1993 Hydrodynamics of Oscillating Cylinder and Disks at Low Keulegan-Carpenter Number

MICROFILMED 1993
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may
be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the

copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely afreet reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by

sectioning the original, beginning at the upper left-hand corner and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in
reduced form at the back of the book.
Photographs included in the original manuscript have been reproduced

xerographically in this copy. Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly
to order.
University Microfilms International

A Bell & Howell Information C o m p a n y
3 0 0 N orth Z e e b Road. Ann Arbor. Ml 4 8 1 0 6 -1 3 4 6 USA
3 1 3 /7 6 1 -4 7 0 0 8 0 0 /5 2 1 -0 6 0 0
O rder N u m b er 9332176
H ydrodynam ics o f oscillating cylinders and disks at low

K eulegan-C arpenter numbers
Thiagarajan, Krishna Prasad, Ph.D.

The University of Michigan, 1993
UMI
300 N. Zeeb Rd.
Ann Arbor, MI 48106
HYDRODYNAMICS OF OSCILLATING CYLINDERS AND DISKS AT
LOW KEULEGAN-CARPENTER NUMBERS
by
Krishna Prasad Thiagarajan
A dissertation submitted in partial fulfillment

of the requirements for the degree of
Doctor of Philosophy
(Naval Architecture and Marine Engineering)
in The University of Michigan
1993
Doctoral Committee:
Associate Professor Annin W. Troesch, Chairman
Professor Robert F. Beck
Associate Professor Luis P. Bernal
Assistant Professor David T. Walker
RULES REGARDING THE USE OF
MICROFILMED DISSERTATIONS
Microfilmed or bound copies of doctoral dissertations sub

m itted to The U niversity of Michigan and m ade available through
University Microfilms International or The University of Michigan are
open for inspection, but they are to be used only with due regard for the
rights of the author. Extensive copying of the dissertation or publication
of m aterial in excess of standard copyright lim its, w hether or not the
dissertation has been copyrighted, m ust have been approved by the
author as well as by the Dean of the Graduate School. Proper credit m ust
be given to the author if any material from the dissertation is used in
subsequent w ritten or published work.
© Krishna Prasad Thiagarajan 1993
All Rights Reserved
To
Gajalakshmi
Life is nothing but a journey..

And you are my companion.
Together, we'll take a long walk,
A wild ride into the brightfuture!
ACKNOWLEDGMENTS
When I first knocked on the doors of the Department of Naval Architecture and
Marine Engineering, three people, after some hesitation and negotiation, let me in:
Professors Michael M. Bemitsas, Michael G. Parsons, and Annin W. Troesch. I owe my
thanks to them for that and several things that followed later.
The research work that formed the basis of this thesis was supported by the
University of Michigan / Michigan Sea Grant / Industry Consortium in Offshore
Engineering, projects number R/T-23 and R/T-29, under Grant number NA85AA-D-
SG045C from the Office of Sea Grant, National Oceanic and Atmospheric Administration
(NOAA), U.S. Department of Commerce and funds from the State of Michigan. Industry
participants include the American Bureau of Shipping; ARCO Gas and Oil Company;
Conoco Inc.; Exxon Production Research; Mobil Research and Development; Shell
Companies Foundation; and the U.S. Coast Guard. Funding for the instrumentation of the
Particle Image Velocimetry setup used during the experiments was provided by the
University Research Initiative, Office of Naval Research, Contract number N00014-86K-
0684. To these sources, several thanks for providing me with the financial assistance.
My research advisor, Professor Armin W. Troesch has been my main source of
inspiration and guidance. The experimental force measurements were conducted very
efficiently and successfully under his careful supervision. Professor Troesch’s ways of
dealing with students is very well known. His compassion, friendly and subtle humor
have kept me going all along. Whenever I went astray, he was always there to remind me
of my direction. Few words can express my gratitude to him for giving me this grand
opportunity to work at UM. I can only hope that I could emulate him sometime in my life.
The hot research topic of Particle Image Velocimetry (PIV) was first introduced to
me by Professor David T. Walker, who had been instrumental in directing the experimental
flow visualization study. I admired Professor Walker for his clarity in thoughts and being
well-versed in several experimental details. Further, the trust and confidence he placed in
me has gone a long way during my work. To him, a special thanks.
I would like to extend my appreciation to Professors Robert F. Beck and Luis P.
Bernal for consenting to be in my Doctoral Committee. Through the discussions I had with
them, I got new ideas which helped shape up my work.
A veiy rewarding experience during my Michigan years was the knowledge gained
through classroom work. Some of the teachers whom I had the benefit of learning from
include: Professors D. T. Walker, R. F. Beck, M. Perlin and M. M. Bemitsas at Naval
Architecture, Professors W. W. Shultz, G. J. Brereton and G. Tryggvason at Mechanical
Engineering, and Professors P. Federbush and R. Pego at Mathematics.
The assistance provided by the staff of the Ship Hydrodynamics Lab was
invaluable while conducting the experimental work. Many thanks go to my friends Rick
Ashcroft, Stuart Cohen, Bob Donaldson, Guenther Kellner, Bill Kirkpatrick, Joe Krasny
and Eric Wright. I would like to acknowledge the help of John Pyhtila during analysis of
data from force measurements. My good man Ryan Jensen helped me out in processing
many flow visualization pictures.
The office staff of NA&ME had always been around whenever I needed help in
adminstrative paperwork. I would like to gratefully acknowledge the support and
friendship of Virginia Konz. Special thanks to Lisa Payton who offered a lot of assistance
during my application process, and for the help and friendship during the first two years.
Thanks axe also due to the graduate secretaries Mary Gibbons and Colleen Vogler. The
friendly smiles of Denise Jones, Karen Trevino, Luella Miller and Cora Wheatley will be
fondly remembered.
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
My entire stay in Ann Arbor would have gone sour had it not been for some of my
wonderful friends Slim Aissi, Amid Ansari, John Cameron, Yusong Cao, Laurent
Foulhoux, Luis Garza-Rios, Brian Haus, John Hicks, Dave Kraemer, Bill Krol, Canhai
Lai, Terry Lee, Allan Magee, Jose Marin, Jong-Hwan Park, Paul Stewart, and Xin Sun.
During those long hours of work in the laboratory, I recall the company of Dave Kraemer,
Chou-Yi Chen, Mary Wise and Chao-Lung Ting.
I would have gotten nowhere without the lively friendship and support of my dear
wife Gaja. My family in Madras, India had been my moral and emotional support and kept
me going through their numerous letters and phone calls.
It has been a pleasure living in this friendly city of Ann Arbor, and in the University
Housing community. It was a dream for me to come to Michigan, one which I have
realized. It is thrilling to think that the name of Michigan will stand by me for the rest of
my life.
K.P.T.
IV
TABLE OF CONTENTS
ACKNOWLEDGMENTS........................................................................................... ii
LIST OF FIGURES.............................................................................................vi
LIST OF TABLES...............................................................................................ix
LIST OF APPENDICES..................................................................................... x
LIST OF SYMBOLS...........................................................................................xi
CHAPTER
I I N T R O D U C T I O N ..............................................................................1
II CYLINDER IN OSCILLATORY CROSS FLOW: REVIEW
AND DISCUSSION............................................................................7
2.1 Inline force on a cylinder 9
2.2 Effect of currents 13
IH AXIAL OSCILLATIONS OF CYLINDERS AND DISKS:
FLOW VISUALIZATION STUDIES....................................... 17
3.1 Introduction 17
3.2 Flow visualization methods 19
3.3 The basic concept 23
3.4 Experimental Setup 27
3.5 Results and discussion 30
IV AXIAL OSCILLATIONS OF CYLINDERS AND DISKS:
FORCE MEASUREMENTS.............................................................60
4.1 Introduction 60
4.2 Description of experimental setup 67
4.3 Results and Discussion 70
V SCALING LAWS AND AN EXAMPLE CALCULATION..........93
VI SUMMARY AND CONCLUSIONS............................................... 100
APPENDICES......................................................................................................... 105
B I B L I O G R A P H Y ............................................................................................. 128
LIST OF FIGURES
Figure
1.1 Schematic diagram of a Tension Leg Platform................................... 5
1.2 Component problems of a TLP.................................................................. 6
2.1 Drag and inertia coefficients vs. KC of (a) a circular cylinder and (b) a
square cylinder from Troesch and Kim (1991); /J=48600..................... 15
2.2 Drag coefficient vs. P for circular cylinders in oscillatory cross flow;
KC=0.5................................................................................................16
3.1 Particle image velocimetry concept; (a) single exposure and (b) double
exposure................................................................................................ 39
3.2 Root mean squared error in velocity vs. frequency................................... 40
3.3 Schematic of the Digital Imaging Velocity Measurement System.............41
3.4 Flow due to the test cylinder in axial oscillations; .4=6.35 mm,/M).5
Hz; (a)-(h) equally spaced over one cycle starting from top dead
center................................................................................................... 42
3.5 Flow due to the test cylinder with disk A in axial oscillations; A=6.35
mm,f=03 Hz; (a)-(h) equally spaced over one cycle starting from top
dead center............................................................................................ 43
3.6 How due to disk A in axial oscillations; 4=6.35 mm,/=0.3 Hz; (a)-
(h) equally spaced over one cycle starting from top dead center................ 44
3.7 Velocity plots obtained by single exposure PIV for the oscillatory flow
due to disk A; 4=6.35 mm,./=0.1 Hz; (a)-(h) equally spaced over one
cycle starting from top dead center; x and y distances in mm................... 45
3.8 Vorticity contours obtained from the velocity plots of Figure 3.7;
vorticity magnitudes in sec1, x and y distances in mm............................ 46
3.9 How due to disk A in axial oscillations; 4=12.70 mm,y=0.5 Hz; (a)-
(h) equally spaced over one cycle starting from top dead center................47
3.10 Velocity plots obtained by double exposure PIV for the oscillatory
flow due to disk A; 4=12.70 mm,y=0.21 Hz; (a)-(h) equally spaced
over one cycle starting from top dead center, x and y distances in mm...... 48
VI
vorticity magnitudes in sec*1, x and y distances in mm............................49
3.12 Evolution of oscillatory flow due to disk B; A=3.2 mm,/=0.5 Hz,
initial condition=BDC; (a)-(x) equally spaced over five cycles of
oscillation, starting from rest......................................................... 50
3.13 Flow due to disk B in axial oscillations; A=6.35 mm,/=0.3 Hz, initial
condition=BDC; (a)-(h) equally spaced over one cycle starting from
bottom dead center................................................................................. 53
3.14 Velocity plots obtained by single exposure PIV for the oscillatory flow
due to disk B; A=12.70 mm,/=0.11 Hz, initial condition=BDC; (a)-
(h) equally spaced over one cycle starting from bottom dead center; x
and y distances in mm............................................................................54
vorticity magnitudes in sec*1, x and y distances in mm............................55
3.16 Flow due to disk B in axial oscillations; A=6.35 mm, 7=0.3 Hz; (a)-
(h) equally spaced over one cycle starting from top dead center............... 56
3.17 Velocity plots obtained by double exposure PIV for the oscillatory
flow due to disk B; A=12.70 mm, 7=0.20 Hz initial condition=TDC;
(a)-(h) equally spaced over one cycle starting from top dead center; x
and y distances in mm............................................................................57
vorticity magnitudes in sec*1, x and y distances in mm............................ 58
3.19 Schematic diagram of the effect of disk thickness on vortex formation
and shedding; (a) KCt = 0, (b) KCt = 0(2tt) and (c) KCt -> ~ ..............59
4.1 Experimental setup of Vertical Motion Mechanism (VMM) and test

cylinder..................................................................................................79
4.2 Time histories of forces and displacement for the test cylinder at
KC^Q3 and £=89236............................................................................ 80
4.3 Time histories of forces and displacement for the test cylinder at
KC^S.% and £=89236............................................................................ 81
4.4 Time histories of forces and displacement for the test cylinder with
disk at KC=0.23 and £=158600............................................................. 82
4.5 Time histories of forces and displacement for the test cylinder with
disk at KC=0.6 and £=158600................................................................83
4.6 Damping ratio of vertical cylinder vs. KC; £=89236................................. 84

4.7 Damping ratio vs. KC for cylinder with and without disk......................... 85
in i
4.8 Damping ratio vs. KC at different thickness KC numbers....................... 85
4.9 Damping ratio for vertical cylinder vs. KC at different p....................86
4.10 Damping ratio vs. KC for cylinder with disk at different p .................... 86
4.11 Variation of damping ratio with KC for the vertical cylinder at
different forward speeds; /J=89236..........................................................87
4.12 Damping ratio for vertical cylinder vs. KC at different forward
speeds; P= 156500..............................................................................87
4.13 Damping ratio vs. KC for vertical cylinder with disk at different
forward speeds; /?=158600......................................................................88
4.14 Damping ratio vs. KC for vertical cylinder with disk at different
forward speeds; jS=278200.................................................................r...88
4.15 Added mass coefficient of vertical cylinder vs. KC', /?=89236................. 89
4.16 Added mass coefficient vs. KC of vertical cylinder at different p. 89

4.17 Added mass coefficient of vertical cylinder vs. KC at different
forward speeds; /5=89236........................................................................90
4.18 Added mass coefficient of vertical cylinder vs. KC at different V;
p=156500...............................................................................................90
4.19 Added mass coefficient vs. KC of cylinder with disk at different p . 91
4.20 Added mass coefficient vs. KC for cylinder with disk at different
forward speeds; j5=158600......................................................................92
4.21 Added mass coefficient vs. KC for cylinder with disk at different
forward speeds; /J=278200......................................................................92
A.1 Nonlinear streaming flow streamlines due to an oscillating circular
cylinder; (a) without uniform stream and (b) with uniform stream. 113
B.l Percentage erroneous vectors vs. seeding density........................... 122
B.2 Percentage erroneous vectors vs. displacement..............................122
B.3 RMS error in pixels of the subpixel methods vs. particle image
diameter..............................................................................................123
B.4 Number of good and erroneous vectors vs. quality parameter. 124
B.5 Effects of varying the threshold parameter on the % erroneous vectors
remaining and the % good vectors lo s t................................................. 124
viii
LIST OF TABLES
Table
3.1 Parameters of flow visualization experiments................................... 30
3.2 Values of parameters for Figures 3.4 - 3.18........................................... T.38
4.1 Parameters of experimental force measurements....................................... 69
5.1 Particulars of example prototype TLP (Petrauskas and Liu, 1987)............ 97
5.2 Damping ratio of prototype TLP components................................... 99
5.3 Damping ratio of prototype TLP (4 columns and 4 pontoons)..............99
ix
LIST OF APPENDICES
Appendix
A Extension of Wang's theory to include a small uniform stream............... 106
B Digital Imaging Velocity Measurement System - details........................... 114
C Flow due to an infinitely long cylinder executing axial oscillations.......... 125
LIST OF SYMBOLS
A Amplitude of oscillation
Awp wateiplane area
B equivalent linear damping coefficient
Cd drag coefficient of Morison’s equation
Cm inertia coefficient of Morison's equation
d diameter of seed particles
D characteristic body length, typically cylinder diameter
Dv diameter of vertical cylinder (Chapter 5)
Dk diameter of horizontal cylinder (Chapter 5)
D\ diameter of a disk
E dimensionless velocity error function
f frequency of oscillation (Hz)
/o,« natural frequency of the test cylinder without and with springs
Fi total damping force
Fk hydrodynamic force
Ft measured total force
Fj amplitude of Fourier components of measured total force
g acceleration due to gravity
/i,2 intensity function of an image or portions thereof
k spring stiffness
KC Keulegan - Carpenter number, 2 jcAJD
L length of horizontal cylinder
xi
m mass of a cylinder
ma added mass
R Reynolds number based on diameter {-KC \9)
Res Reynolds number based on amplitude of oscillation (Jonnson, 1980)
S frontal area of vertical cylinder, itD2t4
Skc Uniform flow velocity parameter

t time
t\ thickness of a disk
T draft of vertical cylinder
u flow velocity
U magnitude of flow velocity
ut terminal velocity of seed particles
Uo magnitude of current velocity
V Reduced velocity parameter (Uq/ coD)
x horizontal length coordinate

y vertical length coordinate, also of springing oscillation
fi dimensionless frequency parameter, D2JJv
S internal angle of an edge
S(x,y) Dirac-delta function
<f>j phase of Fourier components of measured total force
H dynamic viscosity
p water density
pp density of seed particles
t viscous shear stress
v kinematic viscosity
co angular velocity (2^)
£ vorticity
xii
Z dimensionless linear damping ratio (ratio of system damping to critical
damping)
Zv damping ratio of vertical cylinder
Zh damping ratio of horizontal cylinder
Z\ damping ratio of vertical cylinder with disk
CHAPTER I
INTRODUCTION
There is an increasing interest among oil exploration companies towards drilling in

the deep waters of the oceans. One of the more viable structures for deep sea oil
production is the Tension Leg Platform (TLP). Ever since the success of Conoco's Hutton
TLP installed in 1984 at the North Sea, interest in this type of structure has widened.
Typical applications of TLP's in the future are expected to be in the depths of 120 -1500 m
(Patel, 1989). Though a TLP can be classified as a moored system, it is basically a hybrid
between a fixed and a compliant structure (Demirbilek, 1989). The underwater structure of
a TLP consists of vertical cylinders or columns, and horizontal members or pontoons
(Figure 1.1). The TLP derives its vertical stability from taut moorings called tendons, that
attach the structure to the foundation on the sea floor. The tendons are pre-tensioned to
account for the structure’s excess buoyancy over weight The tendons allow the TLP to
move in the horizontal plane, i.e. in surge, sway and yaw, but constrain its motion in the
vertical plane, i.e. in heave, pitch and roll (Demirbilek, 1989). The natural frequencies of
horizontal and vertical motions are also widely different Thus, these two motions can be
treated as two separate classes of problems.
Generally, the natural frequencies of a TLP in heave, pitch and roll are in the order
of 0.2 - 1.0 Hz, high relative to most wave frequencies occurring in the ocean (Heideman,
1989; Faltinsen, 1990). However, second order wave forces at the sum frequencies of
component waves can excite the structure causing vertical oscillations called "springing"
(Faltinsen, 1990). While the magnitude of these exciting forces may be small, the response
of the system may not be negligible, due to very low damping in the system (Chakrabarti,
1991). Springing introduces large tendon loads which is an important design
consideration.
Several ways to augment system damping in the vertical direction have been
proposed. One method easy to implement is to add a disk or a "lip" to the base of the
columns of the TLP (Troesch, 1990). The disk when oscillated normal to its plane,
enhances damping through vortex generation and shedding. The concept is similar to
employing bilge keels to reduce roll motions of ships (e.g. Beck et al. 1989).
The various components of hydrodynamic damping for bodies such as TLP's are
generally due to:
1. time dependent boundary layer flows,
2. flow separation, vortex formation and shedding, and
3. free surface effects such as wave generation or wave diffraction.
These components are influenced by various factors including wall roughness and
incident currents. The resultant damping forces are surface integrations of both shear
stresses and normal stresses or pressures, each of which may follow different scaling laws.
For this reason, model testing of a complete geosim in a model test basin will not provide
useful answers that can be extrapolated to full scale (Couch et al. 1984). Since it is not
possible to satisfy all the relevant scaling laws in one test, various common structural
shapes are tested separately, see Figure 1.2. This suggests that the flow due to a vertically
oscillating TLP can be decomposed into the following sub-problems1:
1. cylinder in oscillatory cross flow,
2. cylinder in oscillatory axial flow, and
3. cylinder-disk configuration in oscillatory axial flow.
JIn the rest of this work, the flow due to an oscillating body in a quiescent
fluid and an oscillating flow around a still body are used interchangeably. It is known that
these two flow situations are kinematically equivalent (see Newman, 1977). Dynamically,
the flows differ only by the Froude-Krylov force, and the added mass and damping values
for the flows are expected to be the same.
3
Based upon engineering necessity, the most significant quantities are then measured and
scaled properly. It is assumed that the total system hydrodynamic damping can be
expressed as the net sum of the individual members, each evaluated separately from the
other structural parts. This is clearly an approximation, for all three of the above damping
components are inter-related and in general not independent However, these interaction
effects are typically of higher order, and expected to be small in magnitude.
Damping mechanisms that are important to each component problem listed above
are considered in this thesis. The two main dimensionless parameters for these oscillatory
flows are the Keulegan-Carpenter number, KC,
KC = — ' (1.1)
D
and the Stokes or frequency parameter, fi,
D2f
P = *L L . ( 1.2 )
v
Here, A is the amplitude,/the frequency of oscillation (Hz) and v the kinematic
viscosity. D is a characteristic length (usually the diameter) of the cylinder. A Reynolds
number (R) can be defined as the product of KC and p.
For a cylinder in an oscillatory cross flow, the flow occurs along a plane parallel to
the cross-section of the cylinder. The three-dimensional flow in this case can be locally
approximated by a two-dimensional flow, within a certain frequency and amplitude range.
This flow has been extensively studied analytically, numerically and experimentally. In
Chapter 2, published results for an oscillatory cross flow are reviewed, and important
results are highlighted- The analytical approach adopted by Wang (1968) is extended to
include the effects of a small uniform stream in Appendix A.
In comparison to oscillatory cross flows, die literature on cylinder or cylinder-disk
configuration in oscillatory axial flows is small. Theoretical treatise of this work is at an
elementary stage. Some numerical and experimental results are available. In order to
4
obtain a detailed understanding of these flows, an extensive flow visualization study was
initiated. A novel variant of the particle image velocimetry technique was developed as part
of this research. This method is described in Chapter 3 and in Appendix B. Qualitative
and quantitative flow visualization results are presented in Chapter 3. Edge flows
associated with oscillating disks of different thickness and a cylinder are considered in
detail
Model tests were conducted to estimate the damping of an oscillating vertical
cylinder with and without a disk attached to the bottom. The tests were conducted at
different frequencies and amplitudes of oscillation and at zero and small forward speeds.
Drag coefficients are calculated and mechanisms of damping are discussed in Chapter 4.
Having developed an understanding of the fundamental flow problems, scaling
laws are derived in Chapter 5 for extrapolation of model results to full scale. Damping
coefficients are calculated for an example TLP prototype, and compared with published
data. It is hoped that results obtained from this research will be useful in developing design
guidelines for estimating the damping of a Tension Leg Platform.
5
Springing
Incident current
Column
Pontoon
•Tendons
Figure 1.1 Schematic diagram of a Tension Leg Platform
6
or
Figure 1.2 Component problems of a TLP
CHAPTER II
CYLINDER IN OSCILLATORY CROSS FLOW: REVIEW AND
DISCUSSION
We consider in this chapter, the case of a cylinder oscillating in a plane parallel to its
cross section, or equivalently that of an oscillatory cross flow incident on a cylinder at rest
The two non-dimensional parameters of the flow are the Keulegan-Caipenter number, KC,
and the frequency parameter, p , defined in Eqs. (1.1) and (1.2). The characteristic length
D is chosen as the diameter of the cylinder.

