Yuan PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 208

Circulation Methods in Unsteady and Three-dimensional Flows

by

Jiankun Yuan


A Dissertation

Submitted to the Faculty

of the

WORCESTER POLYTECNHIC INSTITUTE

In partial fulfillment of the requirements for the

Degree of Doctor of Philosophy

in

Mechanical Engineering

by

________________________________

March 2002



APPROVED:

__________________________
Prof. David J. Olinger, Advisor

__________________________
Prof. Hamid Johari

__________________________
Prof. James C. Hermanson

__________________________
Prof. Mikhail Dimentberg

______________________________
Prof. Peder C. Pedersen

______________________________
Dr. Thomas J. Gieseke
Naval Undersea Warfare Center
Newport, RI

______________________________
Prof. Zhikun Hou
Graduate Committee Representative
i

























2002
Jiankun Yuan
All Rights Reserved
















ii


















Dedicated to the memory of my PARENTS
























iii
Abstract

The largely unstudied extension of ultrasonic circulation measurement techniques
(UCMT) to determine instantaneous lift in unsteady and three-dimensional flows has
been addressed in this work. A combined analytical-numerical-experimental approach
was undertaken with the goal of developing methods to properly convert the measurable
time-dependent bound circulation to instantaneous lift force in unsteady flows. The
measurement of mean sectional lift distribution along structure spans in three-
dimensional flows was also studied.
An unsteady correction method for thin airfoils was developed analytically and
validated numerically (with finite element solutions) to properly convert bound
circulation to instantaneous lift based on unsteady potential flow theory. Results show
that the unsteady correction method can provide increased accuracy for unsteady lift
prediction over the Kutta-Joukowski method used in previous unsteady flow studies. The
unsteady correction model generally should be included for instantaneous lift prediction
as long as the bound circulation is time-dependent.
Using the same framework, we also studied determination of instantaneous lift
forces on stationary bluff bodies (circular cylinders) at low Reynolds number (Re=100).
Various force models, including an approximate vortex force model, were studied. A new
unsteady model, similar to that developed for the thin airfoils, using instantaneous bound
circulation values, was proposed.
Another important issue studied in this thesis is the effect of acoustic path
sensitivity on bound circulation determination, which we found to be crucial for
iv
accurately predicting the instantaneous lift in both unsteady flat plate and cylinder flows.
Proper path selection should take into account the location of boundary layers, attached
and shed vortices. These findings will be useful in future experimental design of UCMT,
PIV and LDV methods.
Finally, we used the UCMT method to experimentally study the mean spatial lift
distribution along structures. Low Reynolds number low aspect ratio (AR) wings that
have application in micro-aerial-vehicles (MAV) were studied. The spanwise circulation
(lift) distribution along the MAV wings exhibits a peak (maximum), and deviates from
predictions of Prandtls lifting line theory. Although only linear lift (due to bound
circulation) was measured, comparison with force balance results showed that reasonable
integrated lift values on low Re, low AR wings can be obtained using UCMT.











v
Acknowledgments

I would like to first express my deep gratitude to my advisor, Prof. David J.
Olinger, for his excellent guidance and support in the development of this work detailed
in this dissertation. Without his outstanding ability as a research advisor, this work would
be impossible.
Thanks are also extended to the members of my thesis committee; Prof. Hamid
Johari for his useful suggestions in the experimental work, and Prof. James Hermanson,
Prof. Mikhail Dimentberg, Prof. Peder C. Pedersen, Dr. Thomas J. Gieseke and Prof.
Zhikun Hou for their time and assistance in this work.
I also wish to thank Prof. Andreas N. Alexandrou for his guidance in using the
finite element code. Thanks to Ganapathi R. Balasubramanian for help in numerical grid
generation. I would like to acknowledge Francis J. Weber and Ken Desabrais for their
help in the UCMT experiments. Thanks to Dr. Qingyun Zhang for his help to
accommodate the lab work in my first year at WPI.
My eternal gratitude to my parents, my brother and sisters, for their endless love
and support, I would not have made it this far without their encouragement. I also wish to
acknowledge support from the Office of Naval Research (ONR) under the direction of
Dr. Thomas Swean and the Mechanical Engineering Department during this work.
Finally, my deepest thanks to my wife Zhiying Hou and my lovely daughter Lin
Yuan for their love, patience and understanding during the past few years. Thanks for
their unconditional support and love despite my moods when the stress and intensity of
this work overwhelmed me. For this, I will always be deeply indebted.
vi
Table of Contents

Abstract...iii
Acknowledgements..v
Table of Contents....vi
List of Figures..x
List of Tables....xvi
Nomenclature...xvii
Ch.1 Introduction..1
1.1 Background1
1.2 Review...5
1.2.1 Theoretical work..6
1.2.2 Experimental work.14
1.3 Major goals..24

Ch.2 A Method for Converting Circulation to Lift Based on Unsteady Potential
Flow Theory...31
2.1 Introduction.31
2.2 Unsteady lift....36
2.2.1 Limitation of UCMT for ) , ( t x determination..39
2.3 Approximations...42
2.4 Results and discussion....47
2.5 Corrected unsteady approximation.55
vii
2.6 Correction in other unsteady flows.....56
2.6.1 Plunging flat plate flow..56
2.6.2 Pitching flat plate flow...58
2.6.3 Impulsively started plate flow59
2.7 summary...60

Ch. 3 Validation of Corrected Unsteady Approximation
Using Numerical Simulation..62
3.1 Introduction..62
3.2 Numerical formulation.64
3.3 Circulation determination67
3.4 Code validation70
3.4.1 Grid effects72
3.4.2 Time step effects...74
3.5 Bound circulation determination..74
3.5.1 Background...74
3.5.2 Effect of closed integration path variation....76
3.6 Results for oscillating plate flow.80
3.6.1 Correction of experimental data from pitching plate study..86
3.7 Summary..88

Ch. 4 Circulation Methods for Instantaneous Lift Force Determination Circular
viii
Cylinder Flow....99
4.1 Introduction..99
4.2 Numerical method..105
4.3 Wake circulation distribution and vortex lift force....106
4.4 Bound circulation determination113
4.5 Results and discussion...119

Ch. 5 Experimental UCMT Study on Three-dimensional Flows..128
5.1 Introduction128
5.2 Experimental apparatus..133
5.2.1 Experimental study for UCMT measurements...133
5.2.2 Experimental setup for force balance measurement...140
5.3 Data analysis..144
5.4 Validation of UCMT measurement...145
5.5 Results for flow over plates with varying chord length ...147
5.5.1 Spanwise lift distributions..148
5.5.2 Comparison between UCMT and force balance results.154
5.6 Low Reynolds, low aspect ratio wing study..157
5.7 Summary165

Ch. 6 Conclusions..168
6.1 Summary of results168
ix
6.2 Future work....172
Appendix A Derivation of Unsteady Correction Factor, R..175
Appendix B Lift Deficiency Factor, C(k)....180
References....181



































x
List of Figures

Figure 1.1 Fluid-cable interaction by direct numerical simulation (Newman &
Karniadakis, 1996).

Figure 1.2 Schematic of nonlinear coupling between fluid force, structural response
and fluid loading.

Figure1.3 (a) Body (xyz) and Initial (XYZ) coordinate systems (Fig13.1 of Katz &
Plotkin (1991)). (b) 2D view with nomenclature (Fig. 3.14 of Katz &
Plotkin (1991)).

Figure1.4 Lift force comparison DPIV data; Force balance data (Fig.2 of Noca
et al (1997)).

Figure 1.5 Schematic of a closed sound path for UCMT.
Figure 2.1 Schematic of a straight path (a) and closed path (b) for UCMT method.
Figure 2.2 Unsteady lift values during pitching motion of thin airfoil measured with
UCMT, from Weber (1995).
Figure 2.3 Nomenclature for the oscillating motion of the flat plate.
Figure 2.4 The plate chordwise circulation distribution required for determination
of (x,t).

Figure 2.5 A typical lift variation with time using Katz & Plotkins (1991) analysis
of an oscillating thin airfoil (U=5m/s, c=4, k=0.35, h
a
/c=0.05, =2
0
).

Figure 2.6 (a) Required set-up to measure (x,t) using UCMT. (b) (t) measurement
using UCMT technique.

Figure 2.7 Typical time traces of lift curves for quasi-steady (K-J) lift L
(1)
, unsteady
approximation L
(2)
compared to true lift curve L for oscillating plate
(U=5m/s, =2
0
, k=0.35, h
a
/c=0.05).

Figure 2.8 Time traces of the relative errors for approximation 1 and approximation
2 for the case of U=5m/s, =2
0
, k=0.1, h
a
/c=0.1.

Figure 2.9 Variation of phase angle with key parameters. a) effect of plate
oscillating amplitude h
a
/c . (b) effect of reduced frequency k. (c) effect
of angle of attack .
xi

Figure 2.10 Variation of rms error with key parameters. (a) effect of plate oscillating
amplitude h
a
/c . (b) effect of reduced frequency k. (c) effect of angle of
attack .

Figure 2.11 3D plots of
) 2 (
rms
e variation with plate oscillating amplitude h
a
/c, reduced
frequency k at different angle of attack .

Figure 2.12 Contour of
) 2 (
rms
e variation with plate oscillating frequency k, amplitude
h
a
/c at different angle of attack .

Figure 2.13 Schematic of a plunging flat plate.

Figure 2.14 Schematic of pitching plate flow.

Figure 2.15 Schematic of a impulsively started plate flow.

Figure 3.1 Numerical grids used in the finite element simulations for a domain
3<x/c<6, -2<y/c<2. (a) grid 1: NM=1043, NND=4330, (b) grid 2:
NM=1970, NND=8124.

Figure 3.2 Schematic of circulation determination.

Figure 3.3 Grids effects on the results with Dt=1/40 and =6
0
. (Grid 1: NM=1043,
NND=4330; Grid 2: NM=1970, NND=8124). (a) For stationary plate.
(b) For oscillating plate at k=0.15 and h
a
/c=0.05.

Figure 3.4 Time step effects on the results with grid 1: NM=1043, NND=4330,
=6
0
. (t
1
=1/40, t
2
=1/100). (a) For stationary plate. (b) For oscillating
plate at k=0.15 and h
a
/c=0.05.

Figure 3.5 Schematic of sound path effects on bound circulation determination.

Figure 3.6 Circulation variation with the movement of each leg of mimic acoustic
path (reference to Fig 3.5, =6
0
). (a) Before leading edge; (b) After
trailing edge; (c) Away from lower surface; (d) Away from upper surface.

Figure 3.7 Schematic of requirement for acoustic path size and locations to make
circulation measurements path independent in oscillating plate flow at
Re=100, =6
0
.

Figure 3.8 Flow visualization represented by contour of u-component of velocity in
the field around a flat plate with =6
0
, h
a
/c=0.15, k=0.15 and Re=100.
xii
The five stations flow fields a, b, c, d, e are corresponding to five lift
values in one cycle of the lift time trace.

Figure 3.9 A typical case of plate instantaneous location and bound circulation
variation with time, k=0.4, h
a
/c = 0.1.

Figure 3.10 Typical lift force obtained from different methods at k=0.4, h
a
/c=0.1.

Figure 3.11 Phase difference variation with plate oscillating amplitude and reduced
frequency at Re=100.

Figure 3.12 Rms error variation with plate oscillating amplitude and reduced
frequency.

Figure 3.13 Fig 3.14 Typical numerically attainable circulation distributions along
plate at an instant. Re=100, =6
0
, k=0.2, h
a
/c=0.1. (a) local chordwise
vortex element strength distribution. (b) Integrated vorticity distributions
from the optimum path and a path just from leading edge (LE) to trailing
edge (TE).

Fig.3.14 Application of corrected unsteady methods to Weber et al (1995)
measurements on a pitching NACA 4418 airfoil. ) 10 sin( 11 7
0 0
t + = ,
k=0.059, Re=5.410
5
. (a) Lift coefficient and correlation measurements of
Weber et al (1995). (b) Behavior of first derivative and unsteady
correction factor R (from eq. (A.19)). (c) Corrected lift coefficients
compared to Weber et al (1995) original results.

Fig 3.15 Application of corrected unsteady methods to Weber et al (1995)
measurements on a pitching NACA 4418 airfoil. ) 54 . 19 sin( 20 9
0 0
t + = ,
k=0.113, Re=5.510
5
. (a) Lift coefficient and correlation measurements of
Weber et al (1995). (b) Behavior of first derivative and unsteady
correction factor R (from eq. (A.19)). (c) Corrected lift coefficients
compared to Weber et al (1995) original results.

Fig.3.16 Application of corrected unsteady methods to Weber et al (1995)
measurements on a pitching NACA 4418 airfoil.
) 23 . 53 sin( 26 10
0 0
t + = , k=0.251, Re=6.710
5
. (a) Lift coefficient and
correlation measurements of Weber et al (1995). (b) Behavior of first
derivative and unsteady correction factor R (from eq. (A.19)). (c)
Corrected lift coefficients compared to Weber et al (1995) original results.

Figure 4.1 Lift force measurement on a oscillating cylinder. (a) Schematic of a square
path enclosing cylinder (Fig. 1 in Obasaju & Bearman (1988)). (b)
xiii
Comparison of lift force from Kutta-Jukowski theorem and that from
pressure average method at KC=6.75 (Fig. 20(a) in Obasaju & Bearman
(1988)).

Figure 4.2 Instantaneous lift force at KC=10 obtained from three methods by Lin &
Rockwell (1996). Fair agreement in amplitude and phase results is found
for the different methods. Measured C
L
means lift obtained with pressure
average method (Fig. 2 in Lin & Rockwell (1996)).

Figure 4.3 Typical results from numerical simulation for flow over a stationary
circular cylinder at Re=100. (a) Velocity vector field and streamlines at
one time instant presented the vortex generation and shedding process. (b)
The time trace of lift force was from the stress integration method.

Figure 4.4 Flow visualization of the near wake of circular cylinder during one lift
cycle at Re=100. (a)-(e) are instantaneous wake vortex structures at five
time instants shown in (f), the periodic time trace of the lift force.

Figure 4.5 (a) Schematic of cylinder near wake and definition of variables needed for
wake circulation determination. In x-direction, wake is divided into N
x

sections. In Y direction, N
y
sections. (b) Wake circulation distribution
in x-direction with y
1
/d=-0.75, y
2
/d=0.75. Wake is divided into a series of
consecutive smaller paths ABCD, with the path extent in x-direction
|AB|=(x/d)=0.5.

Figure 4.6 Time traces of vortex lift force obtained from eq. (4.10) within different
wake domains. The lower, upper and left bounds of the domain are fixed
at y
1
/d= -0.75, y
2
/d= 0.75 and x
1
/d= 0.5. The right bound is movable and
set at four locations x
2
/d= 5, 10, 15 and 20.

Figure 4.7 Schematic of path selection for bound circulation determination around
circular cylinder. Re=100, tU/d=76.875. Five typical paths are
highlighted to study the effect of near-wake vortex structures on bound
circulation values. Define a as the length of one side of the square paths,
for path A, a/d=1.0; path B, a/d=1.6; path C, a/d=2.0; path D, a/d=3.0;
path E, a/d=8.

Figure 4.8 Validation of bound circulation with size of closed paths at one time
instant. A small stable range exists in the plot where is independent of
a/d, These results can serve as a useful guide for closed sound path design
in UCMT experiments. Labels A, B, C, D, E correspond to paths
highlighted in Fig 4.7.

Figure 4.9 Comparison between true lift force
l
C , quasi-steady lift
) 1 (
l
C and vortex lift
xiv

) (v
l
C . Vortex lift is obtained from equation (4.10) within the selected wake
domain. Quasi-steady lift is determined from K-J approximation. is the
phase angle between true lift and
) ( ) 1 ( v
l l
C C + .

Figure 4.10 Phase comparison of true lift
l
C , bound circulation and its first
derivative
dt
d , which will be used for unsteady effect correction.

Figure 4.11 Rms error of lift
) 4 (
l
C approximation with variation in unsteady lift
correction coefficient R. 4 . 0 R for circular cylinder flow yields 0
) 4 (

rms
e .

Figure 4.12 Comparison of lifts between true lift
l
C , vortex lift
) (v
l
C , unsteady
corrected lift
) 3 (
l
C (K-J theorem + corrected unsteady) and total lift
) 4 (
l
C (
K-J theorem + corrected unsteady + vortex force).

Figure 5.1 Schematic of local sectional lift distribution on a finite wing.

Figure 5.2 (a) Comparison of lifting line predictions with spanwise lift distribution
obtained from integrated pressure distribution for Re=210
5
and AR=4
(Fig.9 in Bastedo et al (1985). (b) LDV measurement of bound
circulation along airfoil span at Re=2.5310
5
and AR=9.6 (Fig. 12a in
Bhagwat et al (2000)).

Figure 5.3 Schematic of the experimental arrangement in tunnel 1. Spanwise lift
distributions along plates with chord length variation are studied. (a) Setup
for UCMT measurement. (b) Setup for smoke visualization.

Figure 5.4 Diagram of the closed triangular acoustic path and its relative location to
our allowed region from the path sensitivity study.

Figure 5.5 Schematic of four plates with varying chord length used for study of
sectional lift distribution on flat plates.

Figure 5.6 Schematic of experimental setup for low aspect ratio wing study.
Setup for UCMT measurement of spanwise lift distribution along wings in
tunnel 1. (b) Setup for force balance measurements in tunnel 2.

Figure 5.7 Typical calibration curves for lift measurement before and after test.

Figure 5.8 A typical output voltage signal in HPvee during calibration process.

xv
Figure 5.9 Validation of UCMT measurement for 2D flat plate flow at U=5m/s. (a)
Comparison with force balance data published by Belvin (1984). (b)
Smoke visualization of near wake structure at =30
0
. (c) Near wake
vortex structure at =40
0
, vortex shedding process can be clearly
observed.

Figure 5.10 Mean lift distributions along a plate with sinusoidal curved leading or
trailing edge in streamlined regime (=6
0
) at U=5m/s. (a) Trailing edge
curved; (b) Leading edge curved; (c) Both leading and trailing edge
curved; (d) Comparison of different cases.

Figure 5.11 Smoke visualization of flow at three sections along plate span at =6
0
.
Separation bubble size changes at various spanwise locations. (a) z/z
0
=
1.0. (b) z/z
0
= 0.5. (c) z/z
0
= 0.0

Figure 5.12 Mean lift distributions along plate with sinusoidal curved leading or
trailing edge in a transitional flow regime (=15
0
) at U=5m/s. (a) Trailing
edge curved; (b) Leading edge curved; (c) Both leading and trailing edge
curved; (d) Comparison of different cases.

Figure 5.13 Comparison of UCMT lift measurement (in tunnel 1) and force balance
measurement (in tunnel 2) at U=5m/s for a plate with both sinusoidal
curved leading and trailing edge.

Figure 5.14 Typical circulation distributions along span of a low aspect ratio wing
with AR=2 and Re=3.6510
4
, measured with UCMT in streamlined flow
( = 6
0
) and stalled flow ( = 15
0
).

Figure 5.15 Comparison of lift distribution from UCMT along span of low aspect ratio
wing (AR=2) with Prandtls lifting line results for a wing (AR=6)
(Schlichting, 1979)

Figure 5.16 Effect of aspect ratio on lift distributions measured with UCMT on low
aspect ratio wings at = 6
0
.

Figure 5.17 Comparison of lift coefficient vs aspect ratio in several investigations. (a)
Current UCMT and force balance measurements. (b) Comparison of our
force balance measurements, Pelletier & Mueler (2000)s force balance
data and finite wing theory.



xvi
List of Tables


Table 1.1 Research matrix giving details of proposed work

Table 5.1 Plate parameters

Table 5.2 Geometry of the four low aspect ratio wings

















xvii
Nomenclature
A quasi-steady lift
AOA angle of attack
AR aspect ratio of finite wings
a sound speed
B unsteady lift term
) 2 (
B unsteady lift term for approximation 2
b span of finite wing
c plate chord length
) (z c local sectional chord length in experiments
l
C local sectional lift coefficient for 2D flow or true lift coefficient in numerical
work
L
C total lift coefficient for body in 3D flow
) 1 (
l
C lift coefficient for quasi-steady approximation (1)
) 2 (
l
C lift coefficient for unsteady approximation (2)
) 3 (
l
C lift coefficient for corrected unsteady approximation (3)
) 4 (
l
C lift coefficient including wake vortex lift cylinder flow
) (v
l
C vortex lift coefficient Lighthills concept
) (v
lw
C wake vortex lift coefficient Lighthills concept
l
C mean lift coefficient
N
C mean normal force coefficient on a plate
) (k C lift deficiency factor
d diameter of circular cylinder
) (t E
n
n
th
coefficient in trigonometry expansion of Glauert approach
) (t e time dependent relative error
) 1 (
rms
e root mean square error for
) 1 (
l
C
) 2 (
rms
e root mean square error for
) 2 (
l
C
) 3 (
rms
e root mean square error for
) 3 (
l
C
) 4 (
rms
e root mean square error for
) 4 (
l
C
) (t F instantaneous force
0
F force amplitude
v
F vortex force based on Lighthills concept
g gravitational acceleration
H height of wind tunnel test section
) (t h instantaneous location of oscillating plate
xviii
0
h equilibrium location of the plate
a
h plate oscillation amplitude
K spring stiffness
KC Keulegan-Carpenter number, d y KC
o
/ 2 =
k reduced frequency
L true lift force in numerical work
0
L sectional lift force on plate at constant chord length
) (v
L vortex lift based on Lighthills concept
) (z L local sectional lift force on plate
) 1 (
L lift for quasi-steady approximation (1)
) 2 (
L lift for unsteady approximation (2)
) 3 (
L lift for corrected unsteady approximation (3)
) 4 (
L lift including wake vortex lift cylinder flow
avg
L mean lift
rms
L root mean square amplitude of lift force
p
l

total length of a closed path
M mass per unit span
NM total element number in computational domain
NND total node number in computational domain
n
r
normal vector of surface
p local static pressure in fluids flow

p static pressure of free-stream


l
p static pressure on lower surface of body
u
p static pressure on upper surface of body
R unsteady correction coefficient
Re Reynolds number
r
r
position vector
S integral surface
cw
T transient time of a sound pulse along closed path clockwise
ccw
T

transient time of a sound pulse along closed path counterclockwise
t time
c
t plunging time of a plate
t time-step set in numerical simulation
U free-stream velocity
u velocity component in x-direction
V
r
local fluid velocity
0
V velocity of body reference frame
W width of wind tunnel test section
xix
W(x,t) local downwash
X,Y,Z coordinates in inertia reference frame
z y x , , coordinates in body reference frame
) (t y displacement of body
0
y body vibration amplitude
0
z distance in which local chord length varying

Greek Symbols

angle of attack
0
angle of attack - equilibrium state
a
angle of attack amplitude - pitching plate
viscous damping coefficient
) (t bound circulation
s
circulation measured on closed path enclosing body
a
circulation measured on closed path enclosing wake
k
circulation of the k
th
vortex in wake
(x,t) circulation distribution along plate
(x,t) local vortex element distribution along plate
solid blockage of testing plate
plate shape function
angular coordinate
fluid density
stress tensor
shear stress
k
2
phase angle between lift and vortex shedding event
tensor related to velocity and its derivatives Nocas work
velocity potential
B
velocity potential of body
w
velocity potential of wake
) 1 (
phase angle between
) 1 (
l
C and
l
C
) 2 (
phase angle between
) 2 (
l
C and
l
C
) 3 (
phase angle between
) 3 (
l
C and
l
C
stream function

r
rate of rotation of body reference frame

r
vorticity
angular frequency

1
Chapter 1
Introduction
1.1 Background
The determination of the lift force on aerodynamic bodies in a fluid flow is of
fundamental importance in many engineering applications. These include fluid-structure
interaction problems such as the flow-induced vibration of aircraft wings, tall buildings,
long bridges, and offshore engineering structures including flexible marine risers, cables
and ribbons (Fig.1.1). These flows are characterized by both unsteady effects
(oscillation) and three-dimensional spatial variations (along the structure) in the flow.
These effects also occur in other flows (pitching airfoils, low aspect ratio wings),
making lift measurement under such conditions a topic of general interest. In this
research, a sub-class of these flows that exhibit unsteadiness and spatial variation will
be studied. These include oscillating flat plate flows, cylinder wake flows and low
aspect ratio wings. Though many issues need to be studied for these types of flows, our
focus will be mainly on the topic of lift force determination through circulation
methods.
Determination of local, unsteady lift force in these flows is critical due to the
nonlinear coupling that exists between the structure motion and fluid forces. As shown
in Fig 1.2, the motion of the structure influences the flow dynamics in the wake. The
flow dynamics, in turn, determine the fluid loading (lift force) and hence the structural
motion. As a result, accurate lift determination is a key in modeling of flow-induced
vibration. Modeling of this type has been studied by many investigators. Only a brief
2



Fig. 1.1 Fluid-cable interaction by direct numerical simulation (Newman & Karniadakis, 1996)






y(t)
Structure
response
Wake dynamics
Fluid force F(t)
U



Fig. 1.2 Schematic of nonlinear coupling between fluid force, structural response and fluid loading
3
summary is given here. Detailed review of flow-induced vibration modeling can be
found in Parkinson (1974), Iwan (1975), Blevins (1984), Sarpkaya (1979). For
example, a spring-mass modeling system for the coupling between structure motion and
the exciting fluid force was studied by Bearman (1984). The model can be expressed as
) (t F Ky y y M = + + & & & . (1.1)
where y is the displacement of the body, M the mass per unit span, the viscous
damping coefficient associated with the springs and their mounting, K the stiffness of
the springs, F(t) the time-dependent fluid force. Physically, when a body is responding
to vortex shedding, the fluid force must lead the excitation by some phase angle . The
displacement and force F(t) can be presented as
) sin( ) ( ); sin( + = = t F t F t y y
o o
. (1.2)
where y
o
is the body vibration amplitude, F
o
the fluid force amplitude.

As a result,
accurate determination of lift amplitude F
o
and phase angle are critical for modeling
of flow-induced vibration.
Recently, Olinger (1998) developed a low-order dynamic model, based on circle
maps, to study the dynamics of oscillating flexible cables. A spatial-temporal map
lattice was used to predict the vortex shedding patterns and wake dynamics.
) 2 sin(
2
2
2
2
2
2
k
n
k
n
l
C
z
y
c
t
y

+ +

. (1.3)
Here
l
C is the fluid-induced lift force,
k
n
is the phase between vortex shedding event
and the lift force. As in the more traditional models described above, phase angle
determination is again a critical element.
4
Lift determination is also important for study of three-dimensional spatial flow
variations along the body span. Fluid-cable interaction is one of the cases that involve
the spanwise variation of vortex shedding patterns and fluid forces. Ramberg and
Griffin (1976) studied the effect of vortex coherence on flow-induced forces on
vibrating cables. Newman and Karniadakis (1996) utilized a parallel spectral element
Fourier method to numerically study the flow over a flexible oscillating cable at laminar
and transitional Reynolds numbers. They quantified and compared the coupled cable-
flow response for both forced and flow-induced cable vibrations. Three-dimensional
patterns were also observed in the wake of two-dimensional bluff bodies (Triantafyllou,
1990). One such mechanism includes the three-dimensional patterns resulted from a
two-dimensional non-uniformity, such as the spanwise variation in body sectional size
(Nuzzi et al, 1992). Spatial lift distributions in a similar flow, a flat plate with an
imposed spanwise chord non-uniformity will be described in later sections (Yuan and
Olinger, 1999). Details of the three-dimensional patterns generated from two-
dimensional bluff bodies were summarized in Williamson (1996).
Finally, lift force determination is directly related to the structural stress and
moment distributions at each spanwise sectional location. Fundamental understanding
of lift force on structures will be beneficial in the design and construction of structures
that are safe, functional, economical and able to resist impact from the environment
over a required period.
The fluid forces exerting on body immersed in a flow are derived from two
sources: the normal pressure distribution exerted over the surface of the body and the
5
tangential shear stress distribution over the surface. The resultant force on a body can be
decomposed into lift (transverse force) and drag (in-line force), relative to the flow
direction over the body. Lift is defined as the force component perpendicular to the flow
direction, and drag is the component parallel to the flow direction. In this work, we
focus on the lift force. While experimental techniques exist to measure the time-average
mean lift on structures, little effort has been placed on accurate, non-intrusive
techniques to measure the instantaneous lift at each local section in these types of flows.
This will be a focus in this thesis. The main goal is to address certain aspects of these
issues through extension of the ultrasonic circulation measurement technique (UCMT)
developed by Schmidt (1970, 1975) and Johari & Durgin (1998), to study unsteady and
spatially varying flows. Details of UCMT will be described in a later section.

