Computational Free Wake Analysis of A Helicopter Rotor

The Pennsylvania State University
The Graduate School
Department of Aerospace Engineering
COMPUTATIONAL FREE WAKE ANALYSIS OF A HELICOPTER ROTOR
A Thesis in
Aerospace Engineering
by
Christopher J. Szymendera
Submitted in Partial Fulfillment

of the Requirements
for the Degree of
Master of Science
May 2002
We approve the thesis of Christopher J. Szymendera.
Date of Signature
Lyle N. Long
Professor of Aerospace Engineering
Thesis Advisor
Mark D. Maughmer
Dennis K. McLaughlin
Head of the Department of Aerospace Engineering
ABSTRACT
One of the most important issues in understanding the behavior of rotorcraft is the
accurate prediction of the rotor wake. Understanding the complex nature of the wake is
necessary for the prediction of such factors as blade loading, acoustics, and vibration. A
free wake vortex lattice method was used to predict the wake structure and blade loading
for a rotor in arbitrary motion. The blades were modeled as flat plates, and ring vortices
were distributed on the surface of the blades. As the blades rotated, vortices were shed
into the wake and then moved with a local velocity induced by the effects of the vortices
on the blades and in the wake. The wake was allowed to freely deform over time into its
natural structure. The lift was determined from the strength of each vortex on the blade,
calculated by applying the boundary condition of no flow normal to the blade.
On a helicopter in forward flight, the advancing blades usually experience high
Mach numbers at the tip. Therefore, it is important to be able to accurately predict the
blade loading in compressible flow. The free wake method, though, calculates induced
velocities using the Biot-Savart Law, which is only applicable to incompressible flow. A
Prandtl-Glauert correction was applied then to the Biot-Savart Law, which allowed the
code to accurately model compressible flow. The code was validated using two
experimental hover cases, one in which the flow was entirely incompressible and one in
which the flow was compressible at the tips, and then extended to handle any arbitrary
motion of the helicopter.

TABLE OF CONTENTS
LIST OF FIGURES ...................................................................................................... vi
ACKNOWLEDGMENTS............................................................................................ ix
Chapter 1 INTRODUCTION ....................................................................................... 1
1.1 Overview......................................................................................................... 1
1.2 Rotor Wake Characteristics ............................................................................ 3
1.3 Previous Research in Rotor Wake Simulation................................................ 10
1.3.1 Vortex Methods .................................................................................... 11
1.3.2 Computational Fluid Dynamics ............................................................ 13
1.3.3 Hybrid Methods .................................................................................... 15
1.3.4 Compressibility..................................................................................... 16
1.4 Current Approach............................................................................................ 18
Chapter 2 FUNDAMENTALS OF POTENTIAL FLOW............................................ 19
2.1 Overview......................................................................................................... 19
2.2 Vorticity and Circulation................................................................................. 19
2.3 Two-Dimensional Vortex ............................................................................... 21
2.4 Three-Dimensional Vortex ............................................................................. 23
2.5 Vortex Core..................................................................................................... 28
2.6 Compressibility Correction ............................................................................. 33
2.7 Applying the Prandtl-Glauert Transformation to the Biot-Savart Law........... 41
Chapter 3 FREE WAKE METHOD ............................................................................ 44
3.1 Overview......................................................................................................... 44
3.2 Free Wake Procedure ...................................................................................... 44
3.2.1 Discretization of the Blade and Wake .................................................. 45
3.2.2 Determine the Lift on the Blades .......................................................... 48
3.2.3 Calculate Wake Roll-Up....................................................................... 51
3.3 Implementation of the Free Wake Procedure in the Code .............................. 51
Chapter 4 RESULTS AND DISCUSSION.................................................................. 60
4.1 Overview......................................................................................................... 60
v
4.2 Code Validation .............................................................................................. 61

4.2.1 Case 1: Incompressible Hover .............................................................. 63
4.2.2 Case 2: Compressible Hover................................................................. 68
4.2.2.1 Case 2a: Incompressible Code vs. Compressible Code.............. 68
4.2.2.2 Case 2b: Changing the Chordwise Panel Spacing...................... 73
4.2.2.3 Case 2c: Changing the Vortex Core Radius ............................... 76
4.2.2.4 Case 2d: Changing the Number of Spanwise Panels.................. 79
4.2.2.5 Case 2e: Changing the Number of Chordwise Panels ................ 82
4.3 Wake Visualization ......................................................................................... 85
4.4 Application to Other Cases ............................................................................. 91
4.4.1 Axial Climb .......................................................................................... 91
4.4.2 Forward Climb ...................................................................................... 94
Chapter 5 CONCLUSIONS ......................................................................................... 96
5.1 Summary and Conclusions.............................................................................. 96

5.2 Suggestions for Future Work .......................................................................... 98
BIBLIOGRAPHY ........................................................................................................ 101
Appendix A SAMPLE INPUT FILE............................................................................ 107
Appendix B COMPUTER PROGRAM SUBROUTINES .......................................... 108

LIST OF FIGURES
Figure 1.1: Schematic showing the wake and its interaction with the fuselage in
forward flight ........................................................................................................ 2
Figure 1.2: Traditional model of a rotor wake, showing a concentrated tip vortex
and a trailing vortex sheet ..................................................................................... 5
Figure 1.3: Schematic showing the wake and tip vortex roll-up.................................. 6
Figure 1.4: Tip vortices trailed behind an E/A-6B Prowler ......................................... 8
Figure 1.5: Tip vortices trailed behind the blades of an AH-1W Super Cobra ............ 8
Figure 1.6: Smoke visualization of the tip vortex locations in forward flight ............. 9
Figure 2.1: Relation between surface integral and line integral ................................... 20
Figure 2.2: Two-dimensional flow field around a solid rotating cylinder showing
(a) streamlines and (b) tangential velocity of flow................................................ 22
Figure 2.3: Velocity at point P due to a vortex distribution in volume V .................... 24
Figure 2.4: Velocity at point P induced by a vortex segment....................................... 24
Figure 2.5: Velocity at point P induced by a vortex segment....................................... 26
Figure 2.6: Laser light sheet flow visualization tip vortex and vortex sheet................ 28
Figure 2.7: Velocity field inside a tip vortex shown by (a) idealized view and (b)
model used in vortex method ................................................................................ 30
Figure 2.8: Comparison of different 2-D vortex models.............................................. 32
Figure 2.9: Effect of Prandtl-Glauert transformation: (a) actual domain in

compressible flow; (b) stretched domain in analogous incompressible flow ....... 40
Figure 2.10: Stretching effect on a rotating blade ........................................................ 41

vii
Figure 2.11: Stretching effect of the distance between the vortex and control point
based on the Prandtl-Glauert correction: (a) actual distance; (b) stretched
distance.................................................................................................................. 43
Figure 3.1: Vortex ring model...................................................................................... 46
Figure 3.2: Wake shedding procedure.......................................................................... 47
Figure 3.3: Computer program flowchart..................................................................... 52
Figure 4.1: Configuration of the experimental rotor .................................................... 61
Figure 4.2: Grid spacing on blade ................................................................................ 62
Figure 4.3: Thrust coefficient over time....................................................................... 64
Figure 4.4: Spanwise lift coefficient (normalized by tip speed) .................................. 66
Figure 4.5: Chordwise pressure coefficient at different radial sections ....................... 67
Figure 4.6: Thrust coefficient over time....................................................................... 70
Figure 4.8: Chordwise pressure coefficient at different radial locations...................... 72
Figure 4.10: Chordwise pressure coefficient at different radial sections ..................... 75
Figure 4.11: Spanwise lift coefficient (normalized by tip speed) ................................ 77
Figure 4.13: Spanwise lift coefficient (normalized by tip speed) ................................ 80
Figure 4.15: Sectional lift coefficient (normalized by tip speed)................................. 83
Figure 4.17: Wake trailing behind both blades ............................................................ 86
Figure 4.18: Wake trailing behind one blade ............................................................... 87
Figure 4.19: Path of tip vortices ................................................................................... 89

viii
Figure 4.20: Roll-up of tip vortex at 180°.................................................................... 90
Figure 4.21: Sectional lift coefficient (normalized by tip speed)................................. 92
Figure 4.23: Tip vortices from a 4-bladed rotor in forward climb ............................... 95
Figure B.1: READGRID subroutine ............................................................................ 108
Figure B.2: CALCAIC subroutine ............................................................................... 109
Figure B.3: CALCRHS subroutine .............................................................................. 110
Figure B.4: VORTEX subroutine................................................................................. 111
Figure B.5: PRESSURE subroutine ............................................................................. 112
Figure B.6: WAKEVELOC subroutine........................................................................ 113

ACKNOWLEDGMENTS
I would like to thank the National Science Foundation for funding the CEMBA
program, an IGERT research center. And I would like to thank the students and faculty
of the CEMBA program for providing me with the opportunity and the resources to work
on this project.
I would also like to acknowledge the students of the IHPCA for their suggestions
and technical assistance throughout my work. Many thanks also to Dr. Mark Maughmer
for his comments and suggestions regarding my work. Finally, I would like to express
my appreciation to Dr. Lyle Long for his guidance throughout this project. His support in
this work was invaluable.