Experimental investigations have suggested that an oscillatory cross flow can be
classified into regimes governed mainly by KC, with a weak dependency on the Reynolds
numberR (Bearman et al. 1985; Saipkaya, 1986). When KC is very small {KC —» 0), the
oscillatory boundary layer is attached, two-dimensional and laminar. This flow has been
solved analytically using asymptotic expansions. Earlier works include Stuart (1966),
Wang (1968) and Riley (1975). These theories indicate the existence of an inner region in
the boundary layer called a Stokes layer. Organized Reynolds stress in the Stokes layer
leads to a net mean flow, referred to as steady streaming flow (Telionis, 1981), which
causes closed-cell circulation patterns within the boundary layer. Wang (1968) provided a
composite solution for the oscillatory flow past a circular cylinder, and obtained
expressions for the drag and inertia forces on the cylinder. Cylinders of non-circular cross
section have also been tried analytically. Bearman et al. (1985) derived force expressions
for cylinders of arbitrary cross section using singularity distributions and the Blasius
theorem. Kim (1988) and Kim and Troesch (1989) used conformal mapping and the Lewis
transformation to solve for the Stokes layer of cylinders with circular, rounded-square and
8
square cross sections. Flow in the outer layer was solved numerically using finite
differences. Their theoretical and numerical results compared well with flow visualization
results. Pattani and Olson (1987) used finite elements to model streaming flows due to
oscillating circular, square and airfoil section cylinders. Good agreement was obtained
with published experimental results.
For a given P, as KC increases, the laminar flow becomes unstable. Beyond a
critical value of KC, transition occurs, eventually leading to a turbulent boundary layer. A
three-dimensional instability in the boundary layer of an oscillating circular cylinder was
first noticed during the flow visualization experiments of Honji (1981). At the onset of
instability, streaks composed of mushroom shaped vortices were formed along the length
of the cylinder. Honji (1981) also found that the values of KC at which the flow became
unstable decreased with increasing p. Further investigations using linear stability analysis
by Hall (1984) confirmed Honji’s (1981) findings. An expression for the critical KC
(KCcrit) at which the flow became unstable was obtained by Hall (1984) as
KC^n = 5.71 Sp"0'25(1+0.205/T'°'25+- •-) (2.1)

Another flow event that occurs as KC is increased for a given p is the onset of
boundary layer separation. For any given P and for sufficiently large values of KC, the
flow around the cylinder will have the combined effects of separation, instability and
turbulence. For a circular cylinder, Sarpkaya (1986) found that transition and separation
occurred almost simultaneously for p ~ 2600, KC = 1.25 and R = 3250. For p < 2600,
separation preceded transition. For p > 2600, transition occurred first, and separation was
delayed to a higher KC value. Separation, vortex pairing and their effects on induced
forces were studied using flow visualization by Williamson (1985). Tatsuno and Bearman
(1990) proposed eight flow regimes within 1.6 < KC < 15, and 5 < p < 160, where
distinguishing vortex patterns due to both separation and Honji instability were formed.
For bodies with sharp edges, flow separation will always be expected to occur at
the edges. In the case of an oscillatory flow, the detached vorticity from an edge will be
9
swept by the flow reversal to cause vortex shedding (Bearman et al. 1985). However,
Kim and Troesch (1989), based on numerical and flow visualization results have suggested
that in the limit of KC —» 0, sharp-edged bodies experience attached flow in the mean
sense.
2.1 Inline force on a cylinder
In offshore applications, Morison's equation (Morison et al. 1950) is used to model

inline forces on a cylinder due to oscillatory flow. According to this equation, the
hydrodynamic force on a circular cylinder is expressed as
Fk = ip M ) 2C„ ^ + ± p D C M (2-2)
The first term is the inertia force component in phase with acceleration, and the
second term is the drag force in phase with the velocity u. Cm and C<i are the inertia and
drag coefficients, determined experimentally by the method of least squares or Fourier
averaging (Sarpkaya and Isaacson, 1981). In case of sinusoidally oscillating flow we
have,
u = Ucosd
(2.3)
where U is the amplitude, co the angular frequency and 0 the initial phase. The drag force
then consists of a cos 0]cos 6\ term, which can be approximated by the first term of an
equivalent Fourier series, i.e., (8 / 3rr)cos6 (Sarpkaya and Isaacson, 1981).
Wang’s (1968) asymptotic theory provides expressions forCm and C<j for a
circular cylinder at rest, in the limit of KC « 1, » 1 and (R KC ) « 1. To order
( ttjS)- 1-5, these expressions are
Cm = 2 + 4 (^ )-° -5 + (7T)3)-1-5, (2.4)
Q = ^ [ ( t f / 3 ) ^ - 5 + O r# -1 -0.25(xP)-15]. (2.5)
10
Force measurements were carried out by Sarpkaya (1986) in his U-shaped

oscillating flow tunnel. Three /? values of 1035, 1380 and 11240, and KC ranging from
0.4 - 20 were used. For )3=1035, Sarpkaya (1986) found that below a critical KC of 0.75
marked by the onset of Honji instability, measured Cd values agreed well with the
prediction of Wang's (1968) theory. In the vicinity of KCcru, variation of Cd was not very
clear. Beyond critical KC and until flow separation, Cd values followed the slope of
Wang's (1968) theory line, but with a higher offset. Beyond the separation point, Q
varied in response to the effects of vortex formation and shedding. A similar trend was
noticed at higher ft numbers, though the effect of transition was more pronounced.
Some of Sarpkaya’s (1986) conclusions were in agreement with published
numerical results obtained for two-dimensional flows. Justesen (1991) using a stream
function - vorticity formulation of the Navier - Stokes equation obtained numerical results
for small to moderate KC (0 - 26) and moderate f3 values (196 - 1035). The results for Q
agreed well with Wang's theory until flow separation occurred. The numerical method of
Justesen (1991) modeled a two - dimensional flow and hence Honji instability was not
realized. Smith and Stansby (1991) using the random vortex method have also obtained
similar results. The authors have observed that Cd attains a minimum at the onset of
separation, beyond which it increases, similar to experimental results. Numerical
predictions of the drag coefficient by Graham and Djahansouzi (1989) using a discrete
vortex method also showed similar trends based on KC. However, at ft = 196, predicted
values of Graham and Djahansouzi (1989) were higher than those of Justesen (1991). The
difference was expected to be due to the numerical methods used (Justesen, 1991).
Variation of Q with KC becomes more complex as /J increases. For KCg-it < KC
< KC (separation) Sarpkaya (1986) found that measured values had the same slope as the
theory line but a higher intercept Recent experiments using heave decay test by Bearman
and Mackwood (1992) and Bearman (1992) for /J ranging from 14371 - 29014, showed
that Cd values even at KC = 0.1 were offset from the theory line. For most f3 and
11
0.1 < KC < 0.5, data for Cd followed the slope of the laminar trend very well. Figure
2.1a is reproduced from Troesch and Kim (1991) which shows Cd vs. KC for a smooth
circular cylinder at /J = 48600. Also shown in the figure are Wang's theory line and Cm
values. The trend of measured Cd is similar to Bearman's (1992) and Sarpkaya's (1986)
results for higher /? values. No abrupt changes due to transition are noticeable (KCcrit
using Eq. 2.1 is evaluated at 0.39). At/J = 23200, Troesch and Kim (1991) noticed scatter
in their data, but the measured values were offset from theory line for all KC values in the
range of 0.1 - 0.9.
Figure 2.1b also reproduced from Troesch and Kim (1991) shows the variation of
Cd measured and theory, and Cm as functions of KC at /? = 48600, for an oscillating -
cylinder of square section. For KC < 0.2, Cd follows the trend of laminar theory line,
though the offset is much higher than that in Figure 2.1a. Beyond this KC, separation
effects become important and the measured values agree more closely with the values
obtained by the vortex shedding model of Bearman et al. (1985).
During experiments conducted at high ft values, system vibrations and excitation of
harmonics can be a significant source of external disturbance which can induce transition to
turbulence at a much lower KC than predicted by Eq. 2.1. It thus appears that for all
practically realizable KC numbers, the boundary layer is turbulent Bearman in
Sarpkaya's paper (1985) suggested that if /J is formed using an effective eddy viscosity,
then the trend of Cd with higher offset could be explained. The effective eddy viscosity
would be determined experimentally, and to some extent would inevitably depend on the
experimental setup and the range of /lvalues used. Sarpkaya (1985) determined the eddy
viscosity for a circular cylinder to be about 25 u, whereas Troesch and Kim (1991)
estimated it to be about 20 u. For square section cylinders and for KC < 0.2, Troesch and
Kim (1991) estimated the effective eddy viscosity to be about 50u
Figure 2.2 is a plot of Cd vs. J3 at a KC value of 0.5 for oscillating circular
cylinders. Shown in the figure are results obtained by many authors, along with the
12
laminar theory line of Wang (1968). Troesch and Kim's (1991) result shows an
extrapolated point obtained from the least squares curve fitted to Cd values for lower KC
values. Around f5 = 20000, the drag coefficient suddenly increases, and decreases
gradually afterward. An effective eddy viscosity would generate lines parallel to the curve
denoted as Wang’s theory. Figure 2.2 suggests that an effective eddy viscosity correction,
if valid, would be applicable over only small ranges of {3.
To better understand the influence of boundary layer transition on a circular
cylinder, we can refer to available literature on oscillating flat plates. Oscillatory turbulent
boundary layers on smooth and rough flat surfaces were studied by e.g. Kamphuis (1975),
Jonsson (1980) and Myrhaug (1984). According to these references, a fully turbulent
boundary layer occurs at a Reynolds number (based on amplitude of oscillation, Reg) of
about lOMO5. For a KC of say 0.5, this translates to a p in the range of 105-106.
Friction drag for flat plates due to laminar and turbulent boundary layers were deduced by
Kamphuis (1975) and Jonsson (1980). For the laminar case, the friction drag coefficient
of Kamphuis (1975) agreed with the trend derived by Wang (1968) for a flat plate. Based
on experimental results in the turbulent region, Jonsson (1980), proposed that the friction
AA
factor for a flat plate was proportional to Reg . Using the present notation, this implies
that the coefficient of the friction drag component would vary as
Q (friction) «=KC~0AJ3~0'Z. (2.6)

In Figure 2.2, a straight line fitted to the data points beyond (3 = 20000 has a slope of
-0.186, implying a functional relationship of J3 -0.186. The proximity of the value of the
slope to the exponent of JSin Eq. (2.6) is possibly coincidental due to scarce data at /?>
20000 in Figure 2.2. Further, Eq. (2.6) applies to the friction drag component only and
also does not include effects of surface curvature. Evidence from Figure 2.2 is thus not
entirely conclusive, and more results are required at higher f3values to identify the trend of
Cd.
13
2.2 Effect of currents
Inline oscillations of a cylinder in a uniform flow has attracted attention partly

because of combined occurrence of waves and currents and partly because of vortex -
induced vibrations encountered in tidal waters. When the velocity of oncoming flow is
significant, it leads to vortex shedding off the separation points on the surface of a cylinder.
There exists extensive literature on vortex shedding and the consequent flow-induced
vibrations, e.g. Blevins (1977), Sarpkaya and Isaacson (1981) and Chen (1987).
The effect of adding a uniform stream to an oscillatory flow can be treated in
different ways depending on the magnitude of the imposed stream. Non-zero mean
oscillatory flows have been studied analytically by Lighthill (1954) and Telionis (1975,
1981) and numerically by Tsahalis and Telionis (1974). The mean flow considered by
these references is one order of magnitude larger than the oscillatory part The steady flow
in this case causes a boundary layer of its own, with an imbedded streaming boundary
layer caused by the oscillatory flow (Telionis, 1975 and 1981). Glauert (1956) considered
the boundary layer due to an infinite flat plate normal to an oncoming stream, the plate
oscillating in its plane. The results of this problem were used to describe the boundary
layer in the neighborhood of the stagnation point of an oscillating cylinder.
While it is common to assume that Morison's equation applies equally well to a
periodic flow with a non - zero mean, Sarpkaya and Isaacson (1981) explain why this may
not be entirely true. The authors have presented two widely used modifications of
Morison's equation and have discussed their relative merits. Evidently, the formulation of
the force equation will affect the drag coefficient variability with respect to KC and the
incident current velocity. Arkell et al. (1992) obtained numerical results of drag force
coefficients on circular cylinders placed in a combined oscillatory and mean flow at f} =
200. For the case of the current velocity being smaller than the oscillation velocity, the
authors noted that the effect of current on drag coefficient was negligible prior to
14
separation. This was observed only when the Morison's equation was modified to include
a mean drag and an oscillatory drag component.
For a circular cylinder in two-dimensional oscillatory cross flow, Wang's (1968)
theory can be extended to include the effect of a small transverse current This theory as
presented in Appendix A is applicable for non-separating flows at high values of /J. For
sufficiently small currents, Appendix A shows that the nonlinear streaming flow is
perturbed to an extent sufficient to satisfy the boundary conditions on the cylinder surface.
The time-dependent component of the flow, and hence the drag coefficient to order
(tt/J)-0'5 , is unaffected. The range of validity of this theory is determined by requiring a
velocity parameter, Skc>to be
(2.7)
where V is the reduced velocity parameter (Appendix A). As an example, consider a

horizontal pontoon of diameter 8.5 m (28 ft) oscillating at a period of 2.0 sec and an
amplitude of 0.3 m (1.0 ft). Eq. (2.7) suggests that a current of order l(h2 m/sec (10_1-
10'2 fl/sec) would not significantly affect the drag coefficients.
15
a doo
€
o*
(a)
0-1
KC
QOQ
□
“ fee nQ DB q
B B S BO D q D O DO
o* 1-
(b)
0 - 1 1_______ I____ I___ I. 1 1, M I- 1 ________ I____ 1 ' ! I I I I

0-01 0-1 1
KC
Figure 2.1 Drag and inertia coefficients vs. KC of (a) a circular cylinder
and (b) a square cylinder from Troesch and Kim (1991); /3=48600.
□ Cd', ♦ laminar flow Cd', ■ Cm.
16
•o
u
0.1
1E+2 1E+3 1E+4 1E+5 1E+6
P
■ Troesch & Kim (1991) experiment
H Troesch & Kim (1991) extrapolated
— Theory (Wang, 1968)
♦ Bearman et aL (1985) experiment
□ Sarpkaya (1986) experiment
O Bearman (1992) experiment
A Anaturk (1991) experiment
Figure 2.2 Drag coefficient vs. {3for circular cylinders in

oscillatory cross flow; KC=0.5.
CHAPTER ID
AXIAL OSCILLATIONS OF CYLINDERS AND DISKS: FLOW
VISUALIZATION STUDIES
3.1 Introduction
The bodies under consideration have sharp edges and hence vortex formation and
shedding are expected to dominate the flow characteristics. For this reason it is
advantageous to formulate the problem in terms of vorticity rather than primitive variables.
To lay the foundation for further discussion, let us consider the vorticity transport equation
derived by taking the curl of the unsteady Navier - Stokes equation (e.g. Batchelor, 1973;
Sherman, 1990):
(3.1)
The left hand side of Eq. (3.1) is the Lagrangian or material derivative of vorticity
and the right hand side comprises of the vorticity production/ dissipation terms, i.e. vortex
stretching and viscous diffusion respectively. In a broad sense, the equation states that the
local rate of change of vorticity is governed by the net difference between convection and
generation of vorticity. In two-dimensional flows vortex stretching does not occur
(Batchelor, 1973). In case of high Reynolds number two-dimensional flows, the effects of
viscosity are confined to a thin boundary layer near the surface of a body. Far from the
body, the problem may be approximated by an inviscid flow where the local change of
vorticity is governed solely by convection. However, even at distances far from the body,
17
if steep spatial gradients of vorticity exist in the flow, viscous diffusion effects may come
to play.
When a moving body contains a region of high surface curvature like an edge, the
flow is unable to follow the surface and hence separates from the body forming a free shear
layer. In the case of an oscillatory flow, the free shear layer rolls up into vortex rings,
which are then convected and diffused as the flow reverses. At very small amplitudes of
oscillation, vortices shed from one edge may not move far from the edge, and hence
shedding at each edge may become independent of the other edges in the body. Graham
(1980) has stated that for an isolated edge in oscillatory flow, the pair of vortices formed
during any two half cycles convected away from the edge due to mutual induction. He
observed through flow visualization that convection was asymmetric for a single edge and
anti-symmetric for a body with two edges. Singh in De Bemardinis et al. (1981) noted that
for a body with two edges, vortices moved from one edge to another and vortex pairs
convected at large angles (45 deg) from the edges. As KC was decreased to very low
values, diffusion processes predominated and no convection was noticed (Singh's work in
Graham, 1980).
Graham's (1980) method for an isolated two-dimensional infinite edge used a
distribution of discrete point vortices along the body surface, and the effects of separation
were replaced by a Kutta - Joukowski condition at the edge. Conformal mapping was used
to solve for the inviscid flow around the edge. Some flow visualization experiments were
conducted for comparison with numerical results. The discrete vortex method was applied
by Bearman et al. (1982) and Graham (1985) to study roll damping of a barge, and by
Bearman et al. (1985) to obtain drag coefficients of a flat plate and cylinders of square and
diamond sections. Lian (1988) adopted a similar method to study oscillatory flows around
flat plates and midship sections fitted with bilge keels.
Hows generated by three-dimensional axisymmetric bodies are likely to be more
complex than two-dimensional flows due to vortex stretching and the self-induced velocity
of vortex rings. De Bemardinis et al. (1981) found that axisymmetric oscillatory flows
around slender disks were quite similar to the two-dimensional isolated edge flows. Vortex
pairs formed during one cycle convected unidirectionally, and the direction was dependent
on the starting direction of the flow. De Bernardinis et al. (1981) used a vortex ring
distribution in their calculation of axisymmetric oscillatory flow past a disk. The outer
inviscid flow was solved using a boundary integral method. Vortex shedding was
assumed to be locally two - dimensional, thus enabling the use of some of the concepts of
Graham (1980).
Few flow visualization studies of oscillating cylinders and disks exist in literature.
During the present study, a detailed flow visualization is conducted on these types of
oscillatory flows. Qualitative flow visualization was used to obtain pathlines in the flow.
Quantitative flow visualization was obtained by using a variant of particle image
velocimetry (PIV), developed as part of this research. The next two sections of this chapter
branch away from the main subject to lay the foundation of flow visualization techniques,
in particular to the powerful tool of PIV. Several details of the PIV scheme employed in
the experiments are presented in Appendix B. Numerically simulated images were used to
test the PIV setup before application. Results of these simulations are also presented in
Appendix B. Details of the setup used during the experiments are presented in Section
3.3. Results are presented and discussed in Section 3.4.
3.2 Flow visualization methods
Flow visualization has played an important role in understanding several fluid

mechanics problems. By minimizing physical interference with the flow, many techniques
provide reliable information about the variables of the flow field. Van Dyke (1982) has
assembled photographs of fluid motion ranging from creeping to supersonic flows. A
comprehensive understanding of various flow visualization schemes can be obtained from
texts such as Merzkirch (1987) and Goldstein (1983). Different methods presented here
20
include dye tracers, schlieren photography, hydrogen bubble wires and smoke wires.
While all these methods provide a qualitative understanding of the flow, they do not
provide quantitative data on the flow variables. The scope of these methods is thus
restricted.
For unsteady flows, it is often required to obtain velocity data at several points in
the flow simultaneously. Presently, several "particle imaging" schemes utilize optical
methods to follow the motion of marker particles that are added to the flow. The scattered
light from the particles are recorded and investigated to obtain the required information.
Adrian (1991) has discussed the three modes of operation of particle imaging i.e., laser-
speckle mode, particle tracking mode, and high image density PIV mode. The different
modes are distinguished based on the concentration of scattering particles per unit volume
(Adrian, 1991). Particle tracking velocimetry has been successfully applied to dispersed air
bubbly flows by Hassan and Blanchat (1991) and Hassan and Canaan (1991). Laser
speckle velocimetry has also been applied to many flows, for example laminar thermal
convection (Simpkins and Dudderar, 1978) and thermocapillary convection (Wozniak et
al., 1990). The high image density PIV - generally referred to as PIV in literature - has
found several applications recently. Examples are an impinging wall jet (Landreth and
Adrian, 1990), vortex ring evolution (Willert and Gharib, 1991), turbulent channel flow
(Liu et al., 1991) and vortex - free surface interaction (Gharib et al., 1992).
Velocity estimation in PIV is done by locating the position of a group of particles
within an “interrogation region” in the flow at two successive instants of time. The
assumptions here are that the seed particles follow the local flow closely and also the group
of particles under consideration displace uniformly and linearly over the time interval. PIV
usually involves illuminating the flow at two time instants and recording the particle images
on a photographic medium. Variants of PIV arise from the methods of recording and
analysis of images to obtain the displacements. Many schemes use a 35mm photographic
film as the recording medium. While the film offers a very high spatial resolution,
21
invariably photographs taken at two time instants are recorded on the same frame. This is
because of frame registration errors associated with single exposed multiple frame
techniques (Adrian, 1991). The double exposed picture presents the problem of directional
ambiguity due to loss of phase information of the measured displacement Techniques
used to overcome this problem range from mechanical means like image shifting using a
rotating mirror (Landreth and Adrian, 1988) to pulse tagging (Grant and Liu, 1990).
Alternately, a video camera may be used to record the flow images on successive video
frames, see Kimura and Takamori (1986), Willert and Gharib (1991) and Westerweel and
Nieuwstadt (1991). The flow images are recorded on successive video frames thus
eliminating directional ambiguity.
The complexity of interrogation process and the computational time required to
process an image are important considerations in designing a PIV setup. The Young's
fringe method (Adrian, 1986b and 1991) using opto-mechanical means of analysis has
been employed by Liu et al. (1990). The use of autocorrelation techniques (Prasad et al.,
1992) and cross-correlation schemes (Willert and Gharib, 1991) using fast Fourier
transforms have also found application because of their speed of computation and ease of
automation.
Several aspects of the flow and the measurement technique affect the quality of
velocity measurements obtained. The complexity and speed of the flow inversely affect the
ease of velocity estimation. Complexity, for example, increases the local velocity
gradients, which challenges the assumption that all particles displace uniformly over the
time interval of measurement Willert and Gharib (1991) have mentioned that the effect of
local velocity gradients is to broaden the signal peak in a cross-correlation spectrum,
leading to biasing errors. The seeding density (particles per interrogation region) is
important in obtaining reliable displacement measurements. The size and specific gravity of
seed particles affect the ability of the particles to follow a local flow closely. The particle
image size is expected to have a significant influence on the accuracy of the velocity
22
measurement (Adrian, 1986b and 1991). Adrian and Yao (1985) have analyzed the effect
of the scattering characteristics of various seedings on detectability. The range of
displacements that can be measured by an analysis scheme depends strongly on the size of
the interrogation region used. While the Nyquist criterion restricts the maximum
measurable displacement to half the length of the interrogation region, Willert and Gharib
(1991) have mentioned that in practice this value may be even lower, approximately one-
third its length.
Identifying an optimal set of experimental parameters should yield good
measurements in a PIV setup. However, errors in measurements may still arise due to
inherent limitations of PIV either conceptually or physically. In these situations, pre- or
postprocessing is employed to obtain better results. Particle tracking schemes normally
employ pre-processing of images, like thresholding and filtering (Grant and Liu, 1990;
Guezennec and Kiritsis, 1990) or smoothing and edge detection (Hassan and Blanchat,
1991; Hassan and Canaan, 1991). Postprocessing in PIV usually requires identifying and
eliminating or replacing spurious measurements. Liu et al. (1991) and Willert and Gharib
(1991) have identified erroneous measurements by comparing them with their neighbors.
Linear interpolation was used to replace the removed vector.
The principal limitation of using a standard video camera format (RS-170) to record
images is that the framing rate is set at 30 Hz, thus only relatively low-speed flows could
be visualized (Willert and Gharib, 1991). A new double exposure scheme in conjunction
with image shifting is proposed to alleviate this limitation. The scheme allows for smaller
time interval between successive flow images and can thus be applied to higher flow speeds
than the single exposure scheme of Willert and Gharib (1991). Section 3.3 describes the
underlying concept of the single and double exposure methods. Numerical simulation was
conducted to understand the influence of several variables of the technique. In these
simulation tests, the double exposure and single exposure schemes are compared for
quality and accuracy of measurements. These results are presented in Appendix B.
I
I
23
3.3 The basic concept
The fundamental principles of PIV has been well described (Adrian, 1986b, 1991;
Willert and Gharib, 1991). This concept is described briefly for the specific case of using a
video camera. The process can easily be generalized to include any photographic medium.
If a flow of interest is seeded, and the seed particles are assumed to follow the flow
closely, then the velocity field is estimated by knowing the positions of the particles at two
successive instants of time. The seeded flow is illuminated at two time instants and
recorded on successive video frames. A small portion of the first frame (called an
interrogation region or a sub-image) is correlated with the corresponding region in the
second frame. The interrogation region typically contains several particle images, and the
corresponding particles are assumed to displace uniformly. The correlation is a maximum
at a point where the particle images in the first region match with those in the second. The
position of the correlation maximum with respect to the origin is equivalent to the
displacement of the group of particles in the small region of the flow. This is schematically
shown in Figure 3.1a. Knowing the magnification of the camera and the time interval
between the records, the local velocity of the flow can be estimated.
To present this concept mathematically, let the intensity function of the interrogation
region in the first image be /i(x,y), and that for the second image be I2(x,y). For
simplicity, let the two intensity functions be comprised of contributions from the seed
particles only, devoid of any background noise. If the particles in the second region are the
same as those of the first, but only displaced by a linear amount say (Ax,Ay), then
h Oc,y) = h(x,y)* S(x - A x , y - Ay) (3.2)
where the delta function is defined as:
5(*,y) = 0, x * 0 , y * 0
/> « ) * * - . ( 3 -3 )
—0 0 — 00
24
The cross-correlation function C(X,Y) of/i(x,y) and I2{x,y) is given by

C(X,Y) = Il(x,y)*J2(x,y)
= Il*(l1* S (x - Ax,y - Ay))
= f/i *7j )* 5(x - Ax, y - Ay)
= A(X,Y)* 8(x - Ax,y - Ay)
= A (X -A x ,Y -A y ) (3.4)
where * denotes spatial convolution. The autocorrelation function A(X,Y) is defined as the
convolution of ii(x,y) on itself. Since the maximum of an autocorrelation function occurs
at the origin (Bendat and Piersol, 1986), the above equation states that the maximum of the
cross-correlation function occurs at
X = Ax, Y = Ay. (3.5)
Location of the cross-correlation peak thus gives the displacement of the particles within the
interrogation region.
A video image obtained from a COD camera is an array of grayscale intensity values
typically ranging from 0 - 255 (8 - bit quantization). The PIV concept presented above is
directly valid for this case, with the integrals approximated by summations. The measured
displacements however, are accurate only up to the spatial resolution of the camera, as
determined by the size of a pixel in the image. In order to improve the accuracy of the
measurements, a sub-pixel estimation of displacement is required. Two ways of doing this
are described in Appendix B. The present experiments employ the centroiding method
(Appendix B), which was found to give lower errors in measurement The displacement in
this method is found by locating the centroid of the neighborhood of the discrete correlation
maximum.
With the known displacements, the local flow velocity is given by
25
(3.6)
where / is the length of a pixel. This process is repeated over the entire image to obtain the
velocity field of the flow. The vorticity field can then be estimated by differentiating the
discrete velocity data. By using central differencing, vorticity at location (/ ,/) is given by:
Cij = - f y i - i j ~ AxiJ+1 + A xij-i). (3.7)
Effect of At on measurement accuracy
If the flow is illuminated once every video frame, then the time interval between
two successive images is fixed at 33.33 msec (standard RS-170 video). The effect of this
on velocity measurement is a significant error even at moderate flow speeds. To
demonstrate this, let us consider the difference between measured and actual velocity:
Ox
L it ) it
If the flow under consideration is oscillatory, say

x(t) = A sin(eot) (3.8)
where A and coare respectively the amplitude and angular frequency of oscillation. Let us
define a dimensionless velocity error function as
E (co,t,te) = ~ (3-9)
A(0
The root mean squared error is defined as
(E) = E2(a),t,At)dt (3.10)
Substituting forxfr) from Eq.(3.8) and carrying out the integration, we find
, 1 - cos(toAr) —(oAtsin(ct)At) J_ n in
' (cote)2 +2
26
This result is plotted in Figure 3.2 vs. frequency f=(Dl2n, at At = 33.33 msec and
5.0 msec. The latter At is the value used in the experiments described in Section 3.4. The
RMS error is almost linear with At. This can be perceived by considering the series
expansion of sine and cosine functions in Eq. (3.11). To first order in At, this gives
(E )-^ . (3.12)
The slope of the curves in Figure 3.2 is thus proportional to At. According to the figure,
the RMS error at 33.33 msec exceeds 3% at a frequency of 0.5 Hz increasing up to 7.2% at
1 Hz. The averaged error at 5.0 msec in contrast, remains less than 1% over the entire
frequency range of Figure 3.2. Clearly low At values are desirable for better velocity
measurement
In the present work, a low At value is obtained by capturing two flow images in
one video frame. Each video frame contains two images of each particle, with one image
displaced relative to the other. This leads to directional ambiguity, as one has no
knowledge of which image was recorded first To eliminate this ambiguity, image shifting
using an oscillating mirror is employed. The purpose of image shifting is to move the
second image of each particle to a neighboring interrogation region. Then, instead of
correlating the same spatial region of successive frames (Willert and Gharib, 1991), this
method requires correlating neighboring regions in the same frame. This idea is depicted in
Figure 3.1b.
One side-effect of double exposure is a doubling of image density (number of
particle images per interrogation region). In a typical interrogation region, one can expect
at most half the particle images to correlate with the corresponding images of the
neighboring region to form the signal peak at the desired displacement The other half of
the particle images correlate at random with images of the neighboring sub-image, giving
rise to "noise" peaks in the cross-correlation. If the magnitude of a noise peak is higher
than that of the signal peak, then an erroneous displacement is measured. An error
27
correction scheme, whose principles are described in Appendix B, is used to effectively