1.2 Review
Traditional research efforts in fluid dynamics generally can include three
aspects; theoretical work based on mathematical description of physical phenomena
leading to governing equations of fluid motion, experimental study, and numerical
simulation of the theoretical governing equations. In the following literature review, we
will focus primarily on the theoretical and experimental investigations on lift
determination. However, we will also focus briefly on certain numerical techniques,
namely discrete vortex methods, when necessary to understand important theoretical
aspects of lift determination.

6
1.2.1 Theoretical work
It is well known that finding analytical solutions to equations of viscous fluid
motion, associated with flow-structure interaction problems of practical importance, is
often hindered by insurmountable mathematical difficulties. As a result, many past
investigations involving the prediction of aerodynamic forces were simplified as much
as possible to avoid the entanglement with the details of the viscous fluid motions. The
assumption of potential flow (inviscid, incompressible, irrotational) has been a common
simplification. Even so, theoretical work involving force prediction has been less
common compared to other aspects of research.
The circulation theory of aerodynamic lift, the concept underlying a scientific
breakthrough in understanding and calculation of lift, was first examined by Lanchester
(1926). He suggested that flow over a lifting surface involved a circulating motion
superimposed on the translation motion of the free stream. For example, a flow field
over a lifting airfoil can be considered as two parts, a uniform flow and a circulatory
flow. Lanchester theorized that the circulatory part of the flow could be analyzed by the
conceptual vortex filaments first suggested by Helmholtz (1858), which were imagined
as running along the span of the wing. Though Lanchesters work described the
circulation as the cause of lift, it was essentially qualitative, and no substantive
aerodynamic calculation of lift was provided.
Quantitative formula relating lift to circulation was developed by two
researchers, Kutta (1902) and Joukowski (1906), working independently without any
knowledge of Lanchesters work (Giacomelli, 1934). In their work, a model of the flow
7
over a lifting airfoil was conceived to consist of vortical motions; the bound circulation
vortex was embedded along chord line. A relation for calculating the lift per unit span
of an airfoil was mathematically expressed as
= U L . (1.4)
where is the bound circulation,

= l d V
r r
, expressed as line integral of the flow
velocity taken around any closed curve encompassing the airfoil. Equation (1.4) was a
revolutionary development in theoretical aerodynamics, named as Kutta-Jukowski (K-J)
theorem, a key component leading to development of the UCMT that will be discussed
in detail in a subsequent section.
Use of the K-J theorem has been justified for steady, inviscid incompressible
flows in the past. Because of the inviscid condition, the circulation around any closed
curve enclosed the airfoil will remain constant and equal to the bound circulation within
a curve that only encloses the airfoil for steady flow. In reality, however, the flow is
viscous, and dissipation effects always exist. The circulation calculated within a curve
far away from the model is thus zero, as is the circulation measured in a curve exactly
matching the surface of the body due to the no-slip condition. Hence, an issue in
applying this theorem in practice is selecting the proper closed curve that can provide
the bound circulation for lift determination. Also the question of availability of this
theorem under time dependent conditions is critical for unsteady flow study and is a
major topic of our research. For example, Glauert (1959) and Katz and Plotkin (1991)
found that an additional term to count for the unsteady effect of the circulation has to be
included in the total lift force determination.
8
With incompressible unsteady potential flow theory; Katz and Plotkin (1991)
have summarized earlier work by Theodorsen (1935) and von Karman & Sears (1938)
on a two-dimensional unsteady flow around an oscillating flat plate. We will also use
this flow as a representative unsteady flow in later sections. The unsteady potential
theory assumes the flow is inviscid, time-dependent, and flow can be modeled with
combinations of singularities (point vortex, source, doublet) as in steady potential flow
theory. With the concept of velocity potential and stream function , the complex
potential theory can provide a good mathematical tool for analytical study. In the stress
integration method, unsteady Bernoulli equation can act as the bridge to connect the
circulation to lift force.

t
gZ V
p p


+ + =

2
2
1
r

. (1.5)
where

p is the static pressure of the free-stream flow, p is the static pressure on the
surface of the airfoil, is the velocity potential, V
r
= is the velocity field. gZ is the
gravitational potential energy which can be ignored for air flow. With
), (
1
= =

f l d V
r r
), (
2
= =

f ds p L circulation can be related with lift L. By


using a time derivative transformation between body frame (x,y,z) and inertial frame
(X,Y,Z) (Fig 1.3).

body inertial
t
r V
t

+ + =


] [
0
r
r r
. (1.6)


9



Fig. 1.3 (a) Body (xyz) and Initial (XYZ) coordinate systems (Fig13.1 of Katz &Plotkin(1991)).
(b) 2D view with nomenclature (Fig. 13.14 of Katz & Plotkin (1991))
10
The unsteady Bernoulli equation can also be expressed as

t
r V
p p


+ + =

) ( ) (
2
1
0
2
r
r r

. (1.7)
where
0
V
r
,
r
is the velocity, rate of rotation of the airfoil, r
r
is the position vector in the
bodys frame of reference. The continuity equation can be expressed as
0
2
= . (1.8)
for an airfoil that is not rotating and only moving with velocity i t U V
r r
) (
0
= . With higher
order small quantities eliminated, (1.7) can be simplified as

t x
t U
p p

) (

. (1.9)
hence, the pressures on the lower and upper surface of the airfoil will be
|
.
|

\
|


+ =

t
t x
t x u t U p p
l
) , 0 , (
) , 0 , ( ) ( (1.10)
|
.
|

\
|

+
+ + =

t
t x
t x u t U p p
u
) , 0 , (
) , 0 , ( ) ( . (1.11)
and the pressure difference across airfoil can be expressed as
( ) ( )
(

+ + = = ) , 0 , ( ) , 0 , ( ) , 0 , ( ) , 0 , ( ) ( t x t x
t
t x u t x u t U p p p
u l
. (1.12)
According to the thin airfoil theorem,
2
) , (
) , 0 , (
t x
t x u

= + ,
2
) , (
) , 0 , (
t x
t x u

= ,
and ) , ( ) , ( ) , 0 , ( ) , 0 , (
0
t x ds t s t x t x
x
= = +

. Finally, the lift expression can be
obtained through the pressure integration around the body as
11

4 4 3 4 4 2 1
B
c c
dx t x
t
t t U pdx L

+ = =
0 0
) , ( ) ( ) ( . (1.13)
where part A is the quasi-steady part and B is the unsteady part. ) (t is the total
instantaneous bound circulation around the body, and ) ( ) , (
0
t ds t s
c
=

. The
measurement of ) , ( t x , the circulation distribution along the airfoil chordline, in this
equation is a challenge in experimental studies. This measurement will be discussed
further in future sections.
While the work of Katz & Plotkin (1991) will serve as the theoretical framework
for later work in chapter 2, we also review other previous developments in the theory of
lift determination that will also be applicable to our work. Wu (1981) developed a
general theory for aerodynamic forces and moment, through a rigorous analysis of the
viscous flow equations based on control volume concept, combined with vorticity
transport and decay. When his theory is applied to steady flow over a flat plate or
airfoil, a formula similar to Kutta-Joukowski theorem was obtained

)
(
a s
U L + = . (1.14)
where circulation in two distinct regions is considered.
s
is the total vorticity
(circulation) in the area just before the trailing edge, and
a
is the vorticity in the wake
between the trailing edge and somewhere in the wake where the fluid velocity has
recovered to the free stream velocity. Concerned about the viscous dissipation and
decay, this work showed that for bound circulation determination, the selection of
closed curves enclosing aerodynamic bodies is crucial. Later we will use a similar
12
technique of considering circulation in distinct flow regions when studying
instantaneous lift forces for a circular cylinder flow.
Lighthill (1986) analyzed the fundamental aspects of the wave loading acting on
offshore structures; The main concept leading to an important advance in understanding
of lift on bodies in an unsteady flow was developed to separate the total loading into a
potential-flow force and an vortex-flow force. The potential flow force is related to the
drag on structures and expressed as the well-known Morison equation, which is not a
concern in this research. The vortex flow force is derived from the momentum of the
vortex flow component, which can be expressed as

d r
r r
2
1
, where r
r
is the
position of a vortex element with volume d and vorticity
r
in fluid flow. Similar to
the Newtons second law, the time derivative of the moment equals to the force. A more
accurate vector expression for vortex force is given by

(

=

d r
dt
d
F
v

r r
2
1
. (1.15)
Rockwell (1996) applied Lighthills results and equation (1.15) to the
experimental study of flow around an oscillating cylinder, based on the flow field
information obtained from particle image velocimetry techniques. Govardhan &
Williamsom (2000) used it while investigating the vortex modes and frequency
response for a freely vibrating cylinder. In later section, we will study the application of
Lighthills results to estimate the instantaneous lift force on a circular cylinder at low
Reynolds number. Our interest will be in determining if information measurable using
the UCMT technique can be converted to an accurate lift estimation via equation (1.15).
13
Recently, more advanced techniques to obtain the flow velocity field
experimentally have been used, such as the Particle Image Velocimetry (PIV). While
discussion of experimental lift determination techniques will be reserved in section
1.2.2, use of PIV has motivated a theoretical advance in this area. A theoretical
expression for the evaluation of instantaneous forces on circular cylinder in an
incompressible flow was put forward by Noca (1997). Based on a momentum principle,
it can be expressed as


=
) ( ) ( ) (
) (
1
1
1
1
t V t S t S
b
dS u n r
dt
d
N
dS n d r
dt
d
N
F
r r r r r r
r
. (1.16)
where N is the dimension of the space under consideration (N=2 in a two-dimensional
space), is a tensor related to velocity, vorticity and viscous shear stress, V(t) is a
volume bounded by a non-material surface S(t) moving with velocity u
r
. The first term
in the right hand side of the equation is similar to the vortex force by Lighthill (1986).
Application of Eq. (1.16) to cylinder wake experiments will be discussed in section
1.2.2.
Further investigations to develop refined theories for aerodynamic force
prediction in unsteady flows have been undertaken in the past. Sarpkaya (1975) applied
a numerical discrete vortex method to develop a lift expression for two-dimensional
potential flow. The complex potential function and Joukowski transformation were used
for the mathematical manipulation, and the generalized Blasius theorem (which apply
the integration of derivative of the complex potential along a closed curve enclosing an
aerodynamic body to calculate the normal and tangential force in a potential flow) was
14
used to determine the drag and lift. The lift includes two parts, quasi-steady and
unsteady, it can be expressed as

t
r u u U L
k
k
m
k
k k k
m
k
i i


=

= = 1 1
) ( . (1.17)
where, U is the free stream velocity,
k
and
k
u are the circulation and velocity in the x-
direction of the k
th
vortex in the wake, while
i k
u and
i k
r are the velocity and location of
the corresponding image vortex. Given that this is a numerical technique, it is difficult
to apply in an experiment since the vorticity and velocity of each vortex element would
need to be measured. However, we will later return to this equation to interpret some of
our later results.

1.2.2 Experimental work
Several methods exist to obtain fluid forces acting on an aerodynamic body in
experiments. In the following sections, the force balance method, pressure integration
methods, particle image velocimetry (PIV) method and the ultrasonic circulation
measurement technique (UCMT) are described.
Force balance measurement systems have been commonly used in aerodynamic
research for many years. The term balance is generally used to describe instruments
designed to measure forces on aerodynamic bodies with a relative high degree of
accuracy. The first rough mechanical balance for measuring fluid forces was invented
by Mariotte (Giacomelli, 1934) to study the forces created by a moving fluid impacting
on a flat surface. Based on his investigation, he proved that the force of impact of the
15
fluid on the body varied with the square of the flow velocity. This is an early
breakthrough in the history of aerodynamics. Later, it was derived theoretically based
on the mathematical laws advanced by Newton in his Principia (1867).
Since the invention of wind tunnel, aerodynamic research can be conducted in a
controlled air stream under laboratory conditions. Correspondingly, various forces
balances were developed for different force measurement requirements in various
investigations. Six-component balances (three forces and three moments) are the most
sophisticated set-up for aerodynamic research, but one, two or three-component
balances are also used. However, a deficiency of force balances is that they are
generally limited to measuring of total lift force on a structure, equivalent to integrating
the local sectional forces along the span of an aerodynamic body. Force balance cannot
isolate sectional lift forces that are important for determining the motion of flexible
slender structures. As a result, an alternative method is required to measure the local
sectional lift force. This is a goal of the present work, which will be addressed in a later
section.
Force balances generally operate on the principle of amplifying the effects that
small structural deflections have on the capacitance, inductance, or resistance of
measuring devices, such as strain gauges, load cells, or electromagnetic balances. The
advantage of these types of balance is that the data can be acquired and processed more
easily and automatically with the aid of modern computers.
16
Pressure integration methods involve obtaining forces directly from body
surface pressure distribution, or from the pressure distribution along wind tunnel walls
based on a control volume theorem.
The surface pressure distribution integration method is more suitable for two-
dimensional flow measurements. Beyond the force acquisition capability, pressure
measurement can also provide fundamental information related to the fluid motion
around the body such as the velocity variation, separation effects and shear layer or
vortex development. Experimentally measurement of the pressure distribution is
generally carried out by covering the body with an array of pressure sensors. In some
cases, however, the geometry of the model may be too small to permit the use of a
sufficient number of sensors.
The pressure averaging method is another type of measurement technique
based on pressure integration. In this technique, two sets of pressure taps are located
around each half circumference of the cylinder. Each set of the taps is connected to a
common reservoir (pressure averager), such that the average pressure over half the
cylinder is obtained with insignificant distortion and phase shift. Then, by subtracting
the integrated average pressure obtained on opposing sides of the cylinder, the net
instantaneous transverse force is obtained (Surry & Stathopoulos, 1977; Bearman &
Luo, 1988; Waker, 1990). This technique is more suitable for force measurement on
two-dimensional rigid body that spans the tunnel test section. For sectional force
measurement on flexible slender body, this method may not work.
17
The force on a model in a closed tunnel can also be determined from the
reaction on the boundaries of the working section by measuring the distribution of the
static pressure along the walls. The theoretical basis for this wall pressure method is the
integral momentum equation.
The integral momentum equation states that the time rate of change of
momentum due to unsteady fluctuations of flow properties inside a control volume plus
the net flow of momentum out of the control volume across control surface is equal to
the total force acting on the fluid enclosed in the control volume. When the momentum
theorem is applied to a control volume enclosing an aerodynamic body in a wind tunnel,
the walls will be the control surfaces and the pressure variation on these surfaces can be
used to determine lift, through the integration of the wall pressure difference (Doenhoff
& Abbott, 1947). The only parameter one needs to measure is the pressure distribution
along wall from infinite upstream to infinite downstream theoretically, based on the
location of the test model. Practically, the measured range is only several times of the
model scale. The error caused by this finite measured range can be corrected by a
momentum scheme coefficient. For the measurement of pressure distribution along the
walls, a multi-tube manometer or a series of pressure transducers can be used for steady
flows. For unsteady flow, fast response pressure transducers must be used. The wall
pressure measurement method can also be explained with Newtons third law; the lift
acting on the body is of the same magnitude of that acting on the walls, but with
opposite direction. The fluid is only acting as a medium to transfer the lift force.
18
This method has been broadly applied for about seven decades. Recently, Yuan
(1994) used it to validate a novel velocity integration method for the correction of wind
tunnel wall interference for test models with larger blockages. The corrected force and
pressure distribution on the cylinder was very good even for a 17% blockage. Linn
(1999) used a similar method to measure lift on a oscillating pitching NACA 0012
airfoil by integrating the pressure distribution along the tunnel walls at different angle
of attack. The goal was to investigate the effects of reduced frequency and amplitude on
the final lift force. As for the force balance measuring system, this pressure integration
method is also more suitable for measure the averaged lift for two-dimensional flow
around a rigid body. For local sectional lift measurement, alternative methods are still
required.
As described above, the traditional force measurement methods are not effective
for local sectional lift measurements. Hence, one has to resort to some other methods
that have the capability for sectional lift measurement under non-intrusive (natural)
conditions. Recently developed techniques such as the particle image velocimetry
(PIV) method and ultrasonic circulation measurement method have addressed this issue.
Particle image velocimetry (PIV) is a technique for simultaneously measuring
three components of fluid flow velocity at points on a two-dimensional plane. Generally
PIV includes several components. First, the fluid flow being measured is seeded with
small tracer particles that can follow the fluid velocity fluctuations. Second, the flow is
illuminated by a sheet of laser light at least twice. Third, the particle positions during
each illumination are recorded by camera. The flow speed within the light sheet can be
19
calculated approximately by measuring the distance particles moved between each
illumination. In identify corresponding particles between two images, more advanced
techniques are needed. One of them is the autocorrelation method, in which the
resulting peak represents the most likely particle distance.
Given the measured velocity field, forces on an aerodynamic body in the flow
can be obtained by using various theoretical equations described in previous section.
Lin & Rockwell (1996, 1997, 1999) used laser scanning and PIV techniques to
investigate the relationship between the vortex formation and fluid loading for a two-
dimensional streamwise oscillating rigid cylinder in steady current. Lighthills equation
(1.15), Nocas equation (1.16) and Sarpkayas equation (1.17) were used respectively to
calculate the transverse forces, which were obtained from the integration of the velocity
or vorticity in a control volume enclosing the body (stationary or oscillating). The
results were compared with forces obtained from other methods such as pressure
average technique, Kutta-Joukowski theorem and Blasius formula. The amplitude and
phase shift were studied and the instantaneous vortex structures were related to the
instantaneous force at various times. However, the influence of the size and shape of the
PIV control volume used for integration was not studied.
While the PIV technique provide instantaneous velocity field information which
can illuminate important physical process in a flow, its use in lift determination is less
well developed. Noca (1997) conducted an experiment in a water tunnel with a two-
dimensional rigid circular cylinder as the model. The PIV technique was applied for
measuring velocity field around the cylinder, and force balance was used to obtain the
20

spanwise-average lift force to validate equation (1.16). The force was converted from
the velocity field measured in a square domain enclosing the cylinder and translating
with it. The results show that 500% error exists in lift values between PIV and force
balance results (Fig 1.4). Another disadvantage of the PIV is the cost of instruments
and the complex alignment and operation of the system.
In recent years, a novel method, ultrasonic circulation measurement technique
(UCMT) has been developed that may prove useful in resolving some issues of lift
measurement in unsteady flows. In this section, we will first introduce the basic
principle of this technique, and then review some past investigations that used UCMT.


Fig. 1.4 Lift force comparison. DPIV data; Force balance data (Fig.2 of
Noca et al (1997))
21
The possibility of expanding the capabilities of UCMT for the unsteady and spatial
varying flows is then examined.
It is well known that lift can be related to circulation within an enclosing curve
of an aerodynamic body through the Kutta-Jukowski theorem under appropriate
conditions. This opens a way to experimentally obtain lift data through acquiring
circulation data. Instead of obtaining circulation through the velocity measurement
along a closed path and mathematical integration, the possibility of measuring
circulation directly and non-intrusively using ultrasound was introduced by Schmidt
(1970,1975). In the past thirty years, ultrasound circulation measurement technique
(UCMT) has been improved significantly and verified as an effective way to measure
flow circulation. A detailed review can be found in Johari & Durgin (1998). A simple
introduction to this technique is described below.
UCMT is based on the measurement of the time difference for two sound waves
traveling oppositely along a closed path that encloses a body immersed in moving fluid

Acoustic Reflector






Ultrasonic path

V





T
cw


T
ccw


Transducers




Fig.1.5 Schematic of a closed sound path for UCMT

22
as shown in Fig. 1.5. Two ultrasonic transducers and two acoustic reflectors (mirrors)
are generally employed to establish the closed path that encloses the bluff body. Each
transducer acts as both emitter and receiver. The sound pulse emitted by the first
transducer is redirected by the reflectors and received by the second transducer. The
propagation speed of the sound wave is the sum of the local sound speed and the local
fluid velocity. Hence, if the sound pulse travels in the same (opposite) direction as the
local fluid speed, the sound transit time will be decreased (increased). This process is
repeated by emitting a pulse in the opposite direction along the closed path. The transit
times of the pulse in the two directions around the closed path are
T
cw=
+V a
dl
; T
ccw
=

V a
dl
. (1.18)


T = T
cw
T
ccw
=
2
2
a

|
.
|

\
|

2 2
1 a V
Vdl
=
2
2
a

. (1.19)

T = T
cw
+ T
ccw
=
a
2

|
.
|

\
|

2 2
1 a V
dl
=
a
2
P . (1.20)

From Eq. (1.19) and (1.20), the local sectional circulation can be given as
=

l d V
r r
= 0.5
2
a T = 0.5(
T
l
p

2
)
2
T = 2 T
T
l
p

2
] [ . (1.21)
where l
p
is the total length of the sound path, a is the sound speed, V is the local fluid
velocity. In the derivation of eq (1.19), (1.20), the assumption 1 << a V is applied,
limiting the technique to low speed incompressible flow. A major advantage beyond
the non-intrusive nature of UCMT is its low cost. A typical UCMT system costs one to
23
two orders of magnitude less than a PIV system. While the PIV technique does yield
more detailed local flow information (velocity, vorticity fields); In this thesis, we will
attempt to show how global correlation measurements can also be very useful for lift
determination in unsteady and three-dimensional flows.
In the past, UCMT has been used for experimental studies of various flows
including free surface vortices in aqueous flows, delta wing circulation determination,
wing tip vortex effects and lift study of airfoils with/without oscillation. Smith et al
(1995) conducted experiments to study the swirl strength of free surface vortical flow.
A Rankine vortex model was used to validate the circulations measured from the
UCMT technique. Moreira & Johari (1997) studied the leading edge vortex effects on a
delta wing. Circulation behavior was examined as it related to primary vortex burst
location. Linear and nonlinear regimes were found at different range of angle of attacks
(AOA). Test results highlighted the need for more rigorous models of circulation
incorporating primary vortex behavior. Desabrais & Johari (1998) further studied wing
tip vortex behavior on a sharp-edged NACA 0012 wing. The effects of the size of
closed ultrasound path was examined, they found that the circulation distribution
achieved its maximum level when the sound path surrounds the whole vortex. The
comparison of the lift measured from a load cell and the lift from UCMT showed a
better agreement in streamlined flow with angle of attack of less 10

degree. Beyond that
AOA, fair agreement was observed. For the airfoil lift study, Purutyan (1990) used
UCMT with a simplified single leg path to study the behavior of a plunging airfoil, the
motion of the airfoil was always leading the lift by a phase angle. Weber (1995) used
24
the similar single leg path to investigate the stall behavior of a NACA 4418 airfoil
undergoing pitching oscillation. At lower oscillation frequency, hysteresis loops was
observed. Maximum lift increased linearly with the reduced frequency. Comparison of
these results with other investigators yielded some disagreements in lift measurements.
While past research did provide interesting results using UCMT technique, a
primary concern is the conversion of the measured circulation to the lift force. The
Kutta-Joukowski theorem is applicable to steady flow. However, it was also applied
using a quasi-steady assumption, ) (t U L = , in the work of Purutyan (1990), Weber
(1995). Here ) (t represents the measured unsteady circulation with an acoustic path
enclosing the body. UCMT has not been applied to measure lift distributions along
structure spans in three-dimensional flows or unsteady flows. Hence, extending this
technique for unsteady and spatial varying fluid flows, while addressing the issues
introduced above, is the major intent of this research.

1.3 Major goals
Our introductory sections have highlighted the need for accurate lift
determination on aerodynamic bodies in unsteady flows. In particular, the fundamental
importance of the lift force in understanding nonlinear coupling between structural
motions and flow dynamics in fluid-structure interaction problems has been addressed.
At the same time, deficiencies exist in current experimental techniques available to
measure lift forces. These deficiencies are most acute in the areas of instantaneous lift
measurement, and measurement of local, sectional lift forces along structural spans.
25
However, these two effects, unsteadiness and spatial variation, are also key
characteristics of lift forces in fluid-structure interaction problems.
An experimental method with the potential to address these deficiencies, the
ultrasonic circulation measurement technique (UCMT), has been developed in recent
years. This method has been applied primarily to measure mean (time-averaged) lift in
experiments on various flows. However, the extension of the UCMT methods to
measure time-varying lift forces in unsteady flows has remained largely unstudied. The
UCMT methods have also not been applied to measure local, sectional lift forces along
structural spans. The major goals of this thesis are to address these concerns by;
studying the extension of the UCMT methods to determine instantaneous lift values
in unsteady aerodynamic flows using a combined analytical-numerical approach.
applying UCMT methods to measure mean lift distributions along the spans of
structural elements in experiments on low speed, incompressible flows.
study the issue of dependence of UCMT lift measurements on the specific acoustic
path used to determine bound circulation values.
To accomplish these goals, an integrated analytical-numerical-experimental
approach outlined in the following paragraphs is used. Table 1.1 summarizes the
research matrix. In Chapter 2, an analytical study of the flow over an oscillating thin
airfoil (flat plate) is described. This flow serves as a representative unsteady flow. We
address the issue of the proper method needed to convert instantaneous circulation
measurements (of the type attainable from a UCMT experiment) to instantaneous lift
values. The work of Theodorsen (1935), Karman & Sears (1938), Katz & Plotkin
26
(1991), based on unsteady potential flow theory, is extended to develop this proper
conversion method, which we name the unsteady correction method. The unsteady
correction method incorporates an additional unsteady term in addition to the quasi-
steady Kutta-Jukowski term that has been utilized in past investigations using UCMT.
The accuracy of the unsteady correction method compared to the quasi-steady Kutta-
Jukowski approach is studied. RMS lift amplitudes and phase angles determined
resulting time traces of instantaneous lift force are compared. After establishing this
basic framework, we analytically study additional unsteady flows, including pitching
flat plates, plunging flat plates, and impulsively started flat plates, in a similar manner.
In Chapter 3 we use numerical simulations to verify that the developed unsteady
correction method can properly predict instantaneous lift values in viscous, unsteady
flows. While our eventual goal is application of the unsteady correction method to
circulation data from experiments, we choose to first perform this verification using
numerical techniques. Numerical techniques offer certain advantages including the
capability to obtain independent lift values through integration of surface stress
distributions. Also, in the numerical simulations we can easily mimic the acoustic path
that would be utilized in an experimental UCMT set-up. Study of the effect of variation
of the acoustic path on lift measurements is conducted. The dependence of lift
measurements on the prescribed acoustic path has not been systematically studied prior
to the present work.
The finite element simulation of viscous flow over an oscillating flat plate is
conducted at Reynolds numbers (based on plate chord-length) on the order of 100. The
27
study is conducted at low Reynolds numbers to avoid complexities arising from
turbulent, three-dimensional flows, such as the need for phase-averaging of lift
measurements. Also, since our correction methods are based on inviscid theory, we
believe that a highly viscous flow at low Reynolds number represents a worst-case
scenario, and that the developed the correction method would yield even better results
for higher Reynolds number flows. Also, many open questions regarding vortex
formation and flow-induced vibration still exist even at these low Reynolds numbers.
The velocity flow fields from the simulation are analyzed to determine the
instantaneous circulation by integrating local velocity vectors along the enclosing
acoustic path. This bound circulation is then converted to instantaneous lift using the
developed unsteady correction methods and the quasi-steady Kutta-Joukowski
approximation. These lift values are compared to instantaneous lift values (true lift)
determined from integration of the unsteady surface stress distribution.
Given the success of this validation using the numerical techniques, we next
study the application of the unsteady correction method of Chapter 2 to correct
instantaneous circulation data from a previous experimental UCMT study on rotational
oscillations of a thin airfoil (Weber 1995).
Having achieved a better understanding of lift determination in unsteady thin
airfoil flows, in Chapter 4 we study another important flow, low Reynolds number
vortex shedding from a stationary bluff body (circular cylinder), using similar
techniques. We study whether instantaneous circulation values (of the type attainable
from UCMT experiments) can be properly converted to instantaneous lift values in this
28
flow. Motivated by earlier work of Wu (1981), we study various force models including
a quasi-steady Kutta-Jukowski method, a vortex flow force model (Lighthill, 1986), and
a new unsteady correction model similar to that developed for the thin airfoil flow of
Chapters 2 and 3. Finite element simulations for flow over a circular cylinder at Re =
100 are once again performed in order to obtain true instantaneous lift values (from
surface stress integration), which are then compared to various combinations of the
force models. We study whether UCMT measurements can provide the necessary data
in experiments to properly apply the vortex flow force model.
In the previous chapters, we used theory and simulations to study the issue of
converting circulation to instantaneous lift force for unsteady flows. In chapter 5, we
turn our focus to experimental UCMT studies of lift variation in three-dimensional
flows.
Again, a flat plate airfoil is studied since three distinct regimes exist depending
on the angle of attack (AOA). At small AOA, streamlined flow exists, at intermediate
AOA, stalled flow, and at larger AOA, bluff body flow. To verify our UCMT methods,
various flat plates with constant chord length or imposed chordlength non-uniformity
are studied. The mean lift distributions along the plate span are measured with UCMT.
The near wake vortex structure is also visualized with smoke visualization techniques.
Once the experimental methods are verified, we study low Reynolds, low aspect ratio
(AR) finite wings related to Micro Aerial Vehicle (MAV) applications. MAVs are small
flying wing aircraft (with maximum dimensions less than 12) that operate at low
Reynolds number. Rectangular wings with Re < 42,000 and AR < 3 are studied and

29


Table 1.1 Research matrix giving details of proposed work

Goal Flow cases Parameters Description
Develop correction
method to properly
convert
instantaneous
circulation to lift.



