You are on page 1of 151
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 affect 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.9., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overtaps. ProQuest Information and Leaming 300 North Zeb Road, Ann Arbor, MI 48106-1346 USA. ‘800-621-0600 ® UMI HYDRODYNAMICS OF THIN PLATES by Haiping He A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Naval Architecture & Marine Engineering) in The University of Michigan 2003 Doctoral Committee: Professor Armin Troesch, Co-chair Associate Professor Marc Perlin, Co-chair Professor Robert Beck Professor Robert Krasny UMI Number: 3079458 UMI UM! Microform 3079458 Copyright 2003 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest information and Learning Company '300 North Zeeb Road P.O. Box 1346 ‘Ann Atbor, Ml 48108-1346 Haiping He ‘All Rights Reserved ACKNOWLEDGMENTS My foremost thanks go to my academic advisors: Professor Armin Troesch and Professor Mare Perlin. Without their guidance, encouragement and supervision, 1 couldn’t have accomplished this thesis. 1 enjoyed their discussions on various aspects of my research. Two professors’ insights, creative suggestions have helped shape my research skills, and their good sense of humor have made this four and half years’ study a wonderful journey for me. They have been there for me even in my personal life when I needed help. I thank them for being my advisors, committee chairs and my great teachers. Special thanks are to Professor Robert Beck and Professor Robert Krasny for consenting to be on my doctoral committee and for taking the time to guide me through my dissertation. I would like to acknowledge the following sources for providing financial support: the University of Michigan/Industry Consortium in Offshore Engineering whose industry participants include BP-Amoco Corporation, Conoco Corporation, Chevron Petroleum Technology Company, ExxonMobil Upstream Research Company, Shell Companies Foundation, Aker Maritime, and Petrobras Research and Development; and the discretionary Fellowship from the University of Michigan. Many thanks go to the staff in Marine Hydrodynamics Laboratory for helping with the experimental work. I appreciate Ziyuan Liu, Kevin Maki, and Matt Lake for their help in conducting experimental work. My colleague Keary Lay offered not only valuable suggestions in experiment techniques but also joyful company. My appreciation goes to the entire Naval Architecture and Marine Engineering Department for the faculty’s teaching that has educated me, for the administrative staff's assistance with all the paperwork, and for my fellow students’ friendship. Lowe special thanks to my dear wife Xuemei for her trust and confidence in me. She encourages me and never doubted my abilities for achieving my goals. My family in China have been supporting me and giving me strength all these years. To them, I forever owe a debt of gratitude. iii TABLE OF CONTENTS ACKNOWLEDGMENTS... LIST OF FIGURES............. LIST OF TABLES..... CHAPTER I. INTRODUCTION..................0.... L.1 Motivation...... 1.2 Flow physics background .......csssssoesecessesesseeeesssssesnsssnensennend 1.3 Literature review... 1.3.1 Hydrodynamic damping of TLP column: 1.3.2 Damping-augmenting devices.............. E 1.3.3 Flow physics of oscillating edges. 9 1.3.4 Limitation of previous work... 1.4 Thesis outline...... I. FORCE MEASUREMENTS: DAMPING AND ADDED MASS COEFFICIENTS... 2.1 Introduction. 2.2 Hydrodynamic force coefficients preliminaries............. 2.3 Description of experiment... 2.3.1 Measurement techniques, mathematical model, and data reduction......... 2.3.2 Experimental setup. 