locate and eliminate erroneous measurements. Linear interpolation is used to replace the
removed vectors.
3.4 Experimental Setup
The Digital Imaging Velocity Measurement (DIVM) system was developed at the
Marine Hydrodynamics Laboratory, University of Michigan to visualize several flows of
interest. The schematic of the experimental setup is shown in Figure 3.3. The setup as
shown is applicable to the single or double exposure PIV methods. Results for qualitative
visualization are obtained by a simplification of this setup, described later in this section. A
4-watt Argon-ion laser provides a continuous-wave laser beam of diameter 2.0 mm. Beam
shaping optics convert this light into a planar sheet of 5.08 cm (2.0 in) width, and 1.0 mm
thickness at the waist. Pulsing of the beam is accomplished by an acousto-optic modulator
(AOM) driven by the signal from a software-based pulse generator (Figure 3.3). The
modulator system is a tellurium dioxide crystal with a lithium niobate transducer. The
modulator assembly, when oriented at the Bragg angle, has a diffraction efficiency of 90%
at the first order. An aperture placed in front of the AOM allows only the first order beam
through, thus illuminating the flow according to the pulse pattern fed into the AOM driver.
The image acquisition system is a Pulnix charge coupled device (CCD) camera with a 85
mm lens. The imager has a resolution of 512 x 492 pixels and a pixel size of
17|imx 13pm. A mirror attached to a galvanometer is placed in between the lens and CCD
array, and provides the image shift between successive light pulses. The risetime of the
galvanometer limits the minimum At that can be used to about 5 msec.
During the experiments, the camera was allowed to run freely. The inputs of AOM
and galvanometer drivers were synchronized by the software based pulse generator. The
pulse generator was designed to send two pulses of desired duration and spacing at a
frequency of 30 Hz. To obtain a double exposed image, the duration of the pulses was set
28
at 2.0 msec, and the spacing between the pulses at 5.0 msec. The galvanometer was given
a signal to shift within the time interval between the pulses. If a single exposed image was
desired, the pulse generator set the spacing between pulses to zero and the duration of each
pulse at 1.0 msec. The galvanometer either remained inactive or was removed during this
operation. Alternatively, if qualitative visualization of the flow was required, the AOM and
galvanometer were removed. The laser illuminated the flow continuously and the camera
recorded traces of particles within the field of view.
Images acquired by the camera were written onto a real-time optical disk. The disk
is capable of storing up to 108,000 frames, equivalent to one hour of video at a framing
rate of 30 Hz. The stored images were digitized as 640 x 480 arrays and analyzed by the
DIVM software resident in the Macintosh II fx computer. The analysis uses an
interrogation region of 32 x 32 pixels, with 50% overlap between successive sub-images.
This produces about 103 vectors per image.
The models used for the experiments were a cylinder and two disks, details of
which are given in Table 3.1. The cylinder is made of aluminum, hollow and sealed at the
bottom. Disk A (also referred to as thin disk) is circular and has a uniform thickness of
6.35 mm (0.25 in), while disk B (the sharp disk) has a thickness of 2.54 mm (0.1 in) at
the center, tapering off to 0.1 mm (0.004 in) at the edge. Uniform oscillations of the model
were obtained using a motor - flywheel mechanism. The water tank used has dimensions
0.61m x 0.61m x 0.76m (24in x 24in x 30in). Cross-sectional fields of view of 37.1
mm x 28.2 mm for PIV experiments and, 42.4 mm x 31.7 mm for the continuous video
pictures were chosen in the vicinity of the edge under consideration.
The seeding particles used for qualitative flow visualization were Glass Microcarrier
beads1 of specific gravity 1.02 and diameter ranging from 90 - 150 pm. Due to their large
size and good scattering characteristics, these particles can be seeded sparsely to obtain
ôloHill Engineering Inc., Ann Arbor.
29
particle traces of good contrast The settling velocity of particles is given by (e.g. Douglas
etal., 1985)
(3.13)
where d and pp are the diameter and density of the seed particles respectively. For glass
microcarrier beads, the settling velocity is quite high, about 0.08 - 0.49 mm/ sec. The
particles are thus unsuitable for accurate velocity measurement
For PIV measurements, Titanium dioxide particles of average diameter 2.79 mm
and specific gravity of 4.26 were used. The diffraction limited image diameter (Adrian and
Yao, 1985) is estimated at 3.15 fim, equal to 0.23 pixel. The settling velocity of these
particles is estimated to be 0.013 mm/sec, which is two orders of magnitude smaller than
the typical velocities measured. The 3-dB frequency (Adrian, 1983) of a particle is a
measure of the response time of the particle to a unit step change in velocity. This is given
by a modification of Adrian's (1983) formula to include added mass effects,
9p.
(3.14)
For Titanium dioxide particles in water the 3-dB frequency was found to be 78
KHz. They are thus expected to follow the flow quite well, particularly in the vicinity of
the vortex core. An average image density of 10 - 20 particles was used during the
experiments. This was confirmed by visual analysis of several single exposed sub-images
obtained before the experiments.
30
Table 3.1 Parameters of flow visualization experiments
Water tank area 0.61 m x 0.61 m (24 in x 24 in)

height 0.76 m (30 in)
Cylinder diameter 0.11 m (4.5 in)

draft 0.30 m (12.0 in)
Disk A diameter 0.15 m (6.0 in)

thickness 6.35 mm (0.25 in)
DiskB diameter 0.15 m (6.0 in)

thickness 2.54 mm to 0.1 mm (0.1 in to 0.004 in)
Frequencies of oscillation 0.1 - 1.0 Hz
Amplitudes of oscillation 3.17 m m -25.4 mm

(0.125 in - 1.0 in)
3.5 Results and discussion
Ability of the visualization techniques to produce useful results depends on the

complexity of the flow. The complexity of the flow in turn depends on the body shape,
and the KC and J3 values (or equivalently the amplitude Reynolds number, R) used.
Results in this section are presented for different shapes at different Reynolds numbers
mentioned in Table 3.2.
Results for the cylinder with and without disk A
Figure 3.4a - h show pathlines in the flow due to the test cylinder oscillating at a
frequency of 0.5 Hz and an amplitude of 6.35 mm (0.25 in). This gives KC = 0.31, j8=
7031 and R = 2461. As mentioned in the previous section, the flow seen here occurs along
31
a cross-section of the actual axisymmetric flow. The figures show eight snap shots taken
during one cycle of oscillation, starting from the top dead center. Figure 3.4a shows a
cross section of a vortex ring that has rolled up below the cylinder during the previous half
cycle. Due to the presence of the vortex ring, the flow at the edge reverses ahead of the
ambient flow, as pointed out by Graham (1980). As the cylinder begins its downstroke,
the rolled up vortex is convected initially due to a stagnation-type flow that occurs at the
bottom of the cylinder. This convection however, is short-lived, as diffusion comes into
play. While diffusion annihilates the first vortex, a second vortex ring rolls up around the
cylinder on the side(Figures 3.4b,c). Due to interference with the vertical wall of the
cylinder, the second vortex ring is unstable and breaks down into smaller vortices, which
are subsequently diffused. At this point (Figure 3.4e) the cylinder is almost at its bottom
dead center. During the upstroke, a third vortex ring similar to the first is formed beneath
the cylinder, and the process is repeated cyclically.
The flow described herein is unlike the flows due to oscillating isolated edges of
Graham (1980), where vortices shed at each half cycle are found to form pairs that convect
away from the edge. The first vortex (Figure 3.4a), depleted of its strength due to
diffusion, is not able to convect away from the edge. The remnant of the first vortex is
carried by the flow around the edge where the second vortex is formed (Figure 3.4b,c)
The asymmetry of the edge with respect to the direction of oscillation influences the vortex
formation pattern.
In keeping with one of the objectives of this thesis, modifications in the flow
pattern due to a disk attached to the cylinder are next considered. Disk A was attached to
the base of the cylinder and oscillated at/? = 1965. The flow shown in Figure 3.5, is seen
to be modified significantly due to the presence of the disk. Actually, the flow is
completely dominated by the disk and the features are similar to those of Figure 3.6 due to
disk A alone. This is because at very small amplitudes of oscillation, the length scale of the
local flow around the edge is small. Thus vortices formed off the edges of the disk do not
32
"feel" the presence of the vertical wall of the cylinder. At larger amplitudes of oscillation, it
was observed that the presence of the cylinder leads to skewness of the vortex ring formed
on the upper side of the disk. Asymmetry is thus introduced in the flow as KC is
increased. Since most of our current interest lies in the low KC range, this aspect is not
pursued further.
Results for the oscillating thin disk (disk A)

We next consider the flow due to disk A oscillating normal to its plane, i.e. along
its axis. Periodicity and repeatability of this flow were found to be strongly dependent on
the values of KC and /J, or the amplitude Reynolds number (R). Qualitative visualization
of the flow at i?=1965 is presented in Figure 3.6a-h. The amplitude and frequency of
oscillation for the flow are given in Table 3.2. Figures 3.7 and 3.8 show velocity and
vorticity for the flow at R = 655, obtained by single exposure PIV.
Several flow features of Figure 3.6 are found to be quite similar to those of Figure
3.5. As before, the eight photographs of Figures 3.6-8 are equally spaced along a cycle of
oscillation, with the first photograph in each figure showing the disk at its top dead center.
A rolled up vortex ring is observed below the disk in Figure 3.6a, which was formed
during the previous upstroke. As before for the case of the cylinder, the flow at the edge
reverses ahead of the ambient flow (Figures 3.7a and 3.8a). The first vortex initially
convects through a short distance before being diffused away. Remnants of this vortex are
seen around the second vortex in Figures 3.6-8. As the disk begins its descent, vorticity is
being shed at both edges of the disk. However, shedding and roll - up from the second
edge (the one farther downstream) is not as strong as that from the first edge. This can be
verified from the vorticity plots of Figure 3.8b. The vortex ring formed at the first edge,
initially present along the side of the disk develops in strength and moves further
downstream. Here the weak vortex ring formed at the second edge is rapidly amalgamated,
compounding to its strength. This feature has also been observed by Singh in Graham
33
(1980). At the bottom dead center (Figure 3.6e), the flow resembles the mirror image of
Figure 3.6a. Figures 3.6f - g show the flow as the disk continues it upstroke, and the
processes are very similar to those that occurred during the downstroke.
Figures 3.6-8 point out to several interesting features. The vortex of Figure 3.6a is
larger than the corresponding vortex of the cylinder (Figure 3.4a), even though the
Reynolds number of the cylinder oscillation is higher. This is presumably because of
vortex amalgamation from the two edges of the disk. The disk flow is symmetric about the
mean position of oscillation, which is due to the symmetry of the disk itself, and due to the
KC number being small (see also Graham, 1980). Vorticity shed from the edges rolls up
into vortex rings, which do not convect away from the disk. They remain in the proximity
of the disk, until flow reversal causes a rapid cancellation of vorticity. Careful observation
of the video of the flow showed that occasionally small vortices convect away from the
disk. This did not affect the gross features of the flow.
We next consider what happens to the flow at higher Reynolds numbers. Pathlines
of the flow at a frequency of 0.5 Hz and an amplitude of 12.7 mm (R = 6550) are shown in
Figure 3.9. Velocity and vorticity results for the flow at 0.2 Hz and 12.7 mm (R = 2751),
obtained by the double exposure PIV method are shown in Figures 3.10 and 3.11. Due to
90 deg shift produced by the galvanometer mirror, the disk is seen on the right side of
Figures 3.10 and 3.11. The first vortex formed in Figure 3.10a does not diffuse rapidly,
but instead is carried around the edge of the disk along with the flow (Figure 3.10c). The
presence of this vortex appears to squeeze the second vortex being formed. Significant
vortex stretching splits the second vortex into two vortices of the same orientation. These
two vortices separate away as the disk reverses in direction and diffuse rapidly (Figures
3.10f,g and 3.11f,g).
At low R, it can be seen (Figures 3.6-8) that the gross features of the flow are
periodic and repeatable. Anti-symmetry between the flows occurring during the two half
cycles is also seen quite clearly. For example, core vorticity in Figure 3.8a is about
34
42 sec-1, which compares in magnitude with the core vorticity of -40 sec1 in Figure 3.8e.
As frequency and amplitude are doubled, the strength of vorticity generated by the edges
increases, which diminishes the effects of diffusion. Convection leads to distribution of
vorticity in the vicinity of the edges. Presence of ambient vorticity makes some local
features asymmetric and unpredictable. The flow of Figures 3.9 through 3.11 confirm our
notion that the flow gets more complex as KC and /? are increased.
Results for the oscillating sharp disk (disk B)
The flow due to disk B can be considered to be an extreme case of the flow due to
disk A, when the disk thickness becomes very small. The absence of disk thickness
enhances interaction of the vortices formed during any. two successive half cycles. The
vortices thus formed are of opposite sign, and hence mutual induction and convection can
be expected.
An interesting feature of the vortex shedding pattern in this flow, observed during
the experiments, is its dependence on initial condition. In order to visualize this aspect, let
us consider the evolution of the flow as the disk is started from rest at bottom dead center.
Figure 3.12a-x shows this flow at a frequency of 0.5 Hz and an amplitude of oscillation of
3.2 mm. The first three cycles (a-n) show transients in the flow. Diffusion of vorticity is
seen to be significant as the flow develops. Towards the end of the second cycle (Figure
3.12h), some amount of convection is noticeable. From Figure 3.12 it is seen that the
vortex formed during the upstroke (i.e. below the disk in the figure) is stronger than its
counterpart formed during the downstroke. This difference in strengths is due to a longer
time period available for the former vortex to develop. This in turn is because the bottom
dead center was chosen as the initial condition. The situation is reversed when the initial
condition coincides with the top dead center. Figure 3.12q-x show two cycles of the well-
developed flow with a unidirectional vortex shedding pattern clearly visible (Figure
3.12q,u). This pattern was found to be well-established and reasonably stable over a long
duration of time.
35
Figures 3.13-15 show the well-developed flow due to disk B at two Reynolds
numbers of 1965 and 1441. The initial condition for these figures was chosen to be bottom
dead center. Figures 3.13a and 3.14a show the remnant vorticity from the previous cycle
of oscillation. As the disk moves up, vorticity shed from the edge rolls up into a strong
vortex on the leeward side. As the flow slows down, the flow at the edge reverses ahead
of the ambient flow (Figure 3.13e). De Bemardinis et al. (1981) has mentioned that this is
due to the presence of the vortex ring. As the disk moves down, a second vortex is formed
above the disk. The upstroke vortex meanwhile is convected downstream around the edge.
This vortex interacts with part of the downstroke vortex to form a pair that moves away
from the edge of the disk. With growing distance from the disk, the second vortex picks
up strength and a vortex pair of equal strength is eventually formed. The initial parity in
vortex strengths results in a uni-directional convection that is seen in, e.g. Figure 3.13g.
Due to a large amplitude of oscillation (12.7 mm) chosen for Figure 3.14-15, the vortex
pairing and shedding occurs outside the field of view, and is not seen clearly. As the disk
completes its downstroke, more vorticity is generated, which is diffused as the disk begins
its next upstroke (Figure 3.14a,b).
Figures 3.16-18 show the flow when the disk is initially started from the top dead
center. The flow as expected is the inverse of the flow seen earlier in Figures 3.13-15.
Here, the vortex ring that is formed during the downstroke is strongly developed. This
pairs with part of the upstroke vortex and converts away at about 45 deg below the plane of
the disk (Figure 3.16g,h). Vortex pairing not visible in Figure 3.14 at amplitude of
oscillation of 12.7 mm is seen to some extent in Figure 3.17g. The strengths of these two
vortices appear to be equal in magnitude, about 240 sec-1.
The flow due to disk B in many aspects resembles the flow due to a two-
dimensional isolated edge of Graham (1980). Unidirectional vortex shedding has also been
predicted by De Bemardinis et al. (1981). It must be emphasized that the asymmetric
vortex shedding has its origin in the initial condition chosen for the disk. The initial
36
condition decides the position of the stronger vortex, and hence the direction of vortex
shedding. If the initial condition was chosen at or near the zero crossings, the shed
vortices were found to be disorganized during the first few cycles. Eventually, a vortex
shedding pattern along one of the two stable directions (approximately 45 deg above or
below the disk) was established.
Effect of disk thickness on drag force

The flow phenomena seen in Figures 3.6-18 can be correlated by considering a
Keulegan - Carpenter number based on a characteristic length parallel to the direction of
oscillation (e.g. disk thickness, t j )
KCt = — . (3.15)
td
At an amplitude of 6.35 mm, we get KCt = 6.28 and 392.70 for disks A and B
respectively. For the cylinder, the characteristic length may be the draft, but the local flow
around the edge perceives the cylinder as a disk of infinite thickness. The cylinder may
thus be loosely termed as an infinitely thick disk, and KCt is approximately zero.
The flow patterns evidenced at different KCt are depicted schematically in Figure
3.19. When KCt is zero or very small, flow due to individual edges may be treated
independent of one another. Convection is weak, and occurs only to the extent of
transporting the vortices over short distances (Figure 3.18a). The damping in this case is
mainly due to diffusion of vorticity. As KCt increases to order one (e.g. disk A), the
length scale of vortex roll-up increases, and the interaction of flows due to two edges
becomes important Vortices are convected over longer distances (from one edge to
another, Figure 3.19b) and the diffusion process is slower. The damping in this case has
contributions from both convection and diffusion. When KCt becomes very large (e.g.
disk B) considerable vortex shedding is observed. Very little diffusion is noticeable, and
thus the drag force is primarily due to convection of vorticity. Other features of the flow at
37
large KCt are that vortex formation patterns are not symmetric with respect to the mean
position of the oscillation, and the vortex shedding direction depends on the initial start-up
condition of the disk.
Vortices leaving a control volume surrounding a disk represent an effective means
of energy dissipation and therefore imply increased damping coefficients. From the
schematic of Figure 3.19, it is expected that disk B can have a substantially larger drag than
disk A for a given KC and ft. In comparison with the cylinder, disk A has twice the
number of edges, and hence must experience at least a two-fold increase in drag force. So
the drag force can be expected to increase as KCt increases, with other parameters kept
constant
As a final note to this chapter, since most of our observations arise from
experiments conducted at very low values of KC (based on diameter), the above
discussion pertains only to flows with KC of order one or smaller.
38
Table 3.2 Values of parameters for Figures 3.4 - 3.18
Figure # Object f (Hz) A (mm) KC fi R
3.4 cylinder 0.5 6.35 0.35 7031 2461

3.5 cylinder+ 0.3 6.35 0.262 7500 1965
disk A
3.6 disk A 0.3 6.35 0.262 7500 1965
3.7, 3.8 disk A 0.1 6.35 0.262 2500 655
3.9 disk A 0.5 12.7 0.524 12500 . 6550
3.10,3.11 disk A 0.21 12.7 0.524 5250 2751
3.12 diskB 0.5 3.2 0.131 12500 1637
(BDC)
3.13 diskB 0.3 6.35 0.262 7500 1965
(BDC)
3.14, 3.15 diskB 0.11 12.7 0.524 2750 1441
(BDC)
3.16 diskB 0.3 6.35 0.262 7500 1965
(TDC)
3.17, 3.18 diskB 0.2 12.7 0.524 5000 2620
(TDC)
(BDC = bottom dead center; TDC = top dead center)
39
••’•*
.**.-V.••*
••*•.*•'.*■*.*■**.*
.*-•*•*.**.* Subimage at (x,y,t)
. .* •’ «•
*.*.•/. *.-•/* *.*•

u•*•.’.*.**.*.*-•*•*.**.•*.* *.*•.*.•/. •/. *.*.•**..*%
*.*.*
*••
.*••
•.**• :*•:,*• *.* *.*.;.*•
***.*.♦
*v..*.*
**•..*.*.*.
♦«
Particle Image at lim e t
2-D Cross x&m V(x,y,t) - u I + v J

- (Ax/At) i +(Ay/At) J
Correlation
• .*;• .*;• .*.* .*.♦ .* .• .*„• .*
m m m m i
*.*-•.*•■.*.*;•*•/. *.*••
.•• ,**..♦*-*.*•.**;■
.*.•**.■/. .*•*•••.
,’ •/.*v • Subimage at (x,yft+At) (a)
Particle Image at Time t + At
Subimage at (x,y,t)
/r
, .,
2-D Cross V (x,y,t)»u i + v j
♦.•/--V**.*••jftSNv'* •••*.*•• Correlation « (Ax/At) i + (Ay/At) j
•*.*
.•/.
.*.*•/.
.* •/. .^>sx-7S:* •-V-.*•
•/. •/. *.*•Cv*.*.
•.*••••
..*
V•.
*•
**••*.* /.*.•A -/. • V
*.• •/. •**.•**.•••.••*
.*• ♦**•• / . *.*• •'*. • V / . • v
Double exposed image \
Subimage at (xo^+At)
(b)
Figure 3.1 Particle Image Velocimetry concept;

(a) single exposure (b) double exposure.
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

f (Hz)
Figure 3.2. Root mean squared error in velocity vs. frequency;

□ Ar=5.0 msec, A Ar=33.33 msec.
41
WaterTank
4 Watt Argon-Ion Laser 1

Acousto-Optic Laser Sheet
Modulator /D isk\
Beam-Sharing Beam
Optics Stop
AOM Driver
t
85mm Lens
Image-Shifting Mirror/
Real-Time Galvanometer
Optical Disc Recorder
CCD Video
Camera
— | Frame Grabber j
Galvanometer
Controller
--- >... ■ «>------ L

33Mflop Macintosh nfx
Digital Signal
frocessor «-
Figure 3.3 Schematic of the Digital Imaging Velocity Measurement System
42
Figure 3.4 Flow due to the test cylinder in axial oscillations; A=6.35 mm,f=0.5 Hz;
(a)-(h) equally spaced over one cycle starting from top dead center.
43
Figure 3.S Flow due to the test cylinder with disk A in axial oscillations; .<4=6.35
mm,/=0.3 Hz; (a)-(h) equally spaced over one cycle starting from top dead crater.
44
Figure 3.6 Flow due to disk A in axial oscillations; A=6.35 mm,/=0.3 Hz;
45
2 0 m m /s e c
i ym > i n v m T v r n m 28
x /f » * * » * » a a a a a a aaai »««« a «« ! / / / / • ■
»«««« aaaaaaaa f t t t • • • • • • a a a a a aaaa «al! / / / / • •
aa a a a
a a a a a a a a a a a a a a
k « f «««! %%%%%%«
aaaaaaaa
f t 9 9 9 ........................
if t t *
aaaa
• II!/////■
« aI ! / / / ✓ / -
I U I I / / / / * •»••%%*•%%*»*%*• •»•
f t 99 9
a a a a a a a a a a a a a ..................... I I I I t I I I I «•
_______
UOJLJUJLajULLLftA n » » > ................................
««««««««/f 99 • * * • • • • • • * < 21 - '
. i i m n i i i
m i n i U \a
• II | \ \ | ! • - • • • • • • • • • • • • • « • • •
• M i l l III
• *A 9 9 ............................. a* I • ^ / / 9 t i l l ......................................................................
► %k < k \ \ a v s > ^ •— » • I f | «• • • • • • • • • . • • • « • • • •
» / / /
I X%a '
f I I #/ / / / ✓. *■- • aal M l
I Iff l ll ll l» « i ll ll t« « « % « « « « * * * * « i4 - r a a a a a a I « a •
l l l l l l l l a aai a a a a a
lit* a
>111111 ' III a a a aI I I
a a
I l l II *«•«%%•»«••• a a a a a
III! a
>9 9 9 9 9 «««««%•• a a a a a
III! a
t l l l l l * ! »»•»*•**%%*•*•**•••••*••
f i l l .................................
....... I I I I
aaa «!««%«•« I I I I a a • •
«l t l I I *••»««*•»•%«• a n » «« • a i aaa*
• ■ * %%%• * * * %%%%%%%%%* • %» • • * • “ * • • a«a • • • « • • i / i %%i a a • •
aaa • a*aaaaaaaai
aaa aaaaaaaaaaai
'9 9 9 * 9 9 ■‘J ' l 'l « I I <M M O 'k'H U U t 28