Chapter 1
INTRODUCTION
1.1 Overview
Accurate prediction of the rotor wake is one of the biggest challenges facing the
rotorcraft industry today. Understanding the detailed structure of the wake is
fundamental to the accurate prediction of rotor loads, performance, vibrations, and
acoustics. Away from the rotor, understanding the wake is important in the prediction of
interference effects with the fuselage, empennage, and tail rotor. The position and
strength of the wake are influenced by many factors, including blade geometry, number
of blades, rotor thrust, angle of attack of the tip path plane, and operating state of the
helicopter. In most forward flight operating conditions, the combination of the forward
speed of the helicopter and the rotating speed of the blades results in areas of high-Mach
number flow at the blade tips on the advancing side of the rotor disk. Therefore, it is
important to consider compressibility effects when attempting to model the flow of a
helicopter in any operating state other than the simple hover case and slow-speed forward
flight.
The complexity of the wake of a helicopter in forward flight can be seen in Figure
1.1. It shows some of the aerodynamic issues that exist in the wake as well as the
interaction of the wake with the body of the helicopter. Because of these interaction
2
effects, the ability to predict the performance of the entire helicopter is highly dependent
on the ability to predict the detailed structure of the wake. The details of the wake are
hard to visualize by either experimental or computational models, however, due to the
complex nature of the wake. Therefore, it is necessary to develop more advanced
computational models that will allow engineers to accurately predict the position and
strength of the rotor wake and, in turn, accurately predict the performance and
aeroacoustics of the entire helicopter.
Figure 1.1: Schematic showing the wake and its interaction with the fuselage in forward
flight [22]
3
Because the rotor wake is affected by many different factors, it is important to
develop a method that is extremely versatile in order for it to be effective. The method
should be able to handle any number of rotating blades with any given pitch and coning
angle and should be able to model the wake of a helicopter following any specific flight
path. The simplest approach, which will be presented in detail here, is for a two-bladed
rotor in hover, but it should be noted that the method described herein is easily expanded
to a more general case. Because compressibility effects are considered, the method can
handle any typical helicopter operating state where the flow at the blade tips is subsonic.
As will be presented in detail later, the rotor wake can be modeled as a collection
of a large number of vortex elements that all affect each other. Understanding how these
elements affect each other is fundamental to understanding the wake in general.
Problems of this nature are often termed “many-body” problems. Many-body problems
are found throughout the scientific and engineering world, and include such fields as
molecular dynamics, gas dynamics, charged particles in an electromagnetic field, and
protein folding. While the actual physics behind these problems may be different, they
are for the most part solved computationally using similar algorithms and procedures.
1.2 Rotor Wake Characteristics
The rotor wake consists of a shed vortex sheet and a concentrated vortex at the
tip, as seen in Figure 1.2. There is a bound circulation on the rotor blade associated with
lift, and conservation of vorticity requires that the circulation be trailed into the wake at
the blade tip and root. The strong tip vortices are the dominant feature in the rotor wake.
4
Vorticity is also shed and trailed into the wake, creating the vortex sheet, as a result of
changes in the circulation on the blade. The trailed vorticity is oriented parallel to the
local free stream when it leaves the blade, similar to the tip vortex, while the shed
vorticity is oriented radially in the wake, perpendicular to the trailed vorticity [16].
Because of the rotation of the blade, lift and circulation are highest near the tip. Both
reach a maximum before decreasing rapidly to zero at the tip, which creates a trailing
vorticity of high strength at the edge of the wake and causes the wake to roll-up quickly
into a concentrated tip vortex. The same occurs at the root, but since the circulation drop-
off from the maximum to the root is more gradual, the trailed vortex is weaker and the
roll-up is not as pronounced [16]. Figure 1.3 shows the bound circulation on the blade,
the lift distribution on the blade, the vortex sheet, and the roll-up of the tip vortex.
5
Figure 1.2: Traditional model of a rotor wake, showing a concentrated tip vortex and a
trailing vortex sheet [21]
6
Figure 1.3: Schematic showing the wake and tip vortex roll-up [22]
7
Although the basic structure of the rotor wake is similar to that of a fixed-wing
wake, there is the added complication that each blade might pass through or near the
wakes of previous blades. This can be seen in Figures 1.4 and 1.5, which show the paths
of tip vortices trailed from a fixed-wing and rotary-wing aircraft respectively. The path
of the tip vortices, as well as the geometry and the strength of the rest of the wake,
depends mainly on the operating state of the helicopter. In hover, the tip vortices follow
a helical path as they are convected below the rotor disk. In forward flight, the wake is
skewed behind the rotor, as seen in Figure 1.6. This results in a more complicated
distortion and stronger interactions among the vortices in the wake [22].
8
Figure 1.4: Tip vortices trailed behind an E/A-6B Prowler [40]
Figure 1.5: Tip vortices trailed behind the blades of an AH-1W Super Cobra [40]
9
(a) Top View
Free-Stream
Flow
(b) Side View
Free-Stream
Flow
Figure 1.6: Smoke visualization of the tip vortex locations in forward flight [22]
10
Vorticity is also shed into the wake in the form of a vortex sheet as the blades
rotate. The vortex sheet is convected downward, normal to the rotor disk, by the wake
induced velocity and a component of the free stream velocity normal to the disk in
forward flight and climb. The induced velocity is due to the vortex-vortex interactions in
the wake.
Rotating blades encounter tip vortices shed from proceeding blades, which results
in a phenomenon known as Blade Vortex Interaction (BVI). The blade experiences a
sudden change in the local velocity induced by the approaching vortex. The vortex
produces a downwash velocity as it approaches the blade and then a sudden upwash
velocity as it passes the leading edge of the blade. These unsteady loads are an important
factor in the vibration, noise, and performance of a helicopter. BVIs occur mainly on the
advancing and retreating sides of the rotor disk and are most intense in forward flight
[16].
1.3 Previous Research in Rotor Wake Simulation
The complicated nature of the rotor wake makes studying it, either experimentally
or computationally, a complicated task. Flow visualization on scaled rotor models has
been the most common technique to study the wake experimentally. Many tests have
been completed using smoke flow visualization, shadowgraphy, hot-wire anemometry,
and laser Doppler velocimetry [22]. The rotor wake can even be seen naturally through
condensation in the tip vortices under certain conditions, as shown previously in Figures
1.4 and 1.5. While recent advances have allowed more accurate data to be obtained
11
through experimental techniques, advances in computational power have also allowed for
more accurate mathematical modeling of the rotor wake. Current computational models
usually contain simplifications that limit their usefulness to actual problems of
significance, so there is still much work to be done to improve these models.
1.3.1 Vortex Methods
In a vortex method, the wake is modeled by vortex lines that are discretized into a
lattice with either straight or curved segments. The circulation strength of each element
in the wake is set by the circulation on the blade when the element is shed into the wake.
The advantage of vortex models is that the strength and position of all wake elements are
known, and thus, the induced velocity field on the blade due to the wake can be
calculated. While less computationally intensive than finite-difference methods, vortex
methods can still run into a problem of high computational requirements when a large
number of elements is used to obtain more accurate results. There are two main classes
of vortex methods, distinguished by how they solve for the position of the wake. In
prescribed wake models, the location of the wake is assumed based on experimental or
computational data. In free wake methods, however, the strength and position of the
wake is solved for directly at each time step, which generally results in a more accurate
but more time consuming solution.
Landgrebe [20] modeled the rotor wake in 1969 by starting with the classical rigid
wake, based on undistorted helical sheets, and calculating the distorted geometry of the
wake. He observed significant distortions to the rigid wake, and he saw the wake
12
contract and the tip vortices roll-up, as expected. Later, Landgrebe [21] used
experimental data to create a prescribed wake and observed an improvement in the
accuracy of the results compared to previous work. He used experimental data from a
number of different rotor configurations to create a generalized wake model for a
hovering rotor that could be used as the basis of a prescribed wake model. Egolf and
Landgrebe [10] used a similar procedure to create a generalized wake model for a rotor in
forward flight. Since not much experimental data is available for forward flight cases,
however, the model was based on a combination of experimental data and a mathematical
calculation of the distortion of a classical rigid wake. Beddoes [1] created a simple
prescribed wake model for forward flight by using an undistorted wake and developing a
set of distortion equations to modify the vertical displacement of the wake elements.
This model showed good agreement with free wake models available at the time but was
considerably less computationally intensive.
As the available computational power increased, the prospects of creating a true
free wake model improved. Early work on free wake models was done by Clark and
Lieper [7] in 1970 for heavy-lift helicopters in hover. They started with a basic helical
wake shape represented by straight-line vortex filaments and allowed the wake to
propagate over time, declaring convergence when the wake shape stopped changing
between time steps. They observed that the tip vortex remained close to the tip path
plane until it interacted with the next blade and was convected downward. The computed
wake shape closely matched experimental data for the Sikorsky S-65. At about the same
time, Sadler [31] developed a free wake model by impulsively starting the rotor blades
13
from rest with no wake and allowing the wake to develop behind the blades over time.
Vortices were shed into the wake with strengths corresponding to the blade circulation
strength at each time step and were allowed to translate with a speed based on the
forward speed of the helicopter and the local induced velocity. He was able to obtain
good results with a fine grid, but the computational requirements were too high, so
certain simplifications had to be made. He used only streamwise segments of the vortices
in the far wake to reduce the computational time without significant loss of accuracy and
was able to produce a realistic wake geometry for advance ratios greater than 0.1. The
model was not able to capture the more severe wake distortion and Blade Vortex
Interactions at lower advance ratios. In 1988, however, Egolf and Massar [11] showed
that modern advanced computers could handle free wake models without many of the
simplifying assumptions previously used.
1.3.2 Computational Fluid Dynamics
Computational Fluid Dynamics (CFD) uses finite difference, finite volume, or
finite element methods to solve the Euler and Navier-Stokes equations in the entire rotor
flow field. CFD generally provides a more detailed view of the rotor wake than vortex
methods but still has some outstanding issues that limit its usefulness. Although
available computer power continues to increase, the complex nature of the rotor wake
requires enormous computer resources to provide accurate results. CFD methods also
exhibit numerical dissipation, an artificial viscosity that tends to cause unrealistic
diffusion of the wake. The high grid resolution required to avoid numerical dissipation
14
and to capture the full unsteady aerodynamics of the rotating blades leads to
computational requirements that are several orders of magnitude greater than those of
vortex methods.
Chen and McCroskey [6] first captured the rotor wake by solving the Euler
equations in 1988. The results showed good agreement with experimental data near the
tip but exhibited considerable numerical diffusion. Still, it was a major step in the
development of a method that would capture the rotor wake without using any ad hoc
modeling. Srinivasan et al. [34] developed the TURNS code in 1992, which used the
thin-layer Navier-Stokes equations to capture the rotor wake without any wake modeling.
It showed fair agreement with experimental data except in the inboard region and tip
region of the blade, and it predicted the flow separation seen in the tip region in
experiments and the detailed structure of the tip vortex.
There has been much work done to improve the accuracy of these early CFD
methods. Strawn and Barth [37] used unstructured grids with Euler equations and
adaptive-grid refinement to improve the resolution of interesting flow features.
Similarly, Duque and Srinivasan [9] used overset grids to improve the resolution of
certain flow features by applying separate optimized grids to different areas of the
domain. Despite the improvements of these and other methods, CFD in general is still
too computationally intensive to be used in many rotor wake problems. More
computational power must be available before CFD simulations with finer grids can be
run that could truly capture the physics of the wake and limit the effects of numerical
diffusion.
15
1.3.3 Hybrid Methods
One of the ways around the current limitations of CFD is to use a hybrid method
that combines a finite-difference method in the near wake with a vortex method in the far
wake. Tung et al. [39] combined two independent codes with only minor modifications
to create a hybrid method in 1986. They used a finite-difference code near the blade tip
to calculate the loading and passed that information to an integral code, which calculated
the downwash effects and passed them back to the finite-difference code. Steinhoff and
Ramachandran [36] developed an alternate method where, instead of treating the vortex
sheets separately from the region near the blade, they are embedded into a compressible
potential flow field. The method, which was implemented in a code, HELIX-I,
calculated the free convection of the wake and eliminated the problems of numerical
diffusion seen in true CFD codes. Later, Moultan et al [28] combined an overset version
of HELIX-I with the TURNS Navier-Stokes code. They used the TURNS code to
calculate the viscous effects on the blade and HELIX-I to calculate the free wake
convection. Sezer-Uzol and Long [33] developed an approach at Penn State to preserve
vortices over longer distances in a coupled Euler/vortex method, which also reduced the
effects of numerical diffusion seen in pure CFD codes. It produced a more accurate
model of the tip vortex in the far wake than what is seen in most CFD codes, where the
tip vortex is quickly dissipated.
A different approach was applied by Berkmen et al. [4] in 1997. They separated
the flow field into three different parts. First, the unsteady, compressible Navier-Stokes
equations were solved in an inner zone surrounding each blade to capture the wake and
16
viscous effects. Next, the isentropic potential flow equations were solved in an inviscid
outer zone, which was used to carry pressure waves to the far wake. Finally, a
Lagrangian wake zone inside the outer zone was used to model the wake. It captured the
vorticity leaving the viscous region and convected it to the far field by solving the Biot-
Savart Law.
1.3.4 Compressibility
One of the most important and challenging aspects of modeling the rotor wake is
accounting for the effects of compressibility, since it is common for the tips of advancing
blades to experience high-Mach number flows in forward flight. As was discussed in
previous sections, many CFD and hybrid-CFD methods already account for
compressibility, but there are other ways to consider its effects as well. A supersonic
vortex lattice method was implemented for arbitrary wings in the VORLAX code at
Lockheed-California in 1977, in which the induced velocity equations were rederived
starting with the governing equations for inviscid compressible flow [28].
A method introduced by Morino [29] in 1974 and later extended to arbitrary
motion by Gennaretti and Morino [14] used a new integral equation derived from the
velocity potential equation to account for subsonic, compressible flow. In application of
the method to a hovering rotor, they used the wake geometry obtained from an
incompressible free wake analysis as a prescribed wake for the compressible case. They
compared their results with a crude application of the Prandtl-Glauert correction, which
consisted only of dividing the spanwise pressures by a correction factor, and concluded
17
that their results showed good agreement with the experiment but that the Prandtl-Glauert
correction over-predicted lift.
Long and Watts [26] used an “acoustic analogy approach” to develop an integral
equation based on the Ffowcs Williams-Hawkings equation to solve for compressible
arbitrary motion. They accounted for wake effects by storing the time history of the
blades and calculating the effects of the pressure jump that lingers on at each previous
position of the blade. Because of the finite speed of propagation of the disturbances
caused by the pressure jumps, a time lag exists before the effects of each jump are felt on
the blade. Epstein and Bliss [12] investigated the effect of compressibility during the
initial stages of wake development using a similar technique. New wake elements
produce a pressure wave at the instant they are emitted, and these waves propagate both
upstream and downstream at the speed of sound. They calculated the time required for
newly emitted waves to reach the leading edge of a wing in forward flight and the
distance the wing traveled in that time. They used compressible methods for the near
wake region within that characteristic length but incompressible methods for the far
wake. The combined compressible-incompressible method showed considerable
improvement over a fully incompressible method and produced results very similar to
those of a fully compressible method. Therefore, it was concluded that the effects of
compressibility are related to the wake generation process and are confined to the very
near wake.
18
1.4 Current Approach
The goal of this work is to study vortex-vortex interactions by developing a
method that will model the wake of a multi-bladed rotor following an arbitrary flight
path. While the code will be able to handle such motion, it will only be validated for a
two-bladed hovering rotor, for which extensive experimental data is available. Since
compressibility is a factor in forward flight but usually not a factor in hover, an
unrealistic hover case will be run where the flow is compressible at the tips to show that
the method can account for it. The best way to observe vortex-vortex interactions is
through a vortex method, since vortex methods are based on the effects of these
interactions. In order to obtain the best results, a free wake method was chosen, since the
free wake method provides a good balance between accuracy and computational
efficiency. In general, vortex methods cannot account for the effects of compressibility,
so a Prandtl-Glauert correction will be added to account for these effects. Unlike the
Prandtl-Glauert correction used by Gennaretti and Morino [14] and discussed previously,
this correction will be applied directly to the calculation of vortex-induced effects rather
than to the final pressure calculations on the blade. A detailed description of the free
wake method and the compressibility correction will be discussed, as well as the
development of the code to implement these methods and the results for two hover cases.
The ability of the code to be used to study more complicated helicopter operating states
will also be highlighted through a series of example cases.