U
Low angle of attack
, Low amplitude
h
a
/c,
Low reduced
frequency k


Analytical work.
(chapter 2)





U

Verify the above
methods applied to
both streamlined and
bluff body flows.
Investigate the effect
of the size and
location of the
integral path used to
determine the bound
circulation.



U



Low angle of attack
, Low amplitude
h
a
/c,
Low reduced
frequency k


Numerical work.
Use a finite element
code to obtain
pressure, integrated
lift and velocity field.
Develop logic to
determine circulation
within integral path
and convert to
instantaneous lift.
(chapter3 & 4)





U



3D flat plates flows,
sinusoidal
chord variation for
UCMT verification
studies.








Measure mean
circulation
distribution along
span of spatially
varying flows (3D
flows)







U




Low aspect ratio
b/c, low Re 10000,
flat plate flows with
applications to
micro-aerial
vehicles






Experimental work.

UCMT techniques in
wind tunnel for mean
lift measurement.
(chapter 5)



30
compared with Plandtls lifting line theory. Finally, the integrated lift force on low Re
low AR wings will be compared with results from force balance measurements and
previous investigations.
The conclusions will be presented in chapter 6.
31
Chapter 2
A Method for Converting Circulation to Lift
Based on Unsteady Potential Flow Theory

2.1 Introduction
In this chapter, we develop analytical methods to properly convert instantaneous
circulation measurements (of the type that can be obtained from UCMT experiments) to
determine the instantaneous forces (lift) on a structure oscillating in a two-dimensional
fluid flow. This section will focus on issues related to unsteady aerodynamics by
analyzing various motions of a streamlined aerodynamic body. Following Theodorsen
(1935), von Karman & Sears (1938) and Katz & Plotkin (1991), we primarily study an
oscillating thin airfoil (flat plate) as a representative flow, concentrate on extending the
UCMT techniques to measure lift in unsteady flows, and leave any discussions of spatial
variation effects to later chapters.
The connection between lift and fluid circulation within an area enclosing an
aerodynamic body is given by the Kutta-Joukowski (K-J) theorem for steady flows. An
experimental method based on the K-J theorem, the ultrasonic circulation measurement
technique (UCMT), first proposed by Schmidt (1975) to measure the circulation around
airfoil, has unique advantages. Details of UCMT method have been described in chapter
1. While most of the research based on this method has focused on measuring mean
(time-averaged) circulation and lift, less emphasis has been placed on developing
methods to measure instantaneous circulation and lift required for unsteady fluid flows.
Several studies have been attempted in this direction. Purutyan (1990) applied UCMT
method on a plunging NACA-0012 airfoil at Reynolds number of 5.810
5
, and Weber
32
(1995) investigated the dynamic stall characteristic of an NACA-4418 airfoil undergoing
rotational pitching oscillation with Reynolds number in range of 6.71056.710
5
.
However, these investigations made several simplifications to the basic UCMT set-up
described in chapter 1. First, they used a single acoustic path to enclose the airfoil (Fig
2.1a). This is not a proper assumption for steady or unsteady flows such as pitching and
plunging airfoil where a continuously generated vortex wake forms behind the airfoil. An
closed ultrasound path (Fig 2.1b) used recently is the proper choice ( Johari & Durgin,

Transducer 1 Acoustic Reflector

Vortex


Vortex









plPla




Transducer 2 Transducers
(a) (b)

Fig 2.1 Schematic of a straight path (a) and closed path (b) for UCMT method

1998; Desabrais & Johari, 1998). Also, the conversion from time-dependent circulation
into lift was carried out by using a quasi-steady K-J approximation
) ( ) ( t t U L = . (2.1)
where U(t) and ) (t are the instantaneous free-stream velocity and circulation
respectively. Here, U(t)=U
o
=constant. In a quasi- steady assumption, one assumes that
variations of circulation occur slowly enough so that the K-J theorem (valid for steady
33







Fig. 2.2 Unsteady lift values during pitching motion of thin airfoil measured with
UCMT, from Weber (1995)
34
flows) can be applied at each instant of time. Additional unsteady effects are ignored. A
typical result from Webers work is shown in Fig. 2.2. The deviation of Webers lift
results from other investigations suggests that equation (2.1) is not valid for unsteady
flows. One goal of this chapter is to determine how much error is introduced when a
quasi-steady K-J theorem is applied for unsteady flow.
In this work, we further develop the ultrasonic circulation measurement technique
to improve lift measurement techniques in two-dimensional unsteady flows. Our analysis
will use unsteady potential flow theory to develop a method to properly convert the
instantaneous bound circulation into lift based on the work of Katz and Plotkin (1991),
hence open a new way to experimentally study unsteady flows.
In next section, we will first study the relationship between the instantaneous
circulation and lift based on unsteady potential flow theory. We then apply the results to a
representative two-dimensional unsteady flow, a flat plate airfoil undergoing low
frequency oscillation, plunging, or pitching motions, with small-amplitude in the x-z
plane (Fig 2.3). The instantaneous angle of attack (t) and the vertical displacement h(t)
of the pivot axis located at x=a, are expressed as
) sin( ), sin(
0 0
t h h h t
a a
+ = + = . (2.2)
The subscript 0 denotes the initial value and the subscript a denotes the oscillation
amplitude. denotes angular frequency of the oscillating plate.



35







(x,t)
U

x
c o

h

U
a o
X
Z
c
Pivot axis
Fig 2.3 Nomenclature for the oscillating motion of the flat plate
Fig 2.4 The plate chordwise circulation distribution required for determination of (x,t)
36
2.2 Unsteady Lift
Theodorsen (1935) and Von Karman & Sears (1938) have studied the unsteady
fluid forces on thin airfoils based on two-dimensional unsteady potential flow theory.
Recently Katz & Plotkin (1991) have also summarized this work. In this section, we
briefly review their work (part of their work has been introduced in chapter 1).
The lift force on a plate airfoil is given by

4 4 3 4 4 2 1
3 2 1
B
c
A
c
dx t x
t
t U pdx L

+ = =
0 0
) , ( ) ( . (2.3)
As introduced in chapter 1, the first term A is due to the instantaneous bound circulation
used when a quasi-steady K-J theorem is applied, and the unsteady term B includes the
contribution from time derivative of ) , ( t x , which can be expressed as

=
x
ds t s t x
0
) , ( ) , ( . (2.4)
Here, ) , ( t x is the chordwise circulation distribution (Fig2.4).
Based on the classic approach of Glauert, ) , ( t x can be approximated by a
chordwise trigonometric expansion at any time instant as

=
+
+
=
1
)] sin( ) (
sin
cos 1
) ( [ 2 ) , (
n
n o
n t E t E U t

. (2.5)
Using the standard transformation
) cos 1 (
2
=
c
x . (2.6)
The parameters ) (t E
o
, ) (t E
n
are coefficients of a Fourier series,
37

= =
= =

... 3 , 2 , 1 ) cos(
) , ( 2
) (
0
) , ( 1
) (
0
0
n d n
U
t x W
t E
n d
U
t x W
t E
n
o

. (2.7)
where ) , ( t x W is the local downwash, which is related to the motion of flat plate and the
velocity potential. Applying a time-dependent boundary condition (no normal flow across
the surface of flat plate) and after some mathematical manipulation, it can be expressed
as

t z x
U
t z x
U
x x z
t x W
W
W W B B


=


) ( ) , (
. (2.8)
where the higher order small quantities are neglected.
B
,
W
, are the plate potential,
wake potential and plate shape function respectively (see Fig. 1.3(b)).
Substituting Eq.(2.4) - (2.7) in Eq.(2.3) and after some mathematic manipulation, yields

=
+ =
+ =
) (
2
1
) ( ) ( 3 [
4
)
2
(
2 1
2
1 2
UE
t
UE
t
UE
t
c
B
E
E cU A
B A L
o
o

. (2.9)
For the unsteady plate flow in Fig. 2.3, the chord line shape function can be expressed as
) )( ( ) ( ) ( a x t t h t = . (2.10)
where h(t), (t) and a have been defined in section 2.1. This formulation allows for study
of various plate motions including oscillating, plunging, pitching through the h(t) and
38
(t) terms. Substitute Eq. (2.10) into (2.8) and (2.9), after some mathematic
manipulation, Katz & Plotkin (1991) find

+ =
+ =
+ =
] )
2
1
( ) (
4
3
[
] )
4
3
( )[ (
2


& &
& &
&
&
&
c
c
a
h U c B
c
c
a
h U k cUC A
B A L
. (2.11)
Here
U
c
k
2

= is the reduced frequency, C(k) is called the lift deficiency factor which is
related to wake potential
W
, The lift deficiency factor was originally introduced in
studies of airfoil flutter mechanisms, and later approximated by using an asymptotic
method (Theodorsen, 1935; Von Karman & Sears, 1938). It can be expressed as

|
|
|
|
.
|

\
|
+
+
+
+
|
|
|
|
.
|

\
|
+

|
|
|
|
.
|

\
|
+
+
+
+
|
|
|
|
.
|

\
|
+

=
5 . 0 ;
)
32 . 0
1 (
1072 . 0
)
041 . 0
1 (
0068 . 0
)
32 . 0
1 (
335 . 0
)
041 . 0
1 (
165 . 0
1
5 . 0 ;
)
3 . 0
1 (
1005 . 0
)
0455 . 0
1 (
0075 . 0
)
3 . 0
1 (
335 . 0
)
0455 . 0
1 (
165 . 0
1
) (
2
2 2
2
2 2
2
2 2
2
2 2
k
k
k
k
k
k k
k
k
k
k
k
k k
k C (2.12)
A detailed derivation of C(k)an be seen in appendix B.
Katz & Plotkin then summarized results for a flat plate oscillating vertically with
a constant angle of attack, plate motion equation (2.2) will become as
) t sin( h h h ; t tan cons
a 0
+ = = . (2.13)
Here h
0
is the equilibrium location of the plate before the oscillating, and h
a
is the
oscillation amplitude. Substitution of Eq. (2.13) into Eq.(2.11), yields
39
)]
2
cos(
2
)[ ( t
c
Uk
c
k Uh
U k cUC A
a
= . (2.14)
)]
2
sin( )
2
(
4
3
[
2 2
t
c
Uk
h
c
Uk
c B
a
= . (2.15)

4 4 4 4 4 3 4 4 4 4 4 2 1 4 4 4 4 4 4 3 4 4 4 4 4 4 2 1
B
a
A
a
t
c
Uk
h
c
Uk
c t
c
Uk
c
k Uh
U k cUC t L )]
2
sin( )
2
(
4
3
[ )]
2
cos(
2
)[ ( ) (
2 2
+ = . (2.16)
Typical time traces of the various lift component L, A and B obtained from Eq.
(2.14)-(2.16) are presented in Fig.2.5. The plate angle of attack is
0
0
2 = , the non-
dimensional vibrating amplitude 08 . 0 =
c
h
a
, and the reduced frequency is 35 . 0 = k .
From the plot, we find that phase differences exist between L, A and B, with the phase of
A leading L, and the phase of B always lagging L. It clearly shows that the unsteady lift
term B is a significant portion of the total lift L for typical k
c
h
a
, ,
0
values. As a result,
the unsteady lift term must be carefully studied.

2.2.1 Limitation of UCMT for (x,t) determination
In Eq.(2.3), the unsteady term B includes the time derivative of the circulation
distribution ) , ( t x . When x=c

= =
c
ds t s t t c
0
) , ( ) ( ) , ( . (2.17)
which is the total bound circulation within a region enclosed by the acoustic path in a
UCMT experiment (Fig. 2.6).

40







Fig 2.5 A typical lift variation with time using Katz & Plotkins (1991) analysis of an
oscillating thin airfoil (U=5m/s, c=4, k=0.35, h
a
/c=0.05, =2
0
)
41
Due to the limitations in applying the ultrasonic measurement technique (UCMT),
the measurement of chordwise circulation distribution ) , ( t x and circulation ) , ( t x is a
difficult challenge in UCMT experiments. In an experiment, it would require openings
along the chord line to construct an ultrasound closed path to measure the circulation
distributions ) , ( t x and ) , ( t x (Fig. 2.6a). As a result, one could only measure the local
circulation distributions at limited chordwise locations. The measured circulation
distributions may be inaccurate because the existence of physical openings would affect
the flow pattern and fluid forces on the plate. The experiment would be very challenging
if the plate were oscillating. However, the instantaneous bound circulation ) (t around
the plate is measurable with UCMT technique (Fig 2.6b). This motivates us to study if
the bound circulation ) (t can be used to accurately approximate the instantaneous lift on
the plate.
U
(b)
Fixed
acoustic
path
h
a
plate
Reflector
T
cw
T
ccw

=
c
ds t s t
0
) , ( ) (
U
(a)
Variable
acoustic
path h
a
T
cw
T
ccw
Pulser/Receiver
Reflector

=
x
ds t s t x
0
) , ( ) , (
Fig 2.6 (a) Required set-up to measure (x,t) using UCMT. (b) (t) measurement using UCMT
42
It essence we will make an approximation which studies whether accurate lift
predictions can be made even if the detailed information from the chordwise circulation
distribution (e. g. ) , ( t x and ) , ( t x ) required in equation (2.3) is unavailable. This
detailed information is replaced in favor of the attainable total bound circulation ) (t .
This assumption is reflective of difference between the UCMT and PIV techniques. The
PIV, which provides local flow fields, could be used to measure ) , ( t x . However, UCMT
generally provides global information such as ) (t .

2.3 Approximations
In the previous section, we discussed the inappropriate utilization of K-J theory
in unsteady flows, and the difficulty in applying UCMT methods to experimentally
measure ) , ( t x , a crucial component of the unsteady part of lift B. We next investigate
these issues by developing approximations that will be applied to the oscillating plate
flow of Fig 2.3.
Approximation 1: In this approximation, the quasi-steady part A of Eq. (2.3) is
assumed to describe the lift on the oscillating plate through
)
2
( ) (
1 2 ) 1 (
E
E cU t U A L
o
+ = = = . (2.18)
Inserting Eq. (2.10) into (2.18), yields
] )
4
3
( )[ (
) 1 (
&
&
c
c
a
h U k cUC L + = . (2.19)
Substituting Eq. (2.13) into (2.19), yields
43
)]
2
cos(
2
)[ (
) 1 (
t
c
Uk
c
k Uh
U k cUC L
a
= . (2.20)
Approximation 2: From Eq.(2.17) and Fig 2.6b, we know that bound
circulation ) t , c ( ) t ( = can be measured using the UCMT technique. In approximation
2, we replaced ) , ( t x by ) ( ) , ( t t c = in the unsteady B term of Eq. (2.3). Now ) (t
t


is only a function of time, so the unsteady B term in Eq.(2.3) will be

=
c
t
t
c dx t c
t
B
0
) 2 (
) ( ) , ( . (2.21)
Total lift will be

43 42 1
3 2 1
) 2 (
) (
) ( ) (
) 2 (
B
A
t
t
c t U t L


+ = . (2.22)
We will seek to determine the error that is introduced by applying these approximations
to determine instantaneous lift values.
From Eq.(2.6) and (2.17), we have

o o o o
c
o
d
c
t
t
dx t x
t
t
t


sin
2
) , ( ) , ( ) (
0 0

.
(2.23)
Inserting ) , ( t from Eq.(2.5) into Eq.(2.23), and noticing that
0 ) sin( ) sin( ;
2
) ( sin
0 0
2
= =


d n d
yields
)] (
2
1
) ( [ ) (
1
UE
t
UE
t
c t
t
o

. (2.24)
Substituting this relation into (2.21), result in
44
)] (
2
1
) ( [ ) ( ) , (
1
2
0
) 2 (
UE
t
UE
t
c t
t
c dx t c
t
B
o
c

=

. (2.25)
Again inserting Eq. (2.10) yields
] ) 1 ( ) [(
2 ) 2 (
& &
& &
& c
c
a
h U c B + = . (2.26)

4 4 4 4 4 3 4 4 4 4 4 2 1
& &
& &
&
4 4 4 4 4 3 4 4 4 4 4 2 1
&
&
B A
c
c
a
h U c c
c
a
h U k cUC t L ] ) 1 ( ) [( ] )
4
3
( )[ ( ) (
2 ) 2 (
+ + + = . (2.27)
Inserting Eq. (2.13) yields
)]
2
sin( )
2
[(
) (
2 2 ) 2 (
t
c
Uk
h
c
Uk
c
dt
t d
c B
a
=

= . (2.28)

4 4 4 4 4 3 4 4 4 4 4 2 1 4 4 4 4 4 4 3 4 4 4 4 4 4 2 1
) 2 (
)]
2
sin( )
2
[( )]
2
cos(
2
)[ ( ) (
2 2 ) 2 (
B
a
A
a
t
c
Uk
h
c
Uk
c t
c
Uk
c
k Uh
U k cUC t L + = . (2.29)
Further discussion of Eq. (2.29) can be seen in Yuan & Olinger (2002). A
comparison of the resultant lifts of approximation 1 and approximation 2 to the true lift
are presented in Fig. 2.7 for the case of U=5m/s, = 2
0
, h
a
/c=0.05, k=0.35. It is
observed from the peak lift values that the quasi-steady lift, L
(1)
(Eq. (2.20)),
underestimate much of the true lift, L (Eq. (2.16)). However, the lift of approximation 2,
L
(2)
(Eq.(2.29)), slightly overestimates the true lift L. It is also observed that a larger phase
angle exists between the lift curves L and L
(1)
, and L is always leading the phase of L
(1)
.
However, L is always lagging the phase of L
(2)
and

the phase angle between them is much
small. Based on the above qualitatively comparison, we believe L
(2)
is a better
approximation to the true lift.
45
In order to quantitatively and systematically analyze these approximations, we
define
(1)

as the absolute value of phase angle between lift L
(1)
and L;
(2)
as absolute
value of phase angle between L
(2)
and L. In addition to the phase angle analysis, we also
need to study the amplitude errors between those lift curves. Hence, we define relative
errors as

) (
) ( ) (
) ( ;
) (
) ( ) (
) (
) 2 (
) 2 (
) 1 (
) 1 (
t L
t L t L
t e
t L
t L t L
t e

=

= . (2.30)
Root mean square error as

rms
rms rms
rms
rms
rms rms
rms
L
L L
e
L
L L
e

=

=
) 2 (
) 2 (
) 1 (
) 1 (
; . (2.31)
where the definition of the root mean square lift amplitude is

=
T
rms
dt t L t L
T
L
0
2
] ) ( ) ( [
1
. (2.32)
Mean lift is defined as

=
T
dt t L
T
t L
0
) (
1
) ( . (2.33)
Next, those observations in Fig.2.7 will be further quantified based on the definitions of
phase angle and rms errors.
46





Fig. 2.7 Typical time traces of lift curves for quasi-steady (K-J) lift L
(1)
, unsteady approximation
L
(2)
compared to true lift curve L for oscillating plate (U=5m/s, =2
0
, k=0.35, h
a
/c=0.05)
47
2.4 Results and discussion
In this section we study the errors that are introduced (in rms lift amplitude and
phase angle behavior) when the quasi-steady and unsteady approximations of the
previous section are applied to an oscillating flat plate with chord length of c=10.16cm
and U=5m/s to match typical experimental values attainable in typical wind tunnel tests
(see chapter 5). The effects of the oscillation amplitude, initial angle of attack and the
reduced frequency are studied.
The comparison of the relative errors in Fig. 2.8 shows the phase difference
between ) (
) 1 (
t e and ) (
) 2 (
t e is almost . Another interesting result is that at any time, the
absolute value of ) (
) 2 (
t e is always one third of ) (
) 1 (
t e , or
) (
3
1
) (
) 1 ( ) 2 (
t e t e = . (2.34)
This clearly shows that approximation 2 is a better approximation for the unsteady plate
flow.
Fig. 2.9 presents the phase angle variations with plate oscillating amplitude h
a
/c,
reduced frequency k and plate angle of attack . For all the cases, the phase angle of
approximation 2,
(2)
, is always less than phase angle of approximation 1,
(1)
. At a
typical case of k=0.3, h
a
/c=0.1,
(1)
is almost 40
0
, but
(2)
is just 8
0
, that means
approximation 2 can improve the phase angle by almost 80% ( % 80
) 1 (
) 2 ( ) 1 (


).
This shows that approximation 2 is more accurate than approximation 1 (quasi-steady K-J
approximation). The behavior of phase angle variation shows that they do not change
48
with the plate oscillating amplitude and angle of attack (Fig. 2.9a, Fig 2.9c), but increase
with the increase of the plate reduced frequency (Fig. 2.9b).
Fig. 2.10 presents the variations of rms errors of those approximations with plate
oscillating amplitude h
a
/c, reduced frequency k and plate angle of attack . Fig. 2.10a
and Fig 2.10b show that
) 1 (
rms
e and
) 2 (
rms
e increase with the increase of plate oscillating
amplitude h
a
/c and reduced frequency k, but decrease with the plate angle of attack (Fig.
2.10c). The reason can be found from Eq. (2.16), (2.20) and (2.29), the capability of the
unsteady term contributes much to the total lift with the increase of plate oscillating
amplitude and frequency, compared to the quasi-steady term. However, with the increase
of angle of attack, the contribution from the quasi-steady term increases, hence the effects
of the unsteady term relatively decreases, this cause the rms errors decrease with angle of
attack. From quantitative comparison,
) 1 (
rms
e is always larger than
) 2 (
rms
e for all the cases. At
a typical case of k=0.4, h
a
/c=0.1,
) 1 (
rms
e is almost 0.3, but
) 2 (
rms
e is just 0.191, that means
approximation 2 can improve the rms error by almost 35%. This again confirms that
approximation 2 is a more accurate than the quasi-steady K-J approximation 1 for
unsteady lift prediction. Hence, approximation 2 is more favorable and we will study
) 2 (
rms
e
variation in detail next.
Fig. 2.11 presents the 3D plots of
) 2 (
rms
e variation with plate angle of attack ,
vibration amplitude h
a
/c and reduced frequency k. The general trend is that
) 2 (
rms
e decrease
with plate angle of attack, this can be observed from change of
) 2 (
rms
e peak values from
) 2 (
rms
e =0.22 at =2
0
to
) 2 (
rms
e =0.11 at =8
0
. Another characteristic is
) 2 (
rms
e decreases with
49
decrease of plate oscillating amplitude h
a
/c and reduced frequency k. This 3D plots also
show that
) 2 (
rms
e may reach unacceptable high values (
) 2 (
rms
e >15%) for specified larger plate
oscillating amplitude and reduced frequency. This fact will motivate further development
of a corrected unsteady approximation later in section 2.5.
So far, we know that approximation 2 is favored and the behavior of its rms error
) 2 (
rms
e variation with plate oscillating amplitude, reduced frequency and plate angle of
attack. However, in real UCMT experiments, it should have some guidance for initial test
design so that the errors may be kept in a reasonable range at first. This needs more
quantitative data of
) 2 (
rms
e and we present them in Fig. 2.12, the 2D contour of
) 2 (
rms
e as a
function of k, h
a
/c and . If we define % 5
) 2 (

rms
e as acceptable domain of the
combination of k and h
a
/c at different AOA, these 2D contours can clearly show the
acceptable area and unacceptable area ( % 5
) 2 (

rms
e ). Interesting is that the unacceptable
domain dramatically decreases with the increase of AOA, this contributes to the decrease
of the contribution from unsteady term B and increase of contribution from quasi-steady
term A, with increase of AOA. This kind plot can guide investigators using UCMT for
quick estimation of errors to be introduced in experiments.


50




Fig 2.8 Time traces of the relative errors for approximation 1 and approximation 2 for the case of
U=5m/s, =2
0
, k=0.1, h
a
/c=0.1.
51




Fig 2.9 Variation of phase angle with key parameters. a) effect of plate oscillating amplitude

h
a
/c . (b) effect of reduced frequency k. (c) effect of angle of attack .
52






Fig 2.10 Variation of rms error with key parameters. (a) effect of plate oscillating amplitude h
a
/c .
(b) effect of reduced frequency k. (c) effect of angle of attack .
53




Fig. 2.11 3D plots of
) 2 (
rms
e variation with plate oscillation frequency k, amplitude h
a
/c at different
angle of attack .
54



Fig. 2.12 Contour of
) 2 (
rms
e variation with plate oscillating frequency k, amplitude h
a
/c at different angle
of attack .
55
2.5 Corrected unsteady approximation
The previous section has detailed the effects of including the unsteady term (B) in
Eq. (2.3) on the resultant lift about an unsteady oscillating flat plate. It, therefore, serves
as a guide for determining whether inclusion of the unsteady term B as opposed to a
quasi-steady approximation is appropriate.
Analysis in previous sections does verify that inclusion of the unsteady term is
appropriate and approximation 2 is more favorable. However, in this section, we will
extend our results to develop a corrected unsteady approximation for predicting lift
more accurately. Previously we have
)]
2
sin( )
2
(
4
3
[
2 2
t
c
Uk
h
c
Uk
c B
a
= . (2.15)
)]
2
sin( )
2
[(
) (
2 2 ) 2 (
t
c
Uk
h
c
Uk
c
dt
t d
c B
a
=

= . (2.28)

43 42 1
3 2 1
) 2 (
) (
) ( ) (
) 2 (
B
A
t
t
c t U t L


+ =
. (2.22)
If we define R (see appendix A) as unsteady correction factor as

) 2 (
B
B
R = . (2.35)
Then accurate lift can be expressed as

43 42 1
3 2 1
B
A
dt
t d
c R t U t L
) (
) ( ) (
) 3 (

+ = . (2.36)
From Eq. (2.15) and (2.28), yields R value for oscillating plate as

4
3
) 2 (
= =
B
B
R . (2.37)
56
Eq. (2.36) will be the key equation for properly converting the measurable bound
circulation (t) to instantaneous lift L(t) in circulation based methods such as UCMT,
PIV and LDV. It will serve as an important result in our further studies in unsteady flow.
Comparison of Eq. (2.36) and Eq. (1.17) shows that the B term in Eq. (2.36) is similar to
the final unsteady term in Sarpkayas lift equation. In chapter 3, we will utilize
numerical simulation of flow over an oscillating flat plate at low Reynolds number to
validate the effectiveness of the corrected unsteady approximation of Eq. (2.36) in
determining instantaneous lift values. However, prior to this, we would like to first show
that this framework can be applied to other unsteady flows in next section.

2.6 Correction in other unsteady flows
In this section, we will present our analytical results for plunging flat plate flow,
pitching flat plate flow and summarize results of Katz & Plotkin (1991) for impulsive
started plate flows. The goal is to expand our understanding of lift approximations to
other unsteady flows.

