too.
Otherwise no accelerations could occur according to Newton’s second law.

Inertia and gravity forces

The ratio of gravity forces 𝐹𝑔 = 𝑚 𝑔 would be Gravity forces

j j
𝐹𝑔𝑆 𝜌𝑆 𝑉𝑆 𝑔𝑆 𝜌 𝑔
𝜅𝑔 = = = 𝑆 𝑆 𝜆3 (.)
𝐹𝑔𝑀 𝜌𝑀 𝑉𝑀 𝑔𝑀 𝜌𝑀 𝑔𝑀
In most practical cases, gravitational acceleration at full and model scale will be the
same: 𝑔 = 𝑔𝑆 = 𝑔𝑀 . However, if you some day conduct experiments on the moon …
If both inertia and gravity forces act, dynamic similarity requires that the force ratios 𝜅𝑖 Inertia and
and 𝜅𝑔 are equal. There can be only one force ratio. If 𝜅𝑖 = 𝜅𝑔 then gravity forces

𝜌 𝑆 𝜆4 𝜌 𝑔
= 𝑆 𝑆 𝜆3
𝜌𝑀 𝜏 2 𝜌𝑀 𝑔𝑀
or
𝑔𝑀
𝜏2 = 𝜆. (.)
𝑔𝑆

If 𝑔 = 𝑔𝑆 = 𝑔𝑀 , time scale must be equal to the square root of the geometric scale
√
𝜏 = 𝜆 (.)
Enforcing dynamic similarity of both inertia and gravity forces makes the time scale
dependent on the geometric scale.
The scale of velocities (.) changes to
√
𝑣𝑆 𝜆 𝜆 √ 𝐿𝑆
= = √ = 𝜆 = (.)
𝑣𝑀 𝜏 𝜆 𝐿𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

336 28 Laws of Similitude

Considering the first and last terms in Equation (.), we find the relation of corre-
sponding velocities
𝑣𝑆 𝑣
√ = √𝑀 (.)
𝐿𝑆 𝐿𝑀
Froude number This relation was published by the French marine engineer Ferdinand Reech. However,
Froude () was the first to put it to practical use in ship model testing. If we include
the ratio of gravitational acceleration in Equation (.), we get
√
𝑣𝑆 𝜆 𝜆 𝑔𝑆 𝐿𝑆
= = √ = (.)
𝑣𝑀 𝜏 𝑔𝑀 𝑔𝑀 𝐿𝑀
𝜆
𝑔𝑆

or again combining the first and last terms

𝑣𝑆 𝑣
√ = √ 𝑀 (.)
𝑔𝑆 𝐿𝑆 𝑔𝑀 𝐿𝑀

This is obviously the definition of the Froude number 𝐹 𝑟. Consequently, conducting

model tests where the Froude number is equal for full scale ship and model ensures
dynamic similarity of inertia and gravity forces (Froude similarity).
In the case of Froude similarity, the velocity at model scale is obtained by dividing full
scale velocity by the square root of the geometric scale (assuming equal 𝑔 = 𝑔𝑆 = 𝑔𝑀 ).
j √ j
𝑣
𝑣𝑀 = √𝑆 and 𝑣𝑆 = 𝑣𝑀 𝜆 (.)
𝜆
This velocity ratio is very practical. For example, with a typical geometric scale of
𝜆 = 25, a full scale ship speed of 10 m/s (almost  kn) translates into a model speed of
 m/s. In cases where only inertia and gravity forces are relevant, full scale forces 𝐹𝑆 are
obtained from model scale forces 𝐹𝑀 by the following simple conversion:
𝜌𝑆 𝑔𝑆 3
𝐹𝑆 = 𝜆 𝐹𝑀 for purely inertia and gravity forces (.)
𝜌𝑀 𝑔𝑀

Inertia and friction forces

Friction forces In real fluids a more or less noticeable friction force exists. Friction forces can be
𝜕𝑢
modeled as shear stress times wetted surface: 𝐹𝜈 = 𝜇 𝜕𝑦 𝑆. Here, 𝜇 is the dynamic
𝜕𝑢
viscosity of the fluid and 𝜕𝑦 is the change in velocity perpendicular to the flow direction.
Thus, the friction forces for ship and model form the ratio
𝜕𝑢𝑆 𝜆
𝜇𝑆 𝑆
𝐹 𝜕𝑦𝑆 𝑆 𝜇𝑆 𝜏 2 𝜇 𝜆2
𝜅𝜈 = 𝜈𝑆 = = 𝜆 = 𝑆 (.)
𝐹𝜈𝑀 𝜕𝑢𝑀 𝜇𝑀 𝜆 𝜇𝑀 𝜏
𝜇𝑀 𝑆
𝜕𝑦𝑀 𝑀

Inertia and Assuming inertia and friction forces are present, their respective force ratios must be
friction forces

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.1 Similarities 337

equal to maintain dynamic similarity. For 𝜅𝑖 = 𝜅𝜈 we get

𝜌𝑆 𝜆4 𝜇 𝜆2
= 𝑆 (.)
𝜌𝑀 𝜏 2 𝜇𝑀 𝜏
or by introducing the kinematic viscosity 𝜈 = 𝜇∕𝜌
𝜇𝑆
𝜆2 𝜇𝑆 𝜌𝑀 𝜌 𝜈
= = 𝜇𝑆 = 𝑆
𝜏 𝜌𝑆 𝜇𝑀 𝑀 𝜈𝑀
𝜌𝑀
Thus the necessary time scale is equal to
𝜈𝑀 2
𝜏 = 𝜆 (.)
𝜈𝑆

In contrast to the similarity of combined inertia and gravity forces, the time scale is
now proportional to the squared geometric scale.
𝑣𝑆 𝜆 𝜆 𝜈 1 𝜈 𝐿
= = 𝜈 = 𝑆 = 𝑆 𝑀 (.)
𝑣𝑀 𝜏 𝑀 2
𝜆 𝜈𝑀 𝜆 𝜈𝑀 𝐿𝑆
𝜈𝑆
This yields
𝑣𝑀 𝐿𝑀 𝑣 𝐿
j = 𝑆 𝑆 (.) j
𝜈𝑀 𝜈𝑆
The Reynolds number 𝑅𝑒 = 𝑣𝐿∕𝜈 governs the dynamic similarity of inertia and friction
forces. Equal Reynolds number results in a very impractical velocity ratio. If model
tests are conducted in the same fluid, i.e. 𝜈𝑆 = 𝜈𝑀 , model speed is equal to full scale
speed times geometric scale
𝑣𝑀 = 𝜆 𝑣𝑆 (.)
For a geometric scale of 𝜆 = 25 and a ship speed of 10 m/s, a model speed of 𝑣𝑀 =
250 m/s would be necessary to maintain dynamic similarity of inertia and friction forces.
This is in the range of the cruising speed of a commercial airliner and very unlikely to
be realized in a towing tank!

Inertia, gravity, and friction forces

Our ship resistance problem includes inertia, gravity, and friction forces. In order to Inertia,
maintain dynamic similarity, the scales of all three classes of forces have to be equal: gravitational and
friction forces
𝜅𝑖 = 𝜅𝑔 = 𝜅𝜈 (.)

The second part yields with 𝜅𝑔 = 𝜅𝜈

𝜌𝑆 𝑔𝑆 3 𝜇 𝜆2
𝜆 = 𝑆
𝜌 𝑀 𝑔𝑀 𝜇𝑀 𝜏
𝑔𝑆 𝜈 1
𝜆 = 𝑆 (.)
𝑔𝑀 𝜈𝑀 𝜏

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

338 28 Laws of Similitude

We enforce similarity of inertia and gravitational forces by selecting the time scale
proportional to the square root of the geometric scale –√see Equation (.). This also
leads to more practical model speeds. Substituting 𝜏 = 𝜆 𝑔𝑀 ∕𝑔𝑆 in (.) yields the
condition
√ 𝜈 𝑔
𝜆 𝜆 = 𝑆 𝑆 (.)
𝜈𝑀 𝑔𝑀
With 𝑔𝑆 = 𝑔𝑀 , kinematic viscosity of the model fluid must be
𝜈𝑆
𝜈𝑀 = (.)
𝜆3∕2
in order to maintain similarity of inertia, gravity, and friction forces. For a model scale of
𝜆 = 25, kinematic viscosity 𝜈𝑀 in the model test would have to be as small as 0.008 𝜈𝑆 ,
i.e. eight thousandths of the kinematic viscosity of seawater.
A fluid with a significant smaller kinematic viscosity than water is mercury. Not that
anyone would seriously consider filling a towing tank with this extremely dangerous
substance, but let us play with some numbers. The ratio of kinematic viscosities of
𝜈 𝜈H O
water H2 O and mercury Hg is 𝑆 = 2 ≈ 8. The resultant model scale would be as
𝜈𝑀 𝜈Hg
√
small as 𝜆 = 8 or 𝜆 = 4. Unfortunately, the models would still be way too large for
3
common ship sizes and available test facilities.
Partial dynamic In summary, we will not be able to conduct our ship resistance model test with complete
j similarity dynamic similarity. We will have only partial dynamic similarity for inertia and gravity j
forces. All friction force components measured in the model test will have to be carefully
corrected for scale effects before they can be applied to the full scale ship. For this we
make use of the ITTC  model–ship correlation line, which allows as to compute
the frictional part of ship resistance.
0.075
𝐶𝐹 = [ ]2 (.)
log10 (𝑅𝑒) − 2
We apply this equation to the model as well as to the ship. Of course, the Reynolds
numbers will be quite different.

Inertia and pressure forces

Pressure forces Remaining are the pressure forces which can be modeled as pressure times area 𝐹𝑝 = 𝑝 𝑆.
The ratio of pressure forces is
𝑝𝑆 𝑆𝑆 𝑝
𝜅𝑝 = = 𝑆 𝜆2 (.)
𝑝𝑀 𝑆𝑀 𝑝𝑀
Inertia and Enforcing dynamic similarity for inertia and pressure forces requires 𝜅𝑖 = 𝜅𝑝 . Then
pressure forces
𝜌 𝑆 𝜆4 𝑝
= 𝑆 𝜆2
𝜌𝑀 𝜏 2 𝑝𝑀
𝑝𝑆 𝜌 ( 𝜆 )2
= 𝑆 (.)
𝑝𝑀 𝜌𝑀 𝜏

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.1 Similarities 339

Figure 28.2 Relationship of total static pressures

for model and full scale ship

The ratio 𝜆∕𝜏 is the the scale of velocities, i.e.

2
𝑝𝑆 𝜌 𝑣
= 𝑆 𝑆 (.)
𝑝𝑀 𝜌𝑀 𝑣2
𝑀

𝑝𝑆 𝑝𝑀
= (.)
1 1
𝜌 𝑣2 𝜌 𝑣2
2 𝑆 𝑆 2 𝑀 𝑀

j An Euler number or a dimensionless pressure coefficient 𝐶 𝑃 = 2𝑝∕(𝜌𝑣2 ) can be used j

to characterize dynamic similarity for systems subject to inertia and pressure forces.
We already concluded that we cannot simultaneously attain dynamic similarity for Inertia, gravity
inertia, gravity, and friction forces. What about the combination of inertia, gravity, and and pressure
forces
pressure√forces? Again we choose the time scale according to Froude’s law of similarity,
i.e. 𝜏 = 𝜆 𝑔𝑀 ∕𝑔𝑆 . The required ratio of pressures becomes
𝑝𝑆 𝜌
= 𝑆 𝜆 (.)
𝑝𝑀 𝜌𝑀

Is this possible? Figure . compares the total static pressure for a depth 𝑧𝑀 related to
a model and the corresponding depth 𝑧𝑆 of the full scale ship. Certainly 𝑧𝑆 ∕𝑧𝑀 = 𝜆.
In this case the ratio of total static pressures yields
𝑝𝑆 𝑝 − 𝜌𝑆 𝑔 𝑧𝑆 𝜌
= 𝐴 ≠ 𝑆 𝜆 (.)
𝑝𝑀 𝑝𝐴 − 𝜌𝑀 𝑔 𝑧𝑀 𝜌𝑀

Since the atmospheric pressure is equal for full scale ship and model, we do not have
the required dynamic similarity for the total pressures.
However, pressure forces are the result of pressure differences. The atmospheric pres-
sure acts equally on all parts of the ship surface above and below the waterline. Hence,
its resultant force vanishes. Comparing only the hydrostatic pressures for model and
full scale reveals that the necessary similarity for pressure differences is satisfied.
Δ𝑝𝑆 −𝜌𝑆 𝑔 𝑧𝑆 𝜌
= = 𝑆 𝜆 (.)
Δ𝑝𝑀 −𝜌𝑀 𝑔 𝑧𝑀 𝜌𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

340 28 Laws of Similitude

As a consequence, we are able to maintain simultaneous dynamic similarity for inertia,

gravity, and pressure forces.
Cavitation For some flow problems, similarity of absolute pressure is needed. Cavitation is one of
these flow phenomena. Cavitation occurs when the pressure in the fluid reaches the
vapor pressure level. Water changes from liquid into gaseous phase forming bubbles
with low pressure gas in the flow. The bubbles are transported with the flow and collapse
when they reach areas of higher pressure. This process can destroy propellers, rudders,
or any other material surface nearby.
The inception of cavitation depends on the absolute pressure level. It is usually higher
in the model than it is in the full scale ship, because we do not normally scale back
the atmospheric pressure (equal for both). Therefore models appear cavitation free
whereas the full scale vessel will show cavitation which can have significant effects
on performance. To investigate this phenomenon, special towing tanks or circulating
water tunnels are used that allow the atmospheric pressure above the water surface to
be changed.

28.1.4 Summary

Summary The discussion of the requirements for dynamic similarity discloses one of the greatest
difficulties of marine model testing: We are unable to conduct our model tests under
j complete dynamic similarity. Given the available conditions, we are not able to maintain j
similarity of friction forces. Elaborate procedures have been developed over the past
 years to account for this deficiency. Although practical solutions have been found,
you should always be aware of their limitations. Inertia, gravity, and pressure forces
are modeled correctly – within the limitations regarding geometric and kinematic
similarity – if the time scale is selected based on equal Froude numbers for full scale
ship and model 𝐹𝑟𝑀 = 𝐹𝑟𝑆 .

28.2 Partial Dynamic Similarity

As we have learned in the previous section, full dynamic similarity is not achievable
in practical ship model testing. We are unable to satisfy the requirements of equal
Froude number and equal Reynolds number at the same time. The lack of full dynamic
similarity has severe consequences on our model testing procedures. To emphasize the
problem we first look at the hypothetical case of full dynamic similarity.

28.2.1 Hypothetical case: full dynamic similarity

Hypothetical full Assuming the requirements of geometric similarity and kinematic similarity are satisfied,
dynamic a hypothetical full dynamic similarity is achieved when all force scales are equal:
similarity

𝜅𝑖 = 𝜅𝑔 = 𝜅𝑓 = 𝜅𝑝 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.2 Partial Dynamic Similarity 341

Putting Equation (.) into words, the ratios of full scale to model scale forces have
to be the same for all type of forces important to the problem at hand.

𝐹𝑖𝑆 𝐹𝑔𝑆 𝐹𝑓 𝑆 𝐹𝑝𝑆

= = = (.)
𝐹𝑖𝑀 𝐹𝑔𝑀 𝐹𝑓 𝑀 𝐹𝑝𝑀

Here, scales would have to be equal for inertia, gravity (representing body forces),
frictional (representing surface forces due to viscous stress), and pressure force (repre-
senting surface forces due to normal pressure).
Making full scale predictions from model test results would be easy, if we could have Hypothetical full
full dynamic similarity. Since all forces would scale the same way, the actual force scale scale predictions
could be derived from the physical force unit.
m 𝜌𝑆 3 𝜆
N = kg ⟶ 𝜅 = 𝜆 (.)
s2 𝜌𝑀 𝜏 2
Applying – for practical√reasons – the time scale that follows from similarity of inertia
and gravity forces 𝜏 = 𝜆 (follows from 𝐹𝑟𝑆 = 𝐹𝑟𝑀 ) yields:
𝜌𝑆 3
𝜅 = 𝜆 (.)
𝜌𝑀

Therefore, all model scale forces would scale up to full scale like volumes (𝜆3 ) multiplied
by the scale of densities.
j j
𝜌
𝜅𝑖 = 𝜅𝑔 = 𝜅𝑓 = 𝜅𝑝 = 𝑆 𝜆3 hypothetical (.)
𝜌𝑀

Assume that we are testing a model of geometric scale 𝜆 = 40 in fresh water of density Hypothetical full
𝜌𝑀 = 1000.0 kg/m3 under the condition that 𝐹𝑟𝑀 = 𝐹𝑟𝑆 . In the hypothetical case of dynamic
similarity
full dynamics similarity, a drag force of 𝐹𝐷𝑀 = 25.21 N measured at model scale would example
be equivalent to a full scale drag force of
𝐹𝐷𝑆 𝜌
= 𝑆 𝜆3
𝐹𝐷𝑀 𝜌𝑀
𝜌
𝐹𝐷𝑆 = 𝑆 𝜆3 ⋅ 𝐹𝐷𝑀
𝜌𝑀
1026.0 3
𝐹𝐷𝑆 = 40 ⋅ 25.21 N
1000.0
= 1 655 389.44 N = 1655.4 kN

in salt water of density 𝜌𝑆 = 1026.0 kg/m3 .

It is common practice to make experimental results dimensionless. The selection of Dimensionless
an appropriate normalizing force is up to the user. In ship hydrodynamics forces are coefficients
mostly normalized by dynamic pressure 1∕2𝜌𝑣2 multiplied with the wetted surface at
rest 𝑆. The result is a dimensionless drag coefficient:
𝐹𝐷𝑀
𝐶𝐷 = (.)
1
𝜌 𝑣2 𝑆
2 𝑀 𝑀 𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

342 28 Laws of Similitude

The coefficient would be the same for model and full scale vessel because with
1 1
𝐹𝐷𝑀 = 𝐶𝐷 𝜌𝑀 𝑣2𝑀 𝑆𝑀 and 𝐹𝐷𝑆 = 𝐶𝐷 𝜌𝑆 𝑣2𝑆 𝑆𝑆 (.)
2 2
we get

𝐹𝐷𝑆 𝐶𝐷 21 𝜌𝑆 𝑣2𝑆 𝑆𝑆
=
𝐹𝐷𝑀 𝐶𝐷 12 𝜌𝑀 𝑣2𝑀 𝑆𝑀
2
𝜌𝑆 𝑣𝑆 𝑆𝑆
=
𝜌𝑀 𝑣2 𝑆𝑀
𝑀

𝜌𝑆 𝜆4
= (.)
𝜌𝑀 𝜏 2

Again, assuming we performed the model √ tests under the condition of equal Froude
number, the time scale is fixed to 𝜏 = 𝜆. Substituting the time scale 𝜏 into Equa-
tion (.) results in
𝐹𝐷𝑆 𝜌
= 𝑆 𝜆3 (.)
𝐹𝐷𝑀 𝜌𝑀
which is the force scale factor (.) we derived above.

j Consequently, dimensionless coefficients would apply to both model and full scale, if j
the quantities (forces) could satisfy the conditions of full dynamic similarity.

28.2.2 Real world: partial dynamic similarity

As we have pointed out in the previous section, we cannot achieve full dynamic similarity.
Conducting experiments under the requirement of equal Froude number achieves
dynamic similarity for inertia, gravity, and pressure forces, but not for friction forces.

𝜅 = 𝜅𝑖 = 𝜅𝑔 = 𝜅𝑝 (.)

𝜅 ≠ 𝜅𝑓 (.)

This complicates ship model testing tremendously. The dimensionless total calm water
resistance coefficient
𝑅𝑇𝑀
𝐶𝑇𝑀 = (.)
1
𝜌𝑀 𝑣2𝑀 𝑆𝑀
2
cannot be applied to the full scale vessel without further corrections.

𝐶𝑇 𝑆 ≠ 𝐶𝑇𝑀 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

28.2 Partial Dynamic Similarity 343

28.2.3 Froude’s hypothesis revisited

William Froude was the first to come up with a practical solution to the scaling problem Froude’s
(Froude, ). He postulated that the resistance has to be divided into two parts: hypothesis

𝑅𝑇 = 𝑅𝑅 + 𝑅𝐹 (.)

or in dimensionless form

𝐶𝑇 = 𝐶𝑅 + 𝐶𝐹 (.)

. The first part contains the forces with full dynamic similarity if the tests are Residuary
carried out for corresponding speeds which is equivalent to equal Froude number resistance
with 𝑔𝑆 = 𝑔𝑀 .
𝑣𝑆 𝑣
√ = √𝑀 (.)
𝐿𝑆 𝐿𝑀
This resistance part became known as the residuary resistance 𝑅𝑅 . Its dimen-
sionless coefficient applies to both model and full scale vessel:
𝑅𝑅𝑀
𝐶𝑅 = = 𝐶𝑅𝑀 = 𝐶𝑅𝑆 (.)
1
𝜌𝑀 𝑣2𝑀 𝑆𝑀
2
j j
. The second part is the frictional resistance 𝑅𝐹 which is a function of the Reynolds Frictional
number. Since Reynolds numbers for ship and model are different 𝑅𝑒𝑆 ≠ 𝑅𝑒𝑀 , resistance
viscous forces are not dynamically similar. The frictional resistance coefficient
will differ for model and full scale vessel. Consequently, a method is needed to
scale the model test results.

At model scale the total resistance 𝑅𝑇𝑀 is measured and used to find the residuary
resistance coefficient 𝐶𝑅 which applies to both the model and the full scale vessel:

𝐶𝑅 = 𝐶𝑇𝑀 − 𝐶𝐹𝑀 (.)

The full scale resistance coefficient is derived by summing up the residuary resistance,
the frictional resistance (full scale), and some more or less empirical corrections:

𝐶𝑇 𝑆 = 𝐶𝑅 + 𝐶𝐹𝑆 + corrections (.)

For this procedure to work, we need a method to derive frictional resistance coefficients
𝐶𝐹 for the model and the full scale vessel. They cannot be directly measured with the
ship model. Froude determined a curve for 𝐶𝐹 experimentally by testing flat plates
of different sizes and roughness. Nowadays we employ the ITTC  model–ship
correlation line (.).
Froude knew that his hypothesis was flawed because the residuary resistance still Improving
contains the viscous pressure resistance 𝑅𝑉𝑃 . Since 𝑅𝑉𝑃 is influenced by the viscosity dynamic
similarity
of the fluid it depends at least somewhat on the Reynolds number. Therefore, the
residuary resistance 𝐶𝑅 coefficient does not reflect full dynamic similarity.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

344 28 Laws of Similitude

The ITTC introduced the form factor 𝑘 to eliminate most of the viscous effects from
the force determined under Froude similarity. The dynamically similar force is the wave
resistance.
𝐶𝑊 = 𝐶𝑇𝑀 − (1 + 𝑘)𝐶𝐹𝑀 (.)
This is better in theory but still hard to accomplish in practice, because accurate deter-
mination of the form factor is difficult. We will study the procedures recommended by
the ITTC in the following chapter.

References
Froude, W. (). Observations and suggestions on the subject of determining by
experiment the resistance of ships. Correspondence with the British Admiralty.
Published in The Papers of William Froude, A.D. Duckworth, The Institution of Naval
Architects, London, United Kingdom, pp. –, .
Froude, W. (). Experiments on surface-friction experienced by a plane moving
through water. Read before the British Association for the Advancement of Sci-
ence at Brighton. Published in The Papers of William Froude, A.D. Duckworth, The
Institution of Naval Architects, London, United Kingdom, pp. –, .
ITTC (). Ship models. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .

j j

Self Study Problems

. Provide definitions of Froude and Reynolds number (according to the ITTC).
. What is the greatest difficulty in marine model testing?
. Under the assumption of equal Froude number for model and full scale vessel, how
do the following quantities scale if the scale of the model is 𝜆 = 𝐿𝑆 ∕𝐿𝑀 = 50?

(a) velocity 𝑣𝑆 ∕𝑣𝑀 =?

(b) acceleration 𝑎𝑆 ∕𝑎𝑀 =?
(c) force 𝐹𝑆 ∕𝐹𝑀 =?
(d) torque 𝑄𝑆 ∕𝑄𝑀 =?
(e) wave period 𝑇𝑆 ∕𝑇𝑀 =?
(f ) wave frequency 𝜔𝑆 ∕𝜔𝑀 =?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

345

29
Resistance Test

The resistance test is the first in a series of experiments which together provide the
basis for a performance prediction of the full scale hull–propulsor system. This chapter
summarizes the procedures for measuring the resistance of a scale model. It discusses
test conditions, arrangements, data acquisition, and processing of the test results.

Learning Objectives
At the end of this chapter students will be able to:

j • understand the purpose of resistance tests j

• prepare and set up resistance tests

• process and interpret test results

• calculate form factor and residuary resistance coefficient

29.1 Test Procedure

A resistance test serves two purposes: Purpose

• determine the wave resistance coefficient 𝐶𝑊 as a function of Froude number,

and

• ascertain a form factor 𝑘.

Resistance tests are routinely performed to confirm the estimates of early ship design.
However, strictly speaking, performance predictions could be done without a resistance
test. Results of a load variation test may be used to substitute for the resistance test
data. This will be discussed later in Section ..
The resistance test is part of the ITTC  performance prediction method. The ITTC 1978
procedure has been revised several times since  and the reader is advised to consult performance
prediction
the latest issues of the ITTC recommended procedures. The description here is based method
on the documents issued in  (ITTC, , a,c,d).

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

346 29 Resistance Test

Figure 29.1 Ship model set up for resistance test

Model setup Figure . shows a typical model setup for a resistance test. After water temperature
and density have been measured, the model is ballasted to the desired volumetric
displacement (see Section .).
j j
Nowadays, resistance is measured with a load cell attached to the lower end of a stiff
column which itself is fastened to the towing carriage. The towing force is transferred
from the load cell to the model through a yoke. The yoke pivots in the vertical plane
and allows the model to heave and pitch freely. The towing point is located at the
intersection of the vertical line through the longitudinal center of buoyancy of the
model and the extended line of the propeller shaft. That way, the towing force will not
unduly change the attitude of the model. In some cases, shaft inclination might be too
steep and alternative approaches are needed, e.g. use a line through the center of the
thrust bearing as the height of the towing force attachment point.
The pressure distribution over the hull of a ship underway changes compared to its
hydrostatic equilibrium. A large area of lower pressure over the ship bottom causes
most ships to sink deeper into the water. They also might trim slightly aft or forward
depending on speed and hull shape. Linear transducers or heave potentiometers are
used to measure sinkage close to the fore and aft perpendiculars. The two sinkage
measurements enable us to compute mean sinkage and running trim angle.
Guides at the forward and aft ends of the model keep the model in line with the center
line of the tank. It is important that the guides do not impose any force in longitudinal
direction (resistance) or vertical direction (sinkage, trim). They restrict only sway and
yaw motion of the vessel.
Data flow Figure . shows the principal data flow. During the tests environmental data related
to the tank and the model are directly measured. See ITTC (c) for details.

• Tank water temperature 𝑡𝑊 is measured at least at the beginning and the end of
a series of tests. Water temperature is taken at half the draft of the model with

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

29.1 Test Procedure 347

Figure 29.2 Measured and derived data in the resistance test

at least . ◦ C of accuracy. In a long tank, temperature should be measured at

j several locations. j
The temperature is needed to determine water density 𝜌 and kinematic viscosity
𝜈. See the tables in the ITTC Recommended Procedures (ITTC, ). If a
hydrometer is available, the density should be measured directly.

• The carriage speed 𝑣𝐶 is measured to within .% of the maximum speed or

 mm/s whichever is larger. However, what we really need for our analysis is the
model speed 𝑣𝑀 through the water. Unfortunately, 𝑣𝑀 is impractical to measure.
Since we conduct our test in confined channels, influences of the tank walls and
the tank bottom may have to be considered. Details will be discussed in the
paragraphs on blockage correction below.

• Resistance 𝑅𝑇𝑀 is measured as a horizontal towing force in line with the propeller
shaft and at the longitudinal center of buoyancy location 𝐿𝐶𝐵. As mentioned, this
arrangement avoids trim effects. The tow force should be measured within .%
of the maximum capacity of the dynamometer (load cell) or . N, whichever is
larger (ITTC, c).

• Changes in vessel attitude are recorded as linear changes in vertical position 𝑧1

and 𝑧2 toward the aft and fore ends of the vessel. Mean dynamic sinkage 𝑧𝑉𝑀 at
𝐿𝑃𝑃 ∕2 and trim angle 𝑡𝑉 are calculated from 𝑧1 and 𝑧2 (see page ).

Tests are conducted for several towing velocities covering the whole range of Froude Maintaining
numbers of interest for the full scale vessel. It is important to keep a relatively constant preturbulence
level of preturbulence in the towing tank water throughout the tests. After long periods
of inactivity in the tank, water has settled down and its naturally appearing turbulence

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

348 29 Resistance Test

Figure 29.3 Tank cross section area 𝐴 and blockage factor 𝑚

has diminished. The level of preturbulence is affecting the laminar–turbulent transition

in the boundary layer. Resistance from repeated tests at the same speed will be in
disagreement, if the levels of preturbulence have changed significantly. Therefore, it is
custom to perform a higher speed run (without data recording) or drag a special fork
through the water to attain preturbulence before the actual test sequence commences.
Afterward, carriage speeds should not be simply increased from test to test, but alternate
j between high and low speeds. j
Steady state data In each run the model has to be accelerated from zero to the actual speed. Due to the
transient acceleration phase, certain fluctuations of the resistance can be expected even
once a steady speed has been reached. The resistance 𝑅𝑇𝑀 should be taken as the mean
value of at least  cycles. The time window for averaging should consist of an integer
number of full cycles.

29.2 Reduction of Resistance Test Data

Test results As a result of the tests, we obtain a table with total resistance 𝑅𝑇𝑀 (𝑣𝐶 ), sinkage mea-
surements 𝑧1 (𝑣𝐶 ), 𝑧2 (𝑣𝐶 ) as a function of carriage speed 𝑣𝐶 . In a first step, the actual
model speed through the water has to be determined from carriage speed and tank
dimensions. Once the model speed 𝑣𝑀 is known, form factor and residuary resistance
coefficient may be determined.
Blockage Obviously, a model moves at a speed relative to the towing tank walls which matches
correction the carriage speed 𝑣𝐶 . However, for the performance prediction the speed of a model
relative to the surrounding fluid is the determining factor.
As Figure . illustrates, the cross section of the tank 𝐴 = 𝑏ℎ narrows at the model
slightly to 𝐴−𝐴𝑥 . 𝐴𝑥 is the maximum submerged sectional area of the model. Compara-
ble to a flow in a narrowing pipe, the water will speed up around the model. You can also
imagine the model pushing the water, which creates an indiscernible rise of the water
ahead of the model, and thus causing a flow from bow to stern which is superimposed
to the model movement. As a result of blockage, the relative velocity between model

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

29.2 Reduction of Resistance Test Data 349

and the surrounding body of water is commonly higher than the carriage speed by a
small fraction. How much faster mainly depends on the blockage factor 𝑚 = 𝐴𝑥 ∕𝐴 and
√
the depth Froude number 𝐹𝑟ℎ = 𝑣∕ 𝑔ℎ.
Most model basins have developed their own blockage correction methods considering Blockage
their tank dimensions and the size of models they typically investigate. Blockage correction and
model speed
corrections are expressed as relative increase of model speed over carriage speed Δ𝑣∕𝑣.
For typical towing tank cross sections and model sizes, the increase in velocity is
typically less than %. The carriage velocity 𝑣𝐶 is augmented by the fractional increase
Δ𝑣∕𝑣 to obtain the actual model speed 𝑣𝑀 relative to the surrounding water.
( )
Δ𝑣
𝑣𝑀 = 𝑣𝐶 1 + (.)
𝑣
The former model basin Versuchsanstalt für Wasserbau und Schiffbau (VWS) in Berlin
sometimes simply used Δ𝑣∕𝑣 = 𝑚 which accounts for conservation of mass effects only
(Schuster, ).
ITTC suggests three different blockage correction methods, two of which are fairly Schuster’s
simple to implement (ITTC, c). Schuster () suggests to compute the blockage blockage
correction
correction as a function of the blockage factor 𝑚, the depth Froude number 𝐹𝑟ℎ , and
the ratio of viscous and total resistance.
( )
Δ𝑣 𝑚 2 𝑅𝑉𝑀
= + 1− 𝐹𝑟10
ℎ (.)
j
𝑣 1 − 𝑚 − 𝐹𝑟2ℎ 3 𝑅𝑇𝑀
j
The depth Froude number is defined as
𝑣
𝐹𝑟ℎ = √𝑀 (.)
𝑔ℎ

ℎ is the water depth in the towing tank at the time of the test. 𝑅𝑉𝑀 is the total viscous
resistance of the model.
This obviously involves a ‘chicken or egg’ dilemma. At this point, we only know the
carriage speed 𝑣𝐶 , but we should use the actual model speed 𝑣𝑀 to compute the depth
Froude number 𝐹𝑟ℎ (.) and to estimate the viscous model resistance 𝑅𝑉𝑀 . Due to
the small differences in speed, it will not matter much, if we initially estimate the depth
Froude number based on the carriage speed instead of the model speed.
𝑣
̃ ℎ = √𝐶
𝐹𝑟 (.)
𝑔ℎ

Using this approximation, a first estimate for the model velocity is derived utilizing the
first and dominant part of Schuster’s formula (.):
( )
𝑚
𝑣̃𝑀 = 𝑣𝐶 1+ (.)
̃2
1 − 𝑚 − 𝐹𝑟 ℎ

Equipped with this approximate model speed, the depth Froude number is updated

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

350 29 Resistance Test

and the viscous resistance of the model is estimated:

𝑣̃
Second depth Froude number estimate: 𝐹𝑟ℎ = √𝑀 (.)
𝑔ℎ
𝑣̃ 𝐿
Reynolds number estimate: ̃ = 𝑀 𝑂𝑆𝑀
𝑅𝑒 (.)
𝜈𝑀

0.075
Friction coefficient: 𝐶̃𝐹 𝑀 = [ ] (.)
̃ −2 2
log10 (𝑅𝑒)
1
Viscous resistance estimate: 𝑅𝑉𝑀 = 𝜌 𝑣̃2 𝑆 𝐶̃ (.)
2 𝑀 𝑀 𝑀 𝐹𝑀
The latter does not include a form factor as it is unknown at this point. However, since
the depth Froude number should satisfy 𝐹𝑟ℎ < 0.7, the factor for the correction is small
(𝐹𝑟10
ℎ
< 0.029). All quantities are now known, and Equations (.) and (.) may
be employed to compute the speed correction. This process works well for moderate
speeds with 𝐹𝑟 < 0.35.
Tamura’s Another simple blockage correction method has been proposed by Tamura (). It
blockage computes the fractional increase in speed from the following formula:
correction
[ ]3
Δ𝑣 𝐿𝑊𝐿 4 1
= 0.67 𝑚 ( ) (.)
j 𝑣 𝐵 1 − 𝐹𝑟2ℎ j

Again, model speed follows from Equation (.). Tamura’s blockage correction seems
to overpredict the speed for the relatively small models of about 3 m length used in the
towing tank at the University of New Orleans.
Total resistance With the actual model speed 𝑣𝑀 known, the measured resistance is converted into a
coefficient dimensionless total resistance coefficient:
𝑅𝑇𝑀
𝐶𝑇𝑀 = (.)
1
2
𝜌 𝑣2𝑀 𝑆𝑀

If we would have been able to conduct the test under full dynamic similarity, 𝐶𝑇𝑀
would also apply to the full scale vessel. Unfortunately, this is not the case.
Dynamic trim During the test run, sinkage measurements are taken near stern and bow of the vessel.
and sinkage Figure . shows the general setup. The lengths 𝐿1 and 𝐿2 mark the distances of the
measurement points from midships 𝐿𝑃𝑃 ∕2.
Trim angle The trim angle 𝑡𝑉 follows from the ratio of difference in sinkage fore and aft and the
distance between the measurement points.
( )
−1 −(𝑧2 − 𝑧1 )
𝑡𝑉 = tan (.)
𝐿1 + 𝐿2

If we use a right handed < 𝑥, 𝑦, 𝑧 >-coordinate system with 𝑥-axis pointing forward
and the 𝑧-axis pointing upwards, a positive trim angle 𝑡𝑉 > 0 will cause the bow to trim
down.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

29.3 Form Factor 𝑘 351

Figure 29.4 Measurements and length definitions for the computation of sinkage and trim
(sinkage and trim are exaggerated)

Mean sinkage at midships follows from basic geometrical relationships. Mean sinkage
( )
𝑧2 − 𝑧1
𝑧𝑉𝑀 = 𝑧1 + 𝐿1 (.)
𝐿2 + 𝐿1
Most vessels will have negative sinkage, i.e. they settle deeper into the water when un-
derway. This effect is even more pronounced in shallow water and has to be considered
when sailing fast in shallow waterways in order to prevent grounding.
j j

29.3 Form Factor 𝑘

The form factor 𝑘 has been introduced to augment the flat plate frictional resistance Form factor
coefficient 𝐶𝐹 to include the Reynolds number dependent viscous pressure resistance
and form effects on the boundary layer itself. Once a form factor is known, the wave
resistance coefficient 𝐶𝑊 is found by solving
( )
𝐶𝑇𝑀 = 1 + 𝑘 𝐶𝐹𝑀 + 𝐶𝑊 (.)

for 𝐶𝑊 :
( )
𝐶𝑊 = 𝐶𝑇𝑀 − 1 + 𝑘 𝐶𝐹𝑀 (.)
𝐶𝐹𝑀 is the ITTC  model–ship correlation line coefficient for the Reynolds number
𝑅𝑒𝑀 of the model test (.). Note that the Froude number is computed based on the
waterline length 𝐿𝑊𝐿 , but the Reynolds number is computed based on the length over
wetted surface 𝐿𝑂𝑆 .
𝑣 𝑣𝑀 𝐿𝑂𝑆𝑀
𝐹𝑟 = √ 𝑀 𝑅𝑒𝑀 = (.)
𝑔𝐿𝑊𝐿𝑀 𝜈𝑀

 The residuary resistance coefficient 𝐶 was used instead of 𝐶

𝑅 𝑊 in editions of the ITTC recommended
procedures and guidelines before .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

352 29 Resistance Test

Difficulties Form factors 𝑘 are not universally applied because the uncertainty in their values is
high. Among others, Gross () lists the following difficulties:

• low model speed means low Reynolds number. The boundary layer may be
laminar and the danger of separation increases (which will not happen at the
full scale vessel). Tripping the boundary layer into turbulent flow may be more
difficult. If separation occurs, the increased resistance will cause the form factor
to be too large. On the other hand, if the boundary flow becomes mostly laminar
without separation form factors may become too small. This problem will be
compounded for short models (large scale 𝜆).
• wave breaking may disturb the linearity of the data points used to find 𝑘. This is
especially true for vessels with high block coefficient.
• side wall effects may influence the form factor
• bulbous bow and transom affect the linearity of the Prohaska data points
• hull–propeller interaction may influence the form factor

If the form factor is set too high, resistance of the full scale vessel will be underpredicted,
because the wave resistance estimate will be too low and vice versa. ITTC is discussing
formulas for the form factor, but so far no formula has been found that consistently
improves the performance prediction (see Gross () and Section .). The most
j reliable method to determine the form factor was suggested by Prohaska () and is j
recommended by the ITTC.
Prohaska’s If no boundary layer separation occurs, the difference between total and viscous
method resistance is the wave resistance. Linear wave resistance theory predicts that for slow
speeds 𝐹𝑟 < 0.2 wave resistance is growing with the fourth power of the Froude number
with a constant factor 𝛼.

𝐶𝑊 ≈ 𝛼𝐹𝑟4 for 𝐹𝑟 < 0.2 (.)

This is confirmed by experimental results for Froude numbers 𝐹𝑟 < 0.2. We substitute
assumption (.) into Equation (.) and divide by the ITTC  model–ship
correlation coefficient 𝐶𝐹𝑀 .
𝐶𝑇𝑀 𝐶 𝐹𝑟4
= (1 + 𝑘) 𝐹𝑀 + 𝛼 (.)
𝐶𝐹𝑀 𝐶𝐹𝑀 𝐶𝐹𝑀
Equation (.) may be interpreted as the equation of a straight line, where 𝑏 = (1 + 𝑘)
𝐹𝑟4
becomes the 𝑦-intercept, 𝛼 is the slope, and 𝑥 = 𝐶𝐹𝑀
is the variable.

𝑦 = 𝑏 + 𝛼𝑥 (.)
( 4 )
𝐶
We plot left- and right-hand side of (.) as data points (𝑥𝑖 , 𝑦𝑖 ) = 𝐶𝐹𝑟 , 𝐶𝑇𝑀 .
𝐹𝑀 𝐹𝑀
According to Equations (.) and (.), the data points should align in a straight
line for model speeds with 𝐹𝑟 < 0.2 (Figure .). ITTC recommends to use only data
points in the range of 𝐹𝑟 ∈ [0.1, 0.2]. If the speed is too low (𝐹𝑟 < 0.1), errors in the
resistance measurements may lead to incorrect form factor predictions.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

29.3 Form Factor 𝑘 353

Figure 29.5 Method of Prohaska to determine the form factor 𝑘

Holtrop ()( suggests to carefully)select the data used to compute the form factor.
Data points 𝐶𝑇𝑀 ∕𝐶𝐹𝑀 , 𝐹 𝑟4 ∕𝐶𝐹𝑀 which drop sharply for small 𝐹 𝑟4 ∕𝐶𝐹𝑀 are an
j indication of possible laminar flow effects. If the Reynolds number becomes too small, j
the flow may become laminar ( again downstream of)the turbulence generator (relam-
inarization). If data points 𝐶𝑇𝑀 ∕𝐶𝐹𝑀 , 𝐹 𝑟4 ∕𝐶𝐹𝑀 rise sharply for small 𝐹 𝑟4 ∕𝐶𝐹𝑀
values, it might be an indication of flow separation. However, rising values may also be
caused by a transom that is too deeply submerged for low speeds and, consequently,
creates significant pressure drag. The same is true for a partially submerged bulbous
bow. Careful observation of flow and wave patterns during the tests is required to
choose a suitable data set.
The form factor can be determined graphically from the 𝑦-intercept or better by linear Linear regression
regression. Depending on how slow the model could be towed without flow separation,
data may have to be extrapolated to 𝐹𝑟 ⟶ 0 quite a bit. First, compute the mean
values 𝑥̄ and 𝑦̄ of the used 𝑥𝑖 and 𝑦𝑖 values (𝑀 pairs).

1 ∑ 1 ∑ 𝐹𝑟𝑖
𝑀 𝑀 4
𝑥̄ = 𝑥𝑖 = (.)
𝑀 𝑖=1 𝑀 𝑖=1 𝐶𝐹𝑀𝑖

1 ∑ 1 ∑ 𝐶𝑇𝑀𝑖
𝑀 𝑀
𝑦̄ = 𝑦𝑖 = (.)
𝑀 𝑖=1 𝑀 𝑖=1 𝐶𝐹𝑀𝑖

The slope 𝛼 is given by

∑
𝑀
(𝑥𝑖 − 𝑥)(𝑦
̄ 𝑖 − 𝑦)
̄
𝑖=1
𝛼 = (.)
∑
𝑀
̄ 2
(𝑥𝑖 − 𝑥)
𝑖=1

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

354 29 Resistance Test

The 𝑦-intercept follows from

𝑏 = 𝑦̄ − 𝛼 𝑥̄ (.)
and the form factor 𝑘 is
𝑘 = 𝑏−1 (.)

In some cases, the assumption 𝐶𝑊 = 𝛼𝐹𝑟4 does not work out. If the data points do not
show a linear relationship, higher powers than 4, i.e. 5 or 6, may be tried.
Full scale form The form factor 𝑘 found for the model is also used for the full scale vessel. After all,
factor 𝑘 expresses a form effect, and the shapes of model and full scale vessel ought to be
identical. The boundary layers are, however, not similar. It is relatively thicker on the
model. There is some evidence that the assumption 𝑘𝑀 = 𝑘𝑆 is not entirely correct, but
so far no consistent estimate for the scale effects on the form factor has been developed.
Form factor values typically fall into the range 𝑘 ∈ [0.1, 0.4]. Slender vessels tend to
have smaller values than vessels with high block coefficient. The form factor will vary
for the same vessel from loading condition to loading condition.

29.4 Wave Resistance Coefficient 𝐶𝑊

Wave resistance Finally, with model speed 𝑣𝑀 and form factor 𝑘 known, we can extract the wave
coefficient resistance coefficient 𝐶𝑊 . If not yet done, compute the Reynolds number 𝑅𝑒𝑀 for all
j model velocities 𝑣𝑀 j

𝑣𝑀 𝐿𝑂𝑆𝑀
𝑅𝑒𝑀 = (.)
𝜈𝑀

which leads to the ITTC  model–ship correlation coefficient

0.075
𝐶𝐹𝑀 = [ ]2 (.)
log10 (𝑅𝑒𝑀 ) − 2

Finally, the wave resistance coefficient follows from Equation (.).

( )
𝐶𝑊 = 𝐶𝑇𝑀 − 1 + 𝑘 𝐶𝐹𝑀 (.)

Residuary In cases where a form factor cannot be found experimentally and the estimates do not
resistance apply, the residuary resistance coefficient 𝐶𝑅 replaces the wave resistance coefficient.
coefficient
𝐶𝑅 = 𝐶𝑇𝑀 − 𝐶𝐹𝑀 without form factor (.)

Full scale wave Based on Froude’s fundamental assumption that the residuary resistance depends only
resistance on the Froude number 𝐹𝑟, the wave resistance coefficient 𝐶𝑊 (or 𝐶𝑅 ) derived for the
coefficient
model is applied to the full scale vessel for the same Froude number and corresponding
speed.
𝐶𝑊 = 𝐶𝑊𝑀 = 𝐶𝑊𝑆 (.)
A numerical example is presented at the end of the following chapter.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

29.5 Skin Friction Correction Force 𝐹𝐷 355

29.5 Skin Friction Correction Force 𝐹𝐷

An additional result of the resistance test is the skin friction correction force 𝐹𝐷 (ITTC, Friction
b). For each run of the propulsion test, a skin friction correction 𝐹𝐷 is precomputed deduction force
and applied to the model as an additional towing force. It corrects the otherwise
excessive loading of the model propeller during the propulsion test.
[ ( ) ]
1 2
𝐹𝐷 = 𝜌 𝑣𝑀 𝑆𝑀 (1 + 𝑘) 𝐶𝐹𝑀 − 𝐶𝐹𝑆 − Δ𝐶𝐹 (.)
2
Omission of the skin friction correction Δ𝐶𝐹 results in incorrect values for the hull–
propeller interaction parameters (see Chapter ).
Of course, if the residuary resistance coefficient is determined with Froude’s method
(𝑘 = 0), the form factor is also omitted in the calculation of the skin friction deduction.
[( ) ]
1
𝐹𝐷 = 𝜌 𝑣2𝑀 𝑆𝑀 𝐶𝐹𝑀 − 𝐶𝐹𝑆 − Δ𝐶𝐹 (.)
2
An alternative to applying a specific friction deduction force is to do tests with load
variations on the propeller. We will get back to this when we discuss propulsion.

References
j j
Gross, A. (). Form factor. In Proceedings of the th ITTC, pages –, Ot-
tawa, Canada. International Towing Tank Conference. Report of the Performance
Committee, Appendix .
Holtrop, J. (). Extrapolation of propulsion tests for ships with appendages and
complex propulsors. Marine Technology, ():–.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
ITTC (a).  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
ITTC (b). Propulsion/bollard pull test. International Towing Tank Conference,
Recommended Procedures and Guidelines .---.. Revision .
ITTC (c). Resistance test. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .
ITTC (d). Ship models. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .
Prohaska, C. (). A simple method for the evaluation of the form factor and the low
speed wave resistance. In Proceedings of the th ITTC, pages –, Tokyo, Japan.
International Towing Tank Conference. Resistance Session, Written Contributions.
Schuster, S. (). Untersuchungen über Strömungs- und Widerstandsverhältnisse bei
der Fahrt von Schiffen in beschränktem Wasser. In Jahrbuch der Schiffbautechnischen
Gesellschaft, volume , pages –. Springer-Verlag, Berlin.
Schuster, S. (). Beitrag zur Frage der Kanalkorrektur bei Modellversuchen. Schiff-
stechnik/Ship Technology Research, :–.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

356 29 Resistance Test

Tamura, K. (). Study on the blockage correction. Journal of the Society of Naval
Architects of Japan, ():–.

Self Study Problems

. Summarize the data acquired during a resistance test and state the quantities
derived from the measured data.
. Discuss why models are ballasted based on weight rather than draft.

. Why is the speed of the model relative to the surrounding water higher than the
actual carriage speed?
. Explain why the attitude changes for a moving ship compared to its position at
rest.
. Why is the form factor determined for the model also used for the full scale
vessel?
. What is the major difference between the residuary resistance coefficient and the
wave resistance coefficient? What do they have in common?

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

357

30
Full Scale Resistance Prediction

This chapter continues the discussion of the resistance test. We show how the resistance
values found in the model test are used to predict the resistance of the full scale vessel.
The description follows the ITTC recommended procedure .--- for the resis-
tance test and procedure .---. for the  ITTC performance prediction
method (ITTC, a,b). Additions and corrections to the resistance coefficients of the
model test are explained and discussed. An example closes the chapter.

Learning Objectives

j At the end of this chapter students will be able to: j

• understand the issues associated with scaling model test results to full scale
• make full scale resistance predictions based on model test results
• understand and apply necessary corrections

30.1 Model Test Results

The data of resistance tests result in the form factor 𝑘 and a set of wave resistance Wave resistance
coefficients 𝐶𝑊 as a function of Froude number 𝐹𝑟. In the case a form factor is not and form factor
determined, a set of residuary resistance coefficients 𝐶𝑅 replaces the wave resistance
coefficients.
Wave resistance coefficient (or residuary resistance coefficient) and form factor are
used for the full scale vessel without corrections. This is based on the assertions that
model and ship are geometrically similar and that 𝐶𝑊 only considers forces that are
dynamically similar.
Resistance tests are conducted under Froude similarity, i.e. the full scale ship speed Ship speed
equivalent to the model speed is given by
√
√ 𝑔𝐿𝑊𝐿𝑀 √
𝑣𝑆 = 𝐹𝑟 𝑔𝐿𝑊𝐿𝑆 = 𝑣𝑀 √ = 𝑣𝑀 𝜆 (.)
𝑔𝐿𝑊𝐿𝑆

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

358 30 Full Scale Resistance Prediction

The last term uses the model scale 𝜆 and assumes equal gravitational acceleration for
model and full scale ship.
Total resistance The total resistance for the model has been measured and used to derive the total
coefficient of resistance coefficient 𝐶𝑇𝑀 . During the model test, only partial dynamic similarity is
model
possible. Therefore, the dimensionless total resistance coefficient of the model cannot
be applied to the full scale vessel. The total resistance of the model is split into viscous
and wave resistance 𝐶𝑇𝑀 = (1 + 𝑘)𝐶𝐹𝑀 + 𝐶𝑊 in order to overcome the problem of
partial dynamic similarity. Although the wave resistance coefficient 𝐶𝑊 applies to both,
model and full scale ship, simply replacing the model frictional resistance coefficient
𝐶𝐹𝑀 with the equivalent full scale coefficient 𝐶𝐹𝑆 in the formula is not quite sufficient.
Additional corrections are needed, which are explained in the next section.

30.2 Corrections and Additional Resistance Components

Roughness As stated in our discussion of the frictional resistance, the finished surfaces of models
allowance are hydraulically smooth. However, the full scale ship hull is not. Due to the high
Reynolds number and the resulting thin viscous sublayer, even smooth weld seams
and ever present surface imperfections call for an upward correction of the resistance
compared to the smooth model. Therefore, a roughness allowance Δ𝐶𝐹 is applied to
account for the relative increase of frictional resistance from model to full scale vessel.
ITTC (a) recommends the following correction for surface roughness.
j j
( )1
⎡ 𝑘 3 ( )− 1 ⎤
= 0.044 ⎢ 𝑠 3⎥
Δ𝐶𝐹 − 10 𝑅𝑒𝑆 + 0.000125 (.)
⎢ 𝐿𝑊𝐿𝑆 ⎥
⎣ ⎦

In this formula 𝑘𝑠 is the equivalent sand roughness of the hull and not the form factor!
The standard value for it is 𝑘𝑠 = 150 ⋅ 10−6 m. For advanced hull coatings available
today, lower values might be more appropriate.
Full scale 𝑅𝑒𝑆 is the Reynolds number for the full scale vessel.
Reynolds
number 𝑣𝑆 𝐿𝑂𝑆𝑆
𝑅𝑒𝑆 = (.)
𝜈𝑆

By convention, a kinematic viscosity of 𝜈𝑆 = 1.1892 10−6 m2 /s for seawater of 15 ◦ C

temperature is used for the full scale prediction.
Correlation The correlation allowance accounts for remaining differences between model and full
allowance scale vessel which may be attributed to the partial dynamic similarity of the model
test. Model basins usually develop their own formula by comparing their predictions
with data gathered during ship trials, but ITTC (a) suggests using the following
equation in combination with the roughness allowance (.) above.
[ ]
𝐶𝐴 = 5.68 − 0.6 log10 (𝑅𝑒𝑆 ) ⋅ 10−3 (.)

Surface The sum of D frictional resistance (1 + 𝑘)𝐶𝐹𝑆 , roughness allowance Δ𝐶𝐹 , and corre-
corrections, bilge lation allowance 𝐶𝐴 represent the full scale viscous resistance. It is mainly a function
keels

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

30.3 Total Resistance and Effective Power 359

of Reynolds number 𝑅𝑒𝑆 . Since the viscous resistance is proportional to the wetted
𝑆𝑆 +𝑆𝐵𝐾
surface area 𝑆, an additional correction factor 𝑆𝑆
𝑆
is applied to account for extra
hull surfaces not replicated at model scale. Here, 𝑆𝐵𝐾𝑆 is the wetted surface of the bilge
keels.
Ship models are usually built without deck house and other superstructure like for Air resistance
example cargo containers. Therefore, the full scale resistance prediction contains an
addition for the resistance of a ship moving through air.

𝜌𝐴 𝐴𝑉 𝑆
𝐶𝐴𝐴𝑆 = 𝐶𝐷𝐴 (.)
𝜌 𝑆 𝑆𝑆

For some vessels wind tunnel test data may be available, but in general, 𝐶𝐷𝐴 = 0.8
is the suggested default value for the air drag coefficient. 𝜌𝐴 = 1.225 kg/m3 is the
standard density of dry air at a temperature of 15 ◦ C. 𝐴𝑉 𝑆 is the projected area above
the waterline perpendicular to the direction of motion. It has to be emphasized that
Equation (.) captures only the resistance of a ship moving through air at rest (trial
conditions). Significantly higher forces due to steady wind and wind gusts can occur in
service conditions.
Additional resistance components may have to be considered as needed. Examples Additional
include the additional viscous resistance due to appendages (struts, bossings for twin corrections
propulsion systems, foils, etc.). They are collected in an appendage resistance coefficient
𝐶𝐴𝑃 𝑃𝑆 . Estimates of appendage resistances are provided by Holtrop and Mennen ().
j Additional corrections may be needed for spray resistance, steering resistance, or other j
flow phenomena not captured anywhere else.

30.3 Total Resistance and Effective Power

From the ship speed follows the full scale Reynolds number 𝑅𝑒𝑆 (.) and the corre- Frictional
sponding ITTC  model–ship correlation line coefficient 𝐶𝐹𝑆 for the ship. resistance

0.075
𝐶𝐹𝑆 = [ ]2 (.)
log10 (𝑅𝑒𝑆 ) − 2

The full scale resistance coefficient 𝐶𝑇 𝑆 is build on 𝐶𝐹𝑆 and 𝐶𝑊 including the additional Full scale total
corrections discussed above. resistance
coefficient
𝑆𝑆 + 𝑆𝐵𝐾𝑆 [ ]
𝐶𝑇 𝑆 = (1 + 𝑘) 𝐶𝐹𝑆 + Δ𝐶𝐹 + 𝐶𝐴 + 𝐶𝑊 + 𝐶𝐴𝐴𝑆 + 𝐶𝐴𝑃 𝑃𝑆 (.)
𝑆𝑆

The actual, total resistance of the ship without a propulsor in calm water is obtained Total resistance
by multiplying the total resistance coefficient 𝐶𝑇 𝑆 with dynamic pressure 𝑞 = 𝜌𝑣2𝑆 ∕2
and wetted surface 𝑆𝑆 .
1
𝑅𝑇𝑆 = 𝜌 𝑣2𝑆 𝑆𝑆 𝐶𝑇 𝑆 (.)
2
At steady state, work is equal to force multiplied by distance. The effective power Effective power

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

360 30 Full Scale Resistance Prediction

Table 30.1 Particulars of full scale vessel and model used in the prediction example

Main particular Unit Vessel Model

model number – [–] 0001
model scale 𝜆 – [–] 25
length in waterline 𝐿𝑊𝐿 [m] 172.440 6.8976
length over wetted surface 𝐿𝑂𝑆 [m] 178.040 7.1216
beam 𝐵 [m] 26.529 1.0612
draft 𝑇 [m] 10.204 0.4081
maximum sectional area 𝐴𝑋 [m2 ] 264.466 0.4231
transverse area above WL 𝐴𝑉 𝑆 [m2 ] 464.000 –
wetted surface 𝑆 [m2 ] 5 540.508 8.8648
bilge keel surface 𝑆𝐵𝐾 [m2 ] 85.000 –
displacement (volume) 𝑉 [m3 ] 27 285.788 1.7463
displacement (mass) Δ [t] 27 995.791 1.7444
distance midship–aft sinkage sensor 𝐿1 [m] – 2.400
distance midship–forward sinkage sensor 𝐿2 [m] [–] 2.700
standard sand roughness 𝑘𝑆 [𝜇m] 150.0 –

needed to tow the vessel is defined by the work done per unit time which is equivalent
to force times velocity.
j 𝑃𝐸𝑆 = 𝑅𝑇𝑆 𝑣𝑆 (.) j
Make sure you use consistent units! Especially, do not use the ship speed expressed in
knots with these equations. The whole procedure is best implemented as a spread sheet
or program.

30.4 Example Resistance Prediction

Summary of A resistance test has been conducted for a single screw container ship model at scale
vessel data 𝜆 = 25. Table . shows an excerpt of geometric particulars for vessel and model.
Each model should have a unique identifier within the institution performing the tests
(here the fictitious number ). Typically, a resistance test report will include a full
set of hydrostatic data and potentially a lines plan. ITTC (b) lists the minimum set
of data to be included. The model discussed here features a sand strip as a turbulence
generator and is towed at the 𝐿𝐶𝐵–propeller shaft line intersection point.
Displacement Note that the displacement mass Δ𝑀 of the model is not derived by scaling the full
scale mass. Only the volumetric displacement 𝑉 is scaled. Then, the mass at model
scale follows from Equation (.).
Test environment Important for the evaluation of any model test is the testing environment. Temperature
of the tank water 𝑡𝑊 determines density and kinematic viscosity of the water. Water
depth ℎ and tank cross section area 𝐴 influence the blockage correction. Full scale
resistance predictions are usually performed for standard seawater at temperature  ◦ C.
The relevant environmental data for this example are presented in Table .. Values
are taken from ITTC’s recommended procedures on fresh and seawater properties
(ITTC, ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

30.4 Example Resistance Prediction 361

Table 30.2 Testing and full scale environments for resistance prediction

gravitational acceleration at test site 𝑔 9.793167 m/s2

water depth in towing tank ℎ 4.450 m
cross section area of towing tank 𝐴 35.600 m2
blockage factor 𝑚 0.01189
tank water temperature 𝑡𝑊 16.300 ◦ C
tank water density 𝜌𝑀 998.8968 kg/m3
tank water kinematic viscosity 𝜈𝑀 1.1007 10−6 m2 /s
seawater temperature 𝑡𝑊 𝑆 15.000 ◦ C
seawater density 𝜌𝑆 1026.021 kg/m3
seawater kinematic viscosity 𝜈𝑆 1.1892 10−6 m2 /s
standard air density 𝜌𝐴 1.225 kg/m3

Table 30.3 Measured total resistance and sinkage of model; blockage correction (Schuster, 1956), dynamic sinkage,
and trim

Measured data Blockage correction, dynamic sinkage, and trim

𝑣𝐶 𝑅𝑇𝑀 𝑧1 𝑧2 𝐹𝑟ℎ Δ𝑣∕𝑣 𝑣𝑀 𝐹𝑟 𝑅𝑒𝑀 𝑧𝑉𝑀 𝑡𝑉
[m/s] [N] [cm] [cm] [–] [%] [m/s] [–] [–] [cm] [deg]
1.296 28.41 −1.080 −0.644 0.1988 1.2530 1.3125 0.1597 8 491 701.0 −0.875 −0.049
1.347 30.61 −1.214 −0.600 0.2065 1.2572 1.3635 0.1659 8 822 152.6 −0.925 −0.069
1.396 32.72 −1.286 −0.628 0.2141 1.2615 1.4136 0.1720 9 146 167.8 −0.976 −0.074
j j
1.446 35.16 −1.343 −0.675 0.2219 1.2660 1.4647 0.1782 9 476 709.5 −1.028 −0.075
1.496 37.92 −1.421 −0.718 0.2295 1.2707 1.5148 0.1843 9 800 817.5 −1.091 −0.079
1.546 40.73 −1.547 −0.780 0.2372 1.2756 1.5659 0.1905 10 131 458.6 −1.195 −0.084
1.596 43.63 −1.669 −0.886 0.2448 1.2807 1.6160 0.1966 10 455 668.9 −1.301 −0.088
1.645 46.71 −1.780 −0.907 0.2524 1.2859 1.6661 0.2027 10 779 933.8 −1.369 −0.098
1.694 49.99 −1.895 −0.960 0.2600 1.2914 1.7163 0.2088 11 104 255.9 −1.455 −0.105
1.745 53.59 −1.965 −1.004 0.2677 1.2972 1.7674 0.2150 11 435 126.3 −1.513 −0.108
1.794 57.62 −2.073 −1.085 0.2753 1.3031 1.8175 0.2211 11 759 572.9 −1.608 −0.111
1.846 62.07 −2.192 −1.141 0.2832 1.3096 1.8697 0.2275 12 097 067.9 −1.698 −0.118
1.895 66.46 −2.378 −1.301 0.2908 1.3160 1.9199 0.2336 12 421 653.8 −1.871 −0.121
1.943 71.59 −2.519 −1.389 0.2983 1.3226 1.9690 0.2396 12 739 819.5 −1.987 −0.127

Carriage speed 𝑣𝐶 , total resistance 𝑅𝑇𝑀 , and dynamic sinkage measurements 𝑧1 (aft) Recorded data
and 𝑧2 (forward) recorded during the tests are summarized in Table .. Note that the
total resistance is less than 0.5% of the weight force for the model even at the highest
speed. Therefore, carriage speed has to be kept as steady as possible because even small
fluctuations in velocity may cause accelerations big enough to generate inertia forces of
sufficient magnitude to distort the measurement.
As a first step, a blockage correction is selected and applied in order to find the actual Blockage
speed of the model through the water (see Section .). Here, the blockage correction correction,
sinkage, and trim
by Schuster () is applied using Equations (.) through (.), and Equations (.)
and (.). The velocity correction Δ𝑣∕𝑣 ranges from 1.25% to 1.32% above the carriage
velocity 𝑣𝐶 in this case. With the model velocity 𝑣𝑀 known, Froude and Reynolds
number are computed with Equation (.). Mean sinkage 𝑧𝑉 and running trim 𝑡𝑉 are

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

362 30 Full Scale Resistance Prediction

1.40
Data for regression line 1.35 linear regression
𝐹𝑟4𝑖 𝐶𝑇𝑀𝑖 CTMi /CFMi
𝐹𝑟 𝐶𝐹𝑀𝑖 𝐶𝑇𝑀𝑖 1.30

[−]
𝐶𝐹𝑀𝑖 𝐶𝐹𝑀𝑖
[–] [10−3 ] [10−3 ] [–] [–] 1.25

CTM /CFM
0.1597 3.0871 3.7257 0.2106 1.2069 1.20
0.1659 3.0664 3.7182 0.2471 1.2126 1.15
0.1720 3.0471 3.6987 0.2872 1.2138
1.10 b = 1.1566; form factor k = b − 1 = 0.1566
0.1782 3.0282 3.7021 0.3331 1.2225
0.1843 3.0105 3.7324 0.3833 1.2398 1.05
0.1905 2.9932 3.7514 0.4402 1.2533 1.00
0.1966 2.9769 3.7736 0.5021 1.2676 0.0 0.1 0.2 0.3 0.4 0.5 0.6
4
Fr /CFM [−]

Figure 30.1 Finding the form factor with Prohaska’s method; only data points with 0.1 ≤ 𝐹𝑟 ≤ 0.2 are used

j j

Figure 30.2 Measured mean sinkage and running trim Figure 30.3 Measured total resistance of model
angle of model as function of model speed

found using Equations (.) and (.). Negative values for 𝑧𝑉 and 𝑡𝑉 in Table .
indicate that the model is settling down and trimming aft. Both quantities increase
with increasing velocity (see Figure .).
Form factor, Based on the seven tests with Froude numbers in the range 0.1 ≤ 𝐹𝑟 ≤ 0.2, Prohaska’s
Prohaska’s method from Section . is applied to find the form factor with result 𝑘 = 0.1566.
method
Data points and regression line are shown in Figure ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

30.4 Example Resistance Prediction 363

Table 30.4 Resistance coefficients for the model (𝑘 = 0.1566)

𝑣𝑀 𝐹𝑟 𝑅𝑒𝑀 𝑅𝑇𝑀 𝐶𝑇𝑀 𝐶𝐹𝑀 𝐶𝑊

[m/s] [–] [–] [N] [10−3 ] [10−3 ] [10−3 ]
1.3125 0.1597 8 491 701.0 28.41 3.7257 3.0871 0.1552
1.3635 0.1659 8 822 152.6 30.61 3.7182 3.0664 0.1715
1.4136 0.1720 9 146 167.8 32.72 3.6987 3.0471 0.1743
1.4647 0.1782 94 76 709.5 35.16 3.7021 3.0282 0.1996
1.5148 0.1843 9 800 817.5 37.92 3.7324 3.0105 0.2503
1.5659 0.1905 10 131 458.6 40.73 3.7514 2.9932 0.2894
1.6160 0.1966 10 455 668.9 43.63 3.7736 2.9769 0.3304
1.6661 0.2027 10 779 933.8 46.71 3.8004 2.9612 0.3754
1.7163 0.2088 11 104 255.9 49.99 3.8334 2.9461 0.4258
1.7674 0.2150 11 435 126.3 53.59 3.8749 2.9313 0.4844
1.8175 0.2211 11 759 572.9 57.62 3.9394 2.9173 0.5652
1.8697 0.2275 12 097 067.9 62.07 4.0102 2.9032 0.6522
1.9199 0.2336 12 421 653.8 66.46 4.0727 2.8901 0.7299
1.9690 0.2396 12 739 819.5 71.59 4.1706 2.8777 0.8422

Model speed 𝑣𝑀 through the water, total resistance 𝑅𝑇𝑀 , and form factor 𝑘 make up Resistance
the input for the wave resistance coefficient computation. Following the procedure coefficients of
model
j outlined in Section ., Equation (.) yields the wave resistance coefficient by j
subtracting the viscous model resistance 𝐶𝑉𝑀 = (1 + 𝑘)𝐶𝐹𝑀 from the total resistance
coefficient 𝐶𝑇𝑀 . The results are shown in Table . and Figure .. Note that the
wave resistance coefficient 𝐶𝑊 of the model is also valid for the full scale vessel, because
it is based on a dynamically similar force.
The resistance coefficients for the full scale vessel are derived from Equations (.) Full scale
through (.). Again, full scale resistance predictions are done for seawater at temper- resistance
coefficients
ature 15 ◦ C. The standard seawater properties are included in Table .. First, ship
velocities 𝑣𝑆 are computed for the Froude numbers of the resistance test. From 𝑣𝑆
follow Reynolds numbers and the ITTC  model–ship correlation line coefficient
𝐶𝐹𝑆 (.) for the full scale vessel. Combined with the unchanged form factor 𝑘 and
the roughness allowance Δ𝐶𝐹 , the viscous resistance coefficient 𝐶𝑉𝑆 of the full scale
vessel is built:
𝐶𝑉𝑆 = (1 + 𝑘)𝐶𝐹𝑆 + Δ𝐶𝐹𝑆 (.)
The viscous resistance of the full scale vessel is augmented by the correlation allowance
𝐶𝐴 and by the increase in wetted surfaces due to areas not included in the model test.
Here, the additional surfaces of the bilge keels lead to the following correction factor:

𝑆𝑆 + 𝑆𝐵𝐾𝑆 5540.508 m2 + 85.0003m2

= = 1.0153
𝑆𝑆 5540.508 m2

The total resistance coefficient 𝐶𝑇 𝑆 (.) for the full scale vessel is obtained by adding
wave resistance coefficient 𝐶𝑊 and air resistance coefficient 𝐶𝐴𝐴𝑆 to the augmented vis-
cous resistance coefficient. The example vessel does not feature additional appendages.
Consequently, we set 𝐶𝐴𝑃 𝑃𝑆 to zero. All full scale resistance coefficients are presented

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

364 30 Full Scale Resistance Prediction

4.5 4.5
CTS total resistance
CFS ITTC 1957
4.0 4.0
[10−3 ]

(1 + k)CFS 3D frictional resistance

[10−3 ]
CW wave resistance
3.5 3.5 CA correlation allowance
resistance coefficients at model scale C?M

ΔCF roughness allowance

resistance coefficients at full scale C?S

3.0 3.0

CTM total resistance

2.5 2.5
CFM frictional resistance
(1 + k)CFM viscous resistance
2.0 CW wave resistance 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0
0.16 0.18 0.20 0.22 0.24 0.16 0.18 0.20 0.22 0.24
Froude number Fr [−] Froude number Fr [−]

j Figure 30.4 Resistance coefficients of model Figure 30.5 Resistance coefficients of full scale ship
j

Table 30.5 Predicted resistance coefficients for the full scale vessel

𝐹𝑟 𝑣𝑆 𝑅𝑒𝑆 𝐶𝐹𝑆 (1 + 𝑘)𝐶𝐹𝑆 Δ𝐶𝐹 𝐶𝐴 𝐶𝑊 𝐶𝑇 𝑆

[–] [m/s] [–] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ]
0.1597 6.562 982 468 812.4 1.5340 1.7742 0.1024 0.2846 0.1552 2.4296
0.1659 6.818 1 020 701 245.3 1.5267 1.7659 0.1080 0.2747 0.1715 2.4330
0.1720 7.068 1 058 189 005.1 1.5199 1.7580 0.1132 0.2653 0.1743 2.4236
0.1782 7.324 1 096 431 860.8 1.5133 1.7503 0.1183 0.2560 0.1996 2.4368
0.1843 7.574 1 133 930 354.9 1.5070 1.7431 0.1231 0.2472 0.2503 2.4761
0.1905 7.829 1 172 184 711.5 1.5009 1.7360 0.1277 0.2386 0.2894 2.5039
0.1966 8.080 1 209 695 047.8 1.4951 1.7293 0.1321 0.2304 0.3304 2.5342
0.2027 8.331 1 247 211 689.6 1.4895 1.7228 0.1363 0.2224 0.3754 2.5688
0.2088 8.581 1 284 734 951.6 1.4841 1.7166 0.1403 0.2147 0.4258 2.6092
0.2150 8.837 1 323 015 844.6 1.4788 1.7104 0.1442 0.2071 0.4844 2.6577
0.2211 9.088 1 360 553 510.5 1.4738 1.7046 0.1479 0.1998 0.5652 2.7290
0.2275 9.348 1 399 600 847.2 1.4687 1.6987 0.1517 0.1924 0.6522 2.8064
0.2336 9.599 1 437 154 639.4 1.4640 1.6933 0.1551 0.1855 0.7299 2.8750
0.2396 9.845 1 473 965 621.8 1.4595 1.6881 0.1584 0.1789 0.8422 2.9787
for all speeds: factor (𝑆𝑆 + 𝑆𝐵𝐾𝑆 )∕𝑆𝑆 = 1.0153; air resistance coefficient is 103 𝐶𝐴𝐴𝑆 = 0.08

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

30.4 Example Resistance Prediction 365

Table 30.6 Full scale resistance 𝑅𝑇𝑆 and effective power 𝑃𝐸𝑆

𝐹𝑟 𝑣𝑆 𝑣𝑆 𝐶𝑇 𝑆 𝑅𝑇𝑆 𝑃𝐸𝑆
[–] [m/s] [kn] [10−3 ] [kN] [kW]
0.1597 6.562 12.76 2.4296 297.38 1951.53
0.1659 6.818 13.25 2.4330 321.43 2191.42
0.1720 7.068 13.74 2.4236 344.14 2432.43
0.1782 7.324 14.24 2.4368 371.48 2720.54
0.1843 7.574 14.72 2.4761 403.74 3057.89
0.1905 7.829 15.22 2.5039 436.27 3415.79
0.1966 8.080 15.71 2.5342 470.27 3799.84
0.2027 8.331 16.19 2.5688 506.71 4221.24
0.2088 8.581 16.68 2.6092 546.11 4686.30
0.2150 8.837 17.18 2.6577 589.92 5213.09
0.2211 9.088 17.67 2.7290 640.60 5821.53
0.2275 9.348 18.17 2.8064 697.12 6517.00
0.2336 9.599 18.66 2.8750 753.00 7228.31
0.2396 9.845 19.14 2.9787 820.64 8079.37

in Table . and Figure .. Finally, employing Equations (.) and (.), the dimen-
sional total resistance 𝑅𝑇𝑆 and effective power 𝑃𝐸𝑆 can be predicted for the full scale
j vessel based on the total resistance coefficient 𝐶𝑇 𝑆 . Results are shown in Table . j
and Figures . and ..

References
Holtrop, J. and Mennen, G. (). An approximate power prediction method. Inter-
national Shipbuilding Progress, ():–.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
ITTC (a).  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
ITTC (b). Resistance test. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .
Schuster, S. (). Beitrag zur Frage der Kanalkorrektur bei Modellversuchen. Schiff-
stechnik/Ship Technology Research, :–.

Self Study Problems

. Why is a roughness allowance applied to the full scale vessel but not to the model?

. Explain the purpose of the correlation allowance.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

366 30 Full Scale Resistance Prediction

j Figure 30.6 Full scale total resistance prediction (calm Figure 30.7 Full scale effective power j
water)

. Write a program or create a spread sheet to perform the evaluation of a resistance

test using the latest version of the ITTC recommended procedures. Use the ship’s
geometric data, carriage speed 𝑣𝐶 , and measured total resistance 𝑅𝑇𝑀 of the
model as input. Implement Schuster’s blockage correction.

. Add an option for Tamura’s blockage correction to your program/spread sheet.

How much does the predicted resistance change for the lowest and highest speed
of the example in Section .?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

367

31
Resistance Estimates – Guldhammer
and Harvald’s Method

Data on resistance and powering requirements are required in the ship design process
long before a model test can be conducted. Therefore, methods have been developed
which allow the naval architect to estimate resistance and powering requirements based
on ship speed and a few principal geometric characteristics. Traditionally, data of model
tests with systematically varied hull forms and resulting design charts have been used.
However, the advent of the personal computer has mostly replaced these with formulas
derived from regression analysis of vast sets of model test and trial data. As an example
for such a method, this chapter illustrates a resistance estimate which was originally
developed by Guldhammer and Harvald () and later amended in collaboration with
j other researchers (Andersen and Guldhammer, ; Kristensen and Lützen, ). j

Learning Objectives
At the end of this chapter students will be able to:

• understand the basics of resistance estimates

• program and apply Guldhammer and Harvald’s method

• study the relationships between main particulars and calm water resistance

31.1 Historical Development

One of the earliest attempts to compare and predict performance of ships is based on Admiralty
the admiralty coefficient: coefficient

Δ2∕3 𝑣3𝑆
𝐴𝐶 = (.)
𝑃
In order to estimate the effective power 𝑃𝐸 = 𝑅𝑇 𝑣𝑆 for a new ship design, the admiralty
coefficient of a very similar ship is multiplied
√by the third root of displacement Δ squared
and the design speed 𝑣𝑆 cubed, i.e. 𝑃𝐸 = 3
Δ2 𝑣3𝑆 𝐴𝐶 .

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

368 31 Resistance Estimates – Guldhammer and Harvald’s Method

Unfortunately, admiralty coefficients 𝐴𝐶 are dimensional and their values depend on

the units used. Harvald () suggests the following formula for preliminary design:
(√ )
75
𝐴𝐶 = 3.7 𝐿+ (.)
𝑣𝑆
Here, length of the ship 𝐿 is entered in meter and the speed 𝑣𝑆 in meter per second.
The power is obtained in kilowatt, if the displacement Δ is entered in metric tons, i.e.
1 t = 1000 kg. As ship length 𝐿 the length in waterline 𝐿𝑊𝐿 is applied for vessels
without a bulbous bow, and length over wetted surface 𝐿𝑂𝑆 is used for vessels with
bulbous bow.
Admiralty coefficients capture fundamental relationships between speed, size of the
vessel, and the power required to move them. Due to the simplicity of the formula (.),
power estimates are only good for ships with similar geometry and speed and the results
encompass high uncertainty. If at all, they should be used only as a rough first estimate.
Systematic series With Froude’s hypothesis and the availability of model testing facilities, naval architects
could analyze the resistance of an existing hull form. Obviously, it is desirable to predict
the resistance in advance to eliminate impractical hull forms before time consuming
model tests are conducted. Early on in the th century, model basins developed basic
hull designs and then systematically varied important geometric characteristics like
the prismatic coefficient 𝐶𝑃 , position of the longitudinal center of buoyancy 𝐿𝐶𝐵,
beam–draft ratio, and more.
j Taylor standard Already in  the Experimental Model Basin, which became later known as the David j
series Taylor Model Basin, developed a parent hull model from the lines of the British cruiser
Leviathan.  models were derived from this hull and tested to explore the effect of the
prismatic coefficient on resistance. Over the following five decades more parent hulls
and overall more than  models were added to the data base which is referred to as
Taylor Standard Series. Like the model basin he founded, the Taylor Standard Series is
named after Admiral David W. Taylor (* – †), naval architect, engineer, and
Chief Constructor of the United States Navy during World War I. With the continuous
improvement of model testing procedures, the data have been reanalyzed and published
by Gertler () as residuary resistance curves based on the Schoenherr friction line
Equation (.).
Series 60 In the s, a new series was developed for merchant vessels covering a wide range of
block coefficients. It became known as Series , and the data have been widely used in
ship design (Todd, ). Several other systematic series exist, some of which address
specific ship types like trawlers and high speed planing boats (Lewis, ). All series
provide useful insight into the relationships between resistance and key geometric
characteristics. However, most of the series are decades old and do not reflect modern
hull designs.
Estimates based Some series data are available as regression formulas, but most are only available
on regression as charts that do not fit very well into today’s computerized engineering work flow.
Over the past  years several researchers have developed resistance estimates which
condensed the results of available model test and ship trial data into relatively simple
formula which are easily programmed. Three of these that are applicable to displacement
type vessels are discussed in this book. As a first example, the following section presents
the method by Guldhammer and Harvald () in its latest form. Methods by Holtrop

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.2 Guldhammer and Harvald’s Method 369

Table 31.1 Range of parameters suitable for Guldhammer and Harvald’s method

Permissible range of main dimensions

Parameter Symbol Unit Range
ship speed 𝑣 [m/s] 𝐹𝑟 < 0.33
length-beam ratio 𝐿∕𝐵 [–] 5.0 … 8.0
block coefficient 𝐶𝐵 √ [–] 0.55 … 0.85
length-displacement ratio [–]
3
𝑀 = 𝐿∕ 𝑉 4.0 … 6.0

and Mennen () and Hollenbach () are described in some detail after we have
covered ship propulsion and propeller selection.

31.2 Guldhammer and Harvald’s Method

The method was developed in the late s and early s when most ships did not yet have Origin and
a bulbous bow. In later publications, Harvald () and Andersen and Guldhammer computerization
() updated the procedure and introduced a computation length 𝐿 which made the
method applicable to ships with bulbous bows. Andersen and Guldhammer ()
provide regression formulas to replace the traditional 𝐶𝑅 -charts and extend the method
to include powering estimates. The regression formulas are the basis for this section, as
j they are more applicable to today’s computerized work flow than H.E. Guldhammer and j
S.A. Harvald’s original design charts. However, the design charts provide an important
tool to perform basic error checks during implementation.
In addition, more recent updates reported by Kristensen and Lützen () are refer- Updates to the
enced. The report provides formulas for wetted surfaces of common vessel types and method
an updated bulbous bow correction for the residuary resistance, which promises to
give more accurate results for modern hull forms. Kristensen and Lützen () also
include additional regression formulas for the residuary resistance of ships with high
prismatic coefficient 𝐶𝑃 .

31.2.1 Applicability

The method of Guldhammer and Harvald is applicable to single and twin screw dis- Single and twin
placement type vessels. Based on the regression formulas provided by Andersen and screw vessels
Guldhammer () for the residuary resistance, limits of application for the method
are shown in Table .. The original charts allow predictions for velocities up to Froude
number 𝐹𝑟 = 0.45.

31.2.2 Required input

Guldhammer and Harvald’s method requires only the hull form parameters listed in Input
Table .. A special computation length 𝐿 plays a prominent role for modern hull
forms.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

370 31 Resistance Estimates – Guldhammer and Harvald’s Method

Table 31.2 Required and optional input parameters for Guldhammer and Harvald’s method

Parameter Symbol Remarks

required parameters
length between perpendiculars 𝐿𝑃𝑃
length of aft overhang in waterline 𝐿aft usually 𝐿𝑊𝐿 = 𝐿𝑃𝑃 + 𝐿aft
extension of 𝑆 beyond fore perpend. 𝐿fore
computation length 𝐿 total extension of wetted surface (usually
equal to 𝐿𝑂𝑆 )
maximum molded beam in waterline 𝐵
molded draft 𝑇
block coefficient 𝐶𝐵 or volumetric displacement 𝑉
prismatic coefficient 𝐶𝑃 or midship section area 𝐴𝑀
transverse cross section area of bulb 𝐴𝐵𝑇 at forward perpendicular FP
optional parameters
propeller diameter 𝐷𝑃 for propulsion analysis
longitudinal center of buoyancy 𝐿𝐶𝐵 or assume optimum position
wetted surface (hull+rudder) 𝑆
wetted surface of appendages 𝑆𝐴𝑃𝑃 bilge keels, stabilizer fins, bossings, etc.
form factors for fore and aft body 𝐹𝐹 , 𝐹𝐴 −3 ≤ 𝐹𝐴 , 𝐹𝐹 ≤ +3

j j

Figure 31.1 Definition of the midship section and the computational length 𝐿 for Guldhammer
and Harvald’s resistance estimate (Andersen and Guldhammer, 1986)

Computation Before , the longitudinal extension of a submerged hull was usually equal to its
length length in the waterline 𝐿𝑊𝐿 . Therefore, the waterline length often served as basis for
resistance computations. However, with the introduction of bulbous bows, submerged
volume is stretched over a longer length. The increased length has to be considered in the
resistance estimate. Andersen and Guldhammer () called it the computation length
𝐿 (see Figure .). It is used as reference length throughout the method described
here.
For most modern ships, the computation length will be equal to the length over wetted
surface 𝐿𝑂𝑆 . It is derived from the length between perpendiculars 𝐿𝑃𝑃 by adding

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.2 Guldhammer and Harvald’s Method 371

the length 𝐿fore of the additional wetted parts protruding forward of the forward
perpendicular FP (mostly for bulbous bows) and the length 𝐿aft of wetted surface
extending past the aft perpendicular AP.
𝐿 = 𝐿𝑃𝑃 + 𝐿aft + 𝐿fore (.)
𝐿aft is positive, if the submerged hull extends aft of the aft perpendicular. 𝐿fore is
positive, if the submerged hull extends forward of the fore perpendicular. In unusual
cases, 𝐿fore or 𝐿aft may be negative.
If the length in waterline or the aft overhang are unknown, Kristensen and Lützen
() suggest to use the following relationships:
𝐿𝑊𝐿 = 1.02 𝐿𝑃𝑃 for tanker and bulk carriers
(.)
𝐿𝑊𝐿 = 1.01 𝐿𝑃𝑃 for container ships

Guldhammer and Harvald () assumed that the waterline extends about % of 𝐿𝑃𝑃 Definition of
aft of the aft perpendicular AP. Therefore, midship was located 48.5% of the length in midship location
waterline 𝐿𝑊𝐿 aft of the foremost point of displacement (originally equal to FP). Within
the method, this definition of midship location is kept for vessels with bulbous bow.
Nowadays, the foremost point of displacement is the tip of the bulb. As a result, the
midship location moves forward of the usual position at 𝐿𝑃𝑃 ∕2 (see Figure .).
Consequently, the longitudinal location of the center of buoyancy has to be stated with Longitudinal
respect to the midship position as defined by Andersen and Guldhammer (). center of
buoyancy
j j
If the position of the center of buoyancy is given as length 𝐿𝐶𝐵0 with respect to
the traditional midship position (𝐿𝑃𝑃 ∕2), its location with respect to Guldhammer’s
midship definition is
𝐿𝐶𝐵 = 𝐿𝐶𝐵0 − 0.015𝐿𝑃𝑃 + 0.485𝐿aft − 0.515𝐿fore (.)
Obviously, all quantities have the dimension of the selected length unit. The equation
in Andersen and Guldhammer () corresponding to Equation (.) has a misprint:
0.15 𝐿𝑃𝑃 instead of 0.015 𝐿𝑃𝑃 . Note that in contrast to Andersen and Guldhammer
(), a positive 𝐿𝐶𝐵 indicates that the center of buoyancy is forward of midships
throughout this book.
In early design stages, the wetted surface 𝑆 is often not yet known. Andersen and Wetted surface
Guldhammer () suggest Mumford’s formula (.) but it is likely to underpredict
the wetted surface of modern ship hulls with bulbous bow.
( )
𝑆 = 1.025 𝐿𝑃𝑃 𝐶𝐵 𝐵 + 1.7 𝑇
( ) (.)
𝑉
= 1.025 + 1.7 𝐿𝑃𝑃 𝑇 Mumford’s formula
𝑇

Kristensen and Lützen () modified the constants in Mumford’s formula to improve
the wetted surface prediction for modern hull forms. The formulas employ the length
in waterline instead of the length between perpendiculars.
( )
𝑉
𝑆 = 0.99 + 1.9 𝐿𝑊𝐿 𝑇 bulk carriers, tankers, vessels with high 𝐶𝐵
𝑇
( ) (.)
𝑉
𝑆 = 0.995 + 1.9 𝐿𝑊𝐿 𝑇 container ships (vessels with low to medium 𝐶𝐵 )
𝑇

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

372 31 Resistance Estimates – Guldhammer and Harvald’s Method

( )
𝑉 twin screw ships (Ro-Ro vessels) with
𝑆 = 1.53 + 0.55 𝐿𝑊𝐿 𝑇
𝑇 open shaft lines
( )
𝑉 twin skeg ships (Ro-Ro vessels with twin
𝑆 = 1.2 + 1.5 𝐿𝑊𝐿 𝑇 (.)
𝑇 rudders)
( )
𝑉
𝑆 = 1.11 + 1.7 𝐿𝑊𝐿 𝑇 double ended ferries
𝑇
Kristensen and Lützen () also provide two useful equations to estimate the change
in wetted surface for changes in draft. Assuming that the wetted surface 𝑆1 at draft 𝑇1
is known, the wetted surface 𝑆2 at draft 𝑇2 is approximately equal to
( )( )
𝑆2 = 𝑆1 + 2.0 𝑇2 − 𝑇1 𝐿𝑊𝐿 + 𝐵 tanker, bulk carrier
( )( ) (.)
𝑆2 = 𝑆1 + 2.4 𝑇2 − 𝑇1 𝐿𝑊𝐿 + 𝐵 container ships

31.2.3 Resistance estimate

Total resistance Within the resistance estimation method by Guldhammer and Harvald the total resis-
tance coefficient is defined as
[ ]
𝐶𝑇 = 𝐶𝑅 + 𝐶𝐹′ + 𝐶𝐴 + 𝐶𝐴𝐴 + 𝐶𝐴𝑆 (.)