Chapter 2
FUNDAMENTALS OF POTENTIAL FLOW
2.1 Overview
Vortex methods are based on the interaction of vortices in the wake and the
calculation of vortex-induced velocity fields on the rotor blades and in the wake. It is
necessary to understand the fundamentals of the flow field in and around a vortex,
however, before considering the vortex-vortex interactions. The Biot-Savart Law, which
is the basis for calculating vortex-induced velocities, will be developed in this chapter.
Since the Biot-Savart Law is valid only for incompressible flow, the Prandtl-Glauert
compressibility correction will also be developed and applied to the Biot-Savart Law. It
should be noted that the following work is only valid for regions of inviscid flow.
Although the results cannot be used to account for the effects of the thin, viscous
boundary layer on the surface of the blades, it will be shown later that they do provide a
suitable approximation to the actual flow field.
2.2 Vorticity and Circulation
It is beneficial to start by defining vorticity and circulation, since they will be
used extensively in the vortex method. In general, the motion of a fluid particle consists
20
of translation, rotation, and deformation. The focus of this chapter will be on rotation.
The angular velocity of a fluid particle in vector notation is
1
ω = ∇×q ( 2.1 )
2
where q represents the velocity field of the particle. Vorticity is defined as twice the
angular velocity,
ζ ≡ 2ω = ∇ × q ( 2.2 )
and circulation is defined as
Γ ≡ ∫ q ⋅ dl ( 2.3 )
C
From Figure 2.1, which shows a surface S enclosed by the curve C, the circulation can be
related to the vorticity on the surface using Stokes’ Theorem:
Γ = ∫ q ⋅ dl = ∫ ∇ × q ⋅ ndS = ∫ ζ ⋅ ndS ( 2.4 )

C S S
ζ=2ω
ω
n C
dS
Figure 2.1: Relation between surface integral and line integral [18]
21
2.3 Two-Dimensional Vortex
To illustrate a two-dimensional vortex, it is useful to consider the solid cylinder
shown in Figure 2.2a, with radius R, rotating in a viscous fluid at an angular velocity of
ω. This rotation causes circular streamlines to develop around the cylinder, and the
tangential velocity of these streamlines is given by [18]
Γ
qθ = − ( 2.5 )
2πr
where the circulation is
Γ = 2ωπ R 2 ( 2.6 )
The velocity field will be derived in detail for a three-dimensional vortex in the next
section, and it will be shown that the result simplifies to Equation 2.5 for the two-
dimensional case. The tangential velocity of a point in the solid body, where r is less
than R, is simply given as
qθ = rω ( 2.7 )
The total velocity distribution given by Equations 2.5 and 2.7 is shown in Figure 2.2b.
The significance of the relationship between the velocity inside the solid body and
outside the body will be discussed later when the concept of the vortex core is introduced.
22
Streamlines
r
ω
(a)
qθ
Γ
2πr
rω
R r
(b)
Figure 2.2: Two-dimensional flow field around a solid rotating cylinder showing (a)
streamlines and (b) tangential velocity of flow [18]
23
2.4 Three-Dimensional Vortex
The three-dimensional induced-velocity field can be calculated using the Biot-
Savart Law, which is derived next, based on Refs. 3 and 18. Starting with the assumption
that the flow is incompressible, the continuity equation is
∇ ⋅q = 0 ( 2.8 )
The velocity field q can be expressed as the curl of a vector field B,
q = ∇×B ( 2.9 )
where B is selected such that
∇⋅B = 0 ( 2.10 )
The vorticity can then be expressed in terms of B:
ζ = ∇ × q = ∇ × (∇ × B ) = ∇(∇ ⋅ B ) − ∇ 2 B
Using the condition set in Equation 2.10, the vorticity equation reduces to Poisson’s
Equation for the vector field B
ζ = −∇ 2 B ( 2.11 )
Equation 2.11 can solved using Green’s Theorem to evaluate the vector field at a point P
due to the vorticity in a volume V as shown in Figure 2.3.
1 ⌠ ζ
B=  dV ( 2.12 )
4π ⌡V r0 − r1
Substituting this result into Equation 2.9, the velocity field at P is
1 ⌠ ζ
q=  ∇× dV ( 2.13 )
4π ⌡V r0 − r1
24
•P
r0
r0 – r1
ζ
V
dV
r1
origin
Figure 2.3: Velocity at point P due to a vortex distribution in volume V [18]
dl ζ
r0 – r1 • P
dS
Γ r0
r1
origin
Figure 2.4: Velocity at point P induced by a vortex segment [18]

25
It is necessary to make a substitution for the integrand before evaluating Equation
2.13. Considering the vorticity filament shown in Figure 2.4, where dS and dl are normal
to the filament, the circulation is
Γ = ζdS
and the volume is related to the surface by
dV = dSdl
Therefore, the integrand in Equation 2.13 becomes
ζ dl
∇× dV = ∇ × Γ
r0 − r1 r0 − r1
or, if the curl operation is evaluated while keeping r1 and dl fixed,
dl dl × (r0 − r1 )
∇×Γ =Γ ( 2.14 )
r0 − r1 r0 − r1
3
Substituting this into Equation 2.13 produces the Biot-Savart Law:
Γ ⌠ dl × (r0 − r1 )
q=
4π 
( 2.15 )
⌡ r0 − r1
3
26
Γ
θ2
r0
dl
θ
θ1 rp r2
r
r1
•P
Figure 2.5: Velocity at point P induced by a vortex segment [3]
The general form of the Biot-Savart Law, Equation 2.15, must be integrated
before it can be used in the numerical method. It can be written in differential form in
terms of the new notation shown in Figure 2.5:
Γ dl × r
∆q = ( 2.16 )
4π r 3
The cross product is evaluated,
dl × r = r sin θdl
and the following substitutions are made, based on the geometry shown:
rp
rp = r sin θ , r=
sin θ
rp − rp rp
tan (π − θ ) = , r0 = , dl = dθ
r0 tan θ sin 2 θ
27
Equation 2.16 can then be rewritten in scalar form as
Γ r sin θdl Γ  sin 2 θ  rp  Γ

∆q = = (sin θ ) 
 r  sin θ
dθ 
 4πr sin θdθ
= ( 2.17 )
4π r 3
4π  p
2

2
 p
Equation 2.17 is integrated over the length of the segment:
Γ θ2 Γ
q= ∫θ sin θdθ = (cosθ1 − cosθ 2 ) ( 2.18 )
4πrp 1 4πrp
Converting back to vector notation and making the following geometric substitutions,
r1 × r2 r0 ⋅ r1 r0 ⋅ r2
rp = cos θ 1 = cos θ 2 =
r0 r0 r1 r0 r2
the final expression for the induced velocity at a point due to a straight-line vortex is
Γ r1 × r2  r0 ⋅ r1 r0 ⋅ r2 
q=  −  ( 2.19 )
4π r1 × r2 2  r1 r2 
Equation 2.19 is the general form of the Biot-Savart Law which will be used extensively
in the vortex method to determine the vortex-induced velocities.
To see how this relates to the two-dimensional velocity field presented in the last
section, consider the case of an infinite vortex, where θ1 = 0 and θ2 = π. Equation 2.18
reduces to
Γ
qθ = (cos(0) − cos(π ) ) = Γ ( 2.20 )
4πr 2πr
which matches the result for the solid cylinder. As in the two-dimensional case, though,
the Biot-Savart Law is only valid outside of the vortex core.

28
2.5 Vortex Core
The flow field inside a vortex can be seen in Figure 2.6, which shows a vortex
sheet and the concentrated roll-up of the tip vortex. The flow follows circular streamlines
around the vortex core and is similar to the two-dimensional flow field around the solid
cylinder.
Figure 2.6: Laser light sheet flow visualization tip vortex and vortex sheet [23]
Inside the vortex core, where the flow is viscous, the tangential velocities should
tend to zero, as shown in Figure 2.7a and by the dotted line in Figure 2.7b. The induced
velocities calculated with the potential flow model, however, increase towards infinity
inside the core, as shown by the dashed line in Figure 2.7b. Therefore, it is necessary to
model the flow inside the core using a different method. Since the flow field in the
vortex core is similar to the flow field around a solid cylinder, the simplest approach is to
29
use an approximation based on the solid body rotation developed earlier. Inside the core,
the velocity field is represented by
 Γ  r 
qθ =    ( 2.21 )
 2πrc  rc 
which is a straight line from r = 0 to r = rc. Outside the core, the velocities are calculated
using the Biot-Savart Law, Equation 2.19. The solid line in Figure 2.7b represents the
combination of the solid body rotation approximation and the Biot-Savart Law, which is
known as the Rankine vortex model. This is one of the simplest vortex models and the
one that will be used in this method. The size of the vortex core is known to grow
somewhat as a function of wake age [22], but for simplicity, the core radius was set as a
constant in this model.

30
Outer Potential
Flow Region
rc
Tangential
Velocity Profile Vortex Core
(a)
Potential Flow
Viscous Flow Approximation

(Solid Body Rotation)
(b) rc
Figure 2.7: Velocity field inside a tip vortex shown by (a) idealized view [22] and (b)
model used in vortex method [35]
31
Alternative vortex models have also been developed [22]. The first, created by
Oseen and Lamb, is based on a simplified form of the Navier-Stokes equation:
qθ (r ) =
Γ
2πrc (r rc )
(
1 − e −α (r rc )
2
) ( 2.22 )
where α = 1.25643. Other similar models were also developed by Scully,
 Γ  r rc
qθ (r ) =  
 2πrc (
 1 + (r rc )
2
) ( 2.23 )
and Vatistas et al,
 Γ  r rc
qθ (r ) =   ( 2.24 )
 2πrc  1 + (r rc )4
A comparison of these models is shown in Figure 2.8. Notice that the velocities
calculated by each method are similar away from the core but that there are significant
differences inside and just outside of the core.

32
Tangential Velocity, qθ / (Γ/2πrc)
Non-Dimensional Distance From Center, r / rc
Figure 2.8: Comparison of different 2-D vortex models

33
2.6 Compressibility Correction
The derivation of the compressibility correction is based on Refs. 3, 8, and 13. It
is useful to start with the continuity equation and the Navier-Stokes equation. This time,
however, the assumption of incompressible flow is not made, and the continuity equation
is presented using tensor notation:
∂ρ ∂
+ ( ρu k ) = 0 ( 2.25 )
∂t ∂x k
Ignoring external body forces, the Navier-Stokes equation is
∂u j ∂u j ∂p ∂  ∂uk  ∂   ∂u ∂u j 
ρ + ρu k =− + λ  +  µ  i + 
∂x j ∂x j  ∂xk
( 2.26 )
∂t ∂xk  ∂xi   ∂x j ∂xi 
Assuming that the effects of viscosity are negligible, Equation 2.26 reduces to the Euler
equation:
∂u j ∂u j ∂p
ρ + ρu k =− ( 2.27 )
∂t ∂x k ∂x j
The continuity equation and Euler equation can be linearized by considering small
disturbances in a uniform flow in a coordinate system that is at rest with respect to the
flow (i.e. u0 = 0). The change in the velocity component is ui, where ui is so small that ui2
is negligible compared to ui; p0 and ρ0 are the pressure and density of the undisturbed
flow; and p′ and ρ′ are the small perturbations in the pressure and density, such that
p = p0 + p′ ρ = ρ0 + ρ ′
34
The linear continuity equation then becomes
∂
(ρ 0 + ρ ′) + ∂ [(ρ 0 + ρ ′)(u k )] = 0
∂t ∂x k
or, eliminating higher order terms and assuming ρ0 is a constant,
∂ρ ′ ∂u
+ ρ0 k = 0 ( 2.28 )
∂t ∂x k
Likewise, the linear Euler equation becomes
(ρ 0 + ρ ′) ∂u k + (ρ 0 + ρ ′)(u k ) ∂u k =−
∂
( p0 + p ′)
∂t ∂x k ∂x k
or, eliminating higher order terms,
∂u k ∂p ′
ρ0 =− ( 2.29 )
∂t ∂x k
Taking the time derivative of Equation 2.28 and the spatial derivative of Equation 2.29
yields
∂ 2uk 1 ∂2ρ′
=−
∂x k ∂t ρ 0 ∂t 2
and
∂ 2uk 1 ∂ 2 p′
=−
∂x k ∂t ρ 0 ∂x k ∂x k
Combining these leads to
∂ 2 p′ ∂2ρ′
− 2 =0 ( 2.30 )
∂x k ∂x k ∂t
35
If the following substitution is made,
dp p ′
a2 = =
dρ ρ ′
then Equation 2.30 can be rewritten as
∂ 2 p′ 1 ∂ 2 p′
− 2 =0 ( 2.31 )
∂x k ∂x k a ∂t 2
which is the fundamental wave equation in acoustics for a fluid at rest at infinity. If small
perturbations are again considered, then
dp  dp  d2p
a =
2
=   + ρ  2 
′ +L
dρ  dρ  ρ ′=0  dρ  ρ ′=0
= a0 + O( ρ ′)
2
The term a0 is the speed of propagation of a sound wave in the undisturbed fluid and is
usually written as
 ∂p 
a0 =   ( 2.32 )
 ∂ρ  S
to indicate that the derivative is taken at constant entropy. In a perfect gas, the speed of
sound can be written as
γp 0
a0 = ( 2.33 )
p0
or
a 0 = γRT ( 2.34 )
where γ is the ratio of heat capacities, R is the universal gas constant, and T is
temperature.
36
Using this result and neglecting higher order terms, Equation 2.31 becomes
∂ 2 p′ 1 ∂ 2 p′
− =0 ( 2.35 )
∂x k ∂x k a 0 2 ∂t 2
This is an important result because vortices produce disturbances that propagate from the
trailing edge of the blade with the speed of sound when they are shed into the wake. In
incompressible flow, the speed of sound is assumed to be infinite, and the disturbances
are instantaneously transmitted to infinity in all directions. In compressible flow,
however, the speed of sound is finite, and there is a time delay before the disturbances are
felt throughout the domain [13].
It is necessary at this point to define the velocity potential and derive the potential
equation for incompressible flow. If the flow is irrotational, then
∇×u = 0 ( 2.36 )
The velocity vector can be expressed as the gradient of a scalar function φ, since
∇ × ∇φ ≡ 0
is an identity for any scalar function. Therefore,
u = ∇φ
( 2.37 )
∂φ
uk =
∂x k
The function φ is referred to as the velocity potential of the flow field. For
incompressible flow, the continuity equation reduces to
∂u k
=0
∂x k
( 2.38 )
∂ 2φ ∂ 2φ ∂ 2φ
+ + =0
∂x 2 ∂y 2 ∂z 2
37
Equation 2.38 is the potential equation for incompressible flow.
The velocity potential equation for compressible flow can be derived from the
wave equation. First, it is necessary to show that the velocity potential also satisfies the
wave equation. Recall that
∂φ
uk = ( 2.37 )
∂x k
and
∂u k ∂p ′
ρ0 =− ( 2.29 )
∂t ∂x k
Substituting Equation 2.37 into 2.29 and rearranging, 2.39
∂  ∂φ  ∂p ′ ∂  ∂φ 
ρ0   + =  ρ0 + p′ = 0
∂t  ∂x k
( 2.39 )
 ∂x k ∂x k  ∂t 
Therefore,
∂φ
ρ0 + p′ = 0
∂t
and
∂φ
p′ = − ρ 0 ( 2.40 )
∂t
Since the disturbed pressure is proportional to the velocity potential, the velocity
potential also satisfies the wave equation, and the wave equation can be rewritten as
∂ 2φ 1 ∂ 2φ
− 2 2 =0 ( 2.41 )
∂x k ∂x k a 0 ∂t
38
The wave equation was derived using a coordinate system that was at rest with
respect to the fluid. If the coordinate system is moving with a speed U in the negative x-
direction, it is necessary to apply a coordinate transformation:
x = x ′ − Ut ′, y = y ′, z = z ′, t = t′
x ′ = x + Ut ′
Looking at the time derivative term in Equation 2.41,
∂φ ∂φ  ∂x ′  ∂φ  ∂t ′ 
=  +  
∂t ′ ∂x ′  ∂t ′  ∂t ′  ∂t ′ 
∂φ ∂φ
=U +
∂x ′ ∂t ′
and
∂ 2φ ∂  ∂φ  ∂  ∂φ 
= U +  
∂t ′ 2
∂t ′  ∂x ′  ∂t ′  ∂t ′ 
∂  ∂φ  ∂  ∂φ 
=U  +  
∂x ′  ∂t ′  ∂t ′  ∂t ′ 
∂  ∂φ ∂φ  ∂  ∂φ ∂φ 
=U U + + U + 
∂x ′  ∂x ′ ∂t ′  ∂t ′  ∂x ′ ∂t ′ 
 2 ∂ 2φ ∂ 2φ   ∂ 2φ ∂ 2φ 