2.6.1 Plunging flat plate flow
This type of flow has been studied with UCMT experiments by Purutyan (1990),
we try to confirm if quasi-steady K-J approximation was appropriately used in the UCMT
lift determination.
The motion of the plate can be described with following equation
57

=
<

=
=
c
c
c
t t h t h
t t t
t
h h
h t h
const
, ) (
0 , ) ( ) (
.
1
1 0
0

. (2.38)
Here h
0
, h
1
are the plate initial and final (altitude) locations, t
c
is the plate plunging time
(Fig 2.13). A similar derivation (see Appendix A) as discussed in section 2.3, yields

=
= =
3 2 1
A
t U t L
B R
) ( ) (
0 ; 0

. (2.39)
The result shows that for plunging plate flow, no unsteady correction is necessary, and
the quasi-steady K-J approximation can predict the lift force accurately.



z
U

x
0
h
1
h
0
h(t)
Plunging plate
Fig. 2.13 Schematic of a plunging flat plate
58
2.6.2 Pitching flat plate flow
In this case, the flat plate undergoes a periodic pitching motion about a pivot axis
located at x=a, and the motion of plate can be expressed as
t tan cons h h ); t sin(
0 a o
= = + = . (2.40)
Here
0
is the plate's initial angle of attack before the oscillating, and
a
is the oscillation
amplitude (Fig. 2.14). Weber (1995) studied this flow with UCMT techniques. The lift
force was converted from bound circulation with a quasi-steady K-J approximation.
Using the same mathematical manipulation, yield

|
|
|
|
|
|
|
|
|
|
.
|
= =
=
=
+ + =
) , , , (
)]
2
sin( )
2
( ) 1 ( )
2
cos( )
2
( [
)]
2
sin( )
2
( )
2
1
( )
2
cos( )
2
(
4
3
[
)]
2
cos( )
2
)(
4
3
( )
2
sin( )[ (
) 2 (
2 2 ) 2 (
2 2
t
c
a
U k f
B
B
R
t
c
Uk
c
Uk
c
a
c t
c
Uk
c
Uk
U c B
t
c
Uk
c
Uk
c
a
c t
c
Uk
c
Uk
U c B
t
c
Uk
c
Uk
c
a
c t
c
Uk
U U k cUC A
a a
a a
a a o



. (2.41)
z
U
h
0
(t)=
0
+
a
sin(t)
Pivot axis
x
0 a

Fig. 2.14 Schematic of pitching plate flow
59
Here the unsteady correction coefficient R is a complex function of reduced
frequency, pivot location and freestream velocity. Hence, for this flow, an unsteady
correction is more difficult to apply. However, 0 R means that effect of the unsteady
term B exists and quasi-steady K-J approximation can not predict lift accurately.

2.6.3 Impulsively started plate flow
We have studied unsteady correction term for oscillating plunging and pitching
plates in our work, and found appropriate R values for each flow. Wagner (1925) and
Katz & Plotkin (1991) also studied an impulsively started plate flow using similar
techniques. Wagner (1935) places a single vortex flow at the plate quarter chord point to
model circulation on the plate. Here we summarize their results.
The motion for this flow can be expressed as (Fig 2.15)

)
`

> =
= =
0 ,
0 , 0
t U u
t u
. (2.42)

> =
= =
0 ;
0 , 0
t U u
t u
0
x
z

U


Fig. 2.15 Schematic of a impulsively started plate flow
60
Katz & Plotkin (1991) summarized the work of Wagner (1925), found that

+ = =
= =
43 42 1
3 2 1
) 2 (
) (
) ( ) ( ) (
1
) 2 (
) 2 (
B
A
dt
t d
c t U t L t L
B
B
R

. (2.43)
The results show that for impulsively started plate flows, approximation 2 can be
used for lift prediction.

2.7 Summary
The work in this chapter leads to an important corrected unsteady approximation
for instantaneous lift prediction in unsteady flows with low amplitude and frequency
motions, given by

43 42 1
3 2 1
B
A
dt
t d
c R t U t L
) (
) ( ) (
) 3 (

+ = . (2.36)
This equation is the basis for our research work in this thesis. The value of unsteady
correction factor R has been established for different unsteady flows.
The focus was on oscillating flat plate flow in this chapter. The results for this
flow show that quasi-steady K-J approximation can not predict unsteady lift accurately.
Our corrected unsteady approximation improves the lift prediction. We derived a concise
expression for unsteady correction coefficient as
4
3
= R . The behavior of the phase angle
and rms error can provide guidance for investigators using circulation method (UCMT) to
keep measurement errors in a reasonable range. We found that
61
The phase angle increased with plate reduced frequency, but was
independent of oscillation amplitude and plate angle of attack.
The rms error increased with increase of plate oscillation amplitude and
reduced frequency, but decreases with angle of attack.
We also studied other types of unsteady flow using the framework developed for
oscillating plate flow. The focus was on the determination of R for plunging plate flow,
pitching plate flow and sudden accelerating plate flow. We found R=0 for plunging plate
flow and R=1 for an impulsively started plate flow. However, for a pitching plate, R is a
complex function of plate oscillating frequency, pivot location and freestream speed. It is
more difficult to correct the unsteady lift for this type of flow.
In next chapter, we would like to apply numerical methods to further validate our
developed corrected unsteady method. The advantages of numerical methods used to
accomplish this will be described in chapter 3.



62
Chapter 3
Validation of Corrected Unsteady Approximation
Using Numerical Simulation


3.1 Introduction

In the previous chapter, we developed a method to properly convert the
instantaneous circulation measurements (attainable in UCMT experiments) into the
instantaneous lift force for several unsteady flows. We verified the Kutta-Joukowski
theorem is only suitable for steady flows, though some of researchers still use it to
investigate unsteady flows (Obasaju, Bearman and Graham, 1988; Weber, 1995; Unal,
Lin and Rockwell, 1997). From our work in last chapter, we have learned that unsteady
effects play an important role in lift force determination. Most importantly, we
developed a method for determining the unsteady part of the lift due to the measurable
bound circulation. In this chapter, we intend to further verify this corrected unsteady
approximation by studying numerical simulation of flow over an oscillating flat plate.
In these simulations, we will mimic the acoustic path used in UCMT experiments to
determine bound circulation (t) which will be converted to lift L(t) using the developed
methods of chapter 2.
We pursue numerical simulations for several reasons. While our eventual goal is
application of the corrected unsteady approximation in UCMT experiments, verification
of the results of chapter 2 through numerical simulation has certain advantages. First,
numerical simulation provides an independent method to determine instantaneous lift
force, namely integration of stress tensor along the surface of aerodynamic body. This
63
method does not depend on measurement of circulation. This independent lift
measurement is more difficult in experiments. Current deficiencies in force balance
techniques have already been discussed in chapter 1. Secondly, in a numerical
simulation the geometry of the mimicked acoustic sound path can be easily altered. This
is important because the effect of varying acoustic path on the resulting circulation
measurements has never been systematically studied in UCMT work. Because of the
viscous dissipation effect, the circulation obtained within an enclosed curve far away
from the body is zero, as it is for a curve along the exact surface of the lifting body due
to no-slip boundary condition. This suggests that (t) measurement may be acoustic
path sensitive. The acoustic path variation is difficult to apply in UCMT experiments.
As a result, a primary goal of this chapter is to study this aspect of unsteady UCMT
measurements. Finally, numerical simulations provide detailed velocity/vorticity field
that can be used to determine the instantaneous circulation (t) by integrating tangential
component of velocity vectors along the acoustic path. Hence, numerical simulation
becomes our choice for investigating these issues.
In this section, we used a numerical simulation based on the Galerkin finite
element method as the main tool for our study on an oscillating flat plate flow. The first
goal is to verify the corrected unsteady approximation of chapter 2. The second goal is
to study how the size varying and location of the closed acoustic path alters the bound
circulation determination. The main function of the numerical simulation is to provide
the unsteady flow field. Given this velocity field, an algorithm is developed to obtain
the instantaneous bound circulation along different paths. Finally, we may validate our
64
developed correction method through comparison with the lift force obtained through
pressure and stress integration.

3.2 Numerical formulation
The governing equations for an incompressible viscous flow that are solved in
the finite element simulations are the continuity equation and Navier-Stokes equation
u = 0. (3.1)
=
Dt
u
Re
D
. (3.2)
where I + = p is the stress tensor. Velocity vector u is non-dimensionalized using
freestream velocity U, lengths with the chord length of the flat plate c, and time with
c/U.
The numerical grids are generated through PATRAN software and shown in
Figure 3.1, the mesh near the plate is much finer so that the detailed near wake flow
information can be captured. The boundary conditions of the computational domain are
set as follows. At the inlet, upper and lower grid boundaries, velocity is set as the free-
stream velocity. At the surface of the plate, the no-slip condition is maintained. At the
outlet boundary, a constant pressure condition is utilized following Malamataris (1991)
and Olinger & Alexandrou (1995). The plate is aligned horizontally in line with the x-
axis, the angle of attack is controlled through changing the free-stream incidence
relative to the chord of the flat plate. Based on the principle of relative motion, the
oscillating plate motion ) sin( ) (
0
t h h t h
a
+ = can be applied numerically through an
oscillating free-stream flow at the inlet boundary with velocity
65



(a)


(b)

Fig 3.1 Numerical grids used in the finite element simulations for a domain 3<x/c<6, -2<y/c<2
(a) grid 1: NM=1043, NND=4330, (b) grid 2: NM=1970, NND=8124
66
.
2
; ) cos( ) sin( ; ) cos(
c
Uk
t h U v U u
a
= + = = (3.3)
here is the angle of attack, k the reduced frequency, the angular frequency, U the
free-stream velocity. u, v are the resultant velocity in x and y direction.
The non-linear Navier-Stokes equations are solved by using a classical Galerkin
Finite Element approach, which has the advantage of unstructured meshes, element-by-
element formulation and processing, simplicity and rigor of boundary conditions being
incorporated with sophisticated automatic mesh generation, adaptive meshing and re-
meshing, economical storage and improved solution techniques both in speed and
accuracy (Lohner et al, 1984). The two primary unknowns, pressure and velocity, are
solved using a segregated solution procedure, in which the conservation equations are
solved to provide an intermediate velocity field at each time station through an explicit
step after an initial guess for the pressure field

1
1
1
) (
+
+
+
+ =

n
p e
n
p
e
u u R
t
u
R . (3.4)
here n is the time counter, p the iteration counter. After applying the divergence
theorem and integrating over the whole domain V, equation (3.4) can be written in finite
element form as
dA n dV u u R dV
t
u
R
n
p i
A
i e
V
i
n
p
e
V
i
1
1
1
) ( ) ] [ (
+
+
+
+ =


. (3.5)
The mass conversation equation can be written as
0 ) (
1
1
=
+
+
n
p
u . (3.6)
By combing Eq. (3.6) with (3.4), we can get the following equation
67

1 1
1
) ( 0 ) (
+ +
+
+ = =
n
p e
n
p e
u u R u R &
. (3.7)
Substituting the stress tensor I + = p , and applying the divergence theorem,
equation (3.7) can be expressed as


+ =
+ +
+
A
n
e
V V
n
p e i
ndA u R dV u u R PdV
1 1
1
) ( & . (3.8)
The term can be evaluated separately as = by using finite element form


+ =
A
i
V V
i
ndA dV dV . (3.9)
For each time step, the iteration procedure is repeated within the same iteration
loop until convergence is achieved, and the velocity field and pressure distribution at
that time step is available for post processing. The traction vector is integrated over the
body surface to calculate the resultant lift force at each time instant. The major
parameters input to the FEM solver include Reynolds number, plate angle of attack,
oscillating amplitude and reduced frequency. Further detailed information about the
numerical scheme can be seen in Olinger et al (1995).

3.3 Circulation determination
Generally, circulation can be obtained through the line integration of velocity
along a closed path, as long as the velocity field data is known. With the velocity field
data at each time-step provided by the FEM solver introduced in section 3.2, we can
calculate the circulation value along a mimicked ultrasound path in the flow. A
description of the algorithm for circulation determination follows.

68



A(x
1
, y
1
) D(x
2
, y
1
)
B(x
1
, y
2
) C(x
2
, y
2
)
y
x o
P
k
(x
k
, y
k
) N
i
(x
i
, y
i
)
k
l
plate


Fig 3.2 Schematic of circulation determination
69
Suppose a rectangular integration path is constructed with four points A(x
1
,y
1
),
B(x
1
,y
2
), C(x
2
,y
2
), D(x
2
,y
1
), and the coordinates x
1
, y
1,
x
2
, y
2
are arbitrarily selected in
order to study the size and location of the ultrasound path effects on determination
(Fig.3.2). The circulation in the area enclosed by ABCD can be obtained through the
following expression with discrete method the bound circulation

=
= =
M
k
k
P
ABCD
ABCD
l V dl V
k
1
. (3.10)
where P
k
represents any points which is located on the ultrasound path,
k P V means the
velocity at point P
k
(x
k
,y
k
) (
2 1 2 1
; y y y x x x
k k
). The parameter
k
l is the
infinitesimal distance along the path at point P
k
, and M is the total number of the
discrete points on the closed path.
Due to arbitrarily selection of the ultrasound path ABCD, the coordinates of
point P
k
(x
k
,y
k
) may not match any one of the mesh nodes at which the velocity and
pressure data are determined. Hence, we use the flow information at the nearest node to
represent the flow at P
k.
The nearest node is determined by finding the minimum
distance between P
k
and node N
i
(x
i,
y
i
) through

2 2
) ( ) (
i k i k i k ki
y y x x N P l + = = . (3.11)
Here N
i
represents an arbitrary node position in the computational domain,
ki
l is the
distance between point P
k
and N
i
. When l
kj
reaches a minimum value, then that node in
the domain will be selected as the most representative of point P
k
on the ultrasound
path, The velocity at that node is then used to calculate the circulation around the
arbitrarily selected path ABCD. A subroutine was developed and inserted into the FEM
70
solver introduced in section 3.2 to provide the circulation value at any time step. These
methods will be used for further validation of the analytical results of chapter 2 for
unsteady lift force determination.

3.4 Code validation
The finite element simulation of viscous flow over an oscillating flat plate is
conducted at Reynolds numbers (based on plate chord-length) on the order of 100. The
study is conducted at low Reynolds numbers to avoid complexities arising from
turbulent, three-dimensional flows, such as the need for phase-average of lift
measurements. Also, since our correction methods are based on inviscid theory, we
believe that a highly viscous flow at low Reynolds number represents a worst-case
scenario, and that the developed correction method would yield even better results for
higher Reynolds number flows. The low Reynolds number also implies large boundary
layer thickness. This will become important in our later acoustic path sensitivity study
where we wish to resolve the effect of the boundary layer on determined circulation
values. Also, many open questions regarding vortex formation and flow-induced
vibration still exist even at these low Reynolds numbers.
Before any further study, the numerical codes must be first validated to ensure
that simulation results are grid and time step independent. This validation is conducted
for flow around a stationary flat plate at angle of attack =6
0
first, then for plate
oscillating with non-dimensional amplitude as h
a
/c=0.05 and reduced frequency
k=0.15, with intention to check if the numerical code can provide reasonable lift force
71






















(a)
tU/c
0 2 4 6 8 10 12 14
C
l
0.35
0.40
0.45
0.50
0.55
0.60
Stationary plate, =6
0
Re=100, t=1/40
(b)
tU/c
0 5 10 15 20 25 30 35 40
C
l
0.35
0.40
0.45
0.50
0.55
0.60
Oscillating plate, =6
0
Re=100, t=1/40
k=0.15, h
a
/c=0.05
Fig. 3.3 Grid effect on results with t=1/40, (Grid 1: NM=1043,NND=4330; Grid 2: NM=1970, NND=8124).
(a) Stationary plate. (b) Oscillating plate with k=0.15, h
a
/c=0.05
Grid 1: NM=1043, NND=4330
Grid 2: NM=1970, NND=8124
Grid 1: NM=1043, NND=4330
Grid 2: NM=1970, NND=8124
72
predictions that are independent of grid and time steps resolution. The plate thickness is
set at t/c=0.12. This was done to preliminarily incorporate some airfoil thickness effects
(that will occur in experiments) into the flow, while also eliminating some numerical
difficulties (mesh generation).

3.4.1 Grid effects
The numerical grid resolution may affect the final numerical results
dramatically. Theoretically, higher resolution grids yield more accurate results.
However, increase in grid resolution will increase the CPU run time for the codes.
Hence, a proper grid should be selected that fulfills both requirements of providing
accurate results with reasonable CPU running time.
Here we study two grids as shown in Fig 3.1. The number of elements for grid 1
is NM1=1043, total nodes number is NN1= 4330. Grid 2 essentially doubles the grid
resolution of grid 1, with NM2=1970, NN2=8124.
Fig 3.3 presents this comparison. The trend of the numerical data for a stationary
plate in Fig. 3.3 (a) is similar; the lift forces will gradually converge to a final constant
lift coefficient after the initial transient process. Grid 1 results in 4808 . 0
L
C ; grid 2
leads to 4642 . 0
L
C . That means the difference between them is about 3.6%. Fig.
3.3(b) shows the grid effect on lift for a sinusoidal oscillating plate with amplitude
h
a
/c=0.05 and reduced frequency k=0.15. After the initial transient period of time, the
lift curves gradually follow the sinusoidal nature of the plate motion. The lift coefficient
from grid 1 is always larger than that from grid 2. However, the difference between
73
























(a)
tU/c
0 2 4 6 8 10 12 14
C
l
0.35
0.40
0.45
0.50
0.55
0.60
t
1
=1/40
t
2
=1/100
stationary plate
Re=100, a=6
0
NM=1043, NND=4330
(b)
tU/c
0 5 10 15 20 25 30 35 40
C
l
0.35
0.40
0.45
0.50
0.55
0.60
t
1
=1/40
t
2
=1/100
Oscillating plate
Re=100, =6
0
k=0.15, h
a
/c=0.05
NM=1043, NND=4330
Fig 3.4 Time step effects on the results with Grid1:NM=1043, NND=4330, =6
0
(t
1
=1/40, t
2
=1/100). (a) For stationary plate. (b) For oscilating plate at k=0.15
and h
a
/c=0.05.
74
them is approximately 3.6%, same as the difference in Fig.3.3 (a) for the stationary
plate. With such small difference exist between these two grids and consideration of
shorter running time, we select grid 1 to be used in our numerical simulation.

3.4.2 Time step effects
We also studied the effect of varying the time step on the numerical results. We
seek to determine an optimum time step balancing the conflicting trends of more
accurate solution but longer codes running times with smaller time steps.
Fig 3.4 presents the time traces of the lift forces for t
1
=1/40 and t
2
=1/100.
The numerical solutions are largely independent of the variation in time step, hence
t
1
=1/40 will be used in the present work in order to yield shorter run times. The final
grid and grid are set as grid 1 (NM=1043, NND=4330, t
1
=1/40).
Beyond the validations and convergence studies described here, we also note
that the same numerical formulation and similar resolution grids have been used in
numerical studies of oscillating cylinder flows in Olinger et al (1993) and Olinger et al
(1995).

3.5 Bound circulation determination
3.5.1 Background
In this section, we address in detail the issue of whether UCMT circulation
measurements in viscous flows are dependent on the chosen acoustic path (for example,
paths C
1
and C
2
in Fig 3.5). This acoustic path sensitivity was introduced earlier, but has
75
never been addressed in a systematic fashion. In a study of forces, circulation and
vortex patterns around a circular cylinder in oscillating flows, Obasaju, Bearman and
Graham (1988) determined the bound circulation through velocity measurement along a
square closed path around a circular cylinder, with the leg length as 1.35d (d is cylinder
diameter), however no justification was given for this selection. In a study of evaluation
of time-dependent fluid dynamic forces on bluff bodies, Noca (1997) used a
momentum-based control volume formulation to process velocity fields measured
through the Particle Image Velocimetry technique. Different results were obtained with
three domains of different size, and the author indicated that the largest domain couldnt
yield right answer. In the investigation of bound and wake circulation on a helicopter
rotor, Bhagwat & Leishman (2000) measured the circulation with Laser Doppler
Velocimetry techniques. They pointed out that circulation measurements are very
sensitive to the shape and size of the integration path, but offered no satisfactory

U
C2
C1
A (x
1
,y
1
)
B (x
1
,y
2
) C (x
2
,y
2
)
D (x
2
,y
1
)
x
y
y
2
y
1
x
1
x
2


Fig 3.5 Schematic of sound path effects on bound circulation determination
76
guidelines for choosing the right integration paths. Hence, a systematic study of path
selection in force determination through circulation methods is needed.

3.5.2 Effect of closed integration path variation
As introduced in last section, the lack of previous guidelines for acoustic path
selection, to some extend, is the result of the lack of systematical study of the path size
effects, due to the limitations in experiments to adjust the path size. However, in
numerical simulations, the path size and location can be easily adjusted, as introduced
in section 3.3.
Arbitrarily, we can select any closed path around the body, composed by four
corners at A(x
1
,y
1
), B(x
1
,y
2
), C(x
2
,y
2
), D(x
2
,y
1
) in Fig. (3.5). We chose to fix three of the
legs, and then study the circulation variation when changing the location of the
remaining leg. Here we present the results in Fig 3.6 for flow around a stationary plate
at angle of attack of 6
0
. Fig.3.6 (a) presents the circulation variation with movement of
leg AB. Fig3.6 (b) (c) (d) shows similar results for variation of leg CD, AD and BC. In
Fig 3.6(a), but characteristics of all the plots, we see that a stable region exists where
the circulation result is independent of the precious location of the leg, and these stable
regions can provide a basis for path selection. The fast drop of bound circulation with
the increase if distance away from leading edge may be due to some numerical
boundary effects on bound circulation values. In Fig. 3.6(b), x
2
/c leg must remain in
near wake to avoid effect of shed vorticity in wake. In Fig. 3.6(c) and (d), the stable
77
regions have some relation with the thickness of boundary layers. We estimated the
thickness of boundary layer at the plate trailing edge based on equation

Re
2 . 5
=
c

. (3.13)
it yields 52 . 0 =
c

for Re=100. Considering the thickness of the plate 12 . 0 =


c
t
and
the origin of coordinate system is located in center of the plate, the thickness of
boundary layer above and under plate should be 58 . 0 =
c
y

and this has been marked


in Fig. 3.6c and Fig. 3.6d. The relative location between boundary layer and the stable
regions suggest that the path selection should be at the edge of the boundary layer, and
this is reasonable and consistent with inviscid potential flow theory used for developing
our corrected unsteady method in chapter 2. In potential flow, surface velocity is at the
body surface, but in real viscous flow, it is actually at edge of the boundary layers.
These findings also have implications for PIV and LDV studies. For example, due to
resolution and body interference problem, its hard to obtain the vorticity values within
boundary layers in PIV and LDV studies. For bound circulation determination, it should
be a good method to use

= dl V
r
on a path just outside the boundary layer to replace
the traditional method using

= dA
r
operated within boundary layers. Based on
the results from Fig. (3.6), we schematically present the reasonable region for acoustic
path selection shown in shadow region in Fig. (3.7).
While these results are obtained at a single angle of attack =6
0
, we believe they
will provide reasonable estimates for proper path selection for angle of attack in the
78






















x
1
/c
-3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0

0.08
0.10
0.12
0.14
0.16
0.18
x
2
/c
0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0

-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
y
1
/c
-1.50 -1.25 -1.00 -0.75 -0.50 -0.25 0.00

0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
y
2
/c
0.0 0.3 0.6 0.9 1.2 1.5 1.8

-1.2
-0.8
-0.4
0.0
0.4
0.8
1.2
1.6
y
1
/c = -0.7007
x
2
/c = 1.2079
y
2
/c = 0.7007
x
1
/c = -0.5540
y
1
/c = -0.1896
y
2
/c = 0.3539
x
1
/c = -0.5540
x
2
/c = 1.2079
y
2
/c = 0.7007
x
1
/c = -0.554
x
2
/c = 1.755
y
1
/c = -0.189
stable
area
stable
area
stable area
stable area
L
e
a
d
i
n
g

e
d
g
e
T
r
a
i
l
i
n
g

e
d
g
e
Lower surface
Upper surface
y
x
o
x
1
/c
x
2
/c
U

y
1
/c
y
2
/c
Fig 3.6 Circulation variation with the movement of each leg of 'mimic' acoustic path
(reference to Fig 3.5, =6
0
)(a) Before leading edge; (b) After trailing edge;
(c) Away from lower surface; (d) Away from upper surface
Boundary
layer thickness
Boundary
layer thickness
(a) (b)
(c) (d)
79

y
x
c y /
1
-1.37
-0.4
c y /
2
0.7
1.8
c x /
1
c x /
2
-1.43 -0.7 0.5 1.2
U

Allowed region Dis-allowed region

; 7 . 0
1

c
x
2 . 1 5 . 0
2

c
x
; ; 4 . 0
1

c
y

c
y
2
7 . 0

Optimum path:
7007 . 0 / ; 5689 . 0 /
; 8893 . 0 / ; 7208 . 0 /
2 1
2 1
= =
= =
c y c y
c x c x



Fig 3.7 Schematic of required acoustic path size and locations to make circulation
measurements path independent. Oscillating plate flow at Re=100, =6
0
.
80
streamlined flow region below the stall angle. From this study, the proper closed path
selected is set as x
1
= -0.7208c, x
2
= 0.8893c, y
1
= -0.5689c and y
2
= 0.7007c for the
reminder of the numerical work on the flat plate. In the present work, we have not
repeated the path sensitivity study for oscillating plate flows. However, we point out
that in our numerical simulations we study low amplitude oscillations with h
a
/c < 0.15,
this is the range for which the theory of chapter 2 applied. If this oscillation amplitude is
added to the boundary layer thickness, the new boundary layer thickness will be
7 . 0 /
2
c y , it still remains within the stable region of Fig 3.6(d). As a result, the same
closed path will be used when plate oscillation effects are introduced into the numerical
simulation.
These path selection guidelines will aid in sound path design in future UCMT
experiments. We note the additional criterion that in UCMT experiments, one often
strives to reduce the total acoustic path length in order to maximize acoustic pulse
signal-ratio and sampling frequency. Beyond its usefulness in the present and future
UCMT studies, this systematic study of the sensitivity of results to variation of
integration path has broader implications for PIV and LDV studies. Investigations
where local velocity and vorticity data are integrated over an enclosed region can be
benefit from the present results.