𝐶𝑅 is the residuary resistance coefficient. The frictional resistance coefficient 𝐶𝐹′

is based on the ITTC  model–ship correlation line augmented by the effect of
j appendage surfaces not included in the wetted surface 𝑆. 𝐶𝐴 stands for a combined j
correlation and roughness allowance. As usual, 𝐶𝐴𝐴 is the air resistance coefficient.
The steering resistance coefficient 𝐶𝐴𝑆 originally included in Equation (.) has been
dropped by Kristensen and Lützen ().
All resistance coefficients have been rendered dimensionless by division with the dy-
namic pressure 12 𝜌 𝑣2 and the wetted surface 𝑆 of the hull at rest. Therefore, the total
resistance is computed as usual:

1 2
𝑅𝑇 = 𝜌 𝑣 𝑆 𝐶𝑇 (.)
2

Residuary resistance

Residuary Key to all resistance estimates is the derivation of wave or residuary resistance. The
resistance method does not employ a form factor, hence, the residuary resistance coefficient 𝐶𝑅
is used. In Guldhammer and Harvald’s method, it is a function of Froude number 𝐹𝑟,
prismatic coefficient 𝐶𝑃 , and the length–displacement ratio (.).

𝐿
𝑀 = √ (.)
3
𝑉

For consistency, the prismatic coefficient 𝐶𝑃 should be computed on the basis of the
computation length 𝐿.
Residuary Guldhammer and Harvald () provided charts for 𝐶𝑅 similar to the ones shown in
resistance charts Figures . and .. Charts are selected based on the length–displacement ratio 𝑀,

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.2 Guldhammer and Harvald’s Method 373

and values for 𝐶𝑅 read from the curve with the correct prismatic coefficient 𝐶𝑃 as a
function of Froude number 𝐹𝑟.
In order to accommodate today’s computerized work flow, Andersen and Guldhammer Residuary
() provided regression formulas for 𝐶𝑅 which replace the charts. The example resistance
formula
charts have been redrawn based on the regression formula (.). They represent the
original charts fairly well. Although for Froude numbers above 0.3 and higher prismatic
coefficients (𝐶𝑃 > 0.7), resistance coefficients tend to be larger as found in the original
charts, Andersen and Guldhammer () state that the regression formulas for 𝐶𝑅 are
usable for Froude numbers smaller than 𝐹𝑟 < 0.33. Consequently, curves for 𝐹𝑟 > 0.33
are printed in light gray in Figures . and .. For higher Froude numbers, the
original charts should be consulted which provide residuary resistance coefficients up
to 𝐹𝑟 = 0.45 (Guldhammer and Harvald, ).
Andersen and Guldhammer () stated the following regression formula for a stan- 𝐶𝑅std value for
dard residuary resistance coefficient 𝐶𝑅std : standard form

103 𝐶𝑅std = 𝐸 + 𝐺 + 𝐻 + 𝐾 (.)

The four contributions depend on Froude number 𝐹𝑟, length–displacement ratio 𝑀,

and prismatic coefficient 𝐶𝑃 :

( )⎡ ⎤
2.5 ⎥ ( )4
𝐸 = 𝐴0 + 1.5𝐹𝑟1.8 + 𝐴1 𝐹𝑟𝑁1 ⎢0.98 + ( ) + 𝑀 − 5 (𝐹𝑟−0.1)4 (.)
⎢ 𝑀 −2 ⎥⎦
4
⎣
j j
with variables
𝐴0 = 1.35 − 0.23 𝑀 + 0.012 𝑀 2 𝐴1 = 0.0011 𝑀 9.1
𝐿
𝑁1 = 2 𝑀 − 3.7 𝑀= √
3
𝑉
( )( ) 2
7 − 0.09 𝑀 2 5 𝐶𝑃 − 2.5
𝐺 = [ ]1.5 (.)
( )2
600 𝐹𝑟 − 0.315 + 1
[ ]
𝐻 = e𝐻1 with 𝐻1 = 80 𝐹𝑟 − (0.04 + 0.59 𝐶𝑃 ) − 0.015(𝑀 − 5) (.)
𝐾 = 180 𝐹𝑟3.7 e(20 𝐶𝑃 −16) (.)

The residuary resistance coefficient 𝐶𝑅std is valid only for what was considered a standard Standard hull
hull form in the s: form

. beam–draft ratio of 𝐵∕𝑇 = 2.5

. location of 𝐿𝐶𝐵 at its optimal position (see later)
. no bulbous bow
. no appendages like bossings, struts, etc.

For ship hulls that deviate from the standard shape, five corrections have been intro- 𝐶𝑅 corrections
duced by Guldhammer and Harvald ():

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

374 31 Resistance Estimates – Guldhammer and Harvald’s Method

Figure 31.2 Guidance for the optimum

[% of L]
4
location of 𝐿𝐶𝐵 as a function of Froude
number 𝐹𝑟. Here, negative 𝐿𝐶𝐵 values standard (optimum) LCB
3
indicate a location aft of midship

forward
optimum LCB-corridor

longitudinal center of buoyancy LCB

2

0 G & H midships

2
aft
3

4
0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 0.28 0.30 0.32
Froude number Fr [−]

. Δ𝐶𝑅(𝐵∕𝑇 ) for beam–draft ratios other than 2.5

. Δ𝐶𝑅(𝐿𝐶𝐵) for 𝐿𝐶𝐵 positions deviating from optimum
. Δ𝐶𝑅(form) for section shapes at bow and stern that deviate from normal hull forms
j j
. Δ𝐶𝑅(bulb) for bulbous bows
. Δ𝐶𝑅(app) for appendages other than the typical single rudder and bilge keels

The adjusted residuary resistance coefficient will be

𝐶𝑅 = 𝐶𝑅std + Δ𝐶𝑅(𝐵∕𝑇 ) + Δ𝐶𝑅(form) + Δ𝐶𝑅(bulb)

+ Δ𝐶𝑅(𝐿𝐶𝐵) + Δ𝐶𝑅(app) (.)

Beam–draft ratio For vessels with 𝐵∕𝑇 ≠ 2.5 a simple speed independent correction is applied.
correction ( )
𝐵
103 Δ𝐶𝑅(𝐵∕𝑇 ) = 0.16 − 2.5 (.)
𝑇

Longitudinal In their original work, Guldhammer and Harvald () provided two charts to estimate
center of a correction for 𝐿𝐶𝐵-locations which deviate from the presumed optimum. This
buoyancy
correction
correction bears considerable uncertainties, and Kristensen and Lützen () suggest
to drop it altogether. For the sake of completeness, we detail the procedure outlined in
Andersen and Guldhammer ().
Standard 𝐿𝐶𝐵 Figure . provides an 𝐿𝐶𝐵-location corridor which is considered optimal for a given
design speed. A regression formula for the mean line is reported in Andersen and
Guldhammer (). The formula below has an additional minus sign, reflecting the
convention that a positive 𝐿𝐶𝐵 is forward of midships.
( )
𝐿𝐶𝐵std (𝐹𝑟) = − 0.44𝐹𝑟 − 0.094 𝐿 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.2 Guldhammer and Harvald’s Method 375

Table 31.3 Bulbous bow corrections to the standard residuary resistance coefficient (Andersen
and Guldhammer, 1986)

𝐶𝑃 𝐹𝑟 0.15 0.18 0.21 0.24 0.27 0.30 0.33

0.5 +0.2 +0.2 +0.2 0.0 −0.2 −0.4 −0.4
0.6 +0.2 +0.2 +0.2 0.0 −0.2 −0.3 −0.3
Δ𝐶𝑅(table)
0.7 +0.2 +0.2 0.0 −0.2 −0.3 −0.3 −0.3
0.8 +0.1 0.0 −0.2 −0.2 −0.2 −0.2 −0.2

Based on the actual longitudinal center of buoyancy location 𝐿𝐶𝐵actual of the ship
project, the relative deviation from the standard value for each speed is determined:

𝐿𝐶𝐵actual − 𝐿𝐶𝐵std (𝐹𝑟)

Δ𝐿𝐶𝐵(𝐹𝑟) = (.)
𝐿
The actual 𝐿𝐶𝐵actual is of course considered constant. Effects due to sinkage and trim
are ignored.
A correction for the residuary resistance coefficient is computed, if, for a given Froude 𝐿𝐶𝐵-correction
number 𝐹𝑟, the actual 𝐿𝐶𝐵 is forward of the standard 𝐿𝐶𝐵 location, i.e. Δ𝐿𝐶𝐵 > 0.

⎧
⎪0 if 𝐹𝑟 < (−𝐶𝑃2 + 1.1𝐶𝑃 − 0.0875) or Δ𝐿𝐶𝐵 ≤ 0
⎪ ( )
j 103 Δ𝐶𝑅(𝐿𝐶𝐵) = ⎨ j
𝐹𝑟
⎪90 − 1 Δ𝐿𝐶𝐵 else
⎪ −𝐶𝑃2 + 1.1𝐶𝑃 − 0.0875
⎩
(.)
Again, this correction has been dropped from the newest version of the Guldhammer
and Harvald method (Kristensen and Lützen, ).
A standard hull form has neither pronounced V- nor pronounced U-shaped stations Shape factors
in entrance and run of the hull. A raked stem without a bulbous bow and a moderate
cruiser stern are considered normal. Modern bulk carriers and tankers with high block
coefficient often feature extreme U-shaped stations in the fore body (entrance). Fast
container ships often have entrances with pronounced V-shaped stations. Guldhammer
and Harvald () assigned fore and aft body shape factors 𝐹𝐹 and 𝐹𝐴 between −3 and
+3, respectively. The value −3 is for extreme V-shaped stations, 0 for normal stations,
and +3 for extreme U-shaped stations.
Once values have been assigned to 𝐹𝐹 and 𝐹𝐴 , a correction to residuary resistance Form correction
coefficient is computed from
( )
0.1
103 Δ𝐶𝑅(form) = − 𝐹𝐹 − 𝐹𝐴 (.)
3
Clearly, U-shaped stations in the fore body are deemed beneficial whereas V-shaped
stations increase resistance. Vice-versa for the aft body.
The influence of a bulbous bow on residuary resistance is derived by interpolation Bulbous bow
in 𝐹𝑟- and 𝐶𝑃 -directions between the values provided in Table . (Andersen and correction
Guldhammer, ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

376 31 Resistance Estimates – Guldhammer and Harvald’s Method

Original bulbous The full correction is applied, if the transverse cross section area of the bulb at the
bow correction forward perpendicular 𝐴𝐵𝑇 is larger than 10% of the midship section area 𝐴𝑀 . If
𝐴𝐵𝑇 < 0.1𝐴𝑀 , the correction is linearly scaled down to zero.

⎧103 Δ𝐶𝑅(table) from Table . if

⎪ 𝐴𝐵𝑇 ≥ 0.1𝐴𝑀
103 Δ𝐶𝑅(bulb) = ⎨ (.)
10𝐴𝐵𝑇
⎪103 Δ𝐶𝑅(table) if 𝐴𝐵𝑇 < 0.1𝐴𝑀
⎩ 𝐴𝑀

New bulbous Kristensen and Lützen () analyzed model tests of  newer tanker and bulk carrier
bow correction hull forms and  newer hull forms of vessels with lower block coefficients. As indicated
by Andersen and Guldhammer (), they concluded that the original bulbous bow
correction is too pessimistic. Experience in the design of bulbs and optimization of
their shape has made them more effective. The effectiveness of a bulbous bow depends
on different factors and its exact shape should be determined by shape optimization
later in the design process. Design guidance can be found in Kracht ().
To reflect these improvements, Kristensen and Lützen () proposed new corrections
making a distinction between vessels with high (𝐶𝐵 ≈ 0.8) and low block coefficients
(𝐶𝐵 ≈ 0.6).

⎧ ( ) for tankers and bulk

⎪max − 0.4, −(0.1 + 1.6𝐹𝑟) carriers (high 𝐶 )
⎪ 𝐵
j 103 Δ𝐶𝑅(bulb)
′
= ⎨( ) j
⎪ 250𝐹𝑟 − 90 for container ships (vessels
⎪ 103 𝐶𝑅(no bulb)
⎩ 100 with low 𝐶𝐵 )
(.)
For vessels with high block coefficient (and 𝑀 ∈ [4.4, 5.2] and 𝐶𝑃 ∈ [0.78, 0.87]), the
correction is mostly speed dependent. More slender vessels have larger variations of the
length–displacement ratio 𝑀 and prismatic coefficient 𝐶𝑃 . Therefore, the correction is
expressed as a fraction of the residuary resistance coefficient without a bulb.
( )
103 𝐶𝑅(no bulb) = 103 𝐶𝑅std + Δ𝐶𝑅(𝐵∕𝑇 ) + Δ𝐶𝑅(𝐿𝐶𝐵) + Δ𝐶𝑅(form) (.)

Although not specifically mentioned, it seems safe to assume that the new correction
values (.) have to be scaled for bulbous bows with 𝐴𝐵𝑇 less than % of 𝐴𝑀 .

⎧103 Δ𝐶 ′ from Equation (.) if

⎪ 𝑅(bulb) 𝐴𝐵𝑇 ≥ 0.1𝐴𝑀
103 Δ𝐶𝑅(bulb) = ⎨ (.)
10𝐴𝐵𝑇
⎪103 Δ𝐶 ′ if 𝐴𝐵𝑇 < 0.1𝐴𝑀
⎩ 𝑅(bulb) 𝐴
𝑀

Once all adjustments have been evaluated, the final residuary resistance coefficient can
be computed from Equation (.).
Appendage Guldhammer and Harvald () also include a correction for appendages. Effects
correction for 𝐶𝑅 of single rudder and bilge keels are included in the standard 𝐶𝑅 value. However, if
the vessel features bossings, exposed shafts, and shaft brackets they may affect wave
and viscous pressure resistance which warrants additional corrections to the residuary
resistance coefficient.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.2 Guldhammer and Harvald’s Method 377

• bossings: add 3%–5% of 𝐶𝑅std for full ship forms

• shafts and brackets: add 5%–8% of 𝐶𝑅std for fine hull forms

Andersen and Guldhammer () caution that the corrections should be small com-
pared to the overall 𝐶𝑅 value.

Frictional resistance

The frictional resistance coefficient is computed with the ITTC  model–ship corre- Frictional
lation line. resistance
0.075 coefficient
𝐶𝐹 = [ ]2 (.)
log10 𝑅𝑒 − 2
Again, a form factor is not used in Guldhammer and Harvald’s method.
In case the wetted surface 𝑆 does not include the surface of appendages, the frictional
coefficient should be augmented by the ratio of total wetted surface 𝑆 + 𝑆𝐴𝑃𝑃 and
wetted surface of the hull 𝑆.
𝑆 + 𝑆𝐴𝑃𝑃
𝐶𝐹′ = 𝐶𝐹 (.)
𝑆

Additional resistance components

j j
The incremental resistance coefficient represents a combined correlation and roughness Incremental
allowance 𝐶𝐴 . Guldhammer and Harvald (, (page )) stated only a few discrete resistance
coefficient
values depending on vessel length. In their update to the procedure, Andersen and
Guldhammer () provided the formula (.).
( )2
103 𝐶𝐴 = 0.5 log10 (𝑉 ) − 0.1 log10 (𝑉 ) (.)

It considers that the roughness effect tends to decrease with increasing ship length
and size. The input is unfortunately dimensional, and the displacement 𝑉 must be
entered in cubic meter [m3 ]. This equation should potentially be replaced with the
roughness allowance Δ𝐶𝐹 from the ITTC  performance prediction procedure (see
Equation (.)).
Air resistance coefficient 𝐶𝐴𝐴 and steering resistance coefficient 𝐶𝐴𝑆 are assumed to be Air and steering
constant in Guldhammer and Harvald () and Andersen and Guldhammer (). resistance
Suggested values are

103 𝐶𝐴𝐴 = 0.07 (.)

103 𝐶𝐴𝑆 = 0.04 (.)

The air resistance represents the resistance caused by moving the superstructure of
the vessel through air at rest. Resistance in high winds must be considered separately
for the selection of a service margin and maneuvering assessment. Steering resistance
may occur in vessels which need constant rudder action to counter the wheel effect of
a single propeller.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

378 31 Resistance Estimates – Guldhammer and Harvald’s Method

Table 31.4 Air resistance Vessel type 103 𝐶𝐴𝐴

coefficients for different types of small tanker 0.07
vessels (Kristensen and Lützen, 2012) Handysize tanker 0.07
Handymax tanker 0.07
Panamamax tanker 0.05
Aframax tanker 0.05
Suezmax tanker 0.05
very large crude carrier (VLCC) 0.04
container ship see Equation (.)

Table 31.5 Principal dimensions for resistance estimate example

1000 TEU container ship

length between perpendiculars 𝐿𝑃𝑃 145.0 m
extension of bulb forward of FP 𝐿fore 3.3 m
extension of waterline aft of AP 𝐿aft 2.7 m
beam 𝐵 24.0 m
draft 𝑇 8.2 m
displacement 𝑉 18 872.0 m3
estimated wetted surface 𝑆 4400.0 m2
bulbous bow cross section area at FP 𝐴𝐵𝑇 14.0 m2
wetted area of appendages 𝑆𝐴𝑃𝑃 52.0 m2
j prismatic coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝑃 0.6783 j
location of (with respect to 𝐿𝑃𝑃 ∕2) 𝐿𝐶𝐵0 +0.4 % of 𝐿𝑃𝑃
design speed 𝑣𝑆 17.5 kn
standard shaped fore and aft body

Kristensen and Lützen () analyzed data for different types of vessels and recommend
the constants in Table .. For container ships they developed a simple formula based
on the container carrying capacity of the vessel.
( )
103 𝐶𝐴𝐴 = max (0.28 TEU−0.126 ), 0.09 (.)

TEU stands for Twenty-foot Equivalent Unit, i.e. a  ft long container. The steering
resistance is no longer used in their update to the Guldhammer and Harvald method.
Andersen and Guldhammer () complemented the resistance estimate with equa-
tions for the hull–propeller interaction parameters. We will discuss these in Chapter .

31.3 Extended Resistance Estimate Example

Example input In this section, the Guldhammer and Harvald method in its latest form is illustrated
data with an example. Table . summarizes data selected for the preliminary design of
a 1000 TEU container ship. The task at hand is to determine total resistance 𝑅𝑇 and
effective power 𝑃𝐸 as a function of ship speed 𝑣.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.3 Extended Resistance Estimate Example 379

31.3.1 Completion of input parameters

As a first step we have to complete our set of input data. To perform the resistance
estimate we have to know

• computation length 𝐿,

• length–displacement ratio 𝑀 = 𝐿∕𝑉 1∕3 ,

• midship section area 𝐴𝑀 ,

• prismatic coefficient based on computation length, and

• longitudinal center of buoyancy position in the reference frame defined by Guld-

hammer and Harvald

The computation length is equal to the total length of the wetted surface 𝐿 = 𝐿𝑂𝑆 . Computation
With Equation (.) and the values from Table . we get length

𝐿 = 𝐿𝑃𝑃 + 𝐿aft + 𝐿fore = 145.0 m + 2.7 m + 3.3 m = 151.0 m

The length–displacement ratio 𝑀 for this vessel is Length–

displacement
𝐿 151.0 m ratio
𝑀 = = ( )1∕3 = 5.6716
j 𝑉 1∕3 j
18 872.0 m3

From displacement and given prismatic coefficient 𝐶𝑃 (based on 𝐿𝑃𝑃 ) follows the Midship section
midship section area. area

𝑉 18 872.0 m3
𝐴𝑀 = = = 191.88 m2
𝐿𝑃𝑃 𝐶𝑃 145.0 m ⋅ 0.6783

For the assessment of the residuary resistance coefficient and its corrections, the pris- Prismatic
matic coefficient has to be stated on the basis of the computational length 𝐿. coefficient

𝑉
𝐶𝑃 = = 0.6513 based on computational length 𝐿
𝐿 𝐴𝑀

Redefinition of the longitudinal center of buoyancy position is required only if the 𝐿𝐶𝐵 position
𝐿𝐶𝐵-correction (.) is employed. Although it is not included in the new version of
the method, an example is provided to illustrate the confusing procedure.
As explained by Andersen and Guldhammer (), midships is considered .% of 𝐿
aft of the foremost point of displacement (usually the tip of bulbous bow). Based on
Equation (.) the 𝐿𝐶𝐵 position with respect to the redefined midship section is:

𝐿𝐶𝐵 = 𝐿𝐶𝐵0 − 0.015𝐿𝑃𝑃 + 0.485𝐿aft − 0.515𝐿fore

= 0.004 ⋅ 145 m − 0.015 ⋅ 145 m + 0.485 ⋅ 2.7 m − 0.515 ⋅ 3.3 m
= −1.985 m
= −1.315 % of 𝐿 aft of G+H midship

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

380 31 Resistance Estimates – Guldhammer and Harvald’s Method

Table 31.6 Selected ship speeds and 𝑣 𝑣𝐿

resulting Froude and Reynolds 𝑣 𝑣 𝐹𝑟 = √ 𝑅𝑒 =
𝑔𝐿 𝜈
number
[kn] [m∕s] [−] [−]
15.00 7.717 0.2005 979832380
15.50 7.974 0.2072 1012493460
16.00 8.231 0.2139 1045154539
16.50 8.488 0.2205 1077815618
17.00 8.746 0.2272 1110476698
17.50 9.003 0.2339 1143137777
18.00 9.260 0.2406 1175798856
18.50 9.517 0.2473 1208459936
19.00 9.774 0.2540 1241121015

31.3.2 Range of speeds

The service speed of our vessel is 𝑣𝑆 = 17.5 kn. In our example we will perform the
estimates for speeds in the range of 𝑣 = 15 kn to  kn. In a full application, estimates
would be done for more speeds and different loading conditions if applicable.
Table . summarizes ship speed in knots and meter per second and the corresponding
Froude and Reynolds numbers. We made use of the gravitational acceleration 𝑔 =
9.807 m/s2 in SI-units. Density and kinematic viscosity for salt water at  ◦ C are
j j
taken from the ITTC Recommended Procedure .--- (ITTC, ). Density
is 𝜌 = 1026.021 kg/m3 and kinematic viscosity is 𝜈 = 1.1892 ⋅ 10−6 m2 /s. Note that,
in contrast to ITTC recommendation, the method derives Froude number 𝐹𝑟 and
Reynolds number 𝑅𝑒 based on the computation length 𝐿 rather than length in waterline
and length over wetted surface.

31.3.3 Residuary resistance coefficient

The range of speeds is well within the limit of 𝐹𝑟 < 0.33. There is no chart of our length-
displacement ratio of 𝑀 = 5.6716. If charts were to be used, a linear interpolation is
performed between the standard 𝐶𝑅std values read from the charts with the next lower
and the next higher length–displacement ratio.
( 𝐿 ) ( 𝐿 )
103 𝐶𝑅std 1∕3 =5.6716 = 10 𝐶𝑅
3
std 𝑉 1∕3
=5.5
𝑉
( )
5.6716 − 5.5 [ 3 ( 𝐿 ) ( 𝐿 )]
+ ( ) ⋅ 10 𝐶𝑅std 𝑉 1∕3 3
=6.0 − 10 𝐶𝑅
std 𝑉 1∕3
=5.5 (.)
6.0 − 5.5

𝐶𝑅std values The 𝐶𝑅std values presented in Table . have been derived with the regression for-
mula (.). The contribution 𝐻 (.) is negligibly small in this case (of order 10−8 ).

The last column in Table . lists the residuary resistance coefficients for the presumed
standard hull form. Now, the somewhat tedious process of adjusting the 𝐶𝑅std values
for the actual hull shape begins.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.3 Extended Resistance Estimate Example 381

Table 31.7 Residuary resistance value computation for the standard hull form

𝑣 𝐹𝑟 𝐸 𝐺 𝐻 𝐾 𝐶𝑅std
[kn] [−] [−] [−] [−] [−] [10−3 ]
15.00 0.2005 0.548147 0.089096 0.000000 0.024105 0.6613
15.50 0.2072 0.563610 0.104452 0.000000 0.027215 0.6953
16.00 0.2139 0.581698 0.123384 0.000000 0.030607 0.7357
16.50 0.2205 0.602898 0.146934 0.000000 0.034298 0.7841
17.00 0.2272 0.627774 0.176503 0.000000 0.038303 0.8426
17.50 0.2339 0.656977 0.213986 0.000000 0.042640 0.9136
18.00 0.2406 0.691249 0.261950 0.000000 0.047324 1.0005
18.50 0.2473 0.731443 0.323880 0.000000 0.052374 1.1077
19.00 0.2540 0.778524 0.404461 0.000000 0.057805 1.2408

. 𝐵∕𝑇 −correction: The 𝐶𝑅std values are only valid for a beam–draft ratio of 𝐵∕𝑇 = 𝐵∕𝑇 −correction
2.5. Our ship project has a shallower hull with
𝐵 24.0 m
= = 2.9268. (.)
𝑇 8.2 m
With Equation (.), the speed independent correction becomes
( )
𝐵
103 Δ𝐶𝑅(𝐵∕𝑇 ) = 0.16 − 2.5 = 0.16 (2.9268 − 2.5) = 0.0683 (.)
j 𝑇 j
. 𝐿𝐶𝐵−correction: If an 𝐿𝐶𝐵-correction is computed at all, the 𝐿𝐶𝐵-location 𝐿𝐶𝐵−correction
has to be expressed in the method’s reference frame first. We completed this above
and found that 𝐿𝐶𝐵 is −1.315% of the computation length 𝐿 aft of midships
(48.5% of 𝐿 aft or foremost point of displacement).
Computation of the correction requires monitoring both factors in Equation (.)
𝐿𝐶𝐵actual − 𝐿𝐶𝐵std (𝐹𝑟) !
Δ𝐿𝐶𝐵 = > 0
𝐿
( )
𝐹𝑟 !
𝐶1 = −1 > 0
2
−𝐶𝑃 + 1.1𝐶𝑃 − 0.0875

The correction is zero if either or both values are negative for a given Froude
number. Δ𝐿𝐶𝐵 is positive only if the actual 𝐿𝐶𝐵-position is aft of the standard
position. The optimum 𝐿𝐶𝐵-position moves aft with increasing Froude number.
Therefore, corrections are likely to occur only for higher vessel speeds. In our
example, a positive correction is applied for the two highest velocities 18.5 kn
and 19.0 kn.
Table . presents the values for the 𝐿𝐶𝐵-correction of 𝐶𝑅 for the example at
hand.
. Hull form correction: As stated in Table ., we assume standard section shapes Hull form
(𝐹𝐴 = 𝐹𝐹 = 0). As a consequence, there is no contribution from this adjustment. correction

103 𝐶𝑅(form) = 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

382 31 Resistance Estimates – Guldhammer and Harvald’s Method

Table 31.8 Computation of the 𝐿𝐶𝐵-correction for the residuary resistance coefficient

𝑣 𝐹𝑟 𝐿𝐶𝐵std 𝐿𝐶𝐵actual Δ𝐿𝐶𝐵 𝐶1 Δ𝐶𝑅(𝐿𝐶𝐵)

[kn] [−] [% of 𝐿] [% of 𝐿] [% of 𝐿] [−] [10−3 ]
15.00 0.2005 0.5768 −1.3146 −1.8914 −0.0205 0.0000
15.50 0.2072 0.2827 −1.3146 −1.5973 0.0121 0.0000
16.00 0.2139 −0.0114 −1.3146 −1.3032 0.0448 0.0000
16.50 0.2206 −0.3055 −1.3146 −1.0091 0.0774 0.0000
17.00 0.2273 −0.5996 −1.3146 −0.7149 0.1101 0.0000
17.50 0.2339 −0.8937 −1.3146 −0.4208 0.1427 0.0000
18.00 0.2406 −1.1878 −1.3146 −0.1267 0.1754 0.0000
18.50 0.2473 −1.4819 −1.3146 0.1674 0.2080 0.0313
19.00 0.2540 −1.7760 −1.3146 0.4615 0.2407 0.1000

Table 31.9 Comparison of old and updated bulbous bow correction to the residuary resistance
coefficient

old method updated method

interpolated, Table 31.3 Equations (31.25), (31.27)
𝑣 𝐹𝑟 unscaled Δ𝐶𝑅(bulb) unscaled Δ𝐶𝑅(bulb)
[kn] [−] [10−3 ] [10−3 ] [10−3 ] [10−3 ]
15.00 0.2005 0.1297 0.0947 −0.2909 −0.2122
j j
15.50 0.2072 0.1069 0.0780 −0.2917 −0.2128
16.00 0.2139 0.0713 0.0521 −0.2937 −0.2143
16.50 0.2206 0.0268 0.0195 −0.2971 −0.2168
17.00 0.2273 −0.0178 −0.0130 −0.3023 −0.2205
17.50 0.2339 −0.0623 −0.0455 −0.3094 −0.2258
18.00 0.2406 −0.1058 −0.0772 −0.3190 −0.2327
18.50 0.2473 −0.1389 −0.1014 −0.3401 −0.2482
19.00 0.2540 −0.1721 −0.1255 −0.3734 −0.2724

Bulbous bow . Bulbous bow correction: The necessary corrections Δ103 𝐶𝑅(bulb) can be found
correction by interpolating between the values given in Table . (original method) or
by employing Equation (.) (updated method, Kristensen and Lützen, ).
Table . compares the results for both methods. The values of the updated
correction are employed for this example.
The bulbous bow has a cross section area of less than 10% of the midship cross
section area. Therefore, the values derived from Equation (.) are scaled with
factor 10 𝐴𝐵𝑇 ∕𝐴𝑀 = 0.7296.
Appendage . Appendage correction: Effects of standard appendages like rudder and bilge
correction keels are already included in the 𝐶𝑅 values above. We have a single screw vessel
which usually does not feature bossings, brackets, or exposed shafts. Thus, a
correction considering appendages is not necessary here.

103 Δ𝐶𝑅(app) = 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.3 Extended Resistance Estimate Example 383

Table 31.10 Estimate of residual resistance coefficient

𝑣 𝐹𝑟 𝐶𝑅std Δ𝐶𝑅(𝐵∕𝑇 ) Δ𝐶𝑅(𝐿𝐶𝐵) Δ𝐶𝑅(form) Δ𝐶𝑅(bulb) 𝐶𝑅

[kn] [−] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ]
15.00 0.2005 0.6613 0.0683 0.0000 0.0000 −0.2122 0.5174
15.50 0.2072 0.6953 0.0683 0.0000 0.0000 −0.2128 0.5508
16.00 0.2139 0.7357 0.0683 0.0000 0.0000 −0.2143 0.5897
16.50 0.2206 0.7841 0.0683 0.0000 0.0000 −0.2168 0.6356
17.00 0.2273 0.8426 0.0683 0.0000 0.0000 −0.2205 0.6903
17.50 0.2339 0.9136 0.0683 0.0000 0.0000 −0.2258 0.7561
18.00 0.2406 1.0005 0.0683 0.0000 0.0000 −0.2327 0.8361
18.50 0.2473 1.1077 0.0683 0.0313 0.0000 −0.2482 0.9592
19.00 0.2540 1.2408 0.0683 0.1000 0.0000 −0.2724 1.1366

𝑣 𝐹𝑟 𝑣 𝑅𝑒 𝐶𝐹 𝐶𝐹′ Table 31.11 Frictional resistance estimate

[kn] [−] [m∕s] [−] [10−3 ] [10−3 ]
15.00 0.2005 7.717 979832380 1.5345 1.5526
15.50 0.2072 7.974 1012493460 1.5283 1.5463
16.00 0.2139 8.231 1045154539 1.5223 1.5402
16.50 0.2205 8.488 1077815618 1.5165 1.5344
17.00 0.2272 8.746 1110476698 1.5109 1.5288
17.50 0.2339 9.003 1143137777 1.5055 1.5233
j j
18.00 0.2406 9.260 1175798856 1.5003 1.5180
18.50 0.2473 9.517 1208459936 1.4953 1.5129
19.00 0.2540 9.774 1241121015 1.4904 1.5080

Pulling together the standard value and its corrections, we obtain the residuary resis- Corrected
tance coefficient 𝐶𝑅 for our vessel as listed in Table .. residuary
resistance

31.3.4 Frictional resistance coefficient

Based on the ITTC  model–ship correlation line (.) and the scaling factor
𝑆 + 𝑆𝐴𝑃𝑃 4400 m2 + 52 m2
= = 1.01182
𝑆 4400 m2
we obtain the friction coefficients stated in Table ..

31.3.5 Additional resistance coefficients

Minor contributions to the total resistance result from the correlation allowance, Equa- Additional
tion (.), and the air resistance 𝐶𝐴𝐴 . The displacement has to be entered in cubic resistance
contributions
meter.
( )2
103 𝐶𝐴 = 0.5 log10 (𝑉 ) − 0.1 log10 (𝑉 )
( )2
= 0.5 log10 (18 872) − 0.1 log10 (18 872.0) = 0.3096

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

384 31 Resistance Estimates – Guldhammer and Harvald’s Method

Table 31.12 Resistance coefficients computed with Guldhammer and Harvald’s method using
input from Table 31.5

𝑣 𝑣 𝐹𝑟 𝐶𝐹′ 𝐶𝑅 𝐶𝐴 𝐶𝐴𝐴 𝐶𝐴𝑆 𝐶𝑇

[kn] [m∕s] [−] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ]
15.00 7.717 0.2005 1.5526 0.5174 0.3096 0.1173 0.0000 2.4969
15.50 7.974 0.2072 1.5463 0.5508 0.3096 0.1173 0.0000 2.5240
16.00 8.231 0.2139 1.5402 0.5897 0.3096 0.1173 0.0000 2.5569
16.50 8.488 0.2206 1.5344 0.6356 0.3096 0.1173 0.0000 2.5970
17.00 8.746 0.2273 1.5288 0.6903 0.3096 0.1173 0.0000 2.6460
17.50 9.003 0.2339 1.5233 0.7561 0.3096 0.1173 0.0000 2.7064
18.00 9.260 0.2406 1.5180 0.8361 0.3096 0.1173 0.0000 2.7811
18.50 9.517 0.2473 1.5129 0.9592 0.3096 0.1173 0.0000 2.8990
19.00 9.774 0.2540 1.5080 1.1366 0.3096 0.1173 0.0000 3.0715

Kristensen and Lützen () suggest using Equation (.) for the air drag coefficient
𝐶𝐴𝐴 of container ships.
( ) ( )
103 𝐶𝐴𝐴 = max (0.28TEU−0.126 ), 0.09 = max (0.28 ⋅ 1000−0.126 ), 0.09
( )
= max 0.1173, 0.09 = 0.1173

j The steering resistance is zero 𝐶𝐴𝑆 = 0 in this example. j

31.3.6 Total resistance coefficient

Resistance Finally, we assemble all resistance parts into the total resistance coefficient 𝐶𝑇 according
estimate to Equation (.). Results are collected in Table . and shown in Figure ..

31.3.7 Total resistance and effective power

Required total resistance 𝑅𝑇 and the effective power 𝑃𝐸 at trial conditions (new hull,
calm water) follow from
1 2
𝑅𝑇 = 𝜌 𝑣 𝑆 𝐶𝑇 (.)
2
𝑃𝐸 = 𝑅𝑇 𝑣 (.)

Results are presented in Table . and Figure ..

References
Andersen, P. and Guldhammer, H. (). A computer-oriented power prediction
procedure. In Proc. of Int. Conf. on Computer Aided Design, Manufacture, and
Operation in the Marine and Offshore Industries (CADMO ’), Washington, DC,
USA.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.3 Extended Resistance Estimate Example 385

Table 31.13 Total resistance and effective power computed with Guldhammer and Harvald’s
method using input from Table 31.5

𝑣 𝐹𝑟 𝑣 𝐶𝑇 𝑅𝑇 𝑃𝐸
[kn] [−] [m∕s] [10−3 ] [kN] [kW]
15.00 0.2005 7.717 2.4969 335.62 2589.85
15.50 0.2072 7.974 2.5240 362.25 2888.54
16.00 0.2139 8.231 2.5569 391.03 3218.58
16.50 0.2206 8.488 2.5970 422.36 3585.17
17.00 0.2273 8.746 2.6460 456.82 3995.13
17.50 0.2339 9.003 2.7064 495.13 4457.51
18.00 0.2406 9.260 2.7811 538.28 4984.49
18.50 0.2473 9.517 2.8990 592.72 5641.07
19.00 0.2540 9.774 3.0715 662.39 6474.54

700 7000
3.0
[10−3 ]

600 6000

2.5
resistance coefficients CT , CF , CF′ , (CF′ + CR )

500 5000
j j

[kW]
[kN]

2.0

400 4000
total resistance RT

effective power PE
1.5
300 3000

1.0
200 2000

CT
0.5
CF 100 1000
CF′
RT
CF′ + CR PE
0.0 0 0
0.20 0.21 0.22 0.23 0.24 0.25 0.26 0.20 0.21 0.22 0.23 0.24 0.25 0.26
Froude number Fr [−] Froude number Fr [−]

Figure 31.3 Resistance coefficients for the Figure 31.4 Total resistance and effective power for
Guldhammer and Harvald method example the Guldhammer and Harvald method example

Gertler, M. (). A reanalysis of the original test data for the Taylor Standard Series.
Report , David Taylor Model Basin, Washington, DC.
Guldhammer, H. and Harvald, S. (). Ship resistance – effect of form and principal
dimensions (revised). Akademisk Forlag, Copenhagen.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

386 31 Resistance Estimates – Guldhammer and Harvald’s Method

Harvald, S. (). Resistance and propulsion of ships. John Wiley & Sons, New York,
NY.
Hollenbach, K. (). Estimating resistance and propulsion for single-screw and twin-
screw ships in the preliminary design. In Proc. of th Int. Conference on Computer
Applications in Shipbuilding (ICCAS ’).
Holtrop, J. and Mennen, G. (). An approximate power prediction method. Inter-
national Shipbuilding Progress, ():–.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
Kracht, A. (). Design of bulbous bows. SNAME Transactions, :–.
Kristensen, H. and Lützen, M. (). Prediction of resistance and propulsion power
of ships. Technical Report Project No. -, Work Package , Report No. ,
University of Southern Denmark and Technical University of Denmark.
Lewis, E., editor (). Principles of naval architecture, volume II – Resistance, propul-
sion and vibration. The Society of Naval Architects and Marine Engineers, Jersey
City, NJ, second edition.
Todd, F. (). SERIES  – Methodical experiments with models of single-screw
merchant ships. Technical Report Techn. Report , David Taylor Model Basin.

j Self Study Problems j

. Implement Guldhammer and Harvald’s method for resistance estimates and

compute the total resistance for the PANMAX container ship specified below.
Check the standard 𝐶𝑅 -value against Figures . and .. The total resistance
at design speed should be 𝑅𝑇 = 1114.96 kN and 𝑅𝑇 = 1532.00 kN at  kn if a
correction for 𝐿𝐶𝐵 is included.

Principal dimensions
length between perpendiculars 𝐿𝑃𝑃 . m
length of bulb forward of FP 𝐿fore . m
length of waterline aft of AP 𝐿aft . m
beam 𝐵 . m
draft 𝑇 . m
displacement 𝑉 . m3
container capacity TEU 
block coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝐵 .
prismatic coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝑃 .
longitudinal center of buoyancy 𝐿𝐶𝐵 . % of 𝐿𝑃𝑃 aft of 𝐿𝑃𝑃 ∕2
wetted surface 𝑆 . m2
wetted surface of bilge keels 𝑆𝐴𝑃𝑃 . m2
bulbous bow cross section area at FP 𝐴𝐵𝑇 . m2
ship design speed 𝑣 . kn
ship speeds to consider 𝑣 from  kn to  kn

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

31.3 Extended Resistance Estimate Example 387

L
= 6.0
V 1/3
6
[−]
103 CRstd

for Fr > 0.33 use original charts

residuary resistance coefficient

.80

4
CP = 0

j j

3
0.70
CP =

0.50
.60

CP =
.75

2
=0
=0

CP
CP

5
0.6
=

.55
P

=0
C

1
CP

0
0.10 0.15 0.20 0.25 0.30 0.35
Froude number Fr [−]

Figure 31.5 Charts for standard residuary resistance coefficients 𝐶𝑅std after Guldhammer and
Harvald (1974) for vessels with length-speed ratio 𝐿∕𝑉 1∕3 = 6.0. The values have been computed
and redrawn based on the regression formula provided by Andersen and Guldhammer (1986)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

388 31 Resistance Estimates – Guldhammer and Harvald’s Method

L
= 6.5
V 1/3
6
[−]
103 CRstd

for Fr > 0.33 use original charts

residuary resistance coefficient

4
0.80

j j
CP =

3
0.70
CP =

2
.50
.75

0
0.6

=0
=0

=
CP

CP
P
C

65
0.
1 =
CP .55
=0
CP

0
0.10 0.15 0.20 0.25 0.30 0.35
Froude number Fr [−]

Figure 31.6 Charts for standard residuary resistance coefficients 𝐶𝑅std after Guldhammer and
Harvald (1974) for vessels with length-speed ratio 𝐿∕𝑉 1∕3 = 6.5. The values have been computed
and redrawn based on the regression formula provided by Andersen and Guldhammer (1986)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

389

32
Introduction to Ship Propulsion

The discussion of ship hydrodynamics in this book started out with separating the ship–
propulsor system into ship hull and the propulsor itself. We discussed the flow around
the ship hull and the resultant resistance. Obviously, a propulsion force is needed that
acts in the opposite direction of the resistance in order to maintain a constant speed.
This section provides a general overview on how these systems work together, and
generates a road map for developing the necessary information on propulsor and hull–
propulsor interaction. Basic definitions of power and power ratios are provided.

Learning Objectives
j j
At the end of this chapter students will be able to:

• define power and efficiencies related to ship propulsion

• perform computations with efficiencies

• discuss advantages and disadvantages of propulsors

32.1 Propulsion Task

In our discussion so far, the ship hull miraculously moves through the water at constant Effective power
speed 𝑣 without any propulsion system (Figure .(a)). We defined the force acting
on the hull as the total resistance 𝑅𝑇 . An imaginary tow force of equal strength is
introduced to balance the external forces acting on the ship. The effective power 𝑃𝐸 is
spent towing the hull.

𝑃𝐸 = 𝑅𝑇 𝑣 effective power (hull) (.)

The imaginary tow force is ultimately replaced with the thrust generated by a propulsion Thrust and thrust
system. In Figure .(b) a marine propeller was added to the ship hull. Based on deduction
Newton’s first law one would expect that the propeller thrust is equal to the resistance
but acts in the opposite direction so that the forces are in equilibrium, i.e. 𝑇 = 𝑅𝑇 .
However, adding a propulsor changes the flow regime around the hull. In general, the

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

390 32 Introduction to Ship Propulsion

(a) Ship magically sailing at constant speed without propulsion

(b) Ship sailing at constant speed with propulsion

Figure 32.1 Forces acting on ship without and with propulsion system

drag force acting on the hull–propulsor system is larger than the total resistance 𝑅𝑇
j of the hull alone. A fraction 𝑡 of the propulsor thrust 𝑇 is spent compensating for the j
additional resistance (Figure .(b)).
𝑇 = 𝑅𝑇 + 𝑡 𝑇 (.)
The number 𝑡 is called the thrust deduction fraction and defined as the ratio of increase
in resistance and thrust.
𝑇 − 𝑅𝑇
𝑡 = (.)
𝑇
The thrust deduction fraction and other parameters quantifying the interaction between
hull and propulsor are discussed in detail in Chapter .
Wake fraction For any outside observer a propulsor is moving with speed 𝑣 of its ship. Otherwise it
would fall off! However, due to changes in the flow field caused by the moving body,
a propeller at the stern of a vessel moves with an advance velocity 𝑣𝐴 relative to the
surrounding fluid that is smaller than the ship’s speed. The difference in ship speed and
speed of advance is represented by the dimensionless wake fraction 𝑤.
𝑣 − 𝑣𝐴
𝑤 = (.)
𝑣
Again, details will be illustrated in Chapter .
Thrust power We introduce the thrust power 𝑃𝑇 as the product of thrust and speed of advance. It
represents the power transferred by a propulsor into the fluid.
𝑃𝑇 = 𝑇 𝑣𝐴 thrust power (hull–propeller system) (.)
A major objective of the work of naval architects is to design a hull–propulsor system
that accomplishes this transfer with the highest possible efficiency. The design process
will start with the selection of a suitable propulsion system.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

32.2 Propulsion Systems 391

32.2 Propulsion Systems

In this section we will discuss briefly the properties, advantages, and disadvantages
of the most common mechanical propulsion systems. For a more in depth review see
Carlton ().
Most vessels today are driven by one of these systems:

• marine propeller
• water jet propulsion
• Voith Schneider (cycloidal) propeller

32.2.1 Marine propeller

The most common propulsor is the marine propeller. It consists of a hub and 2–7
radially oriented blades. Marine propellers come in many different variations:

• fixed pitch propellers (Figure .),

• controllable pitch propellers,

j • ducted propellers, and j

• azimuthing propellers.

Propellers provide fairly high efficiency and reliable operation. Changes in draft or Advantages
motions of the vessel have only minor impact on propeller performance as long as it is
fully submerged. This is one of the reasons that propellers superseded paddle wheels so
quickly for seagoing vessels. Combined with the appropriate reduction gears, propellers
can be driven by a wide variety of engines. Typically located at the stern of the vessel,
they are protected and not easily damaged.
Their complex shape poses manufacturing challenges and makes a propeller an expen- Disadvantages
sive piece of ship machinery. Propeller and shaft make up about 7% of the cost for the
whole engine system. The principal characteristics of a propeller have a major impact
on its performance. Therefore, careful tuning of its parameters is required for a given
propulsion task.
Fixed pitch propellers (FPPs) are the most common propulsor because of their high Fixed pitch
reliability and comparatively low price. Fixed pitch refers to the fact that the blades and propellers
hub are rigidly connected. In most cases, hub and blades are cast as one continuous
piece.
The blades of controllable pitch propellers (CPPs) are not rigidly fixed to the hub Controllable
but can rotate about their radial axis. Ship speed may be changed by increasing or pitch propellers
decreasing the angle of the blades rather than changing the engine rate of revolution.
This is beneficial if the propeller shaft is also driving a generator which works best at a
constant rate of revolution. It also allows for faster changes in speed and thrust reversal
resulting in higher maneuverability of the vessel. However, best efficiency is only

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

392 32 Introduction to Ship Propulsion

Figure 32.2 A five-bladed fixed pitch propeller with a Schneekluth nozzle to improve propeller
inflow

j j
achieved at the design point of the propeller. At all other pitch angles efficiency drops
and cavitation becomes more likely. Complexity of the control mechanism increases
cost compared with a fixed pitch propeller. Controllable pitch propellers also require
larger hub diameters, reducing the efficiency. If thrust reversal is desired, expanded
area ratios are limited to values below ..
Ducted For applications where high thrust is needed at low speeds of advance, efficiency can
propellers be improved by placing the propeller in a short duct. In some applications the duct
is fixed but usually the direction (azimuth) of ducted propellers can be changed to
generate thrust in varying directions.Most active dynamic positioning systems feature
azimuthing ducted propellers.
Azimuthing Not all azimuthing propellers feature a duct. In recent years podded propulsion systems
propellers have become popular for diesel–electric propulsion. On vessels with high demand for
hotel power, like cruise ships, electric motors are used to drive the propeller. The electric
motor can be placed in an azimuthing pod providing a high level of maneuverability.

32.2.2 Water jet propulsion

Water jet propulsion might provide better efficiency as a marine propeller in high speed
applications where the propeller is likely to experience significant cavitation. The raised
internal pressure in the pump delays the onset of cavitation to higher speeds. Surface
piercing propellers might be an alternative. In essence, a water jet is a pump typically
designed to create a high mass flow (Figure .). Water is diverted from the bottom
of the vessel through the inlet to the impeller. The impeller accelerates the fluid and

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

32.2 Propulsion Systems 393

Figure 32.3 Schematic of a water jet

pushes it out at high speed above the transom. Steering nozzle and a reverse thrust
bucket can be used to direct and reverse the thrust.
Water jets eliminate appendages that stick out beneath the hull which typically come
with marine propeller installations for high speeds (e.g. exposed shafts, bossings, struts,
and rudders). Delicate machinery is protected inside the hull and can be serviced
without docking the vessel. The inlet and the grating preventing larger objects from
being sucked into the drive require careful design to minimize their impact on resistance.
j j

32.2.3 Voith Schneider propeller (VSP)

Whereas a water jet could still be considered as a type of propeller stuck into a pipe, Cycloidal
the Voith Schneider propeller has a completely different shape. It consists of a rotating propeller
disk with a set of four to five high aspect ratio blades oriented perpendicular to the
disk. The blades themselves rotate about their vertical axes. The angle of attack is
controlled by a mechanism inside the ship. Voith Schneider propellers were invented
in the s, originally conceived as hydrokinetic turbines. They are also known as
cycloidal propellers. Voith Schneider propellers combine propulsion with steering. This
makes them applicable to vessels which require high maneuverability like harbor tugs
and ferries.
Voith Schneider propellers can generate thrust in any direction, which gives the vessels
their high level of maneuverability. Typically, they are installed in pairs, which provides
a high level of redundancy. While propelling the ship, VSPs may also be employed to
increase roll damping in heavy seas and thus eliminate the need for other means of
roll stabilization. They are also suitable for dynamic positioning. Another advantage is
comparatively low noise levels.
VSPs are mechanically complex and servicing the blades can be done only while the
ship is in dry dock. The rotating disk and blades all need watertight seals. Due to the
multitude of moving parts installable, delivered power is limited to about  kW.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

394 32 Introduction to Ship Propulsion

32.3 Efficiencies in Ship Propulsion

Definition of Efficiency is commonly defined as the ratio of output (usable) power to input (total)
efficiency power:
output power usable power
efficiency = = (.)
input power total power spent
Conservation of energy demands that efficiency 𝜂 in a closed system is smaller than
𝜂 ≤ 1. When you switch on a light only a small portion of the energy flowing to the
light bulb (total energy) is converted into light (usable energy). Most of the energy is
converted into heat. We consider this a loss, although the energy is not gone. It has
been converted into anenergy form we cannot use for the intended purpose of lighting
a space. A system that has more usable power than total power spent would violate the
first and second laws of thermodynamics. Ergo, no such system exists!
Hull efficiency As we will see later, some of the efficiencies defined here take values larger than one.
This apparent contradiction is a result of comparing power output from two different
systems rather than a closed system. For instance, forming the ratio of power needed to
tow the vessel (without propulsion system) and the power the propulsor transfers into
the water yields the hull efficiency 𝜂𝐻 . Hull efficiency is the ratio of effective power and
thrust power (.):
𝑃 𝑅 𝑣
𝜂𝐻 = 𝐸 = 𝑇 (.)
𝑃𝑇 𝑇 𝑣𝐴
j Hull efficiency for single screw vessels takes values typically in the range from 1.05 to j
1.1. Of course our hull–propeller system is not a perpetual motion machine. Remember
that effective power is defined for the hull alone whereas thrust power refers to the
combined hull–propeller system. Therefore, hull efficiency is not defined in a closed
system and one may argue that the application of the term efficiency might not be the
best choice.
Delivered power Rotation of the propeller requires application of torque 𝑄 to the shaft. At the propeller
we measure delivered power 𝑃𝐷 with

𝑃𝐷 = 𝜔 𝑄 = 2𝜋 𝑛 𝑄 delivered power (.)

The angular velocity 𝜔 of the propeller is two times 𝜋 times the rate of revolution 𝑛.
Beware that if SI-units are used, rate of revolution has to be entered in 1∕s, although
engine revolutions are commonly stated in rpm (revolutions per minute).
Behind efficiency The efficiency of the propeller working behind the ship is the ratio of thrust power
(usable power) and delivered power (total power spent).
𝑃𝑇 𝑇 𝑣𝐴
𝜂𝐵 = = behind efficiency (.)
𝑃𝐷 2𝜋 𝑛 𝑄

Open water In Chapter  we discuss propeller characteristics under open water conditions, i.e. the
efficiency propeller is working in undisturbed parallel flow instead of the wake of the ship. For
this condition we define the propeller open water efficiency.
𝑃𝑇 𝑂 𝑇𝑂 𝑣𝐴
𝜂𝑂 = = open water efficiency (.)
𝑃𝐷𝑂 2𝜋 𝑛 𝑄𝑂

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

32.3 Efficiencies in Ship Propulsion 395

Figure 32.4 The propulsion system with transmission powers and efficiencies
j j
The difference from the behind efficiency is that thrust 𝑇𝑂 and torque 𝑄𝑂 refer to the
values measured in parallel flow.
The ratio of behind and open water efficiency is known as relative rotative efficiency. Relative rotative
efficiency
𝜂𝐵
𝜂𝑅 = relative rotative efficiency (.)
𝜂𝑂

This is a ratio rather than an efficiency. As will be discussed in Chapter , relative
rotative efficiency is typically a bit larger than one for single screw vessels and a bit
smaller than one for twin screw vessels.
The complete hydrodynamic system of propeller and hull uses only the effective power Quasi-propulsive
to overcome the resistance. The ratio of effective and delivered power defines the efficiency
quasi-propulsive efficiency 𝜂𝐷 (Figure .).
𝑃𝐸
𝜂𝐷 = quasi-propulsive efficiency (.)
𝑃𝐷
𝑃 𝑃
= 𝐸 𝑇 = 𝜂𝐻 𝜂𝐵 (.)
𝑃𝑇 𝑃𝐷

The quasi-propulsive efficiency represents the efficiency of the hydrodynamic system

and is a naval architect’s main responsibility.
To better understand the whole propulsion system, we follow the power to its ship Maximum
based source (Figure .). Each combustion engine, like for example a Diesel engine continuous
rating
or a gas turbine, is rated by the maximum power output it may deliver over prolonged

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

396 32 Introduction to Ship Propulsion

periods of operation without being damaged beyond normal wear and tear. The engine
manufacturer calls this the maximum continuous rating (MCR).
Brake power A motor or turbine converts the energy contained in its fuel into rotation of the engine
shaft. An engine without any load will turn faster and faster until its controls kick in
and prevent a runaway and failure of the engine. Any load (power take off ) attached to
the coupling flange brakes (slows down) the engine. If you take your bicycle from a flat
stretch into a steep climb you will slow down or maintain speed by pedaling harder. The
power necessary to maintain a constant rate of revolution of the engine is known as the
brake power 𝑃𝐵 . Brake power increases with the rate of revolution 𝑛. In their typical
range of operation brake power of large internal combustion engines is approximately
proportional to the rate of revolution cubed (𝑃𝐵 ∼ 𝑛3 ).
Shaft power and Very large diesel engines rotate slowly enough (between  and  rpm) to be con-
gearing nected directly to the propeller shaft without any gearing. Without a gearbox, shaft
efficiency
power is equal to brake power.

𝑃𝑆 = 𝑃𝐵 direct drive without gearbox (.)

Medium and small engines rotate faster than the rate of revolution for best propeller
performance. Therefore, a reduction gear is necessary to transform the engine rate of
revolution down to the optimum propeller rate of revolution. The power take off at the
gearbox becomes the new shaft power 𝑃𝑆 . The ratio of shaft power and brake power
defines the gearing efficiency.
j 𝑃𝑆 j
𝜂𝐺 = with gearbox (.)
𝑃𝐵
Losses in gears are usually smaller than 5%.
Shafting Frictional losses in the shaft bearings reduce the power that the shaft delivers to the
efficiency propeller. The ratio of shaft and delivered power provides the shafting efficiency.
𝑃𝐷
𝜂𝑆 = (.)
𝑃𝑆
In most vessels, engines are installed toward the stern, with short shafts supported by
very few bearings. As a consequence, losses are small and shafting efficiency is only
slightly smaller than one (𝜂𝑆 ≈ 0.98).
Total propulsive Power is transmitted from the engine coupling flange to the propeller through a series
efficiency of subsystems: engine ⟶ gearbox ⟶ shafting ⟶ propeller ⟶ hull. The propulsive
efficiency 𝜂𝑃 combines the efficiencies of all subsystems (Figure .).
𝑃𝐸 𝑃 𝑃 𝑃 𝑃
𝜂𝑃 = = 𝐸 𝑇 𝐷 𝑆 = 𝜂𝐻 𝜂𝐵 𝜂𝑆 𝜂𝐺 (.)
𝑃𝐵 𝑃𝑇 𝑃𝐷 𝑃𝑆 𝑃𝐵

Design of the propulsion system aims to make the propulsive efficiency as high as
possible. Gearing and shafting efficiency are topics for studies in marine engineering.
In the following chapters, we will concentrate on the efficiency 𝜂𝐷 of the hydrodynamic
system consisting of propeller and hull. We will study which properties affect the quasi-
propulsive efficiency 𝜂𝐷 and how it may be determined through model testing and how
it may be estimated for early design purposes.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

32.3 Efficiencies in Ship Propulsion 397

References
Carlton, J. (). Marine propellers and propulsion. Butterworth-Heinemann, Oxford,
United Kingdom.

Self Study Problems

. What is thrust power?
. Where do you measure brake power?
. Explain in your own words the difference between propulsive and quasi-propulsive
efficiency.
. What are the advantages and disadvantages of a Voith Schneider propeller com-
pared with a fixed pitch propeller?
. Explain in your own words the difference between open water and behind effi-
ciency.
. The quasi-propulsive efficiency is estimated to be 𝜂𝐷 = 69% at a speed of 18 kn.
If the resistance of the vessel at this speed is 𝑅𝑇 = 350 kN, what is its effective
j and delivered power? In addition, which quantity or quantities would you have j
to know to also estimate the thrust power?
. For a single screw vessel design project an engine has been selected and a pre-
liminary propulsion analysis resulted in the engine and propulsion data below.
Estimate the quasi-propulsive efficiency 𝜂𝐷 by means of the data. Assume typical
values for data not provided, making especially sure that you do not overestimate
quasi-propulsive efficiency. Also state torque, thrust power, effective power, and
resistance of the vessel.

Ship and engine data

ship design speed 𝑣𝑆 = . kn
engine brake power 𝑃𝐵 =  . kW
rate of revolution 𝑛 =  rpm
hull efficiency 𝜂𝐻 = .
propeller open water efficiency 𝜂𝑂 = .
shaft efficiency 𝜂𝑆 = .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

398

33
Momentum Theory of the Propeller

Propellers are by far the most common mechanically driven propulsors. Before we
take a closer look at their geometry and hydrodynamic performance, we apply the
fundamental principles of conservation of mass and momentum to an idealized model
of the propulsion system.
When a swimmer leaves the starting block, he generates forward momentum by pressing
the block backwards. The block does not move and provides a reaction force equal
to the force exerted by the swimmer’s feet. A propulsor presses against the water to
generate the required forward momentum for the ship. However, unlike the starting
block for the swimmer, water will not stay in place. Therefore, some of the energy
employed to create the forward momentum is lost by accelerating water in the opposite
j direction. This physical principle can be shown in the momentum theory for a propeller j
(Glauert, ). This basic mathematical model of a propulsor ignores actual geometry
and flow details, but nonetheless yields valuable insight into the workings of a propeller
and what is necessary to generate thrust with high efficiency.

Learning Objectives
At the end of this chapter students will be able to:

• apply conservation of mass and momentum principles

• understand the basic workings of a propulsor
• relate thrust loading and propulsor efficiency

33.1 Thrust, Axial Momentum, and Mass Flow

Actuator disk In the model a propulsor is represented as a thin disk of area 𝐴0 (see Figure .). The
disk is oriented perpendicular to the flow and located at position 2 between inlet 1
and outlet 3 . The coordinate system is fixed to the disk. Its positive 𝑥-axis points in
the direction of the ship’s forward movement. Fluid flows through the disk without
interruption. However, the disk ‘magically’ creates a pressure difference Δ𝑝 between its
inflow and outflow side.
Δ𝑝 = 𝑝′2 − 𝑝2 (.)

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

33.1 Thrust, Axial Momentum, and Mass Flow 399

Figure 33.1 Fixed control volume around an idealized propeller (actuator disk)

For the time being, we ignore how this is actually accomplished. The pressure 𝑝′2 on the
outflow side will be higher than the pressure 𝑝2 on the inflow side. As a result, a thrust
𝑇 acts on the disk, which is also known as an actuator disk.
j If we assume that the jump in pressure is constant across the disk area, the thrust is Propulsor thrust j
given as the product of area times pressure difference.

𝑇 = 𝐴0 Δ𝑝 (.)

Next, conservation of momentum and conservation of mass principles are applied to

the control volume, which will reveal the relationship between flow properties and
thrust.
The propeller disk area 𝐴0 and the mean inflow velocity at 1 𝑣1 = 𝑣𝐴 are given as input Input and
to the problem. A known pressure 𝑝1 acts all around the control volume. To proceed, assumptions
we make the following additional assumptions:

(i) The fluid is inviscid (ideal), incompressible, and steady, i.e. independent of time.

(ii) The race boundary is a stream surface, i.e. no fluid enters or escapes the control
volume 𝑉 through the race boundary.

(iii) The flow velocity is perpendicular to the actuator disk and is treated as constant
across 𝐴0 .

(iv) The flow is purely axial in 𝑥-direction throughout the control volume.

The assumptions (ii) and (iii) above generate something of a conflict. As mentioned Lightly loaded
in the introduction, a propulsor – represented by the actuator disk – pushes the fluid propeller
in negative 𝑥-direction. Consequently, one must expect that the flow velocity will be
higher at the outlet 3 than at the inlet 1 . In Figure . this is indicated by expressing

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

400 33 Momentum Theory of the Propeller

the velocity increase by a yet unknown fraction 𝑏.

𝑣3 = 𝑣𝐴 (1 + 𝑏) (.)

If the race boundary is a stream surface, conservation of mass requires that the race
contracts as indicated in Figure .. The cross section 𝐴3 at the outlet is smaller than
the cross section area 𝐴1 at the inlet. Due to the contraction, velocity at the outer ranges
of the actuator disk will be at an angle to the 𝑥-axis rather than normal to the actuator
disk. However, if the contraction is small, flow will be approximately normal to the
actuator disk.
Contraction will be small if the ratio of cross section areas is only slightly smaller
than one (𝐴3 ∕𝐴1 ≈ 1) or, equivalently, the increase in velocity is small (𝑏 ≪ 1). In
addition, a small 𝑏-value is essential for high propulsor efficiency, as will be shown later.
However, smaller 𝑏-values also mean less thrust if all other parameters remain the same.
A propeller that produces low thrust compared with its disk size is lightly loaded. For
now, we will assume that our propulsor satisfies this.
Application of From the conservation of momentum Equation (.) equation follows that the rate of
conservation of change of momentum is equal to the forces acting on the control volume 𝑉 .
momentum
𝜕𝑀 ∑
= 𝐹 (.)
𝜕𝑡 external

The only external force acting on the fluid in 𝑉 is the reaction 𝑇𝑓 to the propulsor
j thrust 𝑇 . ∑ j
𝐹 = 𝑇𝑓 𝑖 = −𝑇 𝑖 (.)
external

The vector 𝑖 = (1, 0, 0)𝑇 is the unit length vector in positive 𝑥-direction.
The rate of change of momentum in 𝑉 is given by
𝜕𝑀 𝜕(𝜌𝑣) ( )
= d𝑉 + 𝜌𝑣 𝑣𝑇 𝑛 d𝑆 (.)
𝜕𝑡 ∭ 𝜕𝑡 ∬
𝑉 𝑆

The local time derivative of the density is zero since the flow is steady. As a consequence,
the volume integral vanishes. Momentum changes in the control volume only by
momentum transport across its boundaries. Momentum transport only happens on
inlet and outlet disks, respectively, and we therefore may restrict the integration to the
cross sections 𝐴1 and 𝐴3 .
𝜕𝑀 ( ) ( )
= 𝜌𝑣1 𝑣𝑇1 𝑛1 d𝑆 + 𝜌𝑣3 𝑣𝑇3 𝑛3 d𝑆 (.)
𝜕𝑡 ∬ ∬
𝐴1 𝐴3

Using the normal and velocity vectors pictured in Figure . and considering that the
velocities are assumed to be constant over the cross sections, we have:
𝜕𝑀
= − 𝜌𝑣𝐴 𝑣1 d𝑆 + 𝜌𝑣𝐴 (1 + 𝑏) 𝑣3 d𝑆
𝜕𝑡 ∬ ∬
𝐴1 𝐴3

= − 𝜌𝑣𝐴 𝐴1 𝑣1 + 𝜌𝑣𝐴 (1 + 𝑏)𝐴3 𝑣3 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

33.1 Thrust, Axial Momentum, and Mass Flow 401

The quantity 𝑄1 = 𝜌𝑣𝐴 𝐴1 represents the mass flow in [kg/s] through the inlet and Application of
𝑄3 = 𝜌𝑣𝐴 (1 + 𝑏)𝐴3 the mass flow through the outlet. conservation of
mass
𝜕𝑀
= − 𝑄1 𝑣1 + 𝑄3 𝑣3 (.)
𝜕𝑡
The continuity equation requires that

𝜕𝜌
d𝑉 = − 𝜌 𝑣𝑇 𝑛 d𝑆 (.)
∭ 𝜕𝑡 ∬
𝑉 𝑆

Again, the volume integral vanishes for steady flow. Since we postulated no flow through
the race boundary, the surface integral reduces to the fact that inflow and outflow must
be equal. No fluid can be lost in between.

𝑄1 = 𝑄3 ⟶ 𝜌𝑣𝐴 𝐴1 = 𝜌𝑣𝐴 (1 + 𝑏) 𝐴3 (.)

The mass flows of inlet or outlet also apply to the actuator disk:

𝑄2 = 𝜌 𝑣𝐴 (1 + 𝑎)𝐴0 = 𝑄1 = 𝑄3 (.)

In this expression only the factor 𝑎 is unknown.

Next, the result for the mass flow Equation (.) is substituted into the rate of change Combine
j of momentum Equation (.). conservation of j
mass and
momentum
𝜕𝑀 ⎛ −𝑣𝐴 ⎞ ⎛ −𝑣𝐴 (1 + 𝑏) ⎞ ⎛ 𝑣𝐴 ⎞
= − 𝑄2 ⎜ 0 ⎟ + 𝑄2 ⎜ 0 ⎟ = − 𝑏 𝑄2 ⎜ 0 ⎟ (.)
𝜕𝑡 ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝ 0 ⎠ ⎝ 0 ⎠

Together with the result for the external force (.), we get:

𝜕𝑀
= −𝑇𝑖
𝜕𝑡
⎛ 𝑣𝐴 ⎞ ⎛1⎞
− 𝑏 𝑄2 ⎜ 0 ⎟ = − 𝑇 ⎜0⎟ ⟶ 𝑇 = 𝑏 𝑄2 𝑣𝐴 (.)
⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ ⎝0⎠

From this expression it becomes clear that a propulsor produces forward thrust (𝑇 > 0)
only if the factor 𝑏 is positive. In other words, thrust is generated only if fluid is
accelerated aft.
The Bernoulli equation provides a relationship between the velocities 𝑣2 and 𝑣3 . For Apply Bernoulli
that we apply Bernoulli’s equation between stations 1 and 2 right in front of the equation
actuator disk and from right behind the disk 2’ to the outlet 3 .

1 2 1 ( )2
1 ⟶ 2 ∶ 𝜌 𝑣𝐴 + 𝑝1 = 𝜌 𝑣𝐴 (1 + 𝑎) + 𝑝2 (.a)
2 2
1 ( ) 2 1 ( )2
2’ ⟶ 3 ∶ 𝜌 𝑣𝐴 (1 + 𝑎) + 𝑝′2 = 𝜌 𝑣𝐴 (1 + 𝑏) + 𝑝1 (.b)
2 2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

402 33 Momentum Theory of the Propeller

Note that the velocity through the actuator disk must be continuous, i.e. 𝑣2 = 𝑣′2 in
order to satisfy the conservation of mass principle.
Adding Equations (.a) and (.b) eliminates 1∕2 𝜌 (1 + 𝑎)2 𝑣𝐴2 + 𝑝1 on both sides:

1 2 1 ( )2
𝜌 𝑣 + 𝑝′2 = 𝜌 𝑣𝐴 (1 + 𝑏) + 𝑝2 (.)
2 𝐴 2
By expansion and rearrangement of the terms we find:
( )
𝑏
Δ𝑝 = 𝑝′2 − 𝑝2 = 𝜌 𝑏 𝑣𝐴2 1 + (.)
2

From Equation (.) follows with Equations (.) and (.) a relationship for the
jump in pressure and the increase of velocity in the slipstream.
[ ]
𝑇 𝑏 𝑄2 𝑣𝐴 𝑏 𝜌 𝑣𝐴 (1 + 𝑎) 𝐴0 𝑣𝐴
Δ𝑝 = = = = 𝑏 𝜌 (1 + 𝑎) 𝑣𝐴2 (.)
𝐴0 𝐴0 𝐴0
Finally, comparing the right-hand sides of Equations (.) and (.) reveals
( )
𝑏 𝑏
(1 + 𝑎) = 1 + or 𝑎 = (.)
2 2

Axially induced Some authors prefer to use an axially induced velocity 𝑢𝐴 instead of the factors 𝑎 and 𝑏.
j velocity The velocities at propeller and outlet are then given as j
( )
𝑏
𝑣2 = 1 + 𝑣 = 𝑣𝐴 + 𝑢𝐴
2 𝐴 (.)
𝑣3 = (1 + 𝑏) 𝑣𝐴 = 𝑣𝐴 + 2𝑢𝐴

Therefore 2𝑢𝐴 = 𝑏𝑣𝐴 . In some references you may also find that the axial induced
velocity has been defined as the whole axially induced velocity 𝑢𝐴 = 𝑏𝑣𝐴 .
Pressure The propulsor accelerates fluid aft. Half of the acceleration takes already place in front
distribution of the actuator disk (Figure .). As the flow velocity increases in front of the actuator
disk, pressure is decreasing to the pressure 𝑝2 :
( )
1 2 𝑏2
𝑝2 = 𝑝1 − 𝜌 𝑣𝐴 𝑏 + (.)
2 4

Surprisingly, the acceleration continues behind the disk as more potential energy (high
pressure) is converted into kinetic energy until the pressure returns to the pressure level
𝑝1 of the undisturbed flow. The propulsor raises the pressure by Δ𝑝. This eliminates the
pressure drop in front of the disk and adds
( )
1 3𝑏2
𝑝′2 − 𝑝1 = 𝜌 𝑣𝐴2 𝑏 + (.)
2 4

to the pressure level. Note that the pressure difference 𝑝1 − 𝑝2 in front of the actuator
disk is smaller than the pressure difference 𝑝′2 − 𝑝1 behind the disk. Thus, the pressure
curve shown in Figure . does not feature point symmetry with respect to the 𝑝1 level
at position 2 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

33.2 Ideal Efficiency and Thrust Loading Coefficient 403

Figure 33.2 Velocity and pressure distribution according to momentum theory

Mass flow 𝑄2 and thrust 𝑇 may now be written as functions of the advance velocity 𝑣𝐴 Mass flow and
and the factor 𝑏 which expresses the fraction by which the velocity is increased in the thrust
j slipstream. j
( )
𝑏
mass flow 𝑄2 = 𝜌 𝑣𝐴 1 + 𝐴0 (.)
2
1 ( )
thrust 𝑇 = 𝜌 𝑣𝐴2 2𝑏 + 𝑏2 𝐴0 (.)
2
Equation (.) shows that there are two ways to increase thrust 𝑇 for a given constant
speed of advance 𝑣𝐴 .

(i) increase the fluid flow velocity at the outlet, i.e. increase the factor 𝑏, or
(ii) increase the propeller disk area 𝐴0 by selecting a larger propeller diameter.

The question arises: which of the two methods yields a higher efficiency?

33.2 Ideal Efficiency and Thrust Loading Coefficient

Answering the question above requires a closer look at the energy balances in the
momentum theory model. As usual, efficiency is defined as the ratio of useful power to
total power spent. Since we neglect viscous losses, it is called ideal efficiency 𝜂id . It is
also known as jet efficiency 𝜂TJ in the ITTC’s nomenclature.
The useful work per unit time obtained from the actuator is the thrust power 𝑃𝑇 (see Useful power
Section .).
𝑃𝑇 = 𝑇 𝑣𝐴 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

404 33 Momentum Theory of the Propeller

For an outside observer the fluid is at rest outside of the propeller slipstream and the disk
is moving with the speed of advance 𝑣𝐴 relative to the observer. Since our coordinate
system is fixed to the moving disk, water is streaming toward it.
Total power The actuator disk adds kinetic energy to the flow. For rigid bodies, kinetic energy is
spent equal to one half times mass times velocity squared. In case of continua, an integration
over the control volume is required. However, since the flow is incompressible, mass
flow is the same in every cross section. Within a time interval Δ𝑡, the mass 𝑚 = 𝑄2 Δ𝑡
flows through any given cross section. Since the velocity is assumed to be constant
across each cross section, we find that a kinetic energy of

1
𝐸kin1 = 𝑄 Δ𝑡 𝑣21 (.)
2 2

is streaming into the control volume. At the outlet, kinetic energy 𝐸kin3 is leaving 𝑉 :

1
𝐸kin3 = 𝑄 Δ𝑡 𝑣23 (.)
2 2

Consequently, the total increase in kinetic energy between inlet and outlet per time
period Δ𝑡 is

j 𝐸total Δ𝐸kin 𝐸kin3 − 𝐸kin1 j

= =
Δ𝑡 Δ𝑡 Δ𝑡
[ ]
1
= 𝑄2 𝑣23 − 𝑣21
2

Substituting the definitions for velocities 𝑣1 and 𝑣3

𝐸total [ ]
1
= 𝑄2 𝑣𝐴2 (1 + 𝑏)2 − 𝑣𝐴2
Δ𝑡 2

and replacing the mass flow 𝑄2 with the right-hand side of Equation (.) yields:

𝐸total ( )
1 𝑏
𝑥 = 𝜌 𝑣𝐴 1 + 𝐴0 𝑣𝐴2 (2𝑏 + 𝑏2 )
Δ𝑡 2 2

Finally, we utilize Equation (.) and collect some factors into the thrust 𝑇 .

𝐸total ( )
𝑏
= 𝑇 𝑣𝐴 1 + (.)
Δ𝑡 2

Ideal efficiency The ideal efficiency of an actuator disk is equal to the ratio of thrust power and the total
energy transferred into the flow.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

33.2 Ideal Efficiency and Thrust Loading Coefficient 405

useful energy per unit time

𝜂𝑇𝐽 =
total energy spent per unit time
𝑃𝑇
=
𝐸total
Δ𝑡
𝑇 𝑣𝐴
= ( )
𝑏
𝑇 𝑣𝐴 1 +
2
1 2
= = (.)
𝑏 2+𝑏
1+
2
Equation (.) provides an answer to the initial question. Clearly, the denominator
of the fraction in Equation (.) is larger than the numerator for all 𝑏 > 0. Therefore,
ideal efficiency reaches its maximum value for 𝑏 = 0. Unfortunately, this would be a
useless propulsor because thrust will vanish too (check Equation (.)). For a given
thrust, the best efficiency is obtained if as much fluid as possible (high 𝑄2 ) is accelerated
aft by as little as possible (low 𝑏). Thus increasing the propeller disk area 𝐴0 results in
better efficiency than increasing 𝑏.
From this follows a general propeller design rule: for a given thrust, a larger propeller
yields better efficiency. In practice, however, propeller diameters are limited due to
j j
ever present viscous losses, which we ignored in our simple propulsor model. If the
propeller diameter is increased beyond its optimum value, expected gains in efficiency
are eaten up by increased frictional losses.
The propeller generates thrust by transferring energy into the fluid which is subsequently Axial and radial
lost to the system. It exits the slipstream. Besides adding kinetic energy in axial direction, losses
propellers also induce a rotational motion in the fluid, which causes an additional drop
in efficiency.
Another important number with respect to propeller efficiency is the thrust loading Thrust loading
coefficient. coefficient
𝑇
𝐶𝑇 ℎ = (.)
1 2𝐴
2
𝜌 𝑣𝐴 0
It expresses how much thrust is generated per unit pressure and propeller disk area.
Most ship propellers are lightly to moderately loaded with thrust loading coefficients in
the order of magnitude one. In heavy weather or bollard pull conditions, where high
thrust is generated at low speeds of advance 𝑣𝐴 , propellers are highly loaded with high
𝐶𝑇 ℎ values.
Next, we investigate the effect of thrust loading on the ideal efficiency. As a first Effect of thrust
step, we establish a relationship between thrust loading and the increase in velocity 𝑏. loading on
efficiency
Substituting the result for the thrust (.) into the definition of the thrust loading
coefficient (.) yields:
1 ( )
𝑇 2
𝜌 𝑣𝐴2 2𝑏 + 𝑏2 𝐴0
𝐶𝑇 ℎ = = = 2𝑏 + 𝑏2 (.)
1 2 1 2
2
𝜌 𝑣𝐴
𝐴0 2
𝜌 𝑣𝐴
𝐴0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

406 33 Momentum Theory of the Propeller

Figure 33.3 Ideal efficiency (jet

efficiency) of propulsor momentum
theory as a function of thrust loading
coefficient

Equation (.) may be rewritten as a quadratic equation to express the factor 𝑏 as a

function of the thrust loading coefficient.
0 = 𝑏2 + 2𝑏 − 𝐶𝑇 ℎ (.)

This equation has the solutions

j j
√
𝑏1,2 = −1 ± 1 + 𝐶𝑇 ℎ (.)
The negative solution does not provide positive thrust. Therefore we only keep case .
√
𝑏 = −1 + 1 + 𝐶𝑇 ℎ (.)
Substituting this relationship into the equation for the ideal efficiency (.) results in
2
𝜂𝑇𝐽 =
2+𝑏
2
= √ (.)
1 + 1 + 𝐶𝑇 ℎ

Figure . pictures the relationship between thrust loading coefficient and ideal effi-
ciency. The curve may be considered as the theoretical limit of attainable efficiency for
a propulsor. Values above the line are not feasible. Considering the neglected radial
and frictional losses, the actual attainable efficiency will fall short of the ideal efficiency.
For that reason any propulsion system should be sceptically scrutinized if it promises
efficiencies close or even above the ideal efficiency.

References
Glauert, H. (). The elements of airofoil and airscrew theory. Cambridge University
Press, Cambridge, United Kingdom, second edition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

33.2 Ideal Efficiency and Thrust Loading Coefficient 407

Self Study Problems

. Compute the ratio of the pressure drop 𝑝′2 − 𝑝1 (.) behind the actuator disk to
the pressure drop 𝑝1 − 𝑝2 (.) in front of the actuator disk as a function of the
factor 𝑏 which represents the increase in fluid velocity far behind the propulsor.
. Compute the slipstream contraction ratio 𝐴3 ∕𝐴1 as a function of propeller disk
area 𝐴0 and the ideal efficiency 𝜂TJ .

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

408

34
Hull–Propeller Interaction

Based on our discussions of the fluid flow around a ship hull, it should be obvious
that the hull affects the inflow to the propeller. A propeller, in turn, affects how fluid
flows around the hull. It almost always increases the drag of the hull–propulsor system
beyond the bare hull resistance. These mutual effects are commonly referred to as
hull–propeller interaction.
Effects of a hull on the propeller are represented by wake fraction 𝑤 and relative rotative
efficiency 𝜂𝑅 . The former expresses changes in inflow velocity and the latter the change
of propeller efficiency in comparison to the open water condition. The effect of a
propeller on the hull is captured in the thrust deduction fraction 𝑡.
Separating the mutual effects is an imperfect model of the real word. It helps to explain
j and, in part, quantify the effects, but keep in mind that we are dealing with a single, j
integrated system.

Learning Objectives
At the end of this chapter students will be able to:

• explain hull–propeller interaction phenomena

• define interaction parameters and their effects

• apply hull–propeller interaction parameters

34.1 Wake Fraction

Effect of hull on The presence of the ship hull causes non-uniform inflow to a propeller behind the ship.
propeller It will significantly differ from the open water condition with its undisturbed parallel
inflow. The propeller is now working in the wake of the hull. As a consequence, the
speed of advance 𝑣𝐴 of a propeller will differ from the ship speed 𝑣. At first glance this
seems impossible. For an outside observer, both hull and propeller move with the same
speed and the propeller, hopefully, stays attached to the ship. So how can its speed differ
from that of the ship? The speed of advance 𝑣𝐴 is the velocity of a propeller relative to
the surrounding water. At the stern of a vessel, the fluid flow slows down relative to the

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.1 Wake Fraction 409

hull, in part to rebuild pressure and in part due to viscous losses in the boundary layer
which is thickest at the stern.
The difference in speed of advance and ship speed is expressed in the wake fraction 𝑤. Wake fraction

wake speed 𝑣 − 𝑣𝐴
wake fraction: 𝑤 = = (.)
ship speed 𝑣

This definition is attributed to Admiral David W. Taylor. The difference between ship
and propeller speed is also referred to as propeller slip. Later we will learn how to
estimate or experimentally determine the wake fraction.
Once the wake fraction 𝑤 is known, the speed of advance 𝑣𝐴 can be computed for the Speed of
propeller. advance

𝑣𝐴 = (1 − 𝑤) 𝑣 (.)
For most ships the wake fraction is positive, which means speed of advance of the
propeller is smaller than ship speed. For example, if 𝑤 = 0.28 and the ship speed is
20 kn, the speed of advance becomes

𝑣𝐴 = (1 − 𝑤) 𝑣 = (1 − 0.28) 𝑣 = 0.72 𝑣

For the speed of advance a representation in meter per second is more appropriate,
since it is used in several equations later on.
j sm 1852 m m j
𝑣 = 20 kn = 20 = 20 = 10.289 ship speed
h 3600 s s

Thus the speed of advance in the example is

m m
𝑣𝐴 = 0.72 ⋅ 10.289 = 7.408 propeller speed of advance
s s

The wake fraction for a ship is an average of inflow speed over the propeller disk. Local wake
fraction and
2𝜋 𝑅 wake field
1 1
𝑤 = ( ) 𝑤(𝑦, 𝑧) d𝐴 = ( ) 𝑤(𝑟, 𝜑) 𝑟 d𝑟 d𝜑 (.)
𝐴0 − 𝜋𝑟2ℎ ∬ 𝐴0 − 𝜋𝑟2ℎ ∫ ∫
𝐴0 ∖hub 0 𝑟ℎ

Figure . shows isolines of local wake fraction as it would be typically found for a
slender single screw vessel with V-type frames in the run. The local wake fraction is
based on the local axial inflow 𝑣𝐴 (𝑦, 𝑧) into the propeller plane:

𝑣 − 𝑣𝐴 (𝑦, 𝑧)
𝑤(𝑦, 𝑧) = (.)
𝑣
 Some old publications use a definition by Froude which divides the wake speed (𝑣 − 𝑣 ) with the speed
𝐴
of advance instead of ship speed 𝑣.
𝑣 − 𝑣𝐴
𝑤Froude = (this is no longer used)
𝑣𝐴

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

410 34 Hull–Propeller Interaction

0.6

0.600

0.600
0.4

0.400

0.500
0.2

0.500
[]

hub
0.0

0
10
0.
z = z/D

0.300

0.500
0.600
0.2

00
0.2
propeller disk
0.4

w(y, z)=0.3192 for point (y, z)=(0.22,0.10)

0.6 isolines of local wake fraction w(y, z) w(y, z)=0.1193 for point (y, z)=(0.25,0.25)

0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8

y = y/D []
j j
Figure 34.1 Example of a nominal wake field of a single propeller ship with moderate block
coefficient

For example, at point (𝑦, 𝑧) = (0.22, 0.1), marked with a dot in Figure ., the local
wake fraction is 𝑤(0.22, 0.1) = 0.3192. Therefore, the difference between local axial
inflow velocity and a ship speed of 𝑣 = 9 m/s at point (0.22, 0.1) is

𝑣 − 𝑣𝐴 (0.22, 0.1) = 𝑣 𝑤(0.22, 0.1) = 9 m/s ⋅ 0.3192 = 2.873 m/s

Consequently, the local axial inflow speed is

[ ]
𝑣𝐴 (0.22, 0.1) = 𝑣 1 − 𝑤(0.22, 0.1) = 9 m/s − 2.873 m/s = 6.127 m/s

Wake hook Figure . shows a wake fraction distribution typical for a full bodied vessel with
U-type sections in the run. The wake field in each half of the propeller disk is formed
like a hook. This feature is caused by a pronounced vortex emanating from the turn of
the bilge. Vessels with high block coefficient tend to have a higher wake fraction, but
parameters like beam–draft ratio 𝐵∕𝑇 , mid ship section coefficient 𝐶𝑀 , shape of the
run, and surface roughness influence the wake distribution as well.
Nominal and Data for plots like Figure . and . stems from flow speed measurements or flow
effective wake speed calculations in the propeller plane. Note that a propeller is not present and its
influence on the flow is not accounted for. The wake field determined in the absence
of a propeller is called nominal wake. Measurements of the wake field are difficult in
the plane of a working propeller and also difficult to interpret because the wake field is

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.1 Wake Fraction 411

0.6

0.400
0.4

0.500

0.100
0.300
0.2

0.500

0.200
[]

hub
0.0
z = z/D

0.6
00

0.2 00
0.5 0.7
00
propeller disk
0.4

w(y, z)=0.6760 for point (y, z)=(0.28,0.05)

0.6 isolines of local wake fraction w(y, z) w(y, z)=0.4277 for point (y, z)=(0.15,0.30)

0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8

y = y/D []
j j
Figure 34.2 Example of a nominal wake field of a single propeller ship with high block
coefficient

locally time dependent. A wake field that includes the influence of a working propeller
is called effective wake.
For design and analysis purposes, it is useful to study contributions to wake or wake Phenomena
fraction individually. Once the dominant influences are identified, procedures and influencing wake
measures may be conceived for wake prediction and control. Remind yourself that
wake is an influence of the hull on the flow pattern in the vicinity of the propeller and
that the wake fraction states how much the flow has slowed down compared to the ship
speed. In our recapture of basic flow phenomena, we came across three flow patterns
that cause local changes of flow speed:

• boundary layers on walls

• potential flow around bodies
• particle movement in waves

Friction between water and ship hull causes fluid to slow down within the boundary
layer. Even in an inviscid flow around a body, molecules slow down toward the aft
stagnation point, converting kinetic into potential energy. Underneath a wave crest,
particles move in the direction of wave propagation, whereas underneath a wave trough
they move in the opposite direction.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

412 34 Hull–Propeller Interaction

j j

(a) Origin of potential wake fraction 𝑤𝑝 (b) Origin of frictional wake fraction 𝑤𝑓

Figure 34.3 Major contributions to the nominal wake fraction

Theoretical wake The phenomena give rise to a simple three part model of the wake fraction (Harvald,
model ).
𝑤 = 𝑤𝑝 + 𝑤𝑓 + 𝑤𝑤 (.)

The total wake fraction is split into a potential wake fraction 𝑤𝑝 , a frictional wake
fraction 𝑤𝑓 , and a wave wake fraction 𝑤𝑤 . Note that the model (.) is not based
on a conclusive theory. For starters, none of the three parts can be truly measured
individually and the additive model expressed by Equation (.) suggests that the parts
have no influence on each other, which is definitely not true. Nevertheless, the model
will prove helpful in explaining what influences the wake and how to deal with scale
effects in model tests.
Potential wake In our discussion of potential flow around a cylinder, we learned that stagnation points
fraction 𝑤𝑝 exist at the front and back of a body. Thus, a propeller at the stern of a vessel operates
close to the virtual stagnation point in a region with small inflow velocity (Figure .(a)).
This is also applicable to viscous flows. For most vessels, the exterior flow will not come
to a complete stop but will slow down nonetheless. This conversion of kinetic energy
into potential energy causes the potential wake which is captured as the wake fraction
component 𝑤𝑝 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.1 Wake Fraction 413

Figure 34.4 Frictional wake for twin screw vessels

The frictional portion of the wake fraction is attributed to the momentum loss caused by Frictional wake
viscous forces within a boundary layer. The propeller of single screw vessels is completely fraction 𝑤𝑓
immersed in the boundary layer of the hull and experiences its decelerated flow over
the entire disk. Note that the propeller is drawn with dashed lines in Figure .(b)
to indicate that we are studying the nominal wake, i.e. the influence of the hull on a
propeller without flow effects caused by the working propeller.
The wave wake fraction 𝑤𝑤 is negligibly small for most vessels. The idea is that, de- Wave wake
pending on the propeller location relative to the ship wave pattern, the water particle fraction 𝑤𝑤
velocities within the stern wave add to or subtract from the wake. Under the crest of a
stern, wave particles move in the direction of the ship, which reduces the local speed of
advance. A slower inflow causes a higher wake fraction. Under a wave trough, particles
move against the motion of the ship and slightly increase the velocity of advance (lower
wake fraction).
j j
In twin screw arrangements, propellers are only partially immersed in the hull’s bound- Wake of twin
ary layer (Figure .). As a consequence, wake fractions of twin screw vessels tend to screw vessels
be smaller than for single screw vessels.
Potential and frictional part are the major contributions to the wake fraction. Chapter  Model and full
discusses how the wake fraction is determined through model tests. Since the frictional scale wake
fraction
part of the wake fraction is of considerable size, and the Reynolds number is a lot smaller
in the model test than for the full scale vessel, wake fractions determined from model
test data have to be corrected before they can be applied to the ship.
Typical wake fraction values for single screw vessels are in the range from 0.2 to 0.35.
Ships with high block coefficients or short runs tend to have higher values toward the
upper end of the range, slender vessels have lower wake fractions.
The resistance estimate by Guldhammer and Harvald (see Chapter ) has been ex- Wake fraction
tended by Andersen and Guldhammer () to provide estimates for the hull–propeller estimate
interaction parameters. For the wake fraction 𝑤 of single screw vessels, Andersen and
Guldhammer () suggest the following method:

(i) First, compute the parameters 𝑎, 𝑏, and 𝑐:

𝐵
𝑎 = 0.1 + 0.149
𝐿
𝐵
𝑏 = 0.05 + 0.449 (.)
𝐿
( )2
𝐵 𝐵
𝑐 = 585 − 5027 + 11700
𝐿 𝐿
Again, 𝐿 stands for the computation length introduced in Section ...