= U +U  + U ∂x ′∂t ′ + ∂t ′ 2 
 
 ∂x ′ 2
∂x ′∂t ′   
∂φ2
∂φ2
∂φ2
= 2 + 2U +U 2
∂t ′ ∂x ′∂t ′ ∂x ′ 2
Therefore, the wave equation can be rearranged into the form
 ∂ 2φ ′ ∂ 2φ ′ ∂ 2φ ′  1  ∂ 2φ ′ ∂ 2φ ′ 2 ∂ φ′
2
 2 + +  −  + 2U + U =0
 ∂x ′ ∂y ′ 2 ∂z ′ 2  a 0 2  ∂t ′ 2 ∂x ′∂t ′ ∂x ′ 2 
1 ∂ 2φ ′ 2U ∂ 2φ ′ ∂ 2φ ′ ∂ 2φ ′ ∂ 2φ ′
− 2 − −
1
U( 2
− 1) + + =0
a 0 ∂t ′ a 0 ∂x ′∂t ′ a 0 ∂x ′ 2 ∂y ′ 2 ∂z ′ 2
2 2 2
or, in terms of the Mach number,

39
1 ∂ 2φ 2 M ∂ 2φ 2 ∂ φ ∂ 2φ ∂ 2φ
( )
2
− 2 − + 1− M + + =0 ( 2.42 )
a 0 ∂t ′
2
a 0 ∂x ′∂t ′ ∂x ′ 2 ∂y ′ 2 ∂z ′ 2
For steady flow, the time derivative terms can be eliminated, which leaves
(1 − M ) ∂∂xφ + ∂∂yφ + ∂∂zφ = 0

2 2 2
2
2 2 2 ( 2.43 )
where the primes have been omitted for clarity. Equation 2.43 is the well-known Prandtl-
Glauert equation, which is a linearized small-perturbation form of the full velocity
potential equation for a flow with a mean velocity in the x-direction. It is in a similar
form as the velocity potential equation for incompressible flow:
∂ 2φ ∂ 2φ ∂ 2φ
+ + =0 ( 2.38 )
∂x 2 ∂y 2 ∂z 2
Applying another coordinate transformation to Equation 2.43, where
x
x′ = ; y ′ = y; z′ = z
1− M 2
leads to
∂ 2φ ∂ 2φ ∂ 2φ
+ + =0 ( 2.44 )
∂x ′ 2 ∂y ′ 2 ∂z ′ 2
Therefore, applying this coordinate transformation creates an analogy between a
compressible flow domain and an incompressible flow domain. This allows for a
solution to the subsonic compressible flow field with small disturbances to be found
using methods that were developed for incompressible flow. The effect of this
transformation is seen in Figure 2.9, which shows that the domain can be stretched in the
direction of the flow to create the analogous domain that can be solved using
40
incompressible methods. This stretching effect is known as the Prandtl-Glauert
compressibility correction and will be used with the Biot-Savart Law to account for
compressibility effects on the blade. The stretching effect on a rotating blade in hover is
shown in Figure 2.10, where the dashed line represents the original geometry and the
solid line represents the stretched geometry of the blade. Since the tangential velocity
varies linearly along the span of the blade, the stretching effect is more pronounced at the
tip where the flow speed is highest.
x
x′ =
x 1− M 2
• • • •
u u
• • • •
(a) (b)
Figure 2.9: Effect of Prandtl-Glauert transformation: (a) actual domain in compressible

flow; (b) stretched domain in analogous incompressible flow
41
Figure 2.10: Stretching effect on a rotating blade
2.7 Applying the Prandtl-Glauert Transformation to the Biot-Savart Law
Because the Biot-Savart Law was derived for incompressible flow, it is necessary
to apply the Prandtl-Glauert transformation before it can be used in problems where the
flow is compressible. The Biot-Savart Law was given as Equation 2.19:
Γ r1 × r2  r0 ⋅ r1 r0 ⋅ r2 
q=  −  ( 2.19 )
4π r1 × r2 2  r1 r2 
If a frame of reference is used where the control point is fixed and the vortex is moving,
it is possible to find the component of the velocity and Mach number of the vortex with
respect to the point P along the perpendicular vector between the vortex and the point.
Two vectors are identified that represent the vector defined by the vortex segment, r0, and
42
the vector from one endpoint of the segment to the control point, r1. The projection of r1
on r0 is
r ⋅r 
projr0 r1 =  0 21 r0 ( 2.45 )
 r 
 0 
and the projection of r1 perpendicular to r0 is
proj⊥r0 r1 = r1 − projr0 r1 ( 2.46 )
Equations 2.45 and 2.46 [2] are used to calculate the components of the perpendicular
vector rp between the vortex segment and the control point. Likewise, Equation 2.47 is
used to find the components of the vector up, which is the projection of the velocity
vector on the perpendicular vector:
 u ⋅r 
 
projrp u = 
p
 rp
2 rp ( 2.47 )
 
The components of the Mach number of the vortex can then be calculated, and the length
of the vector rp is recalculated using the Prandtl-Glauert correction:
rp
rp′ = ( 2.48 )
1− M p
2
This is seen in Figure 2.11, where the vortex is stretched along rp to a new position.
New values for r1′, r2′, r1′, and r2′ can be calculated, and the Biot-Savart Law becomes
Γ r1′ × r2′  r0′ ⋅ r1′ r0′ ⋅ r2′ 

q=  −  ( 2.49 )
4π r1′ × r2′ 2  r1′ r2′ 
43
Equation 2.49 can now be used to solve for induced velocities in a compressible flow.
The translation of the vortex to a new position is only for the purposes of calculating the
induced velocity; the vortex is not physically moved during this process.
up
Γ
r0
rp′ r2′
up
Γ
r0 r1′
rp r2 rp
r2
r1 r1
•P •P
(a) (b)
Figure 2.11: Stretching effect of the distance between the vortex and control point based
on the Prandtl-Glauert correction: (a) actual distance; (b) stretched distance
Chapter 3
FREE WAKE METHOD
3.1 Overview
The equations developed in the last chapter to calculate the velocity induced at a
point by a vortex can be used in a free wake method to determine the lift on the blades
and the distortion of the wake. First, the general methodology of the free wake method
for rotating blades will be discussed in this chapter. Then, the implementation of the
method into a computer program will be described.
3.2 Free Wake Procedure
The free wake method is a method in which the wake is generated behind the
blade at each time step and allowed to freely deform. The blade is discretized into
panels, and a ring vortex is placed at each panel. As the blade advances, the trailing edge
vortices on the wing are shed into the wake, and the wake is allowed to freely deform.
Each ring vortex on the blade and in the wake induces a velocity on all the other vortices
in the domain, which is calculated using the Biot-Savart Law. The Prandtl-Glauert
compressibility correction is applied to the Biot-Savart Law when considering vortices on

45
the blade. The induced velocities on the blades are used to calculate the circulation and
lift on each panel, and the induced velocities in the wake are used to deform the wake.
3.2.1 Discretization of the Blade and Wake
The blade is modeled by a thin wing with no thickness and no camber and is
situated such that its feathering axis lies along the quarter-chord line. The blade is
divided into panels, and a ring vortex is placed in each panel. Each vortex is placed at the
panel’s quarter-chord line so that the two-dimensional Kutta condition is satisfied along
the chord [18]. Figure 3.1 shows the ring vortices on the blade, with the circulation of
each vortex being defined as positive in the clockwise direction. A collocation point,
where the properties of the panel are stored, is also defined at the center of each panel’s
three-quarter-chord line.
Before the blade starts to rotate, there are no free wake elements, but the aft
segments of the vortex rings along the trailing edge of the blade lie in the wake. The end
points of these vortex segments make up one set of corner points for the first row of wake
vortices. As the blade advances during the first time step, the new endpoints of these aft
segments make up the second set of corner points, and the first row of wake elements is
created and oriented at the same angle of attack as the blade. This procedure is repeated
during each subsequent time step, and the wake grows with time as shown in Figure 3.2.
The strength of the most recently created wake element is set equal to the strength of the
trailing edge vortex from the previous time step. In essence, the trailing edge vortex is
shed into the wake as the blade advances from one position to the next. Once a wake
46
vortex is created, its strength remains unchanged, according to the Helmholtz theorem
[18]. Since the wake elements cannot carry aerodynamic loads, they move only with the
local velocity induced by other vortices on the blades and in the wake.
¼ chord
n
Collocation
Point
+Γ
Figure 3.1: Vortex ring model

47
Distance Covered During:
2nd Time Step
Blade
Vortices 1st Time Step
Wake
Vortices
Figure 3.2: Wake shedding procedure

48
3.2.2 Determine the Lift on the Blades
The strengths of the vortices on the blade can be solved for at each time step by
applying the boundary condition that there is no flow normal to the blade. The velocity
of each ring vortex on the blade is comprised of the free stream velocity of the blade, the
velocity induced by other vortices on the blade, and the velocity induced by the vortices
in the wake:
QnK = Qnormal velocity + Qnormal velocity + Qnormal component = 0

induced by blade induced by wake of free stream ( 3.1 )
vortices vortices velocity
Since the strength and position of the wake elements are known, the induced velocities
due to the wake can be calculated at each time step using the Biot-Savart Law. These
velocities, as well as the free stream velocity of the blade, which is also known, can be
transferred to the right hand side of Equation 3.1. It remains then to calculate the normal
velocity induced by the blade vortices as a function of the strength of each vortex and
then to solve for those strengths.
Since Equation 3.1 is used to solve for the strengths of the vortices on the blade, it
is necessary to incorporate these strengths into the equation. This is done in the only
unknown quantity, the velocity induced by the vortices on the blade. At each blade
panel, the velocity induced by every panel is calculated using the Biot-Savart Law. The
Biot-Savart Law can be rewritten as
 1 r × r r ⋅ r r ⋅ r  
q= 1 2 0 1
− 0 2  (Γ )
 4π r × r 2  r1 r2   ( 3.2 )
 1 2 
= a KL Γ
49
where aKL is an influence coefficient accounting for the effects of the Lth panel on the Kth
panel. Equation 3.2 can be summarized in matrix form for all the panels on the blades:
 a11 a12 L a1L  Γ1 

  
 a 21 a 22 L a 2 L  Γ2 
QnB =  a31 a32 L a3 L  Γ3  ( 3.3 )
  