3.6 Results for oscillating plate flow
In this section, we present the results from numerical simulations on a flat plate
at a single intermediate value of angle of attack =6
0
. While a widen variation in
81
could be conducted, we point out that the unsteady potential flow theory leading to the
unsteady correction methods of chapter 2 is only valid within the limited AOA region
below stall ( 10
0
). Also study of AOA variation in the analytical work of chapter 2
showed smooth, monotonic variation in results (rms error amplitude, for example) over
a wider range of 2
0
8
0
. As a result we would not expect to see dramatic
quantitative difference in results in this section if a larger range bracketing the
intermediate AOA was studied. The unsteady correction methods of chapter 2 are
incorporated as follows. The bound circulation (t), from Eq. (3.10), is determined at
each time step. The quasi-steady approximation (K-J theorem) of equation (2.18), along
with the corrected unsteady approximation (Eq. (2.36))
) (
) 1 (
t U L = . (2.18)

dt
t d
c R t U t L
) (
) ( ) (
) 3 (

+ = . (2.36)
are used to convert the numerical bound circulation (t) into appropriate lift values.
The reduced frequency of the oscillating plate was set at four values k=0.1, 0.2,
0.3, 0.4. Amplitude was set at three values h
a
/c=0.05,0.1, 0.15. The numerical grid and
time step are set at NM=1043, NND=4430, and t=1/40 as discussed in our validation
studies.
Fig 3.8 (a)-(e) presents the flow fields, presented with contour of u-component
of velocity field, around the oscillating plate at five time instants of one cycle with
=6
0
, h
a
/c=0.15, k=0.15 and Re=100. The corresponding lift values in Fig. 3.8(f)
shows the periodic variation due to the plate oscillating in a harmonic style, this can be
82




(a) tU/c= 10.325 (b) tU/c= 15.252
(c) tU/c= 20.7 (d) tU/c= 26
(e) tU/c= 31.3
(f) time trace of lift
Fig 3.8 Flow visualization represented by contour of u-component of velocity in the field around a
flat plate with =6
0
, h
a
/c=0.15, k=0.15 and Re=100. The flow fields a, b, c, d, e correspond to five
specific times lift in one cycle of the lift time trace.
83
judged from the almost same values of both lift and flow fields at (a) and (e). From
these flow fields, we see the cavities attached to the plate, not the separation bubbles for
steady flow, due to two reasons. On is larger thickness of the plate t/c=0.12 and another
reason is the unsteady oscillating of the plate. Due to the existence of a cavity behind
the plate instead of a open separated boundary layer, the calculation based on laminar
boundary (from Eq.(3.13)) overestimates the boundary layer thickness. Hence, the
selected optimum closed path is appropriate enough for circulation determination. In
addition, Fig. (3.8) shows that the numerical code can provide good information for our
studies.
Fig 3.9 presents a typical time trace of the numerical bound circulation (t) and
plate instantaneous location h(t). The plate non-dimensional amplitude is set as h
a
/c =
0.1, reduced frequency as k=0.4, =6
0
and Re=100. The bound circulation shows a
periodic variation after a short transient time. The phase of the plate motion is always
leading the phase of the bound circulation. This is the reasonable result because the
plate motion affects the wake dynamics leading to the fluid loading through circulation
as discussed in chapter 1.
Fig 3.10 is a typical time trace of the lift forces and the plate instantaneous
location from the numerical simulation. The quasi-steady lift
) 1 (
l
C , corrected unsteady
approximation
) 3 (
l
C , and the true lift coefficient
l
C obtained from plate surface stress
integration are all presented. Plate non-dimensional amplitude is set at h
a
/c = 0.1,
reduced frequency as k=0.4 and =6
0
. After the initial transient period, the lift forces
also are periodic in nature following the periodic motion of the plate. The phase of the
84
plate location is always leading the phase of the true lift
l
C by approximate a quarter of
a cycle (
2

). The phase difference,


) 1 (
, between the true integrated lift and the quasi-
steady lift is larger than the phase difference,
) 3 (
, between the true lift and corrected
unsteady lift approximation. The phase difference
) 1 (
,
) 3 (
are a measure of the
accuracy of the various unsteady approximation in correctly predicting the proper phase
of the lift forces. We will further quantify these phase difference over a wide range of
flow parameters, but first discuss the behavior of the rms lift amplitude. As in chapter 2,
we use root mean square error for the comparison of the lift force amplitude.
Fig. 3.11 presents the variation of phase difference (
) 1 (
,
) 3 (
) for the quasi-
steady and corrected unsteady approximations for a range of reduced frequency and
amplitude of plate motion. In Fig 3.11a, the reduced frequency is fixed at k=0.3 and
h
a
/c varied for h
a
/c<0.15. The phase difference for the quasi-steady approximation are
in the order of 40
0
, while the corrected unsteady approximation reduces these difference
to approximation 15
0
, an improvement of approximate 60%. The phase difference
(
) 1 (
,
) 3 (
) are found to be independent of the h
a
/c amplitude level. Fig 3.11b
presents the variation in
) 1 (
,
) 3 (
when reduced frequency k is varied at a fixed h
a

/c=0.1, the phase difference
) 3 (
is reduced from the quasi-steady value
) 1 (
over the
entire range of k. While the unsteady corrected method improves the phase prediction
significantly, value of
) 3 (
remains as high as
0 ) 3 (
15 at larger k values. At low k
values, phase differences
0 ) 3 (
5 < are obtainable. The phase difference increases with
increase in reduced frequency k.
85
Fig 3.12 presents the variation of root mean square (rms) errors of the lift force
(defined in chapter 2) for the quasi-steady and corrected unsteady methods over a range
of flow parameters. Here, the analytical lift L is replaced by the true lift from surface
stress integration. Fig 3.12a shows the rms error variation with plate oscillation
amplitude h
a
/c at a fixed reduced frequency of k=0.2. Fig 3.12b shows the rms error
variation with plate oscillation reduced frequency k at a fixed h
a
/c=0.1. Results show
that rms error increase with both the plate oscillating amplitude and reduced frequency.
The rms error for the the unsteady corrected method is much smaller than that from the
quasi-steady approximation for all h
a
/c and k values. Larger amplitude and reduced
frequency leads to the worst case, at k=0.4 and h
a
/c=0.1. The rms error from the quasi-
steady approximation is approximate 31%; from our corrected unsteady method, its
reduced to 12%. The corrected unsteady method can improve the rms error by almost
60%. In additional, we also run the case of higher Reynolds number (Re=200) flow
around the oscillating plate (at k=0.4 and h
a
/c=0.1) to study the Re effects on results,
within limitations of present numerical simulations (not including turbulent model and
3D effects). The results show that the rms error at Re=200 is approximately 8%, smaller
than that for Re=100. This shows that our unsteady correction method can predict rms
lift amplitude within 10% even at the highly viscous Re=200, and confirms our idea that
Re=100 is a worst case of scenario and the unsteady correction method should yield
better results at higher Reynolds number.
Before turning to other issues, we point out that the circulation distributions
(x,t) and (x,t) introduced in chapter 2 are attainable from our numerical simulations.
86
Fig. 3.13 shows typical distributions for k=0.2, h
a
/c=0.1, Re=100, =6
0
at one instant.
In Fig. 3.13a, a higher vorticity region is observed near leading edge. It must be
remembered that the (x,t) distribution in the viscous flow summarizes effects from both
upper and lower surface boundary layers. Near the trailing edge, the (x,t) distribution
becomes negative suggests that lower surface boundary layer with negative vorticity
dominates. In a potential flow, (x,t)=0 at trailing edge due to the Kutta condition. In
Fig. 3.13a, (x,t)=0 has shifted away from the trailing edge. These observations lead to
consideration of the unsteady Kutta condition in these flows. In an unsteady flow, the
location of stagnation point may oscillate and move away from the trailing edge. Our
assumption of low frequency and low amplitude should imply that this movement is
small in an inviscid flow. However, our simulations are highly viscous with thick
boundary layers making interpretation in terms of the Kutta condition difficult. This
needs further study. In Fig. 3.13b, the parameter (x,t) is presented. We perform another
path sensitivity study to determine how (x,t) results are affected. Path 1 is the optimum
path from Fig. 3.7. While in path 2, (x,t) values from leading to trailing edges (-
0.5<x/c<0.5) are integrated. Again, a sensitivity to path geometry is observed. This
could have implications for PIV studies which could possibly measure (x,t). The
enclosed integration region in these studies would have to be carefully selected.

3.6.1 Correction of experimental data from pitching plate study
In this section, we study application of our corrected unsteady method to correct
previous UCMT data from experiments on a pitching airfoil (Weber et al, 1995). Our
87
intent here is to study how much the properly converted lift values will deviate from lift
values determined using the quasi-steady approximation, as was done in Weber et al
(1995). We point out that Weber used a single acoustic leg downstream of the airfoil
(see Fig.1, Weber et al, 1995) and not a closed acoustic path. Also, circulation data was
often obtained for only one-half of an oscillation cycle. Given those facts, we will
accept the measured circulation values as valid, and apply our unsteady correction
methods.
Reduced frequency k is one of the primary parameters studied in Weber et al
(1995) for the pitching airfoil. Fig 3.14 represents a low reduced frequency case
(k=0.059) in Webers work. In Fig 3.14a, measured and lift coefficient values from
Weber et al (1995) are presented for a pitching NACA 4418 airfoil with motion
t t
a
10 sin 11 7 sin
0 0
0
+ = + = . The ) (t f = curve shows that the experimental
data was only acquired in one half of a sinusoidal oscillation cycle. Prior to reaching the
maximum angle of attack, the circulation and lift curves have a similar sinusoidal shape,
but after the maximum point, they drastically deviate from the sinusoidal behavior. The
lift and circulation values lead the angle of attack curve by a fairly large phase angle. In
Fig 3.14b, the time derivative of circulation d/dt and the unsteady correction factor R
(from Eq.(A.19)) for this data are presented. The value of R is generally small in the
area of the maximum angle of attack. Physically, when the pitching airfoil approaches

max,
its motion gradually becomes slowdown and stop, the contribution from the quasi-
steady term A in Eq. (2.3) to the total lift become larger, and correspondingly leads to
the effect of the unsteady term B smaller. In Fig 3.14c, the R and d/dt results from Fig
88
3.15(b) are applied through Eq. (2.36) yielding the corrected lift coefficients which we
compare to the original Weber et al (1995) data. Large correction in lift coefficient is on
the order of 0.3 in C
l
(about 25% error) occurred at low AOA. However, due to the
small values of d/dt for high AOA, only small corrections in peak lift coefficient
values are found.

In Fig 3.15 and Fig. 3.16, Similar results for intermediate and high reduced
frequencies from Webers data are presented. The qualitative trends for these results are
similar to those in Fig. 3.14. The difference between corrected lift and original data
increases with increase of amplitude and reduced frequencies. Further of application of
our unsteady correction methods to experimental data is limited by the lack of UCMT
data at present in unsteady flows. Future anticipated studies to address this lack of data
will be discussed in chapter 6.

3.7 Summary
The results in this chapter show that quasi-steady approximation based on K-J
theorem used by previous investigators in UCMT and oscillating airfoil studies has
limitations for predicting the unsteady lift force due to the lack of consideration of
unsteady effects. Our corrected unsteady method can improve predictions of the phase,
rms lift in oscillating plate flows at low Re number, compared to the quasi-steady
method. Our work in this chapter further verified the methods developed in previous
chapter to properly convert the time dependent bound circulation, which is measurable
with the UCMT technique, to unsteady lift force. We have also shown that specification
89
of the size and location of the closed acoustic path are crucial to bound circulation
determination, and hence will definitely affect the instantaneous lift force measurement.
The knowledge studied here will provide guidance for experimental design when using
the UCMT technique. This path sensitivity also has implications for PIV and LDV
studies that perform integrations within a specified closed path. As an extension of
current developed approximation for the oscillating plate flow, we suggest a general
relation between the time-dependent bound circulation measurable with the UCMT
technique and the time-dependent lift force that may have applications to other unsteady
flows, through this relation

dt
t d
l R t U t L
) (
) ( ) (

+ = . (3.13)
and l is the characteristic length of the body. R will be vary with body shape and
motion, and may not be a concise analytical result as in our studied flow where
4
3
= R .
For example, in the study of the pitching airfoil in chapter 2, R is a complicated function
of time, oscillating amplitude, reduced frequency, angle of attack and pivot location.
But in certain cases, R could be determined and used to develop further unsteady
approximations. For example, in chapter 4, we will study the application of Eq. (3.13)
to a classic unsteady flow over a stationary cylinder at low Reynolds number.
To this point, we have fulfilled several major goals of the present work
including developing a method to properly convert the time dependent bound
circulation, which is measurable with UCMT or PIV techniques, to instantaneous lift
force. We also better understand the importance of closed acoustic path selection on
90
bound circulation determination. Hence, our work has expanded the capability of
UCMT for unsteady flow studies.
Reader may noticed that in our path sensitivity studies, we only considered it at
a fixed =6
0
and Re=100. We believe the path sensitivity will also be affected by other
parameters such as Reynolds number, angle of attack, turbulence intensity, and
systematically investigations about these issues will be an interesting topic in future.
91


Fig 3.9 A typical case of plate instantaneous location and bound circulation variation with time, k=0.4,
h
a
/c = 0.1.
92








Fig. 3.10 Typical lift force obtained from different methods at k=0.4, h
a
/c=0.1
93



K
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45

0
10
20
30
40
50
60
70
h
a
/c
0.04 0.06 0.08 0.10 0.12 0.14 0.16

0
10
20
30
40
50
60
k=0.3

(1)
quasi-steady

(3)
corrected unsteady
h
a
/c=0.1

(3)

(1)
(a)
(b)



Fig 3.11 Phase difference variation with plate oscillating amplitude and reduced frequency
at Re=100
94



Fig 3.12 Rms error variation with plate oscillating amplitude and reduced frequency
95



Fig 3.13 Typical numerically attainable circulation distributions along plate at an
instant. Re=100, =6
0
, k=0.2, h
a
/c=0.1. (a) local chordwise vortex element strength
distribution. (b) Integrated vorticity distributions from the optimum path and a path
just from leading edge (LE) to trailing edge (TE).
96



Fig.3.14 Application of corrected unsteady methods to Weber et al (1995) measurements on a
pitching NACA 4418 airfoil. ) 10 sin( 11 7
0 0
t + = , k=0.059, Re=5.410
5
. (a) Lift
coefficient and correlation measurements of Weber et al (1995). (b) Behavior of first
derivative and unsteady correction factor R (from Eq. (A.19)). (c) Corrected lift coefficients
compared to Weber et al (1995) original results.
97



Fig 3.15 Application of corrected unsteady methods to Weber et al (1995) measurements on a
pitching NACA 4418 airfoil. ) 54 . 19 sin( 20 9
0 0
t + = , k=0.113, Re=5.510
5
. (a) Lift
coefficient and correlation measurements of Weber et al (1995). (b) Behavior of first derivative
and unsteady correction factor R (from Eq. (A.19)). (c) Corrected lift coefficients compared to
Weber et al (1995) original results.
98



Fig.3.16 Application of corrected unsteady methods to Weber et al (1995) measurements on a
pitching NACA 4418 airfoil. ) 23 . 53 sin( 26 10
0 0
t + = , k=0.251, Re=6.710
5
. (a) Lift
coefficient and correlation measurements of Weber et al (1995). (b) Behavior of first derivative
and unsteady correction factor R (from Eq. (A.19)). (c) Corrected lift coefficients compared to
Weber et al (1995) original results.


99
Chapter 4

Circulation Methods for Instantaneous Lift Force
Determination Circular Cylinder Flow

4.1 Introduction
In chapter 2 and 3, we have achieved a better understanding of lift determination
in unsteady thin airfoil flows through analytical and numerical studies. As long as the
bound circulation is varying with time, unsteady effects have to be included for accurate
instantaneous lift force determination. In our developed method, the instantaneous lift
force is composed of the quasi-steady lift expressed in same form as Kutta-Joukowski
(K-J) theorem, and an unsteady term involving the time-derivative of bound circulation,
adjusted with a defined unsteady lift correction coefficient R (for the oscillating thin
airfoil, R=0.75). Another important issue we have identified is that the size and the role
that ultrasonic sound path is critical in determination of time-dependent bound
circulation in the UCMT method. This systematic study will be helpful in design of
sound paths in UCMT experiments. In chapter 4 we study whether the framework
developed in previous chapters can be extended to a classical unsteady flow, low Re
vortex shedding from a stationary bluff body. We again take a similar approach as in
chapter 3 where numerical simulation are preferred and studied over future anticipated
experimental UCMT studies. Before describing our work in this chapter, we will first
present a literature review of past investigations on instantaneous lift measurement in
vortex shedding studies.
The fundamental question addressed in this chapter and by previous
investigations is the proper method for converting flow field data (velocity, vorticity or
100
circulation field) to determine resultant fluid forces on the bluff body. This issue is
important in PIV and LDV in addition to UCMT studies.
Compared to the in-line (drag) force well represented by Morison equation in
certain flow situation, the lift force is more difficult to analyze because it is more
sensitive to how vortices are formed, shed and move in bluff body wake. Several
researchers have suggested that if detailed information about vortex strength and motion
are available, then a Blasius equation could be a good choice to calculate the fluid
forces. Using this equation, Maul & Milliner (1978) have qualitatively studied the
relationship between development of the fluid force and the generation and movement
of shed vortices in the wake. Ikeda & Yamamoto (1981) used similar ideas for a rough
prediction of lift force by estimating the rough motion and strength of vortices in the
wake from particle streak visualization photographs.
With a desire to acquire quantitative information and have a better
understanding of the role of vortices in fluid loading, Obasaju & Bearman (1988)
experimentally studied the sectional and total lift force acting on an oscillating circular
cylinder in a water tunnel. For bound circulation determination, the velocity
measurement around the cylinder at its center span were made by using a thermo system
Laser Doppler velocimetry (LDV) working in forward scatter mode with a frequency
tracker, Measurements were made within a square circuit enclosing the cylinder (Fig
4.1a). The measured bound circulation was converted to instantaneous lift forces based
on the K-J theorem. Simultaneously, the pressure average technique (summarized in
chapter 1) was used to obtain the instantaneous lift force for the purpose of comparison.
101


Mode
sensor
point
0.225D
1.35D
U(t)
To pressure
sensor
(a)
t/T
K-J theorem
Pressure average

(b)
Fig 4.1 Lift force measurement on an oscillating cylinder. (a) Schematic of a square
path enclosing cylinder (Fig. 1 in Obasaju & Bearman (1988)). (b) Comparison of lift
force from Kutta-Jukowski theorem and that from pressure average method at KC=6.75
(Fig. 20(a) in Obasaju & Bearman (1988)).
102










Fig 4.2 Instantaneous lift force at KC=10 obtained from three methods by
Lin & Rockwell (1996). Difference in lift amplitude and phase exist
between different methods (Fig. 2 in Lin & Rockwell (1996)).
KC=10
C
L
t/T
Blasius equation
Lighthills concept
Pressure
average
method
103
A typical result from this work is shown in Fig 4.1(b). The agreement between
the lift obtained from the K-J theorem and that from the pressure average method is not
good, but the gross features of the curves, such as the positions of the peaks were still
well predicted. Hence, they suggested that K-J theorem could be considered as a good
first approximation to determine the unsteady lift on a bluff body, and the authors
thought it was a remarkable finding and could not be explained and may be fortuitous.
Lin & Rockwell (1996) studied the instantaneous lift force on an oscillating
circular cylinder using the PIV technique. After obtaining the vorticity distribution in a
region at successive instants, they used theoretical concepts such as the Blasius equation
and Lighthills concepts to provide an experiment determination of the instantaneous
lift (Fig 4.2). These methods were compared to lift forces obtained using the pressure
average method. Larger difference exists between the instantaneous lift curves obtained
using the different methods. They also mentioned that their future efforts would focus
on studies when varying the size and location of the enclosing region where the
sensitivity of the lift calculation is determined. The goal would be to enhance the
accuracy of their methods through a systematic study similar to these we have
conducted in the presented work.
These show that work still needs to be done to gain a better fundamental
understanding of the proper method for converting flow field measurements into
instantaneous lift curves in wake studies. In this chapter, we seek to extend our
previously developed framework to address issue. We study the low Reynolds number
vortex shedding from a stationary bluff body (circular cylinder) using similar numerical
104
techniques as in chapter 3. We focus on whether instantaneous bound circulation values
(of the type attainable from UCMT experiments) can be properly converted to
instantaneous lift values in this flow. Motivated by earlier work of Wu (1981) and the
cited previous investigations, we study various force models including a quasi-steady
Kutta-Jukowski method, a vortex flow force model (Lighthill, 1986), and a new
unsteady correction model similar to that developed for the thin airfoil flow in chapters
2 and 3. Finite element simulations for flow over a circular cylinder at Re = 100 are
once again performed in order to obtain true instantaneous lift values (from surface
stress integration), which are then compared to various combinations of the force
models. It is determined that stream-wise circulation distributions in the wake behind
the circular cylinder are required to properly apply the vortex flow force model by
Lighthill (1986). These distributions would be attainable in experimental UCMT
studies. Again, the important issue of the proper size and location of the closed path for
bound circulation determination will be studied.
In section 4.2, we describe additional issues related to the numerical methods of
chapter 3 that are now relevant to our circular cylinder flow investigation. We then
study the streamwise circulation distributions in the wake behind the circular cylinder.
These circulation distributions are required to properly determine the vortex force based
on Lighthills model in section 4.3. In section 4.4, the effect of size and location of the
mimic closed ultrasound path on bound circulation determination is investigated.
Finally, lift forces results determined from different fluid force models will be discussed
in section 4.5. We will show that proper superposition of the three previously listed
105
models (K-J quasi-steady, vortex force, and unsteady correction term) can accurately
predict instantaneous lift values on circular cylinders.

4.2 Numerical Method
The numerical code used here is largely the same as that used in chapter 3,
which is based on the classical Garlekin finite element method to solve the Navier-
Stokes equations. Difference includes changing the coordinates of the no-slip
boundaries according to the geometry of the body (circular cylinder here), and
numerical grid. Parameters are all non-dimensionalized with free stream speed U, and
cylinder diameter d, for example, velocity is normalized as U V / V
*
= , time as
d tU t /
*
= . The origin of the xy coordinate system is set at center of the circular
cylinder. The numerical grid is generated through PATRAN in a computational domain
extending 8 diameters in the y (transverse) direction (-4< y/d < 4) and 24 diameters in x
direction (-4< x/d < 20). The mesh in vicinity of cylinder is much finer so that the
detailed near wake flow information can be captured. The total number of nine-node
quadratic elements is NM=984, number of nodes is NND=4068. The input Reynolds
number is Re=100, and time step t=1/40.
The boundary conditions of the computational domain are set as follows; at
inlet, upper and lower grid boundaries, the velocity is set as the free-stream velocity. At
the surface of the circular cylinder, no-slip condition is maintained. At the outlet
boundary, a synthetic (constant pressure) condition is utilized as in chapter 3.
The solution of the problem was conducted using a segregated procedure for the
106
primary unknowns at each node. The algorithm of interpolation with bi-quadratic
polynomials was used for solutions of velocity components and pressure at any arbitrary
spatial location. By using the potential flow solution of pressure field as initial guess to
accelerate the convergence process, the conservation equations were solved at each time
step using an explicit step for an intermediate velocity field. The numerical scheme has
been used previously by Olinger et al (1995) to model flow over an oscillating cylinder
wake at low Reynolds number. In the present work, stationary cylinder wakes are
studied.
A typical result from the numerical simulation is shown in Fig 4.3 for a low
Reynolds number Re=100 flow. Fig 4.3(a) shows the flow field in vicinity of circular
cylinder at tU/d=78.875, which shows clearly the velocity vector field, vortex
generation and shedding process. Fig 4.3(b) is a typical time trace of the lift force acting
on the circular cylinder, obtained from surface stress integration method at each time
instants. The periodical lift force is due to the periodic vortex shedding process in the
wake of cylinder shown in Fig 4.4. The numerical code can provide accurate velocity
data to be used for circulation determination along an arbitrary selected closed path.

4.3 Wake circulation distribution and vortex lift force
In this section, we study how circulation measurements (obtainable with a
UCMT method) can be used to determine the vortex force developed by Lighthill
(1986). The fluid forces in that model are related to wake vortex strength and locations.
Lighthill proposed the following expression.
107

=

d r
dt
d
F
v

r r
r
2
1
. (4.1)
where r
r
is the position vector (relative to the origin) of a fluid element in wake flow
with vorticity
r
. In a two-dimensional flow, j y i x r
r r
r
+ = , k
r
r
= , 1 = dA d . The
cylinder wake is divided into N
x
sections in x direction and N
y
sections in y direction
(Fig 4.5a). The circulation of a fluid element at location (x
i
,y
j
) is given by


= =
ij
A
ij
A d l d V
r
v
r r
. (4.2)
within the infinitesimal area A
ij
, vorticity can be assumed as constant, hence

ij ij ij
A . (4.3)
k
A
ij ij
ij
ij
ij
r
r

, . (4.4)
Discretizing equation (4.1) and substituting equation (4.4), the vortex force will be

+

= = = =
x
y
x
y N
i
N
j
ij ij ij
N
i
N
j
ij ij ij ij v
j x i y
dt
d
A j y i x
dt
d
F
1 1 1 1
) (
2
) 1 ( ) (
2
r r
r
r r r

. (4.5)
j x
dt
d
i y
dt
d
F
x
y
x
y N
i
N
j
ij ij
N
i
N
j
ij ij v
r r r

= = = =
+
1 1 1 1
] ) (
2
[ ] ) (
2
[

. (4.6)
Hence, the vortex lift force will be

= =

x
y N
i
N
j
ij ij v
x
dt
d
L
1 1
) (
2

. (4.7)
The corresponding lift coefficient will be
108


(a)


(b)

Fig 4.3 Typical results from numerical simulation of flow over a stationary circular
cylinder at Re=100. (a) Velocity vector field and streamlines at one time instant
presented the vortex generation and shedding process. (b) The time trace of lift
force was from the stress integration method.

109

(a) tU/d=76.4 (b) tU/d=77.8
(c) tU/d=79.2
(d) tU/d=80.58
Fig 4.4 Flow visualization of the near wake of circular cylinder during one lift cycle at
Re=100. (a)-(e) are instantaneous wake vortex structures at five time instants shown in (f),
the periodic time trace of the lift force.
(e) tU/d=81.98 (f) C
l
time trace
110

(b)


ij
,
ij
r y
2
/d x
2
/d(movable)

F A B N
y

y

x


E D C N
x

x
1
/d y
1
/d
(a)
U
Fig 4.5 (a) Schematic of cylinder near wake and definition of variables needed for wake
circulation determination. In x-direction, wake is divided into N
x
sections. In y direction,
it is N
y
sections. (b) Wake circulation distribution in x-direction with y
1
/d=-0.75,
y
2
/d=0.75. Wake is divided into a series of consecutive smaller paths like ABCD, with
the path extent in x-direction |AB|=(x/d)=0.5
111

) (
) (
2
1 1
d U
x
dt
d
C
x
y N
i
N
j
ij ij
Lv

= =

. (4.8)
Form Eq. (4.8), the vortex lift force only dependent on the x-coordinate of the
fluid element, hence all the fluid elements enclosed by a path ABCD will be with the
same x-coordinates, and


= =

y y
N
j
ij ij
N
j
ij ij
x x
1 1
) ( . (4.9)
After defining

=

y
N
j
ij i
1
) ( as the circulation in area enclosed by path ABCD,
and replacing
ij
x with
i
x , Eq.(4.8) yields

) (
) (
2
1
d U
x
dt
d
C
x
N
i
i i
v
l

=

. (4.10)
once
i
and x
i
are known, lift component of vortex force can be obtained. We note that
Eq. (4.10) approximates the true Lighthill vortex force given by Eq. (4.1), it must be
recognized when interpreting numerical results.
Equation (4.10) is a useful expression for UCMT studies on circular cylinder
wakes. It states that to determine lift component of vortex force in Lighthills model,
one needs only measure
i
and x
i
in the wake of a circular cylinder.
i
can be obtained
in a UCMT study by specifying the closed path ABCD (Fig 4.5(a)). It is particularly
useful that only the location x
i
is needed to find the lift force in Lighthills model, since
112
x
i
is known once the path ABCD is specified. In the schematic of determination of
ABCD
in Fig. 4.5(a), we can calculate

EFAD EFBC ABCD i
= = . (4.11)
Measurement of
i
could be accomplished by using Eq. (4.11) where moving
the downstream leg at x
2
/d incrementally will aid finding
i
. In real UCMT test, the
size of acoustic transducers and the length of sound path may affect the resolution of the
i
distribution measurement. For PIV technique, however, Eq. (4.1) may yield more
accurate results considering that the spatial measurement resolution of PIV is greatly
better than UCMT. Following the spirit of chapter 2, we will validate these ideas using
our numerical simulation in the previous work.
Using the numerical algorithm described in chapter 3 applied to the circular
cylinder flow, we studied circulation variation in both x and y direction in the wake. It
was found that the majority of shear layer vorticity is confined to a region between
y
1
/d=-0.75 and y
2
/d=0.75. Hence, they were selected as upper and lower bound of the
path ABCD with the size in x direction set as (x/d)=0.5. Fig 4.5(b) presents typical
wake circulation distributions in the x-direction at tU/d=76.4 and 79.2 corresponding to
the flow visualizations of Fig 4.4(a) and 4.4(c) respectively. The circulation
distributions at the two instants are clearly out of phase reflecting the shedding process
in Fig 4.4. The wake circulation distribution shows a periodic variation with streamwise
distance reflecting sequentially shed vortices in the wake. Peaks in the circulation
distribution reach a maximum near x/d=3.5 and then decay due to viscous dissipation as
113
the vortices move downstream. These distributions provide all the necessary
i
and x
i

information for vortex lift force determination using equation (4.10) at any time instant,
typical time traces of the wake vortex lift are shown in Fig 4.6.
Recognizing that the extent of the wake over which Eq. (4.10) is applied could
affect resultant vortex lift force values, we fix the region EFBC in Fig 4.5(a) as follows,
the legs EC and FB are set at y
1
/d=-0.75 and y
2
/d=0.75 respectively. The leg EF is fixed
just behind cylinder at x
1
/d=0.5, and the downstream leg BC is variable with x
2
/d
adjusting, Eq. (4.11) is then used to find
i
. Fig.4.6 presents four curves of wake vortex
lifts from four different EFBC regions with x
2
/d=5, 10, 15 and 20. Obviously there exist
phase and amplitude difference between these curves. With an increase in size of EFBC
domain, the amplitude of wake vortex lift gradually increases and the phase gradually
decrease (earlier in time), due to inclusion of additional vortex elements in the far wake
that were shed earlier. However, the difference between results for x
2
/d=15 and x
2
/d=20
is much smaller than that between x
2
/d=5 and x
2
/d=10. This is because the fluid
dissipation effect and the contribution to lift from vortices further away from the
cylinder decrease. For region EFBC with x
2
/d=20, we encompass the downstream
bound of our computational domain for the cylinder flow. Based on Fig.4.6, we
conclude that use of region EFBC with x
2
/d=20 will yield accurate wake vortex lift
forces and that contribution of vortices downstream of our numerical boundary can be
neglected. Hence, all of the wake vortex lift force results presented later will be
obtained from region EFBC with y
1
/d=-0.75, y
2
/d=0.75, x
1
/d=5 and x
2
/d=20.