2.3.3 Experimental models. 2.4 Results and discussions... 2.4.1 The quality of experimental data. 2.4.2 Effect of KC numbers. 2.4.3 Effect of thickness-to-diameter ratio.. 2.4.4 Effect of frequenc; 2.4.5 Effect of reinforcing structures... 2.4.6 Hydroelastic effects... 2.4.7 Effect of corner radius .. 2.4.8 Total force and phase lag between force and displacement..56 2.5 Summary. 58 Ill. FLOW PHYSICS STUDY: VORTEX FORMATION MODES AND SCALING LAWS FOR DAMPING COEFFICIENTS. 3.1 Introduction. 3.2 Experimental methods. 3.2.1 Flow visualization and DPIV.. 3.2.2 Experiment setup and data reduction.. 3.3 Flow visualization results and discussions... 3.3.1 Vortex shedding mode and KC number dependenc 3.3.2 Vortex shedding mode and thickness-to-diameter dependence. 3.3.3 Vortex shedding mode and frequency dependence... 3.3.4 Initial condition influence... 3.4 Scaling laws for damping coefficients... 3.4.1 Friction drag and form drag. 3.4.2 Scaling laws for damping coefficient. 3.4.3 Drag coefficient C, based on a new characteristic velocity... 3.5 Summary... IV. FLOW PHYSICS STUDY: QUANTITATIVE ANALYSIS. 4,1 Introduction. 4.2 Analysis methods for organized quasi-periodic unsteady flow.......103 4.2.1 Phase-averaging method... 4.2.2 Frequency-domain filtering. 4.2.3 Moving averaging. 43 Description of experiments... 4.3.1 DPIV... 4.3.2 Hydrogen bubble flow visualization. 4.4 Coherent structure and phase-averaging analysis. 4.4.1 Cycle number N. 4.4.2 Global mean vorticity field.. 4.4.3 Phase-averaged velocity field and vorticity contour... 4.4.4 Reynolds stresses... 4.4.5 Spectral analysis. 4.5 Cycle-to-cycle variation effects on Reynolds stress... 4.5.1 Spatial variation and vorticity core trajectory... 4.5.2 Moving-averaging and Reynolds stress... 137 4.5.3 Frequency domain filtering and Reynolds stress... 141 4.6 Vortex ring formation, development, decay and Reynolds stresses..152 4.7 Summary. V. CONCLUSIONS AND FUTURE WORK.............. 168 5.1 Summary and conclusions... 5.2 Future work... BIBLIOGRAPHY. ol TS Lt 24 22 23 24 25 2.6 27 2.8 29 2.10 241 2.12 LIST OF FIGURES TLP and spar platform illustration (U.S. Minerals Management Service 2000)... . ‘Schematic of forced vertical oscillation method.. ‘Schematic of experimental setup one... ‘Schematic of experimental setup two (MAM). ‘Two-disk configuration... Time histories of the force and displacement for the 1/87.5 disk (KC=0.185, f-4Hz, B= 1.24x10°) . Time histories of the force and displacement for the 1/87.5 disk (KC=0.375, f=4Hz, B= 1.24x10°)... Amplitude spectrum of force. Amplitude spectrum of displacement... Estimated maximum error in damping coefficient measurement (percentage)... Added mass and damping coefficient comparison (f=4Hz and 24x 10° for 1/25 disk, &3.91Hz and f=1.28x10° for 1/24 disk) Repeatability test results taken one year apart (/D=1/87.5, f=4Hz, B=1.24x10°)... Quality of the MAM system. viii 20 26 27 28 31 32 33 33 35 37 38 39 2.13 214 215 2.16 217 2.18 2.19 2.20 221 2.22 31 32 33 34 35 3.6 KC number dependence (/D=1/87.5, F4Hz, B=1.24x10*). ‘Thickness-to-diameter ratio effect (f=4Hz, A=1.24%10°)... B dependence, disk vd = 87.5, frequency ranging ftom 2, 4, 8 Hz (B=0.62x10*, 1.24x10*, and 2.47x10°) ... Reinforcing structure effect, Dd2/Dd! ranging from 0.85, 0.9, 0.95, 1 (tU/D=1/41, 2/D=1/175, f4Hz, B=1.24x10*)..... Hydroelastic effects: uncorrected for deflection (VD=175, B=0.62x10*, 1.24x10°, and 2.47x10* Hydroelastic effects: corrected for deflection (/D=175, 8=0.62x10°, 124x105, and 2.47x10°).... Round disk model... Comer radius effect (#=1.24x10°).. Total force vs. KC (f4Hz, 8 =1.24x10*)... Value sin(phase angle difference) vs. KC (f=4Hz, f =1.24x10°).. Schematic of experimental setup. PAP2K PIV analysis illustration. Vortex formation Mode 2 (KC=0.049, UD=1/87.5, f-4Hz, (B=1.24x10°). Vortex formation Mode 3 (KC=0.102, vD=1/87. A=1.24x10*).... Schematic of vortex formation Modes 1, 2, 3, 4. 