B r / / / / / / / / «t / / •
I I I / %* « %• • %%* «
I If ! • • « « * « « « % « « « «
................................................................. \ \ \ \ \ I I I I H * * * « k ......................
I I • I I M I !««««% \ \ \ V V* 9 999 9 9 ••» •••••« « •
aaaaak%a«a\a«i
« « « « %» « « « « %» « •
r ................... / / •
///••% •
f 21 / ^ • « - ka a a a
* S S ' 'S .
............................ 1j * ■ ~— . . . . ka a a a
i a a a a a a a a a a k c”* « '/ A - ■■■ •
' 1l///^//>V*A>V> f ^ ^ a a a a k i m i • a a m i a a a
J^^X X a a U K i n i a a m a a a
jk ^ N a a a t i i i u i t i a a i a
99999999999999 Va U n i i i i i i i i i i a a a
a aaaaaaaaaaaaf m I I i i i i i M i a a i i a a
14- a | r**aaaaa«aal
aaaaaaaaaaaaaf i i i l l l i i i i n t i i a a a a
%a ) a a aa a a a a . H M l H i M M i f i M a a
. . . ( i f l l l i i i i i i i t i i t a
a a a a a a a a a a a a f I I / / / 4 / / I M i i i i I i m m i aaa
ka * a a ^ / / / / / / i i l l i i i t a a i i a a a
\a a a a a a K S \a a r^ w / / / / / / / M i n i i 11 i iaaa«
7J ka x a a a a a \ a a » ^ ^ / < > / / / / / 1 aaaaaa
la a a a a a a v a a - ■^'4’^ / ^ / / / / / 1 1 1 ^ a « t a a a a a a a a
i.a ~ » » ^ a a**^»
9111/11
//••aaaaaa
T T v r a a-a a a a a a *
a a aa
28-
r/
r/ ✓//✓//✓/
f 4............................................
4 44 4 44 4
.a W W a a * * • l « a a k * > l
.v M N X v a i i * - I f f a a a a
I M I I a > k *
a
a
a
a
a
a
aa
aa
a
H /
f * 99 9
I f / I I f / / / ^ 4» ^ « a a - * % % \
a
M
M
M a M i i i i i i i n i i i i i i
a M i 1 1 1t m i «i i i i ii
%l I I M I I I I I I <1 I I I I
f44 444 44kkk\ » \i I I I I • % • • • • a aa f / f f f 9/ f XX I I M I M l I M 1 I • I I M II
' / / / I k l l l k l l l i / ’- x V V \ ia*l lI fI ff *f I I M M • • • • a a 1f ft %
t fr 9w • r r v v* M I M M M l l i i i i l l l t k
i i n t m m m i —* r j I I I f l l l l f !> ••• 9 9 9 9 9 9 9 9 9 9 9 9 f 9 f Xl
TTTTTTTTTTTTTT I f I fI •
21 I I H M U I I l i p MA
aaaaaaaaaaaaal M I I I M I ai
aaaaaaaaaaaaa! h i m M II I M f II aaaaaaaa aa aa a f
aa ••«a^l a a a a a a a l I I H I I I I I . taaapfâa aa aa a f
aa a ■• ' I a a a a a a a a >a a f « I a a a aa aa a a
•I ••aaaaaaaaal
TTTTTTTTTTTT H U M I I I I I I I H
f • • • “ H k l k k l l l l/M/MM I I MMi «it
//% | / / |
M l I i« I II I II
aa k « « h V U \ W v / / | I M I I I I l l l l l l l l * 14" r. »a «a »a »a aI aI aI Ta Pa aI aI aI Ia fI - / / 4 iff t * I i n t l l i i l i i i
\ s a s w v ^ / / m u m i l l l f l l l l l • f a k\\\\99999999 f I / / / / / I il /fl i l iI Im Mml I nI I
x n x x M I H I I I H l l l f l l l • If*
a a >a a > a a , » a . , a / / / i i i i m i I I I I I I I H I I • • - \ \ \ \ \ X \ \ v a j^ W / ^ / // // // // 1 I I /i tt •
a i v \ a a k . * . * * . » * / / m i i i i i i I I H I I I I I A \U \S W V ^ " ..............................i n n
a a ....................... h u h I I I H I I * * k X \\\V \W W / - / W / / / / / / / / I I M M I M I M I
v w v v ^ k k k * * * * * * * * / ! h u m H H I I I * * . iX X X K X S w w W / / / / / / / / / / ! • • • • • • « • « aj
kVa\X\aaaa«a«»*a*l I / l l I III H l f H M I IX a \ X \ X S X a j k ^ — • ✓ ✓ ✓ / / / / / / / 1/ / , i i m i i a i i
H W \ \ W »" ■ ■» - W ^ / / / / / | / / | » | a » | I a I II
..................... • ^ ^ ^ / / / / / / / ! • • 1 1 a» a a a a
k a a a a a a a » » a » ^ » a » ^ /» « » * » * « a a a a i a ^ ^ a « • • • a
28-
►a a a a a a I H I H I f f H M I M a a l a i l I m u ii i i • «• • • H
}r / / / / / / / / < * ✓ .
r4 4
r94
444 4
4 4 / 4 4 </✓-'*•
lVV X \ * * a a a .
- \ \\ \ \\ - — ►a a a►
i » k* a a •
»a a a a • •
aaa a • ■
nrrrv—.............
. I f f f t f f f f t f f t
M l V l l l
aaaaaaaaaaaa
f f f
M
af
...........
k M I M I M i i n i i
I k l l l l M M l l i t i
ai I I I I I M M i i i i
••••««
i k i l l l l i
11 1 i i i ii
t44 4 4 4 4 4 4 4 4*~ - a X \ S \ \ X % ™ - ► •* • a —a a a f a a a a a a a a a f a i l l l l l l l l i i i i i i i i i i i i i
fk 4 4 4 4 4 k M l i « *» \ S N \ ' V « ' « I/ f— 2 ] ^ aaaaaaaaaaaa af ai I I II I I I I n 11 i ti h i i n
U M U U M k • I I \ \ \ x x a --------- ' / f a aaaaaa
11111 i n \ \ ^ k " I \ \ \ \ \ X a - - I l k a a • • * a
m m V xvaxx a a a f \ \ \ X X k , * k H a aaaaaa ' a I / • / ’/*/•✓ I M i m m i l l • • t
H I M M X M \ a Xa a 9 \ \ \ \ X X I a 9 9 9 •/ aa a • ■ I * * > / / » a l / / / / / / l 1 1 1 11 i « • « aaa
■ H i m i m M I X t f XXXf f 9 9 9 •/ / aa• 1 11 f f / / / / y ^ . > • • • 1 / / / / / / / / ! ! • t •« « • • aaa
H i l H i H i i i « al t \X « | III 9 99 f / / / •• M l l l l t l f t r f ►• i a/ ’//•✓ /’/'✓W /’t « « t • • « • a aa
a a a a a a a a a a a a kI al t f m M 9 99 / // /• •
14- -*l ( I I I I M I - ►•I «• « « M l a aa
a l N a * a a a a a a a a* M t f M M I 9 99 H M I«
aaaaaaaaaaaa ak M t l H I I I
a a » » a a a a a a a a aa a M I H I M
9 99 H IM *
9 99 999 99•
k>n(UUU-' ►//✓✓Z/////////// «« k / / / / / / / / / / / / / l • t ••
w / / 1
/ / / / / / / / / 1 • / / «
>« • • a a a
a • a aaa
aaa aaa
1AA. * » k / / I I H H I 9 99 H IM * v « v ////ff //« % .» * * ■ • «a a a a
k a a a a a a a a a a a a a aa/ / / I H II I 9 99 f/ 99 •• ^ / / / / / / • • • • • % a aa• aa a a
iX X \x \X \a a « ^ ^ / / / / / • • • * . • • • • • • i • aaa
iX X X X X X a a a — > •«
ia a a a a a x a « A M aa aa
ix x a a a a w aa a a
10 20 30 37 0 10 20 30 37
Figure 3;7 Velocity plots obtained by single exposure PIV for the oscillatory flow
due to disk A; A=6.35 mm,/M).l Hz; (A)-(H) equally spaced over one cycle
starting from top dead center; x and y distances in mm
46
Figure 3.8 Vorticity contours obtained from the velocity plots of Figure 3.7; vorticity
magnitudes in 1/sec, x and y distances in mm.
47
Figure 3.9 Flow due to disk A in axial oscillations; A=12.70 mm,/=0.5 Hz;
48
40 mm/sec
one m u m » i M tlV IM O i i s i w i ii 28-
b* • I
b • ft
I II I I
0 0*4, ,
+ ' ft »ft »»»ft ft f t ft *ft f t f t f t f t f t f t ft *
■ftXft ftft ft ft ft-ftft****ft«««« E
0 0 0 ,,, I kftlftftftftftftftftftftftftftftftftftft*
• • 99 I I 9 9 • « k«l f t ft f t f t ft f t f t ft ft f t f t ft f t l « « « « « »•) ft« I 4 4 ) I I I )
ftftft « • I > I ) b I • • » « « % « » % « « * % « « % « • « % » * ft ! ix
I %ft •• • M l * -ft ft ft ft ft ft ft l ft ' b
k«« ««««•«« bf t f t f t f t f t f t f t f t f t f t f t f t f t f t
k* bftftftftftftftftftftftftftft ................................' ■ W ^
ftft« '•ftftft
' ftftftft . mmmmmm^ VVv o .
• « • « • • • « ! I \ XXX
•>%»>bftft ft ft • • « • « « « • « ft »ft ftftI
ftft\\\\»»^» • • m i ft ft • « # « m m
• • •*•»# « * \ f t f t f t \ \ \ W « - * • • » m « «• - > - • i i / n m i i
kl« • • « % ■ # • # • « « # # • 4 • • «« • • • t i i i i m m
•« ftftftft# « | l # « • \ H I M I W M • 9 0 0 0 0 • • « b*-i 1 1 1 1 1 m 11
»I
b•
b•
ft • • • • • • • • f t ' MMfMMMM
• t « • M b n • * • ) %I 4 ) W K X X X V
« ft«<lt«ftft*' • • • • • f f t f U P t
• # # 4 4# • • «
•«*)•/
ft• %4 4 • .
9 4 4 9 9 * 9 94*
•• « • » / ! I I I M l I
I «# • « I M
b ft ft « % « « « • « « • ' • < • • • • M I M M • 14 4 4 0 • • • ft ft
b ft ft ftftftft* .« « « « / / | , $ , f , %4 4 t t * » » # * » # • « # • • ft• •
ft ft ft f t f t f t f t ^ " * * ' • ' - - — - • 4 4 9 9 9 1 4 4 4 * v ~ t* » • ' # 9 0 9 9 •••«
ft • ft ft m• « « « « « / / / M M
i
/ / / / # ^ 99 !#«•••••
b « ft
• ft ft ■ • i •
• « •« « « •••» 0 0 * • • • • • • f t f t # — * * * p 4 f t f t 4 f t l 4 4 l 4 1 1 ft' 4 4 4 f t 4 4 X 4 4 |
• ft* < 0 * ft* i 9 f l l k*f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t f t <<i
v b b C Tf b u e u i i « T b r i n r bT » * f t i m t
ft—• ft ft f ftft • » • | « ft* ftI ft ft* ftVftftftftftft^ftft
## # - ►» » 9 i * # • » i i f t f t \ i * « f t f t i \ \ f t f t \ m \ f t
I 11 ! % • • # # I I f t f t f t I f t l l 1 9 9 , 9 0 00
* # / / • I I f ft ft t I I I f t f t X f t l f t f t l I X X X X X X X X X X X
* 4 4 4 * * 9 M M f t i « I I f t f t f t X I I I ft ft f t f t l XXXXXf t XX « ft ft • « ft ft k% «b% 0 0 0 * 0 0 0 0 0
' / / - / / • M M « % I I I I ft XXX I l l l l b l i x i f t XXXX - < ( « « • < • - «
» < • • / M M / l l > - « l l f t < l \ I M f t l f t \ l l I Xf t XXXX
-21 ft « % « « • % * % —
* • • • « • • « • • #
mb ft X 1 *
9
-XX* * 9 9 9 9
I M I » / / / / / # * • XXX ft ) ft ) ft I I ) ft 4 t I f t I X X X I X X « • • • • • « • • • • 0 0 * -XXXI 99 9 9 9
9*9 9 9 9 9 4 4 4 * XX X ft 4 ft) I ft ft 4 ft ft I i x x x x x i x x I f t l I l « # t I ( ( • 0 0 * , * « • ft ft ft *f t f t f t X X X X I I I
9 9 9 9
* 4 9 9 9 9 9 4 4 4 ^ * ft X \ I ft ft ft ft ft ft ft ft ft ft f t I X X X X X I f t ft M ( I « ( # # # # • 0 —* • i i i *9 9 9 9 9 9 9 9 9 9 9 9 9 9
* 9 9 9 9 9 9 4 4 4 * 4 % I k f t f t f t f t f t l f t f t f t l f t l f t X X f t X I ft f t • I « • « • • « « # # # • — — + mm ft I a 1
*9 9 9 9 9 9 % t t 9 * 9
>999 9 9 9 9 1 9 9*4
ffftftftftftftlftftftftftftftftftftftftftft m m i f I M % * - * 1 ft ft • • •
9 9 9 9 9 9 9 9 9 9 9 9
ftftf tf tf tf tf tf tf tf tf tf tf tf tf t
9 9 9
t 9 9 9 9 9 9 9 9 | I# / • %%• • • • • • • • • • • • • • • • • • « 14- • l « « « l / M I 9 t
b« • • • 4 ft * / / I 9 t
bJLA I J J U U
4ft ftft ft ft « b ft ft ftft ftf t f t f t ftftft«
9 9 9 9 9 9 9 9 9 h i Mb • • • • • • • • • • % • • • « • • • ! b « ( ( « l l | I M 9 9 r 4* f t * ) • b ftftftft f t f t f t f t f t f t f t ft ft !
»9 9 9 9 9 9 f t 9 9 4 I I ft t • bf t bbf t bbbf t bf t f t f t f t bf t f t k ' M l l l l l l l l i ft* ft ft ft « bftftftft f t f t f t f t f t f t f t ft ft !
9 9 9 M M | I M I
• 9 9 M M M M i
ft %ft • • • • • • • • • • • • • • • • • • I *
ft ft« b b « b b b b b b « b b % < b b b b b <
• 1 ) 1 1 4 1 ) 1 ii >• • ft ft • * bft ft ft ft ftft ftftft ft ft* ft!
) 0 •I IftIIft4ftft ftftf•t kftft ft ft

b ) ) ) ) ) ) 4 ) ) l 4 « i ftft ftftft « b ftftftft f t f t f t f t f t f t ftftft!
t • 9 9 9 9 ft I X X ft « • b b b - - - - - - - - - - - - - - bft« « ) ) • ) ) I ) 4 I bftft ftft f t f t f t f t f t f t • ft ft !
' • 9 9 M l MXXf t f t • X»X> • » b » b i I .*■■» > ■ ■ ! ■ ■ ■ > f t ) ) ) ) ) ) ) ) ) f t ft ftftft ft
• • 9 • M MXf t f t l — •
0 0m
( • • • • • • • • • • • • • • f t ft %X - 7- b*»1 1 ) 1 1 ) 1 ) ) * ftftftft ftftftft
>» ft t f t « « f t f t f t f t f t f t f t f t f t ft ft ft b—ft) I ) ) 4 ) I I ) • ft • f t f t XX 9 9 9 9 I * ftft-ft ftftftft
><! • ! • • • « « %m / /m m# w f t%M w «M •(« « b «« b ftftft • XX 4 1 4 ) 1 1 ) « « — f t XX 9 99 9 I 9 * * 9 • « I • ft
l i t ! 0 0-0
00 0 0
l « • • • • • • '
0 0 0 1 I * « • « mmm
r * * w \ 4 4 4 * * t 4 X ft 4 ft ft ft «
r < r * i f t / * # 9 9 4 ft ft ft • ftftft
' - I X X I X X 4 4 4 41#
- ft—) ) f t Xf t X 4 \ *
9 9 9 9 I 9 * 4 9
I I I 9 9 9 9 * 0
9 * * *
• * * *
w # # f t M t / m t ft«« I*— — ■ X4XXX ) I i I I I 9 9 9 9 9 9 9/ I #
b* • * > » > » » l l » > > < < • > > » ft • • — — -XXXX > »* >« I I I 9 - 9 9 9 9 I 9 * *
■ ft 4 1 nrr a i x r*» 4 41
- ftl I ) - • • « » )
28- H T D U m D e U I * 4 4 *
ftftftftftft«««ftftft«««««a
4
aftaaaaaaaa 0 00 0 0 00
0 ft ft • ft ft f t * ft ft* « f t f t f t f t * * • n 0 m 0 0 0 0 , , 00000 0 , 9 9 0 0 0
• « | • ft I I ft • ) ft %I - « « « « « ■ « « « « • • « ! ! I I I I i I I* I I 0 0 0 0 9 9 9
«• «M
«II ft i n
ftiiftftii
) ) ft
ft 9 9 ft ft k • « « • «
• ft • * • f t * • —* • *
i b i n n * —* * *
! « • « • « # # # — l
ft* * ftft* • ■ • • • • f tft ftft ft ft ft ft ftftft
M i l ft« » I »» »
ft 9 ft ft « » • « ) %«
- 21- : • • • • « « # # # * - l
• • « • • • • * « * # !
- •* — / / / / - - - - « i ft I I I » I ) f t ) ft ft M « •« I I ••< I I
* / / / / / / / # — > * « « « » « • « » « « i ft ft ft ft ft ft ft ft ft ftX M I l« I ft** M l
•*/> I 9 * * 4 r — b b « ft ft ft ft ft ft ft ft ftX ft ft I I I M l • • • • I I
9 9 9 9 9 / / ^ - - - *' • « — » » » » « < « • ft ft ft ft » » f t ) « I ft I M ( I • « ! * f l f ______________ __
• 9 1 # I / / - » «•» , « | ft ft ft ft ft ft « » ft i ft f t «
r / f M > M * * » « » * * * X X ^ X X • ft ft ft ft ft ft ft « f t « « ft ft ft
I I I I I I • - I X X X x —* — — —* * * * ft
jLJkJI
ftftftft
ftftft*ftftft
ftftft
-14- bl I • • « * • « • * *• «
MMMMMi
M M M I f l * * — - x x x x x x n ft ft ft ft ft ft f t t f t x x i f t
! • " f t ft | \ f t ft ft ft ft ft ft I ft ft ft | X » | ft
M M M ! • * • « >11 * 1 4 ) 1 ) 1 ) 4 1 f t f t f t ) ft « ft ft ft ft x
k • ftft! « I * • ! • • • • * * * m \ ^ \ X X ) X • * • b I I I I • •
b ■ ft ft • • • | ! « ! ! » » « * * ^ ^ X X X I I « I X 9 , 0 0 0 0 0
i l l
i f I f l f b , Vk «
• «« ft ft ft ) ) I ft 4 ft ft ft I ft ft ft ft ft ft ft ft
• • • < ) f t ) l f t ) 4 ) ft ft ft ft ft ft ft ft ft ft ft
b i l « ) l # • x x X X M I 111 9 9 9 9 9 * 0
b « « ) 4 ) ) 4 # — * * * . » * s \ X 4 \ X ) ft I 4 1 ) 9
• • • 4 4 ) 4 4 \ 00 0 0 v - V b \ X I X) « I I 9 9 9 9 0 0
99000
n t f f f b « i « « « > # # M M f t < f t l l f t ft ft ft ft ft ft ft ft ft ft ft
># * < < I I I ) ft ft ) ) ft ft ft • ft ft ft ft ft ft ft
M l f t f t l l < * « # < * * * * 9 4 4 ft## ft ft ft ft ft ft i I I I • f t * - 7-
t i « l 4 l ) 4 l ( » i i i b •* • I \ \ % 9 9 9 9 0 0 0
9 990,00
M M M M « ( f
• • • « M M « « i
ftftftftft’ftftxft
ft ft «ftftftft«ftftft
• • H i r m
• •••i % • •• • •
m
9 99 9 0 00
9 99 9 * 9 9
9 9 9 9 9 9 9
l i i i f i f f i . , ••ft f t f t f t f t f t • •««« «ft «ft ft 9 9 9 9 990
• • • >«% 9 , 0 * 0 ft f t f t f t f t f t f t f t f t f t f t %ft • ft %• 99 9 9 999
0 0 ift.ftft M « «
«• «•»• «IM»
T
• ft*ft ftftft ««ft ft
• • » • • > • • > • •
~ r
ft ft ft ft ft *
• • • %»«
T
k ftft f t f t « % 4
>ft— - f t f t f t f t * —'XX 4 —
4 / 9 9 9 9 9 9 9 * 9 * 9 9 9
* 4 4 9 9 9 9 9 9 -*9#1)4
1■■!" >'
18 27 36 0 18 27 36
Figure 3.10 Velocity plots obtained by double exposure PIV for the oscillatory flow
due to disk A; A=12.7 mm, ^=0.21 Hz; (A)-(H) equally spaced over one
cycle starting from top dead center; x and y distances in mm.
49
L ) y ?
-1 4 0
280 r 2W ^
-100
-2 5 O X } ) 50
X -50_Vv^^_-150'/r jr'°
0
-16C
-2i O
m
-so / ; 120
Z?\ ^ - ^ —0
-120
M y g
■Mffr-420W1I M
801 v-i202animi
o- ^ * ii«,r > (/// \U
oo r
—80
^ 120Py. f/ZS
/ 1\v y it oAc -1 0 0
-200
o v/^y-ao
120
.Wf240 f t
1---- 1---- ^-----1--- i f ^ I-----1-----1
36 0
Figure 3.11 Vorticity contours obtained from the velocity plots of Figure 3.10;
vorticity magnitudes in 1/sec, x and y distances in mm.
I
i
Figure 3.12 Evolution of oscillatory flow due to disk B; A=3.2 tnm,/=0.5 Hz,
initial condition=BDQ (a)-(x) equally spaced over five cycles of oscillation,
starting from rest
!
51
Figure 3.12 continued.
52
Figure 3.12 continued.
53
Figure 3.13 Flow due to disk B in axial oscillations; A=6.35 mm,y=0.3 Hz, initial
condition=BDC; (a)-(h) equally spaced over one cycle starting from bottom dead center.
54
100 mm/sec
7X
II | I I 9 9 9 9 9 9 a a a • t
9 9 9 9 P f - m m m m . 9 9 9 9 9 9a • a a a
9 9 9 9 9 t * *— • » • « « a 1 1 1 1 f9 9 9 9 f 9 9 9 9 9 • 9 9 9 •f
9 9 9 9 9 9 » ' —» M 9 9 9 9 9 9 9 9 9 99
9 9 9 9 9 9 0 0 »• «%« « M l I 9 9 9 9 9 9 9 9 9 99
' I 9 ft ft 9 ft ft « - • - •• • M l H I I K t l l 9 9 9 9 9 9 9 9 9 99 21--------------------------------------- • • •
»••«! I I M I M H I 9 9 9 9 9 9 9 9 9 99
n i f m i m #« 9 9 9 9 9 I MM I
>999999999** » * » « « MM( I H1 ( 9 9 9 9 9 9 9 9 9 99 9 0 4 m
M M M f M M I M I f H H I 9 9 9 9 9 9 9 9 9 99 9 9 9 9
> » » « I I M I I M I M 9 9 9 9 9 9 9 9 9 99 9 9 9 9 9 9 * 4 4 4 4 / + —
i M t l U I M M • i •••! MWMMH 9 9 9 9 9 9 9 9 9 99 9 9 9 9
t l l M M I I I *
I 1 %I XXXXI *« • ••# Wiff!* it 999
• •* « M I < j / M M M 9 9 90 9 9 90 0 9 9
9 9 0 0 0 9 9 9 0 0 0 ' 14 9
fMM
IMM
9 9 9
k
—
9 W l * *
• > * - * * 4 4 4 4 1 * 9 Mi 9 0 0 0 9
499 0 0 0 0
00 99 0000
9 9 4 %9 9 0 0 0 0 0 0000
0 000
1M I I iiiiiiiiiiV-iiHJ-
H I M I I I I I I I I l> \« M 1 )
\1MMM
tlMMM
1 ^ — ' ^ ✓ / / / / / / # 9 • *00 0 0 0 0 0 0 9
- + * * * > * * * * * —0 9
II I I I
i n !}K^JW w • MMMII
0 0 000 0
II I I
I I %I i i i M M U
M i l l IV M S W -,,/)!!
M i l M i l l
i IMHIIf
t U H I M
II 11i *I i
| | I I I I 9 9
M i l l m i m m m
M M I ^ / / i
i l l l M i l----------------
I / #1999 99 99 99
M i l I I I 11 I X X X » ~ — * * ///9 #f t 9 9 9 9 9 f 9 9 i
i m M I I I I I I I ///^ / f f 9 9 9 9 9 9 9 9 9 9 '
M i l l M l l l l l \ < r .................................... -
TT TTX*T7TrTTTTTTTTT7TT7T)rTrirT
B ft • 9 » * * %%« « 0 0 0 9 9 9 9 9 9 9 9 9 0 0 0 0 9 0 0 0 0 0 0
M» * * - • • • • # # # # 9 9 9 0 0 0 0 9 9 0 0 9 0 0 0 0 0 0 0 * 0 0 0
••••0 —0 * 0 0 0 0 0 0 0 0 0 0 0 —0 * 0 0 0 0 0 0 mm*
!!«• • • • • • • 0 0 0 0 0 0 0 0 0 0 0 — 0 * * 0 * 0 0 * 0 0 0 0 *
I !«•««• m m m 0 0 m m 0 0 0 0 * » * ««•••••••*••
t 9 0 ■* » * « > » » » > * • • * * 0 0 • 9
Mil" mm m m 0 0 0 0
M I « %I I I I X W V \ X w \ * ' * >* * * » » » • • * • « • • • « « «
!««»•! ( I M M •••« •••••.«
i *M M I II I ..................................
14* %•«««• » » « * M M I I I M * * M * < « < « M M I I * « « a
•l*M9IMM«k*»k««««»ll»lllf«
94 t * * * * - - »«««•«««« # # « M M M M I M M I M M I I I I ( I I «
i M I / / / ' - ’*
i MM 4 * * * * * *
9 9 $ 9 * 4 * * * * * ’>
• - - w * * 4 9 9 9 9 4 4 / / * * * '
— — * * * 9 9 9 I 4
»» f »» » *f » r 9 i ii i i i i i i i i i t i t « i i i # « f f
« ft I ft ■ I I mmm
- > > » » > * ”
* * ///9 |
» » m w » m m 0 0 0 m 0 * 4 I l « ft I ft • % % « % « « • • 0 0 0 0
• 0. 0 0 0 0 0 0 0 0 * 0 0 9 9 ft ! « « « « • • • • 0 0 0 0
|* — Y /7 w ^ iT r r fr r iT r n ttttttttits
* * 1 4 4
- * * * 4 4
M /r ? /r ^KxV ! ! 5* 5!! > ! 5!! 11 • 21
- * * * 4 4
28- •»■1 * I'l I l"l I l l l
l i t MMI III lll lll lll I l ll
I l l l
in tv
1 1 1 • • ft ft ft ft a f
t•ftI
HftiMMI I l ll 111 •« ft II ftftft ftftI
X\nnKK\\i««%I
S\V\WMM MMM M il II ll ll ll I l l l 1111 «• «tII IIfItII ll ll ll
1 1 1 1 1 If
A AWUVVU9
4 AWUUUMMII 9MIII I l ll
Il ll
I l ll
I l ll
I l l l
11111
11111
II II II IIfItII ll ll ll
21- IAUNSUWNSI MM Illl I l l l 11111 I I I I I I If tI
U U S \ U U \ 9 I 1 * 1 I l ll I l l l 11111 I II II Illl
! \ \ \ N \\N \\\\7 M il I l l l I l l l 11111 I I I I I I l l l
I l l l Il ll 111 11 I I I I I I l l l
t u n m t u \ u * 9 i I l l l I l l l 11111 I I I I I I l l l
M i n m u i i • I l l l I l l l 11111 I I I I I I l l l
t MmUMUNUIi
1J r f n
#«\M M MM M Mf
i4* M9MMMMM
MMIMMMM
i u n i M
XXXlr
IX-
M x x x i
• • ■a a a a a a % a a * « «
r» 1 1 1 1 1 1 9 1 1 1 9 % >
!/////V V V /i{
Z A «^^nnnv \ \ x k \ x 11
» *««n vx\V X X \\ \ \ I
MMMIM
MIMMM
'9 9 9 9 9 1 1 1 1 1 1 IMX-
-> 9 9 9 9 9 1 1 1 1 1 1 1 1 « % -
f 0 009 9 9 9 9 9 9 9 9 9 * 9
• 0 0 0 0 4 9 9 9 9 9 9 9 9 9 9
’ 9 ft ft ft 9 I I I | | I •
;??«««< W W \ \ \ |\ M
* * W \N \\U I
t t " I 99 9 9
l i t ! 19 9 99 IIMIHIIIII tlllft!< • • • 000049 9 9 9 9 9 9 9
• • • • 0 0 0 0 0 0 9 9 9 9 9 9
t 4 4 4 4 4 4 4 i
fi 4 4 4 4 i i
I
' * * * - - T I • M M I ft ft ft I ft Mftft<
• 9 « f t 9 9 f t f t f t t f t a « %%%.
• • • • • • • f t f t f t f t Itftftft
• • • • • 0 0 0 0 0 0 t i 9 9 9
• • • • • • 0 0 0 0 0 0 0 0 9 9
H
m
H
a
i t
i l
9 * * *
9 t 9 * IM M I I « I ft !««%«%• • 0 0 0 0 0 • 0 0 0 * * * * 0 0 9 9 9 0 0 9
•000000000000000
Ml M M 1 9 9 9 9 0 9 9 9 9 9 9 9 0 0 0 4 • 0 0 0 0 0 0 0 0 0 0 0 0 0,0 0
I < ( I ft a )
1111111 JJJI
I »I I >I I M U
28 M i i i ) ) 9 9 * f t >i m 1 * 1 1 i n m 1 11
D ■l t « l « l l >9 9 9 9 9 9 9
. 9 9 9 9 9 9 9
9
9
9 9
/ 9
/ — - 1 M M I I I M I M I I I • « M « • ft
/ / « « l I M I I I M I I I M I M ft M « ft I
H
■
•••# I 9 0 — — — — — ^ —>' tl««»MIMII -> 9 9 9 9 1 9 9 9 / /9 / « « l I M I M I M I M M M «1 « • • « .
0 0 0 0 0 0 0 > 9 9 9 9 9 9 9 9 9 / / * — i MI I Ml MMI MMI f t l f t f t f t f t f t
••f m » o e ’ 9 9 9 9 9 9 9 f ? ? 9 « « 9 / 9 I 9 I M M 1 1 1 1 M ft ft M f M l
• • • « ///* * » * » * < a l 9 9 9 9 9 9 9 / 9 9 9 t 9 M 9 M 9 M M I M M M i 9 i i M M
-- I I\ \ \ l I M I I I I
-21
• # 9 9/ / / » — •— • • • ■ •«AVNNVNNV\^\\\X\I 1 1
•# 99 999<f ///« •» ••» « .
* * — w. 0>+. 0>0~»
— I 9 9 9 9 9 9 9 9 f 9 * * * 9 9 9 99 9 9 9 9 9 9 9 9 9 9 9 9 • ft • ■ • •
\ I \ Ii I 9 9 9 9 9 9 9 9 9 9 1' 9 94 9 99 i t 9 9 9 9 9 9 9 9 9 1 a « ft • ft ft ft »
9 9 9 9 f * * — — »m. v.vv\\S N V V W 9 9 9 9 9 9 9 9 9 M < 9 99 9 9 9 9 9 9 9 9 9 9 9 9 9 ft ft a • ft ft
ft » « ft
r9 9 9 9 l\ 9 9 9 9 9 9 9 9 9 I - * 9 99 9 9 9 9 9 9 9 9 9 9 9 9 9 9 » » • • • • • •»
'/I 9 9 9 44 4— — w .W W W U U 9 9 9 9 9 9 9 MM< 0 9 9 9 9 9 9 9 9 9 9 9 9 9 I ft 9 • » * * • » • • •
r/i 9 ( 4 4 4 * —
9 9 9 1 4 4 4 4 0 — -14- 9 9 9 9 9 9
9 9 9 9 9 9
MM« « 999 9
0 9 9 f i t
9 9 9 9 9 9 9 9 9
9 9 9 9 9 9 9
9 9 ft ft a * » « • • « ft *
9 ft ft ft a ft * • • • • »»
9 t m / i • 1 9 9 9 9 MM« > 0 i f f i t 9 9 4 9 9 9 0 0 ft ft ft ft ■ • » * ft • ft ft
9 9 9 4 4 4 4 9 * * * * - ’* 0 - 0' -' • • 9 9 9 9 MM*' 00 9 9 9 9 9 9 9 9 9 9 9 9 Vft ft ft • * ft « • ft ft *
9 9 9 4 4 4 9 * • ■ • 1 1 %I I ■•« • m f 9 9 9 9 9 4 9 9 9 9 0 ft ft ft • • * • • • • • *
9 4 4 4 4 4 4 9 9 9 Z / / / / z ^ » ( » • ( * • ft 1 •«< •0 9 9 9 9 9 9 9 9 ft 9 ft » ft ft ft • • » • • • • • »
9 I 4 44 4 4 9 9 9 4 4 4 / / / ^ ft ft • • ft • » » » »fti 9 9 9 9 9 0 0 ft ft * » ft ft ft • • • » • • « •
ft I ft ft ft ft
iI iI 4 4 4 4 4 9 9 4 4 4 4 4 4 / * -
- 7- ft ft ft ft« «
ft ft ft ft ft I
ft »%*«<
• • • • • • ft !%««<
9
0 9 9
9 9 0 0 0 9 ft ft ft
ft ft ft %ft ft ft
ft ft ft ft » • » • • • • •
* a « * » *»»*
I I ii I u i i u r n i s
I I l l l l t l t t l l l l 1-fcV.
ftftftftftft %%%•• <
«««»**
a • a • a
TTT a a a a a M M
a a a « « a « %« % % % « % a a a « a a
I I U U I I M U w ft 0 • • • a • a a a a a a a a a a a a a
ii
I I I I I I II I MWV m m m ft 00 * » • • a a a a a a a a a a a a a a
—I—
10 10 20 30 37
Figure 3.14 Velocity plots obtained by single exposure PIV for the oscillatory flow due to
disk B; ^4=12.70 mm,^=0.11 Hz, initial condition=BDC; (A)-(H) equally spaced over one
cycle starting from bottom dead center; x and y distances in mm.
55
60
20 - 21-
20---
40
- 14-
80
-120
-4 0
-20 ;
- 21-
- 14- -9 0
-3 0
-20
- 21-
120
-25
21- 120
-2 5 -
25 100
- 14-
60
-2 5
-7 5 20
-1 2 5 25
56
Figure 3.16 Flow due to disk B in axial oscillations; A=6.35 mm,y=0.3 Hz, initial
condition=TDC; (a)-(h) equally spaced over one cycle starting from top dead center.
57
200 mm/sec
"75= iI ftftV W W W ft XI ■
M l 9 r r r r r r p ^
% % 1 I t 9 9 f f t f / ’» * - • * « » » « ■ > <-<
• IM SiiSS
• • I • 1*1
I f f ««»! slUuinxtrSrFflsTHi Hf n
I- . J .*. > ^ > __________a a M
a ai ■
l «
l .k
M / l W / / / M W
•l/Z /W W /A V V }
««%!
t* l I I I X\ M I %l
• *m I M \ U I IM I
X X X X \|l\X X X X X m A X X m m p # 1 *
*<p
I l l l9
• - 21“ ** **S'*f 9f f9 f9 f9 f9 91 1n1
* • • • % • # #' ,.X X K \I\|\K X X X A X # m m m i IM
XX XX XI< |XX XA *X # M M H M IM 9 / 9 1 9 * 1 1 9
1999919 9
%• • «
########
m X X X \\l\X X X X X X # I H I I I I I
i # •H
I M f f1f19t1t\)
t119*t\t\t
1 4 -! 911##\SN\
. VSv
«• • « — XX%\f t f M \ X I I I I I I I H I M M
• % X X \\\f 1 I M M M I I I I M I I I M
%« ft « • W X ^ M u ( f 9 1 1 1 1 1 1 1 1 1 1 M M “ ___
nsS ss:
I I I fxmiKX^VW**'
O i l • M M M U t n f 99 9 99 9 99 9 9
% ••! } M l ? i f 9 99 1 1 1 9 * 1 9 1
I I M
M M
• III XXXSWksv*
I f # #X X \S N V \\V W ~ , ///✓✓✓/-'
•Xft ft
I %I I
« . . | 1 * * 1 1 9 9 9 9 9 9 9 * 9 9 9 9 9 9 M M
| ift
• M l t ^ f l * } } 1 1 1 1 1 1 1 1 1 11 1 11 1 1 I l f # iIIft^ X X X w V W
•\x 9 -W'P'//
W // IftI
%~ 9 > + f * 1 1 9 9 9 9 9 9 9 9 9 9 9 9 9 9 I f f #
.-.* 1 M M I I M I I M M I M I f f # •Xxxm • • • • v
I mmmm* ft
• • V / / I M I M M M V M I M I I f f # ^ ^ ^ / / / f t f t
• * 9 9 9 9 9 9 9 9 * 9 9 9 9 9 9 9 99 9 9
^ . • « < /// H I M I I I I I M M I I • • !
• ■ • • • • • /I ##########•■•### I f f #
M 1 / I 9 PPPPPPI PP I 9 9 9*
...................../ I I I I I M I M M I M M ! ! ! «
• M i l l II I M I M / I ^ t I f f »
r r n r n II 111 ■« err*
I I I I M M
, k W > \\\v .sN ,« ix
rn'Wm-mWVXVxVX\Xx^x*^»—m%
M X x - f c W S N N f t ft x l — X
K M I M
III i w x w x ^ x
• • « « • • ! I f f I I f f »xxx\\v%\\\\\N
wvwx\\\SN\\\\I l%XX\l^
. v v x \ \ \ \ N
IXXXXXX
XX
^V
\ \ \ x \ \ x \x x x x \
smXNWSWX\ XX\X\X X
• « • • / • '/ / / / ! | |
, NS \ \ w x \ / / / / «
« / / •/«
•« « •••••> « ! / • •
M * 1 > * * \ \ N NXWKVXXXxX
• • • O M M •% « •••« ••« !••
L.mmf 9 1 f j i < < < \ \ V I A x X x
■ m m ..f I » f > I
•
IJ\
M I 1 t r \ t \ x x x x x x x
IXX X^*«
P p f p 9 9 9 9 9 XXX— *• ► »•••••••*««% •
\\\^ r . . . , / »//I r
. « • • • • / ! I #9 9 9 9 9 9 \ .mmmmmmmmmm • • •
l l l l l l K » * \ ***•. I
.mmmmmmmmf %mm ft
« •
•//
• I m m »— *
I M I H I
l l l l l l I —
* • • * » * * * » • • f t* * f t
ftftftftft
9 9 9 | • * * 9—»•*
• • • I I I —I — ‘
• • m ||r n » » «% f• • « « • • • • • « ! <
• I !•« ••
m m
• <• f t ( f t | * f t | f t l l M / H I I % l * M I M * m m I I « « P
• ft %ft ft * ft » / * f t f t f t f t * l f t f t » « » fc» * % » * «
••ftftlll I I ftIft »/!//•* ft*.........*......... I
m X % % « « ! ft I ft I I I ft • • * • * • • • • — » * * !
- m m % ft f t l f t f t f t f t f t f t f t » / / / ! ft ft « • ■ ........................................" ' I
• % f t « f t t f t l l l f t f t M l l f t l f t t ) i f t « f t f t « « f t k “ » M I
■^t- . x x x x x %> i m i m u i M f t % * M » * f t f t * * * M i
k x x ' x x x i i i n » i i i i i i M i i * » * * » « * ,* * * * # f i
•X f t \ X f t M l U f t H / M « t M M f t f t * f t * — * f t % M I
M \XftlinU»ll//M IM It»ftft»»*»**% M I
i ’ I f t X I I I I I I I f t l / / / / < » • ft ft ft ft ft ft ft ..................
l l l l l l l l l l * 4 t* * 4 * % • • • % % % • • • • • • * • / / »
......................I 4 4 4 4 4 * * * * * % * * * * • • • • • • • • • ! 9 f
* .......................................................................... 9
4444* * * * * m m % . m m # « • • ! • ' f X f l
* ' Vt f ' / rn— • *11 M r >>»»»
y -n
i f t u
> • • • > • • ftJMLi.
m i lu r i
I I
I I
I I X X X ^N N V S V
# I I X X f t V
• I
« / / / / / / • / « |
_______
f t l l l l l l l M M M I I f t f t f t f t f t f t
1*1 1*1 1 1 111 1 1 1 1 1 1 1
M / M I M M M M M I M / I I
** 1 •
II # ft\\m V ^ /✓ 7 /% l I ; M I If M M M M II M I I I
II »m«**m%
II mmi i MmMMi i mi *****
i / / / r * i m i i i i m i m
l l l l l i m
m m i M
M M M l M M l M I M
_ ix x M H i i M
i mii Mm
j iM
ii ii Mm m m i m i
m ii n#mm
r nJUL
vmuumh *"
ftX -. rr
J . • ft |
119*1%
%•%! l I l l l l l M M M M i m M l f l
k« i f t i i f t « « « f t M i i i m M M m /
ft | « m m m m m ^ f 1 * 9 1 1 9 9 * * * 1 l / « P / / * f t * l
' " t f / / .
•ftftftI/I//W -—PI MMM/f t* ****---+*
’ Lft I ft ft ft ftmmmmmmf I l M M I I
. « a %| « »m p M i n i / / I
! / / » • • . < I I m 1 i «i »U%f t f t f t l # #
^1 ^ l l f t l l / f t f t i / « mm ft | ) ' | I l l l I M M M M M I l .t
kll lftft tftl lft«( M f IMMMMIMIIftftftftlft