 M M O M  M 
a L a KL  ΓL 
 K1 aK 2
The matrix is a square matrix of order m, where m is the total number of panels on all the
blades.
The known quantities in Equation 3.1 can be grouped together and transferred to
the right hand side of the equation. The normal velocity induced by vortices in the wake
can be calculated using the Biot-Savart Law, since the strength and position of each wake
vortex is known at every time step. Summing the effects of all the wake vortices on one
blade panel K, the normal component of the induced velocity is
W
QnWK = ∑ q iK ⋅ n K = [uW , vW , ww ] ⋅ n K ( 3.4 )
i =1
where i covers all W wake vortices. The normal component of the free stream velocity of
each panel is also known, since the motion of the blades is predefined:
QnFSK = [U (t ),V (t ), W (t )]K ⋅ n K ( 3.5 )
Therefore, transferring the known quantities to the right hand side yields
RHS K = −[U (t ) + uW ,V (t ) + vW ,W (t ) + wW ]K ⋅ n K ( 3.6 )
Equation 3.1 can then be rearranged and solved for the unknown vortex strengths:
50
 a11 a12 L a1m  Γ1   RHS1 

    
 a 21 a 22 L a 2 m  Γ2   RHS 2 
a a32 L a3m  Γ3  =  RHS 3  ( 3.7 )
 31    
 M M O M  M   M 
a L a mm  Γm   RHS m 
 m1 am2
Once the circulation of each vortex ring on the blade is known, the lift per unit
span of each panel can easily be calculated using the Kutta-Joukowski theorem,
l K = ρU K ΓK ( 3.8 )
where UK is the free stream velocity of the panel and ΓK is the circulation of the panel [3].
For the leading edge panels, ΓK is just equal to the circulation of the panel, but for all
other panels, ΓK is equal to the difference between the circulation of that panel and the
circulation of the panel directly forward of that panel [18]. The actual lift on each panel
is
LK = l K ∆bK ( 3.9 )
where ∆bK is the spanwise dimension of the panel. Since the blade is modeled by a flat
plate, the pressure difference on the panel is simply
LK
∆p K = ( 3.10 )
∆S K
where ∆SK is the area of the panel. The sectional lift lj is found by summing the sectional
lift of all the panels at a given spanwise location j, and the sectional lift coefficient is
lj
Cl j =
1 ( 3.11 )
ρU ∞2 c
2
51
The thrust coefficient is
T
CT = ( 3.12 )
ρπR 2U ∞2
where the thrust T is equal to the total lift produced by all the blades.
3.2.3 Calculate Wake Roll-Up
The vortices in the wake cannot carry aerodynamic loads, so they move only with
a local velocity induced by the other vortices in the wake and on the blades. The
components of the local induced velocity are calculated by summing the components of
the velocity induced by all of the other vortices. The velocities are calculated at the
corner points of the wake vortices, which are then allowed to translate in space:
(∆x, ∆y, ∆z ) = (u, v, w)w ∆t ( 3.13 )
This is what allows the wake to freely deform by rolling-up, contracting, and convecting
downward below the rotor.
3.3 Implementation of the Free Wake Procedure in the Code
The free wake method was implemented into a Fortran program. Figure 3.3 is a
flow chart showing the structure of the program. Subroutines are shown in capital letters
in brackets, and flow charts for the important subroutines are presented in the Appendix.
The major components of the program will be described in detail in the order in which
they are performed.

52
Start A
Calculate aerodynamic influence

Read input data
coefficient matrix
(INPUT)
(CALCAIC)
Define geometry of the blades Calculate right hand side matrix

(READGRID) (CALCRHS)
Solve matrix equation for

DO I = 1, NSTEPS unknown blade vortex strengths
(SOLVER)
Calculate lift on blades

TIME = TIME + DT
(PRESSURE)
Calculate induced velocity of

Advance blade to next position
wake corner points
(MOVEWING)
(WAKEVELOC)
Shed trailing edge vortices Move wake corner points based

into wake on induced velocity
(SHEDVORTEX) (MOVEWAKE)
Calculate velocity components of Output location and strength of

each blade panel all vortices on blades and in wake
(VELOC) (TECOUT)
Calculate normal vectors for each No

blade panel I = NSTEPS?
(NORMAL)
Yes
A End
Figure 3.3: Computer program flowchart

53
The first step in the program is to input the parameters that define the problem in
the INPUT subroutine. These include the time step size, number of time steps, number of
blades, chord and span of the blades, pitch and coning angle of the blades, number of
elements in the chordwise and spanwise directions, vortex core radius, rotational speed of
the rotors, and velocity of the helicopter fuselage. All units are defined in the metric
system. A sample input file is given in the Appendix. Next, the geometry of the blades
is defined in the READGRID subroutine. The position of one blade is defined and panels
are applied to that blade. Cosine spacing is used in the chordwise direction so that
smaller panels are used near the leading and trailing edges and larger panels are used in
the middle of the blade. Similarly, semi-cosine spacing is used in the spanwise direction
from the root to the tip so that smaller panels are used near the tip. This type of spacing
allows for smaller panels to be used in the areas of interest but saves computational time
since larger panels can be used in areas that are less important. After the panels have
been assigned, the pitch and precone angles are applied to the blade. The ring vortices
are applied at the quarter-chord line of each panel, and the collocation points are applied
at the center of the three-quarter-chord line of each panel. Then, the remaining blades are
created by copying the first blade, and the blades are evenly spaced around the rotor disk
using the blade quarter-chord line, or the feathering axis, as a reference.
After the initial geometry of the blades is defined, the time loop is started that
runs over the length of the simulation. The first step in the loop is to advance each blade
to the current time step position using the MOVEWING subroutine. First, the location of
the center of the rotor disk is subtracted from the coordinates of the panels so that the
54
blade rotation and rectilinear motion of the helicopter can be considered separately. A
transformation matrix is constructed that accounts for the rotation of the blades, and the
old blade locations are multiplied by the transformation matrix to produce the new blade
locations due to rotation. The center of the rotor disk is then translated based on the
rectilinear motion of the helicopter, and the components of the location of the center
point due to translation and of the locations of the panels due to rotation are added to
produce the new locations of the panels. After the blades have been moved, a new row of
vortices is shed into the wake using the SHEDVORTEX subroutine. The new row of
vortices occupies the space previously occupied by the trailing-edge vortices of the
blades before the blades were moved. They are oriented such that they are at the same
angle of attack as the blade when shed. The strengths of the vortices in the first row of
the wake are set equal to the strengths of the trailing edge blade vortices from the
previous time step.
Next, with the blades in their current positions, the velocity of each panel is
calculated in the VELOC subroutine. The velocity is calculated at the collocation point
and is the sum of the rotating velocity of the blades and the translation velocity of the
helicopter. The normal vectors are then found for all blade panels in the NORMAL
subroutine. The normal vectors can easily be calculated using Equation 3.14,
A×B
nK = ( 3.14 )
A×B
where A and B are vectors defined by opposite corner points of the panel [18].
With the velocity and normal vectors calculated for each panel, it is possible to
start the process of solving for the strength of the vortex on each panel. The first step is
55
to calculate the aerodynamic influence coefficient matrix using the CALCAIC
subroutine. Recall that the influence matrix is a square matrix of order m, where m is the
total number of panels on all the blades. Each coefficient in the matrix represents the
effects of a unit-strength vortex ring at a panel L on the panel K. A series of nested loops
is used to calculate the coefficients. The first series of loops iterates over the blades, the
chordwise panels on each blade, and the spanwise panels on each blade. A counter then
assigns the value K to the current panel. With the panel fixed, a second series of loops
iterates over the blades, the chordwise panels on each blade, and the spanwise panels on
each blade. Another counter assigns the value L to the second panel. The subroutine
VORTEX is then used to calculate the induced velocity at K due to a vortex of unit
strength at L. The subroutine is applied to each side of the ring vortex at L and the
components of the velocities are summed. The components are multiplied by the
components of the normal vector at K and summed to produce one influence coefficient
at K due to L.
The second step in calculating the strengths of the vortices on the blades is to
determine the right hand side of the matrix equation using the CALCRHS subroutine.
Recall that the right hand side is made up of the normal component of the free stream
velocity and the normal component of the velocity induced by the wake. First, a series of
loops iterates over the blades, the chordwise panels on each blade, and the spanwise
panels on each blade, and the current panel is assigned the value K. Then, the part of the
right hand side due to the free stream velocity of the panel is calculated by multiplying
the components of the velocity of the panel K by the components of the normal vector at
56
the panel K and summing. Next, the part due to the wake is calculated by looping over
the number of blades, the number of time steps (where each time step represents one row
of panels shed into the wake), and the panels in each row of the wake. Again, the
VORTEX subroutine is used to calculate the velocity induced by each side of the ring
vortex in the wake, but this time, the known strengths of the wake panels are used in the
calculation to get an actual induced velocity rather than a coefficient. The components of
the velocity are summed and then multiplied by the components of the normal vector at
the panel K. Unlike in the CALCAIC subroutine, however, where the influence of each
panel was considered separately, now the effects of all the panels in the wake are added
to the effect of the free stream velocity to produce one total induced velocity field at the
panel K. When the matrix is completed, it is transferred to the right hand side of the
equation by subtracting it from both sides.
It is necessary at this point to describe in detail the VORTEX subroutine that was
used in the calculation of the influence matrix and the right hand side matrix. The
VORTEX subroutine uses the Biot-Savart Law and, when dealing with vortices on the
blade, the Prandtl-Glauert compressibility correction to calculate induced velocities. The
first step is to calculate the compressibility correction, if applicable. The perpendicular
vector between the vortex segment and the control point of the panel K is found. Then
the velocity and Mach number of the vortex along the perpendicular vector can be found
and the length of the vector is recalculated using the Prandtl-Glauert correction.
The Biot-Savart calculation begins by calculating the components of the cross
product of the vectors r1′ and r2′, which define the distance between the endpoints of the
57
virtual location of the vortex and the control point, and the components of the vector r0
that defines the direction of the vortex. If the perpendicular distance rp′ between the
control point and the virtual vortex is less than the size of the core radius, then the
components of the induced velocity are calculated using the solid body approximation:
 rp′  Γ  (r1′ × r2′ )i 

q i =     ( 3.15 )
 
 rc  2πrc  r1′ × r2′ 
If the control point is located outside the vortex core, then the components of the induced
velocity are calculated using the Biot-Savart Law:
 Γ  (r1′ × r2′ )i  r ⋅ r ′ r ⋅ r ′ 