42 45 47 50 52 52 53 35 56 37 65 4 15 16 1 ay 38 39 3.10 BAL 3.12 a7 3.14 3.15 3.16 3.17 3.18 3.19 41 42 43 44 45 Vortex formation Mode 2 (Ke B=1.24x10°). 110, wD: (37, f4Hz, Vortex formation Mode 3 (KC=0.324, wD=1/37, f=4Hz, (B=1.24x10°). Vortex formation modes (VD=1/87.5).. Vortex formation Mode 3 (f =2Hz,8=0.62x10°, KC=0.072, wD=1/87.5).... Vortex formation Mode 4 (f WD=1/87.5)... Hz, 8 =0.62x10°, KC=0.136, Initial condition studies: starting from the top ((VD=1/175, KC=0.09, 8 =0.62x10*, vortex formation Mode 4).......... . Initial condition studies: starting from the bottom (vD=1/175, KC=0.09, 8 =0.6210°, vortex formation Mode 4).. Damping coefficients vs. KC number... Nonlinear damping curve... Damping curves based on KCt... ‘Schwarz-Christoffel mapping. Schematic of velocity at mid point. Schematic of oscillating disk.. Schematic of the disk model (YD=1/112)... ‘Schematic of experimental setup... Ilustration of position 1, 2, 3 (VD=1/87.5, B=0.62x10°, ax = 270", where at = 0 corresponds to the top of stroke).. Phase averaged velocity as a function of phase: U component. Phase averaged velocity as a function of phase: V component... 719 80 81 82 87 91 93 94 98 99 42 12 114 1s 116 46 47 48 49 4.10 4.12 4.13 4.14 4.15 4.16 4.17 418 Evolution of normalized phase averaged velocity as a function of cycle number. Single velocity realization (N=1) for positions 1, 2, and 3: U component (phase averaged velocity (N=32) is plotted as reference).. Single velocity realization (N=1) for positions 1, 2, and 3: V component (phase averaged velocity (N=32) is plotted as reference).. Global vorticity mean of two runs with (a)KC=( .072 and (b)0.136.... Phase averaged velocity and vorticity fields (vD=1/87.5, B=0.62x10°, KC=0.136), Notes: Frames are spaced in 45 degree phase increments, maximum @=+66 sec’' (¢), mininum @=-96 sec @)... Vorticity contours and Reynolds stress contours at phase 0, 90, 180, 270 degrees (YD=1/87.5, 8 =0.62x10°, KC=0.136). Location of sample position (VD=1/87.5, 6 =0.62x10*, ax = 270", where at = 0 corresponds to the top of stroke)... Time evolution of Reynolds stress (u’u’), (v’v’), and (u’v’) as a function of phase (VD=1/87.5, 8 =0.62x10*)... Averaged Reynolds stress (a) (u’u’),., (b) (v’v’),,.. (c)(u'v) Velocity U component amplitude spectra (YD=1/87.5, 6 =0.62x10°).. Velocity V component amplitude spectra (VD=1/87.5, 8 =0.62x10°).. Spatial variation of the vortex core at phase 225 degree (/D=1/87.5, B=0.62x10°)... Trajectories of the core of the negative vortex spanning one cycle for 32 different realizations. The disk is shown at the mean position for 7 11g 119 121 124 126 128 129 130 134 135 138 138 4.19 4.20 421 4.22 4.23 4.24 4.25 4.26 427 428 429 430 Cycle-to-cycle variation at phase 264°(bottom view), wD: B=0.62x10*... 112, Reynolds stresses (u“u"), (v"v"), and (u“v") after removing of cycle- to-cycle variations using moving averaging method (vD=1/87. B=0.62x10*).. Moving averaging M-dependence (averaged Reynolds stresses).. Amplitude spectrum of Position 4 (U component) Amplitude spectrum of Position 4 (V component). Identification of cut-off frequency. Three series of U, V comparison... Reynolds stress (uu), (v*v"), (u"v")based on the improved filtering method (IFFS) (YD=1/87.5, 8 = 0.62x10*). Vorticity contours (/D=1/87.5, KC=0.136, and # =0,62x10*)... Reynolds stresses (1)(u“u”), (2)(v"v") , (3){u*v") (WD=1/87.5, KC=0.136, and § =0.62x10°)... Vorticity