^ 1 1 1 1 1 % * % * * * * % * 9 9 9 9 1 * 1 ** 1 * 1 * 1 * 9 * * * * *
I ft I ft %ft ft ft % *** 9 9 1 9 9 1 9 1 * * 9 * 9 1 1 * 9 1 1 * * * *
iftI ft ftftftftft%***HI**************** ****
« 4_k ft ft ft ft Ift ft f t^^mp i i i M I M M M I M M M M M
1 111
i m M v sj" * " ' "
^
iXftftI I I**^mmXIIMMMMMMMIMMM,
I ft ft ft 1 1 4 * * * * \ 1 9 1 1 9 1 * 9 9 * 9 1 1 * * 9 t * * * 1 * 9
■X t ft 11 4 4 f * m -x x » i m i m m m m i m m m M
-----------------
n i lM V V \^ 11 I ft ft I I I I •y->*X|X\ # +
9 9 9 9 9 1 1 9 * 9 * * 9 9 * 1 9 1
* 4 * 4 4 4 ft x « % \ \ \ \ \ 4 \ \ * 0 * % l K l 1 f f 1 1 1 1 1 1 * 1 1 * 1 1 * 1 * * *
r p I «I I I \ \ \ \ / l \ \ * % * 9 1 1 1 1 * 9 1 *9 1*9 1 * 1 1 * * * ' " *
l n x x \ \ \ \ \ — I I I I I
i - f t I I I %S*# I ft X x W * 9 1 1 1 9 9 * 1 1 9 1 1 1 1 * * * * * 1 1 *
* 7" I I I I 11 I M M M M M \ * * * * * * * *m \ 11
p p\XXX X K X K x x — / / / / / ^ ^ <
lU X X W X M x — PjlllM
x • / / / %%
# I I %«
, | \ t
+~*1191911119*9111**** **11*
« / k l %%* * 1 9 * 1 * 1 1 1 * 1 1 1 1 9 * * 9 * 1 1 * * 1
\ X X \ « . ^ N « g s \ * ^ • • ftf t f t« f tf t f tf t f tf t f t%%l f t« • I l l l i t I * •■ % •* * 1 1 1 1 1 1 1 1 * 1 1 9 * * * 1 1 1 9 * * 1 1
- - * * * * * * ft l l l l l l l l l l i \ \ 4 L k % * ~ 9 1* 9* 9 * * 1 91 9 1 1 9 1 1 * * 1 9 1 1 9 * 9 9
i ■,*,*t 1t*7 t *9r**1**»*11*i**9***111*

km*f tf tf t\f tf t / ft | l l l l l l I I I 7 < l / x f 1 9 1 111 # # # 1 1 1 1 1 1 1 1 * * 1 1 1 * 1 * 1
r nr - ------
IS 27 36 18 27 36
Figure 3.17 Velocity plots obtained by double exposure PIV for the oscillatory flow
due to disk B; 4=12.70 mm,^=0.20 Hz, initial condition=TDC; (A)-(H) equally
spaced over one cycle starting from top dead center; x and y distances in mm.
58
59
(a) KCt ~ 0
'V 'V
/V 'V -
Id q )
(b) tfc, = 0 (1)
(c) K C ,» 0 (1 )
«--------------------->
\ »
« >
Figure 3.19 Schematic diagram of the effect of disk thickness on vortex formation
and shedding; (a) KCt ~ 0, (b) KCt = 0(2 n:) and (c) KCt -» 00.
CHAPTER IV
AXIAL OSCILLATIONS OF CYLINDERS AND DISKS: FORCE
MEASUREMENTS
4.1 Introduction
In the case of a vertical cylinder in springing oscillation, the two main components
of hydrodynamic damping are: 1) friction drag due to viscous shear stress on the surface,
and 2) form drag due to separated flow at the bottom of the cylinder. Contributions from
other sources like wave generation and diffraction are usually insignificant at very low
amplitudes of oscillation. If a disk is attached to the bottom of the cylinder, the form drag
component is further influenced by the interaction of vortices generated by the two edges of
the disk.
In addition to the dimensionless parameters KC and (5 defined in Chapter 3, the
reduced velocity parameter (V) defined in Chapter 1 will be required for the purposes of
this chapter. When discussing the flow due to a cylinder, the dimensionless parameters
KC, ft and V are defined based on the diameter of the cylinder. If a disk is attached to the
cylinder, then the diameter of the disk is used to define these parameters.
For separated flow around bodies where form drag is the dominant damping
component and friction drag is negligible, Morison’s equation provides a way of
formulating the force. This model expresses the drag force as a quadratic function of
velocity. The inherent assumption in the formulation of Morison's equation is that the flow
be two-dimensional. Hence the applicability of this equation for three-dimensional flows is
not obvious. If one were to assume that a three-dimensional oscillatory flow possess
60
61
similar gross characteristics as its two-dimensional counterpart, for example significant

vortex shedding off sharp comers, then one could expect the form of Morison’s equation to
be valid for a body in a three-dimensional flow.
Maull and Milliner (1978) have mentioned that there is no basis for using Morison's
equation at low KC, since a quasi-steady drag generating wake does not occur. Graham
(1980) has endorsed this viewpoint with an argument that the behavior of the drag
coefficient at very low KC varies with cross-sectional shape. For example, as KC —>0,
Cd tends to infinity for a normal flat plate and to zero for a circular cylinder (Graham,
1980).
Several results obtained at low KC by works discussed later in this section have
been ultimately cast in terms of Morison's drag coefficient, even though the formulation
may not be expected to function satisfactorily. The reason for this is that Q is viewed as a
non-dimensional parameter representing the force in phase with velocity. This differs from
the original approach of considering Q as an empirical constant whose value is determined
experimentally. Let us consider a sinusoidal flow whose velocity is given by Eq. (2.3)
u = U cosd
(2.3)
6 = (Ot + <p
Referring to Section 2.1, the drag coefficient, Cd, may be written as
C d = T i r d— • w
—— pSUu
2 3tt
Eq. (4.1) differs from the classical formulation of force coefficient by the factor (8/3tt)
arising from the Fourier decomposition of cos 0|cos 6 \ term.
Alternately, the nonlinear character of the drag force may be represented by a

linearized equation, in a Fourier - averaged or in a least squares sense (Keulegan and
Carpenter, 1958; Sarpkaya and Isaacson, 1981). This gives
Fd =Bu, (4.2)
62
where B is the equivalent linear damping coefficient A dimensionless damping ratio (Z)
may then be defined as
(4.3)
Cc 2mco ’
where Cc is the critical damping and m is the total mass (including the added mass) of the
system.
The two non-dimensional formulations presented above can serve the ultimate
purpose of scaling from model to prototype. This is accomplished by allowing the drag
coefficient and the damping ratio to be functions of KC and f$. The two formulations can
be related in a straight forward manner (e.g. Sarpkaya and Isaacson, 1981). This
relationship for a circular cylinder and a cylinder+disk configuration are obtained by
inserting an expression for the mass in Eq. (4.3). This gives:
cylinder: (4.4)
cylinder+disk: (4.5)
Subscript 1 denotes the variables that pertain to the disk.
Friction drag in oscillatory flow

The classical example of an oscillatory flow is that due to an infinite flat plate
oscillating parallel to its plane, referred to in literature as Stokes flow. See e.g. Batchelor
(1973) or Telionis (1981). Let the plate oscillate at a frequency CO, and a velocity amplitude
U, in an otherwise still fluid In this flow, the Navier - Stokes equation reduces to a balance
between local acceleration and viscous effects. The velocity profile is given by (e.g.
Batchelor, 1973):
(4.6)
63
Here u is the velocity component parallel to the plane and y =0 marks the position of the
plate. The shear stress due to the flow is obtained from Eq. (4.6),
(4-7)
We next consider a circular cylinder of infinite length oscillating parallel to its axis.
This problem is an extension of the flat plate problem, with the inclusion of surface
curvature perpendicular to the direction of motion. Thus a solution to the cylinder problem
must asymptotically tend to the flat plate solution at very large radii of curvature.
Neglecting the effects of convection and pressure gradients, a closed form solution to this
flow can be obtained, e.g. Casarella and Laura (1969), and Chew and Liu (1989).
Independent of these works, this problem has been solved as part of this research, and
details are presented in Appendix C. It can be seen that Eq. (C.14) is identical to Eq. (4.7),
thus confirming the accuracy of the theory. Chew and Liu (1989) note that when the radius
of cylinder to boundary layer thickness ratio,
(4.8)
the transverse curvature of the cylinder can be neglected. The flow is locally two
dimensional, and Eq. (4.7) can be used to evaluate the frictional force experienced by the
cylinder. Applying Eq. (4.7) to a circular cylinder of diameter D and draft T, the force
component in phase with velocity is obtained as:
Fd(friction) = itDTiiU. cos(tof) (4.9)
The interesting aspect of Eq. (4.9) is that the force is linear with velocity. The coefficients
Cd and Z can be derived from Eq. (4.9) in terms of KC and /J. Thus
Z(friction) (4.10)
Q (friction) = 6 7 c2 (4-11)
64
It can be seen from the equation for Cd that the combination KC^fp is directly proportional
to the square root of the Reynolds number based on the amplitude of oscillation (Res).
Form drag due to oscillating sharp-edged bodies
Some of the earlier research on forces experienced by oscillating plates has been
compiled by Dalzell (1978). The inviscid drag force due to vortex shedding off two-
dimensional edges has been studied by Graham (1980) and Bearman et al. (1985).
Graham (1980) has used the Blasius equation to obtain a formula for the inviscid drag force
on an isolated edge. If the internal angle of the edge is 8, then the drag coefficient is
expressed as follows:
Cd = A(KCi1S- W « - 25\ (4.12)
where A is a constant depending on the geometry and calculated from simulations. For a
flat plate (8 = 0 ), diamond (8 = it/2) and circular sections (8 = it), results of this
formulation agreed quite well with experimental results. Shih and Buchanan (1971) have
conducted experiments on flat plates forced to oscillate at low Reynolds numbers (R = 1 -
1000), and moderate values of KC from 1.5-4.7. Based on measured values of drag
coefficients, the following empirical law was proposed as valid over the range of
parameters tested:
Cd = W C r ^ e x p ^ - j ^ j (4.13)
The KC dependence of Q mentioned by this formula is only slightly different from

the dependence predicted by Eq. (4.12), with 8=0. It was pointed out by Graham (1980)
that the correct form was not discernible from Shih and Buchanan's (1971) data because of
experimental scatter.
Bearman et al. (1985) have calculated the vortex shedding force on two-
dimensional sharp-edged bodies using the discrete vortex model of Graham (1980). The
authors have assumed that at very low KC, shedding from each edge of the body became
65
independent of other edges. For a body with multiple edges, the drag force was evaluated
as the algebraic sum of contributions from individual edges. The equation for drag
coefficient obtained by Bearman et al. (1985) is given below:
Cd = if l^ A ( A : C ) (25",r)/(3,r' 25) (4. 14)

k=l
where a# is a characteristic ratio of length scales for edge k, is the direction cosine of
the edge, and A is defined as before. For example, if the included angle is 90 deg, Eq.
(4.14) predicts that Cd for the edge is independent of KC. If 8= 0, then Cd varies as
KC-™.
The dependence of Cd on KC for an isolated edge with zero included angle may be
expected to be valid for a flow due to an oscillating disk, due to similarities that exist
between axisymmetric flows and two-dimensional flows. Discrete vortex ring
computations of De Bemardinis et al. (1981) have shown that Cd for a disk followed the
trend predicted by Eq. (4.14) with 8 = 0, up to a KC of about 3, beyond which the
behavior was very different The proportionality factor for the disk was evaluated to be
about 9.5, giving
Cd = 9.5(KC)~113. (4.15)
Total drag estimation

Huse and Muren (1987) and Huse (1990) have presented a method of calculating
drag on cylinders in an oscillatory flow from the stationary flow value. Their expression is
(4.16)
{ VKCe J
where Q y and are the drag coefficients in oscillatory and stationary flows respectively,
and e=2.71828. Huse (1990) has obtained an expression for the total drag force (friction +
form drag) for a vertical cylinder in heave oscillation. Using the present notation, this is
The drag force predicted by the above equation is linear in velocity because of a
simplification introduced by the author assuming KC to be very small. Experiments were
conducted as heave decay tests at /3 approximately 5 x 106, and KC in the range 0.0005 -
0.01. The measured drag force was also found to be linear with velocity and a fit to
experimental data gave Cds = 0.277. Chakrabarti and Hanna (1990,1991) also conducted
heave decay tests, and found the damping force to be linear with velocity in a/3 range of 25
x 104 - 1 x 106. These authors measured the damping ratio at model scale to be in the
range of 0.3 -1%.
Measurement techniques
Model testing can be conducted in a laboratory by oscillating a body in still fluid or by
oscillating the flow past a stationary body. Experiments performed by oscillating a body in
a still fluid (irrespective of the direction of oscillation) are usually one of two types:
1. The body oscillates freely, and damping is estimated from decrement curves
(e.g. Chakrabarti and Hanna, 1990,1991; and Bearman, 1992).
2. The body is forced to oscillate at a certain frequency, and drag coefficient is
estimated from some type of Fourier analysis (e.g. Troesch and Kim, 1991).
For lightly damped systems, both schemes have inherent difficulties. In decrement tests,
extreme care must be taken to ensure that the system damping is properly accounted for.
See Chakrabarti and Hanna (1990, 1991) and Bearman (1992). In the case of forced
oscillations, straight forward estimation of damping is extremely unreliable, since the total
force is inertia dominated. Troesch and Kim (1991) conducted damping measurements on
cylinders in forced oscillation at or near system resonance. In this case, the system inertia
forces were nearly offset by the restoring forces. The resulting force measurements gave a
67
good estimate of the drag coefficients. Different springs were used to obtain resonance at
several frequencies of interest.
The idea of forced oscillation at resonance is used in the present experiments on a
vertical cylinder oscillating parallel to its axis. Drag and inertia coefficients of the cylinder
are estimated at KC numbers in the range 0.1 - 1.0, and three /J numbers (89236, 156400
and 191700). Further, the effect on the hydrodynamic coefficients of the following are
also investigated:
1. addition of small forward speed, and
3. addition of a disk to the bottom of the cylinder.
The details of the experimental setup is given in^Section 4.2, followed by results and
discussion in Section 4.3.
4.2 Description of experimental setup
The experiments reported here were conducted at the Marine Hydrodynamics

Laboratory, University of Michigan. Table 4.1 gives the dimensions of the tank and the
test models, and the parameter ranges tested. The cylinder and disk are 4:1 scaled models
of the cylinder and disk A used in the flow visualization experiments.
The cylinder was attached to a Vertical Motion Mechanism (VMM) and oscillated in
heave, Figure 4.1. A comprehensive description of the VMM is given in Ashcroft et al.
(1989). The dynamic drive system of the VMM is a ballscrew type with motors and
controllers, capable of driving a stroke of ±0.102 m (4.0 in) at a rate of ±0.076 m/s (3.0
in/s). The mechanical system consists of the fore and aft sections with port and starboard
subsections in each (Ashcroft et al. 1989). The bearing sets of each subsection are so
arranged that only the vertical loads are transmitted to the load cell attached to the
subsection. Since the bearing sets remove all unwanted forces and moments, the VMM is
capable of producing a uniform vertical motion and negligible motion in the other modes of
freedom. The sum of forces measured by the four load cells gives the total heave force on
68
the cylinder. Potentiometers attached to the fore and aft sections give the displacement of
the cylinder.
The oscillation frequency for the experiments was so chosen to correspond closely
to the natural frequency of the system in heave. If a resonant condition is maintained, the
inertia effect of the cylinder and attached dynamometers is offset by the restoring force
associated with the fluctuating buoyancy force. Due to the lack of inertia dominance, the
force measurements give a reliable estimate of the drag coefficient The natural frequency
( / 0) of a heaving cylinder may be obtained approximately by linear system analysis. This
gives
(4.18)
where Awp is the wateiplane area and Cmis the inertia coefficient If a spring of stiffness k
is attached to the cylinder, then the natural frequency is modified as
(4.19)
Assuming Cmto be about 1.15, the natural frequency,/b from Eq. (4.18) is expected to be
about 0.42 Hz. During the experiments, the natural frequency of the oscillating system
was identified as the frequency at which the measured force was out-of-phase with the
measured displacement This gave a value of 0.411 Hz, which agrees with the value
predicted by Eq. (4.18). Two coiled springs of stiffness 36.65 N/cm (20.93 lb/in) and
70.25 N/cm (40.12 lb/in) were also used to give natural frequencies of 0.722 and 0.844 Hz
respectively. These values compare reasonably well with the values of 0.739 and 0.944
Hz obtained from Eq. (4.19).
During a typical experimental run, data for forces and displacements is collected over
16 cycles and sampled at a rate of 128 points / cycle. The signal is filtered through an
[
69
analog filter with a cutoff frequency of 6.4 Hz. The data is then Fourier transformed to
identify the magnitudes and phases of the forces and displacements. The drag force is
identified as the out-of-phase component of the total force at the oscillating frequency. The
repeatability of the data was verified during the experiments.
Table 4.1. Parameters of experimental force measurements
Surface area of water tank 100.5 m x 6.7 m (330 ft x 22 ft)
Water depth 3.0 m (10 ft)

Diameter of cylinder 0.457 m (1.5 ft)
Draft of cylinder 1.219 m (4.0 ft)
Mass of cylinder 26.32 kg (1.8 si)
Mass of cylinder + ballast 198.9 kg (13.63 si)
Diameter of disk 0.609 m (2.0 ft)
Thickness of disk 25.4 mm (1.0 in)
Frequencies of oscillation 0.411,0.722 and 0.884 Hz
Amplitudes of oscillation 0.25 - 6.35 cm (0.1 - 2.5 in)
Forward speed 0.15 - 0.61 m/s
©
o
*•>■4
KC
1
p (cylinder) 89236,156500 and 191700

p (cylinder+disk) 158600 and 278200
V 0.0 - 0.387
70
4.3 Results and Discussion
The dynamic equilibrium of a heaving cylinder gives us an equation for the forces
on the cylinder. We follow the formulation of Morison’s equation with the drag term
replaced by the equivalent linearized damping term. As mentioned in Section 4.1, the
linearized damping coefficient is a representation of the nonlinear drag term in a least
squares or a Fourier-averaged sense (see also Keulegan and Carpenter, 1958; Sarpkaya and
Isaacson, 1981). When no springs are attached, we get:
(m + ma)y + By+pgAwpy = Ft(t). (4.20)
Here F,(t) is the excitation force which is approximately equal to the total force measured
by the four load cells. By using Fourier analysis, this measured force may be decomposed
into its various frequency components. Considering say the first five harmonics, we obtain
Ft(t) = j^ F je iUcot+<l,j\ (4.21)

7=1
where Fj and <pj are the amplitude and phase of the jth component The vertical
displacement is given by a sinusoidal motion of amplitude A and frequency (O. In complex
notation,
y(t) = Aei(0*+* \ (4.22)

The measured time histories of y(t) and F((t) for the test cylinder are plotted in
Figures 4.2 and 4.3. These figures correspond to KC = 0.3 and 0.8 respectively, at f3 =
89236. The measured displacement is found to be very sinusoidal and higher order
harmonics are negligible. The total force on the other hand has significant nonlinear
contributions. The ratio of nonlinear to linear force component is smaller at higher KC.
The total force has a magnitude of about 1.5N to 3N in the figures. In Figure 4.2 points A
- H correspond to the phase points of the flow visualization in Figures 3.4A-H. The KC
numbers of Figures 3.4 and 4.2 are quite close, while their /? numbers are quite different
71
(5208 and 89236). Different flow phenomena are seen during the upstroke and
downstroke of the cylinder in Figure 3.2. Similarly, the total force is also seen to be
different during two consecutive half cycles.
The hydrodynamic force on the cylinder due to the motion can be deduced from Eq.
(4.20) as
Fh(t) = - ( may +By)

t 2 \ (4.23)
= — \ F t ( 0 - { C O m + p g A ^ p )y (t)J
This force as expressed by the above equation is plotted in Figures 4.2 and 4.3 for
the runs considered. This plot is clearly dominated by the added mass component, as can
be seen by comparing the phase of F/, (t) and the displacement The magnitude of the total
force is seen to be much smaller than the hydrodynamic force, which must be expected
because of the resonant oscillation of the cylinder.
In a perfectly tuned system, the first order inertia force is exactly canceled by the
restoring forces. Due to practical difficulties in tuning a system precisely, there will always
be some inertial contribution to the total force. This contribution can be found from Eq.
(4.21) by evaluating the first order component that is in phase with the acceleration. A
residual force can be obtained by subtracting the inertial contribution from the total force.
This force represents the net contribution of the first order drag force and the nonlinear
components. Thus
Fr (r) = —(Fj (r) - F1cos(ft - (4 24)
The residual force plotted in Figure 4.2 and 4.3 resembles the total force time history,
implying that the system has been tuned well.
The time histories for the displacement and forces for the cylinder+disk system at
KC = 0.23 and 0.6 are plotted in Figures 4.4 and 4.5. The frequency of oscillation is
retained at 0.411 Hz, giving a f} of 158600 based on the diameter of the disk. The
magnitude of the total force ranges from 7 to 20 N, much higher than that for the cylinder.
The force time history, unlike that of the cylinder, shows a dominant in phase component
72
This is because the cylinder+disk system does not have the same natural frequency as the
cylinder. The natural frequency of the cylinder is likely to be lowered by the addition of the
disk, as can be inferred from Eq. (4.18). A net effect of this is to detune the system,
leading to a larger inertial contribution to the total force. The hydrodynamic force Fh (?) is
about two times the magnitude of the total force.
It was observed from flow visualization, Figure 3.5, that for a cylinder+disk
combination, the flow during the two half cycles are similar and anti-symmetric. This type
of anti-symmetry is only partially seen in the residual force plot of Figure 4.4. The center
peak during the downstroke is not as pronounced as that in the upstroke. This difference is
however diminished at higher KC, where during the two half cycles, the residual force is
almost symmetric.
Damping ratio for the test cylinder

The damping ratio of the test cylinder as a function of KC at /5=89236 is shown in
Figure 4.6. The scatter in the data is seen to be small, and the trend is linear. A least
squares linear fit to the data gives a slope of 0.0128, an offset of 0.00281 and R2=0.995.
Representing data in terms of the equivalent linear damping ratio Z clearly illustrates the
functional dependence of drag on KC. The good behavior of data indicates the reliability of
the experimental method used. Also plotted in the figure is a solid line corresponding to
Eq. (4.10). This line corresponds to Z = 0.00164, which is approximately 60% of the
offset of the linear fit to data. Another factor that may contribute to the offset is wall
roughness which is not considered in the present study. Based on the above calculation,
the friction drag, being linear with velocity, is the most significant component in the axis
intercept of the least squares curve fit of Figure 4.6. The form drag due to vortex shedding
then contributes to the slope of the curve, i.e. it is linear with KC. This also conforms to
the conclusions of Graham (1980) and Bearman et al. (1985).
73
From their experimental results, Chakrabarti and Hanna (1991) and Huse (1990)
concluded that the drag force was linear with velocity over a KC range of 0.0005 - 0.01.
From Figure 4.6, it appears that for this range, Z is approximately constant because the
form drag contribution is very small. However, Huse (1990) has concluded that the vortex
shedding force is also linear with velocity, in contrast to the present results. The theoretical
expression of Huse (1990), applied to the present case gives an equivalent Z due to vortex
shedding of 0.00058. The total value for "Z" using Huse’s (1990) method is 0.00221.
Effect of adding a disk to the bottom of the cylinder

Values of damping ratio for the cylinder with and without the disk are shown in
Figure 4.7. In this figure the two sets of experimental data converge at near-zero values of
KC . A least squares straight line drawn through the points for the cylinder+disk at low
KC, say less than 0.5, actually shows a marginal decrease in the intercept at KC=0. This
may be because (5is increased due to the presence of the disk (see Eq. 4.10).
At very small KC the experimental data may be compared with some of the results
of isolated edge theory. The two-dimensional results of Bearman et al. (1985) can be
extended to three-dimensions by considering the following relationship between the drag
forces in two and three dimensions:
Fd(3-d) = nDFd(2-d) (4.24)
As before, D in Eq. (4.24) is replaced by D\ if a disk is included. The two-dimensional

force for the cylinder can be obtained as the vortex shedding force due to an isolated edge
of included angle 90 deg. For the disk, this force would be the sum contribution of forces
from two isolated edges of included angle 90 deg. The characteristic length in this
calculation is taken as half the diameter. Expressing in terms of Z using Eqs. (4.4), (4.5)
and (4.24), we get
cylinder: Z=0.0228 KC
cylinder+disk: Z=0.1047 KC.
74
These two expressions are plotted in Figure 4.7. The theory overestimates the damping
ratio by the same amount, about 75%, for both test configurations.
Further insight into Figure 4.7 can be gained by considering the relationship
between the damping ratio of the cylinder and cylinder+disk configuration. If
Dpq « D1! , then we obtain from Eqs. (4.4) and (4.5)
< 4 ' 2 4 )
If the edges are truly isolated, then Cî = 2 C j. Using values of D\ and D from Table 4.1,
we obtain
— = 4.741. (4.25)
Z
In Figure 4.7, if a straight line is fitted to the data points for the case of cylinder
with disk at low KC (less than 0.5), this line would have a slope of 0.0608. Comparing
this to the slope of 0.0128 of Figure 4.6, we get a ratio of 4.75. The proximity of this
value to that predicted by Eq. (4.25) indicates that the physics of the flow is well
represented by considering the edges to be isolated at low KC. The isolated edge theory is
accurate in representing the functional dependence of the drag coefficient on KC.
However, in terms of magnitude, the numerical results based upon inviscid fluid theory,
are significantly higher than the experimental results. This is most likely due to the absence
of viscous diffusion in the numerical model, and the occasional lumping of discrete vortices
in the calculations (see Graham, 1980).
In Figure 4.7, beyond KC =0.5 the trend of Z for the cylinder with disk is
nonlinear. The flow visualization experiments (Chapter 3) shows that at very small KC,
interaction effects of the flows due to the top and bottom edges are negligible. As KC
increases, these interaction effects can no longer be ignored. More nonlinearities are
introduced when the vortices interact with the vertical wall of the cylinder.
75
Figure 4.8 shows the effect of thickness KC number on the variation of damping
ratio. As noted in Section 3.4, the cylinder represents the case of KC ~ 0. Numerical
results of De Bemardinis et al. (1981) for an infinitesimally thin disk has shown that Z
varies as KC2®. This represents the other extreme case of KCt —> However, for a
given disk of finite thickness, KC numbers based on diameter and thickness are linearly
related. For the present case, KCt = 2AKC. It is immediately obvious from Figure 4.8 that
the damping ratio increases as KCt increases. For the cylinder+disk, Z initially varies
linearly up to KC = 0.5, beyond which the trend is almost quadratic. At some point, the
amplitude of oscillation may become large enough that the disk would resemble a flat disk
of infinitesimal thickness. If this happens, then the Z may be expected to vary as KC2®as
predicted by De Bemardinis et al. (1981). The quadratic variation of the cylinder+disk at
moderate KC may be explained as a transition from a linear to a KC2® type trend.
Figure 4.8 supports several of our conjectures made in regard to the effect of disk
thickness on damping (Section 3.4). The model of De Bemardinis et al. (1981), though
inviscid shows a much higher drag than the experimental values obtained at lower KCt.
As expected, convection of vorticity is the dominant mechanism of damping at large values

of KCt . Viscous diffusion in the case of cylinder produces low levels of drag force.
Effect of varying the frequency parameter (B)

Variation of Z vs. KC for three fS values ranging from 89236 - 191700 is presented
in Figure 4.9. It is clear from Figure 4.9 that increasing (3leads to an increase in the slope
of Z - KC curve. Two possible reasons are cited:
1. material damping of the springs used, and
2 . viscous diffusion.
In the setup of Figure 4.1b, the damping of the spring was inevitably measured by the
attached load cell. If this damping is a function of velocity, then it must vary with
frequency and hence f}. Further, the beam attached to the spring was connected to the
76
cylinder by steel ropes, and hence was free to move in other modes of motion. This further
induced uncertainties in the force measured by the load cell attached to the spring.
The isolated edge theory of Graham (1980) is basically inviscid, and is applicable at
high values of /?. The effects of viscous diffusion are not included in the prediction of Cd
variation as a function of KC. Flow visualization experiments (Chapter 3) have shown that
vortices formed during the upstroke of the cylinder were diffused during the downstroke.
Though P values used here are much higher than those used during flow visualization,
Figure 4.9 shows that the effect of diffusion may not be negligible.
Figure 4.10 shows the results obtained at two P values for the cylinder with
attached disk. The slope of Z - KC curve at P = 278200 is only slightly higher than the
slope at P = 158600. The slower increase in slope with p is because the effects of
diffusion in the presence of the disk are likely to be diminished. Diffusion in this case is of
secondary importance relative to vortex shedding. This was also seen in the flow
visualization pictures of Figure 3.5 and was discussed in Section 3.4.
Effects of small forward speeds

One important effect of a current is the onward convection of vorticity shed by the
edges of the cylinder. The net efflux of kinetic energy from a control volume enclosing the
cylinder would thus increase due to the current The resultant increase in damping would
depend on the ratio of the uniform stream to the velocity of oscillation. The reduced
velocity parameter was defined in Chapter 1 as
coD’
where Uq is the incident current For a cylinder+disk combination, D is replaced by D\,
since the disk dominates vortex shedding. Figures 4.11 - 4.14 show the variation of Z
with KC at different forward speeds for the cylinder and cylinder+disk at two P values. At
higher forward speeds, visual observation indicated formation of a wake behind the
77
cylinder. At a speed of 0.61 m/s (2.0 ft/s), there was also a noticeable Karman type vortex
shedding along the direction of the current The effect of these phenomena on damping in
the vertical direction is to cause considerable scatter. The scatter for the cylinder+disk is
seen to be lower than for the case of the cylinder.
It is clear from the figures that the effect of forward speed is more pronounced at
lower KC values. Further, it appears that the damping ratio at different forward speeds
asymptotically approaches the zero speed value at large K C . Then it is possible to define a
velocity parameter, Skc >which would identify the extent to which forward speed would
be important. In the present context, it is more appropriate to define a general velocity
parameter as
(4.26)
where the value of n would depend on the body geometry. Consistent with Chapter 2,
when
Skc ~ ^(l)
the effect of current can be ignored. From Figures 4.11 and 4.13, n is in the range of 1 -
1.5 for the cylinder and cylinder+disk.
Results for the added mass coefficient

Results for the added mass are obtained from the first order force component of the
total force that is in phase with the acceleration. A non-dimensional added mass coefficient
is defined as the ratio of added mass to the displacement
(4.27)
M is plotted vs. KC for the cylinder and cylinder+disk at different p values and at different
forward speeds in Figures 4.15 through 4.20. These results are prone to uncertainties in
the order of 4 - 8% because of:
78
a) tuning the system so as to obtain reliable damping coefficient values; this

diminishes the accuracy of measurement of the inertia component of the force, and
b) an unfortunate water leak in the cylinder which contaminated several added mass
measurements.
For the cylinder at /J = 89236, the value of M is quite constant (Figure 4.15). A
least squares linear fit gives an offset of 0.136. This value seems to increase to about 0.2
at higher f3 (Figure 4.16). When a disk is added, there is a significant increase in M, as
seen in Figure 4.17. At /3=89236, the added mass is about 25 - 30% of displacement, and
the curve has a positive slope. At higher fS, the added mass increases to about 33%. The
increase in added mass due to the presence of the disk was previously noted in the force
time histories, Figures 4.2 - 4.5.
Intuitively, one would expect the added mass to increase with the addition of the
disk to the cylinder. This is because a higher amount of kinetic energy is imparted to the
surrounding fluid as the disk moves. On the other hand, the results of Bearman et al.
(1985) have shown that vortex shedding from an edge actually serves to decrease the added
mass. This decrease may not be crucial, because the potential flow in the presence of the
disk is modified to increase the added mass significantly.
The effect of the forward speed on the cylinder and cylinder+disk added mass
coefficients are shown in Figures 4.18-4.21. The added mass for the cylinder increases
modestly with increasing current speeds. A reverse trend is noticed for the cylinder+disk.
Scatter in data of about 6 - 8 % prevents from quantifying any trend. It appears that the
effect of forward speed on added mass is not very important
79
80
1.5
0.5-
fc" -0-5-
1 - -
-1.5-
- 2 -
61
4-
2 -
-2 -
-4 -
6 - -
0.4-j
0.3-
0 .2 -
0. 1 -
0 1-
- .
- 0. 2 -
-0.3-
-0.4J
3-i
2 -
>»
- 2 -
1.57 3.14 4.71 6.28 7.85 9.42 10.99 12.56

0 = 0) t
Figure 4.2 Time histories of forces (N) and displacement (cm) for the test
cylinder at ATC= 0.3 and p =89236.
81
6-i
4-
- 2 -
-5-
- 10 -
-15-
4-j
2 -
-4-
- 6 -
- 8 -
8 -.
4-
0-
-4-
0 1.57 3.14 4.71 6.28 7.85 9.42 10.99 12.56

0=C O t
cylinder at KC = 0.8 and {3=89236.
82
3-.
2 -
- 1 -
4-
-4-
8 - -
- 12 -
21
1.5-
0.5-
-0.5-
-1 -
-1.5-
-2 J
3-i A
2 -
0-
- 1 -
- 2 -
1.57 3.14 4.71 6.28 7.85 9.42 10.99 12.56

0=001
cylinder with disk at KC =0.6 and P =158600.
83
F.ft)
-5-
10 - -
-15-
32-,
16-
F„(t)
-16-
20 -
10-
F,W
- 10 -
4-
>v
-4 -
0 1.57 3.14 4.71 6.28 7.85 9.42 10.99 12.56

6=(Ot
cylinder with disk at KC = 0.6 and (3=158600.
84
0.016
0.014
O'
0.012
0.01 X
N 0.008
0.006
0.004
0.002
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

KC
Figure 4.6 Damping ratio of vertical cylinder vs. KC; J3=89236.

□ experiments; least squares linear fit;
friction component (Eq. 4.10).
85
0.1
■ experiment - w/o disk
0.09
0.08 • ...... Bearman, one edge
0.07 ▲ experiment -with disk
0.06
/ A -----Bearman, two edges
N 0.05
0.04 // ▲*
/ *
0.03
0.02 // * aiA
r.
s
// »A
/ M ................ ■ ■ ■ "
0.01 - / a * ...... m ■ "
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
KC
Figure 4.7 Damping ratio vs. KC for cylinder with and without disk.
0.18
0.16 experiment - w/o disk
{KC,- 0)
0.14
0.12 experiment -with disk
{KCt =0(1))
0.1
N De Bemardinis et al. (1981)
0.08
0.06
0.04
0.02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

KC
Figure 4.8 Damping ratio vs. KC at different thickness KC numbers.
86
0.018
0.016 ■ {3=89236
0.014 o {3=156493
0.012
a {3=191650
0.01
Nl
0.008
0.006
0.004
0.002
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

KC
Figure 4.9 Damping ratio for vertical cylinder vs. KC at different {3.
8=158600
o 6=278200
N 0.03
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

KC
Figure 4.10 Damping ratio vs. KC for cylinder with disk at different p .
87
0.025
■ V=0.0
0.02 A A* / V A o V=0.129
A
AA° O , A V=0.258
0.015 ❖ V=0.386
0.01
0.005
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

KC
Figure 4.11 Variation of damping ratio with KC for the vertical

cylinder at different forward speeds; P =89236.
0.02
0.018 ■ V=0
0.016
0.014
o V=0.147
0.012
N 0.01
0.008
0.006
0.004
0.002
0 0.1 0.2 0.3 0.4 0.5 0.6

KC
Figure 4.12 Damping ratio for vertical cylinder vs. KC at different V;

p =156500.
88
0.09
■ V=0
0.08
o V=0.097
0.07
▲ V=0.1933
0.06 • A * o V=0.29
0.05
N • V=0.387
o » /
0.04
o ° ■
0.03 ooP _ ■ ■
0 , 0 Ocpo
0.02 ■■■
0.01
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
KC
Figure 4.13 Damping ratio vs. KC for cylinder with disk at different
forward speeds; 0 = 158600.
0.06
■ v=o
0.05
o V=0.110
0.04
A V=0220
N 0.03
0.02
0.01
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

KC
Figure 4.14 Damping ratio vs. KC for cylinder with disk at different
forward speeds; {3=278200.
89
0.16
0.14
— -g — *
0.12
0.1
2 0.08
0.06
0.04
0.02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

KC
Figure 4.15 Added mass coefficient for vertical cylinder vs. KC;
P =89236; A experiments; least squares linear fit
0.25
■ P=89236
0.2
o p=156493
0.15 a p=191650
0.1
0.05
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
KC
Figure 4.16 Added mass coefficient for vertical cylinder vs. KC at
different p.
90
0.2
■ V=0.0
0.18 ' f % ^ £ > 2 o 0 0^ o 0 o 0 o
t 1 o V=0.129
0.16
"o ■ -0 0 ° - - O ^ t> 0_0 ▲ V=0.258
0.14 o& m • ■
■ ■ 0°oO OO
0.12 ❖ V=0.386
2 0.1
0.08
0.06
0.04
0.02
0 I I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
KC
Figure 4.17 Added mass coefficient vs. KC for vertical cylinder at different
forward speeds; {3=89236.
0.25
0.2
o V=0.147
0.15
0.1
0.05
0 0.1 0.2 0.3 0.4 0.5 0.6

KC
Figure 4.18 Added mass coefficient for vertical cylinder vs. KC at
different V; P =156500.
91
0.35
0.3 6=158600
0.25
0.2 o p=278200
0.15
0.1
0.05
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

KC
Figure 4.19 Added mass coefficient vs. KC for cylinder with disk at
different P .
92
0.35
■ V=0
0.3 o V=0.097
0.25 ▲ V=0.1933
•••• H t •••
❖ V=0.29
0.2
• V=0.387
0.15
0.1
0.05
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

KC
different forward speeds; P = 158600.
0.35
■ V=0
0.3
0.25 o V=0.110
0.2 ▲ V=0.220
0.15
0.1
0.05
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

KC
different forward speeds; p =278200.
CHAPTER V
SCALING LAWS AND AN EXAMPLE CALCULATION
Scaling laws are proposed in this chapter that enable damping estimation of a
prototype TLP from model test results. The scaling model includes skin friction due to
viscous shear and form drag due to vortex formation and shedding. Other factors like
roughness, interference between members, and free surface effects are ignored. Most of
the laws presented here are valid at very small values of KC, and moderate values of /?.
Rules are presented for both vertical and horizontal members of a TLP. It is noted that in
this chapter, subscript V and 'K respectively denote quantities associated with vertical and
horizontal cylinders of a TLP.
Scaling laws for vertical cylinders

The damping force experienced by an oscillating vertical cylinder can be identified
to be a function of the following variables:
Fd = Fd {DV,T, co, A, v,p,t). (5.1)
If the motion is sinusoidal, then the dependence of the damping force on time is known. If
the magnitude of the drag force is expressed by the dimensionless equivalent linear
damping ratio (Z), then we obtain the following non-dimensional equation:
(5.2)
For the model cylinder, from Figure 4.6 this relationship is identified to be linear for fixed
geometry and JJ, i.e.
Zv = cq + a2KC (5.3)
93
94
Eq. (5.3) can be used for estimating the damping ratio of the prototype, if cc\ and «2 can be
scaled appropriately. Since the offset (a{) of Figure 4.6 is predominantly friction drag,
Eq. (4.10) can be used as a scaling law. Thus
«icc (p~°5 ). (5-4)
In regard to the slope (a 2) of Figure 4.6, in Chapter 4 we had stated that this slope is
principally due to the form drag component of the total drag force. Further, the drag
coefficient (Q) associated with the form drag is a constant (Graham, 1980; Bearman et al.
1985). Thus from Eq. (4.4), we obtain
<5 -5 >
Scaling law for vertical cylinder with attached disk

It was mentioned in Chapter 4 that the presence of the disk mainly affects the vortex
shedding component of the drag force, and the influence on friction drag is quite small.
From Figure 4.7, we deduce that the drag coefficient of the cylinder is doubled due to the
addition of the disk, i.e.
Cdi = 2 Q
Eq. (4.24) then gives
Z l= 2 ( j f j z v. (5.6)
Eq. (5.6) presents a law for estimating the damping ratio of a cylinder with disk, once the
damping ratio of the cylinder is known.
Scaling laws for horizontal cylinders

Similar to the vertical cylinder case, the drag force for horizontal cylinders depends
on the following variables and parameters
95
Fd = Fd(Dh,L , 0 ),A,v,p,t). (5.7)
Expressing in terms of the damping ratio
Z h = Z i t t L' * c , p ) ' <58)
It is seen from the work of Troesch and Kim (1991) that at low KC, the trend of Cd
is satisfactorily represented by the laminar theory, provided an effective eddy viscosity
correction is included. An expression for Cd is obtained from Eq. (2.5),
Cdh= ^K 2-5(KC)-lp-°-5. (5.9)
Using a Cm value of 2, Eq. (5.8) gives an expression for Zh as
z„ = ^ (5.10)
An effective eddy viscosity may be incorporated into Eq. (5.10) by replacing the parameter
p by /?eff defined as
D2f
* (5-ID
U eff
where ueff is the effective eddy viscosity. The results of Troesch and Kim (1991) for
circular cylinders give
ueff = 21u. (5-12)
For square section cylinders, Troesch and Kim (1991) estimated that
Ueff =51v. (5.13)
The scaling law from Eq. (5.10) is thus

0.564
= —7^— * (5-14)
Example TLP calculation

The scaling laws described in the two previous sections are applied to a prototype
tension leg platform given by Petrauskas and Liu (1987). Table 5.1 gives the dimensions
96
of this prototype TLP. Though this prototype does not scale perfectly with the model
dimensions of Table 4.1, the scaling laws can still be applied, since all quantities used in
the calculations are dimensionless.
Values of the offset and slope of the damping ratio for the model from Figure 4.6
are:
o?i = 2.818 x 10-3
a 2 = 1.286 xlO-2.
For the vertical columns of the prototype,
— = 0.473
T
j3 = 9.324 xlO7.
The scaling law of Eq. (5.3) yields the following equation for the prototype
Zv(prototype) = 8.718 xlO -5 + 0.0162(KC) (5.15)
The friction drag contribution to the total damping is not surprisingly low for the prototype
in comparison with the model. This is to be expected, since the prototype fi is much larger
than that of the model. At /J = 9 x 107 and for very low KC (say 0.01 - 0.1), the
amplitude Reynolds number is in the range 103 - 105, and the boundary layer is laminar or
in transition. Based on the deductions of Kamphuis (1975), the effect of turbulence on
friction drag is negligible over this range of Reynolds number.
The damping ratio for the full scale TLP column using Eq. (5.15) ranges from
0.0249% to 0.1708% for KC = 0.01 - 0.1. If a designer were to use theory (Eq. 4.10) to
scale friction, and Huse's (1990) method for form drag, then the following values would
be obtained for the prototype:
Zv(friction drag) = 0.005%

Zv(form drag) = 0.057%
97
giving a total damping ratio of 0.062% for oscillating vertical cylinders at small KC (<
0.01). In comparison with the present calculations, this method of scaling overestimates
the damping ratio at very low KC and underestimates it at low to moderate KC.
Table 5.1 Particulars of example prototype TLP (Petrauskas and Liu. 1987)
No. of vertical columns 4

Column diameter 15.85 m (52 ft)
Column draft 33.52 m (110 ft)
No. of horizontal pontoons 4
Pontoon diameter 8.53 m (28 ft)
Pontoon length 45.11m (148 ft)
Total displacement 377241 (2.584 x 106 slugs)
Estimated pre-tension 20%
Free-floating displacement 31437 t (2.154 x 106 slugs)

Free-floating draft 25.75 m (84.51 ft)
Mass of each column 52121 (3.571 x 105 slugs)
Mass of each pontoon 2646 1 (1.813 x 105 slugs)
Period of oscillation 2 sec
If a disk of diameter 21.13 m (69.33 ft, Z>i/Z>=1.33) is attached to the column, then
from Eq. (5.6) we see that the damping ratio increases to 0.118% - 0.81% for KC = 0.01 -
0 . 1.
For horizontal circular members using Eqs. (5.12) and (5.14), we have
jSeff = 1-287 x 106
98
Zh = 0.049%
If cylinders of square section replace the circular members, Eq. (5.12) yields
/Jeff = 5.149 x 105
Zh = 0.078%.
From these numerical values, it appears that as KC increases the vertical members
contribute more to the total damping than the horizontal members. The results of this
section are summarized in Table 5.2.
When evaluating the damping ratio for the horizontal and vertical cylinders, the
local mass (of the column or pontoon) is used. This facilitates straight forward
extrapolation from model scale. Hence one cannot sum the Z contributions from the
individual structural components to obtain the total damping ratio of the structure.
Alternately, one could evaluate individual Bv and £/,, which can then be summed to give
the total damping coefficient for the TLP. The relationship between the damping ratio and
the damping coefficient from Eq. (4.3) is given by
(5.16)
(5.17)
Zh cylrtfL
Using the total damping coefficient and the total system mass, the damping ratio for the
TLP can be obtained. These values are tabulated in Table 5.3.
99
Table 5.2 Damping ratio of prototype TLP components
Component Method Damping ratio
Horizontal cylinder - circular Troesch and Kim (1991) 0.049%

section
Horizontal cylinder - square Troesch and Kim (1991) 0.078%

section
Vertical cylinder - circular Huse (1990) 0.062%

section
Vertical cylinder - circular present calculations 0.025% (ATC=0.01)

section 0.171% (ATC=0.1)
Vertical cylinder with disk - present calculations 0.118% (/sTC=0.01)

circular section 0.81% (KC=0.1)
Table 5.3 Damping ratio of the prototype TLP (4 columns and 4 pontoons')
Vertical Pontoon KC=0.C1 KC=0.1

Column section
Cylinder Circular 0.032% 0 . 100 %
Square 0.043% 0 . 111 %
Cylinder with Circular 0.060% 0.383%

disk Square 0.072% 0.394%
CHAPTER VI
SUMMARY AND CONCLUSIONS
This thesis has concentrated on one specific design aspect of Tension Leg
Platforms, namely the hydrodynamics of springing oscillations. These vibrations, usually
of high frequency and small amplitude, have been identified as a potential source of fatigue
failure of tendons. Accurate estimation of damping forces in this mode is important in
estimating the response of the structure in springing.
One of the important design principles for complex structures such as TLP's is
component scaling. This implies that flows due to different components of a TLP in
springing are different, and hence cannot be scaled on the same basis.
In Chapter 2 we considered a cylinder in oscillatory cross flow, and the damping
forces on the cylinder. A literature review has revealed the quality research that has been
conducted in the past on this subject At very low KC and moderate (3 numbers, existing
analytical, numerical and experimental works provide a good understanding of the flow as
KC is varied. Much is still not known about the variation of damping forces as a function
of p. It is clear from Chapter 2 that research needs to be directed towards the variation of
Cd at large values of /?, in the range of 105 - 106, which is typical in an open sea. More
understanding is required on the effects of transition and turbulence on drag. For example,
numerical simulation or turbulence modeling may be used to explain the offset difference
often seen in experimental data.
Particle image velocimetry has proved to be a powerful tool in studying unsteady,
spatially evolving flows. Chapter 3 has presented a variant of PIV which is an
improvement over the existing single exposure digital method. The method is applicable to
100
101
flows whose average velocities are at least twice as large as those handled by the single
exposure method. The new scheme uses double exposure video images, and image
shifting to eliminate directional ambiguity. An effective error checking scheme is used to
eliminate spurious vectors that arise due to higher noise levels in the double-exposed
images.
Chapters 3 and 4 have presented an experimental study of flow due to cylinders and
disks in axial oscillations. Flow phenomena were understood by using qualitative and
quantitative flow visualization. Experimental force measurements have provided
information on the behavior of drag forces with respect to the parameters KC, fi and to
some extent V. The addition of a disk to an oscillating TLP column has been employed to
increase damping in the springing mode.
The flow around the edge of an oscillating cylinder during downstroke was found
to be quite different from the flow during upstroke. This lack of symmetry was because of
the vertical wall of the cylinder. Diffusion was found to dominate over convection at R =
2461. When a disk was attached to the cylinder, the flow was modified significantly. It
was concluded that at low KC, the flow past an oscillating cylinder-disk configuration
resembled the flow past the oscillating disk.
Visualization of the flow due to a disk of uniform thickness showed that the flow
was anti-symmetric about the mean position of oscillation. Vorticity shed from the edges
was found to roll up as vortex rings. These rings were convected around the edge and
diffused over time. At higher Reynolds numbers, the flow became quiie complicated and
dissimilarities were noticed between the flows during successive half-cycles.
Disk B which was of infinitesimal thickness at the edge showed very different flow
patterns compared to disk A. Vortices generated during successive half cycles formed
vortex pairs and convected away from the edges at angles approximately 45 degrees above
or below the plane of the disk. A very important observation in regards to this flow was its
dependence on the initial start-up condition. The initial condition defined the stronger of
102
the rolled up vortices, which in turn determined the direction of vortex shedding.
Willmarth (1991) has suggested that there might exist a critical disk thickness at which flow
transits from an anti-symmetric flow (e.g. disk A) to an unidirectional flow (e.g. disk B).
Identifying such a critical disk thickness would involve a detailed visualization study using
several disks of various thickness and a range of KC values. This idea is left as a topic for
future research.
In discussing the flows due to disks, the thickness of the disks was found to be an
important variable. A Keulegan-Carpenter number (KCt ) based on disk thickness was
defined to characterize some of the observed phenomena. The following conclusions were
obtained in regard to the effect of KCt :
• At near-zero values of KCt , vorticity generated by the edges was rapidly diffused
due to viscous effects, and convection was weak. The magnitudes of damping forces were
small.
• At KCt values in the order of one, diffusion and convection were equally
important The magnitude of damping forces increased in proportion with the two-fold
increase in drag coefficient (Cd).
• At very large values of KCt , convection was the dominant mechanism of vorticity
transport Based on control volume arguments, it was shown that the damping force
increased in comparison with the other cases above.
The first and second hypotheses were also confirmed by the experimental results of
Chapter 4.
Experimental force measurements have shown that the damping force experienced
by an axially oscillating cylinder was composed of two components:
- a friction drag force which varies linearly with velocity, and
- a form drag component which varies quadratically with velocity.
The damping ratio of the cylinder with disk was found to be linear for KC < 0.5. At higher
values, the dependence was nonlinear, which was postulated to be due to interaction of the
103
flows generated by the two edges. The damping ratios of cylinder with and without disk
were found to be related at small KC. This relationship in effect arises out of the deduction
that the drag coefficient for the cylinder is doubled upon addition of the disk. The present
experimental results were in qualitative acceptance with the numerical two-dimensional
results of Graham (1980) and Bearman et al. (1985).
For the tested range of p values, the effect of viscous diffusion on damping was
found to be non-negligible. The effects of viscous diffusion has not been dealt with in
previous numerical works, mainly because of the complexity of the problem. The damping
ratio was found to increase as fi increased. In the forced oscillation tests the most reliable P
obtained was at the natural frequency of the system. Higher p values were obtained with
the addition of springs, but introduced further experimental uncertainties.

The effect of currents on damping ratios of axially oscillating vertical cylinders have
been investigated. Effects of currents were more significant at lower values of KC.
Conversely, there exist large KC values at which damping ratios for different current
speeds are not very different
Several added mass results presented in this thesis were prone to experimental
errors. Typical added mass values for the cylinder were in the range of 12-15% of the
displacement This increased to about 25-30% with the addition of the disk. Forward
speeds did not affect the added mass significantly.
In Chapter 5, scaling laws were derived for the damping ratios of the columns and
pontoons of an example prototype. The scaling laws for the columns did not include the
effects of viscous diffusion. Since increasing p in effect increases the damping ratio, these
scaling laws may be viewed as conservative. Eddy viscosity corrections were applied for
damping estimation of pontoons. The derived values were converted to physical units and
summed up to obtain the total damping ratio for the TLP. These damping ratios range from
0.032 - 0.394% depending upon the cross-section of the pontoon, and addition of a disk to
the columns.
104
The principal contributions of the present research are summarized below:

• Development of a novel variant of the particle image velocimetry technique, which
is applicable to flow fields with velocities twice as fast as those used by the existing
method.
• Extensive visualization of edge flows associated with disks of different thickness.
• Examining the effect of thickness Keulegan-Carpenter number on the vortex
shedding patterns and the resulting drag mechanisms.
• Experimental evidence quantifying the trend of the damping ratio (Z) vs. KC for a
cylinder and cylinder with attached disk.
• Development of a scaling methodology for estimating the damping of a full-scale
TLP from model tests.
In conclusion, empirical design laws that are currently being used are based on past
design successes and failures. These laws, though backed by experience, cannot be an
alternative to a more complete understanding of the fundamental natural phenomena that
physically take place. A flawed design principle turns out to be an expensive error in the
long run, during operation. Notwithstanding the financial losses, there may also be a loss
of human life and environmental hazards.
Engineering is not a perfect science. Engineers must deal with nature, with
unpredictable realities - realities which constantly undergo temporal and spatial changes.
The very size of the oceans and the complex conglomeration of different natural processes
occurring therein inhibit a comprehensive human understanding. However, this cannot
prevent us from navigating these waters and exploiting their available resources. This task
carries with it the responsibility to secure human safety and environmental conservation.
By aiding in a crucial aspect of the design process, it is hoped that this thesis has
contributed to the safe deployment of Tension Leg Platforms in the deep waters of the great
seas.
APPENDICES
105
106
APPENDIX A
EXTENSION OF WANG'S THEORY TO INCLUDE A SMALL UNIFORM

STREAM
Let us consider a circular cylinder placed in a two-dimensional incompressible

flow, and let our coordinate axes be fixed on the cylinder. The oncoming flow velocity has
oscillatory and uniform components, and is given in rectangular coordinates by:
u„=(UeiO3t,U0) (A.1)
Here U is the oscillation velocity amplitude and Uq is the uniform flow velocity.
The velocity of the flow in the vicinity of the cylinder is decomposed into the free stream
and perturbation component
u=(UoJ+up). (A.2)
Introducing vorticity,
£ = (0,0,£) = VxUp, (A.3)
the unsteady Navier-Stokes equation takes the form:
U o % + % - V x ( u p x { ) = v V 2C. (A.4)
We can formulate the perturbation velocity using stream function in polar coordinates,
This formulation satisfies continuity equation identically. The relation between stream
function and vorticity is given by:
V V = -£ (A.6 )
107
The three dimensionless parameters that determine the characteristics of the flow,
namely Reynolds number (/?), Strouhal number (S) and reduced velocity parameter (V)>
are defined as:
« -as.
D
5 =£ * (A.7)
U
v = Vq_.
Deo
The Strouhal number is inversely proportional to the Keulegan - Carpenter number defined
in Chapter 1 of this thesis. More precisely,
S=^ .
KC
Following Wang (1968), the scales chosen for non-dimensionalization of the problem are:
length D
time 1 / co
unsteady velocity U
steady velocity U / S.
Wang (1968) assumes that each flow variable is the sum of a steady (bar) and an unsteady
(tilde) components. Incorporating the non-dimensionaiization described above, the
governing equations become,
V2 vr = - l (A.8 )
. QdC cos9dC
VS smd— -----------
dr r 36 = :is v2f
(A.9)
and
72—_ F
108
+ " * § •t S r % ^ ( A -n )
where the Jacobian is defined as
j™ = W j Ty
The boundary conditions on the cylinder surface (r = l)are the no-through flow conditions:
i£=o
de
(A. 12)
VS2 sin 6 + ^ - = 0
de
and the no-slip conditions:

dy
*£=o
dr
(A. 13)
VS2co s 6 + ^ - = 0
dr
The infinity condition requires that the perturbation velocity become a uniform oscillation.
This is given by:
Vr rsin 9elt, r —» °°. (A. 14)
In order to solve the problem using perturbation analysis, we need to identify a
perturbation parameter (e), which is usually small. This parameter is chosen according to
Wang (1968) as 1/VS. The Reynolds number is assumed to be of order unity and is
given by the constant a such that
*=£?.
S~ a '
The magnitude of V for this problem is chosen such that the terms in the steady boundary
conditions , Eqs. (A.12, A.13) are of the same order. Since the steady velocity in the
boundary layer is one order smaller than the unsteady velocity, we require
VS2 = k b (A. 15)
where k is a constant of order unity.
For the outer flow, we expand the variables as follows:
109
'F = 'P 0 + e ^ i + - (A.16)

Z = Z0 +eZl+ - (A. 17)
The inner flow variables axe obtained by stretching the radial coordinate using a
characteristic boundary layer thickness,
r = l+ e a 05r\ (A. 18)
The flow variables are correspondingly stretched:

yr=£y/0 + e, 2 ,yq+--- (A. 19)
C = £Co + e2£i+" ' (A.20)
Zero order solution

The governing equation for the zero order unsteady problem is a Laplace equation
(Wang, 1968):
ld _ d2w
r (A.21)
r dr dr V de?
with boundary conditions:
a tr = 1: (A.22)
de
at r —>°o T'q = rsin Qek (A.23)
Using separation of variables, the solution is obtained as:
T'o = “ “ ^sin 6ek . (A.24)
First order solution

The unsteady inner flow equation is given by Wang (1968) as:
(A.25)
[dt dr\ dr]
with boundary conditions
110
iV o (o,e)=o
dr]
(A.26)
iV o (o, e)=o
de
Boundedness at infinity can be considered an implicit condition to be satisfied by the stream
function. Using the matching principle described by Van Dyke (1975), we obtain the
solution (Wang, 1968)
yr0 = 2 a 05 sin 0^77 + " *) (A.27)
where
E= (A.28)
The first order unsteady solution satisfies the Laplace equation
d2m
ld _
L i ! 5 i =0
1 _ . 0 u (A.29)
r dr r2 ddz
with
'Fj =0, r - > 00 (A.30)
Following a procedure similar to the zero order solution, we obtain
(A.31)
The first order inner steady flow is driven by the unsteady interaction term, and
follows the equation
d2 \
d ¥ 0
d\
dAW0 __ 1_ df]2 ) (A.32)
drf a 0.5 d( 7],0)
with boundary conditions that include the uniform stream, i.e.
I ll
K s i n e + ^ - ( 0 j e) = 0
dO
_ (A.33)
^ ( ° >e ) = °
Boundedness at infinity gives an expression for the inner streaming (see also Wang, 1968),
Y q = (a)0-5 sin20 + 15 t]+ + V2(3+ 2i)E+ir]E^ + xrcos 0 .

(A.34)
The outer flow can now be solved. The equation is:
V447q = 0 (A.35)
with
—(I,0) = 1.5sin20
dr
(A.36)
^ a e )= °.
The first boundary condition here can be identified as the matching condition on the
tangential velocity. Following Wang (1968), the solution is obtained as:
% = 0.75^1 —-j)sin 2 0 . (A.37)
Composite solution
Composite solutions for the unsteady and steady parts of the stream function valid
for all r are constructed following the additive principle proposed by Van Dyke (1975).
This gives (see also Wang, 1968)
112
^ = ^ r - ^ + (/?5)“0-5V 2 ^ ^ - + ( l - 0 5 sin9eu +0(S~1)
yf = S-10.75^1 - -y^sin 29 + VS cos 9 +
+ ^_0-5S-1-5 s i n 2 0 ^ - ^ L +- ^ e~ ^ RS(r~1}+ ^2(3 + 20 ^ J + 0(S~2)
E =e ' 2J
(A.38)
The free stream contribution to the solution is the second term in the expression for yr.
Figure A.1 shows the steady stream function for two cases (a) R = 1,5 = 100, and V = 0;
and (b) R = 1, 5 = 100, and V = 1x 10-5 . The closed cell patterns of Figure A.la are
modified due to the uniform stream in Figure A.lb. Two limits for the present solution can
be identified:
a) in the limit V -» 0, the solution converges to the expression of Wang (1968).
b) in the limit /?—»<», the solution tends to the potential flow solution of a cylinder in a
flow with oscillatory and steady components.
113
i___
Figure A.1
A. 1 Nonlinear streaming flow streamlines due to an oscillating
circular cylinder at/? = 1 ,5 = 100; (a) without uniform stream (F = 0), and
(b) with uniform stream (V = 1x 10 5).
114
APPENDIX B
DIGITAL IMAGING VELOCITY MEASUREMENT SYSTEM- DETAILS
In this Appendix, we discuss the importance of various experimental parameters in

a particle image velocimetry setup. Results are presented from the numerical simulation
conducted to understand the influence of these variables. Details are also given about
existing sub-pixel methods and the error checking scheme developed during this study.
Influencing variables
Some of the several factors that influence the quality of the velocity measurement
are:
1. size of interrogation region,
2. seeding density (the number of particles per interrogation region),
3. size of a particle image (diameter), and
4. magnitude of displacement
A numerical simulation was conducted to understand the effects of variation of the
above factors. During this test, images were created on computer as an anay of grayscale
intensity values ranging from 0-255. The size of each image was set at 320 x 224 pixels,
which was approximately half the size of images obtained during our experiments. A
maximum sub-image size of 32 x 32 pixels was found to be compatible with the memory
storage and data transfer capabilities of the computing system used. This size, also used by
Willert and Gharib (1991), was found to be adequate for the range of displacements
measured during the present work.
In a typical simulation run, particle images are randomly distributed throughout the
field of view according to a defined seeding density. The position of each particle is
115
generated using a linear congruential random number generator (e.g. Knuth, 1969). The
image of a particle is assumed to have a Gaussian intensity distribution given by
f ( x - x p f + ( y - y o ) 2^ (B.l)
I = I qexp
2a1
where a is the radius and I q is the amplitude (equal to 255 in the present case). This
intensity profile has been used by other investigators, see Adrian and Yao (1985) and
Willert and Gharib (1991). The intensity value of a pixel is obtained by averaging over the
area of the pixel. To obtain the second image, each particle image is given a defined
displacement in the x and y directions (for the present discussion, we will confine
ourselves to displacements which correspond to an integer number of pixels). To obtain a
double exposed image, the particle images are given a uniform offset in the x direction in
addition to the defined x and y displacement The offset value is chosen as the length of the
interrogation region, equal to 32 pixels. The resulting image consists of both the original
particle images and their displaced counterparts.
The results from the simulation are shown in Figures B.l and B.2. For these
results, the particle image diameter is 0.2 pixel, and the displacements in both x and y
directions are 5 pixels. Figure B.l shows the effect of seeding density on the percentage of
displacement measurements that are erroneous. The erroneous vectors are defined as those
that are different from the input displacement This percentage is initially high, when the
seeding is low (3 particles / sub-image). At low seeding, the magnitude of the correlation
peak is small. This is further affected by particle images entering and leaving the
interrogation region. In this situation, quite often a noise peak is detected as the maximum
of the cross correlation array, thus leading to an erroneous displacement measurement
These deductions are in agreement with those obtained by Willert and Gharib (1991). As
seeding increases, the number of erroneous vectors is reduced. For the double exposure
technique, the erroneous vectors average to less than 5%, when the seeding is in the range
10 - 20. This value is found to increase modestly to about 7.5% at a particle image
116
diameter of 2 pixels. The single exposure scheme consistently shows fewer bad
measurements than the double exposure method. The reason for this is the increase in
noise level in the latter method, as described in Section 3.3.
Figure B.2 shows the variation of percentage of erroneous vectors with increasing
displacements for the single and double exposure schemes. These runs were conducted at
a seeding of 19 particles per interrogation region, and a particle diameter of 0.2 pixel. As
displacements increase, there is a higher probability of large number of particles leaving the
interrogation area. This decreases the number of particles that correlate well, and in turn
leads to an increase in spurious vectors. The maximum measurable displacement (MMD) is
that displacement which can be measured by a scheme with a high degree of confidence say
80% or more. The MMD for the single exposure scheme varies between 10 - 14 pixels
depending upon the seeding density. This result is in agreement with Willert and Gharib's
(1991) observation that the maximum recoverable displacement is about one-third the
window size. For the double exposure method, due to inherently higher noise levels, the
MMD ranges from 8 to 10 pixels (about one-fourth the window size). This reduction in
MMD is not a serious handicap, since the latter scheme uses low At values, and hence the
displacements measured are smaller than the former method.
Subpixel methods
It can be inferred from Sec. 3.3 that the cross correlation peak of the continuous
and the discrete intensity distributions are identical only when the displacements in x and y
directions are integer multiples of pixel spacing. In the previous section, since integer
displacements were used, the peak of cross correlation was determined by locating the
maximum value of the cross correlation matrix. In real flows, displacements usually
include fractional and integer parts. In this case, once the maximum of the discrete
correlation function is identified, further sub-pixel analysis has to be performed to locate
the peak of the continuous correlation function. In the centroiding method (Prasad et al.,
1992), the correlation peak is identified as the centroid of the pixel with maximum intensity
117
and its neighbors. Alternately, a parabolic or Gaussian curve or surface may be fitted
around the maximum of the cross correlation array to locate the peak, see for example
Willert and Gharib, (1991). It is believed by the present authors that use of either method
must depend on the size of the particles used. Prasad et al. (1992) has used a particle size
of about 1 pixel, and has found that the curve fitting method gave lower rms error than the
centroiding method. The authors (Prasad et al., 1992) have however used the latter method
because of its superiority in flows with local velocity gradients.
In order to monitor the performance of both schemes, a set of simulations were
performed at different particle image sizes and seeding of nineteen particles per sub-image,
using double exposure. In these simulations, the particle images were given an integer plus
a fractional displacement. The rms error was calculated using the difference between
measured and actual displacements, after removing all spurious vectors. Some of the
results from these runs are shown in Figure B.3. At diameters less than 1.0 pixel, the
centroiding scheme performs better than the curve fitting scheme, and vice versa at
diameters greater than 1.0 pixel. For particle sizes less than one pixel, the correlation peak
is sharp, and lies mostly within the radius of one pixel. Error in curve fitting is likely to be
accentuated because of low intensity values of the neighbors of the correlation maximum.
The centroiding scheme is thus expected to give relatively low errors as particle size gets
small. Figure B.3 shows that for both the methods, the minimum rms error is about 0.21
pixel, at a particle image diameter of 1.0 pixel. Willert and Gharib (1991) in their
simulations, obtained rms fluctuations of 0.02 - 0.08 pixels for a range of displacements
using single exposures. Due to small particle sizes used during the experiments presented
below, the centroiding scheme is used during analysis for the present study.
Error checking
The presence of a spurious vector in the results is principally due to the analysis
program identifying a noise peak rather than the true maximum in a cross correlation matrix
(Sec. 3.3). Due to random positioning of particles in the flow, it is possible for there to
118
exist 'blank spots' i.e., regions with very few particles. Alternately, due to large local
displacements, a sizable fraction of particles may leave the interrogation window. !n either
case, the correlation peak will be diminished, leading to a higher probability that some
noise peak may be detected by the program. These are identified to be inherent limitations
of the particle image velocimetry technique.
A problem rather peculiar to the double exposure technique is the presence of a
large number of uncorrelated particle images in the picture (Sec. 3.3). Typically, each of
these particle images randomly correlate with any one of the particle images in the
neighboring sub-image to form noise peaks in the cross correlation space. However, by
chance, more than one of these particles may correlate at one position to form an enhanced
noise peak. Enhanced noise peaks can be a significant source of spurious vectors because
of their larger magnitude compared to other noise peaks, in relation to the signal peak.
A simple model of the intensity function of a double exposed picture is used below
to evaluate the signal and noise peak in the cross correlation spectrum. The size of each
particle image is set equal to one pixel, and its intensity distribution is represented by a
discrete Dirac- delta function at the pixel location of the particle, of magnitude Iq. The two
dimensional discrete Dirac - delta function is defined as
8{Xi - Xj,yi - y j ) = 0, xt * xy,y(- * yy-