 0 1 − 0 2 
qi =  
 r1′ r2′ 
( 3.16 )
 4π  r1′ × r2′
2

After the influence coefficient matrix and the right hand side matrix have been
established, the unknown vortex strengths are solved for in the SOLVER subroutine. The
SOLVER subroutine uses the LINPACK library to solve the matrix equation by LU
decomposition. LINPACK is a freely available library of Fortran subroutines designed to
solve linear systems [24].
Next, the lift on the blades is calculated in the PRESSURE subroutine. Since the
strength of each ring vortex on the blades is known, the lift on each panel can be
calculated using Equation 3.8. First, a loop iterates over all the blades and the total lift on
each blade is set to zero. The spanwise panel locations on each blade are looped over and
the spanwise lift coefficient at each location is set to zero. Then the chordwise panels are
looped over at each spanwise location. The free stream velocity of each panel is
determined, and the sectional lift and total lift of each panel are calculated. The sectional
58
lift of the panel is added to the cumulative sectional lift of the current spanwise location,
and, after the chordwise panel loop is completed, the sectional lift coefficient can be
calculated for that spanwise location. Similarly, the total lift for the panel is added to the
cumulative lift of the blade, and after the chordwise and spanwise loops are completed,
the total lift on the blade is known. Finally, the pressure and pressure coefficient are
calculated for each panel. Since the blade is represented by a flat plate, the pressure of
each panel represents the difference between the pressure on the upper and lower surfaces
of an airfoil. After the loop over the blades is completed, the thrust coefficient of the
blade configuration can be calculated.
Next, the wake is allowed to freely deform in the WAKEVELOC and
MOVEWAKE subroutines. First, the induced velocity due to the vortices on the blades
and in the wake is calculated at the corner points of each wake element. A series of loops
iterates over all the blades, the number of time steps (where each time step represents one
row of panels shed into the wake), and the panels in each row of the wake. Then, a
second series of loops iterates over all the blades and the panels on each blade, and the
VORTEX subroutine is used to calculate the induced velocity at the current wake corner
point due to all the vortices on the blades. The Prandtl-Glauert compressibility correction
is not used in the VORTEX subroutine when calculating the induced effects of one wake
vortex on another wake vortex. When the second set of loops is completed, a third series
of loops iterates over all the blades and the panels in the wake, and the induced velocity
at the current wake point due to all the vortices in the wake is calculated. The induced
velocities from the blades and the wake are summed to produce one induced velocity at
59
each corner point of the wake panels. Then, that induced velocity is used in the
MOVEWAKE subroutine, where all the wake corner points are looped over and allowed
to move based on their velocity. The first row of wake panels trailing off each blade is
not allowed to move to ensure that the flow leaves tangent to the blade and that the Kutta
condition is preserved.
Finally, the strength and position of the vortices on the blades and in the wake are
outputted to files in the TECOUT subroutine. The files are formatted to be read into
Tecplot where they can be viewed and analyzed. The data output is the last step in the
time loop, and the program is terminated after it loops over the total number of time
steps.
Chapter 4
RESULTS AND DISCUSSION
4.1 Overview
Two experimental cases were chosen as test cases to validate the free wake code.
The experiments were conducted by Caradonna and Tung to provide data to be used in
the validation of future rotor performance codes [5]. Blade pressure measurements were
made for a hovering two-bladed rotor over a wide range of tip Mach numbers from the
incompressible to transonic flow regimes. The two cases chosen to validate this code
represented an incompressible flow case (Mtip = 0.44) and a compressible flow case (Mtip
= 0.88). The compressible hover case is an unrealistic case from the standpoint of normal
helicopter operations, but it is useful as a test case because it introduces flow
characteristics normally only seen in high-speed forward flight. After the code was
validated, a number of other cases were run to demonstrate the versatility of the code.
The experiments were run in the Army Aeromechanics Laboratory hover facility,
a large chamber specially designed to eliminate recirculation. Figure 4.1 shows the set-
up of the rotor blades in the test facility. The blades were NACA 0012 airfoils with no
twist or taper and a half degree of precone. Each blade had a radius of 3.75 ft (1.143 m)
and an aspect ratio of 6. The root cutout was approximately equal to one chord. Each
blade was fitted with pressure taps at five spanwise locations, with more data being taken
61
near the tip and near the leading edge. Both validation cases presented here use a rotor
with 8° of collective pitch, with the first rotating at 1250 rpm, which corresponds to a tip
Mach number of 0.44, and the second rotating at 2500 rpm, which corresponds to a tip
Mach number of 0.88.
Figure 4.1: Configuration of the experimental rotor [5]
4.2 Code Validation
The computational model of the rotor blades used in both cases was the same.
The blades were modeled by a flat plate with 8° of collective pitch and 0.5° of precone.
The dimensions of the blades were set exactly equal to the dimensions of the blades used
62
in the experiment. Each blade had eight panels in the chordwise direction and ten panels
in the spanwise direction. Cosine spacing was used in the chordwise direction such that
the smallest panels were placed near the leading edge and trailing edge of the blade.
Likewise, semi-cosine spacing was used in the spanwise direction from the root such that
the smallest panels were near the blade tip. Figure 4.2 shows the layout of the panels on
one blade. Recall that each ring vortex is placed at the quarter-chord line of its
corresponding panel. The time step size was set in each case such that the blades
advanced 6° during each time step. The number of panels and the time step size were
based on simulations run by Katz and Maskew, who closely matched the data from the
same incompressible test case used here with their own free wake code [17]. The vortex
core radius was set equal to ten percent of the spanwise length of the smallest panels at
the tip to ensure proper roll-up of the tip vortex. The number of panels, the panel
spacing, and the core size were all changed to see what effect they had on the results. All
cases were run on the same machine with one 800 MHz processor and up to 1 GB of
RAM available.
Leading Edge
Root Tip
Trailing Edge
Figure 4.2: Grid spacing on blade
63
4.2.1 Case 1: Incompressible Hover
The first validation case was a case where the flow could be considered
incompressible everywhere, since the Mach number at the tip was only 0.44. Because the
flow is incompressible, a traditional Biot-Savart method can be used to evaluate it, but
the modified code including the Prandtl-Glauert correction should produce accurate
results as well. Therefore, the first case was run using both the incompressible and the
compressible versions of the code, and the results for both are presented and compared.
Each case ran for eight revolutions, or 480 time steps. In actual time, eight revolutions
represents 0.384 s. Convergence was declared after eight revolutions based on the fact
that variations in the thrust coefficient leveled off and that there were no significant
differences in the pressure readings from the one blade to the other. The incompressible
code ran in 33 hrs, while the compressible code ran in 43.67 hrs, with the difference
coming from the extra calculations required by the Prandtl-Glauert correction. Figure 4.3
shows the thrust coefficient over time for each code, as well as the experimentally
determined thrust coefficient of 0.00459. The large increase in thrust at the beginning is
due to the effects of the strong starting vortex shed when the blades start from rest. As
the starting vortex is convected away, its effects diminish and the thrust coefficient
eventually decreases to a steady state value.

64
Thrust coefficient, CT
Angle of rotation (deg)

Figure 4.3: Thrust coefficient over time
65
The most effective way to validate the accuracy of the code is to compare the
calculated lift coefficient and pressure coefficients with experimental data. Figure 4.4
shows the experimental lift coefficient and the lift coefficient computed by each version
of the code over the span of the blade. Both versions of the code produce very similar
results, as expected, and these results closely match the experimental data. Figure 4.5
shows a series of pressure coefficient plots at different spanwise locations. Since the
blade is modeled by a flat plate, the computed pressure represents the difference between
the pressure on the upper surface and the lower surface of an airfoil. Therefore, a delta
pressure coefficient is actually plotted for the each version of the code as well as for the
experimental data. The vortex lattice method for a flat plate is singular at the leading
edge, which is why the pressure the coefficient increases so rapidly there [18]. This
singular can be avoided by using a more accurate panel method. Again, both versions of
the code produce nearly identical results, although these results differ slightly from the
experimental results. The difference is most likely due to the resolution of both the
computational data and the experimental data. It is likely that more panels are required
near the leading edge in the computational method to fully capture the flow, but more
accurate experimental data could also be found using more pressure taps near the leading
edge, especially in the inboard region.

66
Sectional lift coefficient, Cl
Radial position, r/R

Figure 4.4: Spanwise lift coefficient (normalized by tip speed)
67
r/R = 0.50
Delta pressure coefficient, dCp (Cpl – Cpu)
r/R = 0.80
r/R = 0.96
Chordwise position, x/c

Figure 4.5: Chordwise pressure coefficient at different radial sections
68
The results from both the incompressible and the compressible versions of the
code are almost identical in all cases. This is expected since the compressibility
corrections should not have a significant impact on incompressible flow. Even at the tip,
where the Mach number is 0.44, the effect of the Prandtl-Glauert correction on the virtual
distance between a control point and a vortex is only about 10%. But the fact that the
compressible version of the code still works for incompressible flow is an important one
when trying to validate it. Still, since the incompressible version ran approximately 30%
faster than the compressible version while producing essentially the same results, it
makes sense to use it in cases where the flow is known to be incompressible everywhere.
4.2.2 Case 2: Compressible Hover
The second validation case was one where the flow near the tips was
compressible, with the tip Mach number being 0.88. This is the most interesting case,
since traditional Biot-Savart techniques cannot accurately capture the physics of
compressible flow. Several variations of this case were run to determine the best set-up
of the code to achieve a balance between accuracy and speed.
4.2.2.1 Case 2a: Incompressible Code vs. Compressible Code
This case was run with the same set-up as Case 1, using both the incompressible
and compressible versions of the code. Convergence was declared after eight revolutions
(480 time steps) with the incompressible code and ten revolutions with the compressible
69
code (600 time steps). The incompressible code ran in 33 hrs, while the compressible
code ran in 65.5 hrs. The large difference in run time is due to the extra revolutions and
the extra calculations required by the Prandtl-Glauert correction. Figure 4.6 shows the
thrust coefficient over time and the experimental thrust coefficient of 0.00473.
Figure 4.7 shows the experimental lift coefficient and the lift coefficient
computed by both versions of the code. It also shows the effects of applying the Prandtl-
Glauert correction directly to the lift coefficient. There is more lift at the tip in this case,
where the flow is compressible, but the incompressible method fails to predict this lift
increase. Although the compressible version slightly overpredicts lift inboard of the tip,
it does predict the increased lift at the tip. Simply applying the Prandtl-Glauert correction
to the lift coefficient is seen to severely overpredict lift, especially at the tip. This agrees
with the conclusions of Gennaretti and Morino that, while this technique may be useful
for fixed wings, it is not valid for use with rotating blades [14].
Figure 4.8 shows a series of pressure coefficient plots at different radial locations
for both the incompressible and compressible versions of the code. In the inboard region,
the methods produce nearly identical results, which is expected since the flow in this
region is still incompressible. Moving closer to the tip, however, the results begin to
differ as the compressible method predicts more lift than the incompressible method. The
compressible method, which will be used from this point forward, shows good agreement
with the experiment in the inboard region, but the agreement breaks down closer to the
tip, especially near the leading edge. This again is a result of the singularity of the
method at the leading edge.

70
Thrust coefficient, CT
Angle of rotation (deg)

Figure 4.6: Thrust coefficient over time
71

72
r/R = 0.50
r/R = 0.80
r/R = 0.96

Figure 4.8: Chordwise pressure coefficient at different radial locations
73
4.2.2.2 Case 2b: Changing the Chordwise Panel Spacing
Another simulation was run to determine the effect of the panel spacing in the
chordwise direction. The original case used full-cosine spacing, which concentrated
smaller panels and the leading and trailing edges, but the modified case used semi-cosine
spacing to concentrate panels at the just the leading edge. Both cases used 8 panels along
the chord and ran in same amount of time. Figure 4.9, the spanwise lift coefficient,
shows that the modified method predicts more lift at the tip than the original method and
more closely matches the experimental data inboard of the tip. It is difficult to determine
which method is more accurate, however, due to the sparse spacing of experimental data
points. Figure 4.10, the chordwise pressure coefficient, shows that the modified method
does a slightly better job at approaching the experimental data over the forward half-
chord. Again, it is difficult to determine which spacing method is better, but because the
pressure is changing much more near the leading edge than near the trailing edge, it
makes sense to concentrate smaller panels there.

74

75
r/R = 0.50
r/R = 0.80
r/R = 0.96

76
4.2.2.3 Case 2c: Changing the Vortex Core Radius
Two more simulations were run in which the vortex core radius was changed.
Recall that the original simulation used a core radius equal to 0.1% of the spanwise
dimension of the smallest blade panel. The first modified run used a core radius equal to
0.001% of the smallest panel, and the second run used a core radius equal to 0.5% of the
smallest panel. All three cases ran in the approximately same time, with only slight
differences coming from the number of calculations performed inside the vortex core
using the simpler solid body approximation versus the number performed outside the core
using the more complicated Biot-Savart Law for each case. Figures 4.11 and 4.12 show
that the results for all three cases are essentially the same, although the case with the
largest core radius predicts slightly more lift than the other two cases. Increasing the core
radius to a value greater than half the size of the panels at the tip will mean that the
distance along the span between the control point and the vortex ring which it defines
will be within the vortex core, which could adversely affect the roll-up of the tip vortex.
77

78
r/R = 0.50
r/R = 0.80
r/R = 0.96

79
4.2.2.4 Case 2d: Changing the Number of Spanwise Panels
Using semi-cosine spacing in the chordwise direction now and the original core
size, two simulations were run to test the effect of changing the number of panels in the
spanwise direction. The original case used 10 spanwise panels and ran in 65.5 hrs, while
the first modified case used 8 spanwise panels and ran in 43 hrs and the second used 15
spanwise panels but ran in 145 hrs. Figure 4.13, the spanwise lift coefficient, shows that
there is little difference in the spanwise lift between using 10 panels and 15 panels but
that using 8 panels underpredicts the lift. Likewise, Figure 4.14, the chordwise pressure
coefficient, shows that using more than 10 panels in the spanwise direction has almost no
impact on the results but that using less than 10 panels results in a lower prediction of the
pressure. Considering the large amount of extra computer time required to run with 15
panels, it is clear that there is no advantage over using 10 panels. And although using 8
panels allows the code to run in less time, the savings in time does not make up for the
loss in accuracy.
80

81
r/R = 0.50
r/R = 0.80
r/R = 0.96

82
4.2.2.5 Case 2e: Changing the Number of Chordwise Panels
Another case was run to test the effect of using more panels along the chord. The
original case used 8 chordwise panels and ran in 65.5 hrs, while the modified case used
10 chordwise panels and ran in 72 hrs. Figure 4.15 shows that using more panels
changes the lift distribution slightly along the span. Likewise, Figure 4.16 shows a small
difference between the results from each case, with the results from the case using 10
panels more closely matching the experimental data. Still, the improvement in the results
is small and not worth the extra computational cost at this stage.
83

Figure 4.15: Sectional lift coefficient (normalized by tip speed)
84

85
4.3 Wake Visualization
With the code validated using experimental lift and pressure data, it is possible to
look at the interesting characteristics of the wake structure. The wake essentially looks
the same for all the validation cases run, with a few exceptions, since they were all hover
cases. The roll-up of the tip vortex is slightly different depending on the number of
spanwise elements on the blade and the size of the vortex core. Still, looking at the wake
from the original compressible case (Case 2a) provides a good description of the flow
physics seen in the other cases. Figure 4.17 shows the fully developed wake trailing
behind both blades after ten revolutions, where only the near wake region is shown for
clarity. It shows the roll-up of the tip vortex and the vortex sheet being shed from the
blades. Looking at the wake from just one blade from the side in Figure 4.18, it is
possible to see the helical pattern of the wake as it is convected below the rotor disk.
Notice that as the vortex sheet convects downward below the rotor, it convects more
rapidly near the tip than it does inboard of the tip. This is due to the higher induced
velocities that exist at the tip and leads to the vortex sheet becoming increasingly sloped
from the root to the tip over time.