114







Fig 4.6 Time traces of wake vortex lift force obtained from Eq. (4.10) within different wake
domains. The lower, upper and left bounds of the domain are fixed at y
1
/d= -0.75, y
2
/d= 0.75
and x
1
/d= 0.5. The right bound is movable and set at four locations x
2
/d= 5, 10, 15 and 20.
115
4.4 Bound circulation determination
In section 4.3, we focused on the wake vortex force derivation based on
Lighthills concept. In some sense, the contribution of the vortex force to the total lift is
similar to the concept of wake circulation contribution in Wus work (1981). In chapter
2 and 3, we also have shown that contribution from the time-dependent bound
circulation in a region enclosing the cylinder is of significant importance. Our corrected
unsteady approximation, equation (2.9), can properly convert the time-dependent bound
circulation to instantaneous lift force. As a result in the same spirit of chapter 2, issues
related to bound circulation determination will be studied. In this section, we focus on
proper selection of a closed sound path for bound circulation determination in stationary
circular cylinder wakes.
Using the same algorithm for circulation determination introduced in chapter 3,
we can easily calculate the value of circulation around an arbitrary closed path, and
study the effect of path size and location on bound circulation determination. Due to the
geometry of circular cylinders, we select a square shaped path with the center located at
the center of the cylinder as done by Obasaju & Bearman (1988). By varying the side
length of the square path (a), the bound circulation variation with length a can be
obtained.
Fig 4.7 shows this path variation for bound circulation determination. Five
typical paths, A, B, C, D, E are highlighted to study effect of near-wake vortex
structures including the attached and shed vortices labeled in the figure. The attached
vortex develops in close proximity to the cylinder and later sheds down stream (see Fig
116







Fig 4.7 Schematic of path selection for bound circulation determination around circular
cylinder. Re=100, tU/d=76.875. Five typical paths are highlighted to study the effect of near-
wake vortex structures on bound circulation values. Define a as the length of one side of the
square paths, for path A, a/d=1.0; path B, a/d=1.6; path C, a/d=2.0; path D, a/d=3.0; path E,
a/d=8.
117



Fig 4.8 Bound circulation at an instant of time variation with the size of the closed paths. A
small stable range exists in the plot where is independent of a/d, These results can serve as a
useful guide for closed sound path design in UCMT experiments. Labels A, B, C, D, E
correspond to paths highlighted in Fig 4.7.
118
4.4). The smallest path A just encloses the circular cylinder, with each of its legs
tangential to the cylinder.
Fig. 4.8 shows the variation of bound circulation with the size of the paths
defined in Fig 4.7 along with further intermediate path sizes. This plot shows that bound
circulation increases with an increase in path size starting from path A, then it comes to
a stable area in which the bound circulation becomes independent of path length while
reaching its maximum value. With further increase of a/d, area enclosed by the path
begins to include shedding vortices in the wake. The shedding vortices contain opposite
signed vorticity compared to the attached vortex in the other shear layer. This addition
of opposite signed vorticity causes the total bound circulation to decrease with increase
in a/d. For 5 / d a , the bound circulation ) (t approaches zero. As the path becomes
larger and encloses additional shed vortices, the vorticity of these shed vortices
dominate the effect of the attached vortex. Since the downstream vortex street contains
alternating vortices of opposite sign, the measured bound circulation becomes
negligible.
Based on the above results, a proper path should be selected from the stable area
that covers from 5 . 1 / 3 . 1 < < d a In Obasaju & Bearmans experiments (1988), a value
of 35 . 1 / = d a was used, which closely matches our optimum path. We will determine
the bound circulation using this selected path in our later analysis of lift forces on
circular cylinder. Here, we can conclude for the path selection: For bound circulation
determination, sound path should include the attached vortex, but not include any part
of the separated shedding vortex. This fact can act as a guide for future UCMT
119
experiment design. As detailed in chapter 3, the path sensitivity study also has
implications for PIV and LDV studies. We should point out that the optimum path
a/d=1.5 may have some practical limitations. This path implies acoustic sound legs in
close proximity to the cylinder that may require acoustic mirrors that disturb the flow.
However, in an experiment, rectangular acoustic paths, with y
1
/d and y
2
/d legs placed
further away from the cylinder and near the wind tunnel walls, could be used
successfully. These issues require further study.

4.5 Results and discussion
In this section, we present lift force results from our numerical simulations
based on different force models and compared these results with the true lift obtained
from the stress integration method. The wake vortex lift force is based on Lighthills
concept, which is expressed in equation (4.10). The quasi-steady lift force is based on
Kutta-Joukowski theorem, where (t) is the bound circulation within the optimum
sound path (section 4.4). Osbasaju & Bearman (1988) used this approximation in their
experimental work. The lift force from our corrected unsteady approximation (2.22) is
based on unsteady potential flow theory and can count in the unsteady effects. A
variation of this approximation will be applied later in this chapter. Finally, we will put
forward a new lift model in this section to combine the quasi-steady, wake vortex and
unsteady lift models.
Next, we present lift curves from the numerical simulation for the true lift
l
C ,
wake vortex lift
) (v
lw
C , vortex lift
) (v
l
C (including contribution from bound and wake
120
circulation) and the quasi-steady lift
) 1 (
l
C (K-J theorem) in Fig 4.9.
) 1 (
l
C is determined
using the bound circulation obtained from the optimum closed path selected in section
4.4. The vortex lift
) (v
l
C shows the largest phase angle lag behind the true lift
l
C , and
its peak-peak amplitude is larger than
l
C , a 15% deviation of lift rms amplitudes is
observed between
l
C and
) (v
l
C . This deviation should be due to approximation
introduced in derivation of Eq. (4.10). However, it is interesting to note that in
Rockwells experimental work using PIV technique (see Fig.4.2), similar deficiency
between Lighthills vortex lift and true lift was observed. Compared with the vortex lift
) (v
l
C , the quasi-steady lift
) 1 (
l
C better matches the true lift curve. Because
) 1 (
l
C doesnt
include the viscous effects (accounted by vortex lift), we next study the combination of
) 1 (
l
C and
) (v
lw
C . However, due to the large phase difference (
2

) between
) 1 (
l
C and
) (v
lw
C ,
the combination of
) 1 (
l
C and
) (v
lw
C becomes less accurate than
) 1 (
l
C in terms of phase
angle prediction, and little improvement is noted for the rms lift. The best prediction
occurs for the quasi-steady case where values in phase difference of
0
45 between
this case and the true lift are observed. This may serve as some justification for use of
the quasi-steady approximation in earlier wake studies (Obasaju & Bearman, 1988).
However, we will next see that better lift prediction on circular cylinders can be
obtained by incorporating unsteady effects using the framework of chapter 2.
Since the lift approximation
) 1 (
l
C , or various combinations using an
approximation based on Lighthills vortex force concept, cannot predict the true lift
l
C
121




Fig 4.9 Comparison between true lift force
l
C , quasi-steady lift
) 1 (
l
C , wake vortex lift
) (v
lw
C and vortex lift
) (v
l
C . Vortex lift is obtained from Eq. (4.10) within the selected domain.
Quasi-steady lift is determined from K-J approximation. is the phase angle between true lift
and
) ( ) 1 ( v
lw l
C C + .
122







Fig 4.10 Phase comparison of true lift
l
C , bound circulation and its first derivative

dt
d , which will be used for unsteady effect correction.
123
well, we next consider if the unsteady nature of the bound circulation can be used to
increase lift prediction accuracy. In Fig 4.10, we present the results of the time variation
of the bound circulation and its first derivative
dt
d
that was used in our corrected
unsteady approximation in chapter 2. We focus on the phase angle difference
'

between the
dt
d
curve and
l
C curve defined in Fig 4.10, and find
0 '
60 with
dt
d
leading
l
C . In Fig 4.9, we have seen that the phase angle
0 '
45 with
l
C
leading
) ( ) 1 ( v
lw l
C C + . This suggests that
dt
d
curve (which describes the unsteady effect)
has the proper leading phase to correct our lift predictions. As a result, we next study
inclusion of an unsteady term into our lift approximations.
In our corrected unsteady method in chapter 2, we developed the following
approximation

dt
t d
d R t U t L
) (
) ( ) (
) 3 (

+ = . (2.36)
Eq. (2.36) was obtained after applying unsteady potential flow theory, focusing
on the bound circulation (t) that is obtained from a region enclosing the body. Eq.
(2.36) can be converted to non-dimensional form giving

2
) 3 (
) (
2
) ( 2
U
dt
t d
R
Ud
t
C
l

= . (4.12)
We first use Eq. (4.12) to specify a value of the unsteady lift correction
coefficient R, which was found to be for oscillating plate flow in chapter 2. For bluff
circular cylinder flow, R will have a different value and may be a function of various
124
flow parameters such as Re etc. It may be possible to establish R analytically (similar to
our work in chapter 2) using unsteady potential flow theory. Superposition of
freestream flow along with doublet to model cylinder shape and a wake model would be
required. In the present work, we have not undertaken this analysis, but instead we
focus on a systematic study of the effect of R variation on lift results.
To accomplish this, we first choose to study the following lift approximation

{
force vortex
wake
v
lw
force unsteady force steady quasi
l
C
U
dt
t d
R
Ud
t
C
) (
2
) 4 (
) (
2
) ( 2
+

43 42 1 3 2 1
. (4.13)
extending Eq. (4.12), we include the wake vortex force term (although we have shown
it has negligible effect on lift prediction ) to account all physical effects considered by
previous investigators.
We next define a new rms error similar to chapter 2, where
) (rms l
C refers to the
rms amplitude of the true lift value (stress integration)

) (
) (
) 4 (
) (
) 4 (
rms l
rms l rms l
rms
C
C C
e

= . (4.14)
Fig 4.11 presents the results of our systematic study of varying R by showing
the variation of
) 4 (
rms
e with R. From this curve, we found that the best value of R for
circular cylinder flow is 4 . 0 R . The proof that 4 . 0 R should hold for a low Reynolds
cylinder wake is left as a challenge to analysts in potential flow theory, and is an area of
anticipated future work. However, study of our unsteady correction term in Eq. (2.36)
and Sarpkayas generalized Blasius theorem of Eq. (1.17) yields some interesting
125
findings. By comparing the final terms in Eq. (1.17) and Eq. (2.36), we can see that
Rd r
i
k
. If we treat
i
k
r as a constant and
k
as the bound circulation, the parameter
i
k
r represents the location of a vortex element in the flow. Then d Rd r
i
k
4 . 0 = is
obtained using our result from Fig. 4.11. This says that the primary vortex element
determines unsteady lift in Eq. (2.36) is located at 4 . 0 = = =
d
x
d
r
R
i
k
. Now study Fig.
4.7, the attached developing vortex is located at 4 . 0 =
d
x
, suggesting that it is the
attached vortex development dominates the unsteady lift creation. It is physically
plausible that the attached vortex adjacent to the cylinder in the near wake has a
dominant effect on unsteady lift creation, compared to vortices at larger
d
x
values.
Fig 4.12 presents comparison of lift curves using the approximations in Eq.
(4.12) and (4.13). We first observe that the
) 4 (
l
C lift approximation (with 4 . 0 R ) best
approximates the true lift curve. While there is excellent agreement between the two
curves in rms amplitude (as required by Fig 4.11), small phase difference still exists.
For comparison, we also present the results for
) 3 (
l
C (Eq. (4.12)) which deletes the wake
vortex force effect. Degradation in lift prediction is observed. We do however note that
this approximation could also be adjusted to better match the true lift curve through an
additional systematic study of R variation for this approximation, similar to Fig. 4.11.
From this chapter, we may conclude that reasonable prediction of lift force
based on circulation method generally has to include unsteady effects, as long as the
bound circulation is varying with time. Lighthills vortex lift force concept, when
126
applied over the entire flow, using a UCMT based technique, results in approximate
15% error in lift rms amplitude compared to the true integrated lift. As a result, an
alternative lift model based on measuring bound circulation with an acoustic path
enclosing the cylinder is proposed.








127









Fig 4.11 Rms error of lift
) 4 (
l
C approximation with variation in unsteady lift
correction coefficient R. 4 . 0 R for circular cylinder flow yields 0
) 4 (

rms
e .
128




Fig. 4.12 Comparison of lifts between true lift
l
C , vortex lift
) (v
l
C , unsteady corrected lift
) 3 (
l
C (K-J theorem + corrected unsteady) and total lift
) 4 (
l
C ( K-J theorem + corrected unsteady
+ vortex force).
129
Chapter 5

Experimental UCMT study on Three-dimensional Flows

5.1 Introduction
To expand the capability of UCMT for lift determination in unsteady and three-
dimensional flows, we have developed methods to properly convert the time-dependent
bound circulation measurable with UCMT to instantaneous lift in chapter 2 and 3, based
on unsteady potential flow theory. In chapter 4, we studied the instantaneous lift in bluff
body flows, by applying our developed methods along with Lighthills vortex force
concept in numerical simulations. However, we have not yet focused on lift
determination in three-dimensional flows. This will be the main topic in this chapter in
which we will conduct UCMT experiments to study the mean sectional lift distribution
along aerodynamic bodies with three-dimensional features.
In UCMT experiments, it is possible to systematically study the spanwise lift
distribution along structures. For our aerodynamic body, we select the flow around a
flat plate because it has three distinct flow regimes dependent on angle of attack (AOA).
At smaller AOA, it can be characterized as streamlined flow; at intermediate AOA, a
stalled flow; and at larger AOA, it becomes a bluff body flow. As a result, outcomes
from a flat plate flow study should have broad application. Furthermore, it is convenient
to design and fabricate plates with a variation in chord length along the span to study
the three-dimensional lift distribution.
In the past, most investigations on three-dimensional flows have focused on
vortex generation and shedding patterns behind the body. For example, Nuzzi et al
130
(1992) studied the three-dimensional vortex formation from an oscillating, non-uniform
cylinder. The gradual variation of the cylinder diameter acted to detune the highly
coherent vortex formation in the spanwise direction and promoted the occurrence of
non-periodic and period-doubled states. This approach can result in an effective
destabilization of the near wake vortex formation leading to three-dimensional flow.
Recently, considerable attention has also been paid on the development of three-
dimensional flows from stationary plate trailing edges. Lasheras & Meiburg (1988,
1990) employed combined experimental and numerical approaches to study the
evolution of the wake from the trailing edge of a thin flat plate at low AOA having
spanwise perturbations. They found that an important feature of the three-dimensional
flow is the induction of spanwise undulations of the large scale (Karman) vortices. Such
undulations can exhibit either an in-phase or varicose pattern.
An important aspect of the investigation of three-dimensional flow is
U
b
c
AR=2b/c
z
y
x
L(z)
Tip vortex
(3D area)
Bound
vortex
L(z)
o z b

Fig. 5.1 Schematic of local sectional lift distribution on a finite wing
131
measurement of the local sectional fluid forces, which are crucial to aeroelastic analysis
of structures such as aircraft wings (Fig. 5.1). However, past research in this direction is
somewhat limited. Prandtl (1918, 1921) first analytically developed the lift line theory
for predicting the lift distribution and induced drag on a finite lifting surface (wing). In
his theory, he hypothesized that each spanwise section of a finite wing has a local lift
equivalent to that acting on a similar section of infinite two-dimensional wing having
the same circulation. This assumes negligible flow in the spanwise direction along the
wing. However, this assumption is easily violated if three-dimensional flow exists.
Recently, Bastedo et al (1985) experimentally studied the finite wing performance with
a pressure measurement method at low Reynolds number Re = 810
4
~ 210
5
. For a
rectangular wing with AR=4, the agreement between lift from Prandtls finite wing
theory, and the measured local sectional lift based on the integration of the local
pressure distribution became worse with an increase of angle of attack (Fig 5.2a).
Bhagwat et al (2000) employed Laser Doppler Velocimetry to measure the spanwise
bound circulation distribution on a rectangular wing with AR=9.6. As opposite to
Prandtls smooth lift distribution, a peak was found in the measured circulation
distribution at a location near the tip area where strong three-dimensional flow exists
(Fig 5.2b). Due to various experimental limitations, the above investigations could not
measure the lift or circulation distribution along the whole span. Our effort in this
chapter includes experiments to accomplish this to learn more about this type of three-
dimensional flow.
132
Another research area that has been arisen recently is the study of low aspect
ratio (AR) wings at low Reynolds number. These wings have application in the design
of micro aerial vehicles (MAVs), which are small aircrafts with maximum dimension
less than 25cm. MAVs can be used as surveillance or reconnaissance vehicles to carry
visual, acoustic, chemical or biological sensors in military or commercial applications.
For these vehicles, the accompanying chord Reynolds numbers are in the range of
Re=210
4
to 210
5
. This presents numerous aerodynamic challenges to the designer of
MAV. Very little data exists for low Reynolds, low aspect ratio wings, and the
applicability of exist model such as Prandtls lift line theory should be rigorously
examined. Also, details of the spanwise lift distribution are critical to MAV design.
Particularly in swept low AR wings, this lift distribution determines the location of the
MAV center of pressure and neutral point and thus is critical in determining aircraft
stability. Control and stability of MAV aircraft is a current challenge due to their small
size. Experimental studies (Bastedo & Muller, 1985; Pelletier & Muller, 2000) have
been conducted to determine the total lift and drag forces in low Reynolds, low AR
wings. However, little attention has been placed on measurement of local sectional lift
distributions.
In the following sections, we will present experimental lift results obtained from
UCMT and force balance measurements. In section 5.2, we will describe the
experimental apparatus for our three-dimensional flow studying. Data processing and
analysis will be described in section 5.3. In section 5.4, two-dimensional flow over a
flat plate for a wide range of AOA is studied to validate our experimental methods. In
133

(a)

(b)
Fig 5.2 (a) Comparison of lifting line predictions with spanwise lift distribution obtained from
integrated pressure distribution for Re=210
5
and AR=4 (Fig.9 in Bastedo et al (1985)). (b) LDV
measurement of bound circulation along airfoil span at Re=2.5310
5
and AR=9.6 (Fig. 12(a) in
Bhagwat et al (2000)).
134
section 5.5, we present a study of flow over flat plate with local chord length variation
along the span. In section 5.6, we will study lift distribution on low Reynolds number,
low aspect ratio wings applicable to MAV design. Finally, we will summarize our
experimental study in section 5.7.

5.2 Experimental apparatus
Measurements presented in this chapter were conducted in two low-speed wind
tunnels located in the Mechanical Engineering Department in Worcester Polytechnic
Institute. All of the UCMT testing was conducted in an open circuit wind tunnel
(hereafter tunnel 1). Force balance lift measurements were conducted in tunnel 2, a
closed circuit tunnel.

5.2.1 Experimental study for UCMT measurements
The entrance of tunnel 1 includes a honeycomb straighter to uniform flow, and
four wire mesh screens used to reduce turbulence intensity. The contraction ratio is
approximate 12, with entrance size of 165cm 206cm. The test section has dimensions
of 45.7cm (height) by 61cm (width) by 91.4cm (length); The available velocity range is
s m U s m / 64 / 3 , measurable with a Pitot-static tube set at a location near the
entrance of the test section; Free stream turbulence intensity measured from hot-wire
anemometry is approximate 09 . 0 07 . 0
'
. A downstream diffuser connects the test
section to the fan assembly consisting of 75hp 3-phase induction motor driving an axial
135
fan. Tunnel speed control is accomplished by adjusting damper vanes located upstream
of the inlet to the blower.
A schematic of the test setup is shown in Fig. 5.3. A triangular ultrasound path
was constructed for circulation measurement shown in Fig. 5.3a. The triangle path was
also used in the work of Desabrais (1998) who verified UCMT measurements on a 2D
airfoil based on comparison with results measured from a load cell force balance. In
Fig. 5.4, we present a scaled schematic of the triangular sound path. It is noted that we
measured spatial circulation distributions based on this triangular path (which is not the
optimum path) prior to studying the path sensitivity. The triangular path is compared to
the allowed acoustic path locations (Fig. 3.7) from the path sensitivity study of chapter
3. Our main concern is that the downstream acoustic reflector is at an x
2
position larger
than that allowed in Fig 3.7. Study of Fig 3.6b shows that measured circulation values
would be expected to decrease if the downstream reflector is at larger x
2
/c values (here
x
2
/c 3), suggesting the triangular path may slightly underpredict values. We also
note that the Reynolds numbers are different in our experiments compared to the
simulation of chapter 3, so these trends are only speculative.
In real experimental test, the sound path may not keep as straight geometric line
due to the effects of turbulence and wake vortex flow. This may affect the final
measurement precision. However, these effects may be negligible considering the
freestream velocity is just 5m/s, far less than the sound speed (340m/s) in air.
In UCMT test, two 100KHz ultrasonic transducers (3/4 diameter, bandwidth
about 20KHz) from Panametric were used as both transmitter and receiver to generate
136



Fig 5.3 Schematic of the experimental arrangement in tunnel 1. Spanwise lift distributions along plates
with chord length variation are studied. (a) Setup for UCMT measurement. (b) Setup for smoke
visualization.

U
Sound path plate
Side wall of W.T.
Top View
Acoustic reflector
Ultrasound
transducers
cable
computer
preamplifier
o
x
y
W
W.T. size: WxH=24

x18


(a)
camera 300w ELH High
Energy Bulb
bottom wall of W.T.
Smoke Generator
smoke
U
End plate
plate
Light sheet
o
x
z
Rotating Plug Side View
H

(b)
137




y
x
U
Allowed region
" 24
1
= L
" 25
2
= L
4 . 1 /
1
= c x
Transducers
Acoustic
Reflector



Fig. 5.4 Diagram of the closed triangular acoustic path and its relative location to our
allowed region from the path sensitivity study.
138
and receive sound pulses, and two acoustic reflectors (mirrors) were used for redirecting
the acoustic pulses. The broadening of pulse width and deflection from its undisturbed
propagation direction due to freestream flow in the tunnel was carefully calculated and
checked. A Panametric GP68 Flowmeter controlled the emitted and received sound
pulses and measured their average transit times. These transit times are critical for mean
circulation measurement.
A PC based data acquisition system was used for data storage and post-process.
Two Panametric model PRE-120 preamplifiers were used to amplify the signals
received from the transducers. Each measurement of the transit time was obtained by
averaging approximately 100 data acquired over 7 seconds and stored in a data
acquisition file. The relationship between the measured transit times and the circulation
has been described in chapter 1 as

) ( ) ( 2 ) (
2
1
2 2
ccw cw
ccw cw
p
ccw cw
T T
T T
l
T T a
+
= =
. (1.21)
where a is sound speed, l
p
is the total length of sound path, T
cw
, T
ccw
are the transit times
(Johari & Durgin, 1998).
The total length of the closed sound path l
p
was carefully measured as 201cm.
Transducers and acoustic reflectors were recessed into cavities in the test section walls
to reduce disturbances to the flow field. In order to reduce cross-talk between
transducers, rubber O-rings were used to isolate the transducers from each other and the
supporting structure.
The effect of interference from boundary layers on the tunnel walls was assumed
negligible because of the symmetry of the designed sound path. The vorticities in the
139
boundary layers on opposite sides of the tunnel were assumed to have equal and
opposite signs. Furthermore, a small baseline circulation that was measured in the
tunnel with the flow on, but with no flat plate in place, was subtracted from the
circulation measurements.
Calibration of the UCMT setup in still air showed that resolution of the GP68
Flowmeter measurement was about 50 ns, which corresponded to an error about 2.3%
in circulation values, based on the measured transit time at U=5m/s and plate angle of
attack of =6
0
.

In addition, a barometer and a thermometer were used to measure the
ambient temperature and pressure for accurate determination of air density and sound
speed.
In order to correlate the measured lift distribution to the flow structures in flat
plate wake, a smoke flow visualization method was employed (Fig. 5.3b). A propylene
glycol smoke generator was set in front of the wind tunnel entrance to provide smoke as
a flow tracer. A 300W ELH high energy bulb provided a light sheet (with a thickness of
about 1.5mm) to visualize the wake vortex flow in a two-dimensional plane. The
spanwise (z direction) location of the smoke generator probe and the light sheet were
fixed at the mid-point of test section height. The plates were vertically aligned in the
tunnel (span in z direction), and the measurement of local circulation and flow
visualization were implemented by moving the plate in z direction with the aid of two
circular rotating plugs in the test section walls, which were designed to adjust the plate
spanwise position and its angle of attack. In order to reduce interference from top and
bottom test section walls, transparent end plates were designed following Stansby
140
(1974) and mounted on the plate, with a distance of about one chord length from the test
section walls.
Four thin flat aluminum plates with thickness of 3.2mm and maximum chord
length of 89mm were studied (Fig 5.5). Case 1 was a 2D plate used for validation of
two-dimensional flow results. Case 2-4 were plates with sinusoidal varying chord length
in the center section plate from z=0 to z=z
0
=6.35cm. The chord length c(z) for case 2
and 3 was given by ) 4947 . 0 cos( 016 . 1 874 . 7 ) ( z z c = . Case 2 was used to study the
effect of trailing edge chord variation on lift distributions, and case 3 for studying the
effect of leading edge variation. Case 4 was used to examine combined trailing and
leading edge effects. Its chord length variation was ) 4947 . 0 cos( 032 . 2 858 . 6 ) ( z z c = .
The chord length variation was carefully machined on a CNC machine. For local
circulation and lift measurements, 12 stations in z direction were tested with a resolution
of z=6.35mm from z=0 to z=1.1z
0
.
z
o
C
C(z)
o
z
z z
z
o
Case 1 Case 2
Case 3 Case 4
C(z)
o
C(z)
o
z
o
z
o
Table 5.1 Plate parameters

Case 1 Case 2 Case 3 Case 4
z/z
0
c(z)/C c(z)/c c(z)/c c(z)/c
0.0 1.0 0.7714 0.7714 0.5429
0.1 1.0 0.777 0.777 0.5541
0.2 1.0 0.7933 0.7933 0.5865
0.3 1.0 0.8134 0.8134 0.6268
0.4 1.0 0.8504 0.8504 0.7008
0.5 1.0 0.8857 0.8857 0.7714
0.6 1.0 0.921 0.921 0.8421
0.7 1.0 0.9529 0.9529 0.9058
0.8 1.0 0.9782 0.9782 0.9563
0.9 1.0 0.9944 0.9944 0.9888
1.0 1.0 1.0 1.0 1.0
1.1 1.0 1.0 1.0 1.0
Fig 5.5 Schematic of four plates with varying chord length used for study of sectional
lift distribution on flat plates.
141
The spanwise lift distribution along low aspect ratio wings at low Re was also
measured using UCMT in tunnel 1 (Fig 5.6a). The UCMT setup was the same as
described in section 5.2.1. Plates were vertically aligned in the test section using a sting
and support that was designed to move the plate in spanwise direction and change its
angle of attack. The distance from wing tips of the low AR wings to the test section
walls was always greater than one chord length to reduce wall effects on trailing
vortices. Due to the symmetry of the wings, we measured the spanwise circulation and
lift distributions along the upper half of the plates only.