NM (B.2)
^ 8 ( x i -Xj,yi - y j ) = 1
i=i
The interrogation region is an array of size N x M , with n particle images present (seeding
density = 1/2 n), at locations Xj through Xj^. If the sub-image is assumed to have a
uniform background intensity (/*), then its intensity function is given by
^ ( x h yi) = X 7o5(*; ~ xj ^ i - y j ) + h (B-3)

j=h
The mean of this function can be calculated as
119
= nl0 +MN/_b_= _ n _ I I (B 4)
MN MN 0 b
Here ——I q « h and can be neglected. Hence the mean intensity is approximately equal
MN
to the background intensity. The mean-removed intensity function is then given by:
i = 2 / 05(*i - xj*yi - yj) (B-5)

Mi
Equation (B.5) can be used to obtain an expression for the sum of squared intensity of the
sub-image,
NM
Z / 2( ^ ) = /o2«- (B-6)
i=l
Let the intensity functions of the two neighboring interrogation regions be given by
h (xh yi) = Iio(xh yi) + In (xh yi)

(B.7)
h&byd = ho(xi->yd+ (î »y*)
Here I io and J20 represent the particle images of interest at their original and displaced
positions respectively. The other two components In and /21 are uncorrelated particle
images present in the first and second sub-images respectively. The cross correlation
function is the discrete convolution of h and I2 , i.e.
C(xk,yk) = Ii(.xi,yi)*l2(xi,yi)
nm (B.8)
= XAO W 7.-) h(Xi +xk,yi+yk)
i=1
= Ao*^20 + ho* ^2 1 + hi* I2 0 +hi* hi- (B.9)
The first term of Eq. (B.9) can be identified to be the signal term i.e., the term that contains
useful information, and the last three terms contribute to noise. The signal term is zero
everywhere except at the displacement of interest (xo, yo) where it has a value /02 —
ft
, due to
contributions from nj2 particle image pairs. The noise term can contribute up to n2l2 peaks
of magnitude I q. However, if s particles correlate at some point (xs,ys) then the enhanced
noise peak at this location has a value:
120
C(xk = xs,yk = ys) = I q S . (B.10)
We note that the largest noise peak is about s/n times the sum of squared intensity.
In order to implement these results, let us define a threshold parameter for an entire
image
T =~ . (B.l 1)
n
A quality parameter is defined for each sub-image as,
X /2( ^ )
t=1
where Cpeak is the estimated cross correlation peak value. The denominator of Eq. (B. 12)
is the sum of squared intensity for the sub-image. One can then identify and eliminate
measurements according to the criterion
q<T. (B.13)
This criterion if true, implies that the estimated cross correlation maximum is most
probably due to detection of a noise peak, and hence the measurement is unreliable. For a
typical image obtained from experiments, evaluating the threshold is difficult because the
number of particles correlating to form the highest noise peak (s ) is not known. The value
of s depends on the experimental setup used, and also on the seeding density n.
Before presenting a practical approach to evaluate the threshold parameter, we look
at the influence of the quality parameter q on the measurements obtained. Figure B.4
shows a histogram based on q, of the good and erroneous measurements that resulted from
a simulation run of Sec. B.2. This run is characterized by low seeding (n=6), and hence a
large number of spurious vectors initially. The interrogation region is 32 x 32 pixels
(N=M=32). In Figure B.4, the mode of the good measurements is about 0.35 and that of
the erroneous ones is about 0.25. It is clear from this figure that for any value of
threshold, there is always the probability that some good measurements will be removed,
121
and some erroneous measurements will slip through. An optimal threshold is one that lies
in between the two modes, and removes as many erroneous vectors as possible for a
minimal loss in good vectors. In order to maximize the efficiency of this error checking
process,
1. it is desirable to have the modes of the two normal distributions as far apart as
possible, and
2. the variances of the two distributions should be as small as possible.
These two characteristics strongly depend on the experimental setup used.
Once a set of displacement measurements are obtained, one can iteratively evaluate
an optimal value of threshold. Figure B.5 show the variation in the percentage of
erroneous measurements as a function of threshold parameter, T, for the same run used in
Figure B.4. Up to T=0.2, there is no significant reduction in the number of erroneous
vectors. As T is increased to 0.25, there is a marked reduction in the erroneous fraction,
accompanied by a modest increase in the number of good vectors lost Beyond T =0.3,
more good measurements are lost at no comparable reduction in erroneous measurements.
The optimal value of threshold lies in the vicinity of the intersection of the two curves, i.e.
at about 0.25-0.275.
122
5 10 15 20
Seeding density
Figure B.l % erroneous vectors vs. seeding density;

□ double exposure, A single exposure.
0 2 4 6
Displacement (pixels)
Figure B.2 % erroneous vectors vs. displacement;

□ double exposure, A single exposure.
123
0.35-1
^ °-3
□
•«
d. O-25' A
§ 0.2 E
§
§ 0.15
0. 1-
0.05-
IT 1 I 1 I 1 1 I 1
0.1 10
Diameter (pixels)
Figure B.3 RMS error in pixels of the subpixel methods vs.

particle image diameter; □ curve fitting, A centroiding.
124
m #erroneous
0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75

<1
Figure B.4 Number of good and erroneous vectors vs.
quality parameter.
50-
40-
30 f
20 -
10 -
T
Figure B.5 Effects of varying the threshold parameter;
□ % erroneous vectors remaining, A % good vectors lost
125
APPENDIX C
FLOW DUE TO AN INFINITELY LONG CYLINDER EXECUTING AXIAL

OSCILLATIONS
A circular cylinder of radius R and of infinite length is assumed to oscillate along its
axis in still fluid. A cylindrical coordinate system is chosen, whose origin is fixed
somewhere on the axis of the cylinder. Under the assumptions of fully developed and
axisymmetric flow, velocity is directed parallel to the cylinder axis, and is a function of r
and time. The governing equation for this flow deduced from the Navier - Stokes equation
is (see also Casarella and Laura, 1969; Chew and Liu, 1989):
du d2u du
= 0, (C.1)
dt V dr2 rdr
with boundary conditions

u(r = R,t) = Ueiax (C.2)
u(r -> oo,r) be finite. (C.3)
Assuming a solution of the form
u(r,t) = u1(r)eiox, (C.4)
the following equation for u(r) is obtained:
o. (c.5)
dr rdr v
This is a Bessel equation of order zero in terms of the variable [r/<5(7-1)]. 8 is a
characteristic length, typically a boundary layer thickness (e.g. Batchelor, 1973), and is
given by
S’ M . (CO
126
The solution to Eq.(C.5) can be given in terms of Bessel functions of the first and
second kind of order zero, or by Hankel functions of the first and second kind of order
zero (e.g. Abramowitz and Stegun, 1972). Use of Hankel functions enable easy handling
of the complex exponent and the boundary condition Eq. (C.3). The solution to Eq. (C.5)
that remains finite at infinite r is:
(C.7)
where is the Hankel function of the first kind of order zero. The constant c i is
evaluated from the boundary condition Eq. (C.2), and the complete expression for the
velocity distribution is:
(C.8)
The shear stress evaluated on the surface of the cylinder is given by:
(C.9)
where }i is the dynamic viscosity. Substituting for u from Eq. (C.8),
(C.10)
where is the Hankel function of the first kind of first order.

For typical values of frequency of oscillation of a TLP column, the value of S is
much smaller compared to the radius R. Hence equations for velocity and shear stress can
be approximated by using the asymptotic expressions for the Hankel functions. From
Abramowitz and Stegun (1972) we obtain
■r/S+i(r/8+itl 4)
(C.11)
127
Then
u(r,i) = u ^ j e ~ (r~R)/Sei(0X~(r~R)/S), (C.12)
and
v(r = R,t) = Hvb - V eKox-xm (C {3)
o
The real part of Eq. (C. 13) can be evaluated to be
z(r = R,t) = Q ^ c o s (c o t+ 7 r /4 ). (C.14)

d
Eqs. (C.12) differs from the expression for velocity dueto anoscillating flat plate, Eq.
(4.6), by the factor ■yjR/r (see also Casarella and Laura,1969),while .Eq. (C.14) is
identical to the expression obtained for the flat plate problem Eq. (4.7). Thus in the limit of
large radius of curvature, the friction force experienced by an infinite cylinder and an
infinite flat plate are equal.
BIBLIOGRAPHY
128
129
BIBLIOGRAPHY
Abramowitz, M and Stegun, I. A., (1972), "Handbook o f Mathematical Functions", Dover

Pub., New York.
Adrian, R. J., (1983), "Laser velocimetry", in Fluid Mechanics Measurement, R. J.
Goldstein ed., Hemisphere Pub., New York.
Adrian, R. J., (1984), "Scattering particle characteristics and their effect on pulsed laser
measurements of fluid flow: speckle velocimetry vs. particle image velocimetry", Appl.
Optics, 23, 11, pp 1690 - 1691.
Adrian, R. J., (1986a), "Image shifting technique to resolve directional ambiguity in
double pulsed velocimetry", Appl. Optics, 25,21, pp 3855 - 3858.
Adrian, R. J., (1986b), "Multi-point optical measurements of simultaneous vectors in
unsteady flow - a review", Int. J. Heat Fluid Flow, 7, pp 127 - 145.
Adrian, R. J., (1991), "Particle-imaging techniques for experimental fluid mechanics",
Ann. Rev. Fluid Mechanics, 23, pp 261 - 304.
Adrian, R. J. and Yao, C.-S., (1985), "Pulsed laser technique application to liquid and
gaseous flows and the scattering power of seed materials", Appl. Opt., 24,1, pp 44 - 52.
Anaturk, A. R., (1991), "An experimental investigation to measure hydrodynamic forces at
small amplitudes and high frequencies", Appl. Ocean Res. 13,4, pp. 200 - 208
Arkell, R. H., Graham, J. M. R. and Zhou, C.-Y., (1992), "The effects of waves and
mean flow on the hydrodynamic forces on circular cylinders", 6th Intl. Corf. BOSS, 1, pp
445 - 455.
Ashcroft, F. H., Troesch, A. W. and Sullivan, P., (1989), "Design, construction and
initial testing of a Vertical Motion Mechanism", Proc. 22nd ATTC, pp 204 - 210.
Batchelor, G. K., (1973), "An Introduction to Fluid Dynamics ", Cambridge Univ. Press,
London.
Bearman, P. W., (1992), "Fluid loading of cylinders with application to risers: Results
from model tests", NSF Workshop on Riser Mechanics, Ann Arbor.
Bearman, P. W., Downie, M. J. and Graham, J. M. R., (1982), "Calculation method for
separated flows with applications to oscillatory flow past cylinders and roll damping of
barges", Proc. 14th Syrup. Naval Hydro., pp 1151 - 1170.
Bearman, P. W., Downie, M. J., Graham, J. M. R. and Obasaju, E. D., (1985), "Forces
on cylinders in viscous oscillatory flow at low Keulegan - Carpenter numbers", J. Fluid
Mech., 154, pp 337 - 356.
Bearman, P. W. and Mackwood, P. R., (1992), "Measurements of the hydrodynamic
damping of circular cylinders", 6th Intl. Corf. BOSS, 1, pp 405 - 414.
130
Beck, R. F., Cummins, W. E., Dalzell, J. F., Mandel, P. and Webster, W. C., (1989),
"Motion in Waves", in Principles o f Naval Architecture, E. V. Lewis ed., SNAME, Jersey
City.
Bendat, J. S. and Piersol, A. G., (1986), "Random Data Analysis and Measurement
Procedures ", Wiley - Interscience, New York.
Blevins, R. D., (1977), "Flow - induced Vibrations ", Van Nostrand Reinhold Co., New
York.
Casarella, M. J. and Laura, P. A., (1969), "Drag on oscillating rod with longitudinal and
torsional motion", J. Hydronautics, 3, pp 180 - 183.
Chakrabarti, S. K., (1991), "Nonlinear Methods in Offshore Engineering ", Elsevier
Science Pub., Amsterdam.
Chakrabarti, S. K. and Hanna, S. Y., (1990), "Added mass and damping of a TLP column
model", 22nd Annual OTC, Paper 6406, pp 559 - 571.
Chakrabarti, S. K. and Hanna, S. Y., (1991), "High frequency hydrodynamic damping of
a TLP leg", Off. Mech. Arctic Eng., 1-A, pp 147 - 151.
Chew, Y. T. and Liu, C. Y., (1989), "Effect of transverse curvature on oscillatory flow
along a circular cylinder", AIAA Jl, 27, 8, pp 1137 - 1139.
Couch, R. B., Troesch, A. W. and Ashcroft, F. H., (1984), "The use of model basins in
the design of ships and marine structures", SNAME Spring STAR Symposium, pp 93 -
115.
Dalzell, J. F., (1978), "Non-linear forces on oscillating plates: Review and analysis of
literature", Report SIT-DL-78-9-2031, Stevens Institute of Technology, Hoboken.
De Bemardinis, B., Graham, J. M. R. and Parker, K. H., (1981), "Oscillatory flow
around disks and through orifices", J. Fluid Mech., 102, pp 279 - 299.
Demirbilek, Z., (1989), "Tension leg platform: An overview of the concept, analysis and
design", in Tension Leg Platform: A State o f the Art Review, Z. Demirbilek ed., ASCE,
New York, pp 1 - 14.
Douglas, J. F., Gasiorek, J. M. and Swaffield, J. A., (1985), "Fluid Mechanics ", Pitman
Co., London.
Faltinsen, O. M., (1990), "Sea Loads on Ships and Offshore Structures ", Cambridge
Univ. Press, Cambridge.
Gharib, M., Weigand, A., Willert, C. and Liepmann, D., (1992), "Experimental Studies
of vortex reconnection to a free surface: a physical flow model", 19th Symp. Naval
Hydrodynamics, Seoul.
Glauert, M. B., (1956), "The laminar boundary layer on oscillating plates and cylinders",
J. Fluid Mech., 1, pp 97 -110.
Goldstein, R. J., (1983), "Fluid Mechanics Measurements ", Hemisphere Pub. Co.,
Washington.
131
Graham, J. M. R., (1980), "The Forces on Sharp-edged Cylinders in Oscillatory Flow at

Low Keulegan-Carpenter Numbers", J. Fluid Mech., 97, p. 1, pp 331 - 346.
Graham, J. M. R., (1985), "Numerical simulation of steady and unsteady flow about
sharp-edged bodies", Int. Symp. Separated Flow around Marine Structures, pp 347 - 373.
Graham, J. M. R. and Djahansouzi, B., (1989), "Hydrodynamic damping of structural
elements", Proc. 8th Intl. Conf. OMAE., II, pp 289 - 293.
Grant, L. and Liu, A., (1990), "Directional ambiguity resolution in particle image
velocimetry by pulse tagging", Exp. Fluids, 10, pp 71 - 76.
Guezennec, Y. G. and Kiritsis, N., (1990), "Statistical investigation of errors in particle
image velocimetry", Exp. Fluids, 10, pp 138 - 146.
Hall, P., (1984), "On the stability of the unsteady boundary layer on a cylinder oscillating
transversely in a viscous fluid", J. Fluid Mech., 146, pp 347 - 367.
Hassan, Y. and Canaan, R. E., (1991), "Full-field bubbly flow velocity measurements
using a multiframe particle tracking technique", Exp. Fluids, 12, pp 49 - 60.
Hassan, Y. and T., B., (1991), "Full-field bubbly flow velocity measurements by digital
image pulsed laser velocimetry", Exp. Fluids, 11, pp 293 - 301.
Heideman, J. C., (1989), "Environmental design criteria for TLPs", in Tension Leg
Platform: A State of the Art Review, Z. Demirbilek ed., ASCE, New York, pp 21 - 35.
Honji, H., (1981), "Streaked flow around an oscillating circular cylinder", J. Fluid Mech.,
107, pp 509 - 520.
Huse, E., (1990), "Resonant heave dumping of Tension Leg Platforms", 22nd Annual
OTC, Paper 6317, pp 431 - 436.
Huse, E. and Muren, P., (1987), "Drag in oscillatory flow interpreted from wake
considerations", 19th Annual OTC, pp 135 - 142.
Jonsson, I. G., (1980), "A new approach to oscillatory rough and turbulent boundary
layers", Ocean Engng., 7, pp 109 - 152.
Justesen, P., (1991), "A numerical study of oscillating flow around a circular cylinder", J.
Fluid Mech., 222, pp 157 - 196.
Kamphuis, J. W., (1975), "Friction factor under oscillatory waves", J. Waterways
Harbors Coastal Eng. Div., 101, WW2, pp 135 - 144.
Keulegan, G. H. and Carpenter, L. H., (1958), "Forces on cylinders and plates in an
oscillating fluid", J. Res. National Bureau o f Strandards, 60,5, pp 423-440.
Kim, S. K., (1988), "Viscous Flow due to Cylinders Oscillating at Low Keulegan -
Carpenter Numbers", Ph. D. Thesis, Univ. of Michigan.
132
Kim, S. K. and Troesch, A. W., (1989), "Streaming flows generated by high-frequency

small-amplitude oscillations of arbitrarily shaped cylinders", Phys. Fluids, A 1, 6, pp 975 -
985.
Kimura, I. and Takamori, T., (1986), "Image processing of flow around a circular cylinder
by using correlation technique", in Flow Visualization IV, C. Veret ed., Hemisphere Pub.
Co., Washington.
Landreth, C. C. and Adrian, R. J., (1988), "Electro-optical image shifting for particle
image velocimetry", Appl. Opt., 27, 20, pp 4216 - 4220.
Landreth, C. C. and Adrian, R. J., (1990), "Impingement of a low ReynoL.'» number
turbulent circular jet onto a flat plate at normal incidence", Exp. Fluids, 9, pp 74 - 84.
Lian, W., (1988), "A numerical study of two-dimensional separated flow past bluff bodies
at moderate KC-numbers", App. Ocean Res., 10, 3, pp 114-119.
Lighthill, M. J., (1954), "The response of laminar skin friction and heat transfer to
fluctuations in the stream velocity", 224 A, pp 1 - 23.
Liu, Z.-C., Landreth, C. C., Adrian, R. J. and Hanratty, T. J., (1991), "High resolution
measurement of turbulent structure in a channel with particle image velocimetry", Exp.
Fluids, 10, pp 301 - 312.
Maull, D. J. and Milliner, M. G., (1978), "Sinusoidal flow past a circular cylinder",
Coastal Engng., 2, pp 149 - 168.
Merzkirch, W., (1987), "Flow Visualization ", Academic Press Inc., London.
Morison, J. R., O'Brien, M. P., Johnson, J. W. and Schaaf, S. A., (1950), "The force
exerted by surface waves on piles ", Petroleum Trans., 189, pp 149 -157.
Myrhaug, D., (1984), "On the frictional resistance of oscillating rough and smooth
surfaces", Intl. Shipbuilding Prog., 31, 360, pp 196 - 203.
Newman, J. N., (1977), "Marine Hydrodynamics ", MIT Press, Cambridge, MA.
Patel, M. H., (1989), "Dynamics o f Offshore Structures ", Butterworth & Co., London.
Pattani, P. G. and Olson, M. D., (1987), "Periodic solutions of rigid body - viscous flow
interaction", Int. J. for Num. Meth. in Fluids., 17, pp 653 - 695.
Petrauskas, C., (1987), "Springing force response of a Tension Leg Platform", OTC,
Paper 5458 , pp 333 - 341.
Prasad, A. K., Adrian, R. J., Landreth, C. C. and Offutt, P. W., (1992), "Effect of
resolution on the speed and accuracy of particle image velocimetry interrogation", Exp.
Fluids, 13, pp 105 - 116.
Riley, N., (1975), "Unsteady laminar boundary layers", SIAM Review, 17, no 2, pp 274
-297.
Sarpkaya, T., (1985), "Past progress and outstanding problems in time - dependent flows
about ocean structures", Separated Flow Around Marine Structures, pp 1 - 36.
133
Sarpkaya, T., (1986), "Force on a circular cylinder in viscous oscillatory flow at low
Keulegan - Carpenter numbers", J. Fluid Mech., 165, pp 61 - 71.
Sarpkaya, T. and Isaacson, M., (1981), "Mechanics o f Wave Forces on Offshore
Structures ", van Nostrand Reinhold Co., New York.
Sherman, F. S., (1990), "Viscous Flow ", McGraw-Hill Pub., New York.
Shih, C. C. and Buchanan, H. J., (1971), "The drag on oscillating flat plates in liquids at
low Reynolds numbers", J. Fluid Mech., 48, 2, pp 229 - 239.
Simpkins, P. G. and Dudderar, T. D., (1978), "Laser speckle measurement of transient
Benard convection", J. Fluid Mech., 89, pp 665 - 671.
Smith, P. A. and Stansby, P. K., (1991), "Viscous oscillatory flows around cylindrical
bodies at low Keulegan - Carpenter numbers using the vortex method", J. Fluids
Structures, 5, pp 339 - 361.
Stuart, J. T., (1966), "Double boundary layers in oscillatory viscous flow", J. Fluid
Mech., 24, p 4, pp 673 - 687.
Tatsuno, M. and Bearman, P. W., (1990), "A visual study of the flow around an
oscillating circular cylinder at low Keulegan-Carpenter numbers and low Stokes numbers",
J. Fluid Mech., 211, pp 157 - 182.
Telionis, D. P., (1975), "Calculations of time-dependent boundary layers", Unsteady
Aerodynamics, 1, pp 155 - 190.
Telionis, D. P., (1981), "Unsteady Viscous Flows ", Springer - Verlag, New York.
Troesch, A. W., (1990), Personal Communication.
Troesch, A. W. and Kim, S. K., (1991), "Hydrodynamic forces acting on cylinders
oscillating at small amplitudes", J. Fluids Structures, 5, pp 113 - 126.
Tsahalis, D. T. and Telionis, D. P., (1974), "Oscillating laminar boundary layers and
unsteady separation", AJAA Jl., 12,11, pp 1469 - 1476.
Van Dyke, M., (1975), "Perturbation Methods in Fluid Mechanics ", Parabolic Press,
Stanford.
Van Dyke, M., (1982), "An Album of Fluid Motion ", Parabolic Press, Stanford.
Wang, C.-Y., (1968), "On high frequency oscillatory viscous flows", J. Fluid Mech., 32,
p 1, pp 55 - 68.
Westerweel, J. and Nieuwstadt, F. T. M., (1991), "Performance tests on 3-dimensional
velocity measurements with a two-camera digital particle image velocimeter", Fourth Intl.
Coif. Laser Anemometry, Advances and Appl., ASME, 1, pp 349 - 355.
Willert, C E. and Gharib, M., (1991), "Digital particle image velocimetry", Exp. Fluids,
10, pp 181-193.
134
Williamson, C. H. K., (1985), "Sinusoidal flow relative to circular cylinders", J. Fluid

Mech., 155, pp 141 - 174.
Willmarth, W. W., (1991), Personal Communication.
Wozniak, K., Wozniak, G. and Rosgen, T., (1990), "Particle-image-velocimetry applied
to thermocapillary convection", Exp. Fluids, 10, pp 12 -16.

1993 Hydrodynamics of Oscillating Cylinder and Disks at Low Keulegan-Carpenter Number

Uploaded by

Copyright:

Available Formats

1993 Hydrodynamics of Oscillating Cylinder and Disks at Low Keulegan-Carpenter Number

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1993 Hydrodynamics of Oscillating Cylinder and Disks at Low Keulegan-Carpenter Number

Uploaded by

Copyright:

Available Formats

MICROFILMED 1993

The quality of this reproduction is dependent upon the quality of the

Oversize materials (e.g., maps, drawings, charts) are reproduced by

Photographs included in the original manuscript have been reproduced

University Microfilms International

H ydrodynam ics o f oscillating cylinders and disks at low

Thiagarajan, Krishna Prasad, Ph.D.

A dissertation submitted in partial fulfillment

Microfilmed or bound copies of doctoral dissertations sub­

Life is nothing but a journey..

4.1 Experimental setup of Vertical Motion Mechanism (VMM) and test

4.6 Damping ratio of vertical cylinder vs. KC; £=89236................................. 84

4.9 Damping ratio for vertical cylinder vs. KC at different p....................86

4.16 Added mass coefficient vs. KC of vertical cylinder at different p. 89

L length of horizontal cylinder

Skc Uniform flow velocity parameter

x horizontal length coordinate

There is an increasing interest among oil exploration companies towards drilling in

Figure 1.1 Schematic diagram of a Tension Leg Platform

Figure 1.2 Component problems of a TLP

D is chosen as the diameter of the cylinder.

KC^n = 5.71 Sp"0'25(1+0.205/T'°'25+- •-) (2.1)

2.1 Inline force on a cylinder

In offshore applications, Morison's equation (Morison et al. 1950) is used to model

Force measurements were carried out by Sarpkaya (1986) in his U-shaped

that the coefficient of the friction drag component would vary as

Q (friction) «=KC~0AJ3~0'Z. (2.6)

2.2 Effect of currents

Inline oscillations of a cylinder in a uniform flow has attracted attention partly

velocity parameter, Skc>to be

where V is the reduced velocity parameter (Appendix A). As an example, consider a

0 - 1 1_______ I____ I___ I. 1 1, M I- 1 ________ I____ 1 ' ! I I I I

Figure 2.2 Drag coefficient vs. {3for circular cylinders in

3.2 Flow visualization methods

Flow visualization has played an important role in understanding several fluid

3.3 The basic concept

h Oc,y) = h(x,y)* S(x - A x , y - Ay) (3.2)

where the delta function is defined as:

The cross-correlation function C(X,Y) of/i(x,y) and I2{x,y) is given by

= Il*(l1* S (x - Ax,y - Ay))

= f/i *7j )* 5(x - Ax, y - Ay)

= A(X,Y)* 8(x - Ax,y - Ay)

Cij = - f y i - i j ~ AxiJ+1 + A xij-i). (3.7)

Effect of At on measurement accuracy

If the flow under consideration is oscillatory, say

(E) = E2(a),t,At)dt (3.10)

correction scheme, whose principles are described in Appendix B, is used to effectively

3.4 Experimental Setup

^oloHill Engineering Inc., Ann Arbor.

Table 3.1 Parameters of flow visualization experiments

Water tank area 0.61 m x 0.61 m (24 in x 24 in)

Cylinder diameter 0.11 m (4.5 in)

Disk A diameter 0.15 m (6.0 in)

DiskB diameter 0.15 m (6.0 in)

Frequencies of oscillation 0.1 - 1.0 Hz

Amplitudes of oscillation 3.17 m m -25.4 mm

3.5 Results and discussion

Ability of the visualization techniques to produce useful results depends on the

Results for the oscillating thin disk (disk A)

Effect of disk thickness on drag force

Microfilmed or bound copies of doctoral dissertations sub

0 - 1 1_____ I I_ I. 1 1, M I- 1 ____ I 1 ' ! I I I I

= Il(l1 S (x - Ax,y - Ay))

= f/i 7j ) 5(x - Ax, y - Ay)

..•/. .-•/ .•

• .;• .;• .. ..♦ . .• .„• .

i ■,,t 1t7 t 9r1»11i9111

y(t) = Aei(0+ \ (4.22)