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

414 34 Hull–Propeller Interaction

(ii) The wake fraction is split into three parts: 𝑤1 captures the influence of the hull
form, 𝑤2 represents a correction for the stern shape, and 𝑤3 marks the influence
of the propeller.
𝑏
𝑤1 = 𝑎 +
𝑐(0.98 − 𝐶𝐵 )3 + 1
0.025𝐹𝐴
𝑤2 = (.)
100(𝐶𝐵 − 0.7)2 + 1
0.00756
𝑤3 = −0.18 + ( ) however, 𝑤3 ≤ 0.1
𝐷
𝐿
+ 0.002

The parameter 𝐹𝐴 is the shape factor for the aft body. 𝐷 stands for the propeller
diameter. If the computed value for 𝑤3 exceeds . it is set to ..
(iii) The estimate for the model scale wake fraction is the sum of 𝑤1 , 𝑤2 , and 𝑤3 .
% of the model scale value is assigned to the full scale wake fraction for trial
conditions.

𝑤𝑀 = 𝑤1 + 𝑤2 + 𝑤3 model (.)
𝑤𝑆 = 0.7𝑤𝑀 full scale, trial condition (.)

For in service conditions, Andersen and Guldhammer () suggest using the
j model scale value for the full scale ship. This has to be coordinated with the j
selected service margin.

Based on comparisons with model test data for contemporary bulk carrier and tanker
hull designs, Kristensen and Lützen () suggest to correct the value obtained from
Equation (.) based on the length–displacement ratio 𝑀:

𝑤𝑆corrected = 0.7𝑤𝑀 − 0.45 + 0.08𝑀 bulk carriers and tankers (.)

Other methods for wake fraction estimates are presented in Chapters  and .

34.2 Thrust Deduction Fraction

Thrust deduction A working propeller affects the flow around a ship hull, especially in the stern region.
The propeller causes an additional resistance which we expressed as a fraction of thrust
(see Equation (.)). Since part of the propeller thrust is used to compensate for the
increase in resistance, we called the effect thrust deduction.
Let us look into the mechanisms which cause the increase in resistance. As we discussed
in Chapter  on momentum theory, propellers accelerate the fluid aft even before
it reaches the propeller. Changes in flow velocity over the hull surface will affect the
boundary layer development and will change pressure and shear stress distribution. As
a consequence, the drag force on the hull will differ from the case without a propeller.
In addition to changes in hull drag, we have to consider the propeller’s own drag.
Experience shows that the resistance of a hull with propeller is larger than the resistance

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.2 Thrust Deduction Fraction 415

j j

Figure 34.5 Effect of propeller on pressure and velocity distribution at the stern

of a hull alone. Therefore, thrust 𝑇 needs to be larger than the calm water total resistance
𝑅𝑇 .
The effect is quantified by the thrust deduction 𝑇 − 𝑅𝑇 and a corresponding thrust Thrust deduction
deduction fraction 𝑡: fraction

thrust deduction 𝑇 − 𝑅𝑇
𝑡 = = (.)
thrust 𝑇
Like the wake fraction, the thrust deduction fraction is subdivided into three parts for
a discussion of influencing factors.

𝑡 = 𝑡𝑝 + 𝑡𝑓 + 𝑡𝑤 (.)

The potential part of the thrust deduction fraction 𝑡𝑝 is by far the largest contribution to Potential thrust
𝑡. From momentum theory we learned that the propeller is accelerating water through deduction
the propeller disk. The increasing velocity causes a drop in pressure compared with the
flow around the bare hull without propeller (see Figure .). Lower pressure at the
stern results in additional resistance which has to be compensated by thrust larger than
the total resistance 𝑅𝑇 of the hull alone.
The frictional part 𝑡𝑓 may be attributed to the increased flow velocities in the stern Frictional thrust
region (Figure .). Higher velocities will result in higher wall shear stress values. Since deduction

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

416 34 Hull–Propeller Interaction

efficient propellers keep the increase in velocity small, the frictional part of the thrust
deduction fraction is negligibly small for propellers with moderate thrust loading.
Wave thrust The wave portion 𝑡𝑤 of the thrust deduction fraction captures the small effect that
deduction the propeller has on the wave resistance. The working propeller changes the pressure
distribution in its vicinity which in turn will affect the contribution of the stern wave
to the wave pattern. The effect on the wave resistance may be positive or negative. In
general this contribution is also negligible.
Scale effects are Reynolds number effects are very small in the thrust deduction fraction. Therefore, the
negligible value determined in model tests is applied to the full scale ship without corrections.
Thrust deduction fraction values are typically lower than the wake fraction. As a rough
estimate 𝑡 = 0.7𝑤 may be used. Hollenbach () uses a fixed value of 𝑡 = 0.19 for
single screw vessels and 𝑡 = 0.15 for a standard twin screw arrangement with two
rudders.
Thrust deduction For the thrust deduction fraction 𝑡 of single screw vessels Andersen and Guldhammer
estimate () suggest the following procedure:

(i) Compute three parameters 𝑑, 𝑒, and 𝑓 :

𝐵
𝑑 = 0.625 + 0.08
𝐿
𝐵
𝑒 = 0.165 − 0.25 (.)
𝐿
j 𝐵
( )2
𝐵 j
𝑓 = 825 − 8060 + 20300
𝐿 𝐿
𝐿 represents the computation length introduced in Section ...
(ii) The thrust deduction is split into three parts: 𝑡1 captures the influence of the hull
form, 𝑡2 represents a correction for the stern shape, and 𝑡3 marks the influence of
the propeller.
𝑒
𝑡1 = 𝑑 +
𝑓 (0.98 − 𝐶𝐵 )3 + 1
𝑡2 = −0.01𝐹𝐴 (.)
( )
𝐷
𝑡3 = 2 − 0.04
𝐿
The parameter 𝐹𝐴 is the shape factor for the aftbody. 𝐷 is the propeller diameter.
(iii) The estimate for the thrust deduction fraction is the sum of 𝑡1 , 𝑡2 , and 𝑡3 (Andersen
and Guldhammer, ):
𝑡 = 𝑡1 + 𝑡2 + 𝑡3 (.)

Similar to the correction for the wake fraction, Kristensen and Lützen () suggest
correcting the thrust deduction fraction obtained from Equation (.) for bulk carriers
and tankers.
𝑡corrected = 𝑡 − 0.26 + 0.04𝑀 bulk carriers and tankers (.)
Since the influence of viscosity on the thrust deduction fraction is small, we assume the
same value for model and ship.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.3 Relative Rotative Efficiency 417

34.3 Relative Rotative Efficiency

During open water tests, thrust 𝑇𝑂 and torque 𝑄𝑂 are determined as a function of Open water
the advance coefficient 𝐽 . The subscript 𝑂 indicates the open water condition with condition
uniform parallel inflow into the propeller. The ratio of thrust power and delivered power
determines the open water efficiency 𝜂𝑂 .

𝑃𝑇 𝑇𝑂 𝑣𝐴
𝜂𝑂 = = (.)
𝑃𝐷 2𝜋 𝑛 𝑄𝑂

It expresses how effectively the torque 𝑄𝑂 delivered via the shaft to the propeller is
converted into thrust 𝑇𝑂 . Increasing thrust requires increased torque.
As mentioned, a propeller in the behind condition is affected by the hull wake. Provided Behind condition
we use the same advance coefficient 𝐽 , the ratio of average speed of advance 𝑣𝐴 and
rate of revolution 𝑛 will be the same for open water and behind condition. However,
generated thrust and required torque will probably differ in open water and behind
condition. This also affects the efficiency of the propeller. We defined the behind
efficiency 𝜂𝐵 as
𝑃 𝑇 𝑣𝐴
𝜂𝐵 = 𝑇 = (.)
𝑃𝐷 2𝜋 𝑛 𝑄
This is seemingly the same as the open water efficiency. However, the flow conditions
are vastly different: uniform (open water) versus nonuniform flow (behind condition).
j j
This scenario has many influencing factors. To simplify matters we introduce the Thrust identity
assumption of thrust identity: we postulate that the propeller produces the same thrust
𝑇 = 𝑇𝑂 in open water and behind condition for the same advance coefficient 𝐽 . Only
the torque differs 𝑄 ≠ 𝑄𝑂 . We will discuss this further in Chapter .
Under the condition of thrust identity, the ratio of behind and open water efficiency Relative rotative
yields: efficiency

𝑇 𝑣𝐴 |
|
|
𝜂𝐵 2𝜋 𝑛 𝑄 | 𝑄
𝜂𝑅 = = | = 𝑂 assuming thrust identity (.)
𝜂𝑂 𝑇𝑂 𝑣𝐴 | 𝑄
|
|
2𝜋 𝑛 𝑄𝑂 |𝑇 =𝑇𝑂

The ratio 𝜂𝑅 is known as relative rotative efficiency. It expresses the change in propeller
efficiency from open water (uniform flow) to behind condition (nonuniform flow).
Surprisingly, relative rotative efficiency tends to be slightly larger than one for single
screw vessels (𝜂𝑅 ∈ [1.0, 1.05]), i.e. the propeller is more efficient in the nonuniform
wake than in the uniform inflow. For twin screw vessels, relative rotative efficiency is
usually smaller than one (𝜂𝑅 ∈ [0.96, 1.0]). The difference in 𝜂𝑅 values between single
and twin screw vessels is attributed to the fact that propellers of twin screw vessels are
less affected by the hull wake. A portion of the propeller disks usually falls outside of
the boundary layer of the hull.
The definition of the relative rotative efficiency provided with Equation (.) is not Torque identity
unique. We could also define relative rotative efficiency under the assumption of torque
identity: at the same advance coefficient 𝐽 the same torque in behind and open water

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

418 34 Hull–Propeller Interaction

condition 𝑄 = 𝑄𝑂 results in different thrust 𝑇 ≠ 𝑇𝑂 .

𝑇 𝑣𝐴 |
|
|
𝜂𝐵 2𝜋 𝑛 𝑄 | 𝑇
𝜂𝑅 = = | = assuming torque identity (.)
𝜂𝑂 𝑇𝑂 𝑣𝐴 | 𝑇
| 𝑂
|
2𝜋 𝑛 𝑄𝑂 |𝑄=𝑄𝑂

Thrust identity is Torque identity has the advantage that performance predictions are more easily verified
preferred in ship trials. Torque can be measured on board a ship without great effort, thrust less
so. In general, however, thrust identity seems to produce more accurate performance
predictions. Therefore, the ITTC recommends the use of thrust identity.
Estimates In ship design we often know the open water efficiency of a propeller from series data or
propeller design calculations like lifting line and lifting surface computations. Prediction
of the vessel performance requires the efficiency for the behind condition. Consequently,
estimates for the relative rotative efficiency are necessary, which enables us to convert
the measured or computed open water efficiency into the behind efficiency:

𝜂𝐵 = 𝜂𝑂 𝜂𝑅 (.)

As a first conservative estimate 𝜂𝑅 = 1 may be used for single screw vessels and
𝜂𝑅 = 0.96 for twin screw vessels. Hollenbach () suggests the following values:

j Relative rotative efficiency 𝜂𝐑 j

single screw vessel at design draft 1.009
single screw vessel in ballast condition 1.000
twin screw vessel at design draft 0.981

See also Chapters  and  on the performance prediction methods by Holtrop ()
and Hollenbach (). Chapter  explains how the hull–propeller interaction param-
eters are determined during model tests.

References
Andersen, P. and Guldhammer, H. (). A computer-oriented power prediction
procedure. In Proc. of Int. Conf. on Computer Aided Design, Manufacture, and
Operation in the Marine and Offshore Industries (CADMO ’), Washington, DC,
USA.
Harvald, S. (). Resistance and propulsion of ships. John Wiley & Sons, New York,
NY.
Hollenbach, K. (). Verfahren zur Abschätzung von Widerstand und Propulsion
von Ein- und Zweischraubenschiffen im Vorentwurf. In Jahrbuch der Schiffbautech-
nischen Gesellschaft, volume , pages –. Schiffbautechnische Gesellschaft
(STG).
Hollenbach, K. (). Estimating resistance and propulsion for single-screw and
twin-screw ships. Schiffstechnik/Ship Technology Research, ():–.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

34.3 Relative Rotative Efficiency 419

Holtrop, J. (). A statistical re-analysis of resistance and propulsion data. Interna-

tional Shipbuilding Progress, ():–.
Kristensen, H. and Lützen, M. (). Prediction of resistance and propulsion power
of ships. Technical Report Project No. -, Work Package , Report No. ,
University of Southern Denmark and Technical University of Denmark.

Self Study Problems

. Name and explain the three hull–propeller interaction parameters.
. Discuss the major contributions of the wake fraction.
. Discuss the effect the propeller has on the hull.
. What is the relative rotative efficiency?
. Explain the nature of the potential part of the thrust deduction fraction.
. Given a ship speed of 𝑣𝑆 = 18.2 kn and a wake fraction of 𝑤 = 0.242, compute
the speed of advance of the propeller.
. A ship is sailing at a speed of 𝑣𝑆 = 18.2 kn. Resistance and propulsion estimates
j yield the following characteristics for the service speed: j

Ship and engine data

total resistance 𝑅𝑇 = . kN
wake fraction 𝑤 = .
thrust deduction fraction 𝑡 = .
open water efficiency 𝜂𝑂 = .
rate of revolution 𝑛 = . rpm
propeller diameter 𝐷 = . m

Compute the effective power, calm water thrust, speed of advance, advance
coefficient, hull efficiency, and perform an estimate of quasi-propulsive efficiency,
delivered power, and propeller torque. Assume conservative values for data not
provided.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

420

35
Propeller Geometry

Before we discuss the hydrodynamic characteristics of a propeller, it is necessary to

define and explain its principal geometric characteristics: number of blades, diameter,
pitch–diameter ratio, and expanded area ratio, as well as rake, skew-back, and other
properties. The general impact of main propeller characteristics on performance is
discussed.

Learning Objectives
At the end of this chapter students will be able to:
j j
• identify and describe propeller parts

• define principal geometric characteristics of propellers

• understand impact of geometric characteristics on performance

35.1 Propeller Parts

Fixed versus A propeller consists of a hub and two or more blades. Figure . shows two views
controllable of a fixed pitch propeller with four blades. The term ‘pitch’ will be explained shortly.
pitch
The adjective ‘fixed’ refers to the fact that the blades are cast together with the hub as
a single piece. The position of the blades is immutable. Controllable pitch propellers
have blades which can be rotated around a radial axis. About one third of installed
propellers on commercial vessels with more than  horse power are controllable
pitch propellers (Carlton, ). They are commonly used on ships which require high
maneuverability like ferries, tugs, or supply vessels. Fixed pitch propellers are cheaper
and due to the smaller hub often achieve a higher efficiency at their design point.
Hub The hub provides a connection to the propeller shaft. Propellers on pleasure boats
might have a key. On larger vessels the hub has a conical bore which is pressed on the
conical end of the propeller shaft and secured with bolts. The hub diameter is typically
about one fifth of the total propeller diameter. Controllable pitch propellers have hub
diameters of about 25% of the propeller diameter.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.1 Propeller Parts 421

(a) Looking forward (b) Looking aft

Figure 35.1 Parts of a propeller. Shown is a right-handed, fixed pitch propeller with four blades

j j
Like the wings of an aircraft, propeller blades generate a ‘lift’ force. A portion of the lift Blades
force points in direction of the propeller shaft and contributes to the propeller thrust.
The connection of a propeller blade to the hub is called the blade root (Figure .).
Structurally, blades are cantilevered beams. Stress concentrations at the blade root
are avoided by casting fillets. The outer, free end of the blade is the blade tip. Looking
forward, we see the blade faces (Figure .(a)). This is the high pressure side of the
blade, equivalent to the lower side of an aircraft wing. Looking aft, we see the blade
backs (Figure .(b)), which represent the low pressure (suction) side of the blade,
equivalent to the upper side of an aircraft wing. Note that blade faces are oriented aft
relative to the ship and blade backs are oriented forward.
The edge of a blade pointing in the direction of rotation is called the leading edge. By Leading and
convention, single screw vessels operate with right turning (right-handed) propellers, trailing edge
i.e. looking forward along the shaft they turn clockwise when they produce forward
thrust. Therefore, the leading edge is found on the right side of the blade (Figure .(a)).
The downstream edge of the blade is the trailing edge.
Single propellers are always right-handed, i.e. they rotate clockwise when locking Left- and
forward. Due to the clockwise rotation of the propeller, single screw vessels tend to right-handed
steer to port if left uncontrolled. This is known as the wheel effect. Imagine the propeller
being a rolling wheel which shifts the stern to starboard; the vessel will steer to port.
Twin screw vessels use one right-handed (starboard) and one left-handed propeller
(port) to compensate for the wheel effect.
Blades are always arranged in a regular pattern because the propeller needs to be Dynamic balance
dynamically balanced. Any imbalance due to damage in grounding, collision, or high
seas will cause subsequent damage to shaft, bearings, and possibly the engine.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

422 35 Propeller Geometry

35.2 Principal Propeller Characteristics

Principal Ship power prediction in early design requires at a minimum the selection of four
characteristics geometric and one basic operational characteristic of the propeller:

• number of blades 𝑍,
• diameter 𝐷,
• pitch–diameter ratio 𝑃 ∕𝐷,
• expanded area ratio 𝐴𝐸 ∕𝐴0 , and
• rate of revolution 𝑛

The overall objective of matching propeller and ship is to maximize the propulsive
efficiency of the vessel.
The four geometric parameters are explained below. The rate of revolution 𝑛 should be
self-explanatory. We will come back to it when we discuss propeller selection. As a first
step in the design process, principal parameters are selected on the basis of published
propeller series data and later confirmed and refined by lifting line and lifting surface
calculations. We will discuss the selection of propeller characteristics based on series
data in later chapters. For a discussions of lifting line and lifting surface methods the
reader may consult Kerwin and Hadler () and Breslin and Andersen ().
j j
Number of The number of blades ranges from two to seven. Cost of a propeller increases with the
blades 𝑍 number of blades. Therefore, lower blade numbers are preferable. Auxiliary drives for
sailing yachts might feature only two blades. Power boats and small vessels typically
use threebladed propellers. Most propellers for larger merchant vessels have four or
five blades. More than five blades are used only when it is necessary to avoid cavitation
or to reduce vibrations and noise signatures. An important factor in the selection of the
number of blades 𝑍 is the avoidance of resonances with the main engine. Propellers
develop cyclical forces with frequencies that are multiples of 𝑍 𝑛. These frequencies
should not match up with natural frequencies of the main engine. For example, a four
bladed propeller should not be used in conjunction with an engine which has an even
number of cylinders.
Diameter 𝐷 The diameter 𝐷 of a propeller is defined by the outermost circle which circumscribes
the blade tips (Figure .). The enclosed area is the propeller disk area 𝐴0 .
𝜋 𝐷2
𝐴0 = (.)
4
The distance between shaft center and tip of a blade is the blade radius 𝑅. It is equal to
half the diameter and is equivalent to the span of an aircraft wing. In many propeller
drawings and calculation methods a dimensionless radial coordinate 𝑥 is used with
𝑟
𝑥 = with 𝑥 ∈ [0, 1] (.)
𝑅
For conical or streamlined hubs, the hub radius 𝑟ℎ is usually measured where the blade
reference line intersects the hub. This point also defines the longitudinal location of the
propeller plane (see also Figure .).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.2 Principal Propeller Characteristics 423

Figure 35.2 Definition of propeller diameter 𝐷, blade radius 𝑅, hub radius 𝑟ℎ , and disk area 𝐴0

j The diameter has a major impact on propeller performance. As we learned in the axial j
momentum theory of propellers, a larger diameter is usually more efficient. However,
viscous losses and the wake field of the hull usually define an optimum value for the
diameter which has to be found during the design process.
The complicated twisted shape of propeller blades is due to the pitch 𝑃 . Every point Pitch 𝑃
on a propeller travels on a helical curve controlled by the rate of revolution 𝑛 and the
forward movement of the propeller. The geometric pitch 𝑃 is the distance which a
point on the propeller travels in direction of the shaft during a single rotation. This
warrants further explanation.
First, we take a look at the foil sections that make up a propeller blade. On an aircraft Blade sections
wing or a rudder we commonly find the foil section by cutting straight through the
wing/rudder perpendicular to its span. Finding the foil section of a propeller blade is
a bit more complicated. Take a look at Figure .. It shows a four bladed propeller
whose top blade is intersected with a cylinder (shaded surface) of radius 𝑥 = 𝑟∕𝑅 = 0.7.
The axis of the cylinder coincides with the propeller shaft. The intersection of blade
and cylinder reveals the blade foil section (dark gray) at the selected radius.
If we repeat the intersection process for cylinders with different radii and unroll the
resulting foils onto a flat surface, we see the foils as they are shown in Figure . for
the radii 𝑥 = 0.2, 0.7, and 0.975. Notice that the pitch angle 𝜙 is highest close to the
hub (𝑥 = 0.2) and decreases going outward toward the blade tip. This change in pitch
angle gives the propeller blade its characteristic twisted shape.
Now remember that every point on the propeller travels on a helical path. If we view Constant pitch
the helical paths from the top, they look like sine curves – assuming we start with the
blade in a vertical position (Figure .). Since the arc length of the inner radii is much

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

424 35 Propeller Geometry

j Figure 35.3 Hydrofoil section within a propeller blade j

Figure 35.4 Pitch angle variation for a propeller with constant pitch 𝑃

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.2 Principal Propeller Characteristics 425

j j

Figure 35.5 Helical paths for propeller with constant pitch 𝑃

shorter than the helix arc length toward the tip, inner radii must have a greater pitch
angle in order to travel the same pitch distance 𝑃 as the outer radii in one revolution. If
the pitch is independent of the radii, i.e.

𝑃 (𝑟) = 𝑃 = const. (.)

we have a propeller with constant pitch.

Pitch 𝑃 is a length and measured in meters. For design purposes the dimensionless Pitch–diameter
pitch–diameter ratio 𝑃 ∕𝐷 is used to specify the propeller pitch. Pitch–diameter ratios ratio 𝑃 ∕𝐷
vary from 0.5 to 1.4 and sometimes higher for racing boats. High vessel speed is usually
combined with a high pitch–diameter ratio which provides better efficiency compared
with a higher rate of revolution. However, propellers with high 𝑃 ∕𝐷 values need engines
capable of generating high torque at low rate of revolutions.
Besides the diameter, pitch–diameter ratio is also an important parameter in propeller Design impact of
selection and design. 𝑃 ∕𝐷 is fine-tuned to provide the highest possible efficiency for a 𝑃 ∕𝐷
given vessel speed and wake field. Errors in finding the optimal pitch–diameter ratio
will lead to poor performance of the vessel.
The connection between pitch angle 𝜙 and pitch 𝑃 is found by unrolling the helix on a Pitch angle
flat surface (Figure .). Pitch 𝑃 and circumference 2𝜋 𝑟 form a right-angled triangle.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

426 35 Propeller Geometry

Figure 35.6 Relationship between pitch 𝑃 and

pitch angle 𝜙

The hypotenuse is equal to the arc length of the helix.

( )
𝑃 𝑃
tan 𝜙(𝑟) = or 𝜙(𝑟) = tan−1 (.)
2𝜋 𝑟 2𝜋 𝑟

Pitch angle Some examples might help to get comfortable with these geometric relationships.
example Consider a propeller of diameter 𝐷 = 3.2 m and a constant pitch distribution with
pitch–diameter ratio 𝑃 ∕𝐷 = 0.8. Therefore, pitch is equal to
𝑃
𝑃 = 𝐷 = 3.2 m ⋅ 0.8 = 2.56 m
𝐷
j j
Using Equation (.), the following pitch angle distribution is found:

𝑥 𝑟 tan 𝜙 𝜙 𝜙
[–] [m] [–] [rad] [deg]
0.20 0.32 1.2732 0.9050 51.854
0.30 0.48 0.8488 0.7038 40.326
0.40 0.64 0.6366 0.5669 32.482
0.50 0.80 0.5093 0.4711 26.990
0.60 0.96 0.4244 0.4014 22.997
0.70 1.12 0.3638 0.3489 19.991
0.80 1.28 0.3183 0.3082 17.657
0.90 1.44 0.2829 0.2757 15.798
1.00 1.60 0.2546 0.2493 14.287

Note that the pitch angle distribution is nonlinear and that the pitch angle at the tip
may vanish only if the pitch is zero (unlikely case).
Variable pitch Variable pitch means that the pitch is a function of the radius 𝑟. The helix of different
radii no longer end in the same point (Figure .). This is admittedly confusing because
the propeller is certainly not disintegrating. A steel bolt and corresponding thread
in a steel plate with variable pitch would make no sense. The bolt would not be able
to turn. Again, we have to remind ourselves that the propeller moves through a fluid
which gives way. Therefore, blade sections of propellers with variable pitch may travel
different distances relative to the water but an outside observer still sees the propeller
moving with ship speed.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.2 Principal Propeller Characteristics 427

j j

Figure 35.7 Helical paths for propeller with variable pitch 𝑃

It is common practice to reduce the pitch for inner and outer radii to adjust angles
of attack to the wake field and to lower the risk of cavitation. The reduction of pitch
reduces the angle of attack which in turn lowers the amount of lift produced by the foil
sections with reduced pitch angle. Lower lift force requires a smaller drop in pressure
which means less cavitation.
Variable pitch 𝑃 (𝑟) should not be confused with controllable pitch where the whole blade
can be rotated. Both a fixed pitch and a controllable pitch propeller may feature variable
pitch distributions. The pitch 𝑃0.7 at radius 𝑥 = 0.7 is often stated as a representative
pitch value for variable pitch propellers.
The distance actually traveled through the water is smaller than the geometric pitch 𝑃 . Propeller slip
The difference is known as propeller slip.
Last but not least, we explain the expanded area ratio 𝐴𝐸 ∕𝐴0 , which is arguably the Expanded area
most abstract of the propeller parameters. ratio 𝐴𝐸 ∕𝐴0

1
𝑍 𝑅 ∫ 𝑐(𝑥) d𝑥
𝐴𝐸 𝑥ℎ
= (.)
𝐴0 𝜋 𝐷2
4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

428 35 Propeller Geometry

Figure 35.8 Basic geometric properties of a lifting hydrofoil

Foil sections We take another look at the foil sections derived by intersecting the blade with a cylinder
(Figure .). Figure . shows a foil section laid out flat and rotated to zero pitch
angle. The distance from leading to trailing edge of a foil section is called chord length
𝑐. Most foils used in propellers have a camber 𝑓0 , i.e. their center line is slightly bent
toward the back side of the foil. Camber is usually just a few percent (low single digits)
of the chord length. The bigger the camber, the more lift a foil produces at zero angle
of attack. Strength requirements determine the foil section thickness 𝑡. Classification
rules define minimum requirements. Thickness is highest at the root of the blade and
declines toward the tip where bending stresses in the blade vanish.
Expanded blade If all foil sections are extracted, rolled flat, and rotated to zero pitch angle, the expanded
blade is obtained as shown in Figure .. Foil sections would normally appear as thin
j lines in this view but the sections have been rotated upright to make them visible. The j
expanded blade reveals the true chord length distribution 𝑐(𝑥) as a function of the
dimensionless blade radius 𝑥 = 𝑟∕𝑅. The blade reference line marks the position of the
mid chord. For manufacturing purposes points on the blade surfaces are specified with
respect to the generator line. The generator line bisects the chord length on the smallest
radius used to define the blade shape. The corresponding chord length at 𝑥 = 0.15 is
bisected in Figure .. The expanded view also shows a curve marking the location of
maximum foil section thickness. Both reference line and maximum thickness location
are fair curves.
The expanded area of a single blade is obtained by integration of the chord length
distribution from hub radius to blade tip.
1
𝐴𝐸
= 𝑅 𝑐(𝑥) d𝑥 expanded area of propeller blade (.)
𝑍 ∫
𝑥ℎ

The result is equal to the gray shaded area in Figure .. Multiplying Equation (.)
by the number of blades 𝑍 results in the expanded area of the propeller:
1

𝐴𝐸 = 𝑍 𝑅 𝑐(𝑥) d𝑥 expanded area of propeller (.)

∫
𝑥ℎ

Division by the propeller disk area 𝐴0 yields the dimensionless expanded area ratio
𝐴𝐸 ∕𝐴0 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.2 Principal Propeller Characteristics 429

j j

Figure 35.9 The expanded blade for a Wageningen B-Series propeller with four blades and
𝐴𝐸 ∕𝐴0 = 0.85. Drawing is based on data from Oosterveld and van Oossanen (1975)

The expanded area ratio plays a key role in cavitation prevention. A higher expanded Design impact of
area ratio means that thrust generation is spread over a larger blade area. This in turn 𝐴𝐸 ∕𝐴0
requires smaller pressure differences on the blade compared with a propeller with lower
expanded blade area ratio. Smaller pressure differences lowers the danger of cavitation
(see Chapter  on cavitation). On the other hand, larger blade area increases frictional
losses and diminishes propeller efficiency. Therefore, as a general rule, expanded area
ratio 𝐴𝐸 ∕𝐴0 is chosen as low as possible for high efficiency but as high as necessary to
prevent cavitation. Excessive cavitation would lead to even higher efficiency losses and
might ultimately damage the propeller.
Typically, expanded area ratios for fixed pitch propellers vary between 0.5 and 1.1. The
lower the thrust loading coefficient, the lower is the necessary expanded area ratio.
Fast merchant vessels like container ships feature expanded area ratios between 0.8
and 1.0. Expanded area ratios for tankers and bulk carriers range between 0.5 and
0.7. Controllable pitch propellers are available up to about 𝐴𝐸 ∕𝐴0 = 0.75. For higher
expanded area ratios the capability to reverse thrust by turning the blades to negative
pitch angles is lost because the blades will make contact during the rotation. Figure .
illustrates two typical expanded area ratios for propellers with four blades.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

430 35 Propeller Geometry

j j
Figure 35.10 Definition of the expanded area 𝐴𝐸 ∕𝑍 of a propeller blade

Figure 35.11 Two examples of expanded area ratios. 𝐴𝐸 ∕𝐴0 = 0.55 on the left and 𝐴𝐸 ∕𝐴0 = 0.85 on the right

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.3 Other Geometric Propeller Characteristics 431

Figure 35.12 Side elevation of a propeller blade and

definition of propeller rake 𝑖𝐺 (𝑥) and rake angle 𝜃

35.3 Other Geometric Propeller Characteristics

j j
The generator line is drawn upwards on the expanded view and used to define the Rake and rake
blade in its 12 o’clock position. For most propellers the generator line is inclined aft angle
by some degrees when viewed in the side elevation (Figure .). The rake angle 𝜃 of
the generator line is measured in degrees from a normal to the propeller shaft. Older
propeller designs, like the Wageningen B-Series, have rake angles 𝜃 of up to 15 degrees.
The rake was used to increase clearance between propeller and hull. Its influence on
performance is small.
More contemporary designs tend to have smaller rake angles. In some cases the angle
of the generator line to the vertical axis is reduced toward the tip. The generator line is
no longer straight. Due to the rake angle, blade sections are shifted backward parallel
to the propeller shaft direction. The distance is known as the rake 𝑖𝐺 and measured
in meters. The subscript 𝐺 stands for the generator line influence. Rake is considered
positive if blade sections are shifted aft. This is the usual case.
Additional rake of blade sections is created by propeller skew or, more appropriately, Skew-back
skew-back. Figure . illustrates the definitions. A blade is skewed if the projection
of the blade reference line into the propeller plane deviates from the generator line.
Positive skew-back shifts the blade section aft along its pitch helix, i.e. against the
direction of rotation for forward thrust. The skew-back is equal to the arc length of the
backward shift measured along the helix between points 1 and 2 in Figure .. The
backward shift along the helix causes an additional skew induced rake 𝑖𝑆 which is equal
to the component of the skew-back that is parallel to the shaft line. The skew-back is a
function of the radius 𝑟.
Skew-back is commonly measured in degrees as the angle between the intersections of Skew angle
the generator line 1 and intersection of the blade reference line 2 with the cylinder at

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

432 35 Propeller Geometry

Figure 35.13 Definition of propeller skew-back and skew angle 𝜃𝑠

j j

radius 𝑟. The skew angle is sometimes also referred to as warp. Use of the alternative
definition a is discouraged. This alternative replaces point 1 with the intersection of
the cylinder and a radial which is tangent to the blade reference line.
The skew angle 𝜃𝑠 for the blade tip is used to quantify the skew of a propeller. However,
each blade section has its own skew angle as indicated for the radius 𝑥 = 0.7 in Fig-
ure .. Skew angles may be negative for radii close to the hub, before the reference
line bends backwards for positive propeller skew angles toward the blade tip. In the
projected view (looking forward along the propeller
√ shaft) only the arc 𝑟 𝜃𝑠 (𝑟) is visible.
The actual skew-back is equal to the length (𝑟 𝜃𝑠 (𝑟))2 + 𝑖2𝑆 (𝑟).
High skew Propellers with a skew angle of more than  degrees at the tip are considered high
propeller skew propellers. The blade pictured in Figure . falls into this category.
Impact of Blades with high skew-back are considerably more expensive. Assessment of propeller
skew-back strength of highly skewed blades requires detailed analysis. Skew-back is applied to
minimize a propeller’s noise signature. Passage of the propeller blades underneath the
stern causes pressure fluctuations with a pronounced peak at the hull plating closest
to the tip. Those pressure fluctuations excite vibrations in the hull structure which
may be strong enough to be felt and definitely can be heard by sensitive underwater
microphones. Skew-back of blades flattens the pressure peak and helps to reduce
vibrations. Skew-back may also be used to influence cavitation patterns on propeller
blades. The impact of skew-back on the efficiency is fairly small for forward motion.
However, high skew propellers may be quite inefficient for operations with reversed
direction of rotation.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

35.3 Other Geometric Propeller Characteristics 433

j j

Figure 35.14 Cupping of a propeller blade

A feature seemingly popular with powerboat enthusiasts is to cup the propeller blades. Cupping
The blade sections are bent toward the pressure side (face) close to the trailing edge
(Figure .). Similar to flaps at the trailing edge of a wing, cupping increases the
camber of a foil, which results in a higher lift coefficient for the same angle of attack.
Cupping may also be interpreted as a correction of the effective pitch of the propeller.
Increases in effective pitch angle result in a propeller that produces more thrust, but
also requires more torque. In many cases, efficiency of the propulsion system does not
increase. On the contrary, if the effective pitch is increased too much the engine torque
might be insufficient to bring the propeller up to its design rate of revolution, with
detrimental effects on performance. This is especially true in service with additional
weather related resistance components. A properly designed propeller will not need
cupping.

References
Breslin, J. and Andersen, P. (). Hydrodynamics of ship propellers. Cambridge
University Press, Cambridge, United Kingdom.
Carlton, J. (). Marine propellers and propulsion. Butterworth-Heinemann, Oxford,
United Kingdom.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

434 35 Propeller Geometry

Kerwin, J. and Hadler, J. (). Propulsion. The Principles of Naval Architecture Series.
Society of Naval Architects and Marine Engineers (SNAME), Jersey City, NJ.
Oosterveld, M. and van Oossanen, P. (). Further computer-analyzed data of the
Wageningen B-screw series. International Shipbuilding Progress, :–.

Self Study Problems

. State the four most important geometric propeller characteristics.
. Consider the example about pitch angles on page . Assume the data describes
a controllable pitch propeller. Compute the radial pitch distribution 𝑃 (𝑥) if the
blade is rotated by  degrees.

. Consider a propeller with the following radial pitch distribution:

𝑥 𝑟 𝑃 (𝑥)
[–] [m] [m]
0.20 0.40 2.4000
0.30 0.60 2.7200
0.40 0.80 2.9500
j 0.50 1.00 3.0000 j
0.60 1.20 3.0000
0.70 1.40 3.0000
0.80 1.60 3.0000
0.90 1.80 2.9700
1.00 2.00 2.8000

Plot the pitch distribution as a function of 𝑥. Make 𝑥 the vertical axis of the plot.
State pitch–diameter ratio and diameter of the propeller. Compute the pitch
angle distribution 𝜙(𝑥).
. A standard ceiling fan uses inclined but flat boards as wings. Is this a variable
pitch or constant pitch propeller?

. Why is the pitch of propeller blades sometimes reduced toward hub and tip of
the blade?
. Which criterion is used to select the expanded area ratio of the propeller?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

435

36
Lifting Foils

A basic element of a propeller blade is the foil section. Figure . visualizes how foil
sections are embedded in a blade. Foil sections are streamlined shapes designed to
generate a lift force perpendicular to the onflow, hence the name lifting foil. Lifting foils
are also used in rudders, stabilizer fins, and sails. In this chapter we discuss geometry
and basic flow patterns across lifting foils. We also introduce thin foil theory which is
one of the most widely used tools in foil design.

Learning Objectives

j At the end of this chapter students will be able to: j

• explain geometry and flow patterns of lifting foils
• define lift and drag coefficients
• understand the assumptions of thin foil theory

36.1 Foil Geometry and Flow Patterns

The extent of a foil from trailing edge to leading edge is called chord length 𝑐 (Figure .). Foil section
The straight tail–nose line connects trailing and leading edge and serves as the 𝑥-axis of geometry
the foil coordinate system. Here we place the origin in the middle of the chord length 𝑐.
Chord length 𝑐 together with thickness distribution 𝑡(𝑥) and camber distribution 𝑓𝑚 (𝑥)
defines the geometry of a foil section. For the leading edge, a radius 𝑟𝐿𝐸 (nose radius)
may be specified to better define the shape of the hydrodynamically important nose.
The trailing edge of foils is comparatively sharp in most cases. However, thickness at the
trailing edge has to be finite for strength and manufacturing reasons. Superscripts + and
− are used to identify points or properties of upper and lower foil contour, respectively.

Points 𝑓 + on the upper foil side are obtained by moving a distance 𝑡(𝑥) upward in a
direction normal to the camber line. The lower foil side is obtained by moving 𝑡(𝑥)
downward normal to the camber line. The following equations define upper + and
lower − foil contour.
( )
𝑥± = 𝑥 ∓ 𝑡(𝑥) sin 𝜃(𝑥) (.)
( )
±
𝑓 = 𝑓𝑚 (𝑥) ± 𝑡(𝑥) cos 𝜃(𝑥) (.)
Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

436 36 Lifting Foils

Figure 36.1 Definition of foil geometry

The minus sign in the right-hand side of Equation (.) is listed on top because it
applies to the upper side of the foil. 𝜃(𝑥) measures the angle of the camber distribution
𝑓𝑚 (𝑥) with respect to the tail–nose line.
( )
d𝑓𝑚
𝜃(𝑥) = tan−1 (.)
d𝑥

Typically, the ratio 𝑓0 ∕𝑐 of maximum camber 𝑓0 and chord length 𝑐 is less than 2%.
Consequently, the angle 𝜃 is small, which justifies the simplifications sin(𝜃) ≈ 0 and
j cos(𝜃) ≈ 1. This is equivalent to defining the foil contours by moving up or down from j
the line perpendicular to the tail–nose line rather than normal to the camber line.

𝑥± = 𝑥 (.)
𝑓 ± = 𝑓𝑚 (𝑥) ± 𝑡(𝑥) (.)

The documentation of the foil shape should specify which version of the coordinate
definition is employed.
With the simplified foil contour description (.) and (.), thickness and camber
distribution are equivalent to half the difference and half the sum of upper and lower
contour 𝑦-coordinates respectively.

1( + ) 1( + )
𝑡(𝑥) = 𝑓 − 𝑓− 𝑓𝑚 (𝑥) = 𝑓 + 𝑓− (.)
2 2
Owing to the latter equation, the camber line is also referred to as the mean line.
Flow patterns Figure . shows a typical flow pattern and the associated pressure distribution. The
foil is rotated anticlockwise by the angle of attack 𝛼 with respect to the onflow 𝑣0 . On
the upper (back) side of the foil, streamlines bunch together. The continuity equation
tells us that a narrowing of streamlines indicates an acceleration of the flow. Thus, much
of the upper side of the foil sees flow speeds |𝑣| higher than the speed of the onflow |𝑣0 |.
According to Bernoulli’s equation, increased flow speed causes a reduction in pressure.
This is clearly visible in the negative pressure coefficient 𝐶𝑝+ values for the back of the
foil. As stated before, a negative pressure coefficient does not mean the pressure is
negative; rather, it implies that local pressure 𝑝 is smaller than the reference pressure
𝑝0 . Therefore, the alternative name for the back of the foil is suction side. Especially

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36.1 Foil Geometry and Flow Patterns 437

Figure 36.2 Flow pattern and pressure distribution for a 2D foil section at angle of attack 𝛼

j j

toward the leading edge, we encounter low pressure and high flow speeds when the foil
is operating at large angles of attack.
The opposite happens on the underside of the foil, the pressure side. Flow is decelerating
to a stop at the stagnation point and is slower than the onflow across almost all of the
face side of the foil. The pressure coefficient 𝐶𝑝− is positive. The pressure difference
between upper and lower side of a foil determines its lift force.

Δ𝐶𝑝 = 𝐶𝑝+ − 𝐶𝑝− (.)

Friction between fluid and foil develops into a thin boundary layer. The resulting wall
shear stress creates a drag force.
Note that in Figure ., the depicted angle of attack 𝛼 is rather high with about Stall
12 degrees. Typical operational angles of foils are much smaller. Increases in angle of
attack will cause the foil to stall. During a stall event, the flow separates already at the
leading edge, as illustrated in Figure .. Instead of a flow with high speed and low
pressure, the foils produce a large wake filled with large eddies. The lift force all but
vanishes and drag becomes the dominant force.
Stall causes a sudden loss of lift and flight control in airplanes. If the pilots fail to
counteract immediately, the plane will crash. The stalling of marine propeller blades
has less severe consequences but will also result in loss of performance and increased
vibrations.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

438 36 Lifting Foils

Figure 36.3 Foil in stalled flow condition

Figure 36.4 Complete vortex system of the foil section

j j
36.2 Lift and Drag
Kutta- The German mathematician Martin Wilhelm Kutta (*–†) and the Russian sci-
Joukowsky’s lift entist Nikolay Yegorovich Joukowsky (*–†, name is anglicanized from Russian)
theorem
derived independently from each other the Kutta-Joukowsky’s lift theorem.

𝐿 = 𝜌 |𝑣0 | Γ (.)

Note that the section lift 𝐿 has the dimension of force per unit length [N/m]. Kutta
is most famous as one of the inventors of the Runge-Kutta method for solving ordi-
nary differential equations and for the Kutta condition which is necessary to describe
lifting flows with potential flow theory. Joukowsky is one of the founding fathers of
aerodynamics.
Lift may be accurately computed with potential flow theory assuming an inviscid,
incompressible, and irrotational fluid. A nonzero circulation Γ ≠ 0 requires the curl
of the velocity vector to vanish as well, i.e. ∇ × 𝑣 ≠ 0. This seems to contradict the
application of potential flow. However, as explained in Chapter , the condition of
vanishing curl only applies to simply connected regions. If we include the foil, the flow
region becomes multiply connected.
The effect of circulation on a lifting flow can be modeled with vortices. According to
Helmholtz’s theorems, vortex strength must be constant in inviscid fluids. So, where
did the circulation come from if it was not there already (foil at rest)? When the foil
is accelerating and circulation is established around the foil, a start-up vortex is left
behind (Figure .). The start-up vortex is of equal strength but rotates in the opposite

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36.2 Lift and Drag 439

Figure 36.5 Forces acting on the foil section at angle of attack 𝛼

direction compared with the circulation that is traveling with the foil. Therefore, the
sum of vorticity in the entire flow field is still zero.
Although this is a crude model of the physics, start-up vortices are real. The start-up
vortex left behind by a big commercial jet at takeoff may easily flip over a small single
propeller aircraft that happens to fly through it. It is one of the reasons why aircraft of
j different sizes need longer time periods between landing and takeoff than aircraft of j
equal size. The start-up vortex is dissolved by friction over time, and the vortices are
blown away by the wind.
Figure . shows lift force 𝐿 and drag force 𝐷 acting on a foil section. Lift acts Lift and drag
perpendicular to the onflow. However, an additional drag force 𝐷 exists in viscous force
(real) fluids which acts in the direction of the onflow 𝑣0 . Viscous pressure drag for thin
foils is small as long as the angle of attack is small. It increases with increasing angle of
attack 𝛼. Frictional drag is the major drag component and also depends on the surface
roughness of the foil. Increased surface roughness due to fouling or damage is especially
detrimental to performance if it affects the leading edge and the first quarter of the
suction side. See Section . for details.
Besides lift and drag forces there is also a rotational moment which aspires to flip the Pitch moment
foil over. Aircraft have a tail unit to counterbalance the moment. The moment is of
interest for the design of controllable pitch propellers.
Hydrodynamic properties of foil sections are often provided as lift and drag coefficients Lift and drag
𝐶𝑙 and 𝐶𝑑 respectively. coefficients
𝐿
𝐶𝑙 = (.)
1 2
𝜌 𝑣0 𝑐
2
𝐷
𝐶𝑑 = (.)
1 2
𝜌 𝑣0 𝑐
2
These two-dimensional foil section coefficients use the chord length 𝑐 as reference
length. Note that this is different from the three-dimensional resistance coefficients

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

440 36 Lifting Foils

0.025
smooth surface
rough surface

[]
section drag coefficient Cd 0.020

0.015
Typical range
of design lift
coefficients

0.010

Cd =0.008 for initial design

0.005

0.000
1.5 1.0 0.5 0.0 0.5 1.0 1.5
section lift coefficient Cl []

Figure 36.6 Typical lift–drag curves for a thin, cambered foil

used in ship hydrodynamics which use the wetted surface of the hull as reference area
j rather than a projection onto a plane. j
Lift–drag ratio The quality of a foil section may be expressed as the lift–drag ratio 𝐿∕𝐷 or 𝐶𝑙 ∕𝐶𝑑 . The
more lift is generated per unit drag, the more effective is the foil. Form drag of the foil
section increases with the angle of attack 𝛼 and the lift–drag ratio decreases. This may
be best visualized in a plot of drag coefficient 𝐶𝑑 over lift coefficient 𝐶𝑙 as illustrated in
Figure .. The graph shows typical section drag coefficient curves for a smooth and a
rough foil section. The foil has a small camber and its most efficient range of operation
is for section lift coefficients between 0.0 and 0.5. Lift–drag curves for many standard
foil section shapes can be found in Abbott and von Doenhoff ().
In the following we will discuss a method suitable for analyzing the potential flow
around two-dimensional foil sections. It is surprisingly accurate in predicting the lift
force of foil sections. Drag has to be determined from model tests or by semi-empirical
formulas which combine theory and experimental results. A section drag coefficient of
𝐶𝑑 = 0.008 is often used for initial propeller design.

36.3 Thin Foil Theory

Conformal Conformal mapping is the most accurate method for the analysis of two-dimensional
mapping potential flow around foil sections. Conformal mapping transforms the known solution
of the flow around a cylinder into the flow around a foil. The geometries of cylinder and
foil are represented in complex planes 𝑧 = 𝑥 + i𝑦 and 𝜁 = 𝜉 + i𝜂 respectively. At the
heart of the procedure is a complex transformation function which maps points of the 𝑧-
plane (cylinder) into points of the 𝜁-plane (foil). Finding the appropriate transformation

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36.3 Thin Foil Theory 441

Figure 36.7 Setup of boundary value problem for a thin foil operating at angle of attack 𝛼

function for a given geometry is the most difficult task. Readers interested in lifting
flow are encouraged to study conformal mapping. Detailed descriptions are found in
Abbott and von Doenhoff () and Katz and Plotkin ().
Although conformal mapping is a powerful tool for the analysis of two-dimensional Thin foil theory
forms, it is less well suited for design purposes. In conformal mapping, hydrodynamic assumptions
properties are evaluated for a given shape. During design of lifting foils we would like
to specify desired dynamic characteristics first, e.g. a section lift coefficient, and then
find the geometry which produces this characteristic. A suitable tool for this purpose is
thin foil theory. We assume that the foil geometry is given in the form of a mean line
j 𝑓𝑚 (𝑥) and a thickness distribution 𝑡(𝑥). Offsets of upper and lower foil contour 𝑓 ± (𝑥) j
may be derived from Equations (.) and (.). A foil is considered thin if its slope
with respect to the 𝑥-axis (tail–nose line) is small.
𝜕𝑓 ±
≪1 for 𝑥 ∈ [−𝑐∕2, 𝑐∕2] (.)
𝜕𝑥
In addition, thin foil theory assumes that the angle of attack 𝛼 is small.
𝛼 ≪ 1 angle of attack in radians (.)

36.3.1 Thin foil boundary value problem

Analogous to the analysis of the flow around a cylinder, we seek the velocity potential Φ
as the solution of a boundary value problem. We use a body fixed < 𝑥, 𝑦 > coordinate
system, with the 𝑥-axis falling onto the tail–nose line and pointing from trailing edge to
leading edge (Figure .). The velocity potential Φ is initially comprised of two parts:

• a uniform parallel flow in the far field which is oriented at the angle of attack 𝛼
with respect to 𝑥-axis. It has the velocity potential
𝜙∞ (𝑥, 𝑦) = −𝑣0 𝑥 cos(𝛼) + 𝑣0 𝑦 sin(𝛼) (.)
The velocity 𝑣0 is the speed of the foil relative to the fluid in the far field.
• a disturbance caused by the foil which fades away in the far field. Its potential
𝜙(𝑥, 𝑦) has yet to be determined.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

442 36 Lifting Foils

We exploit again the superposition principle grounded in the linearity of the Laplace
equation. The total flow potential is simply the sum of its parts.

Φ(𝑥, 𝑦) = 𝜙∞ (𝑥, 𝑦) + 𝜙(𝑥, 𝑦) (.)

The boundary of the fluid domain 𝑉 consists of the foil contour 𝑆𝑏 , the far field 𝑆∞ ,
and the wake barrier 𝑆𝑤 (Figure .). The latter is needed to determine the circulation
Γ. As before, the normal vector 𝑛 is pointing out of the fluid domain. That means 𝑛 is
pointing into the foil contour on 𝑆𝑏 .
Boundary value The boundary value problem consists of the Laplace equation as the field equation
problem representing conservation of mass of an ideal, irrotational fluid, and the boundary
condition for each boundary.

𝚫Φ = 0 in 𝑉
𝜕Φ
= 0 body boundary condition for 𝑆𝑏 (.)
𝜕𝑛
𝜕𝜙
= 0 far field condition for 𝑆∞
𝜕𝑛
plus a Kutta condition for the wake 𝑆𝑤 . We will specify a Kutta condition in more detail
when we solve the lifting flow problem in Chapter . The body boundary condition
requires that no fluid flows through the foil contour 𝑆𝑏 . This means that the velocity
component normal to the contour has to vanish and the contour itself becomes a
j streamline. The far field condition requires that the disturbance potential 𝜙 vanishes j
far away from the foil.

36.3.2 Thin foil body boundary condition

Physical The body boundary condition holds the key to a significant simplification of the problem.
interpretation Therefore, it is discussed in greater detail. The body boundary condition in the boundary
value problem (.) requests that the normal derivative of the potential vanishes.
The normal derivative is the component of the gradient of the potential that points in
the direction of the normal vector. It is computed via the dot product, and the body
boundary condition requires that this component is of zero length, i.e. the two vectors,
normal and gradient of the potential, are perpendicular to each other.
𝜕Φ
= 𝑛𝑇 ∇ Φ = 0 (.)
𝜕𝑛
Since the gradient of the velocity potential is equal to the velocity vector, the latter must
be tangent to the foil contour. No flow passes through the foil surface.
Normal vector First, we need to relate the normal vector to the geometry of the foil contour. As for the
discussion of the free surface condition in linear wave theory, we make use of implicit
functions. As before, we use superscripts + and − to identify properties of upper and
lower foil contour, respectively.

upper foil surface 𝐹 + (𝑥, 𝑦) = 𝑦 − 𝑓 + (𝑥) = 0 (.)

lower foil surface 𝐹 − (𝑥, 𝑦) = 𝑦 − 𝑓 − (𝑥) = 0 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36.3 Thin Foil Theory 443

Figure 36.8 Definition of normal vectors for

upper and lower foil surface

The implicit definition represents an isoline of 𝐹 ± , here 𝐹 ± = const. = 0. The gradient

of a function is pointing in the direction of steepest ascent which is equivalent to being
normal to the isolines of the function. Then, the normal vector of an implicit function
must be parallel to its gradient.

∇𝐹±
𝑛± = ∓ (.)
|∇ 𝐹 ± |

Since the contours 𝑓 ± are only functions of 𝑥, the gradients of 𝐹 + and 𝐹 − are equal to
j j
⎛ 𝜕𝐹 ±
⎞ ±
⎛ − 𝜕𝑓 ⎞
⎜ 𝜕𝑥 ⎟ ⎜ 𝜕𝑥 ⎟
∇𝐹± = ⎜ ± ⎟ = ⎜ (.)
⎜ 𝜕𝐹 ⎟ ⎟
⎝ 𝜕𝑦 ⎠ ⎝ 1 ⎠

As a consequence, the normal vector 𝑛+ on the suction side becomes

+
⎛ 𝜕𝑓 ⎞
1 ⎜ 𝜕𝑥 ⎟
𝑛 +
= √ (.)
( )2 ⎜ ⎟
𝜕𝑓 + ⎝ −1 ⎠
1+
𝜕𝑥

and the normal vector on the pressure side is

−
⎛ − 𝜕𝑓 ⎞
1 ⎜ 𝜕𝑥 ⎟
𝑛− = √ (.)
( )2 ⎜ ⎟
𝜕𝑓 − ⎝ 1 ⎠
1+
𝜕𝑥

Note the different signs to make the normal vector point into the foil (out of the fluid
domain 𝑉 ) on both suction and pressure side (see Figure .).
Substitution of the normal vector (.) into the body boundary condition (.) Suction side
yields:

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

444 36 Lifting Foils

( ) ⎛ 𝜕Φ ⎞
( )
+ 𝑇 𝜕𝑓 + ⎜ 𝜕𝑥 ⎟
, −1 ⎜ (.)
𝜕Φ ⎟
𝑛 ∇Φ = = 0
𝜕𝑥 ⎜ ⎟
⎝ 𝜕𝑦 ⎠
𝜕𝑓 + 𝜕Φ 𝜕Φ
− = 0 on 𝑓 + (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦
In addition, we split the total velocity potential into parallel flow and disturbance
potential
( )
𝜕𝑓 + 𝜕𝜙+ 𝜕𝜙+
−𝑣0 cos(𝛼) + − 𝑣𝑜 sin(𝛼) − = 0 (.)
𝜕𝑥 𝜕𝑥 𝜕𝑦
For a flat plate the change of velocity 𝜕𝜙∕𝜕𝑥 along the plate would vanish. For a thin foil
it will not be zero but significantly smaller than the onflow velocity 𝑣0 . Together with
the basic assumptions (.) and (.) of thin foil theory we introduce the following
simplifications:
𝜕𝑓 + 𝜕𝜙+
≪ 1, cos(𝛼) ≈ 1, sin(𝛼) ≈ 𝛼, and ≪ 𝑣0 (.)
𝜕𝑥 𝜕𝑥
As a consequence, the first parenthesis in Equation (.) reduces to −𝑣0 .
j 𝜕𝑓 + 𝜕𝜙+ j
− 𝑣0 − 𝛼 𝑣𝑜 − = 0 (.)
𝜕𝑥 𝜕𝑦

We rearrange the terms to have the unknown 𝑦-derivative of the disturbance potential
on the left-hand side.
𝜕𝜙+ 𝜕𝑓 +
= − 𝑣 − 𝛼 𝑣𝑜 (.)
𝜕𝑦 𝜕𝑥 0
This is sometimes called the linearized body boundary condition for thin foil theory,
although technically the body boundary condition (.) is already linear. Nonetheless,
the body boundary condition in its form (.) for the upper foil surface represents a
significant simplification. All terms on the right-hand side are known.
Pressure side Repeating the conversion of Equations (.) through (.) for the pressure side
𝑓 − results in the same simplified body boundary condition except that 𝑓 + is replaced
by 𝑓 − .
𝜕𝜙− 𝜕𝑓 −
= − 𝑣 − 𝛼 𝑣𝑜 (.)
𝜕𝑦 𝜕𝑥 0

or in short the body boundary condition for a thin foil now reads as

𝜕𝜙± 𝜕𝑓 ±
= − 𝑣 − 𝛼 𝑣𝑜 for 𝑦 = ±0, 𝑥 ∈ [−𝑐∕2, 𝑐∕2] (.)
𝜕𝑦 𝜕𝑥 0
We satisfy this condition on the upper and lower sides of the 𝑥-axis with 𝑦 = ±0 because
the right-hand side is a univariate function of 𝑥.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

36.3 Thin Foil Theory 445

Introducing the definition (.) of foil geometry into the thin foil body boundary Thin foil body
condition (.) allows us to split the right-hand side into three distinct parts. With boundary
condition
𝑓 ± = 𝑓𝑚 + 𝑡 we get:
( )
𝜕𝜙± 𝜕𝑓𝑚 𝜕𝑡
= − ± 𝑣0 − 𝛼 𝑣𝑜
𝜕𝑦 𝜕𝑥 𝜕𝑥
𝜕𝑡 𝜕𝑓𝑚
= ∓ 𝑣0 − 𝑣 − 𝛼 𝑣0 for 𝑦 = ±0 (.)
𝜕𝑥 𝜕𝑥 0
This condition has to be satisfied on the tail–nose line between trailing and leading
edge, i.e. 𝑥 ∈ [−𝑐∕2, 𝑐∕2].

36.3.3 Decomposition of disturbance potential

Each term on the right-hand side of Equation (.) depends on a different foil charac- Decomposition
teristic. of disturbance
potential

(i) The first term is a function of the slope of the thickness distribution 𝜕𝑡(𝑥)∕𝜕𝑥.
(ii) The second term is defined by the slope of the camber distribution 𝜕𝑓𝑚 (𝑥)∕𝜕𝑥.
(iii) The last term depends on the angle of attack 𝛼.

j The superposition principle for solutions of the Laplace equation enables us to subdivide j
the disturbance potential into three parts.

𝜙 = 𝜙1 + 𝜙2 + 𝜙3 (.)

The original boundary value problem is split into three separate boundary value prob-
lems. Table . summarizes their key points. In practice, the two lifting flow problems
are solved together. The potential 𝜙1 depends only on the thickness distribution 𝑡(𝑥) Displacement
and describes the flow around a symmetric foil. This is called the displacement flow flow
problem or thickness problem. For zero angle of attack the flow will be symmetric as
well and we do not expect a resultant force acting normal to the flow. The potentials 𝜙2 Lifting flow
and 𝜙3 capture the flow effects caused by camber and the angle of attack, respectively.
In both cases, the flow will be asymmetric to the 𝑥-axis and a lift force will be generated.
Together, 𝜙2 and 𝜙3 describe the lifting flow. In the following two chapters we will solve
displacement and lifting flow problems under the condition that the foil is thin. We
will start with the displacement flow. For the lifting flow we also have to address the
question of the circulation strength Γ.

References
Abbott, I. and von Doenhoff, A. (). Theory of wing sections – including a summary
of airfoil data. Dover Publications, Inc., New York, NY, Dover edition.
Katz, J. and Plotkin, A. (). Low-speed aerodynamics. Cambridge Aerospace Series.
Cambridge University Press, New York, NY, second edition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

446 36 Lifting Foils

Table 36.1 The three subtasks of thin foil theory

displacement flow lifting flow

Symmetric foil at 𝛼 = 0 Cambered plate at 𝛼 = 0 Flat plate at angle of attack 𝛼
𝚫𝜙1 = 0 𝚫𝜙2 = 0 𝚫𝜙3 = 0

𝜕𝜙±
1 𝜕𝑡 𝜕𝜙±
2 𝜕𝑓 𝜕𝜙±
3
= ∓ 𝑣0 on 𝑦 = ±0 = − 𝑚 𝑣0 on 𝑦 = ±0 = −𝛼 𝑣0 on 𝑦 = ±0
𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦

∇ 𝜙1 ⟶ 0 for 𝑥 → ±∞ ∇ 𝜙2 ⟶ 0 for 𝑥 → ±∞ ∇ 𝜙3 ⟶ 0 for 𝑥 → ±∞

Kutta condition for Γ Kutta condition for Γ

Neither camber nor angle of at- The BVP for 𝜙2 involves neither The geometry of the foil (𝑓𝑚 (𝑥)
tack is present in this problem. It thickness distribution nor the an- and 𝑡(𝑥)) does not appear in this
describes the flow around a sym- gle of attack. Only the mean line problem. The potential 𝜙3 repre-
metric foil with thickness distri- 𝑓𝑚 (𝑥) is required. The problem sents the flow over a flat plate at
bution 𝑡(𝑥). defines the flow over a cambered angle of attack 𝛼.
plate at zero angle of attack.

j j

Self Study Problems

. Explain stall in your own words. When does it occur and what are its effects?
. Summarize the underlying assumptions of thin foil theory.
. The body boundary condition for the pressure side of a foil requires that the
velocity normal vanishes.
( )
𝜕𝑓 − 𝜕𝜙− 𝜕𝜙−
− −𝑣0 cos(𝛼) + + 𝑣𝑜 sin(𝛼) + = 0
𝜕𝑥 𝜕𝑥 𝜕𝑦

Apply the assumptions of thin foil theory to derive the body boundary condition
for thin foil theory (.).
. Research the geometry of the NACA -digit series and the NACA -xxx series
of foils. Learn how to extract important foil characteristics from the specification,
i.e. which specifications are embedded in the designation NACA . Use the
provide equations to plot the shape of NACA .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

447

37
Thin Foil Theory – Displacement Flow

In this chapter we continue to discuss the potential flow around a thin foil. The upper
and lower sides of the foil are represented by a mean line 𝑓𝑚 and a symmetric thickness
distribution.
𝑓 ± (𝑥) = 𝑓𝑚 (𝑥) ± 𝑡(𝑥) for 𝑥 ∈ [−𝑐∕2, 𝑐∕2] (.)
By introducing the requirements that

• the angle of attack 𝛼 is small,

𝜕𝑓 ±
• the foil represented by 𝑓 ± (𝑥) is thin with ≪ 1, and
𝜕𝑥
j • any velocities induced by the disturbance potential 𝜙 are small compared to the j
| 𝜕𝜙 |
parallel stream velocity || || ≪ 𝑣0
| 𝜕𝑥 |

we have been able to split the flow problem into three separate boundary value problems.
Here, we solve the first of these boundary value problems which describes the flow
around a symmetric foil with finite but small thickness–chord length ratio 𝑡∕𝑐, no
camber, and at zero angle of attack.

Learning Objectives
At the end of this chapter students will be able to:

• state and solve boundary value problems

• understand and apply singularity distributions

• discuss and compute the pressure distributions around symmetric foils

37.1 Boundary Value Problem

The flow around the symmetric foil at zero angle of attack may be represented by a
velocity potential Φ which consists of two parts: A flow potential 𝜙∞ , which describes

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

448 37 Thin Foil Theory – Displacement Flow

Figure 37.1 The boundary value problem of a symmetric thin foil with finite thickness and zero
angle of attack

the flow in the far field, and a velocity potential 𝜙1 , which describes the disturbance
caused by the foil. The potentials are functions of the planar coordinates 𝑥 and 𝑦.
Φ(𝑥, 𝑦) = 𝜙∞ (𝑥, 𝑦) + 𝜙1 (𝑥, 𝑦) (.)

Figure . shows the control volume and its boundaries. The boundary value problem
j for the disturbance potential consists of the Laplace equation and appropriate boundary j
conditions.
𝚫𝜙1 (𝑥, 𝑦) = 0 for (𝑥, 𝑦) ∈ 𝑉
𝜕𝜙1 [ ]
d𝑡 𝑐 𝑐
= ∓𝑣0 for 𝑥 ∈ − , and 𝑦 = 0± (.)
𝜕𝑦 d𝑥 2 2
√
𝜙1 (𝑥, 𝑦) ⟶ const. for 𝑥2 + 𝑦2 → ∞
Note that we simplified the body boundary condition according to thin foil theory
assumptions.
This problem is similar to the flow around a cylinder or a Rankine oval. The flow around
the latter was described by the potentials of a source, a sink, and parallel. This lends
credit to the idea that more complicated shapes may be represented by a multitude of
sources and sinks. As long as the sum of their strengths vanishes we will obtain a closed
contour.
Line distribution Going one step further, we use a line of distributed sources instead of a set of discrete
of sources sources and sinks. This is the same concept as a distributed load in mechanics of
materials. The source strength 𝜎 is now given as a ‘volume’ flow per unit length. Note
that the ‘volume’ flow in the plane is measured in length units squared per second.
We place the sources on the chord line 𝑦 = 0 between 𝑥 = −𝑐∕2 and 𝑥 = +𝑐∕2. The
total disturbance potential is obtained by integration along the length of the source
distribution.
+𝑐∕2
𝜎(𝜉) (√ )
𝜙1 (𝑥, 𝑦) = ln (𝑥 − 𝜉)2 + 𝑦2 d𝜉 (.)
∫ 2𝜋
−𝑐∕2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

37.1 Boundary Value Problem 449

A pair of coordinates (𝑥, 𝑦) marks the point where we want to know the potential or
velocity. The integration variable 𝜉 marks the position of sources along the tail–nose
line (𝜂 = 0). We have to adjust the yet unknown source strength 𝜎(𝜉) so that the closed
contour looks like our thin foil and represents a streamline of the flow. Therefore we
will employ the linearized body boundary condition for the displacement flow from
Table . to find the source strength.
The derivative of the disturbance potential (.) with respect to 𝑦 is part of the body Transverse
boundary condition. In the potential, both the limits of the integral and the integra- velocity
tion variable are independent of 𝑦, which allows us to interchange integration and
differentiation.
+𝑐∕2
𝜕𝜙1 (𝑥, 𝑦) 𝜎(𝜉) 𝑦
= ( ) d𝜉 (.)
𝜕𝑦 ∫ 2𝜋 (𝑥 − 𝜉)2 + 𝑦2
−𝑐∕2

The linearized body boundary condition is satisfied over the tail–nose line on the 𝑥- Evaluating the
axis with 𝑦 = 0 instead of the actual foil contour. The integrand in (.) vanishes integral
everywhere for 𝑦 = 0 except for the position 𝑥 = 𝜉 for which we obtain an undefined
expression of the type ‘/.’ Without loss of accuracy, we may limit the extent of the
integral to some finite interval of length 2𝛿 around 𝑥. 𝛿 is a small positive number.
The regions outside of this area will make no contribution to the integral because the
integrand vanishes (𝑦 = 0).
𝑥+𝛿
j 𝜕𝜙1 (𝑥, 𝑦) 𝜎(𝜉) 𝑦 j
= ( ) d𝜉 (.)
𝜕𝑦 ∫ 2𝜋 (𝑥 − 𝜉)2 + 𝑦2
𝑥−𝛿

We make 𝛿 so small that it is safe to assume that the source strength takes the value at
𝑥, i.e. 𝜎(𝜉) = 𝜎(𝑥) for 𝜉 ∈ [𝑥 − 𝛿, 𝑥 + 𝛿]. The source strength becomes independent of
the integration variable and can be brought in front of the integral sign.
𝑥+𝛿
𝜕𝜙1 (𝑥, 𝑦) 𝜎(𝑥) 𝑦
= ( ) d𝜉 (.)
𝜕𝑦 2𝜋 ∫ (𝑥 − 𝜉)2 + 𝑦2
𝑥−𝛿

We will proceed with caution and search for the value of integral (.) by letting
𝑦 = 0 ± 𝜀 and compute the limits for 𝜀 → +0 and 𝜀 → −0 if they exist.
𝑥+𝛿
𝜕𝜙1 (𝑥, 𝑦 = ±0) 𝜎(𝑥) 𝜀
= lim ( ) d𝜉 (.)
𝜕𝑦 𝜀→±0 2𝜋 ∫ (𝑥 − 𝜉)2 + 𝜀2
𝑥−𝛿

The following variable substitution simplifies the integration:

d𝜆
𝜆 = (𝑥 − 𝜉) = −1 i.e. d𝜉 = −d𝜆
d𝜉

The limits of the integral transform into

𝜆𝑙 = 𝑥 − (𝑥 − 𝛿) = +𝛿 𝜆𝑢 = 𝑥 − (𝑥 + 𝛿) = −𝛿

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

450 37 Thin Foil Theory – Displacement Flow

Figure 37.2 The inverse tangent function

2

arctan()
0

2
60 40 20 0 20 40 60

Substituting 𝜆 for 𝜉 changes the integral changes into the following form:

−𝛿
𝜕𝜙1 (𝑥, 𝑦 = ±0) 𝜎(𝑥) 𝜀
= lim ( ) (−d𝜆) (.)
𝜕𝑦 𝜀→±0 2𝜋 ∫ 𝜆2 + 𝜀2
+𝛿

j We switch upper and lower limits and consequently remove the minus sign. j
+𝛿
𝜕𝜙1 (𝑥, 𝑦 = ±0) 𝜎(𝑥) 𝜀
= lim ( ) d𝜆 (.)
𝜕𝑦 2𝜋 𝜀→±0 ∫ 𝜆2 + 𝜀2
−𝛿

The antiderivative of the integral of type (.) is the inverse tangent function arctan()
(Figure .).

𝜕𝜙1 (𝑥, 𝑦 = ±0) 𝜎(𝑥) [ ( 𝜂 )]+𝛿

1
= lim 𝜀 arctan
𝜕𝑦 2𝜋 𝜀→±0 𝜀 𝜀 −𝛿
𝜎(𝑥) ( ( ) ( ))
𝛿 −𝛿
= lim arctan − arctan
2𝜋 𝜀→±0 𝜀 𝜀

If 𝜀 approaches zero from the positive side, the first arctan() function’s argument will
grow toward +∞. Then the function value becomes +𝜋∕2 (see Figure .). The
argument of the second arctan() grows toward −∞ and the tangent value approaches
−𝜋∕2.

𝜕𝜙1 (𝑥, 𝑦 = +0) 𝜎(𝑥) ( ( ) ( ))

𝛿 −𝛿
= lim arctan − arctan
𝜕𝑦 2𝜋 𝜀→+0 𝜀 𝜀
𝜎(𝑥) ( 𝜋 −𝜋 )
= −
2𝜋 2 2
𝜎(𝑥)
= + (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

37.1 Boundary Value Problem 451

For the case where 𝜀 approaches zero from the negative side, the signs reverse for the
arguments of the arctan() functions.
𝜕𝜙1 (𝑥, 𝑦 = −0) 𝜎(𝑥) ( ( ) ( ))
𝛿 −𝛿
= lim arctan − arctan
𝜕𝑦 2𝜋 𝜀→−0 𝜀 𝜀
𝜎(𝑥) ( −𝜋 𝜋 )
= −
2𝜋 2 2
𝜎(𝑥)
= − (.)
2
In summary:
𝜕𝜙1 (𝑥, +0) 𝜎(𝑥) 𝜕𝜙1 (𝑥, −0) 𝜎(𝑥)
𝑣+
1
= = + and 𝑣−
1
= = − (.)
𝜕𝑦 2 𝜕𝑦 2

This result needs some physical interpretation: Normal velocity

induced by line
• On the upper side of the foil with 𝑦 = +0 the source distribution generates an sources
upward oriented transverse disturbance velocity of 𝑣+
1
= +𝜎(𝑥)∕2.
• On the lower side of the foil where 𝑦 = −0 the transverse velocity 𝑣−
1
= −𝜎(𝑥)∕2
is oriented downward with the same magnitude.

Thus the velocity discontinuously changes when crossing from one side of the foil to
the other. The total change in velocity is equal to the source strength 𝜎(𝑥). Although
j this was derived here for a line distribution of sources it is a general feature of source j
distributions.
Finally, we introduce the result (.) into the linearized body boundary condition (.). Exploiting the
body boundary
𝜕𝜙1 d𝑡 condition
= ∓𝑣0 (.)
𝜕𝑦 d𝑥
Therefore, the source strength is proportional to the slope of the thickness distribution:
d𝑡(𝑥)
𝜎(𝑥) = −2𝑣0 (.)
d𝑥
Both the upper and lower sides yield the same result which is expected since we have
only one source distribution.
Figure . shows a symmetric foil section and the distribution of source strength over Source strength
the chord of the foil. The source strength will be negative toward the trailing edge; the distribution
sources are actually sinks. At the position of maximum thickness with d𝑡∕d𝑥 = 0 the
source strength is zero and then grows positive toward the leading edge. If the nose is
round and meets the chord line (𝑦 = 0) at a right angle the source strength distribution
will have a singularity. Clearly, |d𝑡∕d𝑥| → ∞ is in violation of our thin foil assumption.
A correction is needed to represent the flow correctly in the vicinity of the leading edge
(see e.g. Lighthill, ).
With the source strength distribution known, we can restate the disturbance potential.
+𝑐∕2
𝑣 d𝑡(𝜉) (√ )
𝜙1 (𝑥, 𝑦) = − 0 ln (𝑥 − 𝜉)2 + 𝑦2 d𝜉 (.)
𝜋 ∫ d𝜉
−𝑐∕2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

452 37 Thin Foil Theory – Displacement Flow

Figure 37.3 The source strength distribution 𝜎(𝜉) as a function of the slope d𝑡∕d𝑥 of the foil
surface

Since the effect of sources declines with growing distance, the disturbance potential
will satisfy the far field condition of the boundary value problem (.).
Total potential Together with the parallel flow we have a velocity potential for the flow around our thin,
for thickness symmetric foil at zero angle of attack.
problem
j +𝑐∕2 j
𝑣 d𝑡(𝜉) (√ )
Φ(𝑥, 𝑦) = −𝑣0 𝑥 − 0 ln (𝑥 − 𝜉)2 + 𝑦2 d𝜉 (.)
𝜋 ∫ d𝜉
−𝑐∕2

37.2 Pressure Distribution

Bernoulli With the potential (.) the velocity distribution in the domain around the foil is
equation for thin known. The Bernoulli equation for steady flow allows us to compute the pressure
foils
distribution. We ignore the gravity forces assuming that foil and fluid are in equilibrium
when at rest.
1 1
𝜌|∇ Φ|2 + 𝑝 = 𝜌 𝑣20 + 𝑝0 (.)
2 2
The magnitude of the velocity squared |∇ Φ|2 is given by
( )2 ( )2
2 𝜕Φ 𝜕Φ
|∇ Φ| = +
𝜕𝑥 𝜕𝑦
( )2 ( )2
𝜕𝜙 𝜕𝜙1 𝜕𝜙1
= 𝑣20 − 2𝑣0 1 + +
𝜕𝑥 𝜕𝑥 𝜕𝑦
The latter two terms must be very small if the foil is thin because they are a function of
(d𝑡∕d𝑥)2 . These small, quadratic terms will be ignored from hereon forward.
𝜕𝜙1
|∇ Φ|2 = 𝑣20 − 2𝑣0 (.)
𝜕𝑥

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

37.2 Pressure Distribution 453

Substitution of this result into (.) yields

𝜕𝜙1
𝑝 − 𝑝0 = 𝜌 𝑣0 (.)
𝜕𝑥
The pressure coefficient 𝐶𝑝 (𝑥) will be
𝑝 − 𝑝0 2 𝜕𝜙1
𝐶𝑝 (𝑥) = = (.)
1 𝑣0 𝜕𝑥
2
𝜌 𝑣20

The 𝑥-derivative of the disturbance potential (.) is

+𝑐∕2
𝜕𝜙1 𝑣 d𝑡(𝜉) (𝑥 − 𝜉)
= − 0 ( ) d𝜉 (.)
𝜕𝑥 𝜋 ∫ d𝜉 (𝑥 − 𝜉)2 + 𝑦2
−𝑐∕2

The pressure coefficient on the foil surfaces, approximated by 𝑦 = ±0, has to be evalu-
ated carefully because the integrand has a singularity for 𝑥 = 𝜉.
+𝑐∕2
2 d𝑡(𝜉) 1
𝐶𝑝 (𝑥) = − − d𝜉 (.)
𝜋 ∫ d𝜉 (𝑥 − 𝜉)
−𝑐∕2

This is called a Cauchy principal value integral, and indicated by the dash through the
middle of the integral sign.
j In this case, it can be shown that the integral exists. We abbreviate d𝑡∕d𝜉 = 𝑡′ and split Cauchy principal j
the integral into three parts. value integral

+𝑐∕2

− 𝑡′ (𝜉) 1 d𝜉
∫ (𝑥 − 𝜉)
−𝑐∕2
𝑥−𝜀 𝑥+𝜀 𝑐∕2
1 1 1
= 𝑡′ (𝜉) d𝜉 + − 𝑡′ (𝜉) d𝜉 + 𝑡′ (𝜉) d𝜉 (.)
∫ (𝑥 − 𝜉) ∫ (𝑥 − 𝜉) ∫ (𝑥 − 𝜉)
−𝑐∕2 𝑥−𝜀 𝑥+𝜀

The first and last integral on the right-hand side do not contain singularities. The case
𝑥 = 𝜉 is not possible. Therefore, the two integrals will exist if 𝑡′ = d𝑡∕d𝜉 is bounded.
The middle integral in Equation (.) needs further investigation. We assume that
𝜀 is so small that we can extract the derivative of the thickness distribution 𝑡′ (𝜉) from
the integral. It will take the value 𝑡′ (𝑥) at position 𝑥 along the tail–nose line. The
remaining integral is a Cauchy principal value integral, and it has a value, if the limit on
the right-hand side exists.
𝑥+𝜀
⎛ 𝑥−𝛿 𝑥+𝜀
⎞
1 ⎜ 1 1
− d𝜉 = lim d𝜉 + d𝜉 ⎟
∫ (𝑥 − 𝜉) 𝛿→+0 ⎜ ∫ (𝑥 − 𝜉) ∫ (𝑥 − 𝜉) ⎟
𝑥−𝜀 ⎝𝑥−𝜀 𝑥+𝛿 ⎠
⎛ 𝑥−𝛿 𝑥+𝜀
⎞
1 1
= lim ⎜ d𝜉 − d𝜉 ⎟ (.)
𝛿→+0 ⎜ ∫ (𝑥 − 𝜉) ∫ (𝜉 − 𝑥) ⎟
⎝𝑥−𝜀 𝑥+𝛿 ⎠

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

454 37 Thin Foil Theory – Displacement Flow

During this conversion the operands switched position in the denominator of the
second integral. The resulting minus sign has been extracted from the integral. This
avoids negative arguments in the integrand, since 𝜉 > 𝑥 for the second integral. This is
helpful in the subsequent integration because the natural logarithm is undefined for
negative arguments.
The remaining integrals can now be solved.
𝑥+𝜀 ([ ]𝑥−𝛿 [ ]𝑥+𝜀 )
− 1
d𝜉 = lim − ln(𝑥 − 𝜉) − ln(𝜉 − 𝑥)
∫ (𝑥 − 𝜉) 𝛿→+0 𝑥−𝜀 𝑥+𝛿
𝑥−𝜀

The additional minus sign in the first antiderivative results from the inner derivative.
Finally, we substitute the limits for the integration variable 𝜉:
𝑥+𝜀
([ ] [ ])
1
− d𝜉 = lim − ln(𝛿) + ln(𝜀) − ln(𝜀) − ln(𝛿) (.)
∫ (𝑥 − 𝜉) 𝛿→+0
𝑥−𝜀
= lim (− ln(𝛿) + ln(𝛿) + ln(𝜀) − ln(𝜀))
𝛿→+0
( )
= lim 0
𝛿→+0
= 0

j The limit for 𝛿 → 0 obviously exists. The value of the integral is zero, i.e. the middle j
integral in Equation (.) makes no contribution to the pressure coefficient. Thus we
have proven that the Cauchy principal value integral for the pressure coefficient (.)
exists, if the derivative of the thickness distribution is well behaved.

37.3 Elliptical Thickness Distribution

Example: As an example, we investigate the elliptical thickness distribution.
elliptical foil
√
section ( )2 [ ]
𝑡max 2𝜉 𝑐 𝑐
𝑡(𝜉) = 1− for 𝜉 ∈ − , (.)
2 𝑐 2 2

Elliptical thickness distributions are not often used, although they are less prone to
cavitation than other profile shapes.
Variable Before substituting the elliptical thickness distribution (.) into the integral for the
substitution pressure coefficient (.), we initiate a variable substitution. The integration variable
𝜉 will be replaced by a polar coordinate 𝜑 such that

𝑐 d𝜉 𝑐 𝑐
𝜉 = cos(𝜑) = − sin(𝜑) d𝜉 = − sin(𝜑) d𝜑 (.)
2 d𝜑 2 2
The integration limits become:
𝑐 𝑐
𝜉𝑙 = − ⟶ 𝜑𝑙 = 𝜋 𝜉𝑢 = ⟶ 𝜑𝑢 = 0 (.)
2 2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

37.3 Elliptical Thickness Distribution 455

The point (𝑥, 0), where we evaluate the pressure coefficient along the foil, is transformed
in the same way, but it has its own angle 𝜃, independent of the integration variable.
𝑐
𝑥 = cos(𝜃) (.)
2
with 𝜃 ∈ [0, 𝜋].
Over the 𝜑-axis, the elliptical thickness distribution (.) is represented by a sine
curve.
𝑡 √ 𝑡
𝑡(𝜑) = max 1 − cos2 (𝜑) = max sin(𝜑) (.)
2 2

We also need the derivative of the thickness distribution with respect to the integration Derivative of
variable 𝜉. thickness
distribution
d𝑡 d𝑡 d𝜑 𝑡 1
= = max cos(𝜑) ( )
d𝜉 d𝜑 d𝜉 2 d𝜉
d𝜑

and with the derivative d𝜉∕d𝜑 from (.)

d𝑡 𝑡 cos(𝜑)
= max ( )
d𝜉 2 𝑐
− sin(𝜑)
2
j 𝑡 cos(𝜑) j
= − max (.)
𝑐 sin(𝜑)

We introduce the variable substitution (.), the transformed field point coordinate Pressure
(.), and the derivative of the thickness distribution (.) into Equation (.) coefficient
for the pressure coefficient.

0
𝑡 cos(𝜑) ( )
2 1 𝑐
𝐶𝑝 (𝜃) = − − − max [ ] − sin(𝜑) d𝜑 (.)
𝜋∫ 𝑐 sin(𝜑) 𝑐 cos(𝜃) − 𝑐 cos(𝜑) 2
𝜋 2 2
We carefully consolidate the three minus signs into one and extract all constants from
the integral.
0
2𝑡 cos(𝜑) sin(𝜑)
𝐶𝑝 (𝜃) = − max − [ ] d𝜑 (.)
𝜋 𝑐 ∫ sin(𝜑) cos(𝜃) − cos(𝜑)
𝜋

Next, we eliminate the sin(𝜑) in numerator and denominator and extract a factor −1
from the denominator which we convert into a reversal of the integration limits.
𝜋
2𝑡 cos(𝜑)
𝐶𝑝 (𝜃) = − max − [ ] d𝜑 (.)
𝜋 𝑐 ∫ cos(𝜑) − cos(𝜃)
0

The Cauchy principal value integral in equation (.) is not quite so easy to solve.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

456 37 Thin Foil Theory – Displacement Flow

Glauert integrals Luckily for us, Glauert (, pp. –) solved this as part of a whole group of
integrals, which are known in hydrodynamics as Glauert integrals.
𝜋
cos(𝑛𝜑) sin(𝑛𝜃)
𝐼𝑛 (𝜃) = − [ ] d𝜑 = 𝜋 (.)
∫ cos(𝜑) − cos(𝜃) sin(𝜃)
0

In our case 𝑛 = 1, and the value of the integral 𝐼1 = 𝜋. For the pressure coefficient of
an elliptical thickness distribution we obtain a constant negative value:
2 𝑡max
𝐶𝑝 (𝜃) = − (.)
𝑐
This is somewhat surprising because we expect to see the value 𝐶𝑝 = 1 at the aft
and forward stagnation points. But then, close to the stagnation points our foil shape
violates the primary condition of thin foil theory. At the rounded ends of the elliptical
foil section, the derivative 𝜕𝑓
𝜕𝑥
𝜕𝑡
= 𝜕𝑥 of the shape is anything but small. In fact, it does
not exist and its values approach ±∞.
Examples Overall, the pressure distribution matches the exact potential theory result obtained by
conformal mapping quite well. Figure . shows the results for an elliptical foil with
% thickness to chord length ratio. The minimum 𝐶𝑝 value is slightly under predicted
by thin foil theory, but it is close to its conformal mapping value. As expected, thin foil
theory fails to predict the correct pressure toward trailing and leading edges of the foil.

j One could argue that a foil with 𝑡max ∕𝑐 = 0.10 is not really a thin foil. Figure . shows j
the result for an elliptical foil section with a thickness to chord length ratio of just %.
The minimum 𝐶𝑝 value is fairly accurate and the deviations are much smaller toward
leading and trailing edges.
Figures . and . also show the velocity magnitude as derived by conformal map-
ping. If one plots the magnitude of the velocity derived from thin foil theory, huge
discrepancies are found toward the ends of the foil if the transverse velocity component
𝑣1 = 𝜕𝜙1 ∕𝜕𝑦 is included. As listed in Equation (.), 𝑣1 is a function of the slope
of the thickness distribution. At the ends of the foil, the slope is ±∞ which renders
the velocity prediction as incorrect. Hence, it is common practice to compute the
magnitude of the velocity as in Equation (.) without the parts 𝑢21 and 𝑣21 .
√
2 𝜕𝜙1
|∇ Φ| = 𝑣0 1 − (.)
𝑣0 𝜕𝑥

References
Glauert, H. (). The elements of airofoil and airscrew theory. Cambridge University
Press, Cambridge, United Kingdom, second edition.
Lighthill, M. (). A new approach to thin foil theory. The Aeronautical Quarterly,
():–.
 Hermann Glauert (* – †), British aerodynamicist

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

37.3 Elliptical Thickness Distribution 457

j j
Figure 37.4 Comparison of thin foil theory and conformal mapping (exact) pressure coefficient
for an elliptical foil with thickness to chord length ratio of 𝑡max ∕𝑐 = 0.10

Self Study Problems

. Summarize the equations that define the linearized boundary value problem for
the displacement flow. State the physical meaning of each equation.

. Given is a line distribution of sources along the tail–nose line with strength 𝜎(𝑥).
By how much changes the transverse velocity 𝑣 as you move from the underside
(𝑦 = 0− ) to the upper side of the source distribution (𝑦 = 0+ )?

. Explain in your own words why the source strength at the leading edge is un-
bounded in Figure ..

. Compute and plot the pressure coefficient over the chord length of an elliptical
foil with 7% thickness–chord length ratio. Use the thin foil approximation.

. Compute the value of the following integral for 𝜃 = 𝜋∕3:

𝜋
cos(3𝜑)
𝐼3 (𝜃) = − [ ] d𝜑 (.)
∫ cos(𝜑) − cos(𝜃)
0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

458 37 Thin Foil Theory – Displacement Flow

j j
Figure 37.5 Comparison of thin foil theory and conformal mapping (exact) pressure coefficient
for an elliptical foil with thickness to chord length ratio of 𝑡max ∕𝑐 = 0.05

. Derive the velocity distribution induced on a thin foil (𝑥 ∈ [−𝑐∕2, 𝑐∕2]) by a

parabolic thickness distribution
[ ( )2 ]
𝑡 2𝑥 𝑐 𝑐
𝑡(𝑥) = max 1 − for − ≤ 𝑥 ≤
2 𝑐 2 2

The parabolic thickness distribution is used to approximate the flow around ogival
foil sections. Compare the velocity and pressure distribution with the elliptical
thickness distribution.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

459

38
Thin Foil Theory – Lifting Flow

The remaining task in solving the flow around thin foils is to find expressions for the
disturbance potentials 𝜙2 and 𝜙3 which describe the lifting flow. We will use the results
to study basic properties of lifting foils. First, we define a combined boundary value
problem for lifting flow which will be solved for the combined lifting flow potential 𝜙23 .

𝜙23 = 𝜙2 + 𝜙3 (.)

The disturbance potential 𝜙23 represents the effects of camber and angle of attack on
the flow. The solution is based on a series expansion developed by Glauert ().

j Learning Objectives j

At the end of this chapter students will be able to:

• understand and apply line distributions of vortices to simulate lifting flows

• comprehend the physical meaning of the Kutta condition
• compute lift coefficients and ideal angle of attack for thin foils

38.1 Lifting Foil Problem

The lifting problem considers an infinitely thin plate with camber operating at an Boundary value
angle of attack 𝛼 (Figure .). We will solve it under the assumption that the fluid is problem
incompressible, inviscid, and irrotational in simply connected regions. Accordingly, we
may apply potential flow theory. The total velocity potential consists of parallel flow
and the disturbance potential 𝜙23 since the thickness is assumed to be zero 𝑡(𝑥) = 0 for
the lifting problem.

Φ(𝑥, 𝑦) = 𝜙∞ + 𝜙23 = −𝑣0 cos(𝛼) + 𝑣0 sin(𝛼) + 𝜙23 (.)

The generation of lift requires a nonzero circulation Γ ≠ 0. Therefore, the flow must
have a component which rotates around the foil (see also Figure .). Obviously, all
velocity potentials must satisfy the Laplace equation.

𝚫Φ = 0 and 𝚫𝜙23 = 0 (.)

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

460 38 Thin Foil Theory – Lifting Flow

Figure 38.1 Boundary value problem for an infinitely thin cambered plate at angle of attack 𝛼

The body boundary condition for the foil contour consists of the parts of Equation (.)
which involve camber 𝑓𝑚 and angle of attack 𝛼.

𝜕𝜙±
23 𝜕𝑓𝑚
= − 𝑣 − 𝛼 𝑣0 for 𝑦 = ±0 (.)
𝜕𝑦 𝜕𝑥 0

As consequence of our thin foil assumptions, the body boundary condition is satisfied
along the 𝑥-axis between trailing and leading edges 𝑥 ∈ [−𝑐∕2, 𝑐∕2]. The disturbance
j j
of the parallel flow caused by the foil must vanish far away from the body. This can be
expressed as

𝜙23 = const. for 𝑟 ⟶ ∞ or ∇ 𝜙23 = 0 for 𝑟 ⟶ ∞. (.)

Kutta condition In addition to the body boundary and far field conditions, we also have to implement
the Kutta condition. The Kutta condition enforces the physical observation that the flow
is leaving the trailing edge smoothly. If, hypothetically, the fluid would flow around the
sharp trailing edge, it would create infinitely high flow velocities. The Kutta condition
may be enforced in different ways.

• One way to enforce the Kutta condition is to request equal pressure on upper
and lower sides of the foil right at the trailing edge i.e. 𝑝+ = 𝑝− at 𝑥 = −𝑐∕2.
Differences in pressure would force the fluid to flow around the trailing edge
from the high pressure side to the side with lower pressure.

• Alternatively, we can impose the Kutta condition as a kinematic requirement, i.e.

the velocity vector at the trailing edge must be equal for upper and lower sides
𝑣+ = 𝑣− at 𝑥 = −𝑐∕2.

Lifting flow We know from our first discussion of the lift force (page ) that a circulation is needed
disturbance to create faster flow speeds on the upper side and slower flow speeds on the lower side of
potential
the foil. In essence, a flow pattern, like the one created by a vortex, must be included in
the flow. A distribution of sources has been used to represent the disturbance potential
associated with the thickness distribution. Likewise, the disturbance potential for the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

38.1 Lifting Foil Problem 461

lifting flow may be based on a distribution of vertices. With the potential of a D-vortex
from Chapter , we define the disturbance potential as
+𝑐∕2 ( )
𝛾(𝜉) 𝑦
𝜙23 (𝑥, 𝑦) = arctan d𝜉 (.)
∫ 2𝜋 𝑥−𝜉
−𝑐∕2

Note that the coordinate 𝜉 marks the position of a vortex. Its effect is probed at the
field point (𝑥, 𝑦). Unknown in the potential is the vortex strength 𝛾(𝜉). By design,
the disturbance potential 𝜙23 will satisfy the Laplace equation because the integral is
a combination of many vortex potentials which all satisfy the Laplace equation. The
potential (.) also satisfies the far field condition because the effect of a vortex vanishes
far away from its origin.
Before we solve the boundary value problem, we take a look at the velocity field defined Velocity field of a
by the disturbance potential 𝜙23 (.). The velocity vector is equal to the gradient of line vortex
distribution
the potential.
⎛ 𝜕𝜙23 ⎞ ( )
⎜ 𝜕𝑥 ⎟ 𝑢23
𝑣23 = ∇ 𝜙23 = ⎜ (.)
𝜕𝜙23 ⎟
=
⎜ ⎟ 𝑣23
⎝ 𝜕𝑦 ⎠

For the horizontal velocity component we have to compute the 𝑥-derivative of Equa- Horizontal
tion (.). velocity
j +𝑐∕2 ( ) component 𝑢23 j
𝜕𝜙23 𝜕 𝛾(𝜉) 𝑦
= arctan d𝜉 (.)
𝜕𝑥 𝜕𝑥 ∫ 2𝜋 𝑥−𝜉
−𝑐∕2

The integration limits are constant and the integration variable is 𝜉, i.e. the position of
the vortex element rather than the location 𝑥 of the field point. Hence, differentiation
and integration may be interchanged. In the integrand, only the inverse tangent function
arctan() is a function of 𝑥.
+𝑐∕2 [ ( )]
𝜕𝜙23 𝛾(𝜉) 𝜕 𝑦
= arctan d𝜉 (.)
𝜕𝑥 ∫ 2𝜋 𝜕𝑥 𝑥−𝜉
−𝑐∕2

The first order derivative of the inverse tangent function is equal to

d 1
arctan(𝑧) =
d𝑧 1 + 𝑧2
Applying this and the chain rule for 𝑧 = 𝑦∕(𝑥 − 𝜉) yields
+𝑐∕2 ( )
𝜕𝜙23 𝛾(𝜉) −𝑦
𝑢± (𝑥, 𝑦) = = [ ] d𝜉 (.)
23 𝜕𝑥 ∫ 2𝜋 (𝑥 − 𝜉)2 + 𝑦2
−𝑐∕2

Except for the minus sign, and the vortex strength 𝛾(𝜉) standing in for the source
strength, this integral is identical to the 𝑦-derivative of the source distribution (.)
from the preceding chapter.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

462 38 Thin Foil Theory – Lifting Flow

Between trailing edge (𝑥 = −𝑐∕2) and leading edge (𝑥 = 𝑐∕2), the integrand in Equa-
tion (.) vanishes for 𝑦 = ±0, except for the point 𝜉 = 𝑥. To evaluate the resulting
undefined expression, we exercise the same limiting process used in Equations (.)
through (.). As a result, the horizontal velocity on the 𝑥-axis 𝑦 = ±0 is equal to
half the vortex strength.
𝜕𝜙23 (𝑥, 𝑦 = ±0) 𝛾(𝑥)
𝑢±
23
(𝑥, 𝑦 = ±0) = = ∓ for 𝑥 ∈ [−𝑐∕2, 𝑐∕2] (.)
𝜕𝑥 2
On the upper side of the foil 𝑦 = +0, the additional horizontal velocity 𝑢+
23
is pointing in
negative 𝑥-axis direction and, thus, augments the horizontal onflow −𝑣0 cos 𝛼. On the
pressure side of the foil, 𝑢−
23
points in positive 𝑥-axis direction and reduces the overall
horizontal flow speed.
Across a line source strength distribution, the vertically induced velocity component
jumps by the source strength 𝜎. If we pass from the lower side to the upper side of a
line vortex distribution, the horizontal velocity changes by the negative of the vortex
strength 𝛾.
Kutta condition Equation (.) bears some significance for the Kutta condition. Since the vortex
revisited distribution induces horizontal velocities which point in opposite directions on the
upper and lower foil surfaces, a Kutta condition must enforce that the vortex strength
vanishes at the trailing edge.
𝛾(𝑥 = −𝑐∕2) = 0 Kutta condition (.)
j Otherwise, we cannot be sure that the flow velocity at the trailing edge is equal on the j
upper and lower sides as the Kutta condition requests.
Transverse The transverse velocity component follows from the 𝑦-derivative of Equation (.).
velocity
component 𝑣23 +𝑐∕2 ( )
𝜕𝜙23 𝜕 𝛾(𝜉) 𝑦
= arctan d𝜉 (.)
𝜕𝑦 𝜕𝑦 ∫ 2𝜋 𝑥−𝜉
−𝑐∕2

Again, interchanging differentiation and integration yields together with the basic arctan
derivative and the chain rule the transverse disturbance velocity
+𝑐∕2 ( )
𝛾(𝜉) 𝑥−𝜉
𝑣23 = d𝜉 (.)
∫ 2𝜋 (𝑥 − 𝜉)2 + 𝑦2
−𝑐∕2

This is the same type of integral which has been found for the 𝑥 derivative of the source
distribution (.).
The body boundary condition is evaluated on the 𝑥-axis. From Equation (.) follows
a Cauchy principal value integral for 𝑦 = ±0.
+𝑐∕2
𝛾(𝜉) 1
𝑣23 (𝑥, 𝑦 = ±0) = − d𝜉 (.)
∫ 2𝜋 (𝑥 − 𝜉)
−𝑐∕2

For 𝑥 = 𝜉 the integrand does not exist, but on pages  and  it has been shown
that the integral has a finite value, as long as the vortex strength is bounded.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

38.2 Glauert’s Classical Solution 463

A1
1 + cos( )

A1 sin( )
sin( )
A0

A0
0 0
0 /2 0 /2

A2 A3
A2 sin(2 )

A3 sin(3 )

0 0

−A2 −A3
0 /2 0 /2
position along chord position along chord
j j
Figure 38.2 The first four elements of Glauert’s trigonometric series for the vortex strength 𝛾

38.2 Glauert’s Classical Solution

The next step is to compute the unknown vortex strength. Since our disturbance Exploiting the
potential satisfies all but the body boundary condition, we can only employ the latter body boundary
condition
to find the vortex strength as a function of foil geometry. Combining Equations (.)
and (.) links the vortex strength with camber distribution and angle of attack.

+𝑐∕2
𝛾(𝜉) 1 𝜕𝑓
− d𝜉 = − 𝑚 𝑣0 − 𝛼 𝑣0 (.)
∫ 2𝜋 (𝑥 − 𝜉) 𝜕𝑥
−𝑐∕2

At this point, we may replace the partial derivative of the camber distribution with an
ordinary derivative because only 𝑥 remains as variable.
Glauert () proposed a trigonometric series to approximate the unknown vortex Glauert’s series
strength. for vortex
[ ] strength
1 + cos(𝜑) ∑
∞
𝛾(𝜑) = 2 𝑣0 𝐴0 + 𝐴𝑛 sin(𝑛𝜑) (.)
sin(𝜑) 𝑛=1

Figure . shows the shape of the first four elements of the series expansion (.).
Note that the series implicitly satisfies the Kutta condition 𝛾(𝑥 = −𝑐∕2) = 0. All

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

464 38 Thin Foil Theory – Lifting Flow

functions in the series vanish at the trailing edge 𝜑 = 𝜋.

𝛾(𝜑 = 𝜋) = 0 Kutta condition (.)

Glauert’s series (.) uses the same variable transformation as was employed in Sec-
tion . for the elliptical thickness distribution. The integration variable 𝜉 is replaced
by a polar coordinate 𝜑 such that

𝑐 d𝜉 𝑐 𝑐
𝜉 = cos(𝜑) = − sin(𝜑) d𝜉 = − sin(𝜑) d𝜑 (.)
2 d𝜑 2 2
Consequently, the integration limits become:
𝑐 𝑐
𝜉𝑙 = − ⟶ 𝜑𝑙 = 𝜋 𝜉𝑢 = ⟶ 𝜑𝑢 = 0 (.)
2 2
The leading edge is 𝜑 = 0 and the trailing edge is 𝜑 = 𝜋. The field point (𝑥, 0) is
rewritten as
𝑐
𝑥 = cos(𝜃) (.)
2
The position angle 𝜃 also takes values from the interval [0, 𝜋].
Body boundary Substituting (.), (.), and (.) into Equation (.) converts the body bound-
condition for ary condition into the following form:
lifting flow
𝜋
j 𝛾(𝜑) sin(𝜑) d𝑓𝑚 j
− − ( ) d𝜑 = − 𝑣 − 𝛼 𝑣0 (.)
∫ 2𝜋 cos(𝜑) − cos(𝜃) d𝑥 0
0

We replace the vortex strength 𝛾 with its series representation (.), and divide the
equation by the onflow velocity 𝑣0 .
𝜋 [ ]
1 + cos(𝜑) ∑
∞
1 − sin(𝜑)
− 𝐴0 + 𝐴𝑛 sin(𝑛𝜑) ( ) d𝜑
𝜋 ∫ sin(𝜑) 𝑛=1 cos(𝜑) − cos(𝜃)
0
d𝑓𝑚
= − −𝛼 (.)
d𝑥

The goal is to define the unknown series coefficients 𝐴𝑛 as functions of the slope of the
camber line and the angle of attack. To this end, the integral on the left-hand side of
Equation (.) is broken up into individual integrals, which each contains only one
unknown coefficient 𝐴𝑛 .
𝜋 [ ]
1 + cos(𝜑) ∑
∞
1 sin(𝜑)
𝐼 = − − 𝐴0 + 𝐴𝑛 sin(𝑛𝜑) ( ) d𝜑
𝜋 ∫ sin(𝜑) 𝑛=1 cos(𝜑) − cos(𝜃)
0

∑
∞
= 𝐼𝑛 = 𝐼0 + 𝐼1 + 𝐼2 + … (.)
𝑛=0

Integral 𝐼0 The first integral involves only the coefficient 𝐴0 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

38.2 Glauert’s Classical Solution 465

𝜋 ( )
1 − 1 + cos(𝜑) sin(𝜑)
𝐼0 = − 𝐴 ( ) d𝜑
𝜋 ∫ 0 sin(𝜑) cos(𝜑) − cos(𝜃)
0
𝜋
𝐴 1 + cos(𝜑)
= − 0 − ( ) d𝜑
𝜋 ∫ cos(𝜑) − cos(𝜃)
0
𝜋 𝜋
𝐴0 ⎡⎢ 1 cos(𝜑)
⎤
= − − ( ) d𝜑 + − ( ) d𝜑⎥
𝜋 ⎢∫ cos(𝜑) − cos(𝜃) ∫ cos(𝜑) − cos(𝜃) ⎥
⎣0 0 ⎦
The two resulting integrals are both Glauert integrals (.) (see Section .) and
result in 0 and 𝜋 respectively.
𝐴0 [ ]
𝐼0 = −
0 + 𝜋 = −𝐴0 (.)
𝜋
That is a surprisingly simple result. Of course, it is connected to the specific choice of
function which accompanies the coefficient 𝐴0 in the series expansion.
The remaining integrals can also be reduced to Glauert integrals. Integrals 𝐼𝑛 ,
𝑛>0
𝜋
1 [ ] sin(𝜑)
𝐼𝑛 = − − 𝐴𝑛 sin(𝑛𝜑) ( ) d𝜑
𝜋 ∫ cos(𝜑) − cos(𝜃)
0
𝜋
j 𝐴 sin(𝑛𝜑) sin(𝜑) j
= − 𝑛 − ( ) d𝜑
𝜋 ∫ cos(𝜑) − cos(𝜃)
0

Further progress needs the help of a trigonometric identity:

[ ( ) ( )]
1
sin(𝑛𝜑) sin(𝜑) = cos (𝑛 − 1)𝜑 − cos (𝑛 + 1)𝜑 (.)
2
This one is hardly in any books but can be derived by using Euler’s formulas:
ei𝑥 − e−i𝑥 ei𝑥 + e−i𝑥
sin(𝑥) = and cos(𝑥) = (.)
2𝑖 2
Employing the formula for the sine function with arguments 𝑛𝜑 and 𝜑 yields:
( ) ( )
1 i𝑛𝑥 1 i𝑥
sin(𝑛𝜑) sin(𝜑) = e − e−i𝑛𝑥 e − e−i𝑥
2i 2i
1 [ i(𝑛+1)𝑥 i(𝑛−1)𝑥
]
= − e −e − e−i(𝑛−1)𝑥 + e−i(𝑛+1)𝑥
4[ ]
1 ei(𝑛+1)𝑥 + e−i(𝑛+1)𝑥 ei(𝑛−1)𝑥 + e−i(𝑛−1)𝑥
= − −
2 2 2
The formula for the cosine allows us to transform the expression back into the real
domain which proves the identity (.).
1[ ( ) ( )]
sin(𝑛𝜑) sin(𝜑) = − cos (𝑛 + 1)𝜑 − cos (𝑛 − 1)𝜑
2
1[ ( ) ( )]
= cos (𝑛 − 1)𝜑 − cos (𝑛 + 1)𝜑
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

466 38 Thin Foil Theory – Lifting Flow

With the help of Equation (.) the integral 𝐼𝑛 is split into two Glauert integrals.
𝜋 ( ) 𝜋 ( )
𝐴𝑛 ⎡⎢ cos (𝑛 − 1)𝜑 cos (𝑛 + 1)𝜑 ⎤
𝐼𝑛 = − − ( ) d𝜑 − − ( ) d𝜑⎥
2𝜋 ⎢∫ cos(𝜑) − cos(𝜃) ∫ cos(𝜑) − cos(𝜃) ⎥
⎣0 0 ⎦
[ ( ) ( )]
𝐴 sin (𝑛 − 1)𝜃 sin (𝑛 + 1)𝜃
= − 𝑛 −
2 sin(𝜃) sin(𝜃)

These, in turn, can be reduced to

[ ]
cos(𝑛𝜃) sin(𝜃)
𝐼𝑛 = 𝐴𝑛 = 𝐴𝑛 cos(𝑛𝜃) (.)
sin(𝜃)

This conversion employs the trigonometric identity:

( ) ( )
sin (𝑛 − 1)𝜃 − sin (𝑛 + 1)𝜃 = −2 cos(𝑛𝜃) sin(𝜃) (.)

Series Collecting the results from Equations (.) and (.), the integral 𝐼 (.) is
coefficients reduced to
∑∞ ∑
∞
𝐼 = 𝐼0 + 𝐼𝑛 = −𝐴0 + 𝐴𝑛 cos(𝑛𝜃) (.)
𝑛=1 𝑛=1
j j
Going back to the body boundary condition in its form (.), the integral on the
left-hand side is replaced with the right-hand side of Equation (.).

∑
∞
d𝑓𝑚
− 𝐴0 + 𝐴𝑛 cos(𝑛𝜃) = − −𝛼 (.)
𝑛=1
d𝑥

This is a purely geometric relationship, which may be exploited to compute the coeffi-
cients 𝐴𝑛 with 𝑛 = 0, 1, 2, … .
Coefficient 𝐴0 In order to gain explicit equations for the coefficients, we first multiply Equation (.)
with cos(𝑘𝜃) on both sides.

∑
∞
d𝑓𝑚
− 𝐴0 cos(𝑘𝜃) + 𝐴𝑛 cos(𝑛𝜃) cos(𝑘𝜃) = − cos(𝑘𝜃) − 𝛼 cos(𝑘𝜃) (.)
𝑛=1
d𝑥

𝑘 is an arbitrary integer. Next, we formally integrate both sides with respect to the
position of the field point 𝜃 from leading edge 𝜃 = 0 to trailing edge 𝜃 = 𝜋.

𝜋 𝜋
∑
∞
− 𝐴0 cos(𝑘𝜃) d𝜃 + 𝐴𝑛 cos(𝑛𝜃) cos(𝑘𝜃) d𝜃
∫ ∫
𝑛=1
0 0
𝜋 𝜋
d𝑓𝑚
= − cos(𝑘𝜃) d𝜃 − 𝛼 cos(𝑘𝜃) d𝜃 (.)
∫ d𝑥 ∫
0 0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

38.2 Glauert’s Classical Solution 467

Equation (.) contains two integrals which may be solved directly:

𝜋 {
𝜋 for 𝑘 = 0
cos(𝑘𝜃) d𝜃 = (.)
∫ 0 for 𝑘 > 0
0
𝜋 {𝜋
for 𝑘 = 𝑛
cos(𝑛𝜃) cos(𝑘𝜃) d𝜃 = 2 (.)
∫ 0 for 𝑘 ≠ 𝑛
0

If one chooses 𝑘 = 0, our thin foil body boundary condition (.) for the lifting Coefficient 𝐴0
problem becomes
𝜋
∑
∞
d𝑓𝑚
− 𝐴0 𝜋 + 𝐴𝑛 0 = − d𝜃 − 𝛼 𝜋 (.)
∫ d𝑥
𝑛=1
0

The integrals in the sum are all zero because 𝑘 = 0 and 𝑛 ≥ 1 and, as a consequence,
𝑘 ≠ 𝑛. This effectively eliminates all coefficients but 𝐴0 , and we obtain
𝜋
1 d𝑓𝑚
𝐴0 = 𝛼 + d𝜃 (.)
𝜋 ∫ d𝑥
j 0 j

For any integer 𝑘 > 0, Equation (.) reduces to Coefficients 𝐴𝑛

𝜋
𝜋 d𝑓𝑚
− 𝐴0 0 + 𝐴𝑘 = − cos(𝑘𝜃) d𝜃 − 𝛼 0 (.)
2 ∫ d𝑥
0

or
𝜋
2 d𝑓𝑚
𝐴𝑛 = − cos(𝑛𝜃) d𝜃 for 𝑛 = 1, 2, 3, … (.)
𝜋 ∫ d𝑥
0

Thus, for a known foil geometry and angle of attack we can compute as many coeffi-
cients 𝐴𝑛 as we need to accurately define the disturbance potential 𝜙23 for the lifting
flow. Equations (.) and (.) may be utilized in numerical approximations. If
the foil mean line is discretized, slope d𝑓𝑚 ∕d𝑥 may be approximated by finite differ-
ences between offsets. Subsequently, the integrals can be solved numerically with the
trapezoidal rule or possibly more accurate methods.

References
Glauert, H. (). The elements of airofoil and airscrew theory. Cambridge University
Press, Cambridge, United Kingdom, second edition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

468 38 Thin Foil Theory – Lifting Flow

Self Study Problems

. Starting with Equation (.), explicitly derive the horizontal velocity

𝛾(𝑥)
𝑢23 (𝑥, 𝑦 = ±0) = ∓
2
induced by the line distribution of vortices on the upper and lower side of the
𝑥-axis between trailing and leading edge.
. Given is a line distribution of vortices along the tail–nose line with strength 𝛾(𝑥).
By how much changes the horizontal velocity 𝑢 as you move from the underside
(𝑦 = 0− ) to the upper side of the vortex distribution (𝑦 = 0+ )?

. Prove the trigonometric identity (.) using Euler’s formula (.).

. You are part of a survivor team and as the team’s naval architect you are tasked
with building a hydrofoil supported boat. You have three straight steel plates that
you can weld into the crude foil shape shown below. The function 𝑓𝑚 describes
the mean line and shape of the foil.

j j

Using thin foil theory results, estimate the section lift coefficient 𝐶𝑙 as a function
of the maximum camber ratio 𝑓0 ∕𝑐 and the angle of attack 𝛼. Hint: integrals can
be split up into integrals over parts of the integration range.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

469

39
Thin Foil Theory – Lifting Flow
Properties

With the coefficients 𝐴𝑛 (𝑛 = 0, 1, 2, … ) known, the vortex strength may be computed

with Equation (.) at all points along the chord of a foil. That in turn defines the
disturbance potential 𝜙23 for the lifting flow. We are most interested in the lift force
acting on the foil. This chapter exploits the solutions of the thin foil boundary value
problems to define and quantify important lifting foil properties.

Learning Objectives
j j
At the end of this chapter students will be able to:

• understand pressure distributions and shock free entry

• compute lift coefficients
• compute the ideal angle of attack

39.1 Lift Force and Lift Coefficient

In our discussion of the flow over a symmetric thin foil (see Section .), we derived Linearized
the pressure difference between local pressure 𝑝 and the reference 𝑝0 from a linearized Bernoulli
equation
Bernoulli equation.
𝜕𝜙1
𝑝 − 𝑝0 = 𝜌 𝑣0 displacement flow (.)
𝜕𝑥
Following the same arguments that led to Equation (.), it may be concluded that,
in the lifting flow problem, the pressure difference is a function of the horizontal
disturbance velocity 𝜕𝜙23 ∕𝜕𝑥.
𝜕𝜙23
𝑝 − 𝑝0 = 𝜌 𝑣0 lifting flow (.)
𝜕𝑥

Within the limits of thin foil theory, the horizontal velocity 𝑢23 induced on the foil is Pressure over the
foil contour
Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

470 39 Thin Foil Theory – Lifting Flow Properties

equal to half the vortex strength.

𝜕𝜙23 𝛾(𝑥)
𝑢±
23
(𝑥, 𝑦 = ±0) = = ∓ for 𝑥 ∈ [−𝑐∕2, 𝑐∕2] (.)
𝜕𝑥 2
Based on this result, the pressure on upper and lower foil surface is derived from
Equation (.)
𝛾
𝑝+ − 𝑝0 = −𝜌 𝑣0 for 𝑦 = +0 (.)
2
𝛾
−
𝑝 − 𝑝0 = +𝜌 𝑣0 for 𝑦 = −0 (.)
2
The difference in pressure between the lower and upper sides becomes

Δ𝑝 = 𝑝− − 𝑝+ = +𝜌 𝑣0 𝛾 for 𝑦 = 0 (.)

Pressure A division of the equations above by the dynamic pressure yields the pressure coefficients
coefficients for upper and lower foil side as well as the pressure difference.

𝑝+ − 𝑝0 𝛾
𝐶𝑝+ = = − for 𝑦 = +0 (.)
1 2 𝑣0
𝜌𝑣
2 0
−
𝑝 − 𝑝0 𝛾
j 𝐶𝑝− = = + for 𝑦 = −0 (.) j
1 2 𝑣0
𝜌𝑣
2 0
𝛾
Δ𝐶𝑝 = 𝐶𝑝− − 𝐶𝑝+ = +2 for 𝑦 = 0 (.)
𝑣0

Lift force from Integrating the pressure difference along the foil yields the lift force.
pressure
𝑐∕2

𝐹𝑝 = Δ𝑝 𝑛 d𝜉 (.)
∫
−𝑐∕2

The normal vector 𝑛 is derived from the slope of the camber line (see Figure .).

⎛ − d𝑓𝑚 ⎞
1 ⎜ d𝑥 ⎟
𝑛 = √ (.)
( )2 ⎜ ⎟
d𝑓𝑚 ⎝ 1 ⎠
1+
d𝑥

Force on flat Consider the case of a flat plate. Its camber is zero and the normal vector simplifies to
plate 𝑛 = (0, 1)𝑇 . The 𝑥-component of the pressure force vanishes, and the 𝑦-component is
perpendicular to the plate.
𝑐∕2

𝐹𝑦 = Δ𝑝 d𝜉 (.)
∫
−𝑐∕2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.1 Lift Force and Lift Coefficient 471

Figure 39.1 The effect of leading edge

suction for a thin plate at angle of attack 𝛼

Introducing Equation (.) yields

𝑐∕2

𝐹𝑦 = 𝜌 𝑣0 𝛾(𝜉) d𝜉 (.)
∫
−𝑐∕2

If the plate operates at angle of attack 𝛼, we find a lift force d𝐿̃ perpendicular to the
onflow and a drag force d𝐷̃ parallel to the onflow 𝑣0 .
j 𝐷̃ = 𝐹𝑦 sin(𝛼) (.) j
𝐿̃ = 𝐹𝑦 cos(𝛼) (.)

This appears to be in contradiction to Kutta-Joukowsky’s lift theorem. For inviscid flow, Leading edge
we expect the drag force to vanish, and the lift force is supposed to be perpendicular to suction
the onflow. However, we derived a pressure force (.) which acts perpendicular to
the plate. Something is obviously amiss. Either Kutta-Joukowsky’s lift theorem does not
apply, or our pressure integration is missing a component of the force. At first glance,
you could argue that there cannot be a pressure force along the infinitely thin plate edge
because there is no area with normal vector along the edge for the pressure to act on.
This argument, however, is flawed. Study the top left graph in Figure .! The vortex
strength 𝛾 becomes infinite at the leading edge (𝜑 = 0), if the series coefficient 𝐴0 is
nonzero. An infinite vortex strength results in an infinite pressure difference Δ𝑝 ⟶ ∞
at the leading edge (Figure .). An infinitely small area multiplied by an infinitely
large pressure may create a finite force 𝐹l.e.s , which acts along the plate and completely
compensates for the apparent drag force 𝐷. ̃ This effect is known as leading edge suction.
The actual leading edge suction force 𝐹l.e.s may be computed with conformal mapping.
This is beyond the scope of this book, but the interested reader may find this in standard
aerodynamics texts like Katz and Plotkin ().
Considering the corrective leading edge suction force, the lift force (perpendicular to Lift force
onflow) is equal to the transverse pressure force 𝐹𝑦 .
𝑐∕2

𝐿 = 𝜌 𝑣0 𝛾(𝜉) d𝜉 (.)
∫
−𝑐∕2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

472 39 Thin Foil Theory – Lifting Flow Properties

This in turn must correspond to Kutta-Joukowsky’s lift theorem (.).

𝐿 = 𝜌 𝑣0 Γ (.)

Circulation A comparison of Equations (.) and (.) leads to the connection between circula-
tion Γ and vortex strength 𝛾:
𝑐∕2

Γ = 𝛾(𝜉) d𝜉 (.)
∫
−𝑐∕2

In Chapter  we expressed the vortex distribution as a Glauert series and linked the
series coefficients to the foil geometry. We replace the integration variable 𝜉 in Equa-
tion (.) with 𝜑 (.), adjust the integration limits according to Equation (.),
and substitute the series expansion (.) for the vortex strength 𝛾(𝜉).
𝜋
𝑐
Γ = 𝛾(𝜑) sin(𝜑) d𝜑
∫ 2
0
𝜋 [ ]
1 + cos(𝜑) ∑
∞
= 𝐴0 + 𝐴𝑛 sin(𝑛𝜑) sin(𝜑)d𝜑
∫ sin(𝜑) 𝑛=1
0

j Next, the integral is split into parts. j

𝜋 𝜋
[ ] ∑
∞
Γ = 𝑐 𝑣0 𝐴0 1 + cos(𝜑) d𝜑 + 𝑐 𝑣0 𝐴𝑛 sin(𝑛𝜑) sin(𝜑)d𝜑
∫ ∫
𝑛=1
0 0

The remaining integrals are of standard form and yield the following results:
𝜋
[ ]
1 + cos(𝜑) d𝜑 = 𝜋 (.)
∫
0

𝜋 ⎧𝜋
⎪ for 𝑛 = 1
sin(𝑛𝜑) sin(𝜑)d𝜑 = ⎨ 2 (.)
∫ ⎪0 for 𝑛 ≠ 1
0 ⎩

Because all integrals vanish for 𝑛 > 1, only the first two terms of the series expansion
remain. [ ]
𝜋 𝐴
Γ = 𝑐 𝑣0 𝐴0 𝜋 + 𝑐 𝑣0 𝐴1 = 𝑐 𝑣0 𝜋 𝐴0 + 1 (.)
2 2

Lift coefficient As a consequence, Kutta-Joukowsky’s lift theorem (.) yields a lift force which
depends on the coefficients 𝐴0 and 𝐴1 only.
[ ]
𝐴1
2
𝐿 = 𝜋 𝜌 𝑐 𝑣0 𝐴0 + (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.1 Lift Force and Lift Coefficient 473

The dimensionless section lift coefficient 𝐶𝑙 follows suit:

[ ]
𝐿 𝐴1
𝐶𝑙 = = 2𝜋 𝐴0 + (.)
1 2 2
𝜌 𝑣0 𝑐
2

We recapture the series coefficients 𝐴0 and 𝐴1 from Equations (.) and (.).
𝜋 𝜋
1 d𝑓𝑚 2 d𝑓𝑚
𝐴0 = 𝛼 + d𝜃 𝐴1 = − cos(𝜃) d𝜃 (.)
𝜋 ∫ d𝑥 𝜋 ∫ d𝑥
0 0

Inserting them into the equation for the lift coefficient, splits 𝐶𝑙 into two distinct parts,
as we would expect considering our body boundary condition (.).

𝜋
⎛ d𝑓𝑚 [ ] ⎞
⎜ 1
𝐶𝑙 = 2𝜋 𝛼 + 2𝜋 1 − cos(𝜃) d𝜃 ⎟ (.)
⎜ 𝜋 ∫ d𝑥 ⎟
⎝ 0 ⎠

- contribution of camber line

- contribution of flat plate at angle 𝛼

j • The first term represents the lift coefficient of a flat plate at angle of attack 𝛼. j
𝐶𝑙 = 2𝜋 𝛼 flat plate (.)

The slope of the camber is zero for a flat plate and, as a consequence, the second
term vanishes. The exact lift coefficient derived from conformal mapping is
𝐶𝑙𝑐.𝑚. = 2𝜋 sin(𝛼). Equation (.) is equivalent to the exact result for small
angles of attack for which sin(𝛼) ≈ 𝛼.
• The second term in Equation (.) captures the effect of camber. Obviously, a
cambered foil may produce a lift force even if it operates at zero angle of attack.

Note that the overall slope of the lift coefficient curve is 2𝜋 for thin foils. Beware that
the angle of attack in these formulas has to be entered in radians rather than in degrees!
The lift coefficient vanishes for the so called zero lift angle 𝛼0 , i.e. 𝐶𝑙 (𝛼0 ) = 0. Setting Zero lift angle 𝛼0
𝐶𝑙 = 0 in Equation (.) yields:
𝜋
⎛ d𝑓𝑚 [ ] ⎞
1
𝐶𝑙 = 0 = 2𝜋 𝛼0 + 2𝜋 ⎜ 1 − cos(𝜃) d𝜃 ⎟
⎜ 𝜋 ∫ d𝑥 ⎟
⎝ 0 ⎠

or
𝜋
1 d𝑓𝑚 [ ]
𝛼0 = − 1 − cos(𝜃) d𝜃 zero lift angle (.)
𝜋 ∫ d𝑥
0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

474 39 Thin Foil Theory – Lifting Flow Properties

j j

Figure 39.2 Section lift coefficient 𝐶𝑙 of thin, symmetric foil sections and a thin, cambered foil
section with zero lift angle 𝛼0 = −4 degree

Zero lift angles are negative for foil sections with positive camber.
Figure . pictures the lift coefficient as a function of the angle of attack. The results
of thin foil theory hold up very well in practice as long as the foils are actually thin
(𝑡max ∕𝑐 < 0.1) and the angle of attack is small (|𝛼| < 5 deg). Thicker foils, however,
have a slightly steeper slope for small angles of attack.

39.2 Moment and Center of Effort

Besides the lift force and lift coefficient, it is useful to know the center of effort for the
lift force. It is important for balancing control surfaces like rudders or stabilizers and
also affects the spindle moment of the blades in a controllable pitch propeller.
Definition of In general, a moment is a vector and computed by integration of the cross product of
Moment 𝑀𝑧 lever 𝑟 and differential force vector d𝐹 .
𝑐∕2

𝑀 = 𝑟 × d𝐹 (.)
∫
−𝑐∕2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.2 Moment and Center of Effort 475

Figure 39.3 Definition of the moment 𝑀𝑧 created by the

pressure force acting on a thin foil

Our flow problem is two-dimensional and only a moment 𝑀𝑧 with axis normal to the
𝑥-𝑦-plane exists (Figure .).
The differential pressure force d𝐹𝑝± = 𝑝± d𝜉 acts on upper and lower foil contours. Resultant
However, in accordance with thin foil theory assumptions, we perform the integration pressure force
distribution
along the tail–nose line from trailing edge to leading edge rather than integrating over
the actual contour. The force acting at a position 𝜉 along the chord is a result of the
pressure difference between lower and upper side:

d𝐹𝑦 = d𝐹𝑝− − d𝐹𝑝+ = Δ𝑝 d𝜉 (.)

Its lever with respect to the 𝑧-axis is equal to 𝜉. Therefore, the moment 𝑀𝑧 becomes
𝑐∕2 𝑐∕2

j 𝑀𝑧 = 𝜉 d𝐹𝑦 = 𝜉 Δ𝑝 d𝜉 (.) j
∫ ∫
−𝑐∕2 −𝑐∕2

Subsequently, the pressure difference is expressed with Equation (.) in terms of

the vortex strength 𝛾. Again, the now familiar variable substitution 𝜉 ⟶ 𝜑 (.) is
employed.
𝑐∕2

𝑀𝑧 = 𝜌 𝑣0 𝜉 𝛾(𝜉) d𝜉
∫
−𝑐∕2
0
[ ]
𝑐 𝑐
= 𝜌 𝑣0 cos(𝜑) 𝛾(𝜑) − sin(𝜑) d𝜑
∫ 2 2
𝜋

Reversal of the integration limits eliminates the minus sign in the integrand.
𝜋
𝑐2
𝑀𝑧 = 𝜌 𝑣0 𝛾(𝜑) sin(𝜑) cos(𝜑) d𝜑 (.)
4 ∫
0

Finally, Glauert’s series expansion (.) is used to represent the vortex strength 𝛾(𝜑). Glauert’s series

𝜋 [ ]
𝑐2 1 + cos(𝜑) ∑∞
𝑀𝑧 = 𝜌 𝑣20 𝐴0 + 𝐴𝑛 sin(𝑛𝜑) sin(𝜑) cos(𝜑) d𝜑 (.)
2 ∫ sin(𝜑) 𝑛=1
0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

476 39 Thin Foil Theory – Lifting Flow Properties

Separating the integrals yields:

𝜋
⎡ ( )
𝑐2 ⎢
𝑀𝑧 = 𝜌 𝑣20 𝐴0 cos(𝜑) + cos2 (𝜑) d𝜑
2 ⎢ ∫
⎣ 0
𝜋
∑
∞ ⎤
+ 𝐴𝑛 sin(𝑛𝜑) sin(𝜑) cos(𝜑) d𝜑⎥ (.)
∫ ⎥
𝑛=1
0 ⎦

Solving the first Once more, we are left with a set of integrals involving products of trigonometric
integral functions. The first integral is of a standard type.
𝜋
( ) [ ]𝜋 [𝜑 ]𝜋
1
cos(𝜑) + cos2 (𝜑) d𝜑 = − sin(𝜑) + + sin(2𝜑)
∫ 0 2 4 0
0
𝜋 𝜋
= 0−0 + −0 = (.)
2 2

Solving the For the second integral in Equation (.), we exploit yet another identity for trigono-
second integral metric functions.
1
sin(𝜑) cos(𝜑) = sin(2𝜑) (.)
2
Therefore
𝜋 𝜋
j 1 j
sin(𝑛𝜑) sin(𝜑) cos(𝜑) d𝜑 = sin(𝑛𝜑) sin(2𝜑) d𝜑 (.)
∫ 2∫
0 0
The integrand of the remaining integral on the right-hand side may be replaced by a
generalization of the trigonometric identity in Equation (.).
[ ( ) ( )]
1
sin(𝑛𝜑) sin(𝑘𝜑) = cos (𝑛 − 𝑘)𝜑 − cos (𝑛 + 𝑘)𝜑 (.)
2

Keeping in mind that 𝑘 = 2, the integral identity (.) converts our integral (.)
into a difference of integrals:
𝜋 𝜋
[ ( ) ( )]
1 1
sin(𝑛𝜑) sin(𝑘𝜑) d𝜑 = cos (𝑛 − 𝑘)𝜑 − cos (𝑛 + 𝑘)𝜑 d𝜑
2∫ 4∫
0 0
𝜋 𝜋
1 ( ) 1 ( )
= cos (𝑛 − 𝑘)𝜑 d𝜑 − cos (𝑛 + 𝑘)𝜑 d𝜑 (.)
4∫ 4∫
0 0

Two cases may occur:

(i) For 𝑘 ≠ 𝑛 we set 𝑘 − 𝑛 = 𝑚. The definite integral of a cosine function vanishes

with the limits  to 𝜋.
𝜋
[ ]𝜋
1
cos(𝑚𝜑) d𝜑 = sin(𝑚𝜑) = 0 (.)
∫ 𝑚 0
0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.2 Moment and Center of Effort 477

The same will be true for the second integral with 𝑘 + 𝑛 = 𝑚.

(ii) If 𝑘 = 𝑛, the argument of the first integrand becomes the constant one.
( )
cos (𝑛 − 𝑘)𝜑 = cos(0) = 1 for 𝑛 = 𝑘

Therefore, the integral takes the value

𝜋 𝜋
( ) [ ]𝜋
cos (𝑛 − 𝑘)𝜑 d𝜑 = 1 d𝜑 = 𝜑 = 𝜋 (.)
∫ ∫ 0
0 0

For 𝑘 = 𝑛 the second integral turns into

𝜋 𝜋
( )
cos (𝑛 + 𝑘)𝜑 d𝜑 = cos(2𝑛𝜑) d𝜑 = 0 (.)
∫ ∫
0 0

which vanishes like the integral in Equation (.).

In summary, the first integral in Equation (.) for the moment 𝑀𝑧 is equal to 𝜋∕2
(.). The integrals in the sum over 𝑛 of Equation (.) vanish except for the case
𝑛 = 2. Based on the results (.) and (.), we get

𝜋 ⎧𝜋
j ⎪ for 𝑛 = 2 j
sin(𝑛𝜑) sin(𝜑) cos(𝜑) d𝜑 = ⎨ 4 (.)
∫ ⎪0 for 𝑛 ≠ 2
0 ⎩

Like in the case of the lift force, only two of the Glauert series coefficients are required Result for
to compute the moment 𝑀𝑧 . The coefficient 𝐴0 is known from the lift force. In addition, moment 𝑀𝑧
we need the coefficient 𝐴2 .
2 [𝜋 ] 2
[ ]
2 𝑐 𝜋 2 𝑐 𝐴2
𝑀𝑧 = 𝜌 𝑣0 𝐴 + 𝐴2 = 𝜋 𝜌 𝑣 0 𝐴0 + (.)
2 2 0 4 2 2

The pressure distribution responsible for the moment 𝑀𝑧 creates the lift 𝐿 as a resultant Center of effort
force. Based on the moment 𝑀𝑧 , we may compute the center of effort 𝑥𝑙 for the lift 𝑥𝑙
force. Using Equation (.) for the lift force and Equation (.) for the moment,
we obtain the coordinate
2
[ ] [ ]
2 𝑐 𝐴2 𝐴2
𝜋 𝜌 𝑣0 𝐴0 + 𝐴0 +
𝑀𝑧 2 2 𝑐 2
𝑥𝑙 = = [ ] = [ ] (.)
𝐿 𝐴 4 𝐴
𝜋 𝜌 𝑐 𝑣20 𝐴0 + 1 𝐴0 + 1
2 2

For the special case of a flat plate, all coefficients 𝐴𝑛 = 0 vanish for 𝑛 > 0 because the Center of effort
derivative of the nonexisting camber is zero d𝑓𝑚 ∕d𝑥 = 0. Therefore, the center of effort for flat plate
is
𝑐
𝑥𝑙 = for a flat plate (.)
4

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

478 39 Thin Foil Theory – Lifting Flow Properties

Figure 39.4 Pressure distribution for a flat 1.5

plate at 2 degrees angle of attack ΔCP = CP− − CP+ singularity
CP+ upper side

[−]
1.0 CP− lower side

pressure coefficients ΔCP , CP+ , CP−

flat plate at =2.0deg

0.5

0.0
T.E. L.E.
center of effort

0.5

singularity
1.0
0.50 0.25 0.00 0.25 0.50
position along chord x/c [−]

The value is positive, which means the center of effort is upstream, halfway between
j origin and leading edge. Result (.) also applies to thin symmetric foils which have j
a small, finite thickness, but no camber. With increasing thickness, the center of effort
tends to move aft toward the origin at mid-chord.
Thus, if you plan to use a small rudder engine, your shaft should be attached to the
rudder at a quarter chord length from its leading edge. In practice, you might place the
shaft a bit forward, closer to the leading edge. This will reduce possible vibrations, and
the rudder will tend to return to midships if left alone.

39.3 Ideal Angle of Attack

Pressure We continue to study the pressure distribution over a flat plate in more detail. Of the
distribution on expansion series coefficients 𝐴0 , 𝐴1 , 𝐴2 , … , from Equations (.) and (.) only
flat plate
𝐴0 has a finite value, because the integrals over the slope of the camber line vanish.
𝜋
1 d𝑓𝑚
𝐴0 = 𝛼 + d𝜃 = 𝛼
𝜋 ∫ d𝑥
0
𝜋
2 d𝑓𝑚
𝐴𝑛 = − cos(𝑛𝜃) d𝜃 = 0 for 𝑛 > 0 (.)
𝜋 ∫ d𝑥
0

The pressure coefficients for upper and lower side of the foil follow from Equations (.)
and (.).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.3 Ideal Angle of Attack 479

[ ]
1 + cos(𝜑) ∑
∞
𝛾 1
𝐶𝑝± = ∓ = ∓ 2𝑣0 𝐴0 + 𝐴𝑛 sin(𝑛𝜑)
𝑣0 𝑣0 sin(𝜑) 𝑛=1
1 + cos(𝜑)
= ∓2 𝛼 (.)
sin(𝜑)

Figure . shows the pressure coefficients of upper and lower foil contour for an angle
of attack of  degrees along with the resulting dimensionless pressure difference Δ𝐶𝑝 .
1 + cos(𝜑)
Δ𝐶𝑝 = 𝐶𝑝− − 𝐶𝑝+ = 4 𝛼 (.)
sin(𝜑)
√
With the help of the identity sin(𝜑) = 1 − cos2 (𝜑) and the inverse of the variable
substitution (.) cos(𝜑) = 2𝑥∕𝑐, we express Δ𝐶𝑝 as a function of 𝑥.
1 + cos(𝜑) 1 + cos(𝜑)
Δ𝐶𝑝 = 4 𝛼 = 4𝛼 √
sin(𝜑) 1 − cos2 (𝜑)
√
√ √𝑐
√ +𝑥
√
= 4𝛼√
1 + cos(𝜑) 2
= 4𝛼 √𝑐 (.)
1 − cos(𝜑) −𝑥
2
j j
Clearly, the pressure distribution has a singularity at the leading edge. Pressure shock

Δ𝐶𝑝 ⟶ ∞ for 𝑥 ⟶ +𝑐∕2 and 𝛼 ≠ 0 (.)

Of course, in a real fluid, the pressure would not rise that far and there is no infinitely
thin plate which could support such a high pressure difference. However, even foils with
a rounded nose will experience a peak in the pressure distribution. The sudden rise and
fall in pressure over the chord wise position has a similar signature as the pressure shock
wave over time after an explosion. Therefore, the peak of the pressure distribution on a
foil operating at a nonzero angle of attack 𝛼 ≠ 0 is referred to as pressure shock. We
will see later that this pressure shock increases the danger of cavitation on a foil and
should be avoided whenever possible.
Mathematically, the singularity stems from the term associated with coefficient 𝐴0 in Ideal angle of
the expansion series for the vortex strength 𝛾 (.). The position angle becomes zero attack
𝜑 ⟶ 0 at the leading edge with 𝑥 ⟶ +𝑐∕2. As a consequence, the term associated
with 𝐴0 grows toward infinity.
( )
1 + cos(𝜑)
lim 𝐴0 ⟶ ∞ (.)
𝜑→0 sin(𝜑)
The only option to avoid the pressure shock is to design and operate the foil such that
the coefficient 𝐴0 vanishes. Recapturing Equation (.) for the coefficient
𝜋
1 d𝑓𝑚
𝐴0 = 𝛼 + d𝜑 (.)
𝜋 ∫ d𝑥
0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

480 39 Thin Foil Theory – Lifting Flow Properties

and enforcing 𝐴0 = 0, leads to a specific angle of attack 𝛼 = 𝛼id :

𝜋
1 d𝑓𝑚
0 = 𝛼id + d𝜑
𝜋 ∫ d𝑥
0
𝜋
1 d𝑓𝑚
𝛼id = − d𝜑 (.)
𝜋 ∫ d𝑥
0

The ideal angle of attack 𝛼id is important because it provides shock free entry for the
leading edge. If the foil is operating at an angle of attack 𝛼 = 𝛼id , pressure is bounded
everywhere and sharp pressure peaks near the leading edge are avoided. It should be
clear that a flat plate is shock free only if 𝛼 = 𝛼id = 0 because there is no camber. Ideally,
a foil will operate at its ideal angle of attack to delay cavitation inception.

39.4 Parabolic Mean Line

Parabolic camber As an example for the use of Glauert’s series expansion (.), we investigate the
parabolic mean line.
[ ( )2 ]
2𝑥
j 𝑓𝑚 (𝑥) = 𝑓0 1 − (.) j
𝑐
The camber line is symmetric to the middle of the chord length and has a maximum
camber value of 𝑓0 for 𝑥 = 0. The parabolic mean line is often used as a stand-in for the
mean line of a circular segment type foil, where the pressure side is flat and the suction
side is a circular arc. This section type was standard for propellers in the first half of
the th century and is still used in off-the-shelf boat propellers.
Coefficients 𝐴0 , We begin by computing the first three coefficients of the series expansion (.).
𝐴1 , and 𝐴2 for
parabolic mean 𝜋
line 1 d𝑓𝑚
𝐴0 = 𝛼 + d𝜑 (.)
𝜋 ∫ d𝑥
0

𝜋
2 d𝑓𝑚
𝐴1 = − cos(𝜑) d𝜑 (.)
𝜋 ∫ d𝑥
0

𝜋
2 d𝑓𝑚
𝐴2 = − cos(2𝜑) d𝜑 (.)
𝜋 ∫ d𝑥
0

In all three cases, we need the derivative of the parabolic mean line.
( [ ( )2 ])
d𝑓𝑚 d 2𝑥 8 𝑓0
= 𝑓0 1 − = − 𝑥 (.)
d𝑥 d𝑥 𝑐 𝑐2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.4 Parabolic Mean Line 481

Substituting the slope into the formula for 𝐴0 and employing the variable substitution
𝑥 = 𝑐 cos(𝜑)∕2 (.) yields:
𝜋 ( )
1 8 𝑓0 𝑐 4 𝑓0 [ ]𝜋
𝐴0 = 𝛼 + − cos(𝜑) d𝜑 = 𝛼 − sin(𝜑)
𝜋∫ 𝑐2 2 𝜋𝑐 0
0
4 𝑓0 ( )
= 𝛼 − 0−0 = 𝛼 (.)
𝜋𝑐
The integral vanishes, which is actually not surprising. The parabolic mean line is
symmetric to the middle of the chord length. Its derivative must be antimetric, and
the integral over antimetric functions is zero, when it is computed over a symmetric
interval. This result obviously holds true for all symmetric mean lines.
As a consequence, the ideal angle of attack is zero for the parabolic mean line or any Ideal angle of
other symmetric mean line. attack

𝐴0 = 0 ⟶ 𝛼id = 0 for symmetric mean lines (.)

The coefficient 𝐴1 becomes

𝜋 ( )
2 8 𝑓0 𝑐
𝐴1 = − − cos(𝜑) cos(𝜑) d𝜑
𝜋∫ 𝑐2 2
j 0 j
𝜋
8 𝑓0 8 𝑓0 [ 1 1
]𝜋
= cos2 (𝜑) d𝜑 = 𝜑 + sin(2𝜑)
𝜋𝑐 ∫ 𝜋𝑐 2 4 0
0
𝑓
= 4 0 (.)
𝑐
And we get for the coefficient 𝐴2 :
𝜋 ( )
2 8 𝑓0 𝑐
𝐴2 = − − cos(𝜑) cos(2𝜑) d𝜑
𝜋∫ 𝑐2 2
0
𝜋
8 𝑓0
= cos(𝜑) cos(2𝜑) d𝜑
𝜋𝑐 ∫
0

The remaining integral is found in tables of standard integrals (e.g. Gradshteyn and
Ryshik, ).
𝜋 [ ]𝜋
sin(−𝜑) sin(3𝜑)
cos(𝜑) cos(2𝜑) d𝜑 = + = 0 (.)
∫ −2 6 0
0

In fact, the integrals vanish for all coefficients 𝐴𝑛 with 𝑛 ≥ 2.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

482 39 Thin Foil Theory – Lifting Flow Properties

Figure 39.5 A prescribed pressure Pressure difference for NACA a =0.8 mean line for Cl =1.0
distribution resulting in the NACA 1.2
𝑎 = 0.8 mean line

[]
1.0

)
0.8

v02
0.6

2(pp
0.4
Cp =
0.2
0.0
0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5

NACA a =0.8 mean line for Cl =1.0

0.20 ideal angle of attack id = 1.540deg
[]

0.15 maximum camber fmax /c = 6.794% of c

0.10
fm ()
c

0.05 TE LE
camber

0.00
0.05
0.5 0.4 0.3 0.2 0.1 0.0 0.1 0.2 0.3 0.4 0.5
position along chord = xc []

Section lift According to Equation (.), the lift coefficient depends on coefficients 𝐴0 and 𝐴1
coefficient 𝐶𝑙 only: [ ] [ ]
j 𝐴1 𝑓0 j
𝐶𝑙 = 2𝜋 𝐴0 + = 2𝜋 𝛼 + 2 (.)
2 𝑐
As stated before, the slope of the lift coefficient curve 𝐶𝑙 (𝛼) is 2𝜋 for thin foils.
Zero lift angle For a foil with positive camber, the curve is shifted to the left. The zero lift angle 𝛼0 is
derived from the condition 𝐶𝑙 = 0.
[ ]
𝑓 𝑓
𝐶𝑙 = 0 ⟶ 0 = 2 𝜋 𝛼0 + 2 0 ⟶ 𝛼0 = −2 0 (.)
𝑐 𝑐
The zero lift angle (.) for the parabolic mean line is computed in radians.
NACA 𝑎 = 0.8 In the practical design of foil sections, engineers start with a desired pressure distri-
mean line bution rather than an equation for the mean line. An example is the NACA 𝑎 = 0.8
mean line. NACA stands for National Advisory Committee for Aeronautics and was
the predecessor to today’s National Aeronautics and Space Administration (NASA).
The 𝑎-mean lines prescribe a constant pressure difference from the leading edge of
a foil to the fraction 𝑎 of the chord length 𝑐 toward the trailing edge. In the case of
𝑎 = 0.8, the pressure difference between lower and upper foil contour will be constant
over % of the chord length starting at the leading edge. From the end of the constant
pressure distribution, the pressure difference linearly declines to zero at the trailing
edge. This pressure distribution and the resulting mean line are presented in Figure .
for 𝑎 = 0.8, which is the most widely used value. Pressure distribution and camber
line will be scaled with the actual design lift coefficient. Usually, lift coefficients for
propeller foil sections are in the range from 𝐶𝑙 = 0.1 to 0.5. The actual equations are
quite cumbersome and go beyond the scope of this book. Details can be found in the
classical text by Abbott and von Doenhoff ().

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

39.4 Parabolic Mean Line 483

References
Abbott, I. and von Doenhoff, A. (). Theory of wing sections – including a summary
of airfoil data. Dover Publications, Inc., New York, NY, Dover edition.
Gradshteyn, I. and Ryshik, I. (). Tables of integrals, series and products. Academic
Press, San Diego, CA, sixth edition. p. .
Katz, J. and Plotkin, A. (). Low-speed aerodynamics. Cambridge Aerospace Series.
Cambridge University Press, New York, NY, second edition.

Self Study Problems

. Derive Equation (.) for the pressure difference in a lifting flow from Bernoulli’s
equation.
1 1
𝜌|∇ Φ|2 + 𝑝 = 𝜌 𝑣20 + 𝑝0 (.)
2 2
Ignore gravity effects and assume the foil has no thickness. Apply appropriate
thin foil theory assumptions.
. Plot the curve for the section lift coefficient 𝐶𝑙 as a function of the angle of attack
𝛼 for a thin foil section with parabolic mean line and a camber–chord length ratio
of 𝑓0 ∕𝑐 = 1.5%.
j j
. Compute and plot the dimensionless pressure distribution over the chord length
for upper and lower side of a flat plate at an angle of attack 𝛼 = 3 degree.
. Revisit Problem  at the end of Chapter . What will be the ideal angle of attack
𝛼id of the foil if the maximum camber ratio is 𝑓0 ∕𝑐 = 0.03? State the result in
degrees.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

484

40
Lifting Wings

So far, our discussion of lifting flow has been restricted to bodies that stretch to infinity
in one of the three spatial coordinates. This allowed us to treat the flow around cylinders
or foil sections as two-dimensional flow problems. This chapter discusses the effects of a
finite span on the flow patterns. We still assume that the fluid is inviscid, incompressible,
and irrotational.

Learning Objectives
At the end of this chapter students will be able to:
j j
• discuss flow patterns of wings with finite span

• understand the origin of downwash

• know the Biot–Savart Law

• explain induced drag

40.1 Effects of Limited Wingspan

From two- to The length of a wing from tip to tip is called span. In the middle of a finite wing the
three- flow pattern may still be planar, and fluid flows from leading edge to trailing edge.
dimensional
flow
However, the closer you get to the wing tips, the more you will see three-dimensional
flow patterns.
Changes in flow Study the wing of finite span 𝑠 in Figure .. Like the two-dimensional flow over
pattern a foil section, the flow across a wing is governed by the onflow 𝑣0 and the effect of
circulation. In order to generate lift, the pressure on the upper side of a wing must be
lower, and the pressure on the underside of a wing must be higher compared to the
free stream pressure far away from the wing. In general, fluid flows from high pressure
areas to low pressure areas attempting to equalize the pressures. At the wing tips, fluid
will move from the underside with high pressure to the upper side of the wing with
low pressure. Therefore, an additional transverse flow component is initiated at the
wing tips for wings with finite span. This transverse flow component will bend velocity

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.1 Effects of Limited Wingspan 485

Figure 40.1 Pressure distribution on a wing of finite span 𝑠

vectors toward the wing tips on the underside and toward the wing center on the upper
side, as indicated by the flow arrows in Figure ..
According to Kutta-Joukowsky’s lift theorem (.), the lifting effect of a foil section Lifting line
may be represented by a single line vortex. The vortex is also known as lifting line, or
bound vortex, because it travels with the wing. For two-dimensional flow, we assume
that the vortex is perpendicular to the flow plane and stretches to infinity on both sides.
The vortex in three-dimensional cases may not extend past the wing tips. However, it
cannot simply end at the wing tips either. Under the condition that all forces acting on
j j
the fluid have a potential, the German physician and physicist Hermann von Helmholtz Helmholtz’s
(* – †) concluded in  that the motion of an inviscid fluid is governed theorems
by three important theorems which are quoted here from Uwe Parpart’s translation
(Helmholtz, ).

(i) No water particle that was not originally in rotation is made to rotate.
(ii) The water particles that at any given time belong to the same vortex line, however
they may be translated, will continue to belong to the same vortex line.
(iii) The product of cross section and the velocity of rotation of an infinitely thin
vortex filament is constant along the entire length of the filament and retains the
same value during all displacements of the filament. Hence, the vortex filaments
must run back into themselves in the interior of the fluid or else must end at the
bounding surface of the fluid.

These are collectively known as Helmholtz’s theorems of fluid mechanics.

From the third theorem follows that the bound vortex cannot simply end at the wing Horseshoe
tips. It finds its continuation in the tip vortices, which stretch downstream from the vortex
tips of the wing to the start-up vortex far behind the lifting wing (Figure .). The
bound vortex, tip vortices, and start-up vortex form a closed vortex ring in the fluid
with constant circulation, as requested by the third theorem.
In modeling a wing, the influence of the start-up vortex is often neglected because it is
assumed to be very far away and its effect has subsided at the location of the wing. This
is a realistic assumption because the viscosity of a real fluid will dissipate the kinetic

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

486 40 Lifting Wings

Figure 40.2 Simplest model of the vortex system of a wing

j j

Figure 40.3 Cross section through the velocity field of the wing tip vortices revealing
downwash and updraft

energy of a start-up vortex causing it to vanish over time. Because of its shape, the
remaining effective vortex system is known as a horseshoe vortex and consists of bound
vortex and tip vortices.
Downwash and The simple line vortex model of a wing helps to explain why migrating birds fly in V-
updraft shaped formations. Bicyclists ride in the slipstream of the riders in front of them to safe
energy. However, birds seem to avoid flying directly behind another bird. Figure .
shows a cross section of the velocity field generated by the tip vortices. Left and right
tip vortex rotate in opposite directions. Both contribute a downward oriented velocity
component in the space between them. This is known as downwash. A bird avoids flying
in the downwash of birds in front because it would have to work harder maintaining
altitude. It rather flies in the updraft, which the tip vortices create in the zone outside
of the wing tips. This transverse shift leads to the V-shaped flight formations.
Trailing vortex A single horseshoe vortex is a rather crude model of a lifting wing. Like a fluid will not
sheet flow around a sharp corner without shedding eddies, a vortex filament will not suddenly
change direction at the wing tips, like it is depicted in Figure .. Also, the bound
circulation will not be constant over the wing span and actually will vanish at the wing

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.1 Effects of Limited Wingspan 487

Figure 40.4 A model of a wing with varying bound circulation Γ𝑏 (𝜂) and the resulting free
vortex sheet

j j

Figure 40.5 Actual shape and roll up of trailing vortex sheet

tips. In order to comply with Helmholtz’ theorem that the vorticity may not change
along a single vortex filament, we picture the bound vortex as the superposition of an
infinite number of vortex filaments which stretch over varying portions of the wing
span (Figure .). Their continuations in the wake of the wing form a continuous sheet
of vortices. It is known as the free vortex sheet or trailing vortex sheet. The vorticity
𝛾𝑓 (𝑦) in the free vortex sheet is a function of the rate of change of bound circulation
(see Section .).
Figure . shows the free vortex sheet as a planar, rectangular area. This is a useful Roll up of free
and often employed approximation. In reality, however, the edges of a trailing vortex vortex sheet
sheet roll up behind the wing deforming the vortex sheet (Figure .). Trailing vortex

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

488 40 Lifting Wings

Figure 40.6 Velocity vectors on upper and

lower wing surfaces

sheets are invisible and seem a fairly abstract concept. However, they do exist and can
be quite strong. In early , a small private jet flipped over several times and almost
crashed because it accidentally flew through the trailing vortex sheet of a giant Airbus
A before it had dissolved (Jamieson, ).
Summary In summary, we may conclude:

• Fluid flows from the high pressure area on the lower side of a wing to the low
pressure area on the upper side of a wing. This causes a transverse velocity
component, which is oriented toward the wing tips on the lower side and oriented
toward the center on the upper side of a lifting wing.
j j
• The overall lifting effect may, like that of a foil section, be described by a single
line vortex which we call bound vortex. In contrast to the two-dimensional foil
section, the bound circulation strength is varying along the wing span.
• The variation of the bound circulation creates a free vortex sheet which extends
behind the wing to the origin of the circulation (start-up vortex).
• The trailing vortex sheet induces a downwash, a velocity component which points
roughly into the opposite direction of the lift force.

In the following section, the relationship between free and bound vorticity is explored
in more mathematical terms.

40.2 Free and Bound Vorticity

Mean velocity Figure . compares the velocities on upper and lower side at an off-center position 𝑦
along a wing of finite span 𝑠. We assume the wing is thin so that we may employ results
of thin foil theory. As explained above, the velocity 𝑣− on the underside is rotated
toward the wing tips and the velocity 𝑣+ on the upper side is pointing toward the center
line of the wing. We define the mean velocity 𝑣𝑚 (Figure .):

1( + )
𝑣𝑚 = 𝑣 + 𝑣− (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.2 Free and Bound Vorticity 489

Figure 40.7 Velocity vectors on upper and

lower wing surface and mean velocity 𝑣𝑚

Figure 40.8 Vorticity vector 𝛾 and the

resulting difference velocity 𝑣𝑑

Upper and lower velocity are now connected via the mean velocity and a difference Difference
velocity 𝑣𝑑 (Figure .). With velocity

2𝑣𝑑 = 𝑣+ − 𝑣− (.)
we may write for upper and lower velocity:
𝑣+ = 𝑣𝑚 + 𝑣𝑑 𝑣− = 𝑣𝑚 − 𝑣𝑑 (.)
j j
In two-dimensional foil theory, a distribution of line vortices is applied to model the
difference in upper and lower side speed (see Section ). Two times the difference ve-
locity 𝑢23 turned out to be equal to the vortex strength 𝛾 = 2|𝑢23 | (see Equation (.)).
The axes of a two-dimensional vortex distribution is perpendicular to the flow plane,
and upper and lower velocity are parallel to each other.
In the three-dimensional case, we may also employ a vortex to model the difference Vortex filament
velocity 𝑣𝑑 . Analog to the two-dimensional case, the strength of the vortex must be
equal to two times the magnitude of the difference velocity:
𝛾 = |𝛾| = 2|𝑣𝑑 | (.)
Upper and lower velocity are no longer parallel though. Straight vortex filaments
generate velocities perpendicular to their axis (see Section .). Therefore, the axis of
the vortex filament 𝛾 must be perpendicular to the difference velocity 𝑣𝑑 . This is shown
in Figure ..
Let the magnitudes of the velocity vectors be: Velocities

𝑣+ = |𝑣+ | 𝑣− = |𝑣− | (.)

𝑣𝑚 = |𝑣𝑚 | 𝑣𝑑 = |𝑣𝑑 | (.)
Using the angles depicted in Figure ., relationships between the velocities follow
from the law of cosines.
( )
𝜋
(𝑣+ )2 = 𝑣2𝑚 + 𝑣2𝑑 − 2𝑣𝑚 𝑣𝑑 cos +𝛿 (.)
(2 )
𝜋
(𝑣− )2 = 𝑣2𝑚 + 𝑣2𝑑 − 2𝑣𝑚 𝑣𝑑 cos −𝛿 (.)
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

490 40 Lifting Wings

Figure 40.9 Bound vorticity 𝛾 and its effect

𝑏

The angle enclosed by mean velocity 𝑣𝑚 and vortex filament 𝛾 is marked by 𝛿.

Pressure The Bernoulli equation (.) provides a connection between velocity and pressure.
difference The pressure difference between lower and upper wing side is
[ ]
1
Δ𝑝 = 𝑝− − 𝑝+ = 𝜌 (𝑣+ )2 − (𝑣− )2 (.)
2
Substituting Equations (.) and (.) for the squared velocities yields:
[ ( ) ( )]
𝜋 𝜋
Δ𝑝 = −𝜌𝑣𝑚 𝑣𝑑 cos + 𝛿 − cos −𝛿 (.)
2 2
Both cosine terms phase shifted by ±90 degrees are identical to a negative sine curve:
( ) ( )
𝜋 𝜋
− sin(𝛿) = cos + 𝛿 = − cos +𝛿 (.)
j 2 2 j
Therefore, the equation for the pressure difference at the wing reduces to

Δ𝑝 = 𝜌𝑣𝑚 2𝑣𝑑 sin(𝛿)

Equation (.) links difference velocity 𝑣𝑑 and vortex strength 𝛾. Replacing 2𝑣𝑑 with 𝛾
expresses the pressure difference as a function of density, mean velocity, and the angle 𝛿.

Δ𝑝 = 𝜌𝑣𝑚 𝛾 sin(𝛿) (.)

We split the vortex filament vector 𝛾 into two components, so that

𝛾 = 𝛾 +𝛾 (.)
𝑏 𝑓

Bound vorticity (i) The first component 𝛾 is perpendicular to the mean velocity 𝑣𝑚 . Thus, it is
𝑏
parallel to the wing axis 𝑦 (Figure .). We call this the bound vorticity, or bound
circulation, because it is tied to the wing. The geometry implies that

𝛾𝑏 = 𝛾 sin(𝛿) (.)

A comparison with Equation (.) reveals that the pressure difference depends
only on the strength of the bound vorticity.

Δ𝑝 = 𝜌𝑣𝑚 𝛾𝑏 (.)

This formula is equivalent to Equation (.), which we obtained for the two-
dimensional foil section where 𝛾 = 𝛾𝑏 (𝛿 = 90◦ ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.2 Free and Bound Vorticity 491

Figure 40.10 Free vorticity 𝛾 and its effect

𝑓

(ii) The second component 𝛾 of the vorticity vector is parallel to the mean velocity. Free vorticity
𝑓
From geometry in Figure ., we gather that

𝛾𝑓 = 𝛾 cos(𝛿) (.)

This component is called free vorticity because the vortex filament exists free of
the wing in the trailing vortex sheet.

Bound vorticity creates a difference velocity ±𝑣𝑑𝑏 parallel to the mean velocity. Thus, Effect of bound
the bound vorticity is responsible for the pressure difference between lower and upper vorticity
wing side and, hence, the lift force.
The free vorticity induces a difference velocity 𝑣𝑑𝑓 perpendicular to the mean velocity. Effect of free
Therefore, free vorticity is responsible for the three-dimensional flow effects that occur vorticity
j on wings with finite span. The lift force may be found with knowledge of the bound j
vorticity alone. However, we must know the free vorticity to quantify three-dimensional
flow effects. Consequently, a connection has to be established between the strengths of
bound and free vorticity.
In two-dimensional foil theory, we computed the circulation created by a distribution Bound
of line vortices by integrating the magnitude of the vorticity 𝛾 along the chord length. circulation
Formally, this is equivalent to the last surface integral in Equation (.). The vortex
strength 𝛾(𝑥) is equal to 𝛾 𝑇 𝑛 along the chord with (𝑥 ∈ [−𝑐∕2, 𝑐∕2]) and is zero every-
where else, which reduces the surface integral to the line integral in Equation (.).
For the three-dimensional wing, we select a path 𝐵 from trailing edge to leading edge
that is always parallel to the mean velocity. By definition, the bound vorticity is perpen-
dicular to this path. Therefore, the bound circulation of a wing at the spanwise location
𝑦 is equal to
Γ𝑏 (𝑦) = 𝛾𝑏 d𝓁 (.)
∫
𝐵

Now, imagine a closed path 𝐶 which consists of two almost closed circles connected by Link between
an infinitesimally thin bridge. The circles are roughly perpendicular to each other, as bound and free
vorticity
depicted in Figure .. This oddly shaped path encloses a simply connected region 𝑆.
Consequently, the circulation over the area 𝑆 vanishes for an inviscid and irrotational
flow according to Equation (.).
The path 𝐶 provides us with a tool to establish the desired connection between bound
and free vorticity. We shift the path so that the area in the first circle lines up with
path 𝐵 across the wing (Figure .). One edge of the thin bridge is just above, the

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

492 40 Lifting Wings

j j
Figure 40.11 Establishing a connection between bound and free vorticity by integration over the boundary of a
simply connected region

other just below the trailing vortex sheet and follows the vortex filament that leaves the
trailing edge of the wing at the end of path 𝐵. The area in the second circle is oriented
perpendicular to the trailing vortex filaments along path 𝐹 . We assume that path 𝐹
is close enough to the wing, so that we may ignore the roll up of the trailing vortex
sheet. The shifted path 𝐶 is still closed and still encloses a simply connected region 𝑆.
Total circulation Therefore, the total circulation is zero.

Γ = 𝑣𝑇 d𝑠 = 𝛾 𝑇 𝑛 d𝑆 = 0 (.)
∮ ∬
𝐶 𝑆

Free circulation However, as is evident from Figure ., both circular areas enclose lines with nonzero
vorticity. The circulation enclosed by the first circle is equal to the bound circulation of
Equation (.). An integral over the second circle encloses the free vorticity from the
edge of the trailing vortex sheet to the vortex filament at spanwise position 𝑦. Integration
along path 𝐹 yields the free circulation:
𝑦

Γ𝑓 (𝑦) = 𝛾𝑓 d𝓁 = 𝛾𝑓 d𝓁 (.)
∫ ∫
𝐹 𝑠∕2

The area of the thin bridge does not contribute any circulation since it is aligned parallel
to the free vortex filament 𝛾 and consequently 𝛾 𝑇 𝑛 = 0.
𝑓 𝑓

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.3 Biot–Savart Law 493

In summary, the total circulation in the enclosed area 𝑆 consists of three parts.

Γ = Γ𝑏 (𝑦) + 0 + Γ𝑓 (𝑦) = 0 (.)

contribution from  circle, Equation (.)
nd

contribution from bridge

contribution from st circle, Equation (.)

The contributions must add up to zero according to Equation (.). Consequently,

the free circulation Γ𝑓 , which is contained in the trailing vortex sheet from its outer
edge to the vortex filament that left the wing at position 𝑦, is equal to the negative
bound circulation Γ𝑏 (𝑦).
Γ𝑓 (𝑦) = −Γ𝑏 (𝑦) (.)

Next, we take the derivative of this equation with respect to the position 𝑦. Free vorticity

d d
Γ (𝑦) = − Γ𝑏 (𝑦) (.)
d𝑦 𝑓 d𝑦

With the integral from Equation (.) we have:

𝑦
d d
𝛾 d𝓁 = − Γ𝑏 (𝑦) (.)
d𝑦 ∫ 𝑓 d𝑦
𝑠∕2

j For the left-hand side, we make use of a rule for the derivative of an integral: if a function j
𝑔 and its antiderivative 𝐺 are given, then the derivative of the integral of 𝑔 with respect
to the upper bound is equal to 𝑔 evaluated at the upper bound.

⎡ 𝑦 ⎤
d ⎢ d[ ] d𝐺(𝑦) d𝐺(𝑎)
𝑔(𝑡) d𝑡⎥ = 𝐺(𝑦) − 𝐺(𝑎) = − = 𝑔(𝑦) (.)
d𝑦 ⎢∫ ⎥ d𝑦 d𝑦 d𝑦
⎣𝑎 ⎦
Note that the value 𝐺(𝑎) is a constant in this context. Therefore, its derivative with
respect to 𝑦 vanishes. This rule is now applied to the integral in Equation (.):
d
𝛾𝑓 (𝑦) = − Γ (𝑦) (.)
d𝑦 𝑏
Thus, the free vorticity 𝛾𝑓 in the trailing vortex sheet depends on the spanwise change
of the bound circulation.
This corresponds to our model of the vortex system of a wing from Figure .. In order
to comply with Helmholtz’ theorems, the vortex system sheds any change of bound
circulation as a vortex filament downstream. As a consequence, the filaments do not
end in the fluid domain and maintain constant strength.

40.3 Biot–Savart Law

Now that we have a means to compute the strength of the trailing vortex filaments, we
may analyze their effects on the flow across the wing. In Section ., the potential

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

494 40 Lifting Wings

Figure 40.12 The velocity induced by an

element of a vortex filament

of a two-dimensional vortex has been introduced. Its planar flow field is shown in
Figure .. The velocity field is expected to be more complex in the three-dimensional
case.
Biot–Savart Law The task at hand is to find the velocity 𝑣 at a point 𝑝 induced by a vortex filament of
strength 𝛾𝑓 . Point 𝑝 is also known as field point or collocation point. Figure . sets
up the domain. Note that, unlike the filament shown in the figure, a vortex filament
forms a closed loop or stretches to the limits of the flow domain. In an unbounded
fluid, it stretches to infinity in both directions. An infinitesimal piece d𝑠 of the vortex
filament of strength 𝛾 is oriented in the direction of a unit length tangent vector 𝑠. It is
located at point 𝑞, which is also known as a source point.
j We want to know the velocity d𝑣 at a field point 𝑝 induced by the infinitesimal piece d𝑠 j
of the vortex filament. A solution to this problem was found by the French scientists
Jean-Baptiste Biot (*–†) and Félix Savart (*–†), who investigated the
magnetic field created by a current through a conductor (Biot and Savart, ). This
problem may also be described by potential theory. Accordingly, many cross references
exist between magneto-electrodynamics and fluid dynamics. The Biot–Savart Law in
fluid mechanics computes the velocity induced by a vortex filament:

𝛾 (𝑟 × d𝑠)
d𝑣 = − Biot–Savart Law (.)
2𝜋 |𝑟|3

Induced velocity The result of a cross product is perpendicular to the plane of the vectors, here 𝑟 and
d𝑠. The vector 𝑟 marks the relative position of field point 𝑝 with respect to the source
point 𝑞. The influence of a vortex filament declines fast with the inverse of the distance
between 𝑝 and 𝑞 squared. Integration along the path 𝐶 of a vortex filament produces
the velocity induced at the field point 𝑝 by the whole vortex filament.

𝛾 (𝑟 × d𝑠)
𝑣(𝑝) = − (.)
2𝜋 ∫ |𝑟|3
𝐶

The integral has to be solved anew for each field point.

Straight vortex Unfortunately, the integral has no explicit solution unless the geometry of the path 𝐶 is
filament known and of reasonably simple shape. One such case is a straight vortex filament like
 Don’t get fooled by the |𝑟|3 in the denominator. 𝑟 also appears in the numerator.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.3 Biot–Savart Law 495

Figure 40.13 The velocity induced by a

straight vortex filament

it is shown in Figure .. For convenience, the filament is aligned with the 𝑥-axis of
our coordinate system, and the field point 𝑝 lies on the 𝑦-axis. Therefore, the vectors 𝑝
and 𝑞 and their difference 𝑟 = 𝑝 − 𝑞 are given by:

⎛0⎞ ⎛𝜉⎞ ⎛ −𝜉 ⎞
𝑝 = ⎜𝑦⎟ 𝑞 = ⎜0⎟ 𝑟 = ⎜ 𝑦 ⎟ (.)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0⎠ ⎝0⎠ ⎝ 0 ⎠
The distance between field and source point is
√
j 𝑟 = |𝑟| = 𝜉 2 + 𝑦2 (.) j

We also need the cross product between 𝑟 and the line element d𝑠 for the integrand in
Equation (.). d𝑠 is stated in Figure ..
| 𝑖 𝑗 𝑘 || ⎛ ⎞ ⎛ 0 ⎞
| 𝑦⋅0−0
| | ⎜ ⎟=⎜ 0 ⎟
𝑟 × d𝑠 = | −𝜉 𝑦 0 | = 0 ⋅ d𝜉 − (−𝜉) ⋅ 0 (.)
| | ⎜ ⎟ ⎜ ⎟
| d𝜉 |
0 0 | ⎝ −𝜉 ⋅ 0 − 𝑦 ⋅ d𝜉 ⎠ ⎝ −𝑦 d𝜉 ⎠
|
Both vectors 𝑟 and d𝑠 lie in the 𝑥-𝑦-plane. Consequently, the result of their cross product
points into the 𝑧-direction.
We substitute the cross product and the distance into the integral of Equation (.). Induced velocity

⎛ 0 ⎞
⎜ 0 ⎟
𝑥2 ⎜ ⎟
𝛾 ⎝ −𝑦 ⎠
𝑣12 = − √ d𝜉 (.)
2𝜋 ∫ ( 𝜉 2 + 𝑦2 )
𝑥1

Our vortex filament stretches to infinity in both directions of the 𝑥-axis. For the moment,
we consider only the velocity induced by the piece located between 𝑥1 and 𝑥2 .
Only the vertical component 𝑤12 of the resulting velocity will have a nonzero value: Effect of partial
filament
𝑥2
𝛾 𝑦
𝑤12 = + d𝜉 (.)
2𝜋 ∫ (𝜉 2 + 𝑦2 )3∕2
𝑥 1

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

496 40 Lifting Wings

This integral is of standard type, and we get:

| |𝑥2
𝛾 | 𝜉 |
𝑤12 = | |
|( )1∕2 |
4𝜋 | 𝜉 2 + 𝑦2 |
| |𝑥1
⎡ ⎤
𝛾 ⎢ 𝑥2 𝑥1 ⎥
= ⎢ √ −√ ⎥ (.)
4𝜋 ⎢
𝑥22 + 𝑦2 𝑥21 + 𝑦2 ⎥
⎣ ⎦
The denominators in the result are equal to the distances 𝑟2 and 𝑟1 from the positions
𝑥2 and 𝑥1 to the field point, respectively.
[ ]
𝛾 𝑥2 𝑥1
𝑤12 = − (.)
4𝜋 𝑟2 𝑟1
From the geometry depicted in Figure ., it follows that the fractions 𝑥1 ∕𝑟1 and
𝑥2 ∕𝑟2 are related to the cosines of the angles 𝜃1 and 𝜃2 . Note that for a negative 𝑥1 < 0,
the fraction 𝑥1 ∕𝑟1 becomes negative and that for a positive 𝑥2 > 0, the angle 𝜃2 becomes
larger than  degrees. Therefore, 𝑥1 ∕𝑟1 = − cos(𝜃1 ) and 𝑥2 ∕𝑟2 = − cos(𝜃2 ) in order to
preserve the signs.
𝛾 [ ]
𝑤12 = − cos(𝜃2 ) + cos(𝜃1 ) (.)
4𝜋
Effect of According to Helmholtz’s theorems the vortex filament cannot simply start at 𝑥1 and
j complete end at 𝑥2 . Beginning and end must be at physical limits of the domain, like a wall, or the j
filament
vortex filament stretches to infinity. We investigate the latter. For 𝑥1 → −∞, the angle
𝜃1 shrinks to zero: lim𝑥1 →∞ (𝜃1 ) = 0. For 𝑥2 → ∞, the angle 𝜃2 approaches 180 degrees:
lim𝑥2 →∞ (𝜃2 ) = 𝜋. Therefore, the limits are:

lim cos(𝜃1 ) = cos(0) = +1

𝑥1 →∞

lim cos(𝜃2 ) = cos(𝜋) = −1

𝑥2 →∞

The resulting induced vertical velocity 𝑤 becomes

𝛾 [ ] 𝛾
𝑤 = lim (𝑤12 ) = − (−1) + 1 = (.)
𝑥1 →−∞ 4𝜋 2𝜋
𝑥2 →∞

This looks familiar, and it is indeed the magnitude of the velocity created by a two-
dimensional vortex at a distance 𝑟 = 𝑦 from the location of the vortex (see Section .).
Effect of trailing The trailing vortex filaments of our wing do not stretch forward of the wing. They merge
vortex filament into the bound vortex at the wing. If we place the wing at position 𝑥 = 0, the trailing
vortex filaments stretch from minus infinity to zero: −∞ < 𝑥 ≠ 0. For 𝑥2 → 0 the
angle 𝜃2 becomes 90 degrees and, as a result, cos(𝜃2 ) = 0. The induced vertical velocity
becomes exactly half of the velocity induced by the vortex filament which stretches to
infinity in both directions.
𝛾 [ ] 𝛾
𝑤 = lim (𝑤12 ) = − (−1) + 0 = (.)
𝑥1 →−∞ 4𝜋 4𝜋
𝑥2 →∞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.4 Lifting Line Theory 497

Figure 40.14 The effect of

downwash on the angle of attack

40.4 Lifting Line Theory

Lifting line theory is the basic mathematical tool used to compute the flow field for Lifting line
wings of finite span 𝑠. The wing is modeled as a single vortex, whose bound circulation theory
Γ𝑏 (𝑦) varies over the span of the wing. The circulation must become zero at the wing
tips to comply with Helmholtz’s theorems. The resulting changes in bound circulation
create a trailing vortex sheet.
Each filament of the trailing vortex sheet induces a velocity at the position of the wing. Downwash
The precise direction and strength of the velocity depends on the shape of the trailing
vortex sheet and the rate of change of bound circulation, which governs the strength
𝛾𝑓 of the trailing vortex filaments. In general, the effect of a trailing vortex sheet looks
j similar to the velocity distribution in Figure .. There is a downward oriented velocity j
𝑤𝑖 across the whole span of the wing, which we call downwash.
Downwash has a noticeable effect on the flow across a wing. Figure . shows a cross Induced angle of
section through a wing and the velocity vectors at a spanwise position 𝑦. The downwash attack
velocity 𝑤𝑖 , induced by the trailing vortex sheet, is superimposed with the onflow 𝑣0
and creates the effective onflow vector 𝑣ef f . As a result, the geometric angle of attack 𝛼
of the foil section is reduced to the effective angle of attack 𝛼ef f . The change in angle of
attack is known as induced angle of attack 𝛼𝑖 , which should not be confused with the
ideal angle of attack 𝛼id from the previous chapter.
Typically, the magnitude of downwash 𝑤𝑖 = |𝑤𝑖 | is considerably smaller than the onflow
velocity 𝑣0 = |𝑣0 | and roughly perpendicular to the onflow, as long as the sectional lift
coefficients remain small. As a consequence, the magnitude of effective velocity 𝑣ef f is
about the same as the undisturbed onflow speed.

𝑣ef f ≈ 𝑣0 (.)

The main effect of downwash is a change in the direction of flow speeds. For cases with
𝑤𝑖 ≪ 𝑣0 , we can state that the induced angle of attack is approximately
𝑤𝑖
𝛼𝑖 ≈ tan(𝛼𝑖 ) = (.)
𝑣0

The change in angle of attack caused by the downwash has substantial consequences. Lift force
From Kutta-Joukowsky’s lift theorem we know that circulation and onflow create a
force perpendicular to the onflow. Figure . shows the force vectors. The total force

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

498 40 Lifting Wings

Figure 40.15 Induced drag caused

by the downwash

d𝐹 on a foil section of the wing is now acting normal to the effective velocity.

d𝐹 = 𝜌 𝑣ef f Γ𝑏 d𝑦 ≈ 𝜌 𝑣0 Γ𝑏 d𝑦 (.)

The usable lift force d𝐿 is the force component perpendicular to the original, undis-
turbed onflow 𝑣0 .

d𝐿 = d𝑓 cos(𝛼𝑖 ) = 𝜌 𝑣0 Γ𝑏 cos(𝛼𝑖 ) d𝑦 ≈ 𝜌 𝑣0 Γ𝑏 d𝑦 (.)

Again, since the induced angle is small, the cosine of 𝛼𝑖 is approximately one. Therefore,
j total and lift force are essentially equal in magnitude but point in different directions. j
Induced drag Due to the induced angle of attack 𝛼𝑖 , a second force component d𝐷𝑖 exists. It points in
the direction of the undisturbed onflow. Since it acts against the motion of the wing, it
is experienced as an additional drag. This force component is called induced drag.

d𝐷𝑖 = d𝐿 tan(𝛼𝑖 ) ≈ 𝜌 𝑣0 Γ𝑏 𝛼𝑖 d𝑦 (.)

Note that induced drag occurs even in an inviscid fluid. In a real fluid with nonzero
viscosity, an additional viscous drag must be considered.
Small induced angles of attack may be replaced by Equation (.). Substituting it
into Equation (.) yields for the induced drag:
𝑤𝑖
d𝐷𝑖 = d𝐿 tan(𝛼𝑖 ) ≈ 𝜌 𝑣0 Γ𝑏 d𝑦 = 𝜌 Γ𝑏 𝑤𝑖 d𝑦 (.)
𝑣0
The last term emphasizes that the higher the downwash at a wing section, the higher
becomes the induced drag.
Summary This short introduction to wing theory may suffice for the purpose of our basic dis-
cussion of ship propulsion. Obviously, further studies are needed to apply lifting line
theory to compute downwash, lift, and induced drag for foils and propeller blades.
Lifting line theory is covered in all texts on aerodynamics, e.g. Moran (). A com-
prehensive description for lifting line theory in the context of propellers is found in
Breslin and Andersen (). A working implementation and an introduction to the
theory is provided by the OpenProp project (Epps, ). Its home page can be found
at https://sage-newt.cloudvent.net/ (last visited in December ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

40.4 Lifting Line Theory 499

References
Biot, J. and Savart, N. (). Note sur le magnétisme de la pile de Volta. Annales de
chimie et de physique, :–.
Breslin, J. and Andersen, P. (). Hydrodynamics of ship propellers. Cambridge
University Press, Cambridge, United Kingdom.
Epps, B. (). An impulse framework for hydrodynamic force analysis: fish propulsion,
water entry of spheres, and marine propellers. PhD thesis, Massachusetts Institute of
Technology (MIT), Cambridge, MA.
Helmholtz, H. (). On the integrals of the hydrodynamic equations that correspond
to vortex motions. Translation by Uwe Parpart. Int. Journal of Fusion Energy, (–
):–. Original published in German in Journal für die reine und angewandte
Mathematik, :–, .
Jamieson, A. (). Private jet flipped over in wake turbulence from
Airbus A. NBC News at http://www.cnbc.com////
private-jet-flipped-over-in-wake-turbulence-from-airbus-a.html. Last
visited June , .
Moran, J. (). An introduction to theoretical and computational aerodynamics. John
Wiley & Sons, New York, NY.

j Self Study Problems j

. Explain why migrating birds fly in a V-shaped formation.

. Given is the following elliptical distribution of bound circulation.
√
( )2
2𝑦
Γ𝑏 (𝑦) = Γ0 1 −
𝑠
Γ0 is the maximum bound circulation, 𝑠 is the wingspan, and 𝑦 is the coordinate
along the wingspan. Γ0 and 𝑠 are constants and known. Compute the strength of
the free vorticity 𝛾𝑓 in the trailing vortex sheet.
. State the Biot–Savart Law.
. Use the solution of Problem  to compute the distribution of downwash 𝑤𝑖 .
𝑠∕2

1 𝛾𝑓 (𝜂)
𝑤𝑖 (𝑦) = d𝜂
4𝜋 ∫ (𝑦 − 𝜂)
−𝑠∕2

This formula is valid under the assumptions that the trailing vortex sheet is
straight, oriented in the direction of the undisturbed onflow, and does not show
noticeable effects due to roll up. Hint: the integral can be converted into a Glauert
integral with the substitution 𝜂 = 𝑠 cos(𝜑)∕2. Details can be found in Section ..
. Explain to a friend the origin of induced drag.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

500

41
Open Water Test

Main objective of propeller design is the creation of a hull–propeller propulsion system

with high efficiency. The system must produce as much thrust as possible for as little
power spent as possible. Although great strides has been made in the computational
assessment of propeller characteristics, model tests with propellers still play an impor-
tant part of design verification. Like the ship hull alone is evaluated in the resistance
test, we first evaluate the propeller without the influence of the hull. These open water
tests are discussed in this chapter.

Learning Objectives
j j
At the end of this chapter students will be able to:

• understand hydrodynamic propeller characteristics

• conduct open water tests and interpret their results

• determine open water efficiency

41.1 Test Conditions

Open water test In Chapter  we subdivided the hull–propeller system into parts to simplify the analysis
of their hydrodynamic characteristics. The ship hull alone and its resistance when
moving with constant speed through deep, calm water are investigated with a resistance
test. The thrust created by a propeller and the required torque to generate that thrust
are analyzed in an open water test. It is called open water test because the propeller
operates in a uniform flow. This may be achieved in two ways:

• a rotating propeller is moved at constant speed through a towing tank where the
water itself is initially at rest

• the test is conducted in a cavitation tunnel where the water streams past the
propeller rotating in place

More details are provided in Section ..

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

41.1 Test Conditions 501

Before measurement equipment is set up, laws of similarity have to be considered which Geometric,
govern the model test. Maintaining geometric, kinematic, and dynamic similarity are kinematic, and
dynamic
critical if the results are used for performance prediction of the full scale propeller. similarity
The thrust 𝑇 of a propeller will depend on water density 𝜌 and kinematic viscosity 𝜈. Dimensional
As usual we assume that they are constant throughout the fluid and over the duration analysis
of a test. The propeller diameter 𝐷 may serve a characteristic length. Flow conditions
are represented by the speed of advance 𝑣𝐴 and the rate of revolution 𝑛 of the propeller.
Gravitational acceleration 𝑔 and pressure 𝑝 indicate the magnitude of external forces.
In summary, the propeller thrust 𝑇 is assumed to be a yet unknown function  of the
seven physical variables (𝜌, 𝜈, 𝑔, 𝐷, 𝑣𝐴 , 𝑝, 𝑛):

𝑇 =  (𝜌, 𝜈, 𝑔, 𝐷, 𝑣𝐴 , 𝑝, 𝑛) (.)

Function  uses three fundamental physical quantities: force, length, and time with
units Newton [N], meter [m], and seconds [s]. Consequently, according to dimensional
analysis and the Buckingham 𝜋-theorem the flow problem is characterized by 7 − 3 = 4
dimensionless numbers.
[ ]
⎛ ⎡ ⎤ [ ] [ ]⎞
1 2 2 ⎜ 𝑔𝐷 𝑝 ⎥ 𝜈 𝑛𝐷 ⎟
𝑇 = 𝜌 𝑣𝐴 𝐷 ⋅ 𝐶 ⋅ 𝑓 , ⎢ , , (.)
2 ⎜ 𝑣2 ⎢ 1 𝜌 𝑣2 ⎥ 𝑣𝐴 𝐷 𝑣𝐴 ⎟
⎝ 𝐴 ⎣ 2 𝐴 ⎦ ⎠
The unknown function 𝑓 is essentially equal to the thrust loading coefficient 𝐶𝑇 ℎ . With
j the constant factor 𝐶 = 𝜋∕4 we may write: j
[ ]
⎛ ⎡ ⎤ [ ] [ ]⎞
𝑇 𝑔𝐷 𝑝 ⎥ 𝜈 𝑛𝐷 ⎟
𝐶𝑇 ℎ = = 𝑓⎜ , ⎢ , , (.)
1
𝜌 𝑣𝐴2 𝜋 𝐷2 ⎜ 𝑣2 ⎢ 1 𝜌 𝑣2 ⎥ 𝑣𝐴 𝐷 𝑣𝐴 ⎟
2 4 ⎝ 𝐴 ⎣ 2 𝐴⎦ ⎠
The unknown function 𝑓 may be found by conducting experiments by systematically
varying the four dimensionless numbers. The numbers represent different laws of
similarity (see Chapter ).
The first dimensionless parameter in Equation (.) is equal to the inverse of a squared Propeller Froude
propeller Froude number number
[ ]
𝑔𝐷 1
=
𝑣𝐴2 𝐹𝑟2

with
𝑣
𝐹𝑟 = √ 𝐴 (.)
𝑔𝐷
Ship speed and length in waterline for the Froude number of the ship are replaced by
speed of advance 𝑣𝐴 and propeller diameter 𝐷, respectively.
The Froude number represents the similarity of inertia and gravity forces. During
resistance tests it ensured that the wave pattern and wave resistance are similar to the
full scale vessel. Most propellers operate fully submerged, in which case, wave making
and gravity forces are of lesser concern.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

502 41 Open Water Test

Figure 41.1 Simplified velocities

triangle for an unrolled propeller blade
section at radius 𝑥 = 0.75 (induced
velocities have been ignored)

Propeller The second dimensionless parameter in Equation (.) is an Euler number and ensures
pressure similarity of inertia and pressure forces.
coefficient
[ ]
2𝑝
𝐸𝑢 = (.)
𝜌 𝑣𝐴2

We already learned that in order to keep 𝐸𝑢 the same for model and full scale, air pressure
above the tank water surface has to be reduced for model tests. This is essential for
cavitation tests (see Chapter ). Fortunately, the thrust generated by the propeller
stems from pressure differences on the propeller blades. Therefore, static pressure
components may be neglected and thrust at model and full scale are similar if the
pressure coefficient is the same, i.e. 𝐶𝑃 𝑀 = 𝐶𝑃 𝑆 = 2Δ𝑝∕(𝜌𝑣𝐴2 𝐴0 ). This will be the case,
j j
if the angles of attack are the same for model and full scale propeller blade sections.
Propeller A form of Reynolds number 𝑅𝑒 is given with the third dimensionless parameter in
Reynolds Equation (.). [ ]
number 𝑣 𝐷
𝜈 1
= with 𝑅𝑒 = 𝐴 (.)
𝑣𝐴 𝐷 𝑅𝑒 𝜈
Again, maintaining both, equal Froude number and equal Reynolds number for model
and full scale vessel, is practically impossible. Since gravity forces are of lesser im-
portance for a deeply submerged propeller, the Reynolds number is kept as high as is
feasible with the available testing equipment in order to ensure turbulent boundary layer
flow over the propeller blades. However, full similarity of viscous forces (𝑅𝑒𝑀 = 𝑅𝑒𝑆 )
is generally not attainable for model scale propellers.
Since Reynolds numbers are typically lower during open water tests for models than
for the full scale propellers, model test results have to be corrected for the full scale
propeller.
Propeller Finally, the fourth and last dimensionless parameter in Equation (.) is the inverse of
advance the advance coefficient 𝐽 . 𝐽 is also referred to as advance ratio.
coefficient [ ]
𝑛𝐷 1 𝑣
= with 𝐽 = 𝐴 (.)
𝑣𝐴 𝐽 𝑛𝐷
𝑣𝐴 is again the speed of advance, 𝑛 the rate of revolution, and 𝐷 the diameter of the
propeller.
Figure . shows a simplified velocity triangle for a propeller blade section. The
transverse velocity due to rotation of the propeller grows linearly from zero at the shaft

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

41.2 Propeller Models 503

center to its maximum value of 2𝜋𝑛𝑅 at the blade tip. At radius 𝑟 the transverse velocity
is 2𝜋𝑛𝑟. The denominator (𝑛𝐷) in Equation . is proportional to the transverse speed
of the blade tip. The factor 𝜋 has been dropped for simplicity.
The ratio of speed of advance 𝑣𝐴 and the transverse velocity (𝑛 𝐷) in Equation (.) is
a measure for the hydrodynamic pitch of the blade section. If the velocity triangles are
similar at model and full scale, blade sections will operate at the same angle of attack.
Therefore, keeping advance coefficient equal for model and full scale propeller ensures
kinematic similarity.
𝑣𝐴𝑀 𝑣𝐴𝑆
𝐽𝑀 = 𝐽𝑆 ⟶ = (.)
𝑛𝑀 𝐷𝑀 𝑛𝑆 𝐷𝑆
Dynamic similarity cannot be obtained without kinematic similarity for a scale model
propeller. Consequently, maintaining an equal advance coefficient for model scale and
full scale propeller is of utmost importance to ensure results of the model test can be
used to make a full scale performance prediction.
If all four parameters 𝐹𝑟, 𝑅𝑒, 𝐶𝑝 , and 𝐽 are equal for two geometrically similar propellers, Hypothetical full
the propellers have the same thrust coefficient 𝐾𝑇 . However, as it is the case with the dynamic
similarity
resistance test, full dynamic similarity is usually not attainable in propeller open water
tests for practical model scales and test conditions. As a consequence, scaling of
the model test results to the full scale propeller requires additional corrections. The
corresponding procedure recommended by the ITTC is discussed in Chapter .

j j
41.2 Propeller Models
Except for small boat propellers, testing of full scale propellers would be unrealistically Propeller model
expensive. However, due to their complicated shape and tight manufacturing tolerances, accuracy
model propellers are typically more expensive than ship models. Accurate leading edge
shapes and thickness distributions are especially difficult to attain. Modern CAD/CAM
and CNC manufacturing tools like D printing help to improve the accuracy of propeller
models. The ITTC distinguishes two classes of model propellers governed by separate
recommended procedures:

• propellers for open water and propulsion tests (ITTC, )

• propellers for cavitation tests (ITTC, ).

Typical model propellers have diameters in the range of  mm ≤ 𝐷 ≤  mm. For Model propellers
open water tests, models should adhere to the following tolerances of their design values for open water
tests
(ITTC, ):

diameter 𝐷 ±0.10 mm
thickness 𝑡 ±0.10 mm
blade width 𝑐 ±0.20 mm
mean pitch at each radius 𝑃 ∕𝐷 ±0.50 %

Surfaces of blades and hub should be polished.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

504 41 Open Water Test

Figure 41.2 Open water test of model propeller with propeller boat in towing tank

j Model propellers Because cavitation phenomena are sensitive to small changes in test parameters, model j
for cavitation propellers for cavitation tests must adhere to more detailed requirements which are doc-
tests
umented in ITTC (). Special attention has to be placed on leading edge geometry
and surface quality.

41.3 Test Procedure

The purpose of the open water test is to determine thrust under different operating
conditions and the associated torque necessary. Open water refers to the fact that
propellers are tested in undisturbed onflow conditions. Tests should be conducted in
accordance with ITTC (). Two different types of testing facilities may be employed:

• towing tanks

• cavitation or circulating water tunnels

In towing tanks undisturbed parallel inflow is generated by moving the rotating propeller
through the water at speed of advance 𝑣𝐴 . In cavitation tunnels water streams with
velocity 𝑣𝐴 past the propeller which is rotating in place.
Towing tank For open water tests in towing tanks, the propeller is mounted on a propeller dy-
namometer, colloquially called propeller boat (Figure .). The propeller boat consists
of a vertical sword with a foil shaped cross section and the streamlined casing, which
contains the propeller shafting and the actual propeller dynamometer. The propeller
dynamometer has sensors for thrust 𝑇 (force in line with shaft), torque 𝑄, and rate

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

41.3 Test Procedure 505

of revolution 𝑛. Advanced propeller dynamometers may feature additional sensors

for transverse bearing forces in the support tube which provide data on propeller side
forces (transverse to axis) and the center of thrust.
Note that the propeller shaft in the open water test is driving the propeller from behind.
The ship hull would be in front of the propeller (see Detail A in Figure .). This allows
the propeller dynamometer to trail the propeller and avoids disturbing the onflow. The
more voluminous parts of the dynamometer should be at least two diameters away
from the propeller to avoid influencing the flow through the propeller disk. During
testing the propeller shaft is submerged by at least 1.5 times the propeller diameter.
This avoids wave generation and other free surface effects like ventilation.
The speed of advance 𝑣𝐴 is measured in front of the propeller at the same depth as the
propeller shaft (see Figure .). In general, it should correspond to the carriage speed.
Before tests with a propeller are conducted, effects of the nose cap are measured by
running the dynamometer with a dummy hub without any blades. Measured forces
and torque are later subtracted from the values found with the actual propeller model.
As mentioned above, full dynamic similarity is usually not attained. In the towing tank Selection of
open water test, the rate of revolution is set to a fixed value 𝑛 so that the local propeller Reynolds
number
Reynolds number 𝑅𝑒𝑐0 is as high as possible.

𝑣1 𝑐0.75
𝑅𝑒𝑐0 = (.)
𝜈
j j
with the approximation of the local blade section velocity 𝑣1

√
𝑣1 = 𝑣𝐴2 + (0.75𝜋 𝑛 𝐷)2 (.)

and the chord length 𝑐0.75 at radius 𝑟 = 0.75𝑅 of the propeller (Figure .). The radius
𝑥 = 0.75 is in the blade region where most of the thrust is generated. Usually the rate
of revolution is limited by the range of the thrust sensor in the propeller dynamometer.
At 𝐽 = 0 a propeller will generate the largest thrust. The advance coefficient is varied
by varying the speed of advance. For each speed of advance setting, values for thrust,
torque, and rate of revolution are measured (see Table . for an example).
If the same propeller is also used in a propulsion test, a second set of open water tests Comparison with
should be conducted for the Reynolds numbers occurring in the propulsion test. In propulsion test
practice, this means the propeller is run at the same Froude numbers and the same rate
of revolutions as the self-propelled model. At a minimum, 𝑅𝑒𝑐0 must be larger than
200 000 to ensure turbulent flow.
For open water tests in a circulating water tunnel or cavitation tunnel, the propeller Cavitation tunnel
dynamometer is fixed to the test section of the tunnel (Figure .). The water is
circulated by a pump. In contrast to the open water test in a towing tank, we keep the
flow speed constant because it is more difficult to control than the propeller rate of
revolution. The advance coefficient is varied by modifying the rate of revolution of the
model propeller. As a consequence, the bollard pull condition (𝐽 = 0) is unattainable.
Depending on the diameter of the model propeller and the size of the cross section of
the tank, wall corrections have to be applied to the test data (Glauert, ).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

506 41 Open Water Test

Figure 41.3 Open water test of model propeller in a circulating water tunnel / cavitation tunnel

41.4 Data Reduction

Thrust and Thrust 𝑇 , torque 𝑄, and rate of revolution 𝑛 are measured at different speeds of advance
torque and converted into three dimensionless coefficients:
coefficients
j 𝑣 j
advance coefficient 𝐽 = 𝐴 (.)
𝑛𝐷

𝑇
thrust coefficient 𝐾𝑇 = (.)
𝜌 𝑛2 𝐷4
𝑄
torque coefficient 𝐾𝑄 = (.)
𝜌 𝑛2 𝐷5

Open water The efficiency of the propeller in open water condition 𝜂𝑂 is given by the ratio of usable
efficiency power and total power spent, i.e. the ratio of thrust power and delivered power.
𝑃𝑇
𝜂𝑂 = (.)
𝑃𝐷

Introducing the definitions for thrust power and delivered power and replacing thrust
and torque using the dimensionless coefficients 𝐾𝑇 (.), 𝐾𝑄 (.), and 𝐽 (.)
yields:

𝑣𝐴 𝑇 𝑣𝐴 𝜌 𝑛2 𝐷4 𝐾𝑇
𝜂𝑂 = =
2𝜋 𝑛 𝑄 2𝜋 𝑛 𝜌 𝑛2 𝐷5 𝐾𝑄
𝑣𝐴 𝐾𝑇 1 𝑣𝐴 𝐾𝑇
= =
2𝜋 𝑛 𝐷 𝐾𝑄 2𝜋 𝑛 𝐷 𝐾𝑄
𝐽 𝐾 𝑇
= (.)
2𝜋 𝐾𝑄

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

41.4 Data Reduction 507

j Figure 41.4 Typical propeller open water diagram j

The results of the open water test are presented in an open water diagram (Figure .). Open water
Thrust coefficient 𝐾𝑇 , torque coefficient 𝐾𝑄 , and open water efficiency 𝜂𝑂 (.) are diagram
plotted as a function of the advance coefficient 𝐽 . The torque coefficient is usually one
order of magnitude smaller than the thrust coefficient. Therefore it is common practice
to plot ten times the torque coefficient. Keep this in mind when you read torque
coefficients off open water diagrams and divide the value by ten before computing
dimensional torque. In Figure . open water efficiency has the same scale as thrust
and torque coefficient but in general they may differ.
Three points in Figure . are marked by circled numbers because they represent
notable operational conditions for the propeller.

1 Bollard pull: the propeller produces maximum thrust for advance coefficient
𝐽 = 0, i.e. initially there is no speed of advance although the propeller is rotating
(𝑛 > 0). This reflects the working condition at the beginning of a tow or simply
when a boat is getting underway. The angles of attack for the blade sections are
very high and close to the geometric pitch angles. High angles of attack will create
high lift forces but also require high torque values. The efficiency is zero because
thrust power vanishes for 𝑣𝐴 = 0. Since the propeller is accelerating fluid in its
onflow it creates its own speed of advance even if the ship is not moving, i.e.
𝐽 > 0 as soon as the propeller starts producing thrust.
2 Maximum efficiency: the efficiency of the propeller grows steadily with increasing
advance coefficient until it reaches a maximum. The propeller shown attains

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

508 41 Open Water Test

its maximum open water efficiency max(𝜂𝑂 ) at 𝐽 = 0.75. Obviously, we would

like the propeller to work close to its maximum efficiency in normal operating
conditions. For practical reasons, the operational advance coefficient is typically
slightly to the left of the maximum efficiency.
3 Zero thrust: at this point the propeller stops producing thrust. Although no
thrust is generated the engine is still providing torque to rotate the propeller. The
torque at zero thrust is a reflection of the propeller’s rotational drag. You may
compare this condition to a bicyclist going downhill. At some point the bike will
roll faster than he or she can pedal. At point 3 the ship is moving so fast that the
propeller is no longer able to accelerate water aft. This condition typically occurs
during stop maneuvers. The engine is powered down but the ship is still going
forward. If the advance coefficient is increased beyond point 3 , the propeller is
generating negative thrust and helps to slow down the vessel.

The open water diagram shows the thrust generating capability of a propeller for all
possible working conditions (advance coefficients 𝐽 ). It plays an important role in
matching a propeller to a given ship hull.

References
Glauert, H. (). The elements of airofoil and airscrew theory. Cambridge University
j Press, Cambridge, United Kingdom, second edition. j
ITTC (). Model manufacture, propeller models, propeller model accuracy. In-
ternational Towing Tank Conference, Recommended Procedures and Guidelines
.---. Revision .
ITTC (). Ship models. International Towing Tank Conference, Recommended
Procedures and Guidelines .---. Revision .
ITTC (). Open water test. International Towing Tank Conference, Recommended
Procedures and Guidelines .---.. Revision .

Self Study Problems

. Why is the advance coefficient 𝐽 used as the equivalent speed of model and full
scale propellers instead of Froude or Reynolds number?
. Discuss the advantages and disadvantages of conducting open water tests in a
towing tank versus conduction them in a circulating water tunnel.
. Explain in your own words the characteristics of the points marked 1 , 2 , and
3 in Figure ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

509

42
Full Scale Propeller Performance

The dimensionless thrust and torque coefficients 𝐾𝑇𝑀 and 𝐾𝑄𝑀 , determined in the
open water test, cannot be directly applied to the full scale vessel. Like the resistance
test, open water tests are conducted under partial dynamic similarity. Characteristic
Reynolds numbers for model propellers are smaller than those for the full scale propeller.
Therefore viscous effects are not replicated to scale in open water tests. In this chapter
we discuss the forces acting on propeller blade sections and the differences between
model scale and full scale caused by the difference in Reynolds numbers. The chapter
closes with a description of the recommended ITTC procedure for full scale predictions
from open water tests.

j Learning Objectives j

At the end of this chapter students will be able to:

• identify forces acting on propeller blade sections

• understand scale effects in open water tests
• predict full scale propeller performance from open water test results

42.1 Comparison of Model and Full Scale Propeller Forces

Figure . shows the forces acting on a foil section of the model propeller blade.
Typically, sections at radius 𝑟 = 0.7𝑅 or 𝑟 = 0.75𝑅 are used as representatives for the
blade because sectional forces are largest in this range of the blade radius.
The total force d𝐹𝑀 on the blade section may be decomposed into two components: Lift and drag
force on model
blade section
(i) a drag force d𝐷𝑀 in the direction of the resultant onflow velocity 𝑣𝑟 , and
(ii) a lift force d𝐿𝑀 perpendicular to 𝑣𝑟 .

The center of effort of blade section forces is located approximately one quarter of the
chord length 𝑐𝑀 from the leading edge. The lift force follows from Kutta-Joukowsky’s
theorem (.), and the drag force is computed from section properties.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

510 42 Full Scale Propeller Performance

Figure 42.1 Fluid forces acting on a propeller blade section at model scale

j Thrust and We are, however, more interested in the thrust generated by the blade section. For this j
torque of model the total blade section force d𝐹𝑀 is split into a component d𝑇𝑀 parallel to the shaft
blade section
and a side force d𝑆𝑀 acting perpendicular to the shaft. The force d𝑇𝑀 is the section’s
contribution to the propeller thrust. The component d𝑆𝑀 perpendicular to the shaft
line is responsible for the torque a propeller exerts on the shaft.
d𝑄𝑀
d𝑆𝑀 =
𝑟𝑀

Forces of the full The same decompositions may be applied to the full scale section force d𝐹𝑆 . The
scale blade decisive difference between the flows over model scale and full scale blade section is the
section
much higher Reynolds number for the full scale flow. Consequently, the drag coefficient
of the full scale blade section is smaller than for the model scale section.

𝐶𝐷𝑆 < 𝐶𝐷𝑀 for propeller blade sections (.)

Figure . shows vectors of both model and full scale section forces. The full scale
forces have been scaled down so that the lengths of the lift force vectors d𝐿𝑀 and d𝐿𝑆
have the same length in the sketch. Since 𝐶𝐷𝑆 < 𝐶𝐷𝑀 , the drag force at full scale is
comparatively smaller than at model scale, i.e. the full scale blade section has a better
lift–drag ratio.
d𝐿𝑆 d𝐿𝑀
> (.)
d𝐷𝑆 d𝐷𝑀
Due to the better lift–drag ratio, the full scale resultant force d𝐹𝑆 points more in the
direction of the propeller shaft than the resultant model force d𝐹𝑀 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

42.2 ITTC Full Scale Correction Procedure 511

Figure 42.2 Comparison of model scale and full scale forces acting on a propeller blade section

As a consequence of the forward rotation, the projection of d𝐹𝑆 onto the shaft line Thrust and
j j
becomes longer and the projection onto the transverse direction is shorter (Figure .). torque of the full
scale blade
Subdividing the full scale blade section force into thrust d𝑇𝑆 and torque d𝑄𝑆 shows section
that the full scale propeller for the equivalent onflow conditions produces relatively
more thrust with less torque than the model propeller.
In summary, we can expect that a full scale propeller is slightly more efficient than the Summary of
model propeller. scale effects

𝐾𝑇𝑀 < 𝐾𝑇𝑆

𝐾𝑄𝑀 > 𝐾𝑄𝑆 (.)
𝜂𝑂𝑀 < 𝜂𝑂𝑆

42.2 ITTC Full Scale Correction Procedure

As part of its  Performance Prediction Method, ITTC provides a procedure to scale
the open water model test results (ITTC, ). The equations yield corrections for
thrust and torque coefficient of the model test. The semi-empirical method is applied
to the test data in Table . to illustrate the procedure.
As outlined in the previous section on open water tests, test results are made dimen- Reduction of
sionless using Equations (.), (.), and (.). model test data

𝑣𝐴𝑀 𝑇𝑀 𝑄𝑀
𝐽= 𝐾𝑇𝑀 = 𝐾𝑄𝑀 = (.–.)
𝑛𝑀 𝐷𝑀 2 4
𝜌 𝑛𝑀 𝐷𝑀 5
𝜌 𝑛2𝑀 𝐷𝑀

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

512 42 Full Scale Propeller Performance

Table 42.1 Example results of an open water test conducted in a towing tank

Model propeller data Test data

number of blades 𝑍= 𝑣𝐴 𝑇𝑀 𝑄𝑀 𝑛𝑀
model scale 𝜆𝑀 =  [m/s] [N] [Nm] [1∕s]
diameter 𝐷𝑀 = . m 0.000 236.74 6.347 15.02
0.330 213.86 5.803 14.96
Section data at radius 𝑥 = 0.75 0.661 191.69 5.304 15.01
chord length 𝑐0.75𝑀 = 0.0733 m 0.990 166.45 4.742 15.02
max. thickness 𝑡0.75𝑀 = 0.0031 m 1.321 142.54 4.218 15.04
pitch–diameter ratio 𝑃 ∕𝐷0.75 = 0.8260 1.651 117.40 3.671 15.04
1.982 91.57 3.071 15.01
Test conditions (fresh water) 2.312 65.75 2.406 14.98
temperature 𝑡𝑤 = . ◦ C 2.641 37.07 1.639 14.98
density 𝜌 = . kg/m3 2.972 5.87 0.794 15.00
kin. viscosity 𝜈 = 1.0157 ⋅ 10−6 m2 /s 3.137 −12.21 0.311 15.02

Table 42.2 Open water characteristics of model propeller (see Table 42.1)

𝐽 𝐾𝑇𝑀 10 𝐾𝑄𝑀 𝜂𝑂𝑀

[–] [–] [–] [–]
j 0.0000 0.44872 0.54683 0.0000 j
0.1003 0.40861 0.50398 0.1294
0.2002 0.36382 0.45758 0.2533
0.2996 0.31549 0.40855 0.3682
0.3498 0.29409 0.38727 0.4228
0.4500 0.24404 0.33775 0.5175
0.5514 0.19600 0.28916 0.5948
0.6490 0.15113 0.23816 0.6555
0.7486 0.09977 0.17787 0.6683
0.8483 0.04340 0.10784 0.5434
0.9493 −0.02314 0.02679 −1.3050

The dimensionless open water characteristics for the propeller from Table . are
listed in Table ..
Model section The corrections for thrust and torque coefficients are based on the difference in drag
drag coefficient coefficients for model and full scale blade sections. Data for the blade section at radius
𝑥 = 0.75 is used to represent the propeller. First, a Reynolds number is computed for
the model blade section using the chord length at radius 𝑥 = 0.75 as reference length.
𝑣1 𝑐0.75𝑀
𝑅𝑒𝑐0 = (.)
𝜈𝑀

𝑣1 is the approximated onflow velocity, which excludes the axial and tangential induced
velocities 𝑢𝑎 and 𝑢𝑡 :
√
𝑣1 = 𝑣𝐴2 + (0.75𝜋 𝑛 𝐷)2 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

42.2 ITTC Full Scale Correction Procedure 513

Table 42.3 Intermediate results for scaling open water characteristics of model propeller (see
Table 42.1)

𝐽 𝑣1 𝑅𝑒𝑐0 𝐶𝐷𝑀 𝐶𝐷𝑆 Δ𝐶𝐷 Δ𝐾𝑇 10 Δ𝐾𝑄

[–] [m/s] [–] [10−3 ] [10−3 ] [10−3 ] [–] [–]
0.0000 7.786 561878 12.0692 7.4928 4.5764 −0.001511 0.01525
0.1003 7.762 560140 12.0779 7.4928 4.5851 −0.001514 0.01528
0.2002 7.809 563527 12.0610 7.4928 4.5682 −0.001509 0.01522
0.2996 7.848 566402 12.0468 7.4928 4.5539 −0.001504 0.01517
0.3498 7.871 568038 12.0387 7.4928 4.5331 −0.001497 0.01510
0.4500 7.921 571653 12.0210 7.4928 4.5114 −0.001490 0.01503
0.5514 7.970 575139 12.0040 7.4928 4.4905 −0.001483 0.01496
0.6490 8.081 583192 11.9654 7.4928 4.4655 −0.001475 0.01488
0.7486 8.180 590341 11.9318 7.4928 4.4316 −0.001464 0.01477
0.8483 8.286 597984 11.8964 7.4928 4.3910 −0.001450 0.01463
0.9493 8.394 605771 11.8610 7.4928 4.3681 −0.001443 0.01455

The drag coefficient for the blade section at model scale is a function of this Reynolds
number.
( )⎡ ⎤
𝑡 0.044 5
𝐶𝐷𝑀 = 2 1 + 2 ⎢ ( − ⎥ (.)
𝑐 ⎢ 𝑅𝑒 )1∕6 (
𝑅𝑒𝑐0
)2∕3 ⎥
⎣ 𝑐0 ⎦
j The Reynolds number 𝑅𝑒𝑐0 should be larger than 200 000. j
The formula for the drag coefficient of the full scale blade section assumes that the Full scale section
Reynolds number (full scale) is so high that even a newly manufactured blade surface drag coefficient
operates in the fully rough flow regime where the drag coefficient becomes independent
of the Reynolds number and solely depends on thickness ratio (𝑡∕𝑐) and roughness ratio
(𝑐0.75𝑆 ∕𝑘𝑝 ) (see Section .).
( )[ (𝑐 )]−2.5
𝑡 0.75𝑆
𝐶𝐷𝑆 = 2 1+2 1.89 + 1.62 log10 (.)
𝑐 𝑘𝑝

The roughness 𝑘𝑝 of a new propeller is assumed to be 0.00003 m (𝑘𝑝 = 30 ⋅ 10−6 m).

Note that the full scale chord length 𝑐0.75𝑆 must be used in Equation (.).
The difference between model (.) and full scale (.) section drag coefficients is Difference in
always positive for typical values of the blade roughness 𝑘𝑝 . drag coefficients

Δ𝐶𝐷 = 𝐶𝐷𝑀 − 𝐶𝐷𝑆 (.)

A negative Δ𝐶𝐷 may occur if the model scale chord length 𝑐0.75𝑀 is erroneously used
in Equation (.) for the full scale section drag coefficient.
With the difference in drag coefficient the following corrections are determined for Corrections for
thrust and torque coefficients. thrust and
torque
𝑃 𝑐𝑍
Δ𝐾𝑇 = −Δ𝐶𝐷 0.3 (.)
𝐷 𝐷
𝑐𝑍
Δ𝐾𝑄 = Δ𝐶𝐷 0.25 (.)
𝐷

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

514 42 Full Scale Propeller Performance

Table 42.4 Predicted open water characteristics of full scale propeller (see Table 42.1)

𝐽 𝐾𝑇𝑆 10 𝐾𝑄𝑆 𝜂𝑂𝑆

[–] [–] [–] [–]
0.0000 0.45023 0.53158 0.0000
0.1003 0.41013 0.48870 0.1339
0.2002 0.36533 0.44236 0.2631
0.2996 0.31700 0.39338 0.3843
0.3498 0.27095 0.34733 0.4957
0.4500 0.22342 0.30040 0.5906
0.5514 0.17528 0.24997 0.6698
0.6490 0.12676 0.19352 0.7314
0.7486 0.07210 0.12720 0.7230
0.8483 0.01261 0.05396 0.3349
0.9493 −0.02170 0.01224 −2.6786

Note that the correction Δ𝐾𝑇 is negative for the thrust coefficient. Table . sum-
marizes the intermediate data, drag coefficients, and thrust and torque coefficient
corrections for the example data from Table ..
Full scale thrust Finally, the full scale thrust and torque coefficients are obtained by subtracting the
and torque corrections from the coefficients obtained in the open water test.
coefficients
j j
𝐾𝑇𝑆 = 𝐾𝑇𝑀 − Δ𝐾𝑇 (.)
𝐾𝑄𝑆 = 𝐾𝑄𝑀 − Δ𝐾𝑄 (.)

Make sure to apply the torque correction to the torque coefficient itself instead of
10 𝐾𝑄𝑀 . Because Δ𝐾𝑇 is negative, the full scale thrust coefficient is larger than its
model scale counter part. As expected, the full scale torque coefficient is smaller. The
full scale open water efficiency 𝜂𝑂𝑆 (.) grows by a few percent.

𝐽 𝐾𝑇𝑆
𝜂𝑂𝑆 = (.)
2𝜋 𝐾𝑄𝑆

Table . shows the predicted full scale open water characteristics for the propeller
and open water test data from Table .. Figure . compares model and full scale
open water characteristics. Note that the thrust and torque coefficient curves for the
full scale propeller lie within the pair of curves for the model. Differences are small and
barely visible for the thrust coefficient. The effect on the torque coefficient is amplified
because we plot ten times its value. An increase in efficiency is clearly visible. In this
example maximum efficiency is % higher at full scale which is significant.

References
ITTC ().  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

42.2 ITTC Full Scale Correction Procedure 515

0.8 0.8
KTM
10KQM
0.7 0.7
OM

KTS
10KQS
[]

0.6 0.6
OS
thrust and torque coefficients KT , 10KQ

[]
0.5 0.5

O
0.4 0.4

open water efficiency

0.3 0.3

0.2 0.2

0.1 0.1

0.0 0.0

0.1 0.1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
advance coefficient J []

Figure 42.3 Comparison of measured open water characteristics and predicted full scale
propeller performance
j j

Self Study Problems

. Explain with a sketch of the forces acting on a blade section why the efficiency of
a full scale propeller is slightly higher than that of its model.
. Given is the following open water test data set. Predict and plot the full scale
open water characteristics.
Model propeller data Test data
number of blades 𝑍= 𝑣𝐴 𝑇𝑀 𝑄𝑀 𝑛𝑀
model scale 𝜆𝑀 =  [m/s] [N] [Nm] [1∕s]
diameter 𝐷𝑀 = . m
0.000 320.50 10.741 14.01
0.387 296.09 10.047 14.00
Section data at radius 𝑥 = 0.75
0.774 267.26 9.226 13.99
chord length 𝑐0.75𝑀 = 0.08281 m
1.161 234.50 8.286 14.02
max. thickness 𝑡0.75𝑀 = 0.00311 m
1.548 198.32 7.234 14.00
pitch–diameter ratio 𝑃 ∕𝐷0.75 = 0.900
1.935 159.20 6.076 13.99
Test conditions (fresh water) 2.322 117.67 4.820 14.00
temperature 𝑡𝑤 = . ◦ C 2.709 74.21 3.473 14.02
density 𝜌 = . kg/m3 3.095 29.32 2.042 14.01
kin. viscosity 𝜈 = 1.0410 ⋅ 10−6 m2 /s 3.482 −16.48 0.534 13.99

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

516

43
Propulsion Test

After discussion of resistance, propeller open water characteristics, and hull–propeller

interaction in general, it is time to explain how the hull–propeller interaction parameters
are determined for a given ship. Deriving values for thrust deduction fraction, wake
fraction, and relative rotative efficiency is the purpose of the propulsion test. In a
propulsion test, propeller thrust, torque, and rate of revolution are measured with a
self-propelled model. This chapter explains the setup of models, methods to conduct
the actual propulsion test, data reduction, and evaluation of hull–propeller interaction
parameters for model and full scale vessel. The chapter closes with an alternative form
of the propulsion test called the load variation test.

j j
Learning Objectives
At the end of this chapter students will be able to:

• understand the purpose of propulsion tests

• prepare and set up propulsion tests

• process and interpret test results

• determine hull–propeller interaction parameters for the model

43.1 Testing Procedure

The propulsion test setup and procedure should follow the guidelines of the Interna-
tional Towing Tank Conference (ITTC, ).
Model setup First, the model used in the resistance test is equipped with a complete drive system of
propeller, shaft, and motor. A propeller dynamometer capable of measuring propeller
thrust, torque, and rate of revolution is integrated into the shaft line. The ship model
will feature the same turbulence generator as during the resistance test. Again, self-
propelled models will be free in surge, heave, roll, and pitch, but sway and yaw motion
are prevented by guides at bow and stern. Typically, sinkage fore and aft is also recorded
in the propulsion test. Figure . shows the typical setup of a model.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.1 Testing Procedure 517

Figure 43.1 Setup of model for propulsion test with single skin friction correction force
(continental method)

Even when conducted with great care, differences occur in dynamic trim between
resistance and propulsion test. Although displacement and velocity are the same, the
propeller will change flow field and pressure distribution especially at the stern. This
introduces considerable uncertainties in thrust deduction fraction and wake fraction
which are determined from a combination of resistance and propulsion test data.
The propeller is often not an exact replica of the actual ship propeller. Model propellers Stock propeller
are quite expensive due to the complicated shape and required tight manufacturing
tolerances. Therefore it is common practice to use a ‘stock propeller’, i.e. a propeller
with the correct general characteristics but not necessarily the exact blade shape and
j section profiles. Most importantly, it must have the correct diameter 𝐷 according to the j
geometric scale 𝜆 of the hull. In fact, the availability of a stock propeller often dictates
the scale 𝜆 of the model. Furthermore, the stock propeller’s pitch–diameter ratio should
match the pitch–diameter ratio of the actual propeller design. Typically number of
blades 𝑍 and expanded area ratio 𝐴𝐸 ∕𝐴0 are of lesser concern but should be matched
whenever possible. If the actual propeller has special features, like for example a highly
varying pitch distribution to reduce tip and hub cavitation or high skew blades, for
which a suitable stock propeller is not readily available, manufacturing a suitable model
should be strongly considered.
We conduct the propulsion test to find the hull–propeller interaction parameters wake Effect of
fraction 𝑤, thrust deduction fraction 𝑡, and relative rotative efficiency 𝜂𝑅 . Besides other propeller loading
parameters these quantities depend on the propeller loading (Harvald, ).
𝑇
𝐶𝑇 ℎ = 2
(.)
1
2
𝜌 𝑣𝐴 𝜋 4𝐷

A smaller propeller diameter 𝐷 for the same ship will increase the thrust loading
coefficient 𝐶𝑇 ℎ (.). Experiments have shown that a larger thrust loading will result in
a higher wake fraction 𝑤. This is somewhat confusing because we introduced the wake
fraction as the influence of a hull onto the propeller. You have to remind yourself that
the hull–propeller interaction parameters are just a model to explain and to quantify the
effects. However, they are not fundamental laws of physics and have to be interpreted
with care.
Ideally, we would conduct model tests under full dynamic similarity. This is, unfortu- Partial dynamic
nately, not possible as was explained in Chapter . Like resistance tests, propulsion similarity

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

518 43 Propulsion Test

Figure 43.2 The relative difference of resistance for model and full scale vessel

j j
tests are performed satisfying Froude’s law. For practical reasons the same Froude
numbers are run as in the resistance test. Consequently, the Reynolds numbers in the
model test are lower than at full scale. Due to the lower Reynolds numbers at model
scale, viscous effects in the resistance are disproportionately larger compared to the
full scale vessel. Thus, the model propeller would have a higher thrust loading than the
full scale propeller because it has to overcome a comparatively higher resistance.
Thrust loading Figure . illustrates the difference in thrust loading between model and full scale
difference vessel. It shows resistance coefficients of model and full scale vessel as a function of
the Reynolds number 𝑅𝑒. Model and ship are connected by the ITTC  model–
ship correlation line coefficient 𝐶𝐹 and the form factor 𝑘. They share the same wave
resistance coefficient 𝐶𝑊 . Compared to the full scale propeller, a model propeller has
to overcome an additional frictional resistance of (1 + 𝑘)(𝐶𝐹𝑀 − 𝐶𝐹𝑆 ) − Δ𝐶𝐹 . Note that
the roughness allowance Δ𝐶𝐹 increases the loading only for the full scale propeller. As
a result, the roughness allowance lowers the difference in frictional resistance between
model and full scale vessel.
Skin friction The difference in loading condition is compensated in the propulsion test by applying
correction force an additional towing force 𝐹𝐷 . It corrects the otherwise excessive loading of the model
propeller during the propulsion test. The computation of this skin friction correction
force 𝐹𝐷 is documented in Section ..
[ ( ) ]
1 2
𝐹𝐷 = 𝜌 𝑣𝑀 𝑆𝑀 (1 + 𝑘) 𝐶𝐹𝑀 − 𝐶𝐹𝑆 − Δ𝐶𝐹 (.)
2
𝐹𝐷 has to be precomputed for each model speed 𝑣𝑀 . This is known as the constant
loading method. In Europe this is also called the continental method. Alternatively, a

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.2 Data Reduction 519

series of tests may be conducted with the same model speed but varying external tow
force 𝐹𝐷 . This load variation method is described in Section ..
For models with extensive appendages like exposed shafts and struts, it is recommended Correction for
to augment the skin friction correction by an appendage drag correction (ITTC, ). appendage drag

[ ( ) ] [ ]
1 2
𝐹𝐷 = 𝜌 𝑣𝑀 𝑆𝑀 (1 + 𝑘) 𝐶𝐹𝑀 − 𝐶𝐹𝑆 − Δ𝐶𝐹 + (1 − 𝛽) 𝑅𝑇𝑀𝐴 − 𝑅𝑇𝑀 (.)
2

𝑅𝑇𝑀𝐴 is the total resistance of the model with appendages and 𝑅𝑇𝑀 the total resistance
without appendages. The correction factor 𝛽 for the appendage drag takes values
between 0.6 and 1.0 with a typical value of 0.75 for conventional twin screw vessels
with bossings, shafts, and brackets.
During testing the model is propelled by its propeller (or propellers). For each test speed Model test
the rate of revolution of the propeller is adjusted so that the model maintains the same
speed as the towing carriage. Only when any initial difference in speed has vanished
is the recording of thrust 𝑇𝑀 , torque 𝑄𝑀 , and rate of revolution 𝑛𝑀 started. In the
past, the skin friction correction force was applied via a pulley and calibrated weights.
Nowadays, the towing staff of the resistance test is used to measure the remaining
tow force and the propeller rate of revolution is adjusted until the remaining tow force
(resistance) is equal to 𝐹𝐷 . This has the additional advantage that the model speed is
accurately defined by the carriage speed.

j j

43.2 Data Reduction

At the end of a test series, model thrust 𝑇𝑀 , torque 𝑄𝑀 , and rate of revolution 𝑛𝑀 have
been measured for given sets of model speed 𝑣𝑀 and skin friction correction values 𝐹𝐷 .
The model speed has been derived from the carriage speed by applying appropriate
blockage corrections (see Section .).
Thrust and torque are rendered dimensionless like the results of an open water test. Thrust and
torque
coefficients
𝑇𝑀 𝑄𝑀
𝐾𝑇𝑀 = 𝐾𝑄𝑀 = (.)
𝜌𝑀 𝑛2𝑀 4
𝐷𝑀 5
𝜌𝑀 𝑛2𝑀 𝐷𝑀

The difference here is that thrust and torque coefficient in Equation (.) have been
measured in the behind condition.
The open water characteristics of the propeller employed in the propulsion test have Behind condition
been plotted in an open water diagram which shows 𝐾𝑇𝑂 and 10 𝐾𝑄𝑂 as a function of
the advance coefficient 𝐽 = 𝑣𝐴 ∕(𝑛𝐷) (see Chapter ). Although we measured thrust,
torque, and rate of revolution in the propulsion test, we are missing a crucial piece of
information. During an open water test the speed of advance is equal to the carriage
speed. Obviously, we cannot make the same determination for the propeller working in
the wake of a hull (behind condition). However, combining results of open water and
propulsion test will allow us to determine the advance ratio and thus speed of advance
of the propeller in the behind condition.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

520 43 Propulsion Test

[−]
0.6
open water
test results

KTM , 10KQM
0.5
KTMO KTMO
10KQTM
0.4 10KQMO

0.3
self propulsion
test result KTM
0.2 10KQMO
self propulsion point
KTM = KTMO
0.1
(thrust identity)
0.0
0.0 0.2 0.4 0.6 0.8 1.0
JTM advance coefficient J [−]

Figure 43.3 Self propulsion point of model propeller under the assumption of thrust identity

43.3 Hull–Propeller Interaction Parameters

Self propulsion From the results of the propulsion test and from the open water test of the stock
j point propeller we have to determine the operating condition of the propeller behind the ship j
model, the self propulsion point. Further progress requires us to make an assumption:

(i) We could assume that the model propeller is producing the same thrust in behind
and open water condition for the same advance coefficient: 𝐾𝑇𝑀 = 𝐾𝑇𝑀 𝑂 . This
is called thrust identity.

(ii) Alternatively, we could assume that the model propeller requires the same torque
in behind and open water condition for the same advance coefficient 𝐽 : 𝐾𝑄𝑀 =
𝐾𝑄𝑀 𝑂 . This is called torque identity.

Probably both, thrust and torque, differ between open water and behind condition but
we do not know exactly by how much. In practice, the assumption of thrust identity is
preferred as it tends to provide more accurate performance predictions for the full scale
vessel. For one, thrust is less affected by different model scales and its measurement
is less susceptible to errors caused by friction in bearings and on propeller blades.
Working under the assumption of torque identity has the advantage that torque is more
easily determined during ship trials to confirm the powering prediction.
Thrust identity Working under the assumption of thrust identity, we take the thrust coefficient 𝐾𝑇𝑀
(.) from the propulsion test and search for the corresponding thrust coefficient
𝐾𝑇𝑀 𝑂 determined under open water condition. Figure . explains the process. We
employ the open water diagram of the model propeller. It represents the capabilities
of a propeller in uniform flow, i.e. open water condition. The intersection of open
water thrust coefficient curve 𝐾𝑇𝑀 𝑂 and a horizontal line drawn through the thrust
coefficient 𝐾𝑇𝑀 of the propulsion test (.) marks the self propulsion point. A vertical

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.3 Hull–Propeller Interaction Parameters 521

0.6
self propulsion point open water
0.5 KQM = KQMO test results
KTMO (torque identity) KTMO
self propulsion 0.4 10KQMO
test result 10KQM
0.3
[−]

0.2 10KQMO
KTQM
KTM , 10KQM

0.1

0.0
0.0 0.2 0.4 0.6 0.8 1.0
JQM advance coefficient J [−]

Figure 43.4 Self propulsion point of model propeller under the assumption of torque identity

line drawn through the self propulsion point defines advance coefficient 𝐽𝑇𝑀 and open
water torque coefficient 10 𝐾𝑄𝑇𝑀 . The additional subscript 𝑇 refers to the assumption
of thrust identity.

j Note that the self propulsion point changes with model speed! j
For the sake of completeness, Figure . shows the equivalent process under the Torque identity
assumption of torque identity. The self propulsion point is defined by the intersection of
the open water torque coefficient curve 10 𝐾𝑄𝑀 𝑂 and the horizontal line drawn through
ten times the torque coefficient 10 𝐾𝑄𝑀 in the behind condition from Equation (.).
Advance coefficient 𝐽𝑄𝑀 and thrust coefficient 𝐾𝑇𝑄𝑀 for the open water condition are
read off the open water diagram. Obviously, the additional subscript 𝑄 refers to the
assumption of torque identity.

43.3.1 Model wake fraction

We continue our data analysis assuming thrust identity. The self propulsion point is
defined by the combination of advance coefficient 𝐽𝑇𝑀 , thrust coefficient 𝐾𝑇𝑀 , and
torque coefficient 𝐾𝑄𝑇𝑀 .
Wake fraction for the model is defined as the difference between model speed 𝑣𝑀 and Speed of
the speed of advance 𝑣𝐴𝑀 . The latter may now be derived from the advance coefficient advance
of the self propulsion point. From 𝐽𝑇𝑀 = 𝑣𝐴𝑀 ∕(𝑛𝑀 𝐷𝑀 ) follows

𝑣𝐴𝑀 = 𝐽𝑇𝑀 𝑛𝑀 𝐷𝑀 (.)

The diameter 𝐷𝑀 of the propeller is of course known and 𝑛𝑀 is the rate of revolution
recorded in the propulsion test for the thrust coefficient x value measured for the model
velocity 𝑣𝑀 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

522 43 Propulsion Test

Wake fraction The definition of the wake fraction (.) provides the path to compute it for the self-
propelled model:
𝐽 𝑛 𝐷
𝑤𝑇𝑀 = 1 − 𝑇𝑀 𝑀 𝑀 (.)
𝑣𝑀
The subscript 𝑇 is again used to emphasize the assumption of thrust identity. The wake
fraction 𝑤𝑄𝑀 based on torque identity is obtained by exchanging 𝐽𝑇𝑀 with 𝐽𝑄𝑀 .
The wake fraction is subject to significant viscous effects (see Chapter ). Since the
Reynolds number is too small at model scale, the model wake fraction has to be corrected
before it is applied to the full scale vessel. The full scale wake fraction is expected to be
smaller than the model wake fraction because viscous effects are proportionally larger
at model scale. The necessary correction of the model wake fraction for the full scale
vessel is explained in Section ...

43.3.2 Thrust deduction fraction

Definition In Chapter  thrust deduction fraction 𝑡 was defined as the portion of thrust that
must be spent to overcome the difference in resistance of the vessel with and without a
propeller.
𝑇 − 𝑅𝑇
𝑡 = (.)
𝑇
Adaption of test We have to adapt this formula to the specific conditions of the propulsion test.
j results j
• During the propulsion test, we reduced the actual model resistance by the skin
friction correction 𝐹𝐷 (.) in order to correct the propeller loading.

• The resistance from the resistance test 𝑅𝑇𝑀 may have to be corrected for changes
in water temperature between the dates of the resistance and propulsion tests. The
correction procedure is documented in the ITTC  Performance Prediction
Method (ITTC, ).
[ ]
(1 + 𝑘)𝐶𝐹𝑀𝐶 + 𝐶𝑅
𝑅𝐶 = 𝑅𝑇𝑀 (.)
(1 + 𝑘)𝐶𝐹𝑀 + 𝐶𝑅

𝐶𝐹𝑀 is the ITTC  model–ship correlation coefficient from the resistance
test. 𝐶𝐹𝑀𝐶 is the equivalent friction coefficient for the water temperature of the
propulsion test. A change in water temperature results in a different kinematic
viscosity 𝜈𝑀 . Thus, the Reynolds number changes despite equal model speeds.

Prediction With the corrections above, the thrust deduction fraction is computed as

𝑇𝑀 − (𝑅𝑇𝑀 − 𝐹𝐷 )
𝑡 = (.)
𝑇𝑀

You may have noticed that the symbol for the thrust deduction fraction 𝑡 does not
feature a subscript 𝑀 for the model. In contrast to the wake fraction, viscous effects
on the thrust deduction fraction are small. In lieu of a better prediction, the thrust
deduction fraction (.) of the model is also used for the full scale ship.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.3 Hull–Propeller Interaction Parameters 523

43.3.3 Relative rotative efficiency

Relative rotative efficiency 𝜂𝑅 is defined as the ratio of propeller efficiency in the be- Change in
hind condition 𝜂𝐵 to the open water efficiency 𝜂𝑂 (uniform flow) at the same advance propeller
efficiency
coefficient 𝐽 .
𝜂
𝜂𝑅 = 𝐵 (.)
𝜂𝑂
Propeller efficiency is equal to the ratio of thrust power to delivered power. For the
behind and open water condition we have

𝑃𝑇 𝑇 𝑣𝐴 𝑃𝑇 𝑜 𝑇𝑂 𝑣𝐴
𝜂𝐵 = = 𝜂𝑂 = = (.)
𝑃𝐷 2𝜋 𝑛𝑄 𝑃𝐷𝑜 2 𝜋 𝑛 𝑄𝑂

Noting that speed of advance 𝑣𝐴 and rate of revolution 𝑛 are the same in both cases
(equal 𝐽 ), relative rotative efficiency is equal to

𝑇 𝑣𝐴
2𝜋 𝑛𝑄 𝑇 𝑄𝑂 𝐾𝑇 𝐾𝑄𝑂
𝜂𝑅 = = = (.)
𝑇𝑂 𝑣𝐴 𝑇𝑂 𝑄 𝐾𝑇 𝑂 𝐾𝑄
2 𝜋 𝑛 𝑄𝑂

We apply this to the results of the propulsion test. Consistent with the selection of Assuming thrust
j the self propulsion point, we use either thrust identity (assuming 𝑇 = 𝑇𝑂 ) or torque identity j
identity (assuming 𝑄 = 𝑄𝑂 ). Under the assumption of thrust identity, we determined
the open water torque 𝐾𝑄𝑇𝑀 . In the propulsion test we measured the torque 𝐾𝑄𝑀 for
the behind condition. We substitute both into Equation (.), and with thrust identity
𝑇 = 𝑇𝑂 we obtain the relative rotative efficiency.

𝐾𝑄𝑇𝑀
𝜂𝑅 = (.)
𝐾𝑄𝑀

If the propulsion test would be analyzed using torque identity the relative rotative Assuming torque
efficiency would be 𝜂𝑅 = 𝐾𝑇𝑀 ∕𝐾𝑇𝑄𝑀 . identity

43.3.4 Full scale hull–propeller interaction parameters

Results of a propulsion test are wake fraction 𝑤𝑇𝑀 , thrust deduction fraction 𝑡, and Thrust deduction
relative rotative efficiency 𝜂𝑅 of the model. The latter two show no significant scale fraction, relative
rotative
effects and the values found at model scale are also applied to the full scale vessel. efficiency

𝑡 = 𝑡𝑀 = 𝑡𝑆 and 𝜂𝑅 = 𝜂𝑅𝑀 = 𝜂𝑅𝑆 (.)

The wake fraction, however, is heavily influenced by viscous effects. As stated before, Full scale wake
viscous effects are more pronounced at model scale than at full scale due to the lower fraction
Reynolds number. Therefore, the model wake fraction 𝑤𝑇𝑀 (𝑤𝑄𝑀 ) found under thrust
(torque) identity is typically larger than the full scale wake fraction 𝑤𝑇𝑆 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

524 43 Propulsion Test

The ITTC recommended procedures supply the following formula to predict the full
scale wake fraction (ITTC, ).
[ ] (1 + 𝑘)𝐶𝐹𝑆 + Δ𝐶𝐹
𝑤𝑇𝑆 = (𝑡 + 𝑤𝑟 ) + 𝑤𝑇𝑀 − (𝑡 + 𝑤𝑟 )
(1 + 𝑘)𝐶𝐹𝑀 (.)
if 𝑤𝑇𝑆 > 𝑤𝑇𝑀 use 𝑤𝑇𝑆 = 𝑤𝑇𝑀
This equation needs some explaining.
Components of Chapter  (page .) described the wake fraction as the sum of three parts: potential
wake fraction wake 𝑤𝑝 , frictional wake 𝑤𝑓 , and wave wake 𝑤𝑤 .
𝑤 = 𝑤𝑝 + 𝑤𝑓 + 𝑤𝑤 (.)
The wave wake fraction is small compared to the other two parts. Since Froude number
for model and ship are equal, we expect the wave systems of model and ship to be
geometrically similar. Omitting the wave wake fraction, we are left with two parts:
𝑤 = 𝑤𝑝 + 𝑤𝑓 (.)
This is also the form of Equation (.). The first part is the potential wake fraction of
the ship and the second part is its frictional wake fraction.
𝑤𝑝𝑆 = 𝑡 + 𝑤𝑟 (.)
[ ] (1 + 𝑘)𝐶𝐹𝑆 + Δ𝐶𝐹
𝑤𝑓 𝑆 = 𝑤𝑇𝑀 − (𝑡 + 𝑤𝑟 ) (.)
j (1 + 𝑘)𝐶𝐹𝑀 j
Potential wake Now, what has the thrust deduction fraction 𝑡 to do with the potential wake fraction?
fraction This is best explained with a thought experiment taken from Harvald (). Imagine
a flow around the ship without any energy losses. We may model this condition as
potential flow. In this hypothetical case, the energy 𝑃𝐸 = 𝑅𝑇 𝑣 spent moving the vessel
through the fluid must be equal to the work 𝑃𝑇 = 𝑇 𝑣𝐴 done by the propulsor.
𝑅𝑇 𝑣 = 𝑇 𝑣𝐴

Resistance may be replaced by 𝑅𝑇 = (1 − 𝑡𝑝 )𝑇 . The thrust deduction fraction is reduced

to its potential part 𝑡 = 𝑡𝑝 because there is no friction in potential flow, and the effect
of the ship wave pattern on the thrust deduction fraction is negligible. Ship speed
multiplied by (1 − 𝑤𝑝 ) may replace the speed of advance 𝑣𝐴 . The wake fraction is also
equal to its potential part 𝑤 = 𝑤𝑝 because there is no frictional wake in potential flow,
and we ignored the small wave wake. We substitute both into the energy balance above.
(1 − 𝑡𝑝 ) 𝑇 𝑣 = 𝑇 (1 − 𝑤𝑝 )𝑣

Dividing by 𝑇 𝑣 yields

(1 − 𝑡𝑝 ) = (1 − 𝑤𝑝 )

For this equation to hold true, it is necessary that wake fraction and potential thrust
deduction fraction are of the same size.

𝑡𝑝 = 𝑤𝑝

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.4 Load Variation Test 525

As outlined in Chapter , the potential part is the largest contribution to the overall
thrust deduction fraction even in viscous flows. Thus, 𝑡𝑝 ≈ 𝑡 and it is reasonable to
substitute the thrust deduction fraction for the potential wake fraction 𝑤𝑝 .

𝑤𝑝 = 𝑡 (.)

The additional term 𝑤𝑟 in Equation (.) represents the wake effect of the rudder.

𝑤𝑝𝑆 = 𝑤𝑝𝑀 = 𝑡 + 𝑤𝑟 (.)

If it is not determined separately, a value of 𝑤𝑟 = 0.04 may be assumed (ITTC, ).

In Equation (.) for the frictional wake fraction of the ship, the first term in brackets Frictional wake
is equal to the difference of model wake fraction 𝑤𝑇𝑀 and (𝑡 + 𝑤𝑟 ). The latter, we just
determined, is the potential wake fraction which applies to model and ship alike (no
scale effects). The difference 𝑤𝑇𝑀 − (𝑡 + 𝑤𝑟 ) is therefore the frictional wake fraction of
the model.
𝑤𝑓 𝑀 = 𝑤𝑇𝑀 − 𝑤𝑝𝑀 = 𝑤𝑇𝑀 − (𝑡 + 𝑤𝑟 ) (.)
The frictional wake of the model cannot be applied to the full scale vessel without
correction. A suitable correction factor is represented by the final fraction in Equa-
tion (.). It represents the ratio of frictional resistances of ship and model including
the form effect. Since 𝐶𝐹𝑆 is considerably smaller than 𝐶𝐹𝑀 for most scale models, the
ratio should be smaller than one.
j (1 + 𝑘)𝐶𝐹𝑆 + Δ𝐶𝐹 j
< 1 (.)
(1 + 𝑘)𝐶𝐹𝑀

The full scale wake fraction must be smaller than the model wake fraction. In some Note:
rare cases, Equation (.) produces a 𝑤𝑇𝑆 larger than 𝑤𝑇𝑀 . This makes no sense 𝑤𝑇𝑆 ≤ 𝑤𝑇𝑀
physically since the frictional effects should always be smaller on the full scale vessel.
Therefore, if the result of Equation (.) is larger than 𝑤𝑇𝑀 , the model wake fraction
should be applied to the full scale vessel instead.

43.4 Load Variation Test

In the continental method described above, a single pre-calculated skin friction cor- Problems with
rection force is employed to ensure a propeller loading which is dynamically similar to continental
method
the loading of the full scale propeller (continental method). Values of the hull propeller
interaction parameters are quite sensitive to propeller loading. Harvald and Hee ()
tested a single hull with different model propellers. Significant scatter was found in
the resulting thrust deduction fraction. Thus, uncertainties in the estimate of the skin
friction correction 𝐹𝐷 will produce uncertainties in the prediction of the hull–propeller
interaction parameters.
Alternatively, propulsion tests may be repeated at a specific speed each time varying British method
the pulling forces 𝐹𝑃 which replaces the skin friction correction force. This varies the
propeller loading and is also known as a load variation test or for short the British
method.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

526 43 Propulsion Test

Figure 43.5 Setup of model for propulsion test with load variation (British method)

Setup of load Figure . shows the required measurement setup. Essentially, the model equipped for
j variation test propulsion is hooked up to the resistance dynamometer. For a selected velocity multiple j
runs are made with the self-propelled model each time increasing the propeller rate of
revolution 𝑛𝑀 . In the first run, the propeller might be standing still (𝑛𝑀 = 0 rpm). The
resistance dynamometer will register a pulling (tow) force 𝐹𝑃 which will be larger than
the resistance 𝑅𝑇𝑀 of the hull alone. The propeller with its blades perpendicular to the
flow generates an additional drag. The propeller dynamometer will measure a negative
thrust because the drag force on the propeller is pulling at the shaft.
In a second run, the propeller RPM is set to a small positive value. Let us say it turns just
fast enough to produce the necessary thrust to compensate for the additional resistance
caused by the propeller. The measured thrust is zero (𝑇𝑀 = 0 N). The propeller is
turning but not yet pushing the model. The resistance dynamometer measures a towing
force which may be interpreted as the total resistance of the complete hull–propeller
system.
𝑅𝑃𝑀0 = 𝐹𝑃 (𝑇𝑀 = 0 N) (.)
This point is marked by 1 in Figure .. The propeller is essentially idling along. Since
no thrust is produced, the propeller loading is zero.
Kracht () reports that the ratio of total resistance with running propeller 𝑅𝑃𝑀0 and
total resistance 𝑅𝑇𝑀 of hull alone is slightly larger than one for single screw vessels and
slightly smaller than one for slender twin screw vessels. Holtrop () states that for
single screw vessels 𝑅𝑃𝑀0 ∕𝑅𝑇𝑀 ∈ [1.01, 1.04] in % of the investigated cases.
In third and subsequent runs, the rate of revolution is increased until the remaining
pulling force 𝐹𝑃 becomes negative. The model starts pushing the towing carriage. What
is astonishing at first is that the relationship between the towing force 𝐹𝑃 and the thrust
𝑇𝑀 is fairly linear. This is true over a wide range of speeds and ship hull forms. Slight

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.4 Load Variation Test 527

j j
Figure 43.6 Typical results of a load variation test (British method)

nonlinearities seem only to occur for very full vessels with small propellers and in the
range of negative 𝐹𝑃 forces (Kracht, ; Holtrop, ).
When the pulling force vanishes (𝐹𝑝 = 0 N), the model is completely self-propelled. It Model self
produces the thrust 𝑇𝑀0 = 𝑇𝑀 (𝐹𝑃 = 0 N) necessary to overcome the resistance of the propulsion point
model hull–propeller system at the given speed. 2 marks this point in Figure ..
The thrust 𝑇𝑀0 must be equal to the resistance of the hull without effects due to the
propeller 𝑅𝑃𝑀0 plus the increase in resistance caused by the propeller. The latter we
defined as the thrust deduction 𝑡 𝑇𝑀0 (see Section .).
𝑅𝑃𝑀0
𝑇𝑀0 = 𝑅𝑃𝑀0 + 𝑡 𝑇𝑀0 or 𝑇𝑀0 = (.)
1−𝑡
The corresponding self propulsion point from the continental method with fixed skin
friction correction force 𝐹𝐷 is marked with 3 .
The slope of the 𝐹𝑝 –𝑇𝑀 line shows no or only insignificant variation across the range
of ship speeds (Holtrop, ). Lines for two smaller Froude numbers have been added
into Figure . at 4 .
How can we determine the thrust deduction fraction 𝑡 based on the result of the Balance of
load variation test? Considering the forces acting in longitudinal direction on the longitudinal
forces
self-propelled model, we have

• thrust 𝑇𝑀 and pulling force 𝐹𝑃 acting in the direction of motion, and

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

528 43 Propulsion Test

• resistance of vessel with running propeller 𝑅𝑃𝑀0 and the thrust deduction 𝑡 𝑇𝑀
acting against the forward movement of the model.

The thrust deduction 𝑡𝑇𝑀 captures the effects on flow pattern and pressure distribution
at the stern.
For a model running at constant velocity these forces must be in equilibrium.
∑
𝐹𝑥 = 0 = 𝑇𝑀 + 𝐹𝑃 − 𝑅𝑃𝑀0 − 𝑡 𝑇𝑀 (.)

We solve Equation (.) for the thrust 𝑇𝑀 of the propeller.

𝑅𝑃𝑀0 − 𝐹𝑃
𝑇𝑀 = (.)
1−𝑡
This is the equation that describes the line in Figure .. Substituting 𝐹𝑃 = 0 N into
Equation (.) yields the self propulsion point:
𝑅𝑃𝑀0
𝑇𝑀 (𝐹𝑃 = 0 N) = 𝑇𝑀0 = (see Equation (.)) (.)
1−𝑡
For vanishing thrust 𝑇𝑀 = 0 N Equation (.) resolves into 𝐹𝑃 = 𝑅𝑃𝑀0 as expected.
Thrust deduction Taking the derivative of the thrust 𝑇𝑀 (Equation (.)) with respect to the pulling
fraction force 𝐹𝑃 results in the expression shown at 5 in Figure ..
j d𝑇𝑀 −1 j
= (.)
d𝐹𝑃 1−𝑡
The thrust deduction fraction defines the slope of the line between the self propulsion
point 2 and the point of vanishing thrust 1 . The fact that the slope does not change for
different Froude numbers indicates that the thrust deduction fraction 𝑡 is independent
of speed which is confirmed by practical experience (Holtrop, ). In fact, one pair of
(𝑅𝑃𝑀0 , 𝑇𝑀0 ) values from a load variation test for a single speed is sufficient to compute
the thrust deduction fraction. Rearrangement of Equation (.) results in:
𝑅𝑃𝑀0
𝑡 = 1− (.)
𝑇𝑀0

Unlike the continental method, load variation tests do not require results from a resis-
tance test. Both, 𝑅𝑃𝑀0 and 𝑇𝑀0 are results from the same series of tests. Uncertainties
from differences in dynamic trim are eliminated. Thrust deduction fractions derived
via Equation (.) show less spread and are considered more physically sound.
The wake fraction is typically increasing with decreasing propeller loading (higher
pulling force 𝐹𝑃 ). Based on the results of the load variation tests, effects of propeller
loading on the wake fraction may be investigated. Again, the concept of thrust identity
will be employed to derive the advance coefficient from the propeller open water data.
The process is described in Section . above.
The following chapter summarizes the ITTC  performance prediction method
which combines results of resistance test, open water test, and propulsion test to predict
the necessary delivered power for trial conditions.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

43.4 Load Variation Test 529

References
Harvald, S. (). Resistance and propulsion of ships. John Wiley & Sons, New York,
NY.
Harvald, S. and Hee, J. (). The components of the propulsive efficiency of ships in
relation to the design procedure. In th Ship Technology and Research Symposium
(STAR), Pittsburgh, PA.
Holtrop, J. (). Are model resistance tests indispensable? In th International
Towing Tank Conference (ITTC), volume , pages –, Madrid, Spain.
Holtrop, J. (). Extrapolation of propulsion tests for ships with appendages and
complex propulsors. Marine Technology, ():–.
ITTC (). Propulsion/bollard pull test. International Towing Tank Conference,
Recommended Procedures and Guidelines .---.. Revision .
ITTC ().  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
Kracht, A. (). Load variation tests improve the reliability of ship power predictions
based on model test results. Schiffstechnik / Ship Technology Research, ():–.

Self Study Problems

j j
. Explain why a skin friction correction force is needed in the propulsion test.
. Discuss the difference between thrust identity and torque identity.
. Why can the thrust deduction fraction of the model be applied to the full scale
vessel but not the model wake fraction.
. In your own words, state the difference between the continental and the British
method.
. Why is the thrust deduction fraction derived from load variation tests considered
more reliable than the thrust deduction derived from resistance and propulsion
test using the continental method?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

530

44
ITTC 1978 Performance Prediction
Method

The ITTC  performance prediction method is a procedure to predict the necessary
delivered power of a vessel for trial conditions. It is part of the extensive set of model
testing procedures recommended by the International Towing Tank Conference (ITTC,
www.ittc.info). The power prediction is based on the results of resistance test, open
water test, and propulsion test. The individual tests have been discussed in detail in
Chapters , , , , and . Even though only very few naval architects conduct
model tests themselves, in order to interpret and use the results in ship design it is
important to understand the process and its inherent limitations.

j j
Learning Objectives
At the end of this chapter students will be able to:

• discuss the prediction of delivered power based on model tests

• determine delivered power for trial conditions based on model tests

44.1 Summary of Model Tests

The results of three different model tests contribute to the prediction of the full scale
vessel performance in the ITTC  performance prediction method (ITTC, a):

(i) Resistance test: a hull without a propeller is towed through the water. Total
resistance 𝑅𝑇𝑀 is measured as a function of model speed (Froude number). From
the model test results follow form factor 𝑘 and total resistance 𝑅𝑇𝑆 of the full
scale vessel.

(ii) Open water test: the performance characteristics of a propeller are observed
while it operates in undisturbed parallel flow. Thrust 𝑇𝑀 , torque 𝑄𝑀 , and rate of
revolution 𝑛𝑀 are measured as a function of advance ratio 𝐽 . Full scale propeller
characteristics are derived from the model test results.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44.2 Full Scale Power Prediction 531

(iii) Propulsion test: a complete hydrodynamics system of hull and propeller is moving
under its own power through the water. Thrust 𝑇𝑀 , torque 𝑄𝑀 , and rate of
revolution 𝑛𝑀 are measured as a function of model speed.

44.2 Full Scale Power Prediction

With the full scale thrust deduction fraction and wake fraction as the final puzzle pieces
we can predict the delivered power for the ship. During self propulsion tests, thrust and
rate of revolution are measured for the model. This enabled us to determine the self
propulsion point of the model propeller: first the advance ratio under the assumption
of thrust (or torque) identity and then the speed of advance from advance ratio and
rate of revolution. Obviously, there are no equivalent full scale measurements available
before the ship is built.
From total resistance and thrust deduction fraction we compute the thrust that the Required thrust
ship requires for each tested speed:
𝑅𝑇𝑆
𝑇𝑆 = (.)
(1 − 𝑡)
The wake fraction allows us to estimate the speed of advance at which the propeller
operates.
𝑣𝐴𝑆 = (1 − 𝑤𝑇𝑆 )𝑣𝑆 (.)
j j
Both required thrust and speed of advance will vary with ship speed 𝑣𝑆 . Next, the
thrust requirement of the ship has to be matched with the thrust capability of the full
scale propeller in order to find the ship’s self propulsion point.
The capabilities of the propeller are recorded in its open water diagram. Specifically, the Propeller
thrust coefficient curve 𝐾𝑇𝑆 𝑂 represents the thrust capability of the propeller under capabilities
open water conditions (Figure .).
Finding the rate of revolution 𝑛𝑆 when propeller diameter 𝐷𝑆 , speed of advance 𝑣𝐴𝑆 , Shortened thrust
and thrust 𝑇𝑆 are known is task  of the propeller design problems (Chapter ). In loading
coefficient
task  the self propulsion point is found by intersecting the parabola 𝐶𝑆 𝐽 2 with the
thrust coefficient curve 𝐾𝑇𝑂 of the propeller. The shortened thrust loading coefficient
𝐶𝑆 is derived from the definition of the thrust coefficient 𝐾𝑇 . The unknown rate of
revolution 𝑛 is eliminated by dividing Equation (.) with 𝐽 2 :
𝐾𝑇𝑆 𝑇𝑆
𝐶𝑆 = = (.)
𝐽2 𝜌𝑆 𝐷𝑆2 𝑣𝐴𝑆
2

All quantities on the right-hand side are known. Applying Equations (.) and (.)
yields
𝑅𝑇𝑆
𝐶𝑆 = ( )2 (.)
𝜌𝑆 𝐷𝑆2 (1 − 𝑡) 1 − 𝑤𝑇𝑆 𝑣2𝑆
Finally, we replace the resistance with the dimensionless total resistance coefficient
𝐶𝑇 𝑆 .
𝑆𝑆 𝐶𝑇 𝑆
𝐶𝑆 = ( ) (.)
2 𝐷𝑆 (1 − 𝑡) 1 − 𝑤𝑇𝑆 2
2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

532 44 ITTC 1978 Performance Prediction Method

0.8 0.8
KTSO
10KQSO
0.7 OTS 0.7
OS
[−]

KTSrequired = CS J 2
0.6 0.6
thrust and torque coefficients KTSO , 10KQSO

[−]
0.5 0.5

OS
open water efficiency
0.4 0.4
thru
st c
apab
ility
of p
0.3 rop
elle
10KQTS 0.3
r
CS J 2
0.2 0.2

self propulsion point KTS = KTSO

0.1 f ship 0.1
ent o
s t re quirem
thru peed
iven s
for a g
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
advance coefficient J [−] JTS

j Figure 44.1 Matching propeller thrust 𝐾𝑇𝑆 𝑂 with the thrust requirement of the ship assuming j
thrust identity

The shortened thrust loading coefficient will vary with ship speed because the resistance
changes.
Self propulsion A match of thrust requirements and thrust capabilities is found by plotting the parabola
point of ship 𝐶𝑆 𝐽 2 into the propeller open water diagram (see Figure .). To that end, we multiply
𝐶𝑆 with increasing values of 𝐽 squared and mark the points in the open water diagram.
The parabola 𝐾𝑇𝑆 = 𝐶𝑆 𝐽 2 represents the thrust requirement for a given vessel speed.
The intersection of the parabola with the propeller 𝐾𝑇𝑆 𝑂 -curve marks the self propul-
sion point. Note that we again employed the concept of thrust identity assuming that
the propeller produces the same thrust for the same rate of revolution and speed of
advance in both, open water and behind condition.
Advance ratio The advance coefficient 𝐽𝑇𝑆 is found via a vertical line drawn into Figure .. Where
under thrust the vertical line intersects the open water torque curve 10𝐾𝑄𝑆𝑂 and the open water
identity
efficiency curve 𝜂𝑂𝑆 we read off the open water torque 10𝐾𝑄𝑇𝑆 and the propeller
efficiency 𝜂𝑂𝑇 𝑆 for the given speed 𝑣𝑆 . Section . explains how to find the intersection
numerically if the open water diagram is given as a set of discrete data points.
𝐽𝑇𝑆 , 𝐾𝑄𝑇𝑆 , and 𝜂𝑂𝑇 𝑆 have the additional subscript 𝑇 because they have been obtained
under the assumption of thrust identity. Computing 𝐶𝑆 and matching thrust require-
ments of the ship with the propeller thrust have to be repeated for each speed.
Rate of From the advance coefficient 𝐽𝑇𝑆 , we retrieve the necessary rate of revolution 𝑛𝑆 .
revolution

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44.2 Full Scale Power Prediction 533

𝑣𝐴𝑆
𝑛𝑆 = (.)
𝐽𝑇𝑆 𝐷𝑆

With the relative rotative efficiency the required torque 𝐾𝑄𝑆 for the propeller in the Torque
behind condition and the behind efficiency 𝜂𝐵 are found. coefficient and
behind efficiency
𝐾𝑄𝑇𝑆
𝐾𝑄𝑆 = (.)
𝜂𝑅
𝜂𝐵 = 𝜂𝑂𝑇 𝑆 𝜂𝑅 (.)

Finally, required torque 𝑄𝑆 and delivered power 𝑃𝐷𝑆 may be determined. Delivered power
and
𝑄𝑆 = 𝜌 𝑛2𝑆 𝐷𝑆5 𝐾𝑄𝑆 torque quasi-propulsive
efficiency
𝑃𝐷𝑆 = 2𝜋 𝑛𝑆 𝑄𝑆 = 2𝜋 𝜌 𝑛3𝑆 𝐷𝑆5 𝐾𝑄𝑆 delivered power (.)
The quasi propulsive efficiency 𝜂𝐷𝑆 of the ship becomes:
𝑃𝐸𝑆
𝜂𝐷𝑆 =
𝑃𝐷𝑆
= 𝜂𝐻𝑆 𝜂𝑂𝑇 𝑆 𝜂𝑅 quasi propulsive efficiency (.)

The predicted values for rate of revolution 𝑛𝑆 (.) and delivered power 𝑃𝐷𝑆 (. are Trial corrections
usually adjusted for the ship trial. Three different methods are described in the ITTC
recommended procedures (ITTC, a):
j • correction factors j
• alternative self propulsion point based on estimates of changes in friction Δ𝐶𝐹
and changes in wake fraction Δ𝑤𝑐
• using torque identity

The first and simplest method is correction factors. Power and rate of revolution at
trial are
𝑃𝐷𝑇 = 𝑃𝐷𝑆 𝐶𝑃 (.)
𝑛𝑇 = 𝑛𝑆 𝐶𝑁 (.)
New editions of the ITTC  performance prediction method (ITTC, a) do not
provide values for the correction factors 𝐶𝑃 and 𝐶𝑁 . It is left to the model basin to
select appropriate values based on its own experience. In older editions values are
stated as 𝐶𝑃 = 1.01 and 𝐶𝑁 = 1.02.
The brake power 𝑃𝐵 of the engine is slightly higher than the delivered power due Brake power
to losses in shaft bearings and gear boxes if present. Brake power is given with the
mechanical efficiency 𝜂𝑆 of the shafting as
𝑃𝐷
𝑃𝐵 = (.)
𝜂𝑆

Note that delivered power and brake power have been computed for trial conditions, i.e.
deep water, calm seas, and no wind or current. However, propeller and engine selection
have to take into account weather and sea state conditions for the intended service area
of the vessel. A discussion of powering margins is provided in ITTC (b).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

534 44 ITTC 1978 Performance Prediction Method

44.3 Summary
This completes our discussion of power prediction from model tests. Development
of this method started with William Froude and is still ongoing today. Every three
years researchers meet at the International Towing Tank Conference to summarize
and discuss the newest developments and to propose and update the recommended
procedures. The current procedure is the essence of decades of experience in power
prediction and produces in most cases accurate and reliable design data. However, it
definitely has its shortcomings which are briefly discussed below.
A fundamental flaw of the power prediction based on resistance, open water, and propul-
sion test is the decomposition of the hull-propeller system into individual parts. Resis-
tance test and open water test investigate hull and propeller in hypothetical conditions.
The hull is not able to move without propeller and the propeller cannot be driven without
a hull housing the prime mover. A possible solution would be to solely rely on load vari-
ation tests (see Section .) which at least would be a complete hydrodynamic system.
This still leaves the problem of partial dynamic similarity. Unless a fluid is found whose
kinematic viscosity is considerably lower than that of fluids known today, Reynolds num-
bers will be significantly lower at model scale than at full scale. Thus, data derived from
load variation tests with models must still be scaled up to the full scale vessel. Possible
solutions have been proposed in the past but so far the international community has not
been able to replace the current performance prediction procedure with a method more
soundly grounded in physics. Some examples of alternatives can be found in Grigson
j (), Holtrop (), and most notably in Schmiechen (, ). Professor M. j
Schmiechen provides updates and in depth discussions about his rational theory of
propulsion on his webpage at http://www.m-schmiechen.homepage.t-online.de.
Less fundamental, but still important, are problems within the current method.

• As mentioned before, the ITTC  model–ship correlation line is not a true flat
plate friction coefficient but rather a compromise to facilitate the scaling of model
test results to full scale vessels. This is a field of ongoing research. A proposal by
Grigson (, ) seems most promising to eventually replace the ITTC 
line with a planar friction coefficient which is based on the best available physical
model.
• Resistance and propulsion tests are evaluated assuming the model has reached a
steady state. That is however rarely the case, and small fluctuations of velocity
do occur. Measured values for resistance, thrust, torque, and rate of revolution
are averaged over time to eliminate the fluctuations. Admittedly, the changes in
velocity are very small, but they are accompanied by small accelerations. The latter
should be less than 1% of the gravitational acceleration 𝑔 in carefully conducted
tests. However, considering how small the resistance values are compared to
the large mass of ships (plus additional added mass), inertia forces (mass times
acceleration) may not be negligible in the equilibrium of forces. This is addressed
by Schmiechen (, ) but not accounted for in the present performance
prediction method.

This being primarily a textbook for future naval architects and marine engineers, the
short discussion above should not be construed as a dismissal of current model testing

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44.4 Solving the Intersection Problem 535

Table 44.1 Propeller open water characteristics as a set of discrete data points

𝑖 𝐽𝑖 𝐾𝑇𝑖 10𝐾𝑄𝑖 𝜂𝑂𝑖

[–] [–] [–] [–]
 0.0000 0.45626 0.65882 0.0000
 0.0242 0.45021 0.65116 0.0267
 0.0485 0.44385 0.64318 0.0533
 0.0727 0.43720 0.63487 0.0797
 0.0970 0.43026 0.62625 0.1060
 0.1212 0.42304 0.61730 0.1322
⋮ ⋮ ⋮ ⋮ ⋮
𝑁 −2 1.0424 0.00299 0.06021 0.0823
𝑁 −1 1.0545 −0.00361 0.05032 −0.1204

practices. Rather, shortcomings of the performance prediction procedure should be

understood as an enticement to study these problems and eventually contribute to their
solution.

j j
44.4 Solving the Intersection Problem

Various numerical methods may be used to solve for the self propulsion point where the
parabola of the ship thrust requirements 𝐶𝑆 𝐽 2 and the open water thrust coefficient of
the propeller intersect. For now, we assume that the curves of the open water diagram
of the full scale propeller are given as discrete points (Table .).
The data points may be derived from open water tests, regression formulas of a suitable
propeller series, or the results of a lifting line or panel code. The numbers in Table .
have been computed with the Wageningen B-Series regression formulas omitting the
additional Reynolds number correction (Oosterveld and van Oossanen, ).
To avoid unnecessary clutter, we omit the subscripts for open water and full scale ship.
It is important to check if the open water torque coefficient itself was stored or ten
times the torque coefficient.
The simplest computation of the self propulsion point approximates the open water Linear
curves as polylines. Data points are connected by straight line segments. The intersec- interpolation
tion point is then found by linear interpolation. This will be accurate enough if your
data points are close together. A quadratic interpolation over three data points may be
more appropriate for a limited number of data points. Here we assume that a sufficient
number of data points is available.
We perform a sequential search through the data starting with 𝑖 = 0 and we stop, once
the product of design constant and squared advance coefficient 𝐶𝑆 𝐽𝑖2 becomes larger
than 𝐾𝑇𝑖 . The intersection point is located between 𝐽𝑖−1 and 𝐽𝑖 , i.e. 𝐽𝑖−1 ≤ 𝐽𝑇𝑆 ≤ 𝐽𝑖 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

536 44 ITTC 1978 Performance Prediction Method

Figure 44.2 Setup for solving the intersection

problem with discrete open water data

We describe the 𝐾𝑇 curve between 𝐾𝑇𝑖−1 and 𝐾𝑇𝑖 as a linear function:

𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 ( )
j 𝐾𝑇 = 𝐾𝑇𝑖−1 + 𝐽 − 𝐽𝑖−1 for 𝐽 ∈ [𝐽𝑖−1 , 𝐽𝑖 ] (.) j
𝐽𝑖 − 𝐽𝑖−1

A corresponding linear function is created for the thrust requirement of the ship:

𝐶𝑆 𝐽𝑖2 − 𝐶𝑆 𝐽𝑖−1
2
( )
𝐶𝑆 𝐽 2 = 𝐶𝑆 𝐽𝑖−1
2
+ 𝐽 − 𝐽𝑖−1 for 𝐽 ∈ [𝐽𝑖−1 , 𝐽𝑖 ] (.)
𝐽𝑖 − 𝐽𝑖−1

In order to find 𝐽𝑇𝑆 , we begin by subtracting Equation (.) from Equation (.):

𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 ( )
𝐾𝑇 − 𝐶𝑆 𝐽 2 = 𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖−1
2
+ 𝐽 − 𝐽𝑖−1
𝐽𝑖 − 𝐽𝑖−1
𝐶𝑆 𝐽𝑖2 − 𝐶𝑆 𝐽𝑖−1
2
( )
− 𝐽 − 𝐽𝑖−1 for 𝐽 ∈ [𝐽𝑖−1 , 𝐽𝑖 ] (.)
𝐽𝑖 − 𝐽𝑖−1

We simplify the resulting equation and make use of the fact that for 𝐽 = 𝐽𝑇𝑆 we get the
intersection point 𝐾𝑇𝑆 = 𝐶𝑆 𝐽𝑇𝑆2 , i.e. the left-hand side of Equation (.) vanishes.

2
0 = 𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖−1 (.)
[ ]
𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 𝐶𝑆 𝐽𝑖2 − 𝐶𝑆 𝐽𝑖−1
2
( )
+ − 𝐽𝑇𝑆 − 𝐽𝑖−1
𝐽𝑖 − 𝐽𝑖−1 𝐽𝑖 − 𝐽𝑖−1

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44.5 Example 537

Multiplying with (𝐽𝑖 − 𝐽𝑖−1 ) eliminates the fractions.

[ ]( )
0 = 2
𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖−1 𝐽𝑖 − 𝐽𝑖−1 (.)
[ ]( )
+ 𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖2 + 𝐶𝑆 𝐽𝑖−1
2
𝐽𝑇𝑆 − 𝐽𝑖−1
for 𝐽𝑇𝑆 ∈ [𝐽𝑖−1 , 𝐽𝑖 ]

Finally, we solve the equation for the desired advance ratio 𝐽𝑇𝑆 at which the propeller
delivers the required thrust.
[ ]( )
2
𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖−1 𝐽𝑖 − 𝐽𝑖−1
𝐽𝑇𝑆 = 𝐽𝑖−1 − [ ] (.)
𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 − 𝐶𝑆 𝐽𝑖2 + 𝐶𝑆 𝐽𝑖−1
2

Substituting the value of 𝐽𝑇𝑆 into Equation (.) yields the thrust coefficient for the
full scale propeller.

𝐾𝑇𝑖 − 𝐾𝑇𝑖−1 ( )
𝐾𝑇𝑆 = 𝐾𝑇𝑖−1 + 𝐽𝑇𝑆 − 𝐽𝑖−1 (.)
𝐽𝑖 − 𝐽𝑖−1

Substituting 𝐽𝑇𝑆 into the corresponding intervals of the polynomials for torque coeffi-
j cient and open water efficiency allows us to compute torque and efficiency for the full j
scale propeller in open water condition.

𝐾𝑄𝑖 − 𝐾𝑄𝑖−1 ( )
𝐾𝑄𝑇𝑆 = 𝐾𝑄𝑖−1 + 𝐽𝑇𝑆 − 𝐽𝑖−1 (.)
𝐽𝑖 − 𝐽𝑖−1
𝜂𝑂𝑖 − 𝜂𝑂𝑖−1 ( )
𝜂𝑂𝑇𝑆 = 𝜂𝑂𝑖−1 + 𝐽𝑇𝑆 − 𝐽𝑖−1 (.)
𝐽𝑖 − 𝐽𝑖−1

Again, if only a few data points define the open water characteristics, a higher order
interpolation method is warranted.

44.5 Example
Let us study the process using an example. As input we need the data listed in Table .
which may be derived from model test results or from a resistance and propulsion
estimate like we discussed.
With the given input data we first compute the required thrust:

𝑅𝑇 1450250 N
𝑇required = = = 1795.71 kN
(1 − 𝑡) 1 − 0.1924

The speed of advance for 𝑣 = 11.472 m/s follows from equation (.).

𝑣𝐴 = (1 − 𝑤) 𝑣 = (1 − 0.2309) 11.472 m/s = 8.823 m/s

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

538 44 ITTC 1978 Performance Prediction Method

Table 44.2 Example input data for the calculation of the self propulsion point for a single ship
speed

Propeller open water data

Resistance and propulsion data 𝑖 𝐽𝑖 𝐾𝑇𝑖 10𝐾𝑄𝑖
ship speed 𝑣 11.472 m/s ⋯
total resistance 𝑅𝑇 1450.250 kN 28 0.6182 0.22329 0.36665
thrust deduction 29 0.6303 0.21744 0.35901
fraction 𝑡 0.1924 30 0.6424 0.21155 0.35131
wake fraction 𝑤 0.2309 31 0.6545 0.20563 0.34354
relative rotative 32 0.6667 0.19968 0.33570
efficiency 𝜂𝑅 0.9939
33 0.6788 0.19369 0.32779
propeller 34 0.6909 0.18767 0.31981
diameter 𝐷 7.200 m
35 0.7030 0.18162 0.31177
density of
salt water 𝜌 1026.021 kg/m3 36 0.7152 0.17554 0.30366
⋯

Table 44.3 Data for required thrust parabola at Required and provided thrust coefficient
𝑣 = 11.472 m/s and 𝐶𝑆 = 0.43372 and propeller open
water thrust coefficient 𝑖 𝐽𝑖 𝐽𝑖2 𝐶𝑆 𝐽𝑖2 𝐾𝑇𝑖
⋯
28 0.6182 0.38217 0.16577 0.22329
j 29 0.6303 0.39728 0.17232 0.21744 j
30 0.6424 0.41268 0.17900 0.21155
31 0.6545 0.42837 0.18579 0.20563
32 0.6667 0.44449 0.19280 0.19968
33 0.6788 0.46077 0.19986 0.19369
34 0.6909 0.47734 0.20705 0.18767
35 0.7030 0.49421 0.21437 0.18162
⋯

With these results we compute the shortened thrust loading coefficient 𝐶𝑆 (.).

𝑇required 1795710 N
𝐶𝑆 = = = 0.43372
𝜌𝐷2 𝑣𝐴2 1026.021 kg/m3 7.22 m2 8.8232 (m/s)2

Next, we compute values of the required thrust parabola 𝐶𝑆 𝐽 2 for the ship using the 𝐽𝑖
values from the open water data in Table .. Table . summarizes the parabola data
and repeats the open water thrust coefficient 𝐾𝑇𝑖 . Comparing 𝐶𝑆 𝐽𝑖2 and 𝐾𝑇𝑖 we find that
for 𝑖 = 33 the required thrust coefficient exceeds the provided thrust coefficient for the
first time, i.e. 𝐶𝑆 𝐽33
2 > 𝐾 . For 𝑖 − 1 = 32 we still have 𝐶 𝐽 2 < 𝐾 . Consequently,
𝑇33 𝑆 32 𝑇32
the self propulsion point 𝐽𝑇 is located between 𝐽32 = 0.6667 and 𝐽33 = 0.6788. With
the help of the interpolation formula (.) and data for 𝑖 = 33, we find

𝐽𝑇 = 0.6731

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

44.5 Example 539

Equipped with the advance ratio 𝐽𝑇 , Equations (.), (.), and (.) deliver
thrust coefficient, torque coefficient, and open water efficiency for the self propulsion
point.

𝐽𝑇 = 0.6731
𝐾𝑇 = 0.1965
𝐾𝑄𝑇 = 0.03315
𝜂𝑂𝑇 = 0.6350

With the data of the propeller at the self propulsion point it is easy to complete the
power prediction for the given speed. We use Equations (.) through (.) and
compute
𝑣𝐴
rate of revolution 𝑛 = = 1.8205 s−1 = 109.231 rpm
𝐽𝑇 𝐷
𝐾𝑄𝑇
torque in behind condition 𝐾𝑄 = = 0.033353
𝜂𝑅
propeller behind efficiency 𝜂𝐵 = 𝜂𝑂𝑇 𝜂𝑅 = 0.6311
propeller torque 𝑄 = 𝜌 𝑛2 𝐷5 𝐾𝑄 = 2194.54 kNm
delivered power 𝑃𝐷 = 2𝜋 𝑛 𝑄 = 25102.56 kW
quasi-propulsive efficiency 𝜂𝐷 = 𝜂𝐻 𝜂𝑂𝑇 𝜂𝑅 = 0.6627
j j

The process described in this example will be repeated for each considered ship speed.
Based on the predicted delivered power, the required engine brake power may be
derived depending on shafting and gearing efficiency as well as suitable engine and sea
margins.

References
Grigson, C. (). Screws working in behind and prediction of the performance of
full ships. Journal of Ship Research, ():–.
Grigson, C. (). An accurate smooth friction line for use in performance prediction.
In Transactions of The Royal Institution of Naval Architects (RINA), volume ,
pages –. The Royal Institution of Naval Architects.
Grigson, C. (). A planar friction algorithm and its use in analysing hull resistance.
In Transactions of The Royal Institution of Naval Architects (RINA), volume ,
pages –. The Royal Institution of Naval Architects.
Holtrop, J. (). Extrapolation of propulsion tests for ships with appendages and
complex propulsors. Marine Technology, ():–.
ITTC (a).  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
ITTC (b). Predicting powering margins. International Towing Tank Conference,
Recommended Procedures and Guidelines .---.. Revision .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

540 44 ITTC 1978 Performance Prediction Method

Oosterveld, M. and van Oossanen, P. (). Further computer-analyzed data of the

Wageningen B-screw series. International Shipbuilding Progress, :–.
Schmiechen, M. (). The method of quasisteady propulsion and its trial on board
the Meteor. Technical Report Report No. /, Versuchsanstalt für Wasserbau
und Schiffbau, Berlin, Germany.
Schmiechen, M. ().  Years rational theory of propulsion – Recent results and
perspectives. In International Symposium on Marine Propulsion smp’, Trondheim,
Norway.

Self Study Problems

. Compute the self propulsion point data for the data in Table . applying the
following ship speed and total resistance combinations:

𝑣 = 20kn 𝑅𝑇 = 1072.76 kn
𝑣 = 24kn 𝑅𝑇 = 1818.65 kn

. What are the advantages of a load variation test compared with a traditional
propulsion test (continental method)?

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

541

45
Cavitation

The substance known as water (H2 O) exists in three phases: as a liquid (water), solid
(ice), or gas (vapor). The sudden transition from its liquid to its gaseous phase is called
cavitation. Cavitation may occur in all flows which have high local flow velocities and
associated low pressures. In marine systems, propellers and lifting foils especially are
subject to the cavitation phenomena. Cavitation may also occur in pumps and piping
systems. This chapter discusses the cavitation phenomenon, where it is likely to occur,
and which negative effects it can have. Design measures to prevent cavitation are
discussed in the following chapter.

Learning Objectives
j j

At the end of this chapter students will be able to:

• explain and identify cavitation

• quantify the likelihood of cavitation with cavitation numbers
• know about detrimental effects of cavitation

45.1 Cavitation Phenomenon

Water (H2 O) exists in three different phases: Phases of Water

(i) solid as ice,

(ii) liquid as water, and
(iii) gaseous as vapor.

Which phase is present depends on the combination of temperature and pressure.

The relationship is visualized in a phase diagram like the one shown in Figure ..
Solid ice exists for temperatures below freezing (0 ◦ C) and (not shown here) at higher
temperatures under very high pressure. For temperatures above 0 ◦ C ice becomes
liquid. If the combination of pressure and temperature falls below the vapor pressure,
ice or water turn into vapor. Vapor pressure is also called saturation pressure.

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

542 45 Cavitation

Figure 45.1 Simplified phase diagram of fresh water

j j
Boiling or We are interested in the transition from the liquid to the gaseous phase. According to
cavitating? the phase diagram Figure ., transition from liquid to gaseous phase may happen in
two different ways:

• If temperature is increased at constant pressure, water will change from liquid

to vapor. On the phase diagram we move horizontally from left to right. At the
standard atmospheric pressure of 101 325 Pa, transition from liquid to vapor
happens at a temperature of 100 ◦ C. This is known as the boiling point. You
observe bubbles popping up from the bottom of a saucepan with boiling water.
The bubbles burst at the free surface and release the vapor into the air. As
mentioned, temperature is considered constant for flow patterns around ships
and propellers. Thus this first method of transition from liquid to gas does not
apply.
• For constant temperature, the transition into the gaseous phase may be triggered
by a drop in pressure. We move vertically in the phase diagram. At the level of the
vapor pressure 𝑝𝑣 , water changes into its gaseous phase. This is called cavitation.
Vapor pressure 𝑝𝑣 increases with temperature, i.e. cavitation is more likely in
warm than in cold water.

Cavitation When the pressure in a flow approaches vapor pressure, the fluid vaporizes in an instant.
Cavitation is the formation of vapor filled cavities in the liquid. Typically, dissolved air
or tiny sand grains serve as nuclei for the bubble. The cavities move along with the flow
and collapse when they reach areas with higher ambient pressure. This phenomenon
occurs in any hydrodynamic or hydraulic system where low pressure is generated by
moving fluids or bodies.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

45.2 Cavitation Inception 543

Cavitation is a highly unsteady process and depends on numerous fluid properties

and flow features, which makes accurate prediction of cavitation and its impact on
performance difficult. This difficulty applies to numerical as well as experimental
methods.

45.2 Cavitation Inception

One objective in the design of hydrodynamic systems is the avoidance or at least Cavitation
the delay of cavitation. Cavitation numbers 𝜎 are used as a measure to quantify the number
likelihood of cavitation to occur.
𝑝 − 𝑝𝑣
𝜎 = 0 (.)
1 2
𝜌𝑣
2
The numerator 𝑝0 − 𝑝𝑣 measures the difference between the static pressure in the flow
𝑝0 and the vapor pressure 𝑝𝑣 . The denominator is the dynamic pressure 𝑞.
1 2
𝑞 = 𝜌𝑣 (.)
2

Cavitation numbers express the resistance of a flow against cavitation. The larger the
cavitation number the less likely is cavitation. High static pressure 𝑝0 lowers the danger
j of cavitation (higher cavitation number 𝜎). This can be achieved by deeply submerging j
a foil or propeller. High flow velocities increase dynamic pressure which increases the
likelihood of cavitation (smaller cavitation number 𝜎).
Consider the two-dimensional flow of an inviscid fluid around a foil section shown Pressure drop
in Figure .. Two points, A and B, are marked along the same streamline. At point
A the fluid has a velocity 𝑣0 and a total static pressure 𝑝0 . According to Bernoulli’s
equation (.) for steady flow, the sum of total static pressure and dynamic pressure
is constant.
1 1
𝑝0 + 𝜌𝑣20 = 𝑝1 + 𝜌𝑣21 = const. (.)
2 2
At point B on the back side of the foil, velocity 𝑣1 will be larger than 𝑣0 , and, as a
consequence, the pressure 𝑝1 smaller than 𝑝0 .
The drop in pressure Δ𝑝 between points A and B is given by
1 ( 2 )
Δ𝑝 = 𝑝1 − 𝑝0 = 𝜌 𝑣0 − 𝑣21 (.)
2
In this case, pressure decreases from A and B, and the drop in pressure Δ𝑝 is negative.
Rewriting the left part of the Equation (.) above yields:
𝑝1 = 𝑝𝑜 + Δ𝑝 (.)

Increasing the angle of attack 𝛼 will lower the pressure 𝑝1 further and |Δ𝑝(𝛼2 )| >
|Δ𝑝(𝛼1 )| when 𝛼2 > 𝛼1 . It may happen that Δ𝑝 becomes equal to −𝑝0 , and the pressure
at point B reaches zero.
𝑝1 = 𝑝0 + Δ𝑝 = 𝑝0 − 𝑝0 = 0 for Δ𝑝 = −𝑝0 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

544 45 Cavitation

j j

Figure 45.2 Flow around a cavitating foil section and the associated pressure distribution for
back (upper) and face (lower) side

This is a theoretical limit. The pressure at point B cannot fall below zero as water
does not support tension. The flow will break down with the formation of cavities. In
practice this will already happen when the vapor pressure or saturation pressure 𝑝𝑣 > 0
is reached:

𝑝1 = 𝑝𝑣 = 𝑝0 + Δ𝑝 for Δ𝑝 = −𝑝0 + 𝑝𝑣 (.)

or

Δ𝑝 = 𝑝𝑣 − 𝑝0 = −(𝑝0 − 𝑝𝑣 ) cavitation inception (.)

The vapor pressure 𝑝𝑣 at which the fluid changes its phase is temperature dependent.
Values have been provided earlier in Chapter .
Division by the dynamic pressure 𝑞 (.) renders Equation (.) dimensionless:

Δ𝑝 −(𝑝0 − 𝑝𝑣 )
= cavitation inception (.)
𝑞 𝑞

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

45.2 Cavitation Inception 545

The absolute value of the right-hand side is equal to the cavitation number 𝜎. 𝑝0 is the
total static pressure, and 𝑞 = 21 𝜌𝑣20 represents the inflow.
The cavitation number 𝜎 characterizes the flow’s resistance to cavitation. If the negative Cavitation
change of pressure becomes larger than the cavitation number, cavitation will occur. inception

Δ𝑝
− > 𝜎 cavitation inception (.)
𝑞
This cavitation inception is indicated in Figure . for an example value of 𝜎 = 1.2.
Cavitation bubbles will appear and grow in the region with
−𝐶𝑝 > 𝜎 = 1.2
The cavities will stretch downstream of this region but will begin to collapse due to the
increased pressure outside of the cavity.
The ability of a flow to withstand cavitation depends on the combination of static Free stream
pressure 𝑝0 and flow velocity. High static pressure 𝑝0 will lower the danger of cavitation cavitation
number
inception. High velocity will increase the risk of cavitation. Both quantities may vary
substantially within a given flow regime. A suitable combination of static pressure and
velocity must be selected to define the cavitation number which is characteristic for
the flow. Common choices are the free stream cavitation number and the propeller
cavitation number. The free stream cavitation number 𝜎0 is formed with the uniform
flow speed 𝑣0 of the far field and the total static pressure at the depth ℎ of the considered
object.
j 𝑝 + 𝜌 𝑔 ℎ − 𝑝𝑣 j
𝜎0 = 𝐴 (.)
1 2
𝜌 𝑣0
2
Note that the atmospheric pressure 𝑝𝐴 in Equation (.) is usually the larger part of
the total static pressure.
Velocity at propeller blades are functions of speed of advance 𝑣𝐴 and rate of revolution Propeller
𝑛. The latter creates a transverse flow speed which linearly increases with growing cavitation
number
distance from the shaft center. Highest flow speeds are encountered at the blade tips.
Since most of the thrust is created over a region near the radius 𝑟∕𝑅 = 0.7, the 70%
radius is often used to calculate a representative flow speed 𝑣1 .
√
𝑣1 = 𝑣𝐴2 + (𝜋 𝑛 0.7𝐷)2 (.)

Submergence also varies considerably over the diameter of the propeller since its axis is
more or less horizontal. Submergence will be smallest for a blade in 12 o’clock position,
and cavitation always appears first at the blade in upright position. In most cases
submergence of the propeller shaft ℎ0 forms the basis to compute the static pressure
but other definitions have been used in the relevant literature. The propeller cavitation
number 𝜎𝑏 is defined as
𝑝𝐴 + 𝜌𝑔ℎ0 − 𝑝𝑣
𝜎𝑏 = ( ) (.)
1
2
𝜌𝑣𝐴2 + (𝜋 𝑛 0.7𝐷)2

We will come back to this in the next chapter to select propeller characteristics which
minimize or prevent cavitation.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

546 45 Cavitation

j j

Figure 45.3 Locations and common types of cavitation

45.3 Locations and Types of Cavitation

Cavitation is identified by visual appearance and location. Figure . depicts common
types of cavitation for a four bladed propeller.
Tip vortex and The highest flow velocities occur near the tip of the propeller blades and especially in the
hub vortex tip vortex. At first, only a portion of the tip vortex of the blade in upright position will
cavitation
cavitate because pressure has dropped to vapor pressure level. The cavitation will occur
on each blade passing through the twelve o’clock position and vanish again until the
next blade passes through. If flow speeds increase further, tip vortex cavitation will start
at all blade tips and persist for sometimes several rotations until decreasing velocities
cause the cavitation bubbles to collapse further downstream (see Figure .). Vortices
are also shed from the root of each blade. The root vortices combine into a single hub
vortex. If the pressure in the hub vortex drops to vapor pressure, a continuous string
of cavitation bubbles will appear behind the hub. Hub vortex cavitation is not visible
in the photo shown in Figure .. The propeller is mounted in open water condition
with the propeller dynamometer downstream of the propeller which disrupts the hub
vortex.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

45.3 Locations and Types of Cavitation 547

j Figure 45.4 Open water test of five-bladed model propeller in a cavitation tunnel; tip vortex j
and bubble cavitation

As the cavitation pattern develops, pressure will also drop on the back (suction) side of Bubble and
the blades and cavitation bubbles appear. At first, individual bubbles can be identified as cloud cavitation
they pop up and collapse pretty soon downstream. Figure . shows bubble cavitation
on the back of the blade. Near the tip of the blades, groups of bubbles form clouds that
are washed downstream. These intermittent congregations of bubbles are called cloud
cavitation.
Figure . shows sheet cavitation on the blades in twelve and three o’clock position. Sheet cavitation
Sheet cavitation develops when the pressure drops below vapor pressure over a whole
stretch of the leading edge of a blade. It starts near the tip and grows as flow speeds
increase and pressure decreases toward the hub and across the blade chord. Due to the
smaller static pressure, sheet cavitation extends over a larger area on the blade in 12
o’clock position compared to other blades. Shown is an approximately 10% coverage of
the blade surface. This is arguably the worst condition still acceptable for continuous
operations. If cavitation is unavoidable, naval architects aim to restrict sheet cavitation
patterns to less than 5% as shown on the blade in three o’clock position.
Vessels with relatively large propellers, which leave only a minimum of clearance be- Propeller–hull
tween blade tip and hull, sometimes experience propeller–hull cavitation. A vortex cavitation
forms briefly between blade tip and adjacent hull surfaces. If the pressure is low enough,
the vortex cavitates like the tip vortices. This highly unsteady process leads to increased
noise and vibrations.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

548 45 Cavitation

j Figure 45.5 Open water test of five-bladed model propeller in a cavitation tunnel; propeller
j
blades completely covered in sheet cavitation

Face cavitation Cavitation is mostly seen on the suction side of the blade. However, face cavitation may
occur, if the angle of attack of the blade section becomes negative. This often happens
when the propeller rate of revolution is winding down but the ship is still going forward
close to its original speed. The same effect may occur on controllable pitch propellers
when they are operated in off-design conditions.
Fully cavitating Cavitation is a problem for all propellers with high loading (high thrust and small
propellers diameter) or high velocities (high ship speed and /or high rate of revolution). As a rule
of thumb: if your vessel goes faster than  knots you will not be able to avoid cavitation.
Figure . shows a propeller which is fully cavitating. All blades are completely engulfed
in sheet cavitation. Serious degradation of performance is to be expected. Propellers
designed for this operating condition have blade sections with a sharp leading edge and
a blunt trailing edge. This arrangement creates a stable sheet cavitity over the whole
chord which collapses downstream of the foil and thus avoids blade erosion.

45.4 Detrimental Effects of Cavitation

Detrimental Cavitation has several negative effects. Lifting foils, pumps, and propellers will experi-
effects ence:

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

45.4 Detrimental Effects of Cavitation 549

Figure 45.6 Loss of thrust and efficiency due to cavitation

j • loss of lift force, thrust, and efficiency j

• vibrations and noise

• destruction of blades due to erosion

In the open water diagram of the propeller effects of cavitation become first visible at Loss of thrust
lower advance ratios where the propeller loading is high. A distinct loss of thrust is and efficiency
observed. Although torque is decreasing as well, thrust drops faster and results in a loss
of propeller efficiency (Figure .). Initially this does not affect the peak efficiency of
the propeller. However, with decreasing cavitation number, effects of cavitation spread
over larger portions of the propeller blade and affect the working conditions at higher
advance ratios 𝐽 . Eventually loss of thrust in the whole operating range will cause a
significant drop in peak propeller efficiency.
Cavitation can actually be heard before it is visible. Imagine the sound of pebbles hitting Noise
a hard surface. As cavitation intensifies, a persistent crackling noise can be heard. Noise
itself is of concern for naval vessels. Skilled sonar operators may use the cavitation
noise to identify classes of vessels. Today, naval architects should also consider the
environmental impact of noise especially its effects on marine mammals.
The highly unsteady nature of cavitation with constant production and collapses of Vibrations
cavities creates high pressure shock waves which are capable of exciting propeller blades,
rudder, and nearby hull plating to measurable vibrations. This adds to the signature of
a vessel, creates discomfort for crew and passengers, and may lead to fatigue damage in
the long term.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

550 45 Cavitation

Figure 45.7 Life of cavitation bubble

Erosion The mechanisms leading to erosion of fairly strong materials are still subject of ongoing
research. Damage occurs when cavitation bubbles collapse, not when they are created.
Two mechanisms have been identified:

• micro-jets and
• high pressure shock waves
j j
Figure . pictures the short life of a cavitation bubble. It starts with a nucleus moving
into an area of pressure equal to the vapor pressure. Air dissolved in the fluid or small
solid particles may serve as nuclei. Water surrounding the nucleus evaporates and
a cavitation bubble is formed. The bubble is filled with vapor at saturation pressure
and grows as long as the pressure stays low. Once the bubble moves into an area with
increasing pressure it collapses. Slow motion pictures revealed that the bubbles fold
inward which creates a high pressure micro water jet. The water jet is strong enough to
break material out of propeller blades and foils. Very hard and polished materials are
less susceptible to damage but persistent cavitation will eventually destroy all materials
typically used in shipbuilding. The collapsing bubbles also emit high pressure shock
waves with short term pressure peaks of several thousand atmospheres. The pressure
waves produce the crackling noise and may cause elastic and plastic deformations on
propeller blades and hull plating.
Cavitation is still a field of ongoing research. Numerous conference proceedings and
a couple of textbooks are available for further study. The book by Franc and Michel
() may serve as a starting point for a more scientific discussion of cavitation and the
dynamics of cavitation bubbles. In the next chapter, we discuss simple design measures
to prevent cavitation under normal operating conditions.

References
Franc, J.-P. and Michel, J.-M. (). Fundamentals of cavitation. Fluid Mechanics and
Its Applications (Book ). Kluwer Academic Publishers, Dordrecht, The Netherlands.
ISBN -.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

45.4 Detrimental Effects of Cavitation 551

Self Study Problems

. Explain in your own words the difference between boiling and cavitation.

. Discuss the effect of submergence and flow speed on the ability of a flow to
withstand cavitation.
. Which areas of a hydrofoil are likely to cavitate?
. Summarize the negative effects of cavitation.
. A propeller is operating under the following conditions.

Propeller data
propeller diameter 𝐷 = . m
rate of revolution 𝑛 = . rpm
speed of advance 𝑣𝐴 = . m/s
shaft submergence ℎ0 = . m
standard air pressure 𝑝𝐴 = . hPa
gravitational acceleration 𝑔 = . m/s2
salt water temperature 𝑡𝑊 = . ◦ C

Compute the propeller cavitation number 𝜎𝑏 according to Equation (.).

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

552

46
Cavitation Prevention

Cavitation has a number of negative effects ranging from noise over performance loss
to destruction of the foil or propeller. It is the naval architect’s task to prevent or at
least limit the negative impacts of cavitation. This chapter introduces the basic design
measures necessary to reduce the risk of cavitation. Specifically, the determination of a
minimum required area ratio for the propeller is explained using Keller’s formula and
Burrill’s chart. It also discusses the alternatives of fully cavitating and surface piercing
propellers in cases where cavitation cannot be avoided.

Learning Objectives
j j
At the end of this chapter students will be able to:

• understand which design measures reduce the risk of cavitation

• compute minimum required expanded area ratios for propellers

46.1 Design Measures

High static Cavitation occurs when the pressure in a fluid drops below vapor pressure. In ship
pressure hydrodynamics this predominantly happens on lifting foils like a propeller. High static
pressure and low flow speeds reduce the risk of cavitation as the cavitation number 𝜎
introduced in the previous chapter makes clear.
𝑝0 − 𝑝𝑣
𝜎0 = (.)
1 2
𝜌𝑣
2 0
High static pressure 𝑝0 may only be achieved by deep submergence of the foil or propeller.
This finds its limits in draft restrictions and geometry of the stern in general.
Low speed Lowering the flow speed 𝑣0 on a foil is possible by lowering the ship speed, reducing
the propeller rate of revolution, or selecting foil sections which create a minimum in
additional flow speed. The latter requires thin foils in combination with the lowest
camber possible to generate the desired lift. We aim to operate foils at their ideal angle
of attack, which is often difficult because in early design phases the actual local flow

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

46.2 Keller’s Formula 553

direction may only be estimated. Paint tests and CFD simulations may help to finalize
alignment of struts, bilge keels, and control surfaces.
Local flow speed for propellers is a function of speed of advance 𝑣𝐴 , rate of revolution
𝑛, and distance 𝑟 from the shaft center. Speed of advance follows from ship speed
and wake fraction. These are typically set when the propeller selection process starts.
Consequently, one might lower the rate of revolution 𝑛 in order to reduce the risk of
cavitation. As a rule of thumb, lowering the rate of revolution leads to propellers with
higher efficiency. However, smaller rate of revolution needs a bigger propeller and/or
increased pitch to generate the same thrust which may counter the effects of lowering
𝑛 on the risk of cavitation.
For a given diameter and rate of revolution, a reduction in the risk of cavitation is Required
possible by increasing the propeller blade area. The increased blade area allows the expanded area
ratio
use of foil sections with lower lift coefficients which leads to smaller pressure drops on
the suction side of the blade. The necessary blade area to avoid excessive cavitation is
specified in the form of a minimum required expanded area ratio (𝐴𝐸 ∕𝐴0 )req . Care is
required not to raise the expanded area ratio too much as this will increase the viscous
losses of the propeller and lower its efficiency. In the following sections, two basic
methods are introduced to compute the required expanded area ratio of the propeller
in early design stages.

j 46.2 Keller’s Formula j

A very basic formula for the required expanded area ratio is attributed to auf ’m Keller
(). J. auf ’m Keller worked for the model basin in Wageningen, The Netherlands.
The original publication is written in Dutch but the formula has been reproduced in
many reference books like Carlton ().

( ) ( )
𝐴𝐸 1.3 + 0.3𝑍 𝑇
= ( ) +𝐾 (.)
𝐴0 req 𝑝0 − 𝑝𝑣 𝐷 2

with constant 𝐾 = 0.2 for single screw vessels

𝐾 = 0.1 for slow twin screw vessels
𝐾 = 0.0 for fast twin screw vessels

As the input to Keller’s formula serve the number of blades 𝑍, thrust 𝑇 in service
condition for design speed, an estimate for the propeller diameter 𝐷, the static pressure
head at the propeller shaft 𝑝0 , and the vapor pressure 𝑝𝑣 . Notable is the absence of any
reference to flow velocity. Therefore, we can apply Keller’s formula without knowing
the rate of revolution or the speed of advance. However, this also limits the accuracy of
the minimum expanded area ratio estimate. Keller’s formula is suitable to set a starting
point, but subsequent design iterations should be checked against Burrill’s cavitation
chart or more sophisticated assessments.
Books on ship design feature regression formulas for typical propeller diameters based Initial diameter
on ship type and principal hull form parameters (Watson, ; Papanikolaou, ). estimate

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

554 46 Cavitation Prevention

Table 46.1 Regression equations for the limiting lines in the Burrill chart (Figure 46.1)

limiting criteria regression equation

based on model tests (Burrill and Emerson, )

% back cavitation 𝜏𝑐 = 0.910 𝜎𝑏0.335 − 0.346
% back cavitation 𝜏𝑐 = 1.147 𝜎𝑏0.195 − 0.672
naval or fast vessels: % back cavitation 𝜏𝑐 = 1.267 𝜎𝑏0.126 − 0.912
merchant vessels: % back cavitation 𝜏𝑐 = 0.715 𝜎𝑏0.184 − 0.437
.% back cavitation 𝜏𝑐 = 0.611 𝜎𝑏0.189 − 0.372
based on experience (Burrill, )
suggested lower limit for tugs, trawlers, etc. 𝜏𝑐 = 0.527 𝜎𝑏0.155 − 0.324
suggested upper limit for merchant vessels 𝜏𝑐 = 0.304 𝜎𝑏0.497 − 0.034
suggested upper limit for naval vessels 𝜏𝑐 = 0.464 𝜎𝑏0.479 − 0.089

The following formulas are from Kristensen and Lützen ():

𝐷 = 0.623 𝑇max − 0.16 m for container ships

j j
𝐷 = 0.395 𝑇max + 1.3 m for bulk carriers and tankers (.)
𝐷 = 0.713 𝑇max − 0.08 m for twin screw Ro-Ro ships

𝑇max is the maximum draft in meters. If it is not known, simply use the design draft.

46.3 Burrill’s Cavitation Chart

Burrill Chart Lennard Constantine Burrill (* – †), a British professor of naval architecture
at Durham University, UK, conducted cavitation experiments and recorded full scale
cavitation observations over many years. The data were converted into a design chart
which is known as the Burrill cavitation chart (Burrill, ; Burrill and Emerson, ).
Figure . shows a chart similar to the original. However, the cavitation limit curves
have been redrawn based on regression formulas. Table . summarizes the equations.
An effort has been made to avoid overpredicting the values derived from the original
chart. Regression formulas better fit the modern computerized work flow. Nevertheless,
design charts provide an easy way to check computational results.
Local cavitation The horizontal axis of the chart is the propeller cavitation number 𝜎𝑏 .
number 𝜎𝑏
𝑝0 − 𝑝𝑣
𝜎𝑏 = (.)
𝑞0.7
It is a measure for the risk of cavitation in the flow. Low 𝜎𝑏 values indicate high cavitation
risks. The propeller cavitation number depends on static pressure head at the propeller

j
Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

46.3 Burrill’s Cavitation Chart 555

Trim Size: mm × mm Single Column Tight

j
Birk — “fshy” — // — : — page  — #
Figure 46.1 Limits for the propeller loading coefficient as a function of cavitation number and acceptable level of cavitation. After Burrill and Emerson (1963),
however, the curves represent the regression equations from Table 46.1

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

556 46 Cavitation Prevention

Figure 46.2 Usage of the Burrill chart

j shaft 𝑝0 , vapor pressure 𝑝𝑣 , and dynamic pressure 𝑞0.7 at blade radius 𝑟 = 0.7𝑅. j

1 2
𝑞0.7 = 𝜌𝑣 (.)
2 1

The reference velocity 𝑣1 is a combination of axial speed of advance and the transverse
velocity at radius 𝑟 = 0.7 𝑅 which results from the rotation of the propeller.
√
𝑣1 = 𝑣𝐴2 + (𝜋 𝑛 0.7𝐷)2 (.)

𝑛 is the propeller rate of revolution.

Thrust loading The vertical axis of the graph measures a thrust loading coefficient 𝜏𝑐 which is based on
coefficient 𝜏𝑐 the projected area of a propeller 𝐴𝑃 .

𝑇prop
𝜏𝑐 = (.)
1 2
𝜌𝑣 𝐴
2 1 𝑃

𝑇prop is the propeller thrust at speed of advance 𝑣𝐴 . The thrust loading coefficient
represents the working conditions of the propeller.
The Burrill chart shows limiting curves as functions 𝜏𝑐 = 𝜏𝑐 (𝜎𝑏 ). The percentages
indicate how much area of a propeller blade back side is covered in cavitation. A
propeller can be expected to be cavitation free or to have cavitation up to the stated
extent if the propeller operates at a (𝜎𝑏 , 𝜏𝑐 ) combination which falls on or below the
limiting curve.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

46.4 Other Design Measures 557

Figure . explains the use of Burrill’s cavitation chart based on the frequently used
%-cavitation limit. First, the local cavitation number 𝜎𝑏 is computed.

𝑝𝐴 + 𝜌𝑔ℎ0 − 𝑝𝑣
𝜎𝑏 = ( ) (.)
1
2
𝜌𝑣𝐴2 + (𝜋 𝑛 0.7𝐷)2

An example value of 𝜎𝑏 = 0.49 is marked in Figure ..

For the value of 𝜎𝑏 , the thrust loading coefficient 𝜏𝑐 is read off the limiting curve or
computed from its respective regression formula. The 5% back cavitation limit is defined
as
𝜏𝑐 = 0.715 𝜎𝑏0.184 − 0.437 (.)
For 𝜎𝑏 = 0.49 we obtain a value of 𝜏𝑐 = 0.19.
The thrust loading coefficient is connected to the minimum required projected area Projected area
𝐴𝑃 of the propeller via its definition (.).

𝑇prop
𝐴𝑃 = (.)
1 2
𝜌𝑣 𝜏
2 1 𝑐

Using a formula attributed to Admiral David W. Taylor, projected area 𝐴𝑃 may be

converted into developed area 𝐴𝐷 .
j j
𝐴𝑃
𝐴𝐷 = ( ) (.)
𝑃
1.067 − 0.229
𝐷
For propellers with zero or moderate rake, expanded blade area is equal to the developed
area.
𝐴𝐸 ≈ 𝐴𝐷 (.)
Substituting these two results into Equation (.) and dividing both sides by the
propeller disk area 𝐴0 = 𝜋𝐷2 ∕4 yields an equation for the required expanded area ratio.
( )
𝐴𝐸 𝑇prop
= ( ) (.)
𝐴0 req 1 2 𝑃 𝜋𝐷2
𝜌 𝑣1 𝜏𝑐 1.067 − 0.229
2 𝐷 4
The propeller will cavitate more than the limit only if the actual expanded area ratio
is smaller than (𝐴𝐸 ∕𝐴0 )req . We will use Burrill’s cavitation criteria in the context of
propeller design examples (see Chapter ).

46.4 Other Design Measures

As mentioned above, cavitation can rarely be avoided for vessels sailing faster than
35 knots. If excessive cavitation is expected, the following three alternatives to a con-
ventional propeller should be considered.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

558 46 Cavitation Prevention

• To avoid damage to the propeller and minimize effects on efficiency, special

wedge shaped foil sections should be used. They have a sharp leading edge which
creates a stable sheet cavitation bubble. The bubbles collapse way downstream of
the propeller and thus do not cause erosion of the blades.

• If the machinery arrangements allow a surface piercing propeller may be con-

sidered. In this case a part of the propeller disk stays above water. This causes
ventilation of the propeller blades. A sheet cavity will develop over the blade
but in contrast to sheet cavitation the bubbles are filled with air at atmospheric
pressure rather than vapor pressure. As a result damaging micro jets and pressure
waves are avoided.

• In some cases water jet propulsion may be a viable solution to extend the speed
range of the vessel without experiencing excessive cavitation.

References
auf ’m Keller, J. (). Enige aspecten bij het ontwerpen van scheepsschroeven. Schip
en Werf, ():–. In Dutch.
Burrill, L. (). Developments in propeller design and manufacture for merchant
ships. Transactions of the Institute of Marine Engineers, :–.
j Burrill, L. and Emerson, A. (). Propeller cavitation: further tests on in propeller j
models in the King’s College cavitation tunnel. Transactions of the North East Coast
Institution of Engineers and Shipbuilders (NECIES), :–.
Carlton, J. (). Marine propellers and propulsion. Butterworth-Heinemann, Oxford,
United Kingdom.
Kristensen, H. and Lützen, M. (). Prediction of resistance and propulsion power
of ships. Technical Report Project No. -, Work Package , Report No. ,
University of Southern Denmark and Technical University of Denmark.
Papanikolaou, A. (). Ship design – Methodologies of preliminary design. Springer,
Dordrecht, The Netherlands.
Watson, D. (). Practical ship design. Elsevier Ocean Engineering Book Series.
Elsevier Science, Oxford, United Kingdom.

Self Study Problems

. Summarize in your own words design measures to decrease the danger of pro-
peller cavitation.

. A propeller has to be checked for cavitation. It is designed to generate a thrust

of 𝑇 = 1250.0 kN. Should the expanded area ratio stated in the table below be
increased to prevent exceedance of the cavitation limit given by the % back
cavitation criteria (.) or should it be decreased to maximize efficiency? Show
all steps to reason your answer.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

46.4 Other Design Measures 559

Propeller data
number of blades 𝑍=
diameter 𝐷 = . m
pitch–diameter ratio 𝑃 ∕𝐷 = .
expanded are ratio 𝐴𝐸 ∕𝐴0 = .
propeller thrust 𝑇prop = . kN
speed of advance 𝑣𝐴 = . m/s
rate of revolution 𝑛 = . rpm
air pressure 𝑝𝐴 =  Pa
vapor pressure 𝑝𝑣 =  Pa
submergence of propeller shaft ℎ0 = . m
water density 𝜌 = . kg/m3
gravitational acceleration 𝑔 = . m/s2

. A propeller is operating under the following conditions:

Propeller data
propeller thrust (service) 𝑇 = . kN
propeller diameter 𝐷 = . m
pitch–diameter ratio 𝑃 ∕𝐷 = .
rate of revolution 𝑛 = . rpm
speed of advance 𝑣𝐴 = . m/s
shaft submergence ℎ0 = . m
j standard air pressure 𝑝𝐴 = . hPa j
gravitational acceleration 𝑔 = . m/s2
salt water temperature 𝑡𝑊 = . ◦ C

Compute the propeller cavitation number 𝜎𝑏 according to Equation (.) and

derive the required expanded area ratio to avoid more than 5% back cavitation.
. How does the required expanded area ratio in the previous problem change if
you employ the .% or the % back cavitation limit?

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

560

47
Propeller Series Data

The propeller is an essential piece in most ship propulsion systems. Its performance
characteristics have to be estimated early in the design because powering requirements
of the ship and the resulting engine selection have repercussions on the ship’s general
layout. Naval architects aim to select a propeller which combines hull and propeller into
a highly efficient propulsion system. Propeller series data is a very useful asset in this
process. A propeller series is a set of propellers in which key propeller characteristics
have been systematically varied. Many different propeller series have been developed
over the years, and data has been published to varying degrees. In this chapter we
concentrate on the Wageningen B-Series (Oosterveld and van Oossanen, ; Kuiper,
). The series covers a wide range of principal propeller characteristics. Its data
is available in the form of charts and regression polynomials. References for other
j propeller series are presented at the end of the chapter. j

Learning Objectives
At the end of this chapter students will be able to:

• understand the definition of systematic propeller series

• compute propeller characteristics from regression polynomials

• know about important propeller series

47.1 Wageningen B-Series

Origin The Wageningen B-Series was developed in the middle of the th century and is
the most extensive propeller series (Troost, , , ). Its name refers to the
town Wageningen, which is the home of the Netherlands Ship Model Basin (NSMB).
NSMB is now called MARIN (Maritime Research Institute Netherlands). In older
books the series might be referred to as Troost series. The series encompasses a total
of 120 propellers with systematic variations of number of blades 𝑍, expanded area
ratio 𝐴𝐸 ∕𝐴0 , and pitch–diameter ratio 𝑃 ∕𝐷. The Wageningen B-Series propeller had
been among the first to use aerofoil shaped section profiles. Circular segment (ogival)
sections have only been used from radius 𝑥 = 0.6 to the blade tip. Figure . shows

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

47.2 Wageningen B-Series Polynomials 561

Table 47.1 Basic characteristics of the propellers in the Wageningen B-Series. For each
combination of 𝑍 and 𝐴𝐸 ∕𝐴0 propellers with 𝑃 ∕𝐷 = 0.5, 0.6, 0.8, 1.0, 1.2, and 1.4 have been
tested

𝑍 expanded area ratios 𝐴𝐸 ∕𝐴0

 .
 . . . .
 . . . . .
 . . . .
 . . .
 . . .

the blade sections within the expanded view. The propellers feature a 15 degree rake
which is rather high by modern standards.
There are twenty different combinations of 𝑍 and 𝐴𝐸 ∕𝐴0 , and for each of the com- Data range
binations six pitch–diameter ratios have been tested. Table . summarizes the 20
basic combinations. Over the years propellers have been added to the series and several
adjustments have been made. For instance, the first set of propellers with 𝑍 = 4 blades
featured a reduction of the pitch by 20% from radius 𝑥 = 0.475 to the hub. This feature
was removed for propellers with less or more than four blades. The original blade
contour was widened later to improve cavitation characteristics. This shape was called
the BB-Series and equations presented here refer to this form (Kuiper, ). The
j j
hydrodynamic characteristics of the Wageningen B-Series tests have been reported in
graphical form as open water diagrams and as regression formulas. The set of data was
extended with open water tests under cavitation conditions.

47.2 Wageningen B-Series Polynomials

Oosterveld and van Oossanen () present a detailed analysis of the available open Regression
water test results for the Wageningen B-Series. Coefficients are provided to retrieve analysis
blade and foil section shapes and to compute the thrust and torque coefficients as
a function of the basic propeller characteristics diameter 𝐷, number of blades 𝑍,
expanded area ratio 𝐴𝐸 ∕𝐴0 , pitch–diameter ratio 𝑃 ∕𝐷, and the advance ratio 𝐽 .
For our studies in ship propulsion, we need the open water thrust and torque curves.
Oosterveld and van Oossanen () provide regression formulas for thrust and torque
coefficients:
∑38 ( )𝑐𝑖 (𝐴 )𝑑𝑖
𝑃 𝐸
𝐾𝑇 = 𝑎𝑖 𝐽 𝑏𝑖 𝑍 𝑒𝑖 (.)
𝑖=0
𝐷 𝐴0

∑
46 ( )𝑐𝑖 (𝐴 )𝑑𝑖
𝑃 𝐸
𝐾𝑄 = 𝑎 𝑖 𝐽 𝑏𝑖 𝑍 𝑒𝑖 (.)
𝑖=0
𝐷 𝐴0

The necessary coefficients 𝑎𝑖 and exponents 𝑏𝑖 , 𝑐𝑖 , 𝑑𝑖 , and 𝑒𝑖 for the thrust coefficient 𝐾𝑇
are listed in Table .. Coefficients and exponents for Equation (.) can be found in
Table ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

562 47 Propeller Series Data

Figure 47.1 Open water chart for a Wageningen B-Series propeller with 𝑍 = 4 and
𝐴𝐸 ∕𝐴0 = 0.85 derived from 𝐾𝑇 and 𝐾𝑄 polynomials Equations (.) and (.)
j j
The open water efficiency 𝜂𝑂 follows from Equation (.): 𝜂𝑂 = (𝐾𝑇 ∕𝐾𝑄 )𝐽 ∕(2𝜋).
Figure . shows an open water chart based on the polynomials. Note that the open
water efficiency has its own scale on the right-hand side of the plot. The key ‘B-’
used in the title identifies the propeller:

• B obviously stands for Wageningen B-Series.

• The number  indicates the number of blades 𝑍 = 4.

• The value after the hyphen (or sometimes a dot) represents the expanded area
ratio multiplied by , i.e. in Figure . it is 𝐴𝐸 ∕𝐴0 = 0.85.

For example, B. or B- stand for a Wageningen B-Series propeller with 𝑍 = 6
blades and an expanded area ratio of 𝐴𝐸 ∕𝐴0 = 105∕100 = 1.05.
Studying thrust, torque, and open water efficiency curves in Figure ., it is obvious
that thrust and torque increase with increasing pitch–diameter ratio 𝑃 ∕𝐷. At the same
advance coefficient 𝐽 , a propeller with higher 𝑃 ∕𝐷 produces more thrust but also
requires more torque, i.e. engine power. The picture is not quite as simple for the open
water efficiency 𝜂𝑂 . The maximum achievable open water efficiency also grows with the
pitch–diameter ratio, at least for the range of 𝑃 ∕𝐷-values shown. However, for a given
advance coefficient 𝐽 , a higher pitch–diameter ratio might not necessarily provide a
better open water efficiency. Compare the open water efficiency curves in Figure .
at 𝐽 = 0.6! The 𝜂𝑂 -curve for 𝑃 ∕𝐷 = 0.8 has the largest open water efficiency whereas
the efficiencies for 𝑃 ∕𝐷 = 1.0, 1.2, and 1.4 are smaller. Thus a propeller with high

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

47.2 Wageningen B-Series Polynomials 563

Table 47.2 Factors and exponents for thrust coefficient polynomials of Wageningen B-Series
propellers (Oosterveld and van Oossanen, 1975)

Open Water Thrust Coefficient 𝐾𝑇

𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒 𝑖 𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒𝑖
 +0.00880496      +0.010465    
 −0.204554      −0.00648272    
 +0.166351      −0.00841728    
 +0.158114      +0.0168424    
 −0.147581      −0.00102296    
 −0.481497      −0.0317791    
 +0.415437      +0.018604    
 +0.0144043      −0.00410798    
 −0.0530054      −0.000606848    
 +0.0143481      −0.0049819    
 +0.0606826      +0.0025983    
 −0.0125894      −0.000560528    
 +0.0109689      −0.00163652    
 −0.133698      −0.000328787    
 +0.00638407      +0.000116502    
 −0.00132718      +0.000690904    
 +0.168496      +0.00421749    
 −0.0507214      +0.0000565229    
 +0.0854559      −0.00146564    
 −0.0504475    
j j

pitch–diameter ratio will be efficient only if it operates at a high advance coefficient. We

will learn how to find the best combination of advance coefficient and pitch–diameter
ratio in the following chapter.
All model propellers in the series have a diameter of 𝐷 = 240 mm and some scale Viscous scale
effects due to the model scale Reynolds numbers are expected. The polynomials (.) effects
and (.) have been normalized to a Reynolds number 𝑅𝑒0.75 of 2 ⋅ 106 . The Reynolds
number 𝑅𝑒0.75 is based on the chord length at radius 𝑥 = 0.75. As resultant flow speed
the combined magnitude of axial and rotational speed is used and induced velocities
are ignored.
√
𝑐0.75 𝑣2𝐴 + (0.75 𝜋 𝑛 𝐷)2
𝑅𝑒0.75 = (.)
𝜈
The chord length 𝑐0.75 at radius 𝑥 = 0.75 may be estimated with Equation (.).

( )
𝐷 𝐴𝐸
𝑐(𝑥) = 𝐶𝑟 (𝑥) (.)
𝑍 𝐴0
Values for the coefficient 𝐶𝑟 are listed in Table ..
The results from thrust and torque coefficient polynomials (.) and (.) should Full scale thrust
be corrected for the full scale Reynolds number. The correction may be computed via and torque

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

564 47 Propeller Series Data

Table 47.3 Factors and exponents for torque coefficient polynomials of Wageningen B-Series
propellers (Oosterveld and van Oossanen, 1975)

Open Water Torque Coefficient 𝐾𝑄

𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒𝑖 𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒𝑖
 +0.00379368      −0.0397722    
 +0.00886523      −0.00350024    
 −0.032241      −0.0106854    
 +0.00344778      +0.00110903    
 −0.0408811      −0.000313912    
 −0.108009      +0.0035985    
 −0.0885381      −0.00142121    
 +0.188561      −0.00383637    
 −0.00370871      +0.0126803    
 +0.00513696      −0.00318278    
 +0.0209449      +0.00334268    
 +0.00474319      −0.00183491    
 −0.00723408      +0.000112451    
 +0.00438388      −0.0000297228    
 −0.0269403      +0.000269551    
 +0.0558082      +0.00083265    
 +0.0161886      +0.00155334    
 +0.00318086      +0.000302683    
 +0.015896      −0.0001843    
 +0.0471729      −0.000425399    
j  +0.0196283      +0.0000869243    
j
 −0.0502782      −0.0004659    
 −0.030055      +0.0000554194    
 +0.0417122    

Table 47.4 Coefficients for the estimate of maximum 𝑥 = 𝑟∕𝑅 𝐴𝑟 𝐵𝑟 𝐶𝑟

thickness and chord length of Wageningen B-Series 0.15 0.0588 0.00425 1.473
propellers (Kuiper, 1992). For convenience values have 0.20 0.0526 0.0040 1.600
been added at radii 𝑥 = 0.15, 0.25, and 0.75 by
0.25 0.0495 0.00375 1.719
interpolation
0.30 0.0464 0.0035 1.832
0.40 0.0402 0.0030 2.023
0.50 0.0340 0.0025 2.163
0.60 0.0278 0.0020 2.243
0.70 0.0216 0.0015 2.247
0.75 0.0185 0.00125 2.208
0.80 0.0154 0.0010 2.132
0.85 0.0123 0.00075 2.005
0.90 0.0092 0.0005 1.798
0.95 0.0061 0.00025 1.434
0.975 0.00455 0.000125 1.122
1.0 0.003 0.000 0.000

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

47.3 Other Propeller Series 565

the ITTC procedure described in Chapter  (ITTC, ) or the following regression
polynomials (Oosterveld and van Oossanen, ).

∑
8 ( )𝑐𝑖 (𝐴 )𝑑𝑖 ( )𝑓
𝑃 𝐸
Δ𝐾𝑇 = 𝑎𝑖 𝐽 𝑏𝑖
𝑍 𝑒𝑖 log10 𝑅𝑒0.75 − 0.301 𝑖 (.)
𝑖=0
𝐷 𝐴0
∑12 ( )𝑐𝑖 𝐴 )𝑑𝑖
(
( )𝑓
𝑏𝑖 𝑃 𝐸
Δ𝐾𝑄 = 𝑎𝑖 𝐽 𝑍 𝑒𝑖 log10 𝑅𝑒0.75 − 0.301 𝑖 (.)
𝑖=0
𝐷 𝐴0

Coefficients for the Reynolds number corrections to thrust and torque coefficients are
provided in Tables .(a) and .(b) respectively.
The ITTC procedure also requires the thickness to chord length ratio 𝑡max ∕𝑐. The Thickness to
maximum thickness of the standard form is given by Equation (.). chord length
ratio
( )
𝑡max (𝑥) = 𝐷 𝐴𝑟 (𝑥) − 𝐵𝑟 (𝑥)𝑍 (.)

Note that structural strength may require thicker blade sections in some cases. Values
for the coefficients 𝐴𝑟 and 𝐵𝑟 are also found in Table .. The original data from
Kuiper () do not contain values for the radii 0.15 (already within hub), 0.25, and
0.75. These values have been found by interpolation through the given data points.
In contrast to the equivalent full scale correction in the ITTC method (see Equa-
tions (.) and (.)), Δ𝐾𝑇 (.) and Δ𝐾𝑄 (.) are added to the values deter-
j mined with Equations (.) and (.). j

𝐾𝑇 𝑆 = 𝐾𝑇 + Δ𝐾𝑇 𝐾𝑄𝑆 = 𝐾𝑄 + Δ𝐾𝑄 (.)

The correction Δ𝐾𝑄 should be negative and Δ𝐾𝑇 should be positive for Reynolds
numbers 𝑅𝑒0.75 > 2 ⋅ 106 .

47.3 Other Propeller Series

This section briefly summarizes other propeller series with references for further study.
An extensive summary of different propeller series is given in Carlton ().
The Gawn series is a set of 37 three-bladed propellers with pitch–diameter ratios be- Gawn series
tween 0.4 and 2.0 and expanded area ratios from 0.2 to 1.1 (Gawn, ). Its extended
blade shapes are elliptical and only ogival sections have been used. Although the geom-
etry is outdated, this series is important because its model propellers are exceptionally
large with a diameter of 20 in (508 mm). The series data therefore has very little to
no scale effects. In addition, the Gawn series also covers a wide pitch–diameter ratio
which makes it applicable to high-speed and naval vessels. Blount and Hubble ()
provide regression polynomials analog to the ones for the Wageningen B-Series. The
polynomials are also published in Carlton ().
The open water tests for the Gawn series were conducted in a towing tank and perfor- King’s College
mance under cavitation conditions was originally not investigated. However, Gawn Admiralty (KCA)
series
and Burrill () tested a set of  slightly smaller propellers with 16 in (406.4 mm)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

566 47 Propeller Series Data

Table 47.5 Factors and exponents for Reynolds number effects on thrust coefficient and torque coefficient of
Wageningen B-Series propellers (Oosterveld and van Oossanen, 1975)

(a) Coefficients and exponents for thrust (b) Coefficients and exponents for torque
correction Δ𝐾𝑇 correction Δ𝐾𝑄

𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒𝑖 𝑓𝑖 𝑖 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 𝑒𝑖 𝑓𝑖
 +0.000353485       −0.000591412     
 −0.00333758       +0.00696898     
 −0.00478125       −0.0000666654     
 +0.000257792       +0.0160818     
 +0.0000643192       −0.000938091     
 −0.0000110636       −0.00059593     
 −0.0000276305       +0.0000782099     
 +0.0000954       +0.0000052199     
 +0.0000032049       −0.00000088538     
 +0.0000230171     
 −0.00000184341     
 −0.00400252     
 +0.000220915     

diameter in the cavitation tunnel of the King’s College, Newcastle, UK . The series was
j conceived for fast naval vessels, hence the name King’s College Admiralty or short KCA j
series. It is also referred to as Gawn-Burrill series. Nowadays it is mostly used as design
basis for high-speed small craft.
Newton–Rader Additional information for propeller performance under cavitation conditions may be
series derived from the Newton–Rader series (Newton and Rader, ). This series consists
of 12 three-bladed propellers. Expanded area ratios of 𝐴𝐸 ∕𝐴0 = 0.48, 0.71, and 0.95
have been tested each with four pitch–diameter ratios between 𝑃 ∕𝐷 = 1.04 and 2.08.
This series is of interest for the design of small high-speed craft where significant
cavitation is suspected.
DTMB skewed The effect of skew on the propeller was investigated by Boswell (). Open water and
propeller series cavitation data is reported for four propellers with skews of 0, 36, 72, and 108 degree
at the blade tip. No other propeller characteristics were modified besides the skew.
Therefore, this mini-series is not suitable for propeller selection. Nevertheless, the
propellers provide important test cases for computer codes that predict effects of skew
on propeller performance.
Ducted Oosterveld () conducted extensive open water and cavitation tests with ducted
propellers, propellers, the so called Ka-Series. Regression polynomials are provided for thrust
Ka-Series
and torque of the propeller as well as the influence of the nozzle (duct). Four different
nozzle types have been investigated.
Controllable Series data are limited for controllable pitch propellers. Data for three-bladed pro-
pitch propellers pellers can be found in Gutsche and Schroeder () and Chu et al. (). Hansen
() reports data for five different controllable pitch propellers with four blades.

 In  King’s College was renamed into University of Newcastle upon Tyne

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

47.3 Other Propeller Series 567

MARIN recently added the Wageningen C- and D-Series to its portfolio which provide
data for controllable pitch propellers with and without duct. Dang et al. () provide
some results in the form of charts but regression polynomials have not yet been released
to the public.
In the following two Chapters we will use propeller series data to select a suitable
propeller for a ship. Chapter  explains the principal propeller selection tasks and
process. Chapter  provides two extended examples.

References
Blount, D. and Hubble, E. (). Sizing segmental section commercially available
propellers for small craft. In Transactions of SNAME, volume . The Society of
Naval Architects and Marine Engineers.
Boswell, R. (). Design, cavitation performance, and open-water performance of
series of research skewed propellers. Report , Naval Ship Research Development
Center, Washington, D.C.
Carlton, J. (). Marine propellers and propulsion. Butterworth-Heinemann, Oxford,
United Kingdom.
Chu, C., Chan, Z., She, Y., and Yuan, V. (). The -bladed JD-CPP series. In th Lips
Propeller Symposium, pages –, Drunen, The Netherlands.
j Dang, J., van den Boom, H., and Ligtelijn, J. (). The Wageningen C- and D-series j
propellers. In Proc. of the FAST  Conference, Amsterdam, The Netherlands.
Paper ID .
Gawn, R. (). Effect of pitch and blade width on propeller performance. In Trans-
actions of The Royal Institution of Naval Architects (RINA), volume . The Royal
Institution of Naval Architects.
Gawn, R. and Burrill, L. (). Effect of cavitation on the performance of a series of
 in. model propellers. In Transactions of The Royal Institution of Naval Architects
(RINA), volume . The Royal Institution of Naval Architects.
Gutsche, F. and Schroeder, G. (). Freifahrversuche an Propellern mit festen und
verstellbaren Flügeln ‘voraus’ und ‘rückwärts’. Schiffbauforschung, ():–.
Hansen, E. (). Thrust and blade spindle torque measurements of five control-
lable pitch propeller designs for MSO . Report , Naval ship Research and
Development Center, Washington, D.C.
ITTC (). Open water test. International Towing Tank Conference, Recommended
Procedures and Guidelines .---.. Revision .
Kuiper, G. (). The Wageningen propeller series. Publication -. Maritime
Research Institute Netherlands (MARIN), Wageningen, The Netherlands.
Newton, R. and Rader, H. (). Performance data of propellers for high speed craft.
In Transactions of The Royal Institution of Naval Architects (RINA), volume . The
Royal Institution of Naval Architects.
Oosterveld, M. (). Wake adapted ducted propellers. NSBM Publication , Nether-
lands Ship Model Basin, Wageningen, The Netherlands. NSBM is now MARIN.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

568 47 Propeller Series Data

Oosterveld, M. and van Oossanen, P. (). Further computer-analyzed data of the

Wageningen B-screw series. International Shipbuilding Progress, :–.
Troost, L. (). Open water test series with modern propeller forms. Transactions of
the North East Coast Institution of Engineers and Shipbuilders (NECIES), .
Troost, L. (). Open water test series with modern propeller forms II. three bladed
propellers. Transactions of the North East Coast Institution of Engineers and Ship-
builders (NECIES), .
Troost, L. (). Open water test series with modern propeller forms III. Two bladed
and five bladed propellers– extension of the three and five bladed b-series. Trans-
actions of the North East Coast Institution of Engineers and Shipbuilders (NECIES),
.

Self Study Problems

. The Wageningen B-Series was developed more than  years ago. Explain why it
is still relevant.
. Why are there polynomials to correct thrust and torque for full scale Reynolds
numbers for the Wageningen B-Series but not for the Gawn series?
. What is the Newton-Rader propeller series known for?
j j
. Implement the Wageningen B-Series polynomials as a program (use Python,
Matlab, or similar) so that you can compute and visualize propeller open water
characteristics. The following table provides some test data:

Prop 1 𝑍 = 4, 𝑃 ∕𝐷 = 0.7000, 𝐴𝐸 ∕𝐴0 = 0.5500

𝐽: 0.000000 0.200000 0.400000 0.600000 0.800000 1.000000
𝐾𝑇 : 0.293034 0.235922 0.163956 0.080357 −0.011655 −0.108858
10𝐾𝑄 : 0.314941 0.265489 0.202513 0.125021 0.032021 −0.077478
Prop 2 𝑍 = 5, 𝑃 ∕𝐷 = 0.7000, 𝐴𝐸 ∕𝐴0 = 0.5500
𝐽: 0.000000 0.200000 0.400000 0.600000 0.800000 1.000000
𝐾𝑇 : 0.304817 0.247985 0.173682 0.084871 −0.015485 −0.124422
10𝐾𝑄 : 0.325948 0.279672 0.218642 0.138134 0.033421 −0.100223
Prop 3 𝑍 = 4, 𝑃 ∕𝐷 = 0.9000, 𝐴𝐸 ∕𝐴0 = 0.5500
𝐽: 0.000000 0.200000 0.400000 0.600000 0.800000 1.000000
𝐾𝑇 : 0.382545 0.327866 0.258221 0.176763 0.086647 −0.008971
10𝐾𝑄 : 0.502623 0.442389 0.366125 0.273048 0.162376 0.033326
Prop 4 𝑍 = 5, 𝑃 ∕𝐷 = 0.9000, 𝐴𝐸 ∕𝐴0 = 0.7000
𝐽: 0.000000 0.200000 0.400000 0.600000 0.800000 1.000000
𝐾𝑇 : 0.413463 0.353818 0.276834 0.186159 0.085442 −0.021668
10𝐾𝑄 : 0.550275 0.483373 0.396737 0.290990 0.166755 0.024655
Prop 5 𝑍 = 6, 𝑃 ∕𝐷 = 1.0000, 𝐴𝐸 ∕𝐴0 = 0.9500
𝐽: 0.000000 0.200000 0.400000 0.600000 0.800000 1.000000
𝐾𝑇 : 0.500113 0.434368 0.349707 0.250456 0.140944 0.025498
10𝐾𝑄 : 0.754051 0.666074 0.552772 0.420127 0.274123 0.120741

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

569

48
Propeller Design Process

Once estimates of resistance and hull–propeller interaction parameters are completed,

a propeller must be selected before propulsive power for a ship can be determined. As a
first step in the propeller design process, basic characteristics are selected on the basis
of propeller series data. Afterwards, details of the propeller design are determined with
additional computational tools like propeller lifting line and lifting surface programs.
In this and the following chapter we study the first step in the propeller design process
and define principal propeller characteristics based on Wageningen B-series data. Here,
the objectives are discussed which govern the selection process and the basic tools are
described. Complete examples are provided in the following chapter.

j j
Learning Objectives
At the end of this chapter students will be able to:

• understand and prepare propeller design tasks

• compute required design constants

• understand the process of hull–propeller matching

• find optimum combinations of propeller characteristics for given thrust or pow-

ering requirements

48.1 Design Tasks and Input Preparation

Propeller design problems may be subdivided into two groups: Basic design
tasks

(i) optimum diameter design problems, and

(ii) optimum rate of revolution design problems.

In the first group (i) the rate of revolution is known, and we search for the optimum
diameter which maximizes efficiency of the propeller. In the second group (ii) the
diameter is known, and we look for the best rate of revolution. In each group we have

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

570 48 Propeller Design Process

Table 48.1 The four basic propeller design tasks

Task given wanted design constant

delivered power 𝑃𝐷 diameter 𝐷 [ ]
𝐾𝑄 𝑃𝐷 𝑛2 𝜂𝑅
1 speed of advance 𝑣𝐴 pitch–diameter ratio 𝑃 ∕𝐷 =
𝐽5 2𝜋 𝜌 𝑣𝐴5
optimum
diameter

rate of revolution 𝑛 open water efficiency 𝜂𝑂

thrust 𝑇 diameter 𝐷 [ ]
𝐾𝑇 𝑇 𝑛2
2 speed of advance 𝑣𝐴 pitch–diameter ratio 𝑃 ∕𝐷 =
𝐽4 𝜌 𝑣𝐴4
rate of revolution 𝑛 open water efficiency 𝜂𝑂

delivered power 𝑃𝐷 rate of revolution 𝑛 [ ]

𝐾𝑄 𝑃𝐷 𝜂𝑅
speed of advance 𝑣𝐴 pitch–diameter ratio 𝑃 ∕𝐷 =
optimum rate

3
of revolution

𝐽3 2𝜋 𝜌 𝐷2 𝑣𝐴3
diameter 𝐷 open water efficiency 𝜂𝑂

thrust 𝑇 rate of revolution 𝑛 [ ]

𝐾𝑇 𝑇
4 speed of advance 𝑣𝐴 pitch–diameter ratio 𝑃 ∕𝐷 =
𝐽2 𝜌 𝐷2 𝑣𝐴2
diameter 𝐷 open water efficiency 𝜂𝑂

two subcategories, depending on whether the delivered power 𝑃𝐷 or the thrust 𝑇 is

provided as input. Table . summarizes the four basic propeller design tasks. Note
that the number of blades 𝑍 and the expanded area ratio 𝐴𝐸 ∕𝐴0 are not listed. The
j number of blades 𝑍 is chosen to detune natural frequencies of propeller, shaft, and j
engine. The expanded area ratio 𝐴𝐸 ∕𝐴0 is determined by a suitable cavitation criterion
(see Chapter ). Besides 𝑍 and 𝐴𝐸 ∕𝐴0 , the input for a propeller design task consists
of a combination of speed of advance 𝑣𝐴 , thrust 𝑇 or delivered power 𝑃𝐷 , and rate of
revolution 𝑛 or diameter 𝐷. The missing parameters pitch–diameter ratio 𝑃 ∕𝐷 and
diameter 𝐷 or rate of revolution 𝑛 are chosen to provide the highest possible open
water efficiency 𝜂𝑂 . How will be discussed below.
Service margin Delivered power 𝑃𝐷 and thrust 𝑇 are commonly determined for trial conditions (calm
and deep water, no wind, no current). For propeller design purposes 𝑃𝐷 and 𝑇 have to
be augmented with a service margin 𝑠𝑀 .

𝑇service = (1 + 𝑠𝑀 ) 𝑇 𝑃𝐷service = (1 + 𝑠𝑀 ) 𝑃𝐷 (.)

The service or sea margin accounts for increases in ship resistance during operation.
Over time fouling will roughen the hull surface and increase its frictional resistance. It
will also increase the wake fraction. Wind, current, and waves will increase resistance
as well. If the propeller is erroneously optimized for trial conditions, it will have too
much pitch. As a consequence, the engine might not reach its design rate of revolution
during normal operations.
Service margins are to a large degree a matter of choice. The margin is influenced by

• owner preferences
• target area of operations
• size and trade of the vessel

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.2 Optimum Diameter Selection 571

Resistance of a vessel will be higher than the calm water resistance even during normal
operation in fair weather. Wind, waves, and currents may cause the resistance to
increase by %–%. In head seas and heavy weather it might be even more; however,
the ship will sail more slowly anyway to avoid damage to its structure. Typical weather
patterns influence the size of the margin. Winter navigation on the rough North Atlantic
between Europe and North America will require a higher service margin than service
in the calmer Mediterranean. On one hand, a smaller margin might suffice for large
vessels because they are less affected by wind and waves. On the other hand, frictional
resistance is the largest resistance component for large and slow vessels. Therefore an
increase in surface roughness due to fouling will have a comparatively larger negative
effect than on a fast container ship for which wave (making) resistance is a sizable part
of resistance. Marine fouling can cause a 25%–50% increase in frictional resistance
depending on hull coating and docking intervals. Vessels which sail on a tight schedule,
like container ships and ferries, may need larger margins to be able to recover lost time.
In summary, it is important that thrust 𝑇 and delivered power 𝑃𝐷 reflect service condi-
tions rather than trial conditions. Current practice is to adjust trial estimates with a
service margin. Service margins are a crude tool considering the effort that goes into
estimating, measuring, and computing ship resistance and hull-propeller interaction
parameters. Many ships arguably sail across the oceans with underutilized engine
power. The ITTC added a manual to its set of recommended procedures with the goal
of making service margin estimates more rational ITTC (). See also the method
proposed by Stasiak ().
j Engine manufacturers provide for their engines the maximum continuous rating (𝑀𝐶𝑅) Engine margin j
and the associated rate of revolution 𝑛. The maximum continuous rating is the brake
power an engine can deliver during prolonged operation. 𝑀𝐶𝑅 and 𝑛 of an engine
may be adjusted within narrow limits. The brake power used for propeller selection
is typically 10% to 15% lower than MCR. This engine margin allows engine operation
with greater fuel economy during normal operation and provides reserve power for
maneuvering and bad weather. It comes on top of the service margin.
The available input data is combined into a dimensionless design constant specific for Design constants
each task. It will allow us to find the self propulsion points. This is essentially the same
process we used during the full scale power prediction (see Section .).

48.2 Optimum Diameter Selection

Let us first discuss the optimum diameter selection tasks. In group (i) the rate of Optimum
revolution 𝑛 has to be determined first. There are two likely scenarios: diameter
selection tasks

• The owner already operates ships of similar type and size. He may specify that
the new ship must have the same engine as an existing vessel. This will simplify
crew training and management. It may also lower maintenance cost.

• A preliminary powering estimate has been completed by assuming desired pro-

peller data. An engine is pre-selected based on the estimated delivered power.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

572 48 Propeller Design Process

48.2.1 Propeller design task 1

Task 1, given: Besides the rate of revolution 𝑛, we have to know the delivered power 𝑃𝐷 and the speed
𝑃𝐷 , 𝑣𝐴 , 𝑛 of advance 𝑣𝐴 (see Table .). Delivered power may be derived from brake power by
wanted:
𝐷, 𝑃 ∕𝐷, 𝜂𝑂
means of the mechanical or shafting efficiency: 𝑃𝐷 = 𝑃𝐵 𝜂𝑆 . Delivered power and rate
of revolution also specify the available propeller torque in behind condition.
𝑃𝐷
𝑄 = (.)
2𝜋 𝑛
Using the relative rotative efficiency (.) yields the torque 𝑄𝑂 for open water condi-
tions:
𝑄𝑂 = 𝑄 𝜂𝑅 (.)
The next step is to find propellers which absorb this torque at the provided combination
of speed of advance and rate of revolution.
Design constant The torque absorbed by the propeller is embodied in the torque coefficient 𝐾𝑄 of the
for task 1 open water diagram:
𝑄𝑂
𝐾𝑄 = (.)
𝜌 𝑛2 𝐷5
In this equation we miss the propeller diameter 𝐷. Dividing Equation (.) by the fifth
power of the advance ratio 𝐽 = 𝑣𝐴 ∕(𝑛𝐷) eliminates the diameter from the right-hand
side:
𝑄𝑂 𝑄𝑂
j j
𝐾𝑄 2
𝜌𝑛 𝐷 5 𝜌 𝑛2 𝐷5 𝑄𝑂 𝑛5 𝐷5 𝑄 𝑛3
= = ( ) = = 𝑂 (.)
𝐽5 𝐽5 𝑣𝐴 5 𝜌 𝑛2 𝐷5 𝑣𝐴5 𝜌 𝑣𝐴5
𝑛𝐷
The last expression in Equation (.) is known. Equations (.) and (.) replace the
open water torque with delivered power and relative rotative efficiency 𝜂𝑅 . The result
is the dimensionless design constant for task :
[ ]
𝐾𝑄 𝑃 𝑛2 𝜂𝑅
= 𝐷 (.)
𝐽5 2𝜋 𝜌 𝑣𝐴5

Setting the relative rotative efficiency to one is an option if a good estimate for it is
unavailable.
Example data Solution of propeller design task  is illustrated based on the data set given in Table ..
Input data has to be carefully converted to a consistent set of 𝑆𝐼-units:

• delivered power 𝑃𝐷 = 23 400.00 kW= 23 400 000 W,

• rate of revolution 𝑛 = 108.0 rpm= 1.8 s−1 , and
• ship speed 𝑣𝑆 = 21.3 kn= 0.958 m/s.

The speed of advance is not directly specified but we can compute it from ship speed
and wake fraction.

𝑣𝐴 = (1 − 𝑤)𝑣𝑆 = (1 − 0.24) 10.958 m/s = 8.295 m/s

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.2 Optimum Diameter Selection 573

Table 48.2 Input data to illustrate task 1: optimum propeller diameter selection based on
delivered power, speed of advance, and rate of revolution

Example ship data – Task 1

number of blades 𝑍=
expanded area ratio 𝐴𝐸 ∕𝐴0 = .
delivered power (service) 𝑃𝐷 =  . kW
rate of revolution 𝑛 = . rpm
design ship speed 𝑣𝑆 = . kn
wake fraction 𝑤 = .
relative rotative efficiency 𝜂𝑅 = .
salt water density (at  ◦ C) 𝜌 = . kg/m3

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves
1.4
[]

D =
P/
thrust and torque coefficients KT , 10 KQ

1.2 0.6

[]O
1.0 0.5

open water efficiency

P/D =
j 0.8 0.4 j
1.0
P/D =1.4
P /D
P/D =0

=1.
0.6 P/D =1.2 4 0.3
.8

P/D =1.0
P/D =

0.4 P /D
0.2
P/D =0.8 =1.2
0.6
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
advance coefficient J []

Figure 48.1 Design task 1 – Input: open water diagram for Wageningen B-series propellers with
𝑍 = 4 and 𝐴𝐸 ∕𝐴0 = 0.85 derived from 𝐾𝑇 and 𝐾𝑄 polynomials (47.1) and (47.2). Torque
coefficient curves 10𝐾𝑄 are emphasized

Based on the input data above, Equation (.) yields the design constant value for Step 1: design
task . constant

[ ] ( )2
𝐾𝑄 𝑃𝐷 𝑛2 𝜂𝑅 23400000 W ⋅ 1.8 s−1 ⋅ 1.03
= = ( )5 = 0.30846 (.)
𝐽5 2𝜋 𝜌 𝑣𝐴5 2𝜋 1026.021 kg m−3 ⋅ 8.295 m/s

Check the units to make sure the design constant is dimensionless. The design constant
for task  measures how much torque is available to turn the propeller.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

574 48 Propeller Design Process

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves 4
=1.
[]

P / D
available torque curve
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

10[ 5 ]J 5
KQ

[]
J

O
1.0 0.5

open water efficiency

P/D =
0.8 0.4

1.0
P/D =1.4
P /D

P/D =0
=1.
0.6 P/D =1.2 4 0.3

.8
P/D =1.0

P/D =
0.4 P /D
0.2
P/D =0.8 =1.2

0.6
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
advance coefficient J []

Figure 48.2 Design task 1 – Step 2: locate self propulsion points ◦ at which the propellers
absorb the delivered power specified with the design constant [𝐾𝑄 ∕𝐽 5 ] from Equation (48.7)
j j
Step 2: self With this design constant a self propulsion point may be estimated for a given propeller.
propulsion Of course, we do not yet have a specific one, but we have the open water diagram for
points
the B-Series propellers B- (Figure .). The curves represent six propellers with
varying pitch–diameter ratios but all have four blades and an expanded area ratio of
𝐴𝐸 ∕𝐴0 = 0.85. The torque coefficient curves have been emphasized in Figure .
because they represent the torque needed to turn the propellers as a function of the
advance ratio 𝐽 . This is equivalent to saying that the propellers absorb the delivered
power 2𝜋 𝑛 𝑄𝑂 .
The dimensionless torque available to the propeller is obtained by multiplying the
design constant value (.) with 𝐽 5 .
[ ]
𝐾𝑄
10 𝐾𝑄available = 10 𝐽 5 = 3.0846 𝐽 5 available torque curve (.)
𝐽5
We compute 10𝐾𝑄available for a set of advance ratios and plot the result as curve into
the open water diagram (Figure .). Do not forget that in most cases the open water
diagram shows ten times the torque coefficient! The intersections of the available torque
curve and the six open water torque curves represent the self propulsion points. They
are marked with circles ◦ in Figure .. At the self propulsion points the torque from
the propeller shaft is in equilibrium with the torque a propeller requires to turn.
Any of the six propellers is a possible solution to our design problem. For example,
for a pitch–diameter ratio 𝑃 ∕𝐷 = 0.8 the intersection point is at the advance ratio

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.2 Optimum Diameter Selection 575

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves 4
=1.
[]

P / D
available torque curve
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

10[ 5 ]J 5
KQ

[]
O at self propulsion points

O
1.0 0.5

open water efficiency

P/D =
0.8 0.4

1.0
P/D =1.4
P /D

P/D =0
=1.
0.6 P/D =1.2 4 0.3

.8
P/D =1.0

P/D =
0.4 P /D
0.2
P/D =0.8 =1.2
0.6
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
advance coefficient J []

Figure 48.3 Design task 1 – Step 3: find open water efficiencies × for self propulsion points ◦

j j
𝐽0.8 = 0.5775. From the advance ratio 𝐽0.8 we may compute the missing propeller
diameter 𝐷0.8 .

𝑣𝐴 𝑣𝐴 8.295 m/s
𝐽0.8 = ⟶ 𝐷0.8 = = = 7.979 m
𝑛 𝐷0.8 𝑛 𝐽0.8 1.8 s−1 ⋅ 0.5775

The open water efficiency of the propeller with 𝑃𝐷 = 0.8 at 𝐽0.8 = 0.5775 is 0.5982.
That is not necessarily bad; however, our objective is to find the pitch–diameter ratio
which will maximize the efficiency. Therefore, a check of the achievable open water
efficiency is needed.
In Figure . the open water efficiencies are marked with crosses × for each self Step 3: open
propulsion point. Of these, the open water efficiency is highest for 𝑃 ∕𝐷 = 1.0. water efficiencies

The best possible open water efficiency seems to be found for a pitch–diameter ratio Step 4: auxiliary
in between 𝑃 ∕𝐷 = 0.8 and 1.0. We draw an interpolation curve through the six open curve
water efficiency points × (see Figure .). This curve is sometimes called the auxiliary
curve. It represents the achievable open water efficiency for this set of B- propellers
and the given design constant (.).
The maximum of the auxiliary curve marks the highest open water efficiency which Step 5: optimum
can be achieved for the chosen data set. Figure . shows the optimum at 𝜂𝑂 = 0.6051 propeller
and the associated self propulsion point □ and optimum advance ratio 𝐽 = 0.6202.
The optimum advance ratio yields an optimum diameter of
𝑣𝐴 8.295 m/s
𝐷𝑂 = = = 7.430 m
𝑛𝐽 1.8 s−1 ⋅ 0.6202

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

576 48 Propeller Design Process

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves auxiliary curve 4
=1.
[]

P / D
available torque curve
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

10[ 5 ]J 5
KQ

[]
O at self propulsion points

auxiliary curve
J

O
1.0 0.5

open water efficiency

P/D =
0.8 0.4

1.0
P/D =1.4
P /D

P/D =0
=1.
0.6 P/D =1.2 4 0.3

.8
P/D =1.0

P/D =
0.4 P /D
0.2
P/D =0.8 =1.2

0.6
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
advance coefficient J []

Figure 48.4 Design task 1 – Step 4: draw auxiliary curve through open water efficiency values

j B-Series propeller B4-85 j

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves optimum O =0.6051 4
=1.
[]

available torque curve P /D

thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

10[ 5 ]J 5
KQ
[]
O at self propulsion points

auxiliary curve
J
O

1.0 0.5
open water efficiency
P/D =

0.8 0.4
1.0

P/D =1.4
P /D
P/D =0

=1.
0.6 P/D =1.2 4 0.3
.8

P/D =1.0 optim

um P
/D =0.92
P/D =

0.4 0.2
P/D =0.8 27 self propulsion pointP/D =1.2
0.6

of optimum propeller
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
optimum J =0.620 2 advance coefficient J []

Figure 48.5 Design task 1 – Result: optimum propeller is defined by maximum of auxiliary curve

This diameter is optimal for open water condition only. The optimum propeller diameter
is 2%–5% smaller for the behind condition. We will come back to this mystery in the
following chapter.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.2 Optimum Diameter Selection 577

Table 48.3 Results for the propeller selection task 1 example

Optimum open water propeller data – Task 1

number of blades 𝑍=
expanded area ratio 𝐴𝐸 ∕𝐴0 = .
delivered power (service) 𝑃𝐷 =  . kW
rate of revolution 𝑛 = . rpm
advance ratio 𝐽 = .
pitch–diameter ratio 𝑃 ∕𝐷 = .
open water efficiency 𝜂𝑂 = .
optimum diameter 𝐷𝑂 = . m

The pitch–diameter ratio of the optimum propeller has to be found by interpolation

between the torque coefficient curves. A 10𝐾𝑄 curve is sketched through the self
propulsion point □ of the optimum propeller. Figure . reveals that the optimum
pitch–diameter ratio is 𝑃 ∕𝐷 = 0.9227. Table . summarizes the optimum propeller
characteristics based on the input from Table ..
In propeller design task  we match the propeller torque to an available torque. There is Next steps
no guarantee that the resulting optimum propeller produces sufficient thrust to achieve
the design speed. For that reason, the thrust generated by the optimum propeller has
to be checked against the thrust required by the ship in design tasks which use the
j delivered power as input. If the produced thrust is insufficient, the rate of revolution j
must be lowered or the available power must be increased.
In addition, the initial value for the expanded area ratio has to be checked against a
suitable cavitation limit. A higher expanded area ratio must be chosen if the limit is
exceeded. For the sake of brevity these steps are skipped here but we will perform them
in the examples of the following chapter.

48.2.2 Propeller design task 2

The second propeller design task uses the same input as task  except that the delivered Task 2, given:
power is swapped for the required service thrust 𝑇 . Since the final propeller produces 𝑇 , 𝑣𝐴 , 𝑛
wanted:
exactly the thrust a ship needs the aforementioned check on ship speed is obsolete. 𝐷, 𝑃 ∕𝐷, 𝜂𝑂
However, one must select an engine which must be capable of generating the required
torque at the selected rate of revolution.
Since the thrust is given, the design constant is based on the definition of the thrust Design constant
coefficient 𝐾𝑇 . The equation is divided by the fourth power of the advance ratio to for task 2
eliminate the unknown diameter 𝐷.
𝑇 𝑇
[ ]
𝐾𝑇 𝜌 𝑛2 𝐷4 𝜌 𝑛2 𝐷4 𝑇 𝑛2
= = = (.)
𝐽4 𝐽 4 4
𝑣𝐴 𝜌 𝑣𝐴4
𝑛4 𝐷4
We use the input data from task  again but exchange the delivered power 𝑃𝐷 in

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

578 48 Propeller Design Process

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
1.4 P/D =1 0.7
O curves optimum O =0.6070 4
=1.
[]

P / D
required thrust curve
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

[]
O at self propulsion points

auxiliary curve

O
1.0 0.5

open water efficiency

P/D =
0.8 0.4

[ J 4 ]P/D
KT 4

1.0
P/D =1.4 J

P/D =0
0.6 =1 .4 0.3
P/D =1.2

.8
P/D =1.0

P/D =
0.4 P /D
0.2
P/D =0.8 =1.2
optimu
m P/D

0.6
=0.906
P/D =

P/D =0.6 0
0.2 P/D =0.5 self propulsion point 0.1
0.5

P/D
of optimum
P/D = propeller =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
optimum J =0.614 6 advance coefficient J []

Figure 48.6 Design task 2 – Result: optimum propeller is defined by maximum of auxiliary
curve
j j

Table . for the estimated thrust for the service condition.

𝑇 = 1762.35 kN new input for task  (.)

The resultant design constant yields the curve of required thrust for the example data.
[ ]
𝐾𝑇
𝐾𝑇required = 𝐽 4 = 1.1755 𝐽 4 required thrust curve
𝐽4

Self propulsion The self propulsion points are now defined by the intersections of the available thrust
points for task 2 curve with the 𝐾𝑇 curves of the propellers in the open water diagram. Rather than
showing all intermediate steps of the process Figure . presents the final result. Note
that in contrast to Figure . from task  the solid lines are now the thrust coefficient
curves 𝐾𝑇 . The reader should complete the process as an exercise.
Although a consistent data set has been used for both tasks  and , it should not
surprise that results slightly differ between the tasks. Locating the maximum of the
auxiliary curve precisely is hard because the curve may be quite flat around the self
propulsion point. In addition, we are working with thrust and torque curves derived
from regression polynomials. They themselves have been produced by extensive fairing
of the available open water data (Oosterveld and van Oossanen, ). In our example,
efficiency is minimally higher for the propeller from task . The diameter is 1% larger
and the pitch–diameter ratio is 2% smaller (see Table .).

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.3 Optimum Rate of Revolution Selection 579

Table 48.4 Results for the propeller selection task 2 example

Optimum open water propeller data – Task 2

number of blades 𝑍=
expanded area ratio 𝐴𝐸 ∕𝐴0 = .
thrust (service) 𝑇 = . kN
rate of revolution 𝑛 = . rpm
advance ratio 𝐽 = .
pitch–diameter ratio 𝑃 ∕𝐷 = .
open water efficiency 𝜂𝑂 = .
optimum diameter 𝐷𝑂 = . m
required delivered power 𝑃𝐷 =  . kW

48.3 Optimum Rate of Revolution Selection

Propeller design tasks  and  mirror design tasks  and , respectively. However, now
the diameter 𝐷 instead of the rate of revolution is known. Consequently, the objective
is to find a combination of rate of revolution 𝑛 and pitch–diameter ratio 𝑃 ∕𝐷 which
provides the best efficiency 𝜂𝑂 .
An initial estimate of the diameter 𝐷 may be derived in two ways:
j j
• In early design stages the diameter may be estimated from regression formulas.
See Section ., page  and Equation (.) for some examples.

• Once preliminary hull lines have been drawn, the shape of the stern defines the
largest possible diameter which may be used as a starting point. Observe the
clearances! A minimum vertical clearance between propeller tip and hull surface
of % should be considered. Smaller vessels may have inclined shafts where the
propeller disk reaches below the baseline. However, larger commercial vessels
maintain a clearance of about % of the diameter to the baseline. It protects the
propeller and simplifies docking procedures.

48.3.1 Propeller design task 3

In propeller design task  we start again with the delivered power 𝑃𝐷 as representation Task 3, given:
of the available torque. The diameter 𝐷 = 7.430 m replaces the rate of revolution on 𝑃𝐷 , 𝑣𝐴 , 𝐷
wanted:
the list of required input. Otherwise the data is taken from Table .. 𝑛, 𝑃 ∕𝐷, 𝜂𝑂
Based on the torque coefficient Design constant
for task 3
𝑄𝑂 𝑃𝐷 𝜂𝑅
𝐾𝑄 = = (.)
𝜌 𝑛2 𝐷5 2𝜋 𝜌 𝑛3 𝐷5

we derive the design constant for task . The unknown rate of revolution 𝑛 is eliminated

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

580 48 Propeller Design Process

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
1.4 P/D =1.2 0.7
curves optimum O =0.6110 4
O
=1.
[]

P /D
available torque
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

[]
O at self propulsion points

auxiliary curve

O
10[ 3 ]J 3
1.0 0.5
KQ

open water efficiency

J

P/D =
0.8 0.4

1.0
P/D =1.4
optim

P/D =0
um P P /D
0.6 =1. 0.3
P/D =1.2 /D = 4
1.054
5

.8
P/D =1.0

P/D =
0.4 self propulsion point 0.2
P/D =0.8 P /D
of optimum propeller =1.2

0.6
P/D =

P/D =0.6
0.2 P/D =0.5 0.1
0.5

P/D
P/D = =1.0
P/D =0.5P/D =0.6 0.8
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
optimum J =0.684 0 advance coefficient J []

Figure 48.7 Design task 3 – Result: optimum propeller is defined by maximum of auxiliary curve

j j
by dividing Equation (.) above by 𝐽 3 .

𝑃𝐷
[ ]
𝐾𝑄 2𝜋 𝜌 𝑛3 𝐷5 𝑃𝐷 𝜂𝑅
= = (.)
𝐽3 𝑣𝐴3 2𝜋 𝜌 𝐷2 𝑣𝐴3
𝑛3 𝐷 3
The curve of available torque is represented by a cubic polynomial.
[ ]
𝐾𝑄
10 𝐾𝑄available = 10 𝐽 3 = 1.1866 𝐽 3 (.)
𝐽3

Self propulsion Figure . shows the polynomial (.) and the self propulsion points at the intersec-
points for task 3 tions with the open water torque 10𝐾𝑄 curves. The maximum achievable open water
efficiency marks the optimum advance ratio 𝐽 = 0.6851. This is about % higher than
in task  although the delivered power is the same as in task  and we used the resulting
diameter 𝐷 = 7.430 m of task  as input. The pitch–diameter ratio is also higher. As a
consequence of the higher advance ratio, the optimum rate of revolution 𝑛𝑂 is smaller
than the input to task .
𝑣𝐴 8.295 m/s
𝑛𝑂 = = = 1.6297 s−1 = 97.782 rpm (.)
𝐷𝐽 7.430 m ⋅ 0.6851
The resultant open water efficiency is % higher than the results from tasks  and .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.4 Design Charts 581

It should be noted that the maximum of the auxiliary curve depends on the interpolation
algorithm used to create the auxiliary curve. Obviously, there is an infinite number of
curves of different shape that could be fitted through the six given efficiency points
marked with crosses.

48.3.2 Propeller design task 4

Finally, in propeller design task  we know thrust 𝑇 , speed of advance 𝑣𝐴 , and diameter Task 4, given:
𝐷, and we are looking for the rate of revolution 𝑛𝑂 and pitch–diameter ratio which 𝑇 , 𝑣𝐴 , 𝐷
wanted:
provides the highest open water efficiency. We again employ the thrust 𝑇 = 1762.35 kN 𝑛, 𝑃 ∕𝐷, 𝜂𝑂
from task  as an example. The input diameter is again 𝐷 = 7.430 m.
The design constant for task  was already introduced when we discussed the prediction Design constant
of full scale power prediction (see Section .). for task 4

𝑇 𝑇
[ ]
𝐾𝑇 𝜌𝑛2 𝐷4 𝜌𝑛2 𝐷4 𝑇
= = = = 𝐶𝑆 (.)
𝐽2 𝐽2 𝑣𝐴2 𝜌 𝐷2 𝑣𝐴2
𝑛2 𝐷2
From the input data we derive the parabola of required thrust:
[ ]
𝐾𝑇
j 𝐾𝑇required = 𝐽 2 = 0.4522 𝐽 2 required thrust curve j
𝐽2

Figure . shows the resulting optimum propeller. Its characteristics are close to the
result of task .

optimum open water efficiency 𝜂𝑂 = 0.6114

advance ratio 𝐽 = 0.6808
optimum rate of revolution 𝑛𝑂 = 1.6369 s−1 = 98.214 rpm
required delivered power 𝑃𝐷 = 23210.89 kW

As for the other tasks, this result has to be corrected for the behind condition and
double checked against the cavitation limit. Details of this process will be explained in
the following chapter.

48.4 Design Charts

Selection of the optimum propeller characteristics based on open water charts is quite
cumbersome. Nevertheless, understanding and being able to follow the process de-
scribed above is a valuable skill. It enables basic checks of computational results and
can be applied to propellers for which design charts are not available.
Early on, naval architects converted available open water data into more practical 𝐵𝑃 - and
propeller design charts. The idea is to redraw the open water data as a function of the 𝐵𝑈 -charts
design constant. Easy to work with are the 𝐵𝑃 - and 𝐵𝑈 -charts. 𝐵𝑃 and 𝐵𝑈 refer to

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

582 48 Propeller Design Process

B-Series propeller B4-85

1.6 0.8
KT curves
10KQ curves
.2
P/D =1
O =0.6114
1.4 0.7
[−]

O curves optimum 4
/D =1.
required thrust curve P
thrust and torque coefficients KT , 10 KQ

1.2 self propulsion points 0.6

[−]
O at self propulsion points

auxiliary curve

O
1.0 0.5

open water efficiency

P/D =
0.8 0.4

1.0
P/D =1.4
P/D

P/D =0
0.6 =1. 0.3
P/D =1.2 4

[PJ 2 ]
KT

.8
P/D =1.0
optimu J2

P/D =
0.4 0.2
P/D =0.8 m P/D =1 /D =
.0452 1.2

0.6
P/D =
P/D =0.6
0.2 P/D =0.5 0.1
0.5

self propulsion point

P/D
ofP/D
optimum
=0.8 propeller =1.0
P/D =0.5P/D =0.6
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
optimum J =0.680 8 advance coefficient J [−]

Figure 48.8 Design task 4 – Result: optimum propeller is defined by maximum of auxiliary
curve
j j
Admiral D.W. Taylor’s propeller coefficients. The original coefficients are dimensional
and their values depend on the system of units employed. They should no longer be
used. Like open water diagrams, design charts show data for a set of fixed number of
blades and expanded area ratio. Figure . shows a 𝐵𝑃1 -chart for task  with the data
for Wageningen B-series propellers B-. The chart has been simplified by omitting
intermediate curves.
Propeller design charts are plotted into an axes system of task specific design constant
and pitch–diameter ratio. Each task therefore has a specific type of chart. Beware that
the horizontal axis uses the fourth root of the design constant! For task  we use:
√[ ] ( )0.25
𝐾𝑄 𝑃𝐷 𝑛2 𝜂𝑅
(.)
4
=
𝐽5 2𝜋 𝜌 𝑣𝐴5
As shown in Figure ., this stretches the horizontal axis and makes it easier to read.
𝐵𝑃 - and 𝐵𝑈 -propeller design charts consist of three distinct sets of curves (Figure .):

(i) curves of constant open water efficiency 𝜂𝑂 = const.,

(ii) curves of constant inverse advance ratio 1∕𝐽 = const., and
(iii) optimum propeller curves.

𝐵𝑃1 -charts solve design task  and 𝐵𝑃2 -charts solve design task . If the thrust is
provided, 𝐵𝑈 -charts are used: 𝐵𝑈1 -charts for design task  and 𝐵𝑈2 -charts for design
task .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.4 Design Charts 583

j j
Figure 48.9 Simplified task 1 propeller design 𝐵𝑃1 -chart for Wageningen B-Series propeller with 𝑍 = 4 and
𝐴𝐸 ∕𝐴𝑜 = 0.85

The highly curved open water efficiency isolines may be interpreted like the lines of Open water
constant altitude on a map. Open water efficiency changes faster where the isolines are efficiency
isolines
closer together and it changes slower where the isolines are farther apart. The open
water efficiency forms a ‘ridge’ that curves from the top left corner to the lower right
corner of the chart. At the end of a selection process we want to end up somewhere
close to the crest of the ridge.
The lines of constant inverse advance ratio 1∕𝐽 appear like a hatching pattern roughly Inverse advance
tracing from the lower left toward the upper right on the chart. In contrast to the open ratio isolines
water efficiency, values of 1∕𝐽 = const. increase from the upper left to the lower right
corner of the chart.
With open water diagrams we found the maximum open water efficiency as the highest Optimum open
point on the auxiliary curve. In Figure . the lower, solid optimum propeller curves water propeller
curve
reflect the pitch–diameter ratios with the maximum open water efficiency 𝜂𝑂,opt . The
‘𝑃 ∕𝐷 for 𝜂𝑂,opt ’-curve marks the points where the open water efficiency isolines have
vertical tangents. Any vertical line in the chart corresponds to a specific design con-
stant value. Deviation from the 𝑃 ∕𝐷 for 𝜂𝑂,opt curve upwards or downwards along
√
4
𝐾𝑄 ∕𝐽 5 = const. causes a decline in open water efficiency. The decline is, however,
much smaller for increases in 𝑃 ∕𝐷-ratio than for lower 𝑃 ∕𝐷-ratios.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

584 48 Propeller Design Process

j j

Figure 48.10 Optimum diameter chart for design task 1

Optimum Many published propeller design charts only feature the curve for the optimum ef-
propeller curves ficiency 𝜂𝑂,opt . However, as mentioned before, this line identifies the best propeller
under open water conditions. It does not necessarily mean that this is the best propeller
for the behind condition with a nonuniform wake field. Design experience shows that
the optimum propeller in the behind condition operates at a reduced 1∕𝐽 -ratio. In
fact, diameter or rate of revolution of the optimum propeller are 1% to 5% smaller for
a nonuniform wake field compared with the optimum propeller for the open water
condition. The additional optimum propeller curves (dash-dotted lines) represent the
proper pitch–diameter ratio for reductions in 1∕𝐽 from 1% to 5%.
Other propeller Even simpler than the 𝐵𝑃 - and 𝐵𝑈 -charts are plots of the data along the optimum
design charts open water efficiency curve. Figure . shows curves for three expanded area ratios.
𝐵𝑃 -charts like Figure . contain data for just one expanded area ratio. However, in
many cases 𝐴𝐸 ∕𝐴0 is adjusted during the selection process according to the chosen
cavitation criterion. The middle curve of each triplet represents the same data as the
𝐵𝑝 -chart in Figure .. Two reports by Bernitsas and Ray (a,b) contain a complete
set of design charts like Figure ..
Harvald () explains the use of logarithmic propeller design charts. Their advantage
is that all four tasks may be solved with a single design chart. They are, however, fairly
complex and data take-off is more difficult compared with the charts discussed above.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

48.5 Computational Tools 585

48.5 Computational Tools

The propeller design charts above have been derived from the thrust and torque coeffi-
cient polynomials of the Wageningen B-Series propellers. It is fairly straight forward
to include the polynomials in an optimization tool which solves the selection process
computationally rather than with charts. For design task , one could minimize the
function ( )
𝑓 𝑃 ∕𝐷, [𝐾𝑄 ∕𝐽 5 ] = 1 − 𝜂𝑂 (𝐽𝑄𝑆 , 𝑃 ∕𝐷, 𝐴𝐸 ∕𝐴0 , 𝑍) (.)
The pitch–diameter ratio serves as a free variable in the optimization process and the
design constant [𝐾𝑄 ∕𝐽 5 ] is a parameter which defines the self propulsion point. The
open water efficiency is computed from the 𝐾𝑇 - and 𝐾𝑄 -polynomials (.) and (.)
𝐽𝑄𝑆 𝐾𝑇 (𝐽𝑄𝑆 , 𝑃 ∕𝐷, 𝐴𝐸 ∕𝐴0 , 𝑍)
𝜂𝑂 = (.)
2 𝜋 𝐾𝑄 (𝐽𝑄𝑆 , 𝑃 ∕𝐷, 𝐴𝐸 ∕𝐴0 , 𝑍)

with the advance ratio 𝐽𝑄𝑆 as the solution of the implicit and nonlinear equilibrium
condition for the self propulsion point
[ ]
𝐾𝑄
0 = 5
𝐽𝑄𝑆 − 𝐾𝑄 (𝐽𝑄𝑆 , 𝑃 ∕𝐷, 𝐴𝐸 ∕𝐴0 , 𝑍) (.)
𝐽5
The process is similar for the other design tasks.

j The result of the propeller selection based on design charts provides the necessary data j
to complete the power prediction for a ship design. The result also forms the basis for
further computations. At a minimum, a propeller lifting line code should be used for
the final design. It helps determining the following details of the propeller design:

• Defines the circulation distribution Γ which in turn determines section lift coeffi-
cients and section camber.
• Adjust pitch distribution to account for axial and tangential wake distribution (if
available) and unload the extreme ends of the propeller blades to reduce tip and
hub vortices.
• Optimize chord length distribution to reduce cavitation issues.

Lifting line codes represent the simplest numerical approximation of hydrodynamic

propeller characteristics. Unfortunately, this topic extends beyond the scope of this
book. Interested readers should take a look at the OpenProp lifting line code (http:
//engineering.dartmouth.edu/epps/openprop/, last visited //) and the asso-
ciated documentation (Epps, a,b). Further analysis with lifting surface and panel
codes may be required if cavitation inception and sound signature are of interest. For
details see Breslin and Andersen () and Kerwin and Hadler ().

References
Bernitsas, M. and Ray, D. (a). Optimal diameter B-Series propellers. Technical
Report , The University of Michigan, Ann Arbor, USA.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

586 48 Propeller Design Process

Bernitsas, M. and Ray, D. (b). Optimal revolution B-Series propellers. Technical

Report , The University of Michigan, Ann Arbor, USA.
Breslin, J. and Andersen, P. (). Hydrodynamics of ship propellers. Cambridge
University Press, Cambridge, United Kingdom.
Epps, B. (a). An impulse framework for hydrodynamic force analysis: fish propulsion,
water entry of spheres, and marine propellers. PhD thesis, Massachusetts Institute of
Technology (MIT), Cambridge, MA.
Epps, B. (b). OpenProp v. theory document. Technical report, Department of
Mechanical Engineering, Massachusetts Institute of Technology (MIT), Cambridge,
MA.
Harvald, S. (). Resistance and propulsion of ships. John Wiley & Sons, New York,
NY.
ITTC (). Predicting powering margins. International Towing Tank Conference,
Recommended Procedures and Guidelines .---.. Revision .
Kerwin, J. and Hadler, J. (). Propulsion. The Principles of Naval Architecture Series.
Society of Naval Architects and Marine Engineers (SNAME), Jersey City, NJ.
Oosterveld, M. and van Oossanen, P. (). Further computer-analyzed data of the
Wageningen B-screw series. International Shipbuilding Progress, :–.
Stasiak, J. (). Service margin – solution of the problem or a problem waiting for
solution? Polish Maritime Research, (Special issue):–.

j j

Self Study Problems

. State input and output for all propeller design tasks.
. Derive the design constant for all propeller design tasks.
. The engine chosen for a ship generates a delivered power of 𝑃𝐷 = 22 000 kW at a
propeller rate of revolution 𝑛 = 99 min−1 . Use the open water diagram for B-
to estimate the optimum propeller diameter 𝐷, pitch–diameter ratio 𝑃 ∕𝐷, and
open water efficiency 𝜂𝑂 when the number of blades is 𝑍 = 4 and the expanded
area ratio is 𝐴𝐸 ∕𝐴0 = 0.7. You can assume the propeller produces sufficient
thrust to sail at a speed of 22.1 kn with a wake fraction of 𝑤 = 0.271 and relative
rotative efficiency 𝜂𝑅 . Density of water is 𝜌 = 1026.021 kg/m3 and gravitational
acceleration is 𝑔 = 9.807 m/s2 . Do not forget to convert everything to consistent
units and that the open water diagram shows 10𝐾𝑄 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

587

49
Hull–Propeller Matching Examples

In this chapter we determine basic propeller characteristics using propeller design

charts and a cavitation criterion. Two examples are presented:

• solution of the optimum rate of revolution problem for a container ship (fourth
propeller design task)
• solution of the optimum diameter problem for a bulk carrier (first propeller
design task)

The examples make use of propeller design charts which have been redrawn on the
basis of the Wageningen B-series polynomials (.) and (.) by Oosterveld and van
j Oossanen (). The 5% back cavitation limit is used as the cavitation criterion. j

Learning Objectives
At the end of this chapter students will be able to:

• set up and solve propeller design tasks

• apply criteria for cavitation prevention
• find the optimum propeller for a ship
• predict propulsive power requirements

49.1 Optimum Rate of Revolution Problem

A performance estimate for a container ship yields the basic data listed in Table ..
We aim for a slow running diesel engine as the main propulsor. An early design formula
for container ships by Kristensen and Lützen () is used to estimate the maximum
propeller diameter.
𝐷 = 0.623 𝑇max − 0.16 m (.)
With the data from Table . this yields a propeller diameter of

𝐷 = 0.623 ⋅ 12.5m − 0.16 m = 7.6275 m (.)

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

588 49 Hull–Propeller Matching Examples

Table 49.1 Optimum rate of revolution problem – example input data for a container ship

Data for design task 4

length in waterline 𝐿𝑊𝐿 = . m
beam 𝐵 = . m
draft 𝑇 = . m
maximum draft 𝑇max = . m
wetted surface 𝑆 = . m2
design speed 𝑣𝑆 = . kn
calm water resistance (design speed) 𝑅𝑇 = . kN
requested service margin Δ𝑠 = %
nonuniform wake adjustment Δ𝛿 = %
relative rotative efficiency 𝜂𝑅 = .
wake fraction 𝑤 = .
thrust deduction fraction 𝑡 = .
number of blades 𝑍=
propeller diameter 𝐷 = . m
shaft submergence ℎ0 = . m
air pressure 𝑝𝐴 = . Pa
vapor pressure (salt water,  ◦ C) 𝑝𝑣 = . Pa
j density (salt water,  ◦ C) 𝜌𝑆 = . kg/m3 j
gravitational acceleration 𝑔 = . m/s2

This diameter should be checked against a preliminary lines plan to ensure that propeller
clearances and shaft submergence are sufficient.

49.1.1 Design constant

Identifying the Our first step is to define the design task by analyzing the given data. We are given ship
design task speed, resistance, hull–propeller interaction parameters, and a maximum propeller
diameter. Resistance at design speed, service margin, and thrust deduction fraction
allow us to calculate the required thrust. From ship speed and wake fraction follows
the speed of advance of the propeller. Summarizing, we know:

𝑅𝑇service
• required thrust 𝑇req =
(1 − 𝑡)
• speed of advance 𝑣𝐴 = (1 − 𝑤)𝑣𝑆 , and

• propeller diameter 𝐷.

This indicates we have to solve propeller design task . We must select a sufficiently
high expanded blade area ratio 𝐴𝐸 ∕𝐴0 to avoid cavitation and find optimum values for
rate of revolution 𝑛, pitch–diameter ratio 𝑃 ∕𝐷, and the open water efficiency 𝜂𝑂 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.1 Optimum Rate of Revolution Problem 589

For any design task it is important to note that the propeller should be optimized for a Service condition
realistic service condition and not for the ideal calm water conditions at the ship’s trial.
The ship contract usually specifies a service margin. It depends on the desired service
area. Small margins (∼ 5%) are used for sheltered waterways, large margins for oceans
with severe weather like, for instance, the North Atlantic in winter (suggested margin
∼ 35%; Harvald, ).
With the service margin given in Table ., the resistance for service condition is Resistance in
( ) ( ) service condition
𝑅𝑇service = 1 + Δ𝑠 𝑅𝑇 = 1 + 0.15 1149.06 kN (.)
= 1321.419 kN

Based on this in service resistance we determine the required thrust:

𝑅𝑇service 1321.419 kN
𝑇req = = ( ) (.)
(1 − 𝑡) 1 − 0.190
= 1631.3815 kN

For the speed of advance we obtain

m
( 1852 M
𝑣𝐴 = (1 − 𝑤)𝑣𝑆 = 1 − 0.2865) 20.5 kn ⋅
3600 hs
m
j = 0.7135 ⋅ 10.5461 (.) j
s
m
= 7.5247
s

The design constant for task  is the shortened thrust loading coefficient which we Design constant
employed before in the ITTC  performance prediction method.
𝐾𝑇 𝑇req
𝐶𝑆 = = (.)
𝐽2 𝜌 𝑣2𝐴 𝐷2
1631381.5 N
= ( )2
kg
1026.021 m3 ⋅ 7.62752 m2 7.5247 ms
= 0.482683
Do not forget to convert everything into a consistent set of units. In this case, the thrust
has to be converted from kilo Newton [kN] into Newton and the speed of advance has
to be expressed in m/s.

49.1.2 Initial expanded area ratio

In order to select an appropriate design chart we have to estimate an initial value for Keller’s formula
the required expanded area ratio. Keller’s formula can be used for this.
( ) ( )
𝐴𝐸 1.3 + 0.3 𝑍 𝑇req
= ( ) +𝐾 Keller’s formula (.)
𝐴0 req 𝑝0 − 𝑝𝑣 𝐷 2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

590 49 Hull–Propeller Matching Examples

The parameter 𝐾 in Keller’s formula (.) is not the form factor but a constant selected
based on ship type. The value 𝐾 = 0.2 is used for single screw merchant vessels. The
total static pressure 𝑝0 is taken at the propeller shaft level. It consists of atmospheric
pressure and the hydrostatic pressure.
𝑝0 = 𝑝𝐴 + 𝜌 𝑔 ℎ0 (.)
= 101325 Pa + 1026.021 ⋅ 9.807 ⋅ 6.6 Pa
= 167735.44 Pa
Note that the atmospheric pressure represents the larger share of the total static pressure
and cannot be ignored.
Substituting all values into Keller’s formula (.) yields
( ) ( )
𝐴𝐸 1.3 + 0.3 ⋅ 5 1631381.5 N
= ( ) + 0.2
𝐴0 req 167735.44 Pa − 1671 Pa 7.62752 m2
= 0.6729 (.)

We will start with a design chart for the expanded blade area ratio 𝐴𝐸 ∕𝐴0 = 0.60 (see
Figure .).

49.1.3 First iteration

j j
Design chart The 𝐵𝑢2 -chart (Figure .) used here has been calculated and redrawn on the basis of
selection the Wageningen B-Series polynomials by Oosterveld and van Oossanen (). Original
charts can be obtained from MARIN (http://www.marin.nl).
Design chart We draw a vertical line for the value of the fourth root of our design constant 𝐶𝑆
application Equation (.).
[ ]1
√ 𝐾𝑇 4
4
𝐶𝑆 = = 0.833519 (.)
𝐽2
The intersection of the vertical line and the curve of the optimum efficiency 𝜂𝑂opt marks
the best achievable self propulsion point for open water conditions. We read off the
following values:

open water efficiency 𝜂𝑂 = 0.6244

pitch–diameter ratio (𝑃 ∕𝐷)𝑂 = 1.0314
reciprocal advance coefficient 𝛿𝑂 = 𝐽1 = 1.4809
𝑂

A value of 𝛿𝑂 = 1∕𝐽𝑂 = 1.4809 is equivalent to an advance coefficient of

1
𝐽𝑂 = = 0.6753 (.)
𝛿𝑂
This, in turn, provides the optimum rate of revolution in open water conditions (uniform
wake):
𝑣𝐴 7.5247 ms
𝑛𝑂 = = = 1.4609 s−1 (.)
𝐷 𝐽𝑂 7.6275 m ⋅ 0.6753

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.1 Optimum Rate of Revolution Problem 591

j j

Figure 49.1 Propeller design 𝐵𝑢2 -chart for Wageningen B-series propeller B5-60. Calculated and plotted based on
the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and van Oossanen (1975)

The thrust coefficient for the open water propeller will be

𝐾𝑇 = 𝐶𝑆 𝐽𝑂2 = 0.482683 ⋅ 0.67532 = 0.2201 (.)

Of course, it produces the required thrust used as input. However, we are not quite
done.
Experience shows that the optimum propeller for the behind condition with a nonuni- Adjustment for
form wake works at a slightly higher advance coefficient. The value of 𝛿 = 1∕𝐽 is behind condition
commonly reduced by 1% to 5% (Lewis, ). Vessels with fine lines or small propeller
loading have small reductions, vessels with high block coefficient or high propeller
loading have larger reductions in 1∕𝐽 . For design tasks where the rate of revolution 𝑛 is
given, a reduction in 1∕𝐽 leads to a lower diameter 𝐷. In our case the diameter 𝐷 is
given. Thus, the rate of revolution 𝑛 is reduced instead.
The container ship under consideration is sailing reasonably fast and will have normal Advance
lines. A medium reduction of Δ𝛿 = 3% is chosen for this example. The new 1∕𝐽 value coefficient for
nonuniform
for the behind condition is wake

1
𝛿 = (1 − Δ𝛿) = 0.97 ⋅ 1.4809 = 1.4364 (.)
𝐽𝑂

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

592 49 Hull–Propeller Matching Examples

This is equivalent to a reduction in the rate of revolution for the behind condition to
( )
𝑛 = 1 − Δ𝛿 𝑛𝑂 = 1.4171 s−1 (.)
Reducing the rate of revolution for constant speed of advance and constant diameter is
equivalent to an increase in advance coefficient 𝐽 .

𝑣𝐴 7.5247 ms
𝐽 = = = 0.6962 (.)
𝑛𝐷 1.4171 s−1 ⋅ 7.6275 m

Adjustment of However, the thrust of a propeller is decreasing with increasing advance coefficient
pitch–diameter 𝐽 . In fact, calculating the thrust coefficient 𝐾𝑇 for a B-series propeller with 𝑍 = 5,
ratio
𝑃 ∕𝐷 = 1.0314, 𝐴𝐸 ∕𝐴0 = 0.60, and 𝐽 = 0.6962 yields only 𝐾𝑇 = 0.2103, which is
lower than the required value of .. Together with the reduced rate of revolution,
this would cause a significant loss in thrust (𝑇 = 1466.75 kN). As a consequence, the
ship would not reach the desired speed.
To compensate for the loss in thrust caused by the reduction of rate of revolution, the
propeller pitch has to be increased. The optimum pitch–diameter ratio for the best self
propulsion
√ point in behind condition can be taken off the 𝐵𝑢2 -chart at the intersection
of the 4 𝐶𝑆 line and the (−3%) 1∕𝐽opt curve plotted above the optimum open water
efficiency curve. For this example we get (see Figure .):
𝑃
= 1.0777 adapted for nonuniform wake (.)
𝐷
j j
Checking with the open water data for the new propeller, we get a thrust coefficient
of 𝐾𝑇 = 0.2339 with a propeller thrust of 𝑇prop = 1631.382 kN, which is equal to the
required thrust.
Adjustment of Plotting the √current propeller data into the 𝐵𝑢2 -chart, we moved upward along the
open water vertical line 4 𝐶𝑆 = 0.833519 and away from the optimum efficiency. However, the
efficiency √
slope of the efficiency curves is small relative to the line of constant 𝐾𝑇 ∕𝐽 2 . As a
consequence, the reduction in open water efficiency will be small compared with the
case without 1∕𝐽 reduction:
𝑃 ( )
𝜂𝑂 = 0.6239 for = 1.0777 and 𝐽 = 0.6962 𝛿 = 1.4364
𝐷
In addition, this reduces the danger of losing efficiency very quickly if the speed of
advance is further reduced due to fouling of the hull and the operating point falls below
the optimum efficiency line.
First iteration In summary, our first iteration results in an optimum propeller with the following
result characteristics:

Result of 1. iteration
rate of revolution 𝑛 = 1.4171 s−1
open water efficiency 𝜂𝑂 = 0.6239
diameter 𝐷 = 7.6275 m
pitch–diameter ratio 𝑃 ∕𝐷 = 1.0777
advance coefficient 𝐽 = 0.6962
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.60

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.1 Optimum Rate of Revolution Problem 593

49.1.4 Cavitation check for first iteration

We now have to check whether the initial guess of the expanded blade area ratio Burrill cavitation
𝐴𝐸 ∕𝐴0 = 0.60 is sufficient to keep the propeller free from excessive cavitation. For chart
this task we employ the empirical cavitation charts by Burrill and Emerson ().
Figure . shows a replication of this chart based on regression curves (see Chapter ).
The curves represent maximum thrust loading coefficients 𝜏𝑐 for different amounts
of cavitation on the blade. We will use the curve labeled 5% back cavitation which is
similar to the suggested upper limit for merchant ship propellers (see Table .). For
programming and calculation purposes, it can be expressed as the following regression
curve.

𝜏𝑐 = 0.715 𝜎𝑏0.184 − 0.437 % back cavitation limit curve (.)

The propeller cavitation number 𝜎𝑏 is computed using the total static pressure at the Static pressure
depth of the propeller shaft and a simplified flow velocity 𝑣1 for the blade section at and flow velocity
radius 𝑟 = 0.7𝑅. The total static pressure is equal to the atmospheric pressure 𝑝𝐴 plus
the hydrostatic pressure at the propeller shaft.

𝑝0 = 𝑝𝐴 + 𝜌 𝑔 ℎ0 (.)

With a standard atmospheric pressure of 𝑝𝐴 = 101325 Pa and a shaft submergence of

ℎ0 = 6.6 m, we get
j j
𝑝0 = 101325 Pa + 1026.021 kg/m3 9.807 m/s2 6.6 m
= 167735.44 Pa

The simplified flow velocity 𝑣1 is

√
𝑣1 = 𝑣2𝐴 + (0.7 𝜋 𝑛 𝐷)2 (.)
√
( )2
= 7.52472 (m/s)2 + 0.7 𝜋 1.4171 s−1 7.6275 m
= 24.9322 m/s

With this input we can compute the propeller cavitation number Propeller
cavitation
𝑝0 − 𝑝𝑣 number
𝜎𝑏 = (.)
1
2
𝜌 𝑣21

The vapor pressure for seawater at 15 ◦ C is 𝑝𝑣 = 1671 Pa (ITTC, ). Therefore, we

get
167735.44 Pa − 1671 Pa
𝜎𝑏 =
1
1026.021 kg/m3 24.93222 (m/s)2
2
= 0.5207 (.)

Based on the cavitation number 𝜎𝑏 = 0.5207, a thrust loading coefficient of 𝜏𝑐 = 0.1971 Required
projected area

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

594 49 Hull–Propeller Matching Examples

is computed with Equation (.). The thrust loading coefficient is connected to the
minimum required projected area 𝐴𝑃 of the propeller by
𝑇prop
𝐴𝑃req = (.)
1 2
𝜌𝑣 𝜏
2 1 𝑐
Projected area may be converted into developed area 𝐴𝐷 using a formula by Admiral
David W. Taylor.
𝐴𝑃
𝐴𝐷 = ( ) (.)
𝑃
1.067 − 0.229
𝐷
or
( )
𝑃
𝐴𝑃 = 1.067 − 0.229 𝐴𝐷 (.)
𝐷

For propellers with zero or moderate rake, expanded blade area is equal to the developed
area.
𝐴𝐸 ≈ 𝐴𝐷 (.)
Required Substituting these two results into Equation (.) and dividing both sides by the
j expanded area propeller disk area 𝐴0 = 𝜋𝐷2 ∕4 yields j
ratio
( )
𝐴𝐸 𝑇prop
= ( ) (.)
𝐴0 req 1 2 𝑃 𝜋𝐷2
𝜌 𝑣 𝜏 1.067 − 0.229
2 1 𝑐 𝐷 4
With the data for the propeller derived in the first iteration, the Burrill criterion asks
for a required minimum expanded area ratio of
( )
𝐴𝐸
= 0.6926 (.)
𝐴0 req

to prevent excessive cavitation. This is more than the value from our first design chart
(𝐴𝐸 ∕𝐴0 = 0.6). We will have to perform a second iteration using a higher expanded
area ratio.

49.1.5 Second iteration

Design chart In the second iteration, we make use of the 𝐵𝑢2 -chart for an expanded blade area ratio
application of 𝐴𝐸 ∕𝐴0 = 0.75. The design constant 𝐶𝑆 remains unchanged because our thrust
requirements for the ship have not changed. Performing the same steps as above:

• read off values for 𝜂𝑂 , (𝑃 ∕𝐷)𝑂 , and 1∕𝐽𝑂

• calculate diameter
• reduce diameter and derive new pitch

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.1 Optimum Rate of Revolution Problem 595

j j

Figure 49.2 Propeller design 𝐵𝑢2 -chart for Wageningen B-series propeller B5-75. Calculated and plotted based on
the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and van Oossanen (1975)

yields the following set of propeller parameters:

open water efficiency 𝜂𝑂 = 0.6224

pitch–diameter ratio (𝑃 ∕𝐷)𝑂 = 1.0348
reciprocal advance coefficient 𝛿𝑂 = 𝐽1 = 1.4787
𝑂

Reduction of 1∕𝐽 by % yields the following optimum propeller data for the behind Second iteration
condition: result

Result of 2. iteration
rate of revolution 𝑛 = 1.4150 s−1
open water efficiency 𝜂𝑂 = 0.6219
diameter 𝐷 = 7.6275 m
pitch–diameter ratio 𝑃 ∕𝐷 = 1.0801
advance coefficient 𝐽 = 0.6972
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.75

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

596 49 Hull–Propeller Matching Examples

Figure 49.3 Auxiliary plot to determine Auxiliary plot for expanded area ratio AE /A0
the expanded area ratio of the final
(AE /A0 )chart design chart value
optimum propeller 2
(AE /A0 )req 5% cavitation limit

AE /A0 =0.6933 for

iteration
(AE /A0 )chart =( AE /A0 )req

0.600 0.625 0.650 0.675 0.700 0.725 0.750

expanded area ratio AE /A0 []

This propeller produces the required thrust of 𝑇prop = 1631.3815 kN. A renewed
cavitation check with Burrill’s chart, i.e. Equation (.), yields a required expanded
area ratio of ( )
𝐴𝐸
= 0.6938 (.)
𝐴0 req
j This is less than the chart value of 𝐴𝐸 ∕𝐴0 = 0.75 we used. A lower expanded area ratio j
typically results in higher efficiency because the reduced blade surfaces have smaller
frictional losses.

49.1.6 Final selection by interpolation

We find the final propeller characteristics by interpolating between the results of the
first and second iterations.
Expanded area First, the expanded area ratio is found by comparing the expanded area ratios we used
ratio (from the charts) and the required expanded area ratios according to Equation (.).
We draw an auxiliary diagram which plots the actually used expanded area ratios and the
required expanded area ratios for the two iterations (Figure .). Actual and required
expanded blade area ratios are equal where the two lines intersect.
( ) ( )
𝐴𝐸 𝐴𝐸
= (.)
𝐴0 chart 𝐴0 req

In our example, the intersection point yields

( )
𝐴𝐸
= 0.6933 (.)
𝐴0

Pitch–diameter For the final optimum propeller characteristics, we plot straight lines connecting the
and advance results of the first and second iterations for pitch–diameter ratio 𝑃 ∕𝐷, advance coeffi-
coefficient
cient 𝐽 , open water efficiency 𝜂𝑂 , and rate of revolution 𝑛 as a function of the expanded

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.1 Optimum Rate of Revolution Problem 597

Auxiliary plot for pitchdia meter ratio P/D Auxiliary plot for advance ratio J
1.0805 0.6972
[]

1.0800 0.6970

[]
J =0.6968
P/D

advance coefficient J
1.0795 0.6968

AE /A0 =0.6933

AE /A0 =0.6933
P/D =1.0792
pitchdiameter ratio

1.0790 0.6966

1.0785 0.6964

1.0780 0.6962

1.0775 0.6960
0.600 0.625 0.650 0.675 0.700 0.725 0.750 0.600 0.625 0.650 0.675 0.700 0.725 0.750

Auxiliary plot for open water efficiency O Auxiliary plot for rate of revolution nS
0.6240 1.4175

[s 1 ]
[]

0.6235 1.4170
O

AE /A0 =0.6933

AE /A0 =0.6933
rate of revolution nS

0.6230 1.4165
open water efficiency

O =0.6226

0.6225 1.4160 1
nS =1.4158s

0.6220 1.4155

j 0.6215 1.4150
j
0.600 0.625 0.650 0.675 0.700 0.725 0.750 0.600 0.625 0.650 0.675 0.700 0.725 0.750
expanded area ratio AE /A0 [] expanded area ratio AE /A0 []

Figure 49.4 Auxiliary plots to determine final optimum propeller characteristics. The ⋆ mark results of first and
second iterations

blade area ratio (Figure .). The graph for the rate of revolution is not really necessary
as 𝑛 could be recomputed from the final advance coefficient 𝐽 via Equation (.).
Final values are read off the plots or found by computing the intersection with the Final selection
vertical line for 𝐴𝐸 ∕𝐴0 = 0.6933.

Result of final selection

rate of revolution 𝑛 = 1.4158 s−1
open water efficiency 𝜂𝑂 = 0.6226
diameter 𝐷 = 7.6275 m
pitch–diameter ratio 𝑃 ∕𝐷 = 1.0792
advance coefficient 𝐽 = 0.6968
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.6933

This propeller provides the required thrust. Checking with the Burrill chart for a third
time yields a required expanded area ratio of 𝐴𝐸 ∕𝐴0 = 0.6937 which confirms our final
selection.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

598 49 Hull–Propeller Matching Examples

Table 49.2 Optimum diameter problem – example input data for a bulk carrier

Data for design task 1

length in waterline 𝐿𝑊𝐿 = . m
beam 𝐵 = . m
draft 𝑇 = . m
maximum draft 𝑇max = . m
wetted surface 𝑆 = . m2
design speed 𝑣𝑆 = . kn
requested service margin Δ𝑠 = %
nonuniform wake adjustment Δ𝛿 = %
brake power of engine 𝑃𝐵 = . kW
target rate of revolution 𝑛 = . rpm
shafting efficiency 𝜂𝑆 = .
relative rotative efficiency 𝜂𝑅 = .
wake fraction 𝑤 = .
thrust deduction fraction 𝑡 = .
number of blades 𝑍 =
shaft submergence ℎ0 = . m
air pressure 𝑝𝐴 = . Pa
j vapor pressure 𝑝𝑣 = . Pa j
density of salt water at  ◦ C 𝜌𝑆 = . kg/m3
gravitational acceleration 𝑔 = . m/s2

Summary Based on this preliminary propeller selection, the propulsive power estimate may be
completed. Typically, wake fraction and thrust deduction fraction depend on some of
the propeller characteristics. If the final estimates deviate from the data used for the
propeller selection (see Table .), the propeller selection process must be repeated
until a converged and consistent solution is obtained. In later design stages the propeller
selection should be refined by employing a higher fidelity computational method, for
instance a lifting line calculation which would allow to adapt the pitch distribution to
the estimated, measured, or computed wake field of the hull. At this point Reynolds
number effects may be considered as well.

49.2 Optimum Diameter Problem

Using the Guldhammer and Harvald method, a performance estimate has been com-
pleted for a handy size bulk carrier. Table . lists the basic ship data and Table .
summarizes the resistance estimate in the range of the design speed.
Owner specifies The future owner of the bulk carrier already has similar ships in his or her fleet and
engine wants the new design to use the same main engine. Therefore, brake power and rate of
revolution are used as input.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 599

49.2.1 Design constant

Our first step is to define the design task by analyzing the given data. We are given speed, Identify Design
brake power, rate of revolution, and hull-propeller interaction parameters. Based on Task
brake power and shaft efficiency we can readily compute the available delivered power.
The resistance, service margin, and thrust deduction fraction allow us to calculate the
required thrust. From ship speed and wake fraction follows the speed of advance of the
propeller. In summary, we know:

• delivered power 𝑃𝐷 = 𝑃𝐵 𝜂𝑆

• speed of advance 𝑣𝐴 = (1 − 𝑤)𝑣𝑆 , and

• rate of revolution 𝑛.

This indicates we have to solve propeller design task . We must select a sufficiently
high expanded blade area ratio 𝐴𝐸 ∕𝐴0 to avoid cavitation and find optimum values for
the propeller diameter 𝐷, the pitch–diameter ratio 𝑃 ∕𝐷, and the open water efficiency
𝜂𝑂 .
For any design task it is important to note that the propeller should be optimized for a
realistic service condition and not for the ideal calm water conditions at the ship’s trial.
The service margin specified for this design is Δ𝑠 = 20%.
j With the data given in Table ., we obtain the available delivered power j
𝑃𝐷 = 𝑃𝐵 𝜂𝑆 = 7650 kW ⋅ 0.98 = 7497.00 kW (.)

The speed of advance is:

𝑣𝐴 = (1 − 𝑤)𝑣𝑆 (.)
m
( 1852 M
= 1 − 0.3413) 14.5 kn ⋅
3600 hs
m
= 0.6587 ⋅ 7.4594
s
m
= 4.9135
s

The design constant for task  is Design constant

[ ]
𝐾𝑄 𝑃𝐷 𝑛2 𝜂𝑅
= (.)
𝐽5 2𝜋 𝜌 𝑣5𝐴

7497000 W ⋅ 1.93332 s−2 ⋅ 1.0

=
kg ( )5
2𝜋 1026.021 m3 4.9135 ms
= 1.517729

Do not forget to convert everything into a consistent set of units. In this case, the
delivered power has to be converted from kilo Watt (kW) into Watt and the rate of
revolution has to be expressed in 1∕s.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

600 49 Hull–Propeller Matching Examples

Table 49.3 Resistance estimate for bulk carrier Resistance estimate

from Table 49.2 𝑣𝑆 𝑅𝑇 (trial)
[kn] [kN]
13.5 427.54
14.0 468.92
14.5 514.62
15.0 565.36
15.5 623.51

49.2.2 Initial expanded area ratio

Keller’s formula In order to select an appropriate design chart, we have to estimate an initial value for
the minimum expanded blade area ratio. Keller’s formula can be used for this. The
parameter 𝐾 in Keller’s formula Equation (.) is not the form factor but a constant
selected based on ship type. The value 𝐾 = 0.2 is used for single propeller merchant
vessels. The total static pressure 𝑝0 is taken at the propeller shaft level. It consists of
atmospheric pressure and the hydrostatic pressure.

𝑝0 = 𝑝𝐴 + 𝜌 𝑔 ℎ0 (.)
= 101325 Pa + 1026.021 ⋅ 9.807 ⋅ 7.0 Pa
= 171760.32 Pa
j j
The required thrust 𝑇 at design speed is estimated using the total resistance for trial
conditions 𝑅𝑇 , the service margin Δ𝑠, and the thrust deduction fraction 𝑡.

(1 + Δ𝑠)𝑅𝑇
𝑇req = (.)
1−𝑡
(1 + 0.2) 514620 N
=
1 − 0.2541
= 827.92 kN

The total resistance for the design speed 𝑣𝑆 = 14.5 kn has been taken from Table ..
In cases where a resistance estimate for the design speed is not yet available, one may
first assume a quasi-propulsive efficiency 𝜂𝐷 (for instance 𝜂𝐷 = 0.6) and then compute
effective power from delivered power.

𝑃𝐸 = 𝑃𝐷 𝜂𝐷 (.)

From the effective power follows the total resistance.

𝑃𝐸
𝑅𝑇service = (.)
𝑣𝑆

The resistance is considered the resistance for service condition since we started using
all available delivered power. At the end of the selection process, we will check whether
the propeller provides sufficient thrust.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 601

An early design formula for bulk carriers by Kristensen and Lützen () is used to Initial propeller
estimate an initial propeller diameter. diameter

𝐷 = 0.395 𝑇max + 1.3 m (.)

With the data from Table . this yields a propeller diameter of

𝐷 = 0.395 ⋅ 10.65m + 1.3 m = 5.5068 m (.)

Again, a check of this estimate against a preliminary lines plan is warranted to ensure
sufficient propeller clearances.
Substituting all values into Keller’s formula (.) yields
( ) ( )
𝐴𝐸 1.3 + 0.3 ⋅ 4 827918 N
= ( ) + 0.2
𝐴0 req 171735.32 Pa − 1671 Pa 5.50682 m2
= 0.6013 (.)

We will start with a design chart for the expanded blade area ratio 𝐴𝐸 ∕𝐴0 = 0.55 (see
Figure .).

49.2.3 First iteration

j The 𝐵𝑝1 -chart (Figure .) used here has been calculated and redrawn on the basis of Selecting the j
the Wageningen B-Series polynomials by Oosterveld and van Oossanen (). Original design chart
charts can be obtained from MARIN (http://www.marin.nl).
We draw a vertical line for the value of the fourth root out of our design constant 𝐶𝑛 Reading values
Equation (.). off design chart
√ [ ]1
4 𝐾𝑄 4
𝐶 = = 1.109938 (.)
𝐽5
For the intersection of the vertical line and the curve of the optimum efficiency 𝜂𝑂opt
we read off the following values:

open water efficiency 𝜂𝑂 = 0.5330

pitch–diameter ratio (𝑃 ∕𝐷)𝑂 = 0.6876
1
reciprocal advance coefficient 𝛿= 𝐽𝑂
= 2.4053

The reciprocal advance coefficient is equivalent to an advance coefficient of

𝐽𝑂 = 0.4158 (.)

This, in turn, gives us the optimum diameter for open water conditions (uniform wake):

𝑣𝐴 4.9135 ms
𝐷𝑂 = = = 6.1130 m (.)
𝐷 𝐽𝑂 1.9333 s−1 ⋅ 0.4158

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

602 49 Hull–Propeller Matching Examples

j j

Figure 49.5 Propeller design 𝐵𝑝1 -chart for Wageningen B-series propeller B4-55. Calculated and plotted based on
the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and van Oossanen (1975)

Selection of a As stated before, experience shows that the optimum propeller for the behind condition
nonuniform with a nonuniform wake works at a slightly higher advance coefficient. The value of
wake adjustment
𝛿 = 1∕𝐽 is commonly reduced by % – % (Lewis, ). Vessels with fine lines or
small propeller loading have small reductions, vessels with high block coefficient or
high propeller loading have larger reductions in 1∕𝐽 . For design tasks where the rate of
revolution 𝑛 is given, a reduction in 1∕𝐽 leads to a lower diameter 𝐷. In our case, the
rate of revolution 𝑛 is given and the diameter 𝐷 is reduced. The bulk carrier is a fairly
full vessel which usually leads to larger variations in the wake. Therefore, a reduction of
5% is chosen for this example.
Nonuniform The new 1∕𝐽 value for the behind condition is
wake adjustment
1
𝛿𝐵 = (1 − Δ𝛿) = 0.95 ⋅ 2.4053 = 2.2850 (.)
𝐽𝑂

This results in an equivalent reduction of the propeller diameter for the behind condition
of ( )
𝐷 = 1 − Δ𝛿 𝐷𝑂 = 5.8073 m (.)
Reducing the diameter for constant speed of advance and rate of revolution is equal to

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 603

an increase in advance coefficient.

𝑣𝐴 4.9135 ms
𝐽 = = = 0.4376 (.)
𝑛𝐷 1.9333 𝑠 ⋅ 5.8073 m
With increasing advance coefficient 𝐽 , the torque a propeller absorbs is decreasing. As
a consequence, the engine would speed up until the torque absorbed by the propeller
and the torque delivered by the engine are in equilibrium again. To compensate for
the loss in torque caused by the reduction of the diameter, pitch has to be increased.
The optimum pitch–diameter ratio for the behind condition can be taken off the 𝐵𝑝1 -
√
chart at the intersection of the 𝐶 line and the (−5%) 1∕𝐽opt curve plotted above the
4

optimum open water efficiency curve. Here

𝑃
= 0.7752 adapted for behind condition (.)
𝐷

If you have a propeller design chart without these curves, the following rule of thumb Rule of thumb
may be applied: When changing diameter or pitch, keep the sum of pitch and diameter
constant to achieve the same thrust:

𝑃 + 𝐷 = 𝑃𝑂 + 𝐷𝑂 (.)

We solve this for the new pitch for the behind condition.
j [ ( ) ] j
𝑃
𝑃 = 1+ 𝐷𝑂 − 𝐷 (.)
𝐷 𝑂
= [1 + 0.6876] ⋅ 6.1130 m − 5.8073 m
= 4.5090 m (.)

This, in turn, results in a new (increased) pitch–diameter ratio for the propeller of
𝑃 ∕𝐷 = 0.7764 in behind condition. It is close to the value we read off the chart with
the (−5%) 1∕𝐽opt curve.
By adjusting 1∕𝐽 for nonuniform wake and subsequently increasing the pitch–diameter
ratio, we moved away from the optimum efficiency. The new open water efficiency will
be slightly smaller:

𝑃 ( )
𝜂𝑂 = 0.5283 for = 0.7752 and 𝐽 = 0.4376 𝛿𝐵 = 2.2850
𝐷

In summary, the first iteration yields the following optimum propeller characteristics: First iteration
result

Result of 1. iteration
rate of revolution 𝑛 = 1.9333 s−1
open water efficiency 𝜂𝑂 = 0.5283
diameter 𝐷 = 5.8073 m
pitch–diameter ratio 𝑃 ∕𝐷 = 0.7752
advance coefficient 𝐽 = 0.4376
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.55

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

604 49 Hull–Propeller Matching Examples

49.2.4 Cavitation check for first iteration

Cavitation limit The optimum diameter obtained during the first iteration is quite a bit larger than our
initial estimate for Keller’s formula. Therefore, the initially selected expanded blade area
ratio 𝐴𝐸 ∕𝐴0 = 0.55 might be too pessimistic. Again, we employ the empirical cavitation
charts by Burrill and Emerson () and use the curve labeled 5% back cavitation as
limiting criteria (see Table .). For programming and calculation purposes, it can be
expressed as a regression curve (see Equation (.)).
Static pressure The local cavitation number 𝜎𝑏 is computed using the total static pressure at the depth
and flow velocity of the propeller shaft and a simplified flow velocity 𝑣1 for the blade section at radius
𝑟 = 0.7𝑅. The total static pressure is equal to the atmospheric pressure 𝑝𝐴 plus the
hydrostatic pressure at the propeller shaft.

𝑝0 = 𝑝𝐴 + 𝜌 𝑔 ℎ 0 (.)

With a standard atmospheric pressure of 𝑝𝐴 = 101325 Pa and a shaft submergence of

ℎ0 = 7.0 m, we get:

𝑝0 = 101325 Pa + 1026.021 kg/m3 9.807 m/s2 7.0 m

= 171 760.32 Pa

The simplified flow velocity 𝑣1 is

j √ j
𝑣1 = 𝑣2𝐴 + (0.7 𝜋 𝑛 𝐷)2 (.)
√
( )2
= 4.91352 (m/s)2 + 0.7 𝜋 1.9333 s−1 5.8073 m
= 25.1748 m/s

Local cavitation With this input we compute the local cavitation number:
number
𝑝0 − 𝑝𝑣
𝜎𝑏 = (.)
1
2
𝜌 𝑣21

With a vapor pressure of 𝑝𝑣 = 1671 Pa (ITTC, ) we get

171760.32 Pa − 1671 Pa
𝜎𝑏 =
1
1026.021 kg/m3 25.17482 (m/s)2
2
= 0.5231

Required Based on the cavitation number 𝜎𝑏 = 0.5231, we get a thrust loading coefficient of
projected area 𝜏𝑐 = 0.1976 from Equation (.). The thrust loading coefficient is connected to the
minimum required projected area 𝐴𝑃 of the propeller by

𝑇prop
𝐴𝑃req = (.)
1 2
𝜌𝑣 𝜏
2 1 𝑐

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 605

In order to proceed, we need an accurate estimate of the thrust generated by the

propeller defined by the first iteration. Given 𝐽 , 𝑃 ∕𝐷, 𝐴𝐸 ∕𝐴0 , and 𝑍, the Wageningen
B-series polynomials from Chapter . can be used to compute the thrust coefficient
𝐾𝑇 and thus the propeller thrust. Using the polynomial (.), we obtain

𝐾𝑇 = 0.1848
𝑇prop = 806.10 kN

If polynomials are not available, direct computational methods like blade element
theory or lifting line theory might be employed.
As before, we use Equation (.) to compute the required expanded area ratio Required
expanded area
( )
𝐴𝐸 𝑇prop ratio
= ( ) (.)
𝐴0 req 1 2 𝑃 𝜋𝐷2
𝜌 𝑣1 𝜏𝑐 1.067 − 0.229
2 𝐷 4
With the data for the propeller derived in the first iteration, the 5% back cavitation
criterion asks for a required minimum expanded area ratio of
( )
𝐴𝐸
= 0.5325 (.)
𝐴0 req

to prevent excessive cavitation. This is less than the value from our first design chart
j (𝐴𝐸 ∕𝐴0 = 0.55). A lower expanded area ratio typically results in higher efficiency j
because the reduced blade surfaces have smaller frictional losses. We will have to
perform a second iteration using a lower expanded area ratio.

49.2.5 Second iteration

In the second iteration, we make use of the 𝐵𝑝1 -chart for an expanded blade area ratio Design chart
of 𝐴𝐸 ∕𝐴0 = 0.40. The design constant 𝐶 remains unchanged because the available application
delivered power for the ship has not changed. Performing the same steps as above, i.e.

• reading off values for 𝜂𝑂 , (𝑃 ∕𝐷)𝑂 , and 1∕𝐽𝑂 ,

• calculating diameter, and
• reducing diameter and deriving new pitch

yields the following set of propeller parameters:

open water efficiency 𝜂𝑂 = .

pitch–diameter ratio (𝑃 ∕𝐷)𝑂 = .
1
reciprocal advance coefficient 𝛿= 𝐽𝑂
= .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

606 49 Hull–Propeller Matching Examples

j j

Figure 49.6 Propeller design 𝐵𝑝1 -chart for Wageningen B-series propeller B4-40. Calculated and plotted based on
the 𝐾𝑇 , 𝐾𝑄 polynomials by Oosterveld and van Oossanen (1975)

Adjustment for Reduction of 1∕𝐽 by Δ𝛿 = 5% yields:

nonuniform
wake
Result of 2. iteration
rate of revolution 𝑛 = 1.9333 s−1
open water efficiency 𝜂𝑂 = 0.5316
diameter 𝐷 = 5.7825 m
pitch–diameter ratio 𝑃 ∕𝐷 = 0.7866
advance coefficient 𝐽 = 0.4395
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.40

Cavitation check This propeller produces a thrust of 𝑇prop = 811.14 kN. The cavitation check with
Burrill’s chart, i.e. Equation (.), yields a required expanded area ratio of
( )
𝐴𝐸
= 0.5438 (.)
𝐴0 req

This is more than the 𝐴𝐸 ∕𝐴0 = 0.40 from the chart we used.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 607

Auxiliary plot for expanded area ratio AE /A0 Figure 49.7 Auxiliary plot to determine
the expanded blade area ratio of the final
2 optimum propeller
iteration

(AE /A0 )chart design chart value AE /A0 =0.5337 for

1 (AE /A0 )chart =( AE /A0 )req
(AE /A0 )req 5% cavitation limit

0.400 0.425 0.450 0.475 0.500 0.525 0.550

expanded area ratio AE /A0 [−]

49.2.6 Final selection by interpolation

We find the final propeller characteristics by interpolating between the results of the
first and the second iterations.

j First, the expanded area ratio is found by comparing the expanded area ratios we used Expanded area j
(from the charts) and the required expanded area ratios according to Equation (.). ratio
The auxiliary plot for the expanded blade area ratio (Figure .) compares the actually
used expanded blade area ratios and the required expanded blade area ratios for the
two iterations. Where the two lines intersect, we find:
( )
𝐴𝐸
= 0.5337 (.)
𝐴0

For the final optimum propeller characteristics, we plot straight lines connecting the Pitch–diameter
results of the first and the second iterations for pitch–diameter ratio 𝑃 ∕𝐷, advance ratio, advance
coefficient
coefficient 𝐽 , open water efficiency 𝜂𝑂 , and diameter 𝐷 as a function of the expanded
area ratio (Figure .). The fourth graph for the diameter may be omitted and the
diameter recomputed from the final advance coefficient 𝐽 via Equation (.).
Final values are read off the plots or found by computing the intersection with the Final selection
vertical line for 𝐴𝐸 ∕𝐴0 = 0.5337.

Result of final selection

rate of revolution 𝑛 = 1.9333 s−1
open water efficiency 𝜂𝑂 = 0.5287
diameter 𝐷 = 5.8046 m
pitch–diameter ratio 𝑃 ∕𝐷 = 0.7764
advance coefficient 𝐽 = 0.4378
expanded area ratio 𝐴𝐸 ∕𝐴0 = 0.5337

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

608 49 Hull–Propeller Matching Examples

Auxiliary plot for pitch−diameter ratio P/D Auxiliary plot for advance ratio J
0.7880 0.4400
[−]

0.7860

[−]
0.4395
pitch−diameter ratio P/D

0.7840

advance coefficient J
AE /A0 =0.5337

AE /A0 =0.5337
0.7820 0.4390

0.7800 0.4385
0.7780
P/D =0.7764 0.4380 J =0.4378
0.7760

0.7740 0.4375
0.400 0.425 0.450 0.475 0.500 0.525 0.550 0.400 0.425 0.450 0.475 0.500 0.525 0.550

Auxiliary plot for open water efficiency O Auxiliary plot for propeller diameter D
0.5320 5.8100

D =5.8046m
[−]

0.5315
[m]

5.8050
0.5310
O

5.8000
propeller diameter D
AE /A0 =0.5337

AE /A0 =0.5337
0.5305
open water efficiency

0.5300 5.7950

0.5295
5.7900
0.5290
O =0.5287
5.7850
0.5285
j j
0.5280 5.7800
0.400 0.425 0.450 0.475 0.500 0.525 0.550 0.400 0.425 0.450 0.475 0.500 0.525 0.550
expanded area ratio AE /A0 [−] expanded area ratio AE /A0 [−]

Figure 49.8 Auxiliary plots to determine final optimum propeller characteristics. The ⋆ mark results of first and
second iterations

This propeller provides a thrust of 807.09 kN. Checking with the Burrill chart for a
third time yields a required expanded area ratio of 𝐴𝐸 ∕𝐴0 = 0.5340, which confirms
our final selection.

49.2.7 Attainable speed check

Check of Propeller design tasks  and  use the required thrust as input. This ensures that the
attainable speed propeller is capable of bringing the vessel up to its design speed, assuming that the
engine is producing the necessary torque. Propeller design tasks  and , however,
use the available delivered power (or torque) as input. We selected a propeller which
absorbs the available torque with high efficiency. Unfortunately, this does not guarantee
that the propeller is producing sufficient thrust to attain the desired design speed.
Therefore, a check of the achievable speed is required for design tasks  and . At this
point, a resistance estimate is necessary like the one shown in Table ..
Available vs. Figure . shows the ship speed as a function of required thrust in service. 𝑇req
required thrust has been computed from the resistance data in Table . and Equation (.). For

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

49.2 Optimum Diameter Problem 609

Figure 49.9 Auxiliary plot to determine the

attainable ship speed

comparison, the curve of thrust for trial conditions is shown as well. In order to find
the attainable speed, we mark the available propeller thrust 𝑇prop = 807.09 kN in the
graph by a vertical line. The intersection with the required thrust curve reveals the
actual attainable ship speed.
j In this example, a speed of 𝑣𝑆 = 7.387 m/s or 14.36 kn is achieved, which is 99% of the Attainable j
target design speed of 14.5 kn. Note that the same thrust would suffice for a speed of velocity
about 15.3 kn at trial. If the lower speed is unacceptable to the owner, the following
remedies may be pursued:

• The rating or power output of diesel engines may be adjusted within certain limits.
If the specified brake power of 𝑃𝐵 = 7650 kW is not already the maximum range
of the engine model, the rating may be increased. In the example, an increase
of brake power by about 3% at the same rate of revolution would be sufficient to
attain the desired design speed.
• In some cases, it may be possible to achieve the same power output at a lower
rate of revolution. In the example above, a propeller optimized for 𝑛 = 104 rpm
would drive the ship at design speed with the stated brake power of 7650 kW. A
combination of power increase and lower rate of revolution may be used if the
engine layout diagram permits it.
• If an adjustment of engine rating is impossible, design measures have to be taken
to improve the overall propulsive efficiency. Optimizations of bulbous bow and
hull shape in the run may lower resistance and/or improve hull efficiency.

The propeller selection process must be repeated once a higher brake power, lower rate
of revolution, lower resistance, or improved hull–propeller interaction parameters are
known.
Obviously, the whole propeller selection process may be implemented as a program, Process
eliminating the need to read values off charts. In fact, both examples here were solved automation

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

610 49 Hull–Propeller Matching Examples

by a computer program which also creates the data plots. Plotting or using the charts is
a good way to check for errors in the input and for finding ways to improve the design.

References
Burrill, L. and Emerson, A. (). Propeller cavitation: further tests on in propeller
models in the King’s College cavitation tunnel. Transactions of the North East Coast
Institution of Engineers and Shipbuilders (NECIES), :–.
Harvald, S. (). Resistance and propulsion of ships. John Wiley & Sons, New York,
NY.
ITTC (). Fresh water and seawater properties. International Towing Tank Confer-
ence, Recommended Procedures and Guidelines .---. Revision .
Kristensen, H. and Lützen, M. (). Prediction of resistance and propulsion power
of ships. Technical Report Project No. -, Work Package , Report No. ,
University of Southern Denmark and Technical University of Denmark.
Lewis, E., editor (). Principles of naval architecture, volume II – Resistance, propul-
sion and vibration. The Society of Naval Architects and Marine Engineers, Jersey
City, NJ, second edition.
Oosterveld, M. and van Oossanen, P. (). Further computer-analyzed data of the
Wageningen B-screw series. International Shipbuilding Progress, :–.
j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

611

50
Holtrop and Mennen’s Method

In the late s and early s, J. Holtrop and G.G.J. Mennen developed a resistance
and propulsion prediction method based on the regression analysis of model tests and
trial data of MARIN, the model basin in Wageningen, The Netherlands (Holtrop, ;
Holtrop and Mennen, , ; Holtrop, , ). All necessary equations are
presented in their newest published form and an example data set with intermediate
results allows readers to check their own implementations.

Learning Objectives

j At the end of this chapter students will be able to: j

• understand applicability and limitations of Holtrop and Mennen’s resistance and

propulsion estimate

• implement the method as a computer program or spreadsheet

• apply the method in design projects

50.1 Overview of the Method

Holtrop and Mennen’s method is arguably the most popular method to estimate resis-
tance and powering of displacement type ships. It is based on the regression analysis of
a vast range of model tests and trial data which give it a wide applicability. It is – to the
best of the author’s knowledge – the only method that adopted the use of the ITTC
form factor 𝑘. Resistance is calculated as a dimensional force. The method also provides
formulas to estimate the hull–propeller interaction parameters thrust deduction, full
scale wake fraction, and relative rotative efficiency.

50.1.1 Applicability

Not much information is provided in the publications about the range of application Range of
of the method. Holtrop () provides a table with ranges for prismatic coefficient, application
length to beam ratio, and Froude number of the ship types considered in the original

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

612 50 Holtrop and Mennen’s Method

Table 50.1 Required and optional input parameters for Holtrop and Mennen’s method

Parameter Symbol Remarks

required parameters
length in waterline 𝐿𝑊𝐿
molded beam 𝐵
1( )
molded mean draft 𝑇 typically 𝑇 = 2
𝑇𝐴 + 𝑇𝐹
molded draft at aft perpendicular 𝑇𝐴
molded draft at forward perpendicular 𝑇𝐹
volumetric displacement (molded) 𝑉 alternatively use the block coefficient as
𝐶𝐵 = 𝑉 ∕(𝐵 𝑇 𝐿𝑊𝐿 )
prismatic coefficient (based on 𝐿𝑊𝐿 ) 𝐶𝑃
midship section coefficient 𝐶𝑀 or use 𝐶𝑀 = 𝐶𝐵 ∕𝐶𝑃
waterplane area coefficient 𝐶𝑊𝑃 may have to be estimated in early design
stages
longitudinal center of buoyancy 𝓁𝐶𝐵 positive forward; with respect to 𝐿𝑊𝐿 ∕2
in percent of 𝐿𝑊𝐿
area of ship and cargo above waterline 𝐴𝑉 projected in direction of 𝑣𝑆
immersed transom area 𝐴𝑇 measured at rest
transverse area of bulbous bow 𝐴𝐵𝑇 measured at forward perpendicular
height of center of 𝐴𝐵𝑇 above basis ℎ𝐵 has to be smaller than 0.6𝑇𝐹
j propeller diameter 𝐷 j
propeller expanded area ratio 𝐴𝐸 ∕𝐴0
stern shape parameter 𝐶stern typical values provided with Equa-
tion (.) below
optional parameters
wetted surface (hull) 𝑆
wetted surface of appendages 𝑆𝐴𝑃 𝑃𝑖 bilge keels, stabilizer fins, etc.
half angle of waterline entrance 𝑖𝐸
diameter of bow thruster tunnel 𝑑𝑇𝐻

regression analysis. Summarizing the numbers, reasonable estimates can be expected

for cases that fit the following conditions:

𝐹𝑟 ≤ 0.45
0.55 ≤ 𝐶𝑃 ≤ 0.85 (.)
𝐿
3.9 ≤ ≤ 9.5
𝐵

50.1.2 Required input

Input to Holtrop and Mennen’s method consists of principal dimensions and a few
basic hull form parameters. The necessary parameters are listed in Table ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.1 Overview of the Method 613

Note that the method uses the length in waterline 𝐿𝑊𝐿 as the characteristic length of Characteristic
the vessel. Do not forget to recompute the block coefficient 𝐶𝐵 and prismatic coefficient length
𝐶𝑃 based on 𝐿𝑊𝐿 . ( )
𝐿𝑃𝑃
𝐶𝐵(w.r.t 𝐿 ) = 𝐶𝐵(w.r.t 𝐿 ) (.)
𝑊𝐿 𝐿𝑊𝐿 𝑃𝑃

Do the same for the prismatic coefficient, if it is provided on the basis of length between
perpendiculars 𝐿𝑃𝑃 .
In early design stages some input parameters may still be unknown. They may be Missing
initially derived from design formulas and later replaced with the actual values. parameters

If, for instance, the longitudinal center of buoyancy is not yet known, the following Longitudinal
suggestion for the optimum 𝐿𝐶𝐵 location by Guldhammer and Harvald () may be center of
buoyancy
used. ( )
𝓁𝐶𝐵 = − 0.44 𝐹𝑟design − 0.094 (.)
It is important to enter 𝓁𝐶𝐵 as a percentage. Say a ship with a waterline length of
𝐿𝑊𝐿 = 182 m has its longitudinal center of buoyancy 𝐿𝐶𝐵 = 1.5 m aft of 𝐿𝑊𝐿 ∕2. Thus
𝓁𝐶𝐵 will be negative (aft of 𝐿𝑊𝐿 ∕2) and

𝐿𝐶𝐵 1.5 m
𝓁𝐶𝐵 = − ⋅ 100 = − ⋅ 100% = −0.8242%
𝐿𝑊𝐿 182 m

Only the value −0.8242 will be entered into the formulas. If 𝐿𝐶𝐵 is given with respect
j to 𝐿𝑃𝑃 ∕2, it first has to be referenced to 𝐿𝑊𝐿 ∕2. j
A regression equation based on a graph by Jensen () may help with the midship Midship section
section coefficient: coefficient
1
𝐶𝑀 = (.)
1 + (1 − 𝐶𝐵 )3.5
The waterplane area coefficient 𝐶𝑊𝑃 is often unknown until the lines plan is completed. Waterplane area
Bertram and Wobig () offer this formula for initial estimates: coefficient

⎧0.763 (𝐶 + 0.34) for tanker, bulk carrier, general

⎪ 𝑃
⎪ cargo with 0.56 < 𝐶𝑃 < 0.87
𝐶𝑊𝑃 = ⎨ ( ) (.)
⎪3.226 𝐶𝑃 − 0.36 for container ships with
⎪ 0.57 < 𝐶𝑃 < 0.62
⎩

Similar formulas may be found in Watson () and Papanikolaou ().

The wetted surface of the hull at rest 𝑆 may be estimated by the following formula, if it Wetted surface
is not yet known from hydrostatic analysis (Holtrop and Mennen, ).
√ 𝐴
𝑆 = 𝑐23 𝐿𝑊𝐿 (2𝑇 + 𝐵) 𝐶𝑀 + 2.38 𝐵𝑇 see also Equation (.) (.)
𝐶𝐵

with the factor

[ ]
𝐵
𝑐23 = 0.453 + 0.4425 𝐶𝐵 − 0.2862 𝐶𝑀 − 0.003467 + 0.3696 𝐶𝑊𝑃 (.)
𝑇

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

614 50 Holtrop and Mennen’s Method

Equation (.) was later updated to provide a more accurate prediction, especially for
slender hull forms with small midship section coefficients (Holtrop, ).
[
√ 3
𝑆 = 𝐿𝑊𝐿 (2𝑇 + 𝐵) 𝐶𝑀 0.615989 𝑐23 + 0.111439 𝐶𝑀
]
𝑐23
+ 0.000571111 𝐶stern + 0.245357
𝐶𝑀
( )
𝐴𝐵𝑇 0.5839497
+ 3.45538 𝐴𝑇 + 1.4660538 + (.)
𝐶𝐵 𝐶𝑀

Waterline For the half angle of the waterline entrance, Holtrop and Mennen () provide the
entrance angle following formula:
𝑖𝐸 = 1 + 89 e𝑎 (.)
The necessary exponent 𝑎 is found by evaluating
[( )
𝐿𝑊𝐿 0.80856( )0.30484 [ ]0.6367
𝑎 = − 1 − 𝐶𝑊𝑃 1 − 𝐶𝑃 − 0.0225 𝓁𝐶𝐵
𝐵
( )0.34574 ( )0.16302 ]
𝐿𝑅 100 𝑉
(.)
𝐵 𝐿𝑊𝐿 3
j j
Note that the angle 𝑖𝐸 is returned in degrees like it is needed in formulas below.
Length of run The length of the run 𝐿𝑅 is required as an additional input for Equation (.). It may
be estimated by the following formula (Holtrop, ):
( )
1 − 𝐶𝑃 + 0.06 𝐶𝑃 𝓁𝐶𝐵
𝐿𝑅 = 𝐿𝑊𝐿 (.)
4 𝐶𝑃 − 1

50.2 Procedure
The Holtrop () method computes a dimensional total resistance which is broken
down into several components:

• frictional resistance 𝑅𝐹 ,

• appendage resistance 𝑅𝐴𝑃 𝑃 ,

• wave resistance 𝑅𝑊 ,

• resistance due to bulbous bow near the water surface 𝑅𝐵 ,

• pressure resistance due to immersed transom 𝑅𝑇 𝑅 ,

• model–ship correlation resistance 𝑅𝐴 , and

• air resistance 𝑅𝐴𝐴 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.2 Procedure 615

As mentioned, Holtrop and Mennen’s method is the only early design estimate for Form factor is
resistance and propulsion that has adopted the ITTC form factor approach. Since the used
use of a form factor affects the estimate of the residuary resistance, or, in this case, the
wave resistance, the method should not be used without a form factor.
Resistance components will be computed as functions of Froude and Reynolds numbers Froude and
for the range of speeds in question. Reynolds
number
𝑣𝑆 𝑣𝑆 𝐿𝑊𝐿
𝐹𝑟 = √ 𝑅𝑒 = (.)
𝑔 𝐿𝑊𝐿 𝜈

Note that both are based on the length in waterline 𝐿𝑊𝐿 in the context of this method.

50.2.1 Resistance components

The frictional resistance 𝑅𝐹 is computed on the basis of the ITTC  model–ship Frictional
correlation line coefficient 𝐶𝐹 (.) as the resistance of a flat plate with wetted surface 𝑆. resistance

1 2
𝑅𝐹 = 𝜌 𝑣 𝑆 𝐶𝐹 (.)
2 𝑆
Do not forget to convert density 𝜌, ship speed 𝑣𝑆 , and wetted surface 𝑆 into a consistent
set of units.
j The flat plate resistance 𝑅𝐹 is later augmented by a form factor 𝑘 when the resistance Form factor j
components are assembled into the total resistance. First, a constant labeled 𝑐14 has to
be found which captures the influence of the aft body shape.

Aft body shape 𝐶stern

pram with gondola −25
𝑐14 = 1.0 + 0.011 𝐶stern with V-shaped sections −10 (.)
normal sections 0
U-shaped sections +10

Using the constant 𝑐14 , the length of run 𝐿𝑅 (.), and input values from Table .,
the hull form factor 𝑘 may be estimated (named 𝑘1 in Holtrop, ).
[( )1.06806 ( )0.46106
𝐵 𝑇
𝑘 = −0.07 + 0.487118 𝑐14
𝐿𝑊𝐿 𝐿𝑊𝐿
( ) ( )0.36486 ]
𝐿𝑊𝐿 0.121563 𝐿𝑊𝐿 3 ( )−0.604247
1 − 𝐶𝑃 (.)
𝐿𝑅 𝑉

The original formula for the form factor provided by Holtrop () computes (1 + 𝑘).
In a later update, Holtrop () introduced a speed dependent correction for the form
factor. However, this only seems to yield improvements for very fast vessels (𝐹𝑟 > 0.5).
A method to estimate the appendage resistance 𝑅𝐴𝑃 𝑃 is provided in Holtrop and Men- Appendage
nen (). The form factor values for appendages in Table . have been updated in resistance
Holtrop (). Appendages mostly affect the viscous resistance. As discussed at length,

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

616 50 Holtrop and Mennen’s Method

Table 50.2 Approximate values for appendage form factors 𝑘2𝑖 according to Holtrop (1988)

Appendage 𝑘2𝑖 value

rudder behind skeg .–.
rudder behind stern .
twin screw rudder (slender) .
twin screw rudder (thick) .
shaft brackets .–.
skeg .–.
strut bossing .–.
hull bossing .
exposed shafts (angle with buttocks about  degrees) .
(angle with buttocks about  degrees) .
stabilizer fins .
dome .
bilge keels .

Reynolds numbers are considerably smaller in model tests than at full scale. Conse-
quently, model tests are not best suited to quantify appendage resistance. Unfortunately,
they are often the only option. In addition, effects of appendages are usually registered
as a whole and not individually. Practice has shown that reasonable estimates can be
j made based on the individual form factors listed in Table .. A single, equivalent j
form factor is determined for the calculation of appendage resistance (.).
∑
(1 + 𝑘2𝑖 )𝑆𝐴𝑃𝑃𝑖
( ) 𝑖
1 + 𝑘2 eq
= ∑ (.)
𝑆𝐴𝑃𝑃𝑖
𝑖

The resistance due to a bow thruster tunnel opening may be computed according to

𝑅𝑇𝐻 = 𝜌 𝑣2𝑆 𝜋 𝑑𝑇𝐻

2
𝐶𝐷𝑇𝐻 (.)

The drag coefficient 𝐶𝐷𝑇𝐻 for the thruster tunnel takes values between . and
.. The smaller values are for thrusters which are in the cylindrical part of the
bulbous bow, i.e. the rim of the opening is fairly parallel to the midship plane. See also
Equation (.).
The appendage resistance is calculated as the sum of thruster resistance and all consid-
ered appendages.

1 2( ) ∑ ∑
𝑅𝐴𝑃 𝑃 = 𝜌 𝑣𝑆 1 + 𝑘2 eq 𝐶𝐹 𝑆𝐴𝑃𝑃𝑖 + 𝑅𝑇𝐻 (.)
2 𝑖

𝐶𝐹 is the ITTC  model–ship correlation line coefficient which was already com-
puted for the frictional resistance 𝑅𝐹 .
Wave resistance Wave resistance is a function of Froude number 𝐹𝑟. For the estimate of 𝑅𝑊 , Holtrop

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.2 Procedure 617

() subdivided the range of Froude numbers into three sections.

⎧
⎪𝑅𝑊𝑎 (𝐹𝑟) Equation (.) if 𝐹𝑟 ≤ 0.4
⎪
𝑅𝑊 (𝐹𝑟) = ⎨interpolation, see Equation (.) if 0.4 < 𝐹𝑟 ≤ 0.55 (.)
⎪
⎪𝑅𝑊 𝑏 (𝐹𝑟) Equation (.) if 𝐹𝑟 > 0.55
⎩

The wave resistance for Froude numbers 𝐹𝑟 < 0.4 is computed from (Holtrop, )
[ ( )]
𝑅𝑊𝑎 (𝐹𝑟) = 𝑐1 𝑐2 𝑐5 𝜌 𝑔 𝑉 exp 𝑚1 𝐹𝑟𝑑 + 𝑚4 cos 𝜆 𝐹𝑟−2 (.)

The expression exp[𝑥] is used here for the function e𝑥 to make the formula more readable.
Equations for the coefficients in Equation (.) are provided in Table ..
The wave resistance for Froude numbers 𝐹𝑟 > 0.55 is computed from (Holtrop, )
[ ( )]
𝑅𝑊 𝑏 (𝐹𝑟) = 𝑐17 𝑐2 𝑐5 𝜌 𝑔 𝑉 exp 𝑚3 𝐹𝑟𝑑 + 𝑚4 cos 𝜆 𝐹𝑟−2 (.)

Formulas for the coefficients in Equation (.) are provided in Tables . and ..
For the remaining range of Froude numbers 0.4 > 𝐹𝑟 ≤ 0.55, a linear blending between
Equations (.) and (.) is performed (Holtrop, ). Ships typically do not
operate in this uneconomical speed range for prolonged periods of time.
j (20 𝐹𝑟 − 8) [ ]
j
𝑅𝑊 (𝐹𝑟) = 𝑅𝑊𝑎 (0.4) + 𝑅𝑊 𝑏 (0.55) − 𝑅𝑊𝑎 (0.4) (.)
3
The expressions 𝑅𝑊𝑎 (0.4) and 𝑅𝑊 𝑏 (0.55) mean that the wave resistance is evaluated
with the equation for 𝑅𝑊𝑎 or 𝑅𝑊 𝑏 at Froude number 𝐹𝑟 = 0.4 or 𝐹𝑟 = 0.55 respectively.
Do not forget to recompute the factor 𝑚4 with the respective Froude numbers 𝐹𝑟 = 0.4
or 𝐹𝑟 = 0.55.
Holtrop () presents additional reduction factors to capture the effect of bulbous
bow and transom on the ship’s wave resistance. However, only few data sets were
available where models had been tested with and without bulbous bow or transom. For
early design purposes one should venture on the side of caution not to underpredict
the wave resistance.
The formula for the pressure resistance of a bulbous bow close to the water surface was Resistance of
updated in Holtrop () to include the effects of forward sinkage ℎ𝐹 and local wave bulbous bow
height at the bow ℎ𝑊 .
𝐵𝑇 ( )
ℎ𝐹 = 𝐶𝑃 𝐶𝑀 136 − 316.3 𝐹𝑟 𝐹𝑟3 but ℎ𝐹 ≥ −0.01𝐿𝑊𝐿 (.)
𝐿𝑊𝐿
𝑖𝐸 𝑣2𝑆
ℎ𝑊 = but at most ℎ𝑊 ≤ 0.01𝐿𝑊𝐿 (.)
400 𝑔
The values of ℎ𝐹 and ℎ𝑊 are used in the definition of an immersion Froude number
𝐹𝑟𝑖 for the bulbous bow.
𝑣𝑆
𝐹𝑟𝑖 = √ (.)
( √ )
𝑔 𝑇𝐹 − ℎ𝐵 − 0.25 𝐴𝐵𝑇 + ℎ𝐹 + ℎ𝑊

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

618 50 Holtrop and Mennen’s Method

Table 50.3 Coefficients for the wave resistance computation in Equation (50.20) if 𝐹𝑟 ≤ 0.4
(Holtrop, 1984)

⎧ ( )(1∕3)
⎪0.229577 𝐵
if 𝐵∕𝐿𝑊𝐿 ≤ 0.11
⎪ 𝐿𝑊𝐿
⎪ 𝐵
𝑐7 = ⎨ if 0.11 < 𝐵∕𝐿𝑊𝐿 ≤ 0.25
⎪ 𝐿𝑊𝐿
⎪ 𝐿𝑊𝐿
⎪0.5 − 0.0625 if 𝐵∕𝐿𝑊𝐿 > 0.25
⎩ 𝐵
( )1.07961 ( )−1.37565
𝑇
𝑐1 = 2223105 𝑐73.78613 90 − 𝑖𝐸
𝐵
𝐴1.5
𝐵𝑇
𝑐3 = 0.56 [ ( √ )]
𝐵 𝑇 0.31 𝐴𝐵𝑇 + 𝑇𝐹 − ℎ𝐵
√
𝑐2 = e(−1.89 𝑐3 )

𝐴𝑇
𝑐5 = 1 − 0.8
𝐵 𝑇 𝐶𝑀

⎧ 𝐿𝑊𝐿 3
j ⎪−1.69385 if ≤ 512 j
𝑉
⎪ 𝐿𝑊𝐿
⎪ −8
𝐿𝑊𝐿 3
𝑐15 = ⎨−1.69385 + 𝑉 (1∕3) if 512 < ≤ 1726.91
⎪ 2.36 𝑉
⎪
⎪0 𝐿𝑊𝐿 3
⎩ if > 1726.91
𝑉
{
8.07981 𝐶𝑃 − 13.8673 𝐶𝑃2 + 6.984388 𝐶𝑃3 if 𝐶𝑃 ≤ 0.8
𝑐16 =
1.73014 − 0.7067 𝐶𝑃 if 𝐶𝑃 > 0.8

𝑑 = −0.9

⎧ 𝐿𝑊𝐿 𝐿𝑊𝐿
⎪1.446 𝐶𝑃 − 0.03 𝐵 if ≤ 12
𝜆 = ⎨ 𝐵
⎪1.446 𝐶𝑃 − 0.36 𝐿
⎩ if 𝑊𝐿 > 12
𝐵
𝐿𝑊𝐿 𝑉 (1∕3) 𝐵
𝑚1 = 0.0140407 − 1.75254 − 4.79323 − 𝑐16
𝑇 𝐿𝑊𝐿 𝐿𝑊𝐿
−3.29 )
𝑚4 = 0.4 𝑐15 e(−0.034 𝐹𝑟

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.2 Procedure 619

Table 50.4 Additional coefficients for the wave resistance computation in Equation (50.21) if
𝐹𝑟 > 0.55 (Holtrop, 1984)

𝐴𝑇
𝑐5 = 1 − 0.8
𝐵 𝑇 𝐶𝑀
( )2.00977 ( )1.40692
−1.3346 𝑉 𝐿𝑊𝐿
𝑐17 = 6919.3 𝐶𝑀 − 2.0
𝐿𝑊𝐿 3 𝐵
( )0.326869 ( )0.605375
𝐵 𝑇
𝑚3 = −7.2035
𝐿𝑊𝐿 𝐵
−3.29 )
𝑚4 = 0.4 𝑐15 e(−0.034 𝐹𝑟

The parameter 𝑃𝐵 quantifies the emergence of the bulb from the still water line.
√
𝐴𝐵𝑇
𝑃𝐵 = 0.56 (.)
𝑇𝐹 − 1.5 ℎ𝐵 + ℎ𝐹
The additional resistance 𝑅𝐵 is computed with (Holtrop, ):
j j
(√ )3 𝐹𝑟3𝑖 (−3.0 𝑃𝐵−2 )
𝑅𝐵 = 0.11 𝜌 𝑔 𝐴𝐵𝑇 e (.)
1 + 𝐹𝑟2𝑖

Again, these formulas should be applied with caution since they are based on a limited
number of data points (Holtrop, ).
The effects of a bulbous bow are only captured in a very broad way. One should refrain
from optimizing the bulbous bow based on the presented formulas. This task should
be done with a nonlinear wave resistance code in later design stages (see for instance
Raven, ).
An immersed transom may cause an additional pressure resistance 𝑅𝑇𝑅 . It is a function Transom
of a depth Froude number 𝐹𝑟𝑇 which considers the immersion of the transom. If a resistance
transom area 𝐴𝑇 > 0 is given as input, 𝐹𝑟𝑇 is defined by
𝑣𝑆
𝐹𝑟𝑇 = √ (.)
2 𝑔 𝐴𝑇
(𝐵 + 𝐵 𝐶𝑊𝑃 )

The expression 𝐴𝑇 ∕(𝐵 + 𝐵 𝐶𝑊𝑃 ) is a measure for the average draft of the transom. If
this average draft is small compared to the speed, the flow will separate cleanly at the
transom edge and the additional transom drag vanishes. This is expressed with the
coefficient 𝑐6 .
{ ( )
0.2 1 − 0.2 𝐹𝑟𝑇 if 𝐹𝑟𝑇 < 5
𝑐6 = (.)
0 if 𝐹𝑟𝑇 > 5

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

620 50 Holtrop and Mennen’s Method

The transom drag 𝑅𝑇𝑅 is deemed negligible above the limit 𝐹𝑟𝑇 = 5.
1 2
𝑅𝑇𝑅 = 𝜌𝑣 𝐴 𝑐 (.)
2 𝑆 𝑇 6

Correlation The correlation allowance considered here includes effects of roughness and additional
allowance phenomena not captured in other resistance components. Note that correlation al-
resistance
lowance and roughness effects have been separated in the current ITTC performance
prediction procedure (ITTC, ). First, we need the additional coefficient 𝑐4 .
⎧ 𝑇𝐹
⎪ if 𝑇𝐹 ∕𝐿𝑊𝐿 ≤ 0.04
𝑐4 = ⎨ 𝐿𝑊𝐿 (.)
⎪0.04 if 𝑇𝐹 ∕𝐿𝑊𝐿 > 0.04
⎩
𝑇𝐹 is the draft at the forward perpendicular.
Then, the correlation allowance coefficient follows from Equation (.) below (Holtrop,
).
( )−0.16
𝐶𝐴 = 0.00546 𝐿𝑊𝐿 + 100 − 0.002
√
𝐿𝑊𝐿 4
+ 0.003 𝐶 𝑐 (0.04 − 𝑐4 ) (.)
7.5 𝐵 2
The coefficient 𝑐2 is found in Table .. Holtrop () states that with modern hull
j coatings, values of 𝐶𝐴 may be achieved that are 0.1⋅10−3 lower than predicted. However, j
this will not make a significant difference for early design estimates. The effect of surface
roughness higher than the standard value of 𝑘𝑆 = 150 𝜇m may be estimated by an
addition to 𝐶𝐴 (Holtrop and Mennen, ).

⎧0 if 𝑘𝑆 = 150 𝜇m
⎪
Δ𝐶𝐴 = ⎨ 0.105 𝑘(1∕3) − 0.005579 (.)
𝑆
⎪ if 𝑘𝑆 > 150 𝜇m
⎩ 𝐿𝑊𝐿 (1∕3)
In contrast to other formulas in Holtrop and Mennen’s method, Equation (.) is not
dimensionless. 𝑘𝑆 and 𝐿𝑊𝐿 have to be entered in meters to obtain correct results.
The correlation resistance is then given by
1 2( )[ ∑ ]
𝑅𝐴 = 𝜌 𝑣𝑆 𝐶𝐴 + Δ𝐶𝐴 𝑆 + 𝑆𝐴𝑃𝑃 (.)
2

Air resistance The resistance caused by moving the ship above the waterplane through air at rest is
calculated according to the standard ITTC procedure.
1
𝑅𝐴𝐴 = 𝜌 𝑣2 𝐶 𝐴 (.)
2 A 𝑆 𝐷𝐴 𝑉
𝐴𝑉 is the area of the lengthwise projection of hull and superstructure above the water-
line. The density of air is 𝜌𝐴 = 1.225 kg/m3 for standard atmospheric pressure and a
temperature of 15 ◦ C. The default air drag coefficient is 𝐶𝐷𝐴 = 0.8. Note that 𝐶𝐷𝐴 itself
cannot simply be added to the other ship resistance coefficients since it is not based on
the wetted surface.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.2 Procedure 621

50.2.2 Total resistance

The resistance components from the previous subsection are assembled into the total
resistance.

𝑅𝑇 = (1 + 𝑘) 𝑅𝐹 + 𝑅𝐴𝑃 𝑃 + 𝑅𝐴 + 𝑅𝑊 + 𝑅𝐵 + 𝑅𝑇𝑅 + 𝑅𝐴𝐴 (.)

This calm water resistance should be augmented by a service margin before it is used
to select an optimum propeller for the ship.

50.2.3 Hull–propeller interaction parameters

Holtrop () provides different sets of formulas of the propulsion parameters for Single screw
single and twin screw vessels. The formulas for single screw vessels are more complex vessels
and are discussed first.
In both cases a viscous resistance coefficient 𝐶𝑉 is needed, which combines all fric- Viscous
tion related components of the resistance and the correlation resistance from Equa- resistance
coefficient
tion (.).
(1 + 𝑘) 𝑅𝐹 + 𝑅𝐴𝑃 𝑃 + 𝑅𝐴
𝐶𝑉 = ∑ (.)
1 2( )
𝜌 𝑣𝑆 𝑆 + 𝑆𝐴𝑃 𝑃𝑖
2 𝑖

j The effect of the hull onto propeller inflow is expressed as a wake fraction 𝑤𝑆 for the Full scale wake j
full scale vessel. fraction

[ ]
𝐿𝑊𝐿 𝑐11 𝐶𝑉
𝑤𝑆 = 𝑐9 𝑐20 𝐶𝑉 0.050776 + 0.93405
𝑇𝐴 1 − 𝐶𝑃 1
√
𝐵
+ 0.27915 𝑐20 + 𝑐19 𝑐20 (.)
𝐿𝑊𝐿 (1 − 𝐶𝑃 1 )

Formulas for the necessary coefficients are listed in Table ..

Thrust deduction fraction for single screw vessels is estimated as Thrust deduction
fraction
()0.28956 ( √ )0.2624
𝐵 𝐵𝑇
0.25014
𝐿𝑊𝐿 𝐷
𝑡 = ( )0.01762 + 0.0015 𝐶stern (.)
1 − 𝐶𝑃 + 0.0225 𝓁𝐶𝐵

Finally, Holtrop () states the following formula for the relative rotative efficiency of Relative rotative
single screw vessels. efficiency

𝐴𝐸 ( )
𝜂𝑅 = 0.9922 − 0.05908 + 0.07424 𝐶𝑃 − 0.0225 𝓁𝐶𝐵 (.)
𝐴0

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

622 50 Holtrop and Mennen’s Method

Table 50.5 Coefficients for the full scale wake fraction of single screw vessels in Equation (50.38)
(Holtrop, 1984)

⎧ ( )
⎪ 𝑆 𝐵
if 𝐵∕𝑇𝐴 ≤ 5
⎪ 𝐿𝑊𝐿 𝐷 𝑇𝐴
⎪ ( )
⎪
𝑐8 = ⎨ 𝑆 7 𝐵 − 25
⎪ 𝑇𝐴
⎪ ( ) if 𝐵∕𝑇𝐴 > 5
⎪𝐿 𝐵
𝐷 − 3
⎪ 𝑊𝐿 𝑇𝐴
⎩

⎧𝑐8 if 𝑐8 ≤ 28
⎪
𝑐9 = ⎨ 16
⎪32 − if 𝑐8 > 28
⎩ 𝑐8 − 24

⎧ 𝑇𝐴 if 𝑇𝐴 ∕𝐷 ≤ 2
⎪𝐷
𝑐11 = ⎨ ( )3
𝑇𝐴
⎪0.0833333 + 1.33333 if 𝑇𝐴 ∕𝐷 > 2
⎩ 𝐷
j ⎧ 0.12997 0.11056 j
⎪ − if 𝐶𝑃 ≤ 0.7
⎪ (0.95 − 𝐶𝐵 ) (0.95 − 𝐶𝑃 )
𝑐19 = ⎨
⎪ 0.18567
− 0.71276 + 0.38648 𝐶𝑃 if 𝐶𝑃 > 0.7
⎪ (1.3571 − 𝐶𝑀 )
⎩

𝑐20 = 1 + 0.015 𝐶stern

𝐶𝑃 1 = 1.45 𝐶𝑃 − 0.315 − 0.0225 𝓁𝐶𝐵

Twin screw The respective formulas for twin screw vessels are:
vessels

𝐷
𝑤𝑆 = 0.3095 𝐶𝐵 + 10 𝐶𝑉 𝐶𝐵 − 0.23 √ (.)
𝐵𝑇
𝐷
𝑡 = 0.325 𝐶𝐵 − 0.1885 √ twin screw vessels (.)
𝐵𝑇
( ) 𝑃
𝜂𝑅 = 0.9737 + 0.111 𝐶𝑃 − 0.0225 𝓁𝐶𝐵 − 0.06325 (.)
𝐷

With the stated estimates for resistance and hull–propeller interaction parameters, a
power prediction can be completed following the procedure outlined in Section ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.3 Example 623

50.3 Example
As an example, we repeat the resistance estimate for the container ship from Section .
and add an estimate for the powering requirements.

50.3.1 Completion of input parameters

The basic input parameters from Table . are used again to facilitate a comparison.
However, block coefficient and prismatic coefficient have to be recomputed with respect
to the length in waterline 𝐿𝑊𝐿 as explained in Section ...
The waterline length is taken as the sum of length between perpendiculars and the aft Length in
overhang, assuming that the waterline reaches all the way to the unwetted transom waterline
(𝐴𝑇 = 0 m2 ).

𝐿𝑊𝐿 = 𝐿𝑃𝑃 + 𝐿aft = 145 m + 2.7 m = 147.7 m (.)

Based on this length in waterline, we obtain for block and prismatic coefficients: Block and
prismatic
𝑉 coefficient
𝐶𝐵 = = 0.6492 (.)
𝐿𝑊𝐿 𝐵𝑇
𝑉
𝐶𝑃 = = 0.6659 (.)
𝐿𝑊𝐿 𝐴𝑀
j j
The position of the longitudinal center of buoyancy also has to be stated as a fraction of 𝐿𝐶𝐵 position
length in waterline. The given 𝐿𝐶𝐵0 position of .% 𝐿𝑃𝑃 ∕2 before midships translates
into
𝓁𝐶𝐵 = 1.3067% forward of 𝐿𝑊𝐿 ∕2. (.)
Furthermore, Holtrop and Mennen’s method requires the waterplane area coefficient Waterplane area
𝐶𝑊𝑃 . With the help of the first case in Equation (.), we estimate: coefficient

𝐶𝑊𝑃 = 0.7675 (.)

For consistency, the presented results are based on the given wetted surface of 𝑆 = Wetted surface
4400 m2 . Equation (.) yields a slightly smaller value of 𝑆 = 4380.5947 m2 , but we
rather employ given data when it is available.
For the air resistance estimate, a transverse area above the waterline of 𝐴𝑉 = 383.76 m2 Air drag input
has been used. The air drag coefficient is 𝐶𝐷𝐴 = 0.8, and air density is . kg/m3 .
With the provided data, Equations (.) and (.) predict a half angle of the waterline Half angle of
entrance and length of run of 𝑖𝐸 = 19.231 degree and 𝐿𝑅 = 53.982 meters. entrance

The propulsion estimate is based on the propeller data from Table . and open water Propeller data
characteristics according to the 𝐾𝑇 and 𝐾𝑄 polynomials presented in Section ..

50.3.2 Resistance estimate

A resistance and powering estimate is executed using the equations presented in this Computational
coefficients

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

624 50 Holtrop and Mennen’s Method

Table 50.6 Propeller data for powering estimate; see Propeller data
also Table 31.5
number of blades 𝑍 5
diameter 𝐷 4.9000 m
pitch–diameter ratio 𝑃 ∕𝐷 0.9924
expanded area ratio 𝐴𝐸 ∕𝐴0 0.7692

Table 50.7 Holtrop and Mennen resistance and powering estimate example; speed
independent procedural coefficients

𝑐1 =2.045485 𝑐7 = 0.162492 𝑐15 =−1.693850 𝑑 =−0.900000

𝑐2 =0.707301 𝑐8 =17.793996 𝑐16 = 1.293587 𝜆 = 0.778264
𝑐3 =0.033572 𝑐9 =17.793996 𝑐17 = 1.731558 𝑚1 =−2.135451
𝑐4 =0.040000 𝑐11 = 1.673469 𝑐19 = 0.042998 𝑚3 =−2.076076
𝑐5 =1.000000 𝑐14 = 1.000000 𝑐20 = 1.000000 𝑃𝐵 = 2.555278
𝐶𝑃 1 = 0.621151

Table 50.8 Holtrop and Mennen resistance and powering estimate example; speed dependent procedural
coefficients

𝑣𝑆 𝐹𝑟 𝐹𝑟𝑖 𝐹𝑟𝑇 𝑐6 𝑚2 𝑚3 𝐹𝑟𝑑 𝑚4 𝑚4 cos(𝜆∕𝐹𝑟2 )

[𝑘𝑛] [−] [−] [−] [−] [−] [−] [−] [−]
j 15.00 0.20275 1.36572 0.00 0.20 −0.06596 −8.72909 −0.00104 −0.00104 j
15.50 0.20951 1.39805 0.00 0.20 −0.07697 −8.47525 −0.00202 −0.00088
16.00 0.21627 1.42948 0.00 0.20 −0.08855 −8.23651 −0.00359 0.00214
16.50 0.22303 1.46001 0.00 0.20 −0.10060 −8.01153 −0.00595 0.00594
17.00 0.22979 1.48968 0.00 0.20 −0.11303 −7.79915 −0.00926 0.00524
17.50 0.23655 1.51848 0.00 0.20 −0.12576 −7.59831 −0.01369 −0.00310
18.00 0.24331 1.54644 0.00 0.20 −0.13869 −7.40808 −0.01933 −0.01617
18.50 0.25006 1.57358 0.00 0.20 −0.15177 −7.22764 −0.02627 −0.02608
19.00 0.25682 1.59991 0.00 0.20 −0.16491 −7.05623 −0.03452 −0.02485

chapter. Table . lists values for all procedural coefficients that do not depend on
ship speed. Table . presents intermediate coefficients which are a function of ship
speed.
Form factors Equation (.) yields an ITTC form factor of 𝑘1 = 0.193743. The appendage resistance
is based on the surface of the bilge keels 𝑆𝐴𝑃𝑃 = 52 m2 and a form factor of 𝑘2 = 0.4.
Correlation The estimate for the model–ship correlation coefficient (.) results in 103 𝐶𝐴 =
coefficient 0.433837.
Resistance Based on the input and the coefficients listed in Tables .and ., resistance com-
components ponents and total resistance have been computed for trial conditions. Holtrop and
Mennen’s method computes dimensional resistance values rather than resistance coef-
ficients. Table . states resistance values for the same set of Froude numbers that has
been used in Section .. The values lie about % above the estimate from Section .
but the gap closes for higher speeds.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.3 Example 625

Table 50.9 Holtrop and Mennen resistance and powering estimate example; resistance
components and total resistance

𝑣𝑆 𝐹𝑟 𝑅𝐹 𝑅𝐴 𝑅𝑊 𝑅𝐵 𝑅APP 𝑅𝐴𝐴 𝑅𝑇
[kn] [−] [kN] [kN] [kN] [kN] [kN] [kN] [kN]
15.00 0.20275 206.82 58.31 34.60 32.56 3.42 11.20 386.98
15.50 0.20951 219.94 62.27 44.93 33.87 3.64 11.96 419.21
16.00 0.21627 233.44 66.35 57.61 35.15 3.86 12.74 454.37
16.50 0.22303 247.31 70.56 72.88 36.39 4.09 13.55 492.70
17.00 0.22979 261.56 74.90 90.61 37.61 4.33 14.38 534.07
17.50 0.23655 276.18 79.37 110.48 38.79 4.57 15.24 578.14
18.00 0.24331 291.18 83.97 132.61 39.93 4.82 16.12 625.06
18.50 0.25006 306.55 88.70 158.08 41.05 5.07 17.03 675.88
19.00 0.25682 322.29 93.56 188.80 42.13 5.33 17.97 732.51

Table 50.10 Holtrop and Mennen resistance and powering estimate example; wake fraction and self propulsion
point analysis

𝑣𝑆 𝑣𝑆 𝐹𝑟 𝑤𝑇𝑆 𝑣𝐴 𝑇req 𝐶𝑆 𝐽𝑇𝑆 𝐾𝑇𝑆 10𝐾𝑄𝑇𝑆

[kn] [m∕s] [−] [−] [m∕s] [kN] [−] [−] [−] [−]
15.0 7.717 0.2028 0.2696 5.636 482.69 0.61677 0.6108 0.2301 0.3790
15.5 7.974 0.2095 0.2694 5.826 522.89 0.62544 0.6082 0.2314 0.3807
j j
16.0 8.231 0.2163 0.2693 6.015 566.75 0.63591 0.6052 0.2329 0.3827
16.5 8.488 0.2230 0.2691 6.204 614.56 0.64813 0.6016 0.2346 0.3850
17.0 8.746 0.2298 0.2690 6.393 666.15 0.66156 0.5978 0.2365 0.3874
17.5 9.003 0.2365 0.2688 6.583 721.13 0.67555 0.5940 0.2383 0.3899
18.0 9.260 0.2433 0.2687 6.772 779.64 0.69009 0.5900 0.2403 0.3924
18.5 9.517 0.2501 0.2685 6.961 843.04 0.70616 0.5858 0.2423 0.3952
19.0 9.774 0.2568 0.2684 7.151 913.68 0.72532 0.5808 0.2447 0.3983

50.3.3 Powering estimate

The powering estimate starts with the computation of thrust deduction fraction and Constant
relative rotative efficiency. Both are treated as constants in this method. hull–propeller
interaction
parameters
thrust deduction fraction 𝑡 = 0.1983
relative rotative efficiency 𝜂𝑅 = 0.9940

With the thrust deduction fraction, the required thrust can be computed via Equa-
tion (.). Once the estimate for the wake fraction is completed, the procedure in
Section . is followed to predict the necessary delivered power for trial conditions.
Table . shows the results for wake fraction estimates and the self propulsion point Delivered power
analysis based on the propeller characteristics provided in Table .. Table .
summarizes propulsive efficiencies, the propeller rate of revolution, and the necessary
delivered power 𝑃𝐷 .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

626 50 Holtrop and Mennen’s Method

Table 50.11 Holtrop and Mennen resistance and powering estimate example; efficiencies,
propeller rate of revolution, and delivered power

𝑣𝑆 𝑣𝑆 𝐹𝑟 𝜂𝐻 𝜂𝑂 𝜂𝐷 𝑛 𝑛 𝑃𝐷
[kn] [m∕s] [−] [−] [−] [−] [1∕s] [rpm] [kW]
15.0 7.717 0.2028 1.0976 0.5902 0.6440 1.883 112.990 4637.23
15.5 7.974 0.2095 1.0974 0.5884 0.6418 1.955 117.277 5208.37
16.0 8.231 0.2163 1.0971 0.5861 0.6392 2.028 121.700 5850.65
16.5 8.488 0.2230 1.0969 0.5835 0.6363 2.104 126.269 6573.49
17.0 8.746 0.2298 1.0967 0.5808 0.6331 2.182 130.949 7378.31
17.5 9.003 0.2365 1.0965 0.5780 0.6299 2.262 135.703 8263.98
18.0 9.260 0.2433 1.0963 0.5750 0.6265 2.342 140.542 9239.76
18.5 9.517 0.2501 1.0961 0.5717 0.6229 2.425 145.520 10328.13
19.0 9.774 0.2568 1.0959 0.5680 0.6187 2.512 150.747 11573.40

References
Bertram, V. and Wobig, M. (). Simple empirical formulae to estimate main form
parameters of ships. Schiff & Hafen, ():–.
Guldhammer, H. and Harvald, S. (). Ship resistance – effect of form and principal
dimensions (revised). Akademisk Forlag, Copenhagen.
j Holtrop, J. (). A statistical analysis of performance test results. International j
Shipbuilding Progress, ():–.
Holtrop, J. (). A statistical re-analysis of resistance and propulsion data. Interna-
tional Shipbuilding Progress, ():–.
Holtrop, J. (). A statistical resistance prediction method with a speed depen-
dent form factor. In Scientific and Methodological Seminar on Ship Hydrodynamics
(SMSSH ’), Varna, Bulgaria.
Holtrop, J. and Mennen, G. (). A statistical power prediction method. International
Shipbuilding Progress, ():–.
Holtrop, J. and Mennen, G. (). An approximate power prediction method. Inter-
national Shipbuilding Progress, ():–.
ITTC ().  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .
Jensen, G. (). Moderne Schiffslinien. In Keil, H., editor, Handbuch der Werften,
volume XXII, pages –. Schiffahrts-Verlag Hansa.
Papanikolaou, A. (). Ship design – Methodologies of preliminary design. Springer,
Dordrecht, The Netherlands.
Raven, H. (). A solution method for the nonlinear ship wave resistance problem.
PhD thesis, Technical University Delft, Delft, The Netherlands.
Watson, D. (). Practical ship design. Elsevier Ocean Engineering Book Series.
Elsevier Science, Oxford, United Kingdom.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

50.3 Example 627

Self Study Problems

. Discuss the principal difference between Guldhammer and Harvald’s method for
the resistance estimate and Holtrop and Mennen’s method.
. For a ship design project, the following data is provided:

Ship data
length between perpendiculars 𝐿𝑃𝑃 = . m
length in waterline 𝐿𝑊𝐿 = . m
molded beam 𝐵 = . m
molded draft 𝑇 = . m
block coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝐵 = .
prismatic coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝑃 = .

Compute the input values for block coefficient 𝐶𝐵 and prismatic coefficient 𝐶𝑃
for Holtrop and Mennen’s method.
. Implement the Holtrop and Mennen method as a program in Python, Matlab, or
similar, and test it with the data presented in the last section.

j j

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

628

51
Hollenbach’s Method

Hollenbach’s method for resistance and powering estimates of single and twin screw
vessels was developed in the s based on test data of the Schiffbau Versuchsanstalt
in Vienna, Austria. This method has a narrower range of applicability than Holtrop
and Mennen’s method, but seems to provide more reliable results, especially for twin
screw vessels. In addition to the equations for a resistance prediction, formulas provide
estimates of upper and lower bound of the resistance. An example and comparison
with results from Sections . and . close out the discussion of resistance and
propulsion.

j
Learning Objectives j

At the end of this chapter students will be able to:

• understand applicability and limitations of Hollenbach’s resistance and powering

estimate

• implement the method as a computer program or spreadsheet

• apply the method in design projects

51.1 Overview of the method

Origin In the s, U. Hollenbach evaluated test data of the Vienna Model Basin (www.sva.at)
and developed regression formulas for resistance and propulsion estimates of single and
twin screw vessels. The formulas were tested against model test data of the Hamburg
Ship Model Basin (www.hsva.de) and, in general, showed less standard deviation than
other methods. This is an indication that Hollenbach’s method is more reliable than
previous methods (Hollenbach, ). The improvement is particularly visible for twin
screw vessels.
Underlying data The formulas are based on experiments for  displacement type vessels. Since some of
the vessels had variants and were tested on different drafts, a total of  resistance and
 propulsion tests form the basis for Hollenbach’s method. The developed formulas
have been published in Hollenbach (a) and a series of papers with updates to the

Fundamentals of Ship Hydrodynamics: Fluid Mechanics, Ship Resistance and Propulsion, First Edition.
Lothar Birk.
© 2019 John Wiley & Sons Ltd. Published 2019 by John Wiley & Sons Ltd.
Companion website: www.wiley.com/go/birk/hydrodynamics

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.1 Overview of the method 629

Table 51.1 Recommended limits for principal dimensions and form parameters of single screw
vessels on design draft

Permissible range of main dimensions

Parameter Symbol Unit Range
ship speed 𝑣𝑆 [m/s] displacement type vessel
length between perpendicular 𝐿𝑃𝑃 [m] . … .
length–beam ratio 𝐿𝑃𝑃 ∕𝐵 [–] . … .
beam–draft ratio 𝐵∕𝑇 [–] . … .
propeller diameter–draft ratio 𝐷∕𝑇𝐴 [–] . … .
block coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝐵√ [–] .(∗) … .
length–displacement ratio [–] . … .
3
𝐿𝑃𝑃 ∕ 𝑉
(∗) The resistance estimate for vessels on design draft was later expanded to smaller values of the
block coefficient (see Equation (.))

method (Hollenbach, , b, a,b, ). In addition to the mean resistance
at trial conditions, the method provides estimates for the minimum resistance (best
case scenario) and for the maximum resistance (worst case scenario). Formulas for the
resistance in ballast condition are presented as well. A complete propulsion estimate
is possible based on estimates for the hull efficiency, thrust deduction fraction, and
relative rotative efficiency, if open water propeller characteristics are known.

j 51.1.1 Applicability j

Based on the dimensions of the vessels used for the regression analysis, Hollenbach Range of
recommends staying within the limits for the main dimensions listed in Table . for application
single screw vessels. Similar limiting values for single screw ships in ballast condition
and twin screw vessels are provided in Hollenbach ().
𝑉 = 𝐶𝐵 𝐿𝑃𝑃 𝐵𝑇 is the volumetric displacement. All other parameters are explained
below in Section ... Results are expected to be less reliable for vessels that fall
outside these ranges. In addition to the restrictions on geometry, the method cd
provides estimates for the range of Froude numbers for which the predictions are
considered valid.

51.1.2 Required input

One of the advantages of Hollenbach’s method is that it requires only a few basic hull Input data
form parameters. The parameters are listed in Table ..
If the waterline length is not known, it can be set to . 𝐿𝑃𝑃 .
A calculation length 𝐿𝑐 is used in some of the formulas. It is determined on the basis of Computational
length between perpendiculars 𝐿𝑃𝑃 and length over wetted surface 𝐿𝑂𝑆 . length

⎧𝐿𝑂𝑆 if 𝐿𝑂𝑆 < 𝐿𝑃𝑃

⎪ ( )
𝐿𝑐 = ⎨𝐿𝑃𝑃 + 23 𝐿𝑂𝑆 − 𝐿𝑃𝑃 if 𝐿𝑃𝑃 < 𝐿𝑂𝑆 < 1.1𝐿𝑃𝑃 (.)
⎪1.0667𝐿 if 𝐿𝑂𝑆 > 1.1𝐿𝑃𝑃
⎩ 𝑃𝑃

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

630 51 Hollenbach’s Method

Table 51.2 Required and optional input parameters for Hollenbach’s method

Parameter Symbol Remarks

required parameters
length between perpendiculars 𝐿𝑃𝑃
length in waterline 𝐿𝑊𝐿 check definition for ballast condition
length over wetted surface 𝐿𝑂𝑆 check definition for ballast condition
molded beam 𝐵
molded draft at aft perpendicular 𝑇𝐴
molded draft at forward perpendicular 𝑇𝐹
propeller diameter 𝐷
block coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝐵
transverse vertical area above waterline 𝐴𝑉 for air resistance
number of rudders 𝑁rudders  or  (for twin screw vessels)
number of shaft brackets 𝑁brackets ,  or  (for twin screw vessels)
number of shaft bossings 𝑁bossings ,  or  (for twin screw vessels)
number of side thrusters 𝑁thrusters between  and 
optional parameters
wetted surface (hull) 𝑆
wetted surface of appendages 𝑆𝐴𝑃 𝑃𝑖 bilge keels, stabilizer fins, etc.
diameter(s) of transverse thruster tunnels 𝑑𝑇𝐻 for appendage resistance fins, etc.
j j

The draft 𝑇 of the vessel midships is taken as mean value of the drafts at fore and aft
perpendiculars.
𝑇 + 𝑇𝐹
𝑇 = 𝐴 (.)
2

Wetted surface In early design stages the wetted surface 𝑆 is often not yet known. Hollenbach (a)
estimates 𝑆 in two steps: First, a shape factor 𝑘 is computed according to the formula:
( ) ( ) ( )
𝐿𝑂𝑆 𝐿𝑊𝐿 𝐿𝑃𝑃
𝑘 = 𝑠0 + 𝑠1 + 𝑠2 + 𝑠3 𝐶𝐵 + 𝑠4
𝐿𝑊𝐿 𝐿𝑃𝑃 𝐵
( ) ( ) ( ) ( )
𝐵 𝐿𝑃𝑃 𝑇 − 𝑇𝐹 𝐷
+ 𝑠5 + 𝑠6 + 𝑠7 𝐴 + 𝑠8 (.)
𝑇 𝑇 𝐿𝑃𝑃 𝑇
+ 𝑘rudders 𝑁rudders + 𝑘brackets 𝑁brackets + 𝑘bossings 𝑁bossings

Then, the actual wetted surface of the hull is approximated by

( )
𝑆 = 𝑘 𝐿𝑃𝑃 𝐵 + 2𝑇 (.)

The shape factor should not be confused with the form factor used in the ITTC perfor-
mance prediction method. Obviously, the factors 𝑘rudders , 𝑘brackets , and 𝑘bossings are only
used for twin screw vessels. Possible values for number of rudders 𝑁rudders , brackets
𝑁brackets , and bossings 𝑁bossings are listed in Table .. The estimated wetted surface
includes rudder and skeg for single screw vessels.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.2 Resistance Estimate 631

Table 51.3 Coefficients for an estimate of the wetted surface

single propeller vessels twin propeller vessels

ballast design draft design draft w/o
design draft
condition with bulb bulb
𝑠0 −0.6837 0.8037 −0.4319 −0.0887
𝑠1 0.2771 0.2726 0.1685 0.0000
𝑠2 0.6542 0.7133 0.5637 0.5192
𝑠3 0.6422 0.6699 0.5891 0.5839
𝑠4 0.0075 0.0243 0.0033 −0.0130
𝑠5 0.0275 0.0265 0.0134 0.0050
𝑠6 −0.0045 −0.0061 −0.0005 −0.0007
𝑠7 −0.4798 0.2349 −2.7932 −0.9486
𝑠8 0.0376 0.0131 0.0072 0.0506
𝑘rudders — — 0.0131 0.0076
𝑘brackets — — −0.0030 0.0036
𝑘bossings — — 0.0061 0.0049

If other formulas are used for the wetted surface of the hull, the wetted area of ap- Wetted surface
pendages 𝑆𝐴𝑃 𝑃 for twin screw vessels may be approximated based on the following of appendages
formula (Hollenbach, a):
j ( [ ]) j
𝐿 𝑇
𝑆𝐴𝑃 𝑃 = 𝑆 𝑝1 + 𝑝2 exp − 𝑃𝑃 (.)
1000m2

with the following properties and constants

𝑆 = wetted surface of hull without appendages

{
. as upper limit
𝑝1 =
. as lower limit
{
. as upper limit
𝑝2 =
. as lower limit

51.2 Resistance Estimate

Hollenbach’s method divides the resistance into five components.

. frictional resistance 𝑅𝐹 without a form factor,

. residuary resistance 𝑅𝑅 ,
. correlation allowance 𝑅𝐴 ,
. appendage resistance 𝑅𝐴𝑃 𝑃 , and an

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

632 51 Hollenbach’s Method

. environmental resistance 𝑅env .

A dimensionless coefficient is computed for each component.

51.2.1 Frictional resistance coefficient

Frictional The ITTC  model–ship correlation line is used for the frictional resistance coeffi-
resistance cient.
without form 0.075
factor 𝐶𝐹 = [ ]2 (.)
log10 (𝑅𝑒) − 2
The Reynolds number 𝑅𝑒 is computed based on the computational length 𝐿𝑐 (.) and
the kinematic viscosity of seawater 𝜈 = 1.1892 ⋅ 10−6 m/s2 at a temperature of  ◦ C.
𝑣𝑆 𝐿𝑐
𝑅𝑒 = (.)
𝜈

51.2.2 Mean residuary resistance coefficient

Mean residuary The central part of almost any resistance estimate is the determination of the residuary
resistance resistance or wave resistance. Since Hollenbach’s method does not employ a form
factor, the residuary resistance will encompass wave resistance and the major part of
the viscous pressure resistance.
j j
In his regression analysis, Hollenbach (a) found that a better correlation between
formula and data could be found if the residuary resistance coefficient is taken to a basis
of beam times draft (𝐵 𝑇 ) instead of the usual wetted surface 𝑆. Therefore, Hollenbach’s
residuary resistance coefficient is defined as
10 𝑅𝑅
𝐶𝑅𝐵𝑇 = (.)
1
2
𝜌 𝑣2𝑆 𝐵𝑇
The additional scaling factor  adjusts the decimals so that values of 𝐶𝑅𝐵𝑇 usually
range from . to ..
Formulas for 𝐶𝑅𝐵𝑇 have been provided for several cases:

. minimum and mean residuary resistance of single screw vessels at design draft
. mean residuary resistance of single screw vessels at ballast draft
. minimum and mean residuary resistance of twin screw vessels at design draft

The computation of 𝐶𝑅𝐵𝑇 is performed in two steps. First, a standard value 𝐶̃𝑅std is
determined and then multiplied with several correction factors.
𝐶𝑅 for standard The standard value is computed with a parabolic polynomial.
hull form
𝐶̃𝑅std = 𝑏11 + 𝑏12 𝐹𝑁 + 𝑏13 𝐹𝑁2
( )
+ 𝑏21 + 𝑏22 𝐹𝑁 + 𝑏23 𝐹𝑁2 𝐶𝐵 (.)
( )
+ 𝑏31 + 𝑏32 𝐹𝑁 + 𝑏33 𝐹𝑁2 𝐶𝐵2

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.2 Resistance Estimate 633

Table 51.4 Coefficients for computation of the standard residuary resistance coefficient in
Hollenbach’s method

mean residuary resistance minimum residuary resistance

single screw twin screw single screw twin screw
design draft ballast draft design draft design draft design draft
𝑏11 −0.57424 −1.50162 −5.3475 −0.91424 3.27279
𝑏12 13.3893 12.9678 55.6532 13.3893 −44.1138
𝑏13 90.596 −36.7985 −114.905 90.596 171.692
𝑏21 4.6614 5.55536 19.2714 4.6614 −11.5012
𝑏22 −39.721 −45.8815 −192.388 −39.721 166.559
𝑏23 −351.483 121.82 388.333 −351.483 −644.456
𝑏31 −1.14215 −4.33571 −14.3571 −1.14215 12.4626
𝑏32 −12.3296 36.0782 142.738 −12.3296 −179.505
𝑏33 459.254 −85.3741 −254.762 459.254 680.921

The polynomial uses the block coefficient 𝐶𝐵 as a parameter and is a function of the
special Froude number
𝑣
𝐹𝑁 = √ 𝑆 (.)
𝑔𝐿𝑐
j Note that the ITTC recommended length in waterline has been replaced by the compu- j
tational length 𝐿𝑐 from Equation (.).
The necessary coefficients 𝑏𝑖𝑗 in (.) are listed in Table . for all cases. In the case of
single screw vessels at design draft, the coefficient 𝑏11 for the mean residuary resistance
is corrected for vessels with block coefficients smaller than . (Hollenbach, a).

⎧−0.87674 if 𝐶𝐵 < 0.49

⎪ ( )2
𝑏11 = ⎨−0.57424 − 25 0.6 − 𝐶𝐵 if 0.49 ≤ 𝐶𝐵 < 0.6 (.)
⎪−0.57424 see Table .
⎩
Note that this correction applies only to the mean residuary resistance of single screw
vessels on design draft. In all other cases, the 𝑏11 values of Table . apply.
The standard value is then multiplied with eight factors representing additional influ- Correction
ences on residual resistance. factors

𝐶𝑅𝐵𝑇 = 𝐶̃𝑅std ⋅ 𝑘𝐹𝑟 ⋅ 𝑘𝐿 ⋅ 𝑘𝐵𝑇 ⋅ 𝑘𝐿𝐵 ⋅ 𝑘𝐿𝐿 ⋅ 𝑘𝐴𝑂 ⋅ 𝑘𝑇 𝑟 ⋅ 𝑘𝑃 𝑟

𝑎 𝑎 𝑎 𝑎
7
⋅ 𝑁rudders 8
⋅ 𝑁brackets 9
⋅ 𝑁bossings 10
⋅ 𝑁thrusters (.)

The correction factors are:

. High Froude number factor 𝑘𝐹𝑟

For ship speeds exceeding a critical Froude number 𝐹𝑟crit , the wave resistance is
growing faster than predicted by formula (.). With the estimate for the critical
Froude number
𝐹𝑟crit = 𝑑1 + 𝑑2 𝐶𝐵 + 𝑑3 𝐶𝐵2 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

634 51 Hollenbach’s Method

the high Froude number factor is defined as

⎧1 if 𝐹𝑁 < 𝐹𝑟crit
⎪( )
𝑘𝐹𝑟 = ⎨ 𝐹𝑁 𝑐1 (.)
⎪ 𝐹𝑟 if 𝐹𝑁 ≥ 𝐹𝑟crit
⎩ crit

. Length factor 𝑘𝐿
The residuary resistance is still dependent on length
( )
𝐿𝑃𝑃 𝑒2
𝑘𝐿 = 𝑒1 (.)
[m]
𝑘𝐿 obviously has to be dimensionless. Thus, the equation is valid only if the length
between perpendiculars is entered in meters.
. Beam–draft ratio factor 𝑘𝐵𝑇

⎧(1.99)𝑎1 if
𝐵
< 1.99
⎪ 𝑇
𝑘𝐵𝑇 = ⎨( )𝑎 (.)
⎪ 𝐵 𝐵
1

⎩ 𝑇 if ≥ 1.99
𝑇
. Length–beam ratio factor 𝑘𝐿𝐵
j ( )𝑎2 j
⎧ 𝐿𝑃𝑃 if
𝐿𝑃𝑃
≤ 7.11
⎪
𝑘𝐿𝐵 = ⎨ 𝐵 𝐵 (.)
⎪(7.11)𝑎2 𝐿
⎩ if 𝑃𝑃 > 7.11
𝐵
. Wetted length ratio factor 𝑘𝐿𝐿

⎧( 𝐿 )𝑎3 𝐿𝑂𝑆
𝑂𝑆
⎪ if ≤ 1.05
⎪ 𝐿𝑊𝐿 𝐿𝑊𝐿
𝑘𝐿𝐿 = ⎨ (.)
⎪ 𝐿
⎪(1.05)
𝑎3
if 𝑂𝑆 > 1.05
⎩ 𝐿𝑊𝐿

. Aft overhang ratio factor 𝑘𝐴𝑂

⎧(𝐿 )𝑎4 𝐿𝑊𝐿

⎪ 𝑊𝐿 if ≤ 1.06
⎪ 𝐿𝑃𝑃 𝐿𝑃𝑃
𝑘𝐴𝑂 = ⎨ (.)
⎪ 𝐿
⎪(1.06)
𝑎4
if 𝑊𝐿 > 1.06
⎩ 𝐿𝑃𝑃

. Trim correction factor 𝑘𝑇 𝑟

[ ]
𝑇𝐴 − 𝑇𝐹 𝑎5
𝑘𝑇 𝑟 = 1+ (.)
𝐿𝑃𝑃

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.2 Resistance Estimate 635

. Propeller factor 𝑘𝑃 𝑟

⎧(0.43)𝑎6 if
𝐷
< 0.43
⎪ 𝑇𝐴
⎪( )𝑎
⎪ 𝐷 6 𝐷
𝑘𝑃 𝑟 = ⎨ if 0.43 ≤ ≤ 0.84 (.)
⎪ 𝑇𝐴 𝑇𝐴
⎪( ) 𝑎6 𝐷
⎪ 0.84 if > 0.84
⎩ 𝑇𝐴

The additional constants 𝑎1 through 𝑒2 are listed in Table .. Note that the coeffi-
cients 𝑎1 and 𝑎2 have different signs compared with Hollenbach’s papers because the
corresponding ratios in Equations (.) and (.) have been inverted to simplify
notation.
The formulas for the residuary resistance are valid in specific Froude number ranges Range of Froude
𝐹𝑟min ≤ 𝐹𝑁 ≤ 𝐹𝑟max (Hollenbach, ). numbers

( )
lower limit 𝐹𝑟min = min 𝑓1 , 𝑓1 + 𝑓2 (𝑓3 − 𝐶𝐵 ) (.)
upper limit 𝐹𝑟max = 𝑔1 + 𝑔2 𝐶𝐵 + 𝑔3 𝐶𝐵2 (.)

Predictions for Froude numbers outside of the indicated range should be interpreted
with great care and confirmed by other methods.
j j

51.2.3 Minimum residuary resistance coefficient

In essence, the best case scenario resistance estimate relies on the same formulas (.) Minimum
and (.) for 𝐶̃𝑅std and 𝐶𝑅𝐵𝑇 . However, the following modifications to the coefficients residuary
resistance
are made:

. The coefficients 𝑏𝑖𝑗 change; see Table .. In the case of the single screw vessel,
only 𝑏11 changes. All coefficients 𝑏𝑖𝑗 change for the minimum residuary resistance
of twin screw vessels.

. Coefficients 𝑎5 and 𝑎6 vanish, i.e. no corrections for trim and propeller are em-
ployed.
. No correction is made for higher Froude numbers, the factor 𝑘𝐹𝑟 is equal to one:
𝑘𝐹𝑟 = 1.
. Also the length factor is set to 𝑘𝐿 = 1 (𝑒1 = 1 and 𝑒2 = 0 respectively).

The lower and upper limits for the Froude number are calculated with the same formulas
but with changed coefficients (see Table .). For Froude numbers 𝐹𝑁 < 𝐹𝑟min the
lower limit 𝐹𝑟min should be used in the formula for 𝐶̃𝑅std .

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

636 51 Hollenbach’s Method

Table 51.5 Coefficients for correction factors of the standard residuary resistance coefficient in
Hollenbach’s method

mean residuary resistance minimum residuary resistance

single screw twin screw single screw twin screw
design draft ballast draft design draft design draft design draft
𝑎1 0.3382 0.7139 0.2748 0.3382 0.2748
𝑎2 −0.8086 −0.2558 −0.5747 −0.8086 −0.5747
𝑎3 −6.0258 −1.1606 −6.761 −6.0258 −6.761
𝑎4 −3.5632 0.4534 −4.3834 −3.5632 −4.3834
𝑎5 9.4405 11.222 8.8158 0.0 0.0
𝑎6 0.0146 0.4524 −0.1418 0.0 0.0
𝑎7 0.0 0.0 −0.1258 0.0 0.0
𝑎8 0.0 0.0 0.0481 0.0 0.0
𝑎9 0.0 0.0 0.1699 0.0 0.0
𝑎10 0.0 0.0 0.0728 0.0 0.0
( ) ( ) ( )
𝐹𝑁 𝐹𝑁 𝐹𝑁
𝑐1 10 𝐶𝐵 −1 0.0 0.0
𝐹𝑟crit 𝐹𝑟crit 𝐹𝑟crit
j 𝑑1 0.854 0.032 0.897 0.0 0.0 j
𝑑2 −1.228 0.803 −1.457 0.0 0.0
𝑑3 0.497 0.739 0.767 0.0 0.0
𝑒1 2.1701 1.9994 1.8319 1.0 1.0
𝑒2 −0.1602 −0.1446 −0.1237 0.0 0.0

Table 51.6 Factors for lower and upper limit formulas of the range of Froude numbers in which
the 𝐶𝑅 formulas are valid

mean residuary resistance minimum residuary resistance

single screw twin screw single screw twin screw
design draft ballast draft design draft design draft design draft
𝑓1 0.17 0.15 0.16 0.17 0.15
𝑓2 0.20 0.10 0.24 0.20 0.0
𝑓3 0.60 0.50 0.60 0.60 0.0
𝑔1 0.642 0.42 0.83 0.614 0.952
𝑔2 −0.635 −0.20 −0.66 −0.717 −1.406
𝑔3 0.150 0.0 0.0 0.261 0.643

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.2 Resistance Estimate 637

51.2.4 Residuary resistance coefficient

As mentioned above, Hollenbach’s residuary resistance coefficients are based on the Residuary
unusual reference surface 𝐵 𝑇 ∕10. Therefore, the coefficients 𝐶𝑅𝐵𝑇 for mean and resistance
coefficient
minimum residuary resistance have to be rescaled based on the wetted surface 𝑆 before
they can be added to the other resistance coefficients:
𝑅𝑅 𝐵𝑇
𝐶𝑅 = = 𝐶𝑅𝐵𝑇 (.)
1 10 𝑆
2
𝜌 𝑣2𝑆 𝑆

51.2.5 Correlation allowance

A formula developed at the Schiffbautechnische Versuchsanstalt Wien (Vienna Model Correlation

Basin) is used for the correlation allowance. allowance
{( )
0.35 − 0.002𝐿𝑃𝑃 10−3 if 𝐿𝑃𝑃 < 175 m, 𝐿𝑃𝑃 in meter
𝐶𝐴 = (.)
0 if 𝐿𝑃𝑃 ≥ 175 m

The formula yields negative values for ships with a length between perpendiculars larger
than  m. For these cases, we set the correlation allowance to zero.

j 51.2.6 Appendage resistance j

Hollenbach’s method considers two distinct appendage resistance categories:

. frictional and form resistance of bilge keels, stabilizer fins, and exposed shafts,
shaft brackets and bossings especially for twin screw vessels, and
. the resistance of transverse thrusters if present.

The formulas for both resistance components are borrowed from Holtrop and Mennen
().
The viscous resistance of appendages is estimated to Viscous
resistance of
1 2 ( ) appendages
𝑅𝐴𝑃 𝑃 = 𝜌 𝑣𝑆 𝑆𝐴𝑃 𝑃 1 + 𝑘2 𝑒𝑞 𝐶𝐹 (.)
2
The equivalent factor (1 + 𝑘2 )𝑒𝑞 is derived from the weighted sum of individual ap-
∑
pendage contributions and the total wetted area 𝑆𝐴𝑃 𝑃 = 𝑆𝐴𝑃 𝑃𝑖 of all appendages:
∑
(1 + 𝑘2𝑖 )𝑆𝐴𝑃 𝑃𝑖
( ) 𝑖
1 + 𝑘2 𝑒𝑞 = ∑ (.)
𝑆𝐴𝑃 𝑃𝑖
𝑖

Suitable form factors 𝑘2𝑖 for streamlined and flow-oriented appendages are also pro-
vided in Holtrop () and repeated in Table . for completeness. Again, rudder(s)
and skegs are included in the hull surface estimate for Hollenbach’s method. Conse-
quently, they are not considered part of the appendage resistance. Equation (.) uses

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

638 51 Hollenbach’s Method

the same ITTC  model–ship correlation coefficient 𝐶𝐹 (.) as the hull frictional
resistance.
Resistance due The resistance from transverse thrusters may be approximated by Holtrop’s method
to thruster (Holtrop and Mennen, ):
tunnels
𝑅𝑇𝐻 = 𝜌 𝑣2𝑆 𝜋 𝑑𝑇𝐻
2
𝐶𝐷𝑇𝐻 (.)

Resistance coefficients for bow thruster openings 𝐶𝐷𝑇𝐻 are in the range . to ..
Hollenbach suggests the following estimate:
( )
10𝑑𝑇𝐻
𝐶𝐷𝑇𝐻 = 0.003 + 0.003 −1 (.)
𝑇

Finally, the appendage resistance coefficient is derived from the formula:

𝑅𝐴𝑃 𝑃 + 𝑅𝑇𝐻
𝐶𝐴𝑃 𝑃 = (.)
1
2
𝜌𝑣2𝑆 𝑆

51.2.7 Environmental resistance

Air resistance For ship trial conditions, the resistance of a ship moving through air at rest is estimated
according to ITTC recommended procedures (ITTC, ).
j j
𝜌 𝐴
𝐶𝐴𝐴𝑆 = 𝐶𝐷𝐴 air 𝑉 𝑆 (.)
𝜌𝑆

A standard value of 𝐶𝐷𝐴 = 0.8 can be used for the air drag coefficient, if wind tunnel
test data are unavailable.
Additional resistance contributions from wind and waves may be included here as well
for estimates of resistance in service conditions.
[ ]
𝐶env = 𝐶𝐴𝐴𝑆 + 𝐶wind + 𝐶wave (.)

51.2.8 Total resistance

Mean, minimum, The total resistance coefficient for the full scale vessel is given as the sum of the five
and maximum component coefficients:
resistance
𝐶𝑇mean = 𝐶𝐹 + 𝐶𝑅mean + 𝐶𝐴 + 𝐶𝐴𝑃 𝑃 + 𝐶env
𝐶𝑇min = 𝐶𝐹 + 𝐶𝑅min + 𝐶𝐴 + 𝐶𝐴𝑃 𝑃 + 𝐶env (.)
𝐶𝑇max = ℎ1 𝐶𝑇mean

Make sure the residuary coefficient has been converted to its standard form based on
the wetted surface 𝑆. Obviously, we employ the mean residuary resistance coefficient to
form the mean total resistance and use the minimum residuary resistance coefficient for
the minimum total resistance. An estimate for the total maximum resistance coefficient

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.3 Hull–Propeller Interaction Parameters 639

Vessel type Draft Average 𝜂𝑅 Table 51.7 Suggested values for the relative rotative efficiency 𝜂𝑅 , if
the propulsion estimate is based on Wageningen B-Series propeller
single screw design 1.009 data (Hollenbach, 1999)
single screw ballast 1.000
twin screw design 0.981

is obtained by multiplying the mean total resistance coefficient with the factor ℎ1 =
1.204 for a single screw vessel on design draft, ℎ1 = 1.194 for a single screw vessel on
ballast draft, and ℎ1 = 1.206 for a twin screw vessel.
The total calm water resistance is computed in the usual way using the respective total
resistance coefficient.
1
𝑅𝑇 = 𝜌𝑣2𝑆 𝑆 𝐶𝑇 (.)
2
The limiting minimum and maximum resistance values will not be exceeded with a
probability of %.

51.3 Hull–Propeller Interaction Parameters

Keep in mind that the estimated resistance is for trial conditions and for a vessel without
a propulsor. In order to combine the results of the resistance estimate with data from
j open water tests (propeller without ship), we have to specify the changes in hull and j
propeller properties if the propeller is working behind the hull. The parameters used to
describe this hull–propeller interaction are:

. relative rotative efficiency 𝜂𝑅 ,

. thrust deduction fraction 𝑡, and

. wake fraction 𝑤.

Hollenbach found that relative rotative efficiency 𝜂𝑅 and thrust deduction fraction 𝑡
showed very little correlation to the main dimensions of the vessel. They depend on
shape details of the stern of a vessel.

51.3.1 Relative rotative efficiency

Due to the lack of correlation between relative rotative efficiency and the ship’s principal Relative rotative
dimensions, Hollenbach proposed constant values for 𝜂𝑅 based on ship type and draft. efficiency
Hollenbach (a) notes that the efficiency of stock propellers used in the propulsion
tests influence the result. If the stock propeller has less efficiency than a comparable
Wageningen B-Series propeller, relative rotative efficiency becomes larger and vice
versa.
In case the propulsion estimate is based on a Wageningen B-Series propeller, the values
of Table . can be used for the relative rotative efficiency.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

640 51 Hollenbach’s Method

Table 51.8 Suggested values for the thrust deduction fraction 𝑡 (Hollenbach, 1999)

Vessel type Condition Appendage arrangement 𝑡 Std. deviation

single screw design – 0.190 .%
single screw ballast – 0.195 .%
twin screw design twin rudder, shaft brackets 0.150 .%
twin screw design twin rudder, twin skegs 0.186 .%
twin screw design single rudder, shaft brackets 0.130 .%
twin screw design single rudder, shaft bossings 0.113 .%

51.3.2 Thrust deduction fraction

Thrust deduction Since no correlation between main dimensions and the thrust deduction fraction could
fraction be found from the model test results, Hollenbach () suggests using a constant value
for 𝑡 in early design stages according to Table ..
Since the thrust deduction fraction does not appear to show significant viscous effects,
the value of 𝑡 is applied to the full scale vessel without corrections. The stated standard
deviations for the thrust deduction fraction indicate that there is considerable uncer-
tainty in the values. However, this by no means indicates that more complex formulas –
like the ones presented by Holtrop () – are more accurate.

j 51.3.3 Wake fraction j

Hull efficiency of Hollenbach does not include explicit formulas for the wake fraction 𝑤. Instead, he
model developed formulas for the hull efficiency 𝜂𝐻𝑀 at model scale. The hull efficiency
combines effects of wake fraction 𝑤𝑇𝑀 and thrust deduction fraction 𝑡.
1−𝑡
𝜂𝐻𝑀 = (.)
1 − 𝑤𝑇𝑀
According to Hollenbach’s method, hull efficiency of models for single screw vessels on
design draft is given by
( ) ( )
𝑅𝑇 mean −0.58( 𝐵 )0.1727 𝐷2 −0.1334
𝜂𝐻𝑀 = 0.948 𝐶𝐵0.3977 (design) (.)
𝑅𝑇 𝑇 𝐵𝑇

and on ballast condition by:

( )0.2991( )−3.2806( )−0.2317
𝐿𝑃𝑃 𝐿𝑊𝐿 𝐷
𝜂𝐻𝑀 = 1.055 𝐶𝐵1.0099 (ballast) (.)
𝐵 𝐿𝑃𝑃 𝑇

For twin screw vessels in design condition, the following formula should be used.
( )−0.0285
𝐷
𝜂𝐻𝑀 = 𝐶 𝐶𝐵0.1202 (twin screw) (.)
𝑇
In the formulas above, 𝐵 is the ship molded beam, 𝐷 the propeller diameter, 𝐿𝑃𝑃 the
length between perpendiculars, 𝐿𝑊𝐿 the length in the water line, and 𝑇 the mean
molded draft.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.3 Hull–Propeller Interaction Parameters 641

Table 51.9 Suggested values for constant 𝐶 for the hull efficiency of twin screw vessel models
(Hollenbach, 1999)

Vessel type Condition Appendage arrangement 𝐶 value

twin screw design twin rudder, shaft brackets 1.125
twin screw design twin rudder, twin skegs 1.224
twin screw design single rudder, shaft brackets 1.086
twin screw design single rudder, shaft bossings 1.096

The factor 𝑅𝑇 mean ∕𝑅𝑇 in the hull efficiency estimate for the single screw vessels is
meant to adjust the propulsion prediction for varying confidence in the resistance
estimate. If there is an indication (by comparing with similar vessels or other methods)
that the actual resistance is closer to 𝑅𝑇 minimum rather than the estimate 𝑅𝑇 mean , the
factor becomes 𝑅𝑇 mean ∕𝑅𝑇 minimum > 1. If you think 𝑅𝑇 mean is too optimistic, the
factor could be 𝑅𝑇 mean ∕𝑅𝑇 maximum < 1 or anything in between. If no additional data is
available, use 𝑅𝑇 mean ∕𝑅𝑇 = 1.
The factor 𝐶 in the 𝜂𝐻𝑀 formula for twin screw vessels should be taken from Table ..

Once the hull efficiency of the model has been determined, we compute the wake Wake fraction of
fraction 𝑤𝑇𝑀 of the model. Solving Equation (.) for the wake fraction yields model

1−𝑡
𝑤𝑇𝑀 = 1 − (.)
j 𝜂𝐻𝑀 j
The additional subscript 𝑇 for the wake fraction is a reference to thrust identity which
is commonly assumed for the assessment of self-propulsion tests.
In contrast to the thrust deduction fraction, wake fraction is affected by the viscous Wake fraction of
boundary layer of the ship hull. Therefore, the model value has to be adjusted for the full full scale ship
scale vessel. Following the formula provided by the ITTC  Performance Prediction
Method, the full scale wake fraction 𝑤𝑇𝑆 is given as
( ) [𝐶 + 𝐶 ]
𝐹𝑆 𝐴
𝑤𝑇𝑆 = (𝑡 + 0.04) + 𝑤𝑇𝑀 − (𝑡 + 0.04) (.)
𝐶𝐹𝑀
The thrust deduction fraction is used to represent the potential wake fraction 𝑤𝑝 . The
summand . accounts ( for the influence
) of a rudder onto the wake (ITTC suggested
value). The difference 𝑤𝑇𝑀 − (𝑡 + 0.04) is an approximation of the frictional wake 𝑤𝑓
𝐶𝐹𝑆 +𝐶𝐴
which is scaled by the ratio of frictional forces 𝐶𝐹𝑀
of full scale vessel and model.

Obviously, we need a Reynolds number 𝑅𝑒𝑀 for the model to compute the ITTC  ITTC 1957
model–ship correlation coefficient 𝐶𝐹𝑀 for the model. According to Hollenbach, the model–ship
correlation line
Reynolds number is based on length in waterline and kinematic viscosity of fresh water
at  ◦ C temperature.

0.075 𝑣𝑀 𝐿𝑊𝐿𝑀
𝐶𝐹𝑀 = ( )2 with 𝑅𝑒𝑀 = (.)
log10 (𝑅𝑒𝑀 ) − 2 𝜈𝑀

Hollenbach () states that the equations (.), (.), and (.) are for models
of length 𝐿𝑃𝑃 𝑀 = 6.5 m.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

642 51 Hollenbach’s Method

Using the actual length between perpendiculars 𝐿𝑃𝑃 of our ship project, we derive the
scale of our virtual model
𝐿𝑃𝑃 𝐿𝑃𝑃
𝜆virtual = = (.)
𝐿𝑃𝑃𝑀 6.5 m

Based on the virtual scale, we obtain length in waterline and speed of the virtual model.

𝐿𝑊𝐿𝑀 = 𝐿𝑊𝐿 ∕𝜆virtual (.)

√
𝑣𝑀 = 𝑣𝑆 ∕ 𝜆virtual (.)

Assuming fresh water at a temperature of  ◦ C, we take the kinematic viscosity of the

ITTC water property tables and compute 𝑅𝑒𝑀 , 𝐶𝐹𝑀 , and the full scale wake fraction
𝑤𝑇𝑆 (.).
Hull efficiency of From the full scale wake fraction follows the full scale hull efficiency.
full scale vessel
1−𝑡
𝜂𝐻𝑆 = (.)
1 − 𝑤𝑇𝑆

51.4 Resistance and Propulsion Estimate Example

j j
Finally, we repeat the resistance and propulsion estimate for the  TEU container
ship from Sections . and . using the formulas presented in this chapter.

51.4.1 Completion of input parameters

Lengths The principal dimensions for the example are listed in Tables . and .. As stated
in Equation (.) in the previous chapter, length over water line is 𝐿𝑊𝐿 = 147.7 m.
Length over wetted surface is 𝐿𝑂𝑆 = 151.0 m. The calculation length 𝐿𝑐 (.) is
𝐿𝑐 = 149.0 m and is used to compute the Froude number 𝐹𝑁 (.).
Dimensionless The block coefficient of the vessel based on its length between perpendiculars 𝐿𝑃𝑃 is
parameters
𝑉
𝐶𝐵 = = 0.6613 (.)
𝐿𝑃𝑃 𝐵𝑇

The following dimensionless ratios apply:

𝐿𝑃𝑃 𝐵
= 6.0417 = 2.9268
𝐵 𝑇
(.)
𝐷 𝐿𝑃𝑃
= 0.5972 √ = 5.5668
𝑇𝐴 3
𝑉

In its design condition the ship floats on even keel; therefore, 𝑇𝐴 = 𝑇 . The resulting
ratios fall within the limits stated in Table ..

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.4 Resistance and Propulsion Estimate Example 643

Table 51.10 Residuary resistance coefficients for minimum and mean resistance cases

Constant factors
𝑘𝐿 𝑘𝐵𝑇 𝑘𝐵𝐿 𝑘𝐿𝐿 𝑘𝐴𝑂 𝑘𝑃 𝑟 𝑘𝑇 𝑟
Minimum 𝐶𝑅𝐵𝑇 factors: . . . . . . .
Mean 𝐶𝑅𝐵𝑇 factors: . . . . . . .
Speed dependent results
minimum residuary resistance mean residuary resistance
𝑣𝑆 𝐹𝑁 𝐶𝑅std 𝑘𝐹𝑟 𝐶𝑅𝐵𝑇 𝐶𝑅min 𝐶𝑅std 𝑘𝐹𝑟 𝐶𝑅𝐵𝑇 𝐶𝑅
[kn] [−] [−] [−] [−] [10−3 ] [−] [−] [−] [10−3 ]
15.0 0.2019 0.3851 1.0000 0.1060 0.4741 0.7251 1.0000 0.1937 0.8663
15.5 0.2086 0.4251 1.0000 0.1170 0.5234 0.7651 1.0000 0.2044 0.9141
16.0 0.2153 0.4705 1.0000 0.1295 0.5792 0.8105 1.0000 0.2165 0.9683
16.5 0.2221 0.5212 1.0000 0.1435 0.6417 0.8612 1.0000 0.2300 1.0289
17.0 0.2288 0.5773 1.0000 0.1589 0.7107 0.9173 1.0000 0.2450 1.0959
17.5 0.2355 0.6387 1.0000 0.1758 0.7863 0.9787 1.0000 0.2614 1.1692
18.0 0.2422 0.7055 1.0000 0.1942 0.8685 1.0455 1.0000 0.2792 1.2490
18.5 0.2490 0.7775 1.0000 0.2140 0.9572 1.1175 1.0000 0.2985 1.3351
19.0 0.2557 0.8550 1.0000 0.2353 1.0526 1.1950 1.0000 0.3192 1.4276

j j
For the calculations we use the given wetted surface of 𝑆 = 4400.0 m2 . The wetted Surfaces
surface estimate according to Equations (.) and (.) yields:

𝑘 = 0.75938 𝑆 = 4448.453 m2 (estimate) (.)

The latter deviates from the actual value by just .%. For the transverse area above the
water line, we use again 𝐴𝑉 = 383.76 m2 .

51.4.2 Powering estimate

The resistance estimate starts with the computation of residuary resistance coefficients. Residuary
Unlike other procedures, Hollenbach’s method provides values for a best scenario resistance
minimum and a mean residuary resistance coefficient. The necessary formulas and
coefficients are stated in Sections .., .., and ... The Froude numbers 𝐹𝑁
stay within the limits provided by Equations (.) and (.) for the mean residuary
resistance estimate: 0.1577 < 𝐹𝑁 < 0.2877. In the best case scenario (minimum) 𝐶𝑅 ,
the highest Froude number exceeds the respective upper limit by less than %. We will
ignore this because it applies only to values above the design speed. Table . lists the
constant and speed dependent factors and final residuary resistance coefficient values
for the example ship.
Equations (.) and (.) are employed to estimate correlation allowance and the Correlation
air resistance. allowance and
environmental
resistance
𝐶𝐴 = 0.06 ⋅ 10−3 𝐶env = 𝐶𝐴𝐴𝑆 = 0.0833 ⋅ 10−3 (.)

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

644 51 Hollenbach’s Method

Table 51.11 Resistance coefficients for example vessel by Hollenbach’s method

𝑣𝑆 𝑣𝑆 𝐹𝑟 𝐶𝑅min 𝐶𝑅 𝐶𝐹 𝐶𝐴𝑃 𝑃 𝐶𝑇min 𝐶𝑇 𝐶𝑇max

[kn] [m∕s] [−] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ] [10−3 ]
15.0 7.717 0.2019 0.4741 0.8663 1.5370 0.0254 2.1799 2.5720 3.0967
15.5 7.974 0.2086 0.5234 0.9141 1.5308 0.0253 2.2228 2.6135 3.1467
16.0 8.231 0.2153 0.5792 0.9683 1.5248 0.0252 2.2725 2.6616 3.2046
16.5 8.488 0.2221 0.6417 1.0289 1.5190 0.0251 2.3291 2.7163 3.2704
17.0 8.746 0.2288 0.7107 1.0959 1.5134 0.0250 2.3924 2.7776 3.3442
17.5 9.003 0.2355 0.7863 1.1692 1.5080 0.0250 2.4625 2.8455 3.4260
18.0 9.260 0.2422 0.8685 1.2490 1.5028 0.0249 2.5394 2.9199 3.5156
18.5 9.517 0.2490 0.9572 1.3351 1.4977 0.0248 2.6230 3.0009 3.6131
19.0 9.774 0.2557 1.0526 1.4276 1.4928 0.0247 2.7134 3.0885 3.7185
Total resistance coefficients include constant 𝐶env = 0.0833 ⋅ 10−3 .

Table 51.12 Comparison of total calm water resistance estimates

Guldhammer Holtrop
Hollenbach and Harvald and Mennen
Table . Table .
𝑣𝑆 𝐹𝑟 𝑅𝑒 𝑅𝑇min 𝑅𝑇 𝑅𝑇max 𝐹𝑟 𝑅𝑇 𝐹𝑟 𝑅𝑇
[kn] [−] [−] [kN] [kN] [kN] [−] [kN] [−] [kN]
j j
15.0 0.2019 966854468.0 293.00 345.71 416.24 0.2005 335.62 0.2028 386.98
15.5 0.2086 999082950.3 319.02 375.10 451.62 0.2072 362.25 0.2095 419.21
16.0 0.2153 1031311432.5 347.54 407.04 490.08 0.2139 391.03 0.2163 454.37
16.5 0.2221 1063539914.8 378.80 441.78 531.90 0.2205 422.36 0.2230 492.70
17.0 0.2288 1095768397.1 413.04 479.54 577.37 0.2272 456.82 0.2298 534.07
17.5 0.2355 1127996879.3 450.52 520.58 626.78 0.2339 495.13 0.2366 578.14
18.0 0.2422 1160225361.6 491.51 565.16 680.45 0.2406 538.28 0.2433 625.06
18.5 0.2490 1192453843.9 536.29 613.56 738.72 0.2473 592.72 0.2501 675.88
19.0 0.2557 1224682326.1 585.16 666.05 801.92 0.2540 662.39 0.2568 732.51

Both are assumed to be independent of ship speed. An appendage resistance component

has been added for the bilge keels. For the formula (.) provided by Holtrop and
Mennen (), an area of 𝑆𝐴𝑃 𝑃 = 52.0 m2 and a form factor of 𝑘2 = 0.4 has been used.
Subsequently, the dimensional appendage resistance is converted into a dimensionless
coefficient with Equation (.).
Resistance The estimated resistance coefficients are listed in Table .. The maximum total resis-
coefficients tance coefficient has been obtained by multiplying the mean total resistance coefficient
with the factor ℎ1 = 1.204 (see Section ..).
Comparison of Table . and Figure . compare the total resistance estimates derived with the three
resistance methods presented in this book: Hollenbach, Guldhammer and Harvald (Chapter ),
estimates
and Holtrop and Mennen (Chapter ). The Froude numbers vary slightly because each
method uses a different reference length. All estimates fall within the range defined by
the minimum and the maximum resistance predicted by Hollenbach’s method. Holtrop

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.4 Resistance and Propulsion Estimate Example 645

900
mean RT Hollenbach
800 RT Guldhammer and Harvald
RT Holtrop and Mennen
RTmin − RTmax range Hollenbach
[kN]

700
total resistance RT

600

500

400
results for
design speed 17.5 kn
300

200
0.20 0.21 0.22 0.23 0.24 0.25 0.26
Froude number Fr [−]

j Figure 51.1 Comparison of total resistance estimates for the methods by Hollenbach, j
Guldhammer and Harvald, and Holtrop and Mennen

and Mennen’s estimate is higher than the mean total resistance by Hollenbach. Guld-
hammer and Harvald’s estimate is close to Hollenbach’s but features a steeper increase
at the end of the Froude number range. At the design speed of . kn, Guldhammer and
Harvald’s estimate is actually .% lower than Hollenbach’s. However, due to the longer
reference length and, consequently, slightly lower Froude numbers, Guldhammer and
Harvald’s resistance curve appears to lie above Hollenbach’s mean resistance curve.
Holtrop and Mennen’s resistance estimate lies about % above Hollenbach’s for the
design speed. These deviations are fairly typical for different methods.
Hollenbach’s method defines constant values for thrust deduction fraction 𝑡 and relative Thrust deduction
rotative efficiency 𝜂𝑅 . From Tables . and . we retrieve fraction, relative
rotative
𝜂𝑅 = 1.009 and 𝑡 = 0.19 efficiency

The powering estimate with Holtrop and Mennen’s method resulted in a relative rotative
efficiency of 𝜂𝑅 = 0.9940 and a thrust deduction fraction of 𝑡 = 0.1983 (Section ..).
Andersen and Guldhammer () use a relative rotative efficiency of 𝜂𝑅 = 1.0 and
with Equation (.) follows a thrust deduction fraction of 𝑡 = 0.2019. Although, the
values of the thrust deduction vary quite a bit between the three methods, the resulting
factor 1∕(1 − 𝑡) for the required thrust varies only by about .%.
The length of water line of a virtual model is needed for the estimate of the full scale Full scale wake
wake fraction. According to Equation (.), a virtual model scale 𝜆virtual = 22.3077 fraction
is derived for a length between perpendiculars of 𝐿𝑃𝑃 = 145 m. Then, the waterline

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

646 51 Hollenbach’s Method

Table 51.13 Prediction of model and full scale wake fraction

H and M
Hollenbach Table
.
𝑣𝑆 𝑣𝑆 𝑣𝑀 𝑅𝑒𝑀 𝐶𝐹 +𝐶𝐴 𝐶𝐹𝑀 𝑤𝑇𝑆 𝑤𝑇𝑆
[kn] [m∕s] [m∕s] [−] [10−3 ] [10−3 ] [−] [−]
15.0 7.717 1.634 9500732.4 1.59703 3.02687 0.30285 0.26959
15.5 7.974 1.688 9817423.5 1.59079 3.00963 0.30298 0.26942
16.0 8.231 1.743 10134114.6 1.58477 2.99307 0.30311 0.26926
16.5 8.488 1.797 10450805.7 1.57898 2.97715 0.30323 0.26911
17.0 8.746 1.852 10767496.8 1.57339 2.96183 0.30335 0.26896
17.5 9.003 1.906 11084187.9 1.56799 2.94707 0.30346 0.26882
18.0 9.260 1.961 11400878.9 1.56277 2.93282 0.30357 0.26868
18.5 9.517 2.015 11717570.0 1.55772 2.91907 0.30368 0.26854
19.0 9.774 2.069 12034261.1 1.55283 2.90578 0.30378 0.26841

length of the virtual model is 𝐿𝑊𝐿𝑀 = 6.6210 m. Model speeds 𝑣𝑀 that correspond
to the investigated full scale speeds follow from Equation (.). Based on 𝐿𝑊𝐿𝑀
and 𝑣𝑀 , Reynolds numbers for the model 𝑅𝑒𝑀 are computed, followed by the friction
coefficient 𝐶𝐹𝑀 for the model (.). Once 𝐶𝐹𝑀 is known, the model wake fraction is
corrected for the full scale vessel via Equation (.).
j j
The hull efficiency of the model is treated as a speed independent constant as long as
the factor 𝑅𝑇 mean ∕𝑅𝑇 is constant. Lacking other guidance, we use 𝑅𝑇 mean ∕𝑅𝑇 = 1.
The estimated hull efficiency and wake fraction at model scale are:

𝜂𝐻𝑀 = 1.28179 𝑤𝑇𝑀 = 0.36807

Table . summarizes the remaining values for the full scale wake fraction prediction
based on Equation (.). For comparison, the last column in Table . repeats the
result from Table . obtained by Holtrop and Mennen’s method. Equation (.) by
Andersen and Guldhammer () yields a considerably smaller full scale wake fraction
of 𝑤𝑆 = 0.21661 if the model value is converted with the suggested factor of . which
applies to the ship trial condition.
Self propulsion In Hollenbach’s method the full scale power propulsion analysis of Section . is
point repeated three times because the required thrust will be different for minimum, mean,
and maximum resistance cases. Detailed results are presented for the mean resistance
case only. Table . summarizes the data for the self propulsion point analysis based
on the estimated mean total resistance. Characteristics of a Wageningen B-Series
propeller are assumed, and its principal dimensions are listed in Table .. The self
propulsion points change if the analysis is based on the minimum or maximum total
resistance curves.
Delivered power Once the self propulsion points are known, Equations (.) through (.) yield rate
of revolution, torque, delivered power, and efficiencies. Table . shows the values
of torque, rate of revolution, and delivered power for the self propulsion points from
Table .. Table . shows the corresponding system efficiencies. A quasi-propulsive

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.4 Resistance and Propulsion Estimate Example 647

Table 51.14 Self propulsion point based on mean resistance curve

𝑣𝑆 𝑣𝑆 𝐹𝑟 𝑤𝑇𝑆 𝑣𝐴 𝐶𝑆 𝐽𝑇𝑆 𝐾𝑇𝑆 10𝐾𝑄𝑇𝑆

[kn] [m∕s] [−] [−] [m∕s] [−] [−] [−] [−]
15.0 7.717 0.2019 0.3028 5.380 0.59864 0.6163 0.2274 0.3754
15.5 7.974 0.2086 0.3030 5.558 0.60852 0.6133 0.2289 0.3774
16.0 8.231 0.2153 0.3031 5.736 0.61995 0.6099 0.2306 0.3796
16.5 8.488 0.2221 0.3032 5.914 0.63291 0.6061 0.2325 0.3821
17.0 8.746 0.2288 0.3033 6.093 0.64741 0.6019 0.2345 0.3848
17.5 9.003 0.2355 0.3035 6.271 0.66345 0.5973 0.2367 0.3878
18.0 9.260 0.2422 0.3036 6.449 0.68103 0.5925 0.2391 0.3909
18.5 9.517 0.2490 0.3037 6.627 0.70014 0.5874 0.2416 0.3942
19.0 9.774 0.2557 0.3038 6.805 0.72078 0.5820 0.2442 0.3976

Table 51.15 Prediction of rate of revolution and delivered power for trial condition based on
mean resistance curve

𝑣𝑆 𝑣𝑆 𝐹𝑟 𝑇 𝑄 𝑛 𝑛 𝑃𝐷
[kn] [m∕s] [−] [kN] [kNm] [1∕s] [rpm] [kW]
15.0 7.717 0.2019 426.80 342.15 1.781 106.877 3829.36
15.5 7.974 0.2086 463.08 370.76 1.849 110.964 4308.25
16.0 8.231 0.2153 502.52 401.76 1.919 115.169 4845.40
j 16.5 8.488 0.2221 545.40 435.36 1.992 119.497 5447.98 j
17.0 8.746 0.2288 592.02 471.78 2.066 123.955 6123.89
17.5 9.003 0.2355 642.69 511.23 2.142 128.546 6881.85
18.0 9.260 0.2422 697.73 553.96 2.221 133.276 7731.49
18.5 9.517 0.2490 757.48 600.22 2.302 138.149 8683.37
19.0 9.774 0.2557 822.28 650.25 2.386 143.170 9749.06

Table 51.16 Predicted efficiencies based on mean resistance curve

𝑣 𝑣 𝐹𝑟 𝜂𝑂 𝜂𝐵 𝜂𝐻 𝜂𝐷
[kn] [m/s] [−] [−] [−] [−] [−]
15.0 7.717 0.2019 0.5942 0.5996 1.1619 0.6967
15.5 7.974 0.2086 0.5921 0.5974 1.1621 0.6942
16.0 8.231 0.2153 0.5896 0.5949 1.1623 0.6915
16.5 8.488 0.2221 0.5868 0.5921 1.1625 0.6883
17.0 8.746 0.2288 0.5837 0.5890 1.1627 0.6848
17.5 9.003 0.2355 0.5804 0.5856 1.1629 0.6810
18.0 9.260 0.2422 0.5768 0.5820 1.1631 0.6769
18.5 9.517 0.2490 0.5729 0.5781 1.1633 0.6725
19.0 9.774 0.2557 0.5689 0.5740 1.1634 0.6678

efficiency of .% is achieved at the design speed of . kn. This could possibly be
improved by designing a wake adapted propeller using lifting line theory and other
methods.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

648 51 Hollenbach’s Method

Table 51.17 Comparison of predicted rate of revolution and delivered power

Guldhammer Holtrop
Hollenbach and and Mennen
Harvald Table .
𝑣𝑆 𝑛min 𝑃𝐷min 𝑛mean 𝑃𝐷mean 𝑛max 𝑃𝐷max 𝑛 𝑃𝐷 𝑛 𝑃𝐷
[kn] [rpm] [kW] [rpm] [kW] [rpm] [kW] [rpm] [kW] [rpm] [kW]
15.0 101.84 3134.85 106.88 3829.36 113.18 4816.33 111.81 4067.26 112.99 4637.23
15.5 105.81 3540.31 110.96 4308.25 117.54 5420.96 115.88 4545.39 117.28 5208.37
16.0 109.91 3998.65 115.17 4845.40 122.03 6099.78 120.05 5077.00 121.70 5850.65
16.5 114.14 4516.78 119.50 5447.98 126.66 6861.99 124.34 5671.91 126.27 6573.49
17.0 118.52 5102.39 123.96 6123.89 131.44 7717.78 128.79 6343.20 130.95 7378.31
17.5 123.04 5764.01 128.55 6881.85 136.36 8678.38 133.42 7108.52 135.70 8263.98
18.0 127.71 6511.05 133.28 7731.49 141.44 9756.18 138.31 7992.05 140.54 9239.76
18.5 132.53 7353.89 138.15 8683.37 146.68 10964.81 143.86 9121.81 145.52 10328.13
19.0 137.52 8303.96 143.17 9749.06 152.08 12319.22 150.28 10598.40 150.75 11573.40

14000
mean PD Hollenbach
PD Guldhammer and Harvald
12000
j PD Holtrop and Mennen j
PDmin − PDmax range Hollenbach
[kW]

10000
delivered power PD

8000

6000
results for
design speed 17.5 kn
4000

2000
100 110 120 130 140 150 160
rate of revolution n [rpm]

Figure 51.2 Comparison of predicted rate of revolution and delivered power for the methods
by Hollenbach, Guldhammer and Harvald, and Holtrop and Mennen

Comparison Table . and Figure . present a comparison of the estimated rate of revolution and
delivered power for the prediction methods discussed in this book. Although the curves
are close together there are differences, especially in the predicted rate of revolution

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

51.4 Resistance and Propulsion Estimate Example 649

for each speed. Circles mark the values predicted for the design speed. Holtrop and
Mennen’s method predicts the highest delivered power of 𝑃𝐷 = 9414.4 kW at a rate
of revolution of 𝑛 = 170.53 rpm. Hollenbach’s estimate is the most optimistic with a
delivered power of . kW at . rpm.
An engine may be selected based on the powering prediction. The delivered power is Engine selection
converted into the engine brake power 𝑃𝐵 via Equations (.) and (.). Proper sea
and engine margins have to be added to the brake power predicted for trial conditions.
The final combination of rate of revolution and brake power is matched with the engine
layout diagram. This is a marine engineering rather than a hydrodynamic problem. The
reader can find details in the engine selection guides published by engine manufacturers.
The spread of the predicted power values is an indication of the uncertainty intrinsic Conclusion
to resistance and propulsion estimates used in early design phases. Better results can
hardly be expected since only a few form parameters are used to describe the hull shape.
Too much of the flow patterns depends on details of the hull geometry, which will not
be known until the lines plan is completed. Once the lines are faired, a model may be
manufactured and tested. The hull geometry may also serve as the starting point for a
CFD analysis if computational resources and expertise are available.

References
j Andersen, P. and Guldhammer, H. (). A computer-oriented power prediction j
procedure. In Proc. of Int. Conf. on Computer Aided Design, Manufacture, and
Operation in the Marine and Offshore Industries (CADMO ’), Washington, DC,
USA.
Hollenbach, K. (). Verfahren zur Abschätzung von Widerstand und Propulsion
von Ein- und Zweischraubenschiffen im Vorentwurf. In Jahrbuch der Schiffbautech-
nischen Gesellschaft, volume , pages –. Schiffbautechnische Gesellschaft
(STG).
Hollenbach, K. (a). Beitrag zur Abschätzung von Widerstand und Propulsion von
Ein- und Zweischraubenschiffen im Vorentwurf. PhD thesis, Institut für Schiffbau,
Universität Hamburg, Hamburg, Germany.
Hollenbach, K. (b). Beitrag zur Abschätzung von Widerstand und Propulsion
von Ein- und Zweischraubenschiffen im Vorentwurf. IfS Report , Institut für
Schiffbau, Universität Hamburg, Hamburg, Germany.
Hollenbach, K. (a). Estimating resistance and propulsion for single-screw and
twin-screw ships. Schiffstechnik/Ship Technology Research, ():–.
Hollenbach, K. (b). Weiterentwicklung eines verfahrens zur Abschätzung von
Widerstand und Propulsion von Ein- und Zweischraubenschiffen im Vorentwurf. In
Jahrbuch der Schiffbautechnischen Gesellschaft, volume , pages –. Berlin.
Hollenbach, K. (). Estimating resistance and propulsion for single-screw and twin-
screw ships in the preliminary design. In Proc. of th Int. Conference on Computer
Applications in Shipbuilding (ICCAS ’).
Holtrop, J. (). A statistical re-analysis of resistance and propulsion data. Interna-
tional Shipbuilding Progress, ():–.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

650 51 Hollenbach’s Method

Holtrop, J. (). A statistical resistance prediction method with a speed depen-

dent form factor. In Scientific and Methodological Seminar on Ship Hydrodynamics
(SMSSH ’), Varna, Bulgaria.
Holtrop, J. and Mennen, G. (). An approximate power prediction method. Inter-
national Shipbuilding Progress, ():–.
ITTC ().  ITTC performance prediction method. International Towing Tank
Conference, Recommended Procedures and Guidelines .---.. Revision .

Self Study Problems

. Discuss the principal differences and similarities between the resistance estimates
based on Guldhammer and Harvald’s method, Holtrop and Mennen’s method,
and Hollenbach’s method.
. For a ship design project the following data is provided.

Ship data
length between perpendiculars 𝐿𝑃𝑃 = . m
length in waterline 𝐿𝑊𝐿 = . m
molded beam 𝐵 = . m
j molded draft 𝑇 = . m j
block coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝐵 = .
prismatic coefficient (based on 𝐿𝑃𝑃 ) 𝐶𝑃 = .

Compute the input values for block coefficient 𝐶𝐵 and prismatic coefficient
𝐶𝑃 for Hollenbach’s method. Compare the results with the values from the
corresponding problem at the end of Chapter .
. Implement Hollenbach’s resistance and propulsion estimate as a program in
Python, Matlab, or similar, and test it with the data presented in the last section.

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

651

Index

A Blockage 
Actuator disk  correction , , 
Added mass  factor 
Admiralty coefficient  Schuster’s correction 
Advance coefficient , , , Tamura’s correction 
, –,  Boiling point 
Airy, Sir George Biddell  Bollard pull 
American Towing Tank Conference Boundary condition 
see ATTC body , , , , ,
Angle of attack , , , , , , , , 
,  Dirichlet 
effective  far field , , , 
ideal –,  free surface see Free surface
induced  kinematic , , 
j j
zero lift ,  mixed 
Appendage  Neumann , 
Archimedes  ocean bottom 
Archimedes’ principle , , , , Boundary layer , , , , 
 buffer layer 
ATTC  equations 
Averaging  inner law 
inner scaling 
B laminar 
Behind condition , , , , log–wake law 
, , ,  logarithmic overlap law 
adjustment  modified log–wake law 
Bernoulli equation , , no slip condition 
–, , , ,  outer scaling 
linearized , , , , overlap layer , 
 separation 
potential flow –, , thickness , , , 
, ,  turbulent , –, 
steady flow , , ,  viscous sublayer , , 
unsteady flow , ,  wall layer 
Bertrand, Joseph  wall shear stress 
Bilge keels  Boundary layer theory , –
Biot, Jean-Baptiste  assumptions 
Biot–Savart law – Boundary value problem 
Blade see Propeller, blade cylinder 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

652 Index

displacement flow (thin foil) Chord length , , , ,
,  
lifting flow (thin foil) ,  Circulation , , , 
linear wave theory  bound , 
moving cylinder ,  free 
thin foil  Coefficient
Boussinesq’s block , , , , ,
eddy viscosity hypothesis  , , , , , ,
Boussinesq, Joseph V.  , 
British method see Propulsion test, midship section 
load variation prismatic , , , ,
Buckingham 𝜋-theorem  , 
Buckingham, Edgar  waterplane area 
Bulbous bow , , , ,  Collocation point 
resistance ,  Computational Fluid Dynamics see
Burrill CFD
% back cavitation criterion Condition
, , ,  behind , , , 
cavitation chart ,  calm water , , 
Burrill, Lennard Constantine  loading , , 
open water , , , , ,
, , , , , ,
C 
j Camber , , ,  j
service , 
maximum  trial , , 
Cauchy Conformal mapping 
principal value integral Conservation
–,  of mass , , , , , ,
Cauchy, Augustin-Louis  , , 
Cavitation ,  of momentum –, , ,
bubble ,  , , , 
cloud  integral form 
effects – Conservative 
face  Continuity equation –, , ,
hub vortex  , , 
inception  differential, conservative , 
prevention – differential, nonconservative 
propeller–hull  incompressible, steady flow ,
sheet  
test  integral form 
tip vortex  integral, nonconservative 
tunnel  integral, conservative 
Cavitation criterion see Burrill Contraction nozzle , 
Cavitation number  Control volume 
free stream  differential , 
propeller ,  fixed 
CFD , , , , ,  moving 
Chapman, Fredrik H. af  finite see Control volume,

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

Index 653

integral Divergence –, , , , , ,

infinitesimally small see 
Control volume, differential theorem , 
integral , ,  Downwash , , 
fixed  Drag , , , 
moving  coefficient , , , ,
Coordinate system , , 
Cartesian xvii, ,  induced 
cylindrical 
polar , 
E
Efficiency 
spherical 
behind , 
Correlation lines 
gearing 
Cupping see Propeller blade,
hull , , , 
cupping
ideal , , 
open water , , ,
D –, , , , ,
d’Alembert’s paradox , , , 
 quasi-propulsive , 
d’Alembert, Jean-Baptiste le Rond relative rotative , ,
,  –, , , , 
Density  shafting 
of air  total propulsive 
j of fresh water  Energy j
of seawater  kinetic , , 
Derivative potential 
convective  Entrance 
directional  half angle of 
local  Euler
normal see Normal derivative equations , 
partial  formula , 
substantial ,  number , 
Differential equation Euler, Leonard , , 
Eulerian formulation , 
ordinary , , 
Expanded area 
partial , , , , , ,
Expanded area ratio , , 

required , , 
Differential equations
partial  F
Dimensional analysis –,  Field
Dipole  point see Collocation point
Direct numerical simulation (DNS) theory 
,  Flat plate friction coefficient
Dispersion relation , ,  ATTC, Schoenherr 
Displacement Grigson 
effect , , , , , ITTC  see ITTC, 
 model–ship correlation line
thickness , , ,  Prandtl–Schlichting 
volumetric ,  White 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

654 Index

Flow Form factor , , , , 

cylinder ,  appendage 
displacement , , , Granville 
 Grigson 
exterior , ,  Holtrop and Mennen 
interior  Prohaskas’s method 
irrotational  Watanabe 
laminar ,  Free surface , , , 
lifting , , ,  dynamic boundary condition
parallel , –, , , , , 
, ,  elevation 
Rankine oval  generalized boundary condition
source  , 
steady , , , , , kinematic boundary condition
 , , , 
turbulent ,  Froude
unsteady , , , , , , depth number 
 number , , , , ,
viscous , , , , , 
,  similarity 
vortex  Froude’s
wave ,  hypothesis , , 
Fluid  law of similarity 
j method ,  j
ideal , , 
incompressible , ,  Froude, William , , , , 
isotropic 
G
Newtonian , , 
Gas 
viscous , , , , 
Gauss’
Flux
integral theorem 
mass , , , , , 
Glauert
momentum , , , ,
integral , , 
, , 
series , , , , 
Foil see Lifting foil
Glauert, Hermann 
Foil section , ,  Gradient 
camber see Camber Gravitational acceleration , 
chord length see Chord length function of latitude 
geometry  local 
thickness  standard value 
Force Group velocity , , 
body , , ,  Guldhammer, H.E. 
conservative 
external , , ,  H
friction  Harvald, Svend Aage 
gravity  Helmholtz’s theorems , , ,
inertia , ,  , 
pressure , , , , ,  Helmholtz, Hermann von 
surface , ,  Hollenbach, Uwe 
viscous  Holtrop, Jan 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

Index 655

Hooke’s law  Kutta-Joukowsky’s

Hooke, Robert  lift theorem , , , ,
Hot wire anemometer  , 
Hull–propeller interaction 
Hydraulically L
actuated  Lagrange, Joseph-Louis 
smooth ,  Lagrangian formulation 
Hydrometer  Landau’s symbol 
Hydrostatic equilibrium  Laplace equation , –, ,
Hydrostatics , , , , , ,
basic theorem of  , , , , , 
cylindrical coordinates 
I polar coordinates , 
IAPWS  spherical coordinates 
Interaction Laplace operator , , 
hull–propeller , ,  dimensionless 
Intermittency factor  Laplace, Pierre-Simon , 
International Association for the Leading edge , 
Properties of Water and suction 
Steam see IAPWS Length
International Towing Tank Conference characteristic 
see ITTC computation 
ITTC xvii, , , , , , , in waterline 
j , , , , ,  over wetted surface  j
 model–ship correlation line Length–displacement ratio 
, , , , , , Lift , , , , 
, , , , , , Lift coefficient , , , ,
, , , , , , 
,  Lift force 
 performance prediction Lift–drag ratio , 
method , , ,  Lifting foil , –, 
 performance prediction Lifting line 
procedure  Lifting line theory , 
Liquid 
J Load variation test 
Jet efficiency see Efficiency, ideal
Joukowsky, Nikolay Yegorovich  M
Margin
K engine 
Keller’s formula , ,  service , , 
Keller, J. auf ’m  Mass
Kelvin transport (in waves) 
angle  Mass flow rate see Flux, mass
wave pattern – Mass flux see Flux, mass
Kelvin, Lord  Mathematical models , 
Kinematic  Matrix 
Kinetic  multiplication 
Kutta condition ,  Maximum continuous rating 
Kutta, Martin Wilhelm  Mean line , , 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

656 Index

NACA 𝑎 = 0.8  Normal vector , , , , , ,
parabolic  , , , , , –,
Mennen, G.G.J  , , , , , ,
Michell’s integral – , , , , 
Michell, John Henry  Normal velocity see Velocity, normal
Model
basin ,  O
test  Open water condition see Condition,
testing – open water
Moment Open water diagram , , ,
pitch  , , , , , 
Momentum  Open water efficiency see Efficiency,
flux see Flux, momentum open water
thickness , , , , , Open water test –, , ,
 , , , 
Moody chart  P
Motion Paint flow test 
steady  Panel methods 
unsteady  Pathline 
Performance prediction , , ,
N , –
Nabla operator ,  Perturbation 
dimensionless  Phase velocity , , , 
j j
NACA  deep water 
NASA  Pitch , , 
Naval architect  angle , , 
Navier, Claude L.M.H.  constant 
Navier, Claude Louis Marie Henri  effective 
Navier-Stokes equations –, , variable , , 
,  Pitch–diameter ratio , , ,
conservative, differential  , 
dimensionless  optimum , 
incompressible flow  Pitot, Henri 
incompressible, steady flow  Pitot-static tube , 
Navier-Stokes equations, Reynolds Potential see also Velocity potential,
averaged see RANSE 
Newton’s of gravity 
first law  Potential flow , , –, ,
laws of motion  , , , 
second law , , , , , Potential theory , , , ,
 
Newton, Sir Isaac , ,  Power
Newton-Raphson method  brake , 
Newtonian fluid see Fluid, delivered , , , ,
Newtonian , 
Nomenclature xvii effective , , , 
Nonconservative  shaft 
Normal derivative , ,  thrust , , , 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

Index 657

Power law  leading edge 

Powering estimate  radius 
comparison  tip 
example , – trailing edge 
Hollenbach – Propeller boat see Propeller
Holtrop and Mennen – dynamometer
Prandtl, Ludwig , , ,  Propeller design
Pressure ,  constant , , 
atmospheric ,  task  –, –
coefficient , , , , task  
, , , , , , task  
,  task  , –
difference  Propeller design chart –
shock  𝐵𝑃1 -chart , 
vapor , , ,  𝐵𝑃2 -chart 
Preturbulence , ,  𝐵𝑈1 -chart 
Propeller  𝐵𝑈2 -chart , 
azimuthing  logarithmic chart 
clearance  Propeller dynamometer , ,
controllable pitch , ,  
developed area  Propeller hub 
diameter , , ,  radius , 
ducted  vortex 
j j
expanded area ratio see Propeller selection
Expanded area ratio optimum diameter –,
fixed pitch ,  –
high skew  optimum rate of revolution
left-handed  –, –
lightly loaded  Propeller series –
loading see Thrust loading controllable pitch 
number of blades ,  ducted 
pitch–diameter ratio see Gawn 
Pitch–diameter ratio KCA 
podded  Newton–Rader 
projected area ,  skew 
rake see Rake Wageningen B-Series –
right-handed  Propulsion test –
skew see Skew continental method , 
slip ,  load variation 
stock  Propulsor , , , , , 
Voith Schneider 
warp  R
wheel effect  Rake , 
Propeller blade angle 
back  skew induced 
cupping  Rankine
expanded  oval , , 
face  source , 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

658 Index

Rankine, William J.M.  residuary , , , , ,
RANSE , , – 
Rate of revolution , , , steering 
, ,  total , , , , 
Reech, Ferdinand ,  wave , , , 
Region Resistance estimate 
multiply connected  comparison 
simply connected  example –, –,
Relaminarization  –
Resistance – Guldhammer and Harvald
air ,  –
appendage , , , , , Hollenbach –
 Holtrop and Mennen –
bow thruster  Reynolds
bow thruster tunnel ,  averaging 
bulbous bow  stress tensor , , –
components ,  stresses 
correlation allowance  Reynolds number , , , ,
eddy  , , , , 
frictional , –, , at radius 𝑥 = 0.75 , 
–, ,  local , 
hollows ,  Reynolds, Osborne , , , 
humps ,  Rheology 
j j
induced  Roughness
residuary , , ,  equivalent sand , 
roughness allowance , , propeller 
 technical 
shape factors  Roughness allowance see Resistance,
spray  roughness allowance
steering  Run , , , 
total , , , , , , , length of , , 
, , , 
transom ,  S
viscous , , , , ,  Sagitta 
viscous pressure , , , Salinity 
 Savart, Félix 
wave , , , , , , , Scale
, –, , – acceleration 
wave breaking ,  factor 
wave pattern ,  force , 
Resistance coefficient  geometric 
air , ,  length , , 
appendage  model 
bow thruster tunnel  surface , 
correlation allowance , , time , 
,  velocity 
environmental  volume , 
frictional , , , ,  Schlichting, Hermann 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

Index 659

Schneekluth nozzle  wave theory 

Self propulsion point , , , Stokes, Sir George Gabriel , ,
, , , , , , , 
,  Streakline 
Self-propelled ,  Streamline , , , 
Separation  dividing , 
laminar flow  Stress
point  apparent 
turbulent flow  normal 
Separation of variables ,  shear , 
Series   tensor 
Shock free entry  wall shear , 
Shortened thrust loading coefficient Strouhal number 
, ,  Strouhal, Vinzenz 
Similarity Suction side 
dynamic  Surface
Froude  free see Free surface
full dynamic  fully rough 
geometric  hydraulically smooth , 
kinematic  roughness , , 
laws of 
partial  T
Singularity ,  Tail–nose line , , , ,
j j
Sink  , , 
Sinkage Taylor
aft  series , , , , , 
dynamic ,  series, several variables 
fore  series, single variable 
mean , ,  Taylor Standard Series 
Skew  Taylor, Brook 
angle  Taylor, David Watson , , ,
Skew-back  
Skin friction correction  TEU , , 
force ,  Thickness distribution 
Source  elliptical , 
distribution  ogival 
point  Thickness ratio 
strength ,  Thin foil theory –
Speed Thomson, William see Kelvin, Lord
corresponding ,  Thrust , , , , , ,
design  , , , , , ,
Speed of advance , , , , , , , , , 
 available 
Stagnation point , , , , bearing 
, , , , ,  blade section 
Stall  coefficient , , , ,
Stokes , , , , 
hypothesis  coefficient correction 

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

660 Index

effect of cavitation  dot product 

identity , , , , , magnitude , 
,  Vector field
required , , , , irrotational 
 Velocity
Thrust deduction ,  attainable 
Thrust deduction fraction , , mean , 
–, , –, , normal , , , –,
 , 
Guldhammer and Harvald  turbulent 
Hollenbach  wall friction , 
Holtrop and Mennen  Velocity defect law 
Thrust loading , , , , Velocity potential , , , 
 D dipole 
coefficient , , , , D source 
,  D vortex , 
Torque , , , ,  D dipole 
available  D source 
blade section  cylinder flow 
coefficient , , , , of deep water wave , 
, ,  disturbance , , , ,
coefficient correction  , , , , , 
effect of cavitation  moving cylinder 
j identity ,  j
parallel flow , , 
Towing carriage  Rankine oval 
Towing point  of regular wave , 
Towing tank , , ,  Vessel
beach  displacement type , , ,
Trailing edge , ,  , , , 
Trial corrections  planing type , , 
Trim  Viscosity 
angle , ,  apparent see Viscosity, eddy
running  dynamic , 
Trim tank  eddy , 
Turbulence ,  eddy, kinematic 
generation , ,  kinematic , , , , 
isotropic , 
second 
longitudinal 
Vortex 
mean kinetic energy , 
bound 
model , 
distribution 
transverse 
filament 
U flow 
Updraft  horseshoe 
start-up , 
V strength , , , , ,
Vector  , , 
component  tip 
cross product  Vortex sheet

j
 Trim Size: mm × mm Single Column Tight Birk — “fshy” — // — : — page  — #
j

Index 661

roll up  period 

trailing  phase , 
Vorticity  profile 
bound  radiation 
free ,  regular , 
superposition , 
W transverse , 
Wake , , ,  trough 
effect of rudder  Wave energy 
effective  density 
frictional , ,  kinetic , , 
nominal 
potential , , 
nonuniform 
transport , 
potential , 
Wave maker , 
wave , 
Wave number , , , 
Wake fraction , –, ,
deep water , 
, , , 
Wave theory
frictional 
Airy see Wave theory, linear,
full scale 

Guldhammer and Harvald 
boundary value problem 
Hollenbach –
linear , –
Holtrop and Mennen 
linearized boundary value
potential 
problem 
j Wake function  j
Wake hook  ocean bottom condition 
Water radiation condition 
fresh ,  Stokes 
sea ,  Wetted surface , 
Water jet  appendages 
Wave  Hollenbach 
amplitude ,  Holtrop and Mennen 
crest  Kristensen and Lützen 
diffraction  Mumford’s formula 
dispersion ,  Wheel effect 
divergent ,  Wigley hull , 
dynamic pressure  Wind tunnel 
elevation ,  Wing
frequency ,  finite span 
height ,  tip 
length , , , ,  tip vortex 
long-crested ,  Wingspan 
particle acceleration 
particle path  Z
particle velocity  Zero lift angle see Angle of attack,
pattern ,  zero lift