5.2.2 Experimental setup for force balance measurement
Force balance measurements were conducted in tunnel 2 to further verify
UCMT measurements (Fig5.6b). Wind tunnel 2 is a 24 re-circulating tunnel.
Dimensions of its test section are 61cm wide by 61cm high by 240cm long. The free-
stream velocity in test section can vary from 3m/s to 55m/s. Turbulent intensity is
approximate 0.5% ~ 0.8%.
Force measurements were conducted with a LVDT (Linear Variable
Displacement Transducer) dynamometer that was incorporated into the wind tunnel.
The dynamometer is externally mounted to the bottom of the test section to
independently measure lift and drag based on a internal beam deflection principle.
These deflections are translated into a voltage change in the LVDT output signal
filtered and amplified by a signal-conditioning unit (Popp, 2000). Calibration of the
142





Fig. 5.6 Schematic of experimental setup for low aspect ratio wing study. (a)
Arrangement for UCMT measurement of spanwise lift distribution along wings in
tunnel 1. (b) Setup for force balance measurements in tunnel 2.
143
LVDT is using known external weight before and after each test run was
conducted. Fig 5.7 shows a typical calibration curve used for lift measurements. Once
the output voltage is collected by the signal-conditioning unit, it was sent to a Cyber
Research

CYDAS 6402 HR data acquisition card, which interfaced with a PC through


a commercial software package, HPVee

, to visualize and analyze the acquired signal.


It permits the user to alter the scan rate and time period for the recorded signal. Fig 5.8
is a typical acquired voltage signal in HPVee

with a sampling rate as 1000HZ and scan


time of 6 seconds during the calibration process.


Four low aspect ratio plates made of aluminum were studied in wind tunnel 2 as
part of our focus on low Re, low AR aerodynamics study. The plate thickness was
3.2mm, plate chord length c and span b are varying with aspect ratio, which were shown
in Table 5.2. Free-stream speed of oncoming airflow in test section was set at U=5m/s,
yielding Reynolds number (based on chord length of the plates) in the range of 2.110
4

to 4.1 10
4
. The distance from wing tips of the low AR wings to the test section walls
was always greater than one chord length to reduce wall effects on trailing vortices.
Plates were horizontally aligned in wind tunnel test section and attached to a pivot strut
Table 5.2 Geometry of the four low aspect ratio wings
AR b(cm) c(cm) Re
1 10.16 10.16 41,753
1.5 15.24 10.16 41,753
2 17.78 8.89 36,534
3 15.24 5.08 20,876
144



Fig 5.7 Typical calibration curves for lift measurement before and after test


Fig 5.8 A typical output voltage signal in HPvee during calibration process.
145
used to vary plate angle of attack. One of the flat plates with leading and trailing edge
chord variation (case 4 from section 5.2.1) was also studied in wind tunnel 2 using the
force balance system.

5.3 Data analysis
All the test data measured by both UCMT and the force balance system was
processed using standard statistic analysis. The Chauvenet method was used the reject
abnormal data points. Data points where the absolute error (from mean value) exceeded
three times standard deviation were eliminated.
Great care was also given to account for a base circulation value that occurred
when the wind tunnel was running without the flat plates installed. This base circulation
value, which was found to be approximately,
base
= 0.0143 m
2
/s, was subtracted from
all measured circulation values. When measuring this base circulation value, all
experimental parameters were carefully adjusted to match later experiments with the
flat plate present.
Finally, all results were corrected for solid blockage, wake blockage and
streamlined curvature by using techniques presented by Pankhurst & Holder (1952) and
Rae & Pope (1984). For example, the solid blockage was accounted for using
sin ]; 48 / ) ( 1 )[ 2 1 (
2
w
c
C C
lm l
= = . (5.1)
where is solid blockage, c plate chord length, w wind tunnel width, the plate angle
of attack. C
lm
is the raw lift value and C
l
the corrected value. Other correction
expressions for 2D and 3D (finite wings) flows can be found in Rae & Pope (1984). The
146
required blockage correction for the low aspect ratio wings was only on the order of
2.2% because of their small thickness and volume.

5.4 Validation of UCMT measurement
The UCMT measurements were compared with force balance measurements
either from previous investigations or from our LDVT measurement in tunnel 2. In this
section, we present the lift force comparison for a 2D plate (with no chordlength
variation) with the force balance data of Blevin (1984). We will also provide this type
of comparison for the plates with chord length variation and low aspect ratio wings in
later sections. The measured mean circulation values were converted to mean lift values
using Kutta-Joukowski theorem. Smoke flow visualizations are also presented in this
section.
Fig 5.9 presents the validation for a stationary 2D plate with chord length
c=8.89cm in the test section of wind tunnel 1. The lift was converted from the
circulation measured at the center of the plate in the spanwise direction. The lift
comparison is shown in Fig 5.9a. The lift measured with UCMT shows that lift
increases linearly with angle of attack in the range of
0 0
8 0 (streamlined flow
regime). For
0 0
15 8 , the lift decreases with angle of attack, it is in stalled flow
regime. In the regime
0 0
30 15 , the mean lift increases with angle of attack again,
it is in transitional flow regime. After >30
0
(bluff body regime), error bars (based on
standard deviation of the UCMT measurements) abruptly increase, and the mean lift
decreases after it approaches the maximum value at =40
0
. Fig. 5.9b presents smoke
147
(a)
(b) =30
0
(c) =40
0


Fig 5.9 Validation of UCMT measurement for 2D flat plate flow at U=5m/s. (a) Comparison with force
balance data published by Blevin (1984). (b) Smoke visualized near wake structure at =30
0
. (c) Near
wake vortex structure at =40
0
, vortex shedding process can be clearly observed.
148
visualization of the near wake vortex structure behind the plate at =30
0
. A stable
vortex is attached to the rear surface of the plate, and vortex shedding was not observed.
In Fig 5.9c, the vortex shedding process in near wake at =40
0
is clearly shown. We
believe the larger error bars accompanying the mean lift value in the bluff body flow are
due to the effect of shed vortices on circulation measurements. The mean lift measured
with UCMT agrees very well with the data of Blevin (1984), except in the stalled flow
regime. However, the trends of both measurements are the same in this regime, with lift
decreasing with angle of attack. Underprediction of lift values in the stalled flow region
was also observed in Desabrais & Johari (1998), who studied mean lift on 2D airfoils.
Based on the error bars on lift measurement in Fig 5.9, we can conclude that for
flow around a 2D stationary plate, the Kutta-Joukowski theorem can be used to predict
the mean lift in the streamlined and transitional flow regimes. In bluff body flow, the
instantaneous lift values should be predicted with the methods developed in chapter 4.
However, for the mean lift prediction, K-J theorem can be applied (although with larger
error bars) because the unsteady lift fluctuation generated by the periodic vortex
shedding process is averaged to zero during the period of data acquisition over many
shedding cycles.

5.5 Results for flow over plates with varying chord length
In this section, we present the mean lift force distribution along the span of the
plates with chord length variation. The four plates described in section 5.2.1 were used
in UCMT test. Each data point presented in this section is the mean lift averaged from 5
149
separate measurements. Lift distributions for streamlined flow (=6
0
) and transitional
flow (=15
0
) are presented.

5.5.1 Spanwise lift distributions
We next present lift distributions along the flat plate spans (z direction) in Fig
5.10 and 5.12. The local lift L(z) was non-dimensionalized with L
0
,

the measured lift
value for z > z
0
where the chord length is constant. In order to determine whether three-
dimensional flow is induced by curved leading and trailing edges, we also present the
nominal two dimensional lift distribution which was calculated based on local chord
length with

c
z c
L
z L ) ( ) (
0
= . (5.2)
Here c(z) is the local chord length which is a sinusoidal function of the coordinate z, c is
the constant chord length for z > z
0.
If the plate flow remains purely 2D, we would
expect that its distribution would follow Eq. (5.2). As a result deviations of measured
lift distributions for various plates will be a measure of the level of three-dimensionality
in the flow.
In Fig 5.10(a) for the streamlined flow, the deviation of the 3D lift distribution
from the nominal 2D is small. This shows that curved trailing edge generates slight 3D
effect on the spanwise lift distribution. However, in Fig 5.10b, a curved leading edge
yields a larger 3D variation, especially near center span area where the plate chord
length is minimum. Here the deviation between the 2 curves is almost 26%. This
demonstrates that leading edge dominates the flow development in streamlined flow.
150




Fig 5.10 Mean lift distributions along a plate with sinusoidal curved leading or trailing edge in
streamlined regime (=6
0
) at U=5m/s. (a) Trailing edge curved; (b) Leading edge curved; (c) Both
leading and trailing edge curved; (d) Comparison of different cases.

(a)
(b)
(c) (d)
151








(a)
(b)
(c)
Fig 5.11 Smoke visualization of flow at three sections along plate span at =6
0
. Separation bubble
size changes at various spanwise locations. (a) z/z
0
= 1.0. (b) z/z
0
= 0.5. (c) z/z
0
= 0.0
152

Fig 5.10c presents the lift distribution from the plate with both sinusoidal
leading and trailing edge (case 4). the deviation between the curves is the largest among
all cases. The deviation of 3D lift from the nominal 2D results in Fig 5.10a-c serves to
determine the relative contribution of the curved leading or trailing edges in
development of 3D effects in the flow. In Fig 5.10d, we summarize the results from Fig
5.10a-c and include case 1 (constant chordlength plate). The maximum deviation occurs
at the midspan with largest 3D effect for case 4.
A possible physical mechanism for these results could be related to the chord
length variation and separation bubble effects. In general, the plate lift will be larger if
there is a bubble attached on its upper surface, because the bubble effect seems to make
the flow around a cambered airfoil (Tani, 1964). When the bubble bursts and no longer
reattaches, a leading edge stall will occur resulting in lift reduction. This point can be
used to partially explain our results combined with our smoke flow visualization at
=6
0
. Fig. 5.11 presents the flow patterns at three local sections of z/z
0
= 1, 0.5, 0.0
respectively for the plate with combined leading and trailing edges curved. At section
z/z
0
=1, a separation bubble clearly exists on plate upper surface extending from plate
leading edge x/c(z) = 0.0 to a place with x/c(z) = 0.7. At z/z
0
=0.5, the separation bubble
covers the whole plate upper surface and extends beyond the trailing edge, a burst may
happen. At plate middle span section z/z
0
=0.0, the bubble no longer reattaches to the
plate and leading edge stall occurs. These flow observations combined with Tanis
viewpoint could explain the trend of the spanwise variation of lift distribution. Of
153
course, the effect of chord length variation also contributes to the lift variation. A larger
chord length may result in larger circulation and lift. Finally it should be noted that the
complex 3D flow interaction between shear layers at neighboring sections along plate
span may also affect the lift distribution on final results. This is an interesting topic in
the future.
The spanwise lift distribution along these curved plates for typical transitional
flow (=15
0
) is presented in Fig 5.12. In Fig 5.12(a), the deviation of the 3D lift
distribution from the nominal 2D is still small, but it is larger when compared with the
same case for streamlined flow (Fig 5.10(a)). From Fig 5.12(b), the curved leading edge
yields much stronger 3D effect than in streamlined flow (Fig 5.10(b)). The difference
between the 3D lift distribution and the nominal 2D lift increases with the reduction of
chord length, and approaches its maximum value at the minimum chord length section
as 62%, which is a dramatically variation compared with the same case for streamlined
flow as of 26% in Fig 5.10(b). This further demonstrates the leading edge effect on 3D
flow generation. Fig 5.12(c) presents the results for the plate with both sinusoidal
leading and trailing edges. The difference between 3D and 2D nominal lift has the same
trend as that for case 3. The maximum deviation is about 67%, showing that the
combined leading and trailing edges still generate the strongest 3D effects in the flow.
However, this is not a large change compared with case 3, which has a 62% deviation.
Fig 5.12(d) presents the 3D lift distribution of all cases along with the 2D plate (case 1).
The deviation of case 2-4 are much larger compared to case 1 than in the streamlined

154


Fig 5.12 Mean lift distributions along plate with sinusoidal curved leading or trailing edge in a
transitional flow regime (=15
0
) at U=5m/s. (a) Trailing edge curved; (b) Leading edge curved; (c) Both
leading and trailing edge curved; (d) Comparison of different cases.



(a) (b)
(c) (d)
155
flow case. This shows curved leading and trailing edges inducing stronger 3D effects in
transitional flow than in streamlined flow. However, we reach the same conclusion as in
streamlined flow, i.e., the leading edge predominates in creating 3D effects in the flow.
This is further illuminated by the fact that the difference in lift distributions between
case 3 and case 4 is much smaller compared with that in streamlined flow (Fig 5.10(d)).
A possible physical mechanism may relate to the leading edge stall at =15
0
for all
cases. The shear layers have separated from the plate leading edges as observed in
smoke flow visualizations, hence results in smaller lift value compared to the 2D
results.

5.5.2 Comparison between UCMT and force balance results
In order to further validate the UCMT measurements, we also used the force
balance system in tunnel 2 to measure the total lift force on a flat plate (case 4) and
compared it with the integrated lift obtained by integrating the lift distributions (in Fig.
5.10 and 5.12) across the plate span. An independent force measurement is necessary to
confirm that the lift distributions measured with UCMT in section 5.5.1 are valid. The
plate from case 4 (with both sinusoidal curved leading and trailing edges) was selected
for the test, because it induced the strongest 3D effects on the flow as discussed
previously. In previous section, our focus was mainly on the streamlined and
transitional flow regimes, so in force balance measurement, the range of plate angle of
attack was set from =0
0
to =20
0
, which includes the streamlined, stalled and part of
the transitional flow regimes. The free stream airflow speed was U=5m/s as in the
156
UCMT test. For each data point, we repeated the test for 5 times and present the mean
value and error bars based on standard deviation. The UCMT lift distribution was
integrated over plate span to give to the total mean lift from UCMT for comparison to
the LVDT force balance measurements in tunnel 2, with following expression

=
+
=
+
+
+ +
= =
+
N
i
i i
N
i
i i l l
A
A
A
A
l
L
z c z c
z c z c C C
dA
dA z C
C
i i
c
c
0
1
0
1
)] ( ) ( [ 2
)] ( ) ( )[ (
) (
1
0
0
. (5.3)
Where C
L
is the UCMT total lift; C
l
(z) the measured local lift coefficients;
i
l
C the lift coefficient at each local section with chord length ) (z c
i
, setting i=0
corresponds to the middle span of the plate and i=N=12 represents the section with
constant chord length.
An identical plate (case 4) used to determine the lift distribution (Fig 5.10c and
Fig 5.12c) in tunnel 1 was now tested in tunnel 2. Since the test section dimension is
slightly larger in tunnel 2 than tunnel 1, the plate did not extend to the test sectional wall
in tunnel 2. Endplates (Stansby, 1974) were added to the plate in tunnel 2 to eliminate
any wing tip vortex effects.
Fig 5.13 presents the comparison of lift force measured from the UCMT lift
distributions and the force balance system of tunnel 2 for the case 4. The integrated lift
from the UCMT lift distribution shows slightly lower values than the force balance
measurements. In the streamlined flow regime at = 6
0
, the difference is about 7%. In
the transitional flow regime at = 15
0
, the difference increases to almost 12%. The

157









Fig 5.13 Comparison of UCMT lift measurement (in tunnel 1) and force balance
measurement (in tunnel 2) at U=5m/s for a plate with both sinusoidal curved leading and
trailing edge.

158
error may result from the discrete process when using integration and from
measurement error in both the UCMT and force balance system. Another concern is that
if Kutta-Joukowski theorem developed for 2D flow may yield some error in a 3D flow.
Considering all these issues, the 6% deviations are still acceptable. In addition, we also
present the UCMT lift for a 2D plate (Case 1) for comparison. The trend of lift variation
with angle of attack is similar for case 1 (2D plate) and case 4. The difference between
the lift curves for case 1 and case 4 reflects the data in Fig 5.10c and Fig. 5.12c.

5.6 Low Reynolds, low aspect ratio wing study
Given that we have developed methods in section 5.5 to study lift distributions
in 3D flows, in this section, we present UCMT measurement results for low aspect ratio
wings of rectangular planform shape. The wings consist of flat plate with no camber.
These low AR wings may induce 3D flow naturally through the effect of wing tip
vortices. The range of plate aspect ratio varied from AR=1 to AR=3. For such low
aspect ratios, 3D effects are expected to be strong. Given this, we recognize that in
UCMT experiments, we are only measuring bound circulation related to the so-called
linear lift on the wing (Muller, 2001). Lift due to low-pressure regions near the wing
tips created by trailing wing tip vortices is not measured with UCMT. Still, we perform
UCMT experiments similar to that in section 5.5 to study the linear lift, to see if UCMT
can be used to accurately predict lift on low Re, low AR wings applicable to MAV
design. This isolation of the linear lift from the vortex lift may also useful in low AR
studies. We will discuss this in more detail later. Here, we present lift distributions
measured with UCMT and determine the integrated total lift on the plates, then compare
159
our UCMT measurements with force balance measurement results. Our focus is again
on a typical streamlined flow at = 6
0
and a transitional flow at = 15
0
. Circulation
values were again converted using the Kutta-Joukowski theorem, although its
applicability for lift determination in highly 3D flows such as low AR wings must be
verified.
Fig 5.14 presents a typical bound circulation distribution measured with UCMT
along span of a low aspect ratio wing with AR=2.0 and Re=3.6510
4
. Streamlined flow
( = 6
0
) and transitional flow ( = 15
0
) cases are shown. Circulation were measured at 8
sections varying from z/(b/2) =0.0 to 1.0. The results shown here have processed as
described in section 5.2. The uniform spanwise circulation distribution at = 0
0
shows
negligible bound circulation as expected. The circulation distribution for = 6
0
and =
15
0
are similar. The primary characteristic is a maximum (peak) at a spanwise location
near wing tip area. This deviates from the typical monotonic decreasing of lift
(circulation) distributions for high AR wings through Prandtls lift line theory (see Fig
5.15). This phenomenon may result from the strong three-dimensional flow induced
near the tip vortex. It was also found by Bhagwat (2000) who measured the bound
sectional circulation along the span of a wing with AR=9.6 at Re=2.5310
5
using the
Laser Doppler Velocimetry method (see Fig 5.2(b)). We should point out that Bhagwat
(2000) was studying rotating helicopter blades where a resultant freestream velocity
variation existed along the wing span. This may explain his observation of a peak in
circulation along the span of fairly high aspect ration wing. As a result, direct
160








Fig 5.14 Typical circulation distributions along span of a low aspect ratio wing with AR=2
and Re=3.6510
4
, measured with UCMT in streamlined flow ( = 6
0
) and stalled flow ( =
15
0
).
161
comparison to our flow may not be proper. However, the existence of a circulation
distribution peak in another flow is interesting. The spanwise location of the circulation
peak in our work seems to be independent of AOA for = 6
0
and = 15
0
. Final
resolution in UCMT measurements along the span may yield some variation. However,
ratio of
0

peak
(see Fig 5.14) is larger for = 15
0
compared to = 6
0
suggesting
larger induced 3D effects at higher AOA.
Fig 5.15 presents our measured lift distribution for a low aspect ratio wing
(AR=2) at = 6
0
compared with Prandtls lifting line theory (schlicting, 1979) for a
wing with AR=6 (Prandtls lifting line result for AR=2 is not available in literatures to
the best of our knowledge). The sectional lift C
l
is normalized with the mean lift of the
wing C
L
. The difference between the UCMT measurement and results form Prandtls
lifting line theory is dramatic. This is attributed to the strong three-dimensional effect
induced by tip vortex in low aspect ratio wings. Similar results were also found by
Bastedo et al (1985) (see Fig 5.2a) for a low aspect ratio wing with AR=4 and
rectangular planform and in Bhagwat (2000) (Fig 5.2b). These results indicate that
Prandtls lifting line theory is not applicable to low aspect ratio wings due to violation
of the assumption of negligible flow in spanwise direction (Phillips et al, 2000).
The effect of varying aspect ratio on the lift distribution along the span of low
aspect ratio wings is presented in Fig 5.16. The trends are the same for streamlined
flow and transitional flow, here we only present the case for streamlined flow at = 6
0
.
We find that larger aspect ratios generally yield larger local lift across the span
including the local lift value at the peak. The relative location of the peak lift value also
162











Fig 5.15 Comparison of lift distribution from UCMT along span of low aspect ratio
wing (AR=2) with Prandtls lifting line results for a wing (AR=6) (Schlichting, 1979)
163










Fig 5.16 Effect of aspect ratio on lift distributions measured with UCMT on low aspect
ratio wings at = 6
0
.
= 6
0
164
varies with aspect ratio. For larger aspect ratios, the relative distance of the peak away
from the wing tip decreases.
We have described the characteristics of lift distribution on low Re, low AR
wings (measured with UCMT) and compared them to results from Prandtls lifting line
theory. Next, we will study the integration of the lift distributions (Fig 5.16) and
compare these results with force balance measurements, results from other
investigations (Pelletier & Mueller, 2000), and results from Prandtls lifting line theory.
The total lift measured with UCMT was generated from integration of sectional lift
distributions (Fig 5.16) using Eq. (5.3) with N = 8. The lifting line theory result is based
on

)] 1 )(
3 . 57
( 1 [
0
0

+
=
=
Ar
a
a
C
C C
L
L L
. (5.4)
from Anderson (1991), where
0
a = 2(rad
-1
) = 0.1(deg
-1
) is the lift curve slope for two-
dimensional airfoil,
L
C the lift curve slope for finite wing. is a parameter related to
wing planform shape, where typically 25 . 0 0 , following Pelletier & Mueller
(2000), we set = 0.164 for a rectangular plate in our calculations.
We first present our lift data from UCMT measurements and our force balance
measurement in Fig 5.17(a) at = 6
0
. Each data point from balance measurement is the
average of five independent runs. Mean value and error bars based on standard
deviation are shown. The UCMT lift values are less than the force balance values, and
difference between the two values increase slightly with increase of aspect ratio, with a
165

(a)

(b)

Fig 5.17 Comparison of lift coefficient vs aspect ratio in several investigations. (a)
Current UCMT and force balance measurements. (b) Comparison of our force balance
measurements, Pelletier & Mueler (2000)s force balance data and finite wing theory.
166
maximum difference of approximate 12%. The lower lift values for the UCMT
measurements seems reasonable, since the linear lift due to bound circulation is
accounted for in the UCMT data. The data in Fig. 5.17a suggests that integration of
bound circulation from UCMT measurement can provide reasonable estimates of lift on
low Re, low AR wings. The 12% difference between the curves in Fig 5.16a is
acceptable at least for preliminary design work on MAVs.
In Fig 5.17(b) we compare our force balance measurements to data from
Pelletier & Mueller (2000) and Prandtls lifting line (finite wing) theory in order to
check the reliability of our measurements. All these data sets show an increase in lift
coefficient with increasing AR as expected. Results from Pelletier & Mueller (2000)
and the present work generally bracket Prandtls lifting line theory, however, fairly
difference in lift coefficient value is observed. These differences may be attributed in
part to different test conditions. For example, the Reynolds number for our test was
about Re = 410
4
, but Re = 1.410
5
in the work of Pelletier & Mueller (2000). Also in
their work, a wing configuration with only one wingtip exposed to the flow (i.e., a half
aspect ratio configuration) was used which is different from our work. The interaction
between the two wingtip vortices plays an important role in low aspect ratio wings, and
will certainly affect measured lift values. Unfortunately, limited lift measurements on
low Re, low aspect ratio wings exists in the literature. This is an area for future work.



167
5.7 Summary
Two types of three-dimensional flow were studied in this chapter using UCMT
and force balance techniques. First, flow over flat plate (non-uniformity) with
curvedleading or/and trailing edges was studied. Next flow over low Re, low aspect
ratio wing was considered. Focus was on measuring mean lift distributions along the
span of plates leading to total integrated lift force measurement in the streamlined and
transitional flow regimes.
For the flow over the curved flat plates, the measurements of lift distributions
showed that leading edge effect dominates the 3D flow development. Plate with both
curved leading and trailing edges induce the strongest three-dimensionality in the flow.
For streamlined flows, the lift variation in span along the plates was conjectured to be
related to the relative size of the separation bubble, and leading edge stall and
reattachment phenomena. Smoke visualization of the flow was conducted to aid in this
interpretation. Total lift data obtained by integrating the lift distributions was compared
with force balance measurements. A deviation of 6% to 12% was observed and
discussed.
For the low Reynolds, low aspect ratio wing study applicable to MAV design,
local maximum (peaks) were observed in spanwise lift distributions measured with
UCMT. This was attributed to the strong three-dimensionality in flow near wingtips.
Only the bound circulation, or linear lift effect is accounted for in UCMT
measurements. The UCMT technique therefore can serve as a way to isolate the linear
lift effect verse the lift induced by wingtip vortices. This may prove useful in future
168
studies of low AR wings. The relative location and value of the peak lift varies with
aspect ratio. Comparison between total integrated lift from UCMT lift distributions and
force balance measurements showed that UCMT measurements under-predicted lift
slightly as expected since only the linear lift is measured. The qualitative trends from
these different measurements were the same, with larger aspect ratios leading to larger
lift coefficients.
In calculating mean lift forces from UCMT measurements, the Kutta-Joukowski
theorem was used to convert circulation to lift. This may contribute some error in lift
determination since the Kutta-Jouskowski theorem is valid principally for two-
dimensional flows. However, we have found that UCMT techniques can provide
reasonable measured lift distributions and integrated lift values as long as the imposed
three-dimensionality is not too large. Another concern is whether Prandtls lifting line
theory is applicable to low aspect wings because its assumption of negligible flow in
spanwise direction may not hold near the wing tip region. Future investigations are still
needed to further understand these issues.
169
Chapter 6

Conclusions

6.1 Summary of Results

Lift force determination is a key issue in many engineering applications,
especially in understanding nonlinear coupling between structural motions and flow
dynamics in fluid-structure interaction problems. However, deficiencies exist in past
experimental techniques available to measure lift forces, particularly in areas of
unsteady instantaneous lift measurement, and measurement of spatial lift distributions
variation of sectional forces along structural spans.
The major goal in our work is to explore a method that can be used in future
experiments to address the above issues. The developed ultrasonic circulation
measurement technique (UCMT) has shown potential for addressing those issues.
UCMT was developed for lift determination based on the quasi-steady K-J theorem that
properly converts the circulation to lift for steady flow. Hence, it has been mainly used
in the past for mean lift measurement in two-dimensional steady flows around various
aerodynamic bodies, including airfoils and delta wings. The extension of the UCMT
methods to measure time-varying lift forces in unsteady flows has remained largely
unstudied, and UCMT methods have also not been applied to measure lift distributions
along structural spans. The focus of this research is to study extension of UCMT for lift
determination in unsteady and three-dimensional flows.
We used a combined analytical-numerical-experimental approach to fulfill this
major goal. We first developed analytical methods that can properly convert time-
170
dependent circulation (measurable with UCMT) to instantaneous lift forces based on
unsteady potential flow theory. Then we verified the developed correction method
numerically with a finite element solution. Finally, we experimentally used UCMT to
study lift distributions along spans of structures that can induce three-dimensional
flows.
The corrected method was developed based on the work of Katz & Plotkin
(1991), who deduced a general expression to relate unsteady circulation to
instantaneous lift for an unsteady flow around a thin plate.