86
Figure 4.17: Wake trailing behind both blades

87
Figure 4.18: Wake trailing behind one blade
The convection of the tip vortices can be seen more clearly in Figure 4.19, where
the slipstream boundary is added for emphasis. The slipstream boundary separates the
88
region of turbulent flow in the wake from the region of quiescent flow outside the wake.
Notice that the tip vortices convect along the boundary and that the boundary contracts in
the far wake. Contraction of the wake is measured by the wake contraction ratio, which
is the ratio of the far wake radius to the radius of the blades. In this case, the wake
contraction ratio is approximately 0.73, which corresponds favorably to the value of
0.707 obtained using momentum theory and the more practical value of approximately
0.78 seen in experiments [22]. The roll-up of the tip vortex after 180° is shown in Figure
4.20. Although this method does not allow for a detailed description of the tip vortex, it
does at least show the presence of the roll-up.

89
Tip vortices
convect along
slipstream
Slipstream boundary
boundary
Figure 4.19: Path of tip vortices

90
Tip vortex roll-up
Figure 4.20: Roll-up of tip vortex at 180°

91
4.4 Application to Other Cases
Now that the code has been validated for hover using experimental data, it is
possible to use the code to analyze flow in other operating conditions, such as climb and
forward flight. Since the code was shown to be able to handle compressible flow, it can
be used in typical forward flight conditions, where the flow on the advancing blade is
compressible. The code does not currently trim the helicopter (i.e. account for blade flap,
pitch control, or lead-lag), however, so the results in forward flight should only serve as
an approximation.
4.4.1 Axial Climb
Using the same set-up as the validation cases, the code was used to simulate axial
climb at a rate of approximately 1000 ft/min (5 m/s). In a helicopter, the rotor would
need to provide extra thrust to climb, but in this case, since there is no weight to be lifted,
the rotor is kept at a constant thrust while being vertically translated at the climb speed.
This is equivalent to wind tunnel testing of a climbing rotor where the rotor is kept at
constant thrust and an external fan blows air down over the rotor [30]. The vertical flow
effectively reduces the angle of attack of the blade, which should result in less lift being
produced. Since the climb velocity (5 m/s) is relatively small compared to the free
stream velocity of the blades due to rotation (300 m/s at the tip), the lift in this case
should only be slightly less than the lift in the hover case. Figures 4.21 and 4.22 show
this is in fact what is predicted by the code.

92
Sectional lift coefficient, C

Figure 4.21: Sectional lift coefficient (normalized by tip speed)
93

94
4.4.2 Forward Climb
Another more interesting case is the combination of forward flight and climb,
which is more typical for a helicopter than a pure axial climb. Since this code does not
account for rotor trim, it cannot be used to accurately predict blade loading in forward
flight, but it can still approximate the wake generated in such a case. A 4-bladed rotor
with a diameter of 44 ft (13.4 m), aspect ratio of 16, collective pitch of 8°, and coning
angle of 3° was rotated at 325.5 rpm (Vtip = 750 ft/s, 230 m/s). The rotor had a forward
speed of 60 kts (31 m/s) and a climb speed of 3000 ft/min (15 m/s). Figure 4.23 tracks
the tip vortices coming off the blades seen from above and from the side. Recall that this
is very similar to the smoke visualization image shown in Chapter 1 (Figure 1.6).
95
(a) Top View
(b) Side View
Figure 4.23: Tip vortices from a 4-bladed rotor in forward climb

Chapter 5
CONCLUSIONS
5.1 Summary and Conclusions
A code has been developed that accurately predicts the wake of a hovering rotor
using a free wake vortex lattice method. The code solves the Biot-Savart Law to
calculate vortex-induced velocities and uses a Prandtl-Glauert based correction to account
for compressibility effects on the blade.
The accuracy of the code was determined by comparing the results with
experimental data for two hover cases. The code was seen to closely match the
experimental data for the first case where the flow was incompressible everywhere. It
was also shown to predict the increased lift in the compressible region at the tip in the
second case, whereas an incompressible version of the code did not. The results due not
agree well near the leading edge of the blade, however, because of the singularity
associated with the vortex lattice method and flat plate assumption. Using a panel
method would eliminate this singularity, but it would also increase the computational
time. Although the overall accuracy of the code was slightly less for the compressible
case than the incompressible case, the results are still encouraging considering the
complexity of the flow, the coarse grid used on the blades, and the assumptions made in
the modeling process.

97
Several variations of the compressible case were run to test the sensitivity of the
code to panel size, panel spacing, and vortex core radius. The first modified run used
semi-cosine spacing along the chord instead of full-cosine spacing. Although the results
were similar, it was determined that, because of the higher pressure gradients near the
leading edge as compared to the trailing edge, semi-cosine spacing provides for the best
use of computational resources. The second set of runs used different vortex core radii.
The code was seen to be relatively insensitive to the core radius, as long as the radius was
not larger than half the spanwise dimension of the smallest panel, which affected the roll-
up of the tip vortex. Increasing the number of panels in the spanwise direction did not
have any effect on the results, but it did greatly increase the computational time of the
simulation. Decreasing the number of panels improved the computational time but did
not provide the same level of accuracy. Increasing the number of panels in the chordwise
direction, however, did improve the results slightly but with an increase in the
computational time. Therefore, it was found that using eight panels along the chord and
ten panels along the span with semi-cosine spacing in each direction to cluster smaller
panels at the leading edge and the tip provides the best balance between accuracy and
computational time.
After the code was validated, it was used to evaluate two other test cases to
illustrate the versatility and potential of the code. First, the code was used to evaluate the
flow during an axial climb of the rotor. The vertical motion of the rotor decreases the
effective angle of attack of the blades, which produces slightly less lift than the hover
case. Second, the code was used to evaluate the flow during a forward climb of the rotor,
98
which is more typically seen in helicopters than pure axial climb. Although the code
does not account for blade flapping, pitch control, or lead-lag, it does provide an
approximation of the wake structure and position in forward flight.
5.2 Suggestions for Future Work
The code showed good accuracy when compared to experimental data in both
validation cases, but the agreement for the compressible case shows room for
improvement. The most obvious choice for improving the results would be to use more
panels on the blade or a smaller time step. Unfortunately, this does not always increase
the accuracy of the code and can, in some cases, actually make the results worse. Using
more panels near the leading edge will not eliminate the singularity associated with using
the vortex lattice method on a flat plate either, although this could be dealt with by using
a panel method. The downside to using more panels, a smaller time step, or a panel
method, however, is the increased computational time. So before attempting to improve
the results, it is necessary to look at ways to improve the speed of the code. The easiest
way to do that would be to allow the code to run on multiple processors in parallel.
Another common technique to improve the speed of rotor codes is to truncate the wake
after a set number of revolutions, keeping only the tip vortex. Because the effect of the
wake on the blades diminishes the as the wake moves farther downstream, truncating the
far wake should have little effect on the lift of the blades. Another possible way to
improve the speed would be to only use the compressible correction in the near wake
region, where the compressibility effects are greatest.

99
Although the goal in this case was to determine the lift on the rotor blades,
another common use of the free wake method is to predict wake interference effects with
the fuselage or tail rotor. Unfortunately, in its current state, this method creates unnatural
distortions of the far wake over time. So, in order to be able to use the far wake geometry
for practical purposes, the code must be modified to eliminate these distortions. This is
usually done by allowing the vortex core radius to grow over time, which reduces the
induced velocities and damps out the wake, or by some other means of artificial damping.
As was shown by the axial climb and forward climb cases, the code has the
potential to evaluate many different types of operating conditions of a helicopter. To
truly analyze forward flight, though, the code must first be able to account for rotor trim.
Understanding the complex coupling of blade flapping, cyclic pitch, and lead-lag is not
an easy task, however, as it requires an advanced knowledge of rotorcraft dynamics.
Likewise, implementing such a scheme into the code is not completely straightforward.
Solving for the rotor trim is an iterative process, which must then be applied to the blade
grids to recalculate the positions of all the panels at each time step. If a trim code is
added, though, any prescribed flight path could be input into the code, and the wake
could be accurately modeled.
The code could easily be adapted to handle multiple rotors for a tilt-rotor or
tandem-rotor helicopter. Likewise, a mirror image of the rotor could be used to simulate
ground effect by placing it below the ground plane at a distance equal to the height of the
rotor. Combining the use of multiple rotors and ground effect would allow tilt-rotor and
tandem-rotor helicopters to be studied in ground effect or, more interestingly, in the

100
approach to a naval ship or elevated helipad where one rotor is in ground effect and the
other is not. Because the code can handle rotation of the blades about more than one
axis, it could also be used to analyze the wake during the transition phase of a tilt-rotor as
it changes from helicopter to airplane or vice versa.

BIBLIOGRAPHY
1. Beddoes, T.S. “A Wake Model for High Resolution Airloads.” Proceedings of
the 2nd International Conference on Basic Rotorcraft Research. Triangle Park,
NC, 1985.
2. Berkey, D.D. and Blanchard, P. Calculus, 3rd Edition. Saunders College
Publishing, 1992.
3. Bertin, J.L. and Smith, M.L. Aerodynamics for Engineers, 3rd Edition. Prentice
Hall, Upper Saddle River, NJ, 1998.
4. Berkman, M.E., Sankar, L.N., Berezin, C.R., and Torok, M.S. “A Navier-
Stokes/Full Potential/Free Wake Method for Advancing Multi-Blade Rotors.”
Proceedings of the 53th Annual Forum of the American Helicopter Society.
Virginia Beach, VA, May 1997.
5. Caradonna, F.X. and Tung, C. “Experimental and Analytical Studies of a Model
Helicopter Rotor in Hover.” NASA TM-81232, 1981.
6. Chen, C.L. and McCroskey, W.J. “Numerical Simulation of Helicopter Multi-
Bladed Rotor Flow.” AIAA 26th Aerospace Sciences Meeting. AIAA-88-0046.
Reno, NV, Jan. 1988.

102
7. Clark, D.R. and Leiper, A.C. “The Free Wake Analysis, A Method for the
Prediction of Helicopter Rotor Hovering Performance.” Journal of the American
Helicopter Society. Vol. 15, No. 1, Jan. 1970, pp. 3-11.
8. Currie, I.G. Fundamental Mechanics of Fluids. McGraw-Hill, 1993.
9. Duque, E.N. and Srinivasan, G.R. “Numerical Simulation of a Hovering Rotor
Using Embedded Grids.” Proceedings of the 48th Annual Forum of the American
Helicopter Society. Washington, DC, June 1992.
10. Egolf, T.A. and Landgrebe, A.J. “Generalized Wake Geometry for a Helicopter
Rotor in Forward Flight and Effect of Wake Deformation on Airloads.”
Arlington, VA, May 1984.
11. Egolf, T.A. and Massar, J.P. “Helicopter Free Wake Implementation on
Advanced Computer Architectures.” 2nd International Conference on Basic
Rotorcraft Research. College Park, MD, Feb. 1988.
12. Epstein, R.J. and Bliss, D.B. “Wake Generation Compressibility Effects in
Unsteady Lifting Surface Theory.” 15th AIAA Applied Aerodynamics Conference.
Atlanta, GA, June 1997.
13. Fung, Y.C. An Introduction to the Theory of Aeroelasticity. John Wiley & Sons,
Inc., New York, NY, 1955.

103
14. Gennaretti, M. and Morino, L. “A Boundary Element Method for the Potential,
Compressible Aerodynamics of Bodies in Arbitrary Motion.” Aeronautical
Journal. Vol. 96, Jan. 1992, pp. 15-19.
15. Hall, C.M. “High-Order Accurate Simulations of Wake and Tip Vortex
Flowfields.” M.S. Thesis, Pennsylvania State University, Department of
Aerospace Engineering, Dec. 1998.
16. Johnson, W. Helicopter Theory. Dover Publications, Inc., New York, 1994.
17. Katz, J. and Maskew, B. “Unsteady Low-Speed Aerodynamic Model for
Complete Aircraft Configurations.” Journal of Aircraft. Vol. 25, No. 4,
Apr.1988, pp. 302-310.
18. Katz, J. and Plotkin, A. Low-Speed Aerodynamics, 2nd Ed. Cambridge University
Press, 2000.
19. Kinsler, L.E., Frey, A.R., Coppens, A.B., Sanders, J.V. Fundamentals of
Acoustics, 4th Edition. John Wiley & Sons, Inc., 2000.
20. Landgrebe, A.J. “An Analytical Method for Predicting Rotor Wake Geometry.”
Journal of the American Helicopter Society. Vol. 14, No. 4, Oct. 1969, pp. 20-32.
21. Landgrebe, A.J. “The Wake Geometry of a Hovering Helicopter Rotor and its
Influence on Rotor Performance.” Journal of the American Helicopter Society.
Vol. 17, No. 4, Oct. 1972, pp. 3-15.

104
22. Leishman, J.G. Principle of Helicopter Aerodynamics. Cambridge University
Press, 2000.
23. Leishman, J.G., Martin, P.B, and Pugliese, G.J. “Surface and Wake Flow
Characteristics of a Hovering Helicopter Rotor.” 9th International Symposium on
Flow Visualization. Edinburgh, Scotland, UK, Aug. 2000.
24. LINPACK web page. (http://www.netlib.org/linpack/index.html).
25. Long. L.N. “The Compressible Aerodynamics of Rotating Blades Based on an
Acoustic Formulation.” NASA TP-2197, 1983.
26. Long, L.N. and Watts, G.A. “Arbitrary Motion Aerodynamics Using an
Aeroacoustic Approach.” AIAA Journal. Vol. 25, No. 11, Nov. 1987, pp. 1442-
1448.
27. Miranda, L.R., Elliot, R.D., and Baker, W.M. “A Generalized Vortex Lattice
Method for Subsonic and Supersonic Flow Applications.” NASA CR-2865, 1977.
28. Moulton, M.A., Wenren, Y., and Caradonna, F.X. “Development of an
Overset/Hybrid CFD Method for the Prediction of Hovering Performance.”
Virginia Beach, VA, May 1997.
29. Morino, L. “A General Theory of Unsteady, Compressible, Potential
Aerodynamics.” NASA CR-2462, 1974.

105
30. Prouty, R.W. “Helicopter Aerodynamics.” Course Notes from A Comprehensive
Short Course in Rotary Wing Technology. University Park, PA, Aug. 2001.
31. Sadler, S.G. “A Method for Predicting Helicopter Wake Geometry, Wake-
Induced Flow and Wake Effects on Blade Airloads.” Proceedings of the 27th
Annual Forum of the American Helicopter Society. Washington, DC, May 1971.
32. Schreier, S. Compressible Flow. John Wiley & Sons, Inc., 1982.
33. Sezer-Uzol, N. “High Accuracy Wake and Vortex Simulations Using a Hybrid
Euler/Discrete Vortex Method.” M.S. Thesis, Pennsylvania State University,
Department of Aerospace Engineering, May 2001.
34. Srinivasan, G.R., Baeder, J.D., Obayashi, S., and McCroskey, W.J. “Flowfield of
a Lifting Rotor in Hover: A Navier-Stokes Simulation.” AIAA Journal. Vol. 30,
No. 10, Oct. 1992, pp. 2371-2378.
35. Stahl-Cucinelli, H. “Vortex-Lattice Free Wake Model For Helicopter Rotor
Downwash.” 20th European Rotorcraft Forum. Amsterdam, NL, Oct. 1994.
36. Steinhoff, J. and Ramachandran, K. “Free Wake Analysis of Compressible Rotor
Flows.” AIAA Journal. Vol. 28, No. 3, Mar. 1990, pp. 426-431.
37. Strawn, R.C. and Barth, T.J. “A Finite-Volume Euler Solver for Computing
Rotary-Wing Aerodynamics on Unstructured Grids.” Proceedings of the 48th
Annual Forum of the American Helicopter Society. Washington, DC, June 1992.
106
38. Tauszig, L. “Numerical Detection and Characterization of Blade Vortex
Interactions Using a Free Wake Analysis.” M.S. Thesis, Pennsylvania State
University, Department of Aerospace Engineering, Aug. 1998.
39. Tung, C., Caradonna, F.X., and Johnson, W. “The Prediction of Transonic Flows
on Advancing Rotors.” Journal of the American Helicopter Society. Vol. 31, No.
3, July 1986, pp. 4-9.
40. U.S. Navy Photo Gallery web page.
(http://www.chinfo.navy.mil/navpalib/.www/digital.html).
41. Ward, G.N. Linearized Theory of Steady High-Speed Flow. Cambridge
University Press, 1955.

Appendix A
SAMPLE INPUT FILE
! Timestep size, dt (s)

0.0004
! Number of time steps, nsteps

600
! Rotational speeds of rotor, rmpx, rpmy, rpmz (rpm)

0, 0, 2500
! Number of blades, Nb
2
! Number of nodes (panels + 1) along chord and span, nx, ny

9, 11
! Maximum size of wake, nwakemax

600
! Aspect Ratio, AR
6
! Radius, R (m)
1.143
! Root cutout, e_by_c (number of chords)

1.0
! Core size as fraction of smallest spanwise element, rc

0.1
! Fuselage velocity in x, y, and z directions, u, v, w (m/s)

0, 0, 0
! Rotation point of rotor in space, hx, hy, hz

0, 0, 0
! Collective pitch, theta (deg)

8.0
! Precone Angle, beta (deg)

0.5
! Interval for Writing Tecplot Data (60 iterations = 1 revolution)

60
Appendix B
COMPUTER PROGRAM SUBROUTINES
WAKE
Set angular location of each blade

by spacing evenly around rotor
disk
Define positions of panels on the

first blade
Calculate size of each panel on

the first blade
Adjust position of first blade

based on pitch and precone angle
Calculate position of collocation

points on the first blade
Copy all information from the

first blade and paste at the
locations of the other blades
Stop
Figure B.1: READGRID subroutine

WAKE
K=0
DO N = 1, NB
DO I = 1, NX - 1
L=0
DO J = 1, NY - 1
K=K+1
DO N2 = 1, NB
DO I2 = 1, NX - 1
DO J2 = 1, NY - 1 L=L+1
Calculate induced velocity at vortex

K due to vortex L (VORTEX)
No No
J = NY – 1? J2 = NY – 1? AIC(K,L) = (u,v,w)·nK
Yes Yes
No No
I = NX – 1? I2 = NX – 1?
Yes Yes
No Yes No
N = NB? N2 = NB?
Yes
Stop
Figure B.2: CALCAIC subroutine

WAKE
K=0
DO N = 1, NB
DO I = 1, NX - 1
DO J = 1, NY - 1 RHS(K) = (U,V,W)·nK
DO N2 = 1, NB
DO I2 = 1, ITER - 1
DO J2 = 1, NY - 1 Calculate induced
velocity at panel K
due to vortex L
(VORTEX)
No No RHS(K) = (u,v,w)·nK
J = NY – 1? J2 = NY – 1? + RHS(K)
Yes Yes
No No
I = NX – 1? I2 = ITER – 1?
Yes Yes
No Yes No
N = NB? N2 = NB?
Yes
Stop
Figure B.3: CALCRHS subroutine

WAKE
Calculate dot products,

r0·r1, r0·r2
Calculate length of vortex

segment, r0
Calculate perpendicular
vector between vortex and
point
Calculate component of
velocity of vortex segment
along perpendicular vector
Stretch perpendicular vector

using Prandtl-Glauert
correction and locate virtual
control point
Recalculate dot products and

cross product, r1 × r2
Calculate induced velocity

Yes
rp ≤ rc using solid body rotation
approximation
No
Calculate induced velocity

using Biot-Savart Law
Stop
Figure B.4: VORTEX subroutine

WAKE
DO N = 1, NB
LIFT(N) = 0
DO J = 1, NY - 1
DCL(J,N) = 0
Determine free stream velocity

DO I = 1, NX - 1
of panel
DL(I,J,N) = ρUΓ
No DCL(J,N) = DL(I,J,N)
I = NX – 1?
+ DCL(J,N)
Yes
CL(J,N) = LIFT(N) = LIFT(N) +
DCL(J,N)/(0.5cρU2) DL(I,J,N)*DY(I,J,N)
No P(I,J,N) =
J = NY – 1?
DL(I,J,N)*DY(I,J,N)/DS(I,J,N)
Yes
CT = T/ρAΩ2R2 CP(I,J,N) = P(I,J,N)/(0.5ρU2)
No
N = NB?
Yes
Stop
Figure B.5: PRESSURE subroutine

WAKE
DO N = 1, NB
DO I = 1, ITER - 1
DO J = 1, NY - 1 DO N2 = 1, NB
DO I2 = 1, NX - 1 DO I2 = 1, ITER - 1
DO J2 = 1, NY - 1 DO J2 = 1, NY - 1
Calculate induced velocity at Calculate induced velocity at

wake corner point due to wake corner point due to
blade vortices (VORTEX) wake vortices (VORTEX)
UWAKE(I,J,N) = (u,v,w) + UWAKE(I,J,N) = (u,v,w) +

UWAKE(I,J,N) UWAKE(I,J,N)
No No
J2 = NY – 1? J2 = NY – 1?
No
J = NY – 1? Yes Yes
No No
Yes I2 = NX – 1? I2 = ITER – 1?
No
I = NX – 1? Yes Yes
Yes No
Yes N2 = NB?
No
N = NB?
Yes
Stop
Figure B.6: WAKEVELOC subroutine

Computational Free Wake Analysis of A Helicopter Rotor

Uploaded by

Copyright:

Available Formats

Computational Free Wake Analysis of A Helicopter Rotor

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computational Free Wake Analysis of A Helicopter Rotor

Uploaded by

Copyright:

Available Formats

The Pennsylvania State University

The Graduate School

Department of Aerospace Engineering

COMPUTATIONAL FREE WAKE ANALYSIS OF A HELICOPTER ROTOR

Submitted in Partial Fulfillment

calculated by applying the boundary condition of no flow normal to the blade.

On a helicopter in forward flight, the advancing blades usually experience high

motion of the helicopter.

LIST OF FIGURES ...................................................................................................... vi

Chapter 1 INTRODUCTION ....................................................................................... 1

Chapter 2 FUNDAMENTALS OF POTENTIAL FLOW............................................ 19

Chapter 3 FREE WAKE METHOD ............................................................................ 44

Chapter 4 RESULTS AND DISCUSSION.................................................................. 60

4.2 Code Validation .............................................................................................. 61

Chapter 5 CONCLUSIONS ......................................................................................... 96

5.1 Summary and Conclusions.............................................................................. 96

BIBLIOGRAPHY ........................................................................................................ 101

Appendix A SAMPLE INPUT FILE............................................................................ 107

Appendix B COMPUTER PROGRAM SUBROUTINES .......................................... 108

Figure 1.4: Tip vortices trailed behind an E/A-6B Prowler ......................................... 8

Figure 2.3: Velocity at point P due to a vortex distribution in volume V .................... 24

Figure 2.4: Velocity at point P induced by a vortex segment....................................... 24

Figure 2.5: Velocity at point P induced by a vortex segment....................................... 26

Figure 2.8: Comparison of different 2-D vortex models.............................................. 32

Figure 2.9: Effect of Prandtl-Glauert transformation: (a) actual domain in

Figure 2.10: Stretching effect on a rotating blade ........................................................ 41

Figure 3.1: Vortex ring model...................................................................................... 46

Figure 3.2: Wake shedding procedure.......................................................................... 47

Figure 3.3: Computer program flowchart..................................................................... 52

Figure 4.1: Configuration of the experimental rotor .................................................... 61

Figure 4.2: Grid spacing on blade ................................................................................ 62

Figure 4.3: Thrust coefficient over time....................................................................... 64

Figure 4.4: Spanwise lift coefficient (normalized by tip speed) .................................. 66

Figure 4.5: Chordwise pressure coefficient at different radial sections ....................... 67

Figure 4.6: Thrust coefficient over time....................................................................... 70

Figure 4.7: Spanwise lift coefficient (normalized by tip speed) .................................. 71

Figure 4.8: Chordwise pressure coefficient at different radial locations...................... 72

Figure 4.9: Spanwise lift coefficient (normalized by tip speed) .................................. 74

Figure 4.10: Chordwise pressure coefficient at different radial sections ..................... 75

Figure 4.11: Spanwise lift coefficient (normalized by tip speed) ................................ 77

Figure 4.12: Chordwise pressure coefficient at different radial sections ..................... 78

Figure 4.13: Spanwise lift coefficient (normalized by tip speed) ................................ 80

Figure 4.14: Chordwise pressure coefficient at different radial sections ..................... 81

Figure 4.15: Sectional lift coefficient (normalized by tip speed)................................. 83

Figure 4.16: Chordwise pressure coefficient at different radial sections ..................... 84

Figure 4.17: Wake trailing behind both blades ............................................................ 86

Figure 4.18: Wake trailing behind one blade ............................................................... 87

Figure 4.19: Path of tip vortices ................................................................................... 89

Figure 4.20: Roll-up of tip vortex at 180°.................................................................... 90

Figure 4.21: Sectional lift coefficient (normalized by tip speed)................................. 92

Figure 4.22: Chordwise pressure coefficient at different radial sections ..................... 93

Figure B.1: READGRID subroutine ............................................................................ 108

Figure B.2: CALCAIC subroutine ............................................................................... 109

Figure B.3: CALCRHS subroutine .............................................................................. 110

Figure B.4: VORTEX subroutine................................................................................. 111

Figure B.5: PRESSURE subroutine ............................................................................. 112

Figure B.6: WAKEVELOC subroutine........................................................................ 113

this work was invaluable.