4 4 3 4 4 2 1
43 42 1
B
c
A
dx t x
t
t u U t L

+ =
0
) , ( ) ( ) ( ) ( . (2.3)
In real experiments, ) , ( t x is very difficult to measure using UCMT. However, the
bound circulation ) (t is attainable. Hence, we developed a corrected unsteady method
that can properly convert measurable unsteady bound circulation to instantaneous lift
through

43 42 1
43 42 1
B
A
dt
t d
c R t u U t L
) (
) ( ) ( ) (

+ = . (2.36)
Here R is defined as an unsteady lift correction factor. The value of R was determined
for several unsteady flow cases; R=3/4 for an oscillating plate; R=0 for a plunging
plate; R=1 for an impulsively started plate. For a pitching plate, R is a complicated
function of plate oscillation frequency, pivot location and freestream velocity. These
results are valid for low amplitude, low frequency motions of thin airfoils. For flow
around a stationary circular cylinder at Re=100, R=0.4 was determined empirically
171
from finite element solutions. Our corrected unsteady method expands the capability of
the UCMT technique for unsteady flow studies may have application in other
experimental methods that can acquire bound circulation such as the PIV techniques.
Our corrected unsteady method was validated numerically using a finite element
solution for flow around an oscillating flat plate at low angle of attack and at Re=100.
Time-dependent bound circulation, based on the instantaneous velocity field integration
along a mimicked closed sound path, was converted to instantaneous lift force using our
developed methods. These results were compared to the true lift forces (from stress
integration). It was shown that our unsteady correction method provided improved
results compared to the quasi-steady K-J approximation for unsteady flow studies.
Another important issue was the effect of variation in the acoustic path on bound
circulation determination, which has not been addressed in previous UCMT studies. We
systematically studied this issue with numerical finite element solutions and found that
proper selection of the enclosing acoustic path is critical for accurate lift measurements.
The information obtained in this work will be useful for closed path design in future
UCMT, PIV or LDV experiments.
Having achieved a better understanding of lift determination in unsteady plate
flows, we next studied another important flow, low Reynolds number vortex shedding
from a stationary bluff body (circular cylinder) using similar methods. We study various
force models including a quasi-steady K-J method, an approximate Lighthills vortex
force model, and a new unsteady model similar to that developed for plate flows. Finite
element simulations for flow over a circular cylinder at Re = 100 were once again
172
performed. Various combinations of force models were studied. An approximate vortex
force model based on an UCMT concept yielded lift predictions differing from the true
lift values. As a result, an alternative model similar to that developed for unsteady flat
plate flows was proposed and studied. This model requires measurement of
instantaneous bound circulation in a region enclosing the cylinder. Some of our results
also will have future impacts on PIV and LDV studies where flow data is integrated
over regions enclosing aerodynamics bodies.
To expand the capability of the UCMT technique for three-dimensional flow
studies, we experimentally studied the mean lift distribution along structures that can
induce three-dimensional flows. Flow around plates with sinusoidal curved leading
or/and trailing edges and low Reynolds, low aspect ratio wings applicable to micro-
aerial-vehicle (MAV) design were studied. The size of the local chord length, size of
separation bubbles and leading edge stall and reattachment phenomena were suggested
to understand the results on the curved plates. Investigation of the lift distribution on the
curved plates, and comparison to force balance results validated the UCMT methods
used.
Study of low Reynolds low aspect ratio wings with the UCMT technique
showed that spanwise lift distributions are dramatically different from those obtained
from Prandtls lifting line theory. UCMT measured results showed that a peak
circulation (lift) value exists at a location near the wing tip. The UCMT measurements
only measure the bound circulation (e.g. linear lift). Lift due to low pressure region
created by wing tip vortices are not included. However, integration of the spanwise lift
173
distributions yielded reasonable (but lower) total lift values compared to independent
force balance measurements.
Based on the deviation between lift values from integrated lift distribution and
force balance measurements, some concern still exist concerning the applicability of the
K-J theorem in the MAV studies. Our results suggest that the K-J theorem is valid as
long as the three-dimensionality in the flow is small (curved plates), but care must be
taken with highly three-dimensional flows such as low Reynolds, low aspect ratio
wings.

6.2 Future Work
Future research in this area would first concentrate on applying what has been
learned in the present study to experimental UCMT studies on unsteady flows. From the
present work, we have not only learned how to properly convert measured unsteady
circulation values to instantaneous lift, but also gained insight into a wide variety of
unsteady flows. These results are valid for low amplitude, low frequency motions of flat
plate airfoils. We have learned that plunging plates and impulsively started plates are
the best candidate flows for UCMT experiments. Plunging plates, for example, require
no-unsteady correction (R=0), and the quasi-steady K-J theorem is applicable. Pitching
airfoils, on the other hand, require complex unsteady corrections ( ) , , , ( t U
c
a
k f R = )
and may be more difficult to study. Systematic determination of R values for other
unsteady flows will be of interest in future work.
174
We have also gained insight into the dependence of lift measurements on the
geometry of acoustic path. As a result, we now can design proper acoustic path in
experiments on unsteady flows and low Re cylinder wakes and possibly re-interpret
previous UCMT results. In future, we would like to systematically study variation of the
allowed region for path selection as a function of Reynolds number, body oscillation
amplitude and reduced frequency, so that the results become a true resource for
guidance of circulation measurement experiments such as UCMT, PIV and LDV. We
also foresee studying combinations of K-J theorem, unsteady correction and vortex flow
models in UCMT experiments on cylinder wakes.
In three-dimensional flows, we plan to focus on low Reynolds, low aspect ratio
wings for MAV studies. We would like to better understand the observed circulation
distribution peaks, while also measuring trailing wing tip vortex strength (circulation)
using UCMT. All parameters may depend on the wing tip chord-length, freestream
velocity, angle of attack, and airfoil span at extreme low aspect ratios.
On the numerical side, we would like to repeat the simulation of chapter 3 and 4
at higher Reynolds numbers. Simulations on oscillating plates at Re=10
3
, 10
4
would
confirm our unsteady correction methods at Reynolds numbers of many engineering
applications. Simulations on a pitching flat plate could also be conducted to verify and
better understand the complex unsteady correction factor R in this flow. The application
of the cylinder wake work to data from numerical simulation of freely vibrating cables
(NEKTAR code of G. Karniadakis) is also a possibility.
175
For analytical studies, confirming the empirically determined corrected factor,
R=0.4 for cylinder flow, remains as a challenge. In the development of our correction
method, higher order time derivatives of bound circulation were not included in Eq.
(2.36), which only expressed the quasi-steady and one unsteady term. Those high order
terms may improve our results for body oscillations with higher frequency and
amplitude. The methods of chapter 2 could also be extended to develop unsteady
corrections for pitching moments in unsteady aerodynamic flows.


176
Appendix A

Derivation of Unsteady Correction Factor, R

(1) General plate motion
The unsteady correction coefficient is defined in chapter 2 as

) 2 (
B
B
R = . (2.35)

=
c
dx t x
t
B
0
) , ( . (A.1)

=
c
dx t
t
B
0
) 2 (
) ( . (A.2)

The integrated circulation ) , ( t x is expressed as

=
x
ds t s t x
0
) , ( ) , ( . (2.4)

Based on the classic approach of Glauert and with a standard transform ) cos 1 (
2
=
c
x ,
) , ( t s can be expressed as a chordwise trigonometric expansion at any time instant as

=
+
+
=
1
)] sin( ) (
sin
cos 1
) ( )[ ( 2 ) , (
n
n o
n t E t E t U t

. (2.5)

Hence, the derivative of ) , ( t x will be


)
`

+ +

=
)
`

+
+

=
1
0 0 0
0
0
0
0
0
0 0
1
0
0
0
0
0 0 0
0
) sin( ) sin( ) ( ) cos 1 )( ( )[ (
sin
2
)] sin( ) (
sin
cos 1
) ( [ ) ( 2
sin
2
) , ( ) , ( ) , (
n
n o
n
n o
x
d n t E d t E t cU
t
d
c
n t E t E t U
t
d
c
t
t
ds t s
t
t x
t



(A.3)
177
With the aid of following integrals


4
2 sin
2
sin
0
0
0
2

d . (A.4)


) 1 ( 2
) 1 sin(
) 1 ( 2
) 1 sin(
) sin( ) sin(
0
0
0 0
+
+

n
n
n
n
d n

. (A.5)

and considering U(t)=U=constant in our work, (A.3) will be

)
`

+
+

+ +

=2
1 0
]
) 1 ( 2
) 1 sin(
) 1 ( 2
) 1 sin(
)[ (
)
4
2 sin
2
)( ( ) sin )( ( ) , (
n
n
n
n
n
n
t E
t E t E
t
cU t x
t



. (A.6)

For bound circulation = = = c x t c t ) , ( ) ( , substitute into (A.6), yields

]
2
) (
) ( [ ) (
1
0
t E
t E
t
cU t
t
+

. (A.7)

Substitute (A.6) into (A.1), yields

)] (
2
1
) ( ) ( 3 [
4
1
sin
2
]
) 1 ( 2
) 1 sin(
) 1 ( 2
) 1 sin(
)[ (
)
4
2 sin
2
)( ( ) sin )( ( ) , (
2 1 0
2
2
0
1 0
0
t E t E t E
t
U c
d
c
n
n
n
n
t E
t E t E
t
cU dx t x
t
B
n
n
c
+ +

=
)
`

+
+

+ +

=






. (A.8)

Substitute (A.7) into (A.2), yields


]
2
) (
) ( [
sin
2
]
2
) (
) ( [ ) (
1
0
2
0
1
0
0
) 2 (
t E
t E
t
U c
d
c t E
t E
t
cU dx t
t
B
c
+

=
+

=




. (A.9)

Substitute (A.8), (A.9) into (2.35), yields
178

]
2
) (
) ( [
)] (
6
1
) (
3
1
) ( [
4
3
1
0
2 1 0
) 2 (
t E
t E
t
t E t E t E
t
B
B
R
+

+ +

= = . (A.10)

From eq. (2.7), it gives

=
=
=

0
2
0
1
0
) 2 cos(
) (
) , ( 2
) (
) cos(
) (
) , ( 2
) (
,
) (
) , ( 1
) (
d
t U
t x W
t E
d
t U
t x W
t E
d
t U
t x W
t E
o
. (A.11)

Local downwash is given in eq. (2.8) as


t z x
U t x W
W

=

) , ( . (2.8)

For the motion studied in this work, plate shape function is given in eq. (2.10)

) )( ( ) ( ) ( a x t t h t = . (2.10)

Substitute (2.10) into (2.8), yields

) )( ( ) ( ), ( a x t t h
t
t
x
=

&
&
. (A.12)

Substitute (A.12) in (2.8), yields


t
a x t t h t U t x W
W


+ = ) )( ( ) ( ) ( ) , ( &
&
. (A.13)

The wake potential
W
is a complex function of the plate motion history. From the work
of Theodorsen (1935) and Von Karman & Sears(1938), effect of
W
can be corrected by
using a lift deficiency factor C(k) (see Appendix B) to the quasi-steady term. Hence, here
it can be temporarily neglected. (A.13) will be
179
) )( ( ) ( ) ( ) , ( a x t t h t U t x W + = &
&
. (A.14)

Substitute (A.14) in (A.11), and notice ) cos 1 (
2
=
c
x , yields

=
=
+ =
0 ) (
,
2
) (
)],
2
1 (
2
) ( ) ( [
1
) (
2
1
t E
U
c
t E
c
a c
t h t U
U
t E
o


&
&
&
. (A.15)

Substitute (A.15) in (A.10), yields




& &
& &
&
& &
& &
&
)
4
3
(
)
3
2
(
4
3
) 2 (
a
c
h U
a
c
h U
B
B
R
+
+
= = . (A.16)


(2) Application to typical unsteady flows

(2.1) Oscillating plate

For an oscillating plate, eq. (2.13) gives

) t sin( h h h ; t tan cons
a 0
+ = = . (2.13)

Substitute it in (A.16), yields


4
3
) 2 (
= =
B
B
R . (2.17)


(2.2) Plunging plate

For a plunging plate, its motion is given by eq. (2.38)

180

=
<

=
=
c
c
c
t t h t h
t t t
t
h h
h t h
const
, ) (
0 , ) ( ) (
.
1
1 0
0

. (2.38)

yields 0 , 0 = = = h
& &
& & & , substitute in (A.16), get

0 = R . (2.39)


(2.3) Pitching plate

For a pitching plate, the plate motion is given as

t tan cons h h ); t sin(
0 a o
= = + = . (2.40)

yields
) sin( ), cos( , 0
2
t t h
a a
= = = & & &
& &
. (A.17)

Substitute (A.17) in (A.16), yields


) sin( )
4
3
( ) cos(
) sin( )
3
2
( ) cos(
4
3
) 2 (
t a
c
t U
t a
c
t U
B
B
R




= = . (A.18)
Substitute
c
Uk 2
= in (A.18), yields


)
2
sin( )
4
3
( 2 )
2
cos(
)
2
sin( )
3
2
( 2 )
2
cos(
4
3
) 2 (
t
c
Uk
k
c
a
t
c
Uk
t
c
Uk
k
c
a
t
c
Uk
B
B
R


= = . (A.19)









181
Appendix B

Lift Deficiency Factor, C(k)

Lift deficiency factor is also called Theodorsens function (Fung, 1993), it is used
to correct the lift obtained from a mathematical manipulation in which the wake effect is
temporarily neglected due to the complex mathematical difficulty. It is a function of
reduced frequency and its exact expression is

) ( ) (
) (
) ( ) ( ) (
1 0
1 '
ik K ik K
ik K
k iG k F k C
+
= + = . (B.1)
Where K(ik) are modified Bessel functions. Obviously (B.1) is a complex
mathematical expression. An approximate expression of C(k) can be expressed as

=
5 . 0 ;
32 . 0
1
335 . 0
041 . 0
1
165 . 0
1
5 . 0 ;
3 . 0
1
335 . 0
0455 . 0
1
165 . 0
1
) (
'
k
i
k
i
k
k
i
k
i
k
k C & . (B.2)
After some mathematical manipulation, its module can be expressed as

|
|
|
|
.
|

\
|
+
+
+
+
|
|
|
|
.
|

\
|
+

|
|
|
|
.
|

\
|
+
+
+
+
|
|
|
|
.
|

\
|
+

=
5 . 0 ;
)
32 . 0
1 (
1072 . 0
)
041 . 0
1 (
0068 . 0
)
32 . 0
1 (
335 . 0
)
041 . 0
1 (
165 . 0
1
5 . 0 ;
)
3 . 0
1 (
1005 . 0
)
0455 . 0
1 (
0075 . 0
)
3 . 0
1 (
335 . 0
)
0455 . 0
1 (
165 . 0
1
) (
2
2 2
2
2 2
2
2 2
2
2 2
k
k
k
k
k
k k
k
k
k
k
k
k k
k C . (B.3)
(B.3) was used in our work in chapter 2 (equation (2.12)).
182
References


Anderson, D. (1991), Fundamental of Aerodynamics, 2
nd
ed., McGraw-Hill, New York,
343.

Anderson, D. (1997), A History of Aerodynamics and Its Impact on Flying Machines,
Cambridge Aerospace Series 8, Cambridge University Press, London, England.

Bastedo, et al (1985), Performance of finite wings at low Reynolds numbers,
Proceeding of the Conference on Low Reynolds Number Airfoil Aerodynamics, edited
by Muller, T.J, University of Notre Dame, Notre Dame, IN, 195-205.

Bearman, P.W.(1984), Vortex shedding from oscillating bluff bodies, Ann. Rev.
Fluid Mech., 16, 195-222.

Bearman, P.W., Graham, J.M.R., & Obasaju, E.D. (1984) A model equation for the
transverse forces on cylinders in oscillatory flows. Appl. Ocean Res. 6, 166-172

Bearman, P.W. & Luo, S.C. (1988), Investigation of the aerodynamic instability of a
square-section cylinder by forced oscillation, Journal of Fluids and Structures, 2, 161-
176.

Bhagwat, M.J. & Leishman, J.G. (2000), Measurement of bound and wake circulation
on a helicopter rotor, Journal of Aircraft, 37 (2), 227-234.

Blevins, R.D. (1984), Flow-induced Vibrations, New York, Van Nortranl Reinhold.

Blevins, R.D. (1988), Applied Fluid Dynamics Handbook, New York, Von Nortranl
Reinhold, 322.

Cetiner, O. & Rockwell, D. (2001), Streamwise oscillation of a cylinder in a steady
current. Part 1. Locked-on states of vortex formation and loading, Journal of Fluid
Mechanics, 427, 1-28.

Cetiner, O. & Rockwell, D. (2001), Streamwise oscillation of a cylinder in a steady
current. Part 2. Free-surface effects on vortex formation and loading, Journal of Fluid
Mechanics, 427, 29-59.

Desabrais, K.J. and Johari, H. (1998), "Direct Circulation Measurement of a Wing Tip
Vortex Using Ultrasound", AIAA paper 98-0609.


183
Fung, Y. C. (1993), An Introduction to the Theory of Aeroelasticity, Dover Publication
Inc., New York.

Gad-el-Hak, M. (2001), Micro-Air-Vehicles: Can they be controlled better? Journal
of Aircraft, 38(3), 419-429.

Giacomelli, R. & Pistolesi, E. (1934), Historical Sketch In Aerodynamic Theor, Vol.1,
ed. Durand, W.F. Berlin, Springer.

Glauert, H. (1959), The Elements of Aerofoil and Airscrew Theory, 2
nd
ed., Cambridge
University Press.

Govardhan, R. & Williamson, C.H.K. (2000), Modes of vortex formation and
frequency response of a freely vibrating cylinder, Journal of Fluid Mechanics, 420, 85-
130.

Helmholtz, H. L. (1858), On the integrals of the hydrodynamic equations
corresponding to vortex motions, Crelles Journal fur die reine und angewandte
Mathematik, 60, 23-55.

Hover, F. S., Techet, A. H. & Triantafyllou, M. S. (1998), Forces on oscillating
uniform and tapered cylinder in crossflow, Journal of Fluid Mechanics, 363, 97-114.

Ikeda, S. & Yamamoto, Y. (1981), Lift force on cylinders in oscillating flows, Report
of Department Foundation of Engineering and Coastal Engineering, Saitama University,
Japan, 10, 1-16.

Iwan, W.D. (1975), The vortex induced oscillation of elastic structural elements,
Journal of Engineering for Industry, Transaction of ASME, 1378-1382.

Johari, H. & Durgin, W. W. (1998), Direct measurement of circulation using
ultrasound, Experiments in Fluids, 25, 1-10.

Joukowski, N.E., 1906, Sur les Tourbillons Adjionts, Traraux la Section Physique de
la Societe Imperiale des Science Naturales, Vol.13, No. 2.

Kutta, M.W., 1902, Auftriebskrafte in Stromenden Flussigkeiten, Illustrierte
Aeronautische Mitteilungenm Vol. 6, p133.

Katz, J. & Plotkin, A. (1991), Low-speed Aerodynamics, from wing theory to panel
methods, McGraw-Hill series in Aeronautical and Aerospace Engineering, McGraw-
Hill, Inc.

184
Lancaster, F.W. (1926), Sustentation in flight, Journal of Royal Aeronautical Society,
30, 587-606.

Lasheras, J.C. & Meiburg, E. (1990), Three-dimensional vorticity modes in the wake
of a flat plate, Phys. Fluids A, 2, 371-379.

Lighthill, J. (1986), Fundamentals concerning wave loading on offshore structures,
Journal of Fluid Mechanics, 173, 667-681.

Lin, J.-C. & Rockwell, D. (1996), Force identification by vorticity fields: Techniques
based on flow imaging, Journal of Fluids and Structures, 10, 663-668.

Lin, J.-C. & Rockwell, D. (1997), Quantitative interpretation of vortices from a
cylinder oscillating in quiescent fluid, Exps. Fluids, 23, 99-104.

Lin, J-C & Rockwell, D. (1999), Horizontal oscillations of a cylinder beneath a free
surface: Vortex formation and loading, Journal of Fluid Mechanics, 389, 1-26.

Linn, A. (1999), Determination of Average Lift of a Rapidly Pitching Airfoil, M. S.
Thesis, Worcester Polytechnic Institute.

Lohner, R. et al (1984), The solution of non-linear hyperbolic equation systems by the
finite element method, International Journal of Numerical Methods in Fluids, 4, 1043-
1052.

Malamataris, N.G. (1991), Computer-aided analysis of flows on moving and unbounded
domains: Phase-chage fronts and liquid leveling, PhD thesis, The University of
Michigan.

Maull, D.J. & Milliner, M.C. (1978), Sinusoidal flow past a circular cylinder, Coastal
Engineering, 2, 149-168.

Meiburg, E. & Lasheras, J.C. (1988), Experimental and numerical investigation of the
three-dimensional transition in plane wakes, Journal of Fluid Mechanics, 190, 1-37.

Moreira, J. & Johari, H. (1997), Direct measurement of delta wing vortex circulation
using ultrasound, AIAA paper 97-0743.

Newman, D. & Karniadakis G. E. (1996), Simulations of flow over a flexible cable: A
comparison of forced and flow-induced vibration, Journal of Fluids and Structures, 10,
439-453.

185
Noca, F., Shiels, D. & Jeon, D. (1997), Measuring instantaneous fluid dynamic forces
on bodies, using only velocity fields and their derivatives, Journal of Fluids and
Structures, 11, 345-350.

Nuzzi, F. et al (1992), Three dimensional vortex formation from an oscillating non-
uniform cylinder , Journal of Fluid Mechanics, 238, 31-54.

Obasaju, E. D., Bearman, P. W. & Graham, J.M.R. (1988), A study of forces,
circulation and vortex patterns around a circular cylinder in oscillating flow, Journal of
Fluid Mechanics, 196, 467-494.

Olinger, D. J. & Sreenivasan, K.R. (1988), Nonlinear dynamics of the wake of an
oscillating cylinder, Physical Review Letters , 60, 797-800.

Olinger, D.J., Alexandrou, A.N. and Dutka, J. (1993), Vortex shedding lock-on in a
periodic freestream flow past a circular cylinder, ASME Forum on Unsteady Flows,
ed. Keith W. L. and Wei T. 157, 183-188.

Olinger, D.J., Alexandrou, A.N. and Keller, D.R. (1995), Non-harmonic forcing of a
cylinder wake by a periodic freestream flow, Proceeding of the 1995 Symposium on
Fluid-Structure Interaction, ed. Coke, T.C. & Williams, D.R., 1995 ASME International
Mechanical Engineering Congress & Exposition, San Francisco, CA, 41-52.

Olinger, D.J. (1998), A low-order model for vortex shedding patterns behind vibrating
flexible cables, Physics of Fluids, 10 (8), 1953-1961.

Pankhurst, R.C. & Holder, D.W. (1952), Wind Tunnel Technique, Pitman, London,
327-427.

Parkinson, G.V. (1974), Mathematical Models of Flow Induced Vibrations, ed. E.
Naudascher, 81-127, Berlin, Springer.

Pelletier, A. & Mueller, T. (2000), Low Reynolds number aerodynamics of low aspect
ration thin/flat/cambered plate wings, Journal of Aircraft, 37 (5), 825-832.

Popp, K. (2000),Experimental investigation of Reynolds number and scale effects on
parachute inflation, M.S. Thesis, Worcester Polytechnic Institute, Worcester, MA.

Prandtl, L. (1918), Tragflugel Theorie, Nachrixhten von der Gesellschaft der
Wisseschaften zu Gottingen, Geschaeftliche Mitteilungen, Klasse. 451-477.

Prandtl, L. (1921), Applications of modern hydrodynamics to aeronautics, NACA
116.

186
Pritchard, J. Laurence (1958), Francis Herbert Wenham, Honorary Member, 1824-
1908; An Appreciation of the First Lecture to the Aeronautical Society, Journal of
Royal Aeronautical Society, 62, 571-596

Purutyan, H. (1990), Ultrasonic Measurement of Circulation Around a Plunging Airfoil,
M. S. Thesis, Worcester Polytechnic Institute.

Rae, W. H. & Pope, A (1984), Low-speed wind tunnel testing, 2
nd
ed., Wiley, New
York, 176-344.

Ramberg, S.E. & Griffin, O.M. (1976), The effects of vortex coherence, spacing and
circulation on flow-induced forces on vibrating cables and bluff structures, Naval
Research Laboratory 7945.

Sarpkaya, T. (1975), An inviscid model of two-dimensional vortex shedding for
transient and asymptotically steady separated flow over an inclined plate, Journal of
Fluid Mechanics, 68, 109-128.

Sarpkaya, T. (1979), Vortex induced oscillation, Journal of Applied Mechanics, 46,
241-258.

Sarpkaya, T. & Isaacson, M. (1981) Mechanics of Wave Forces on Offshore Structures,
Van Nostrand Reinhold Co.

Sarpkaya, T. (1986), Force on a circular cylinder in viscous oscillatory flow at low
Keulegan-Carpenter numbers, Journal of Fluid Mech. 165, 61-71

Schlichting, H. (1979) Aerodynamics of Airplane, New York : McGraw-Hill

Schmidt, D.W. & Tilmann, P. M. (1970), Experimental study of sound wave phase
fluctuations caused by turbulent wakes, Journal of Acoustic Society, 47, 1301-1324.

Schmidt, D.W. (1975), Acoustic methods for fast detection and measurement of
vortices in wind tunnel, ICIASF-75 Record, 216-228.

Smith, R. H., Durgin, W. W. & Johari, H. (1995), The direct measurement of
circulation in free surface vortices, AIAA paper 95-0104.

Stansby, P.K. (1974), The effects of end plates on the base pressure coefficient of a
circular cylinder, Aeronautical Journal,78, 36-37.

Surry, D. & Stathopoulos, T. (1997), An experimental approach to the economical
measurements of spatially averaged wind loads, Journal of Ind. Aerodynamics, 2, 385-
397.
187

Szepessy, S. & Bearman, P.W. (1993), Analysis of a pressure averaging device for
measuring aerodynamic forces on a circular cylinder, Experiments in Fluids, 16, 120-
128.

Tani, I. (1964), Low-speed flows involving bubble separations, Progress in
Aeronautical Science, 5, ed. Kuchemann D. & Sterne, L.H.G., Pergamon, New York,
70-103.

Theodorson, T. (1935), General theory of aerodynamics instability and the mechanism
of flutter, NACA Rep. 496.

Triantafyllou, G.S.(1990), Three dimensional flow patterns in two dimensional wake,
Proceeding of the ASME Symposium on Unsteady Flows, University of Toronto,
Toronto, Canada, 395-.

Unal, M.F., Lin, J-C & Rockwell, D. (1997), Force prediction by PIV imaging: A
momentum-based approach, Journal of Fluids and Structures, 11, 965-971.

Van Atta, C.W. & Gharib, M. (1987), Ordered and chaotic vortex streets behind
circular cylinders at low Reynolds numbers, Journal of Fluid Mechanics, 174, 113-
133.

Von Karman, T. & Sears, W.R. (1938), Airfoil theory for non-uniform motion,
Journal of Aeronautical Science, 5 (10), 379-390

Wagner, H., (1925), ber die Entstehng des Dynamischen Aftriebes von
Tragflgeln, Zeitschrift fr Angewandte Mathematik und Mechanik, 5 (1), 17-35.

Walker, P. (1990), The forces on a circular cylinder in oscillatory flow plus a current,
Ph.D. Thesis, Department of Aeronautics, Imperial College of Science and Technology,
London, England.

Weber, F. J., Durgin, W. W. & Johari, H. (1995), Direct circulation measurements
about a rapidly pitching airfoil using an ultrasonic system, AIAA paper 95-2269.

Williamson, C.H.K. (1996), Vortex dynamics in the cylinder wake, Ann. Rev., Fluid
Mechanics, 28, 477-.

Wu, J.C. (1981), Theory for aerodynamic force and momentum in viscous flows,
AIAA Journal, 19(4), 432-441.

Wu, J. Z. et al. (1998), Post-stall flow control on an airfoil by local unsteady forcing,
Journal of Fluid Mechanics, 371, 21-58.
188

Yuan, J. K. et al (1994), "The Corrected Investigation of Wall Interference on the Flow
Field around Cylinders", ACTA AERODYNAMIC SINICA, Vol. 12, No.2, pp219-224,
1994.

Yuan, J.K. & Olinger, D.J.(1999), An ultrasonic measurement technique for spatial lift
distributions in bluff body flows, Proceeding of the 13
th
ACSE Engineering
Mechanics, Fluid-Structure Interaction Sessions, Paper W1H-3, pp375. John Hopkins
University, Baltimore, MD.

Yuan, J.K. & Olinger, D.J. (2002), Circulation methods for unsteady aerodynamics
flows, to appear in 20
th
AIAA Applied Aerodynamics Conference, Sessions of
Unsteady Aerodynamics; St. Louis, June 24-27, AIAA-2002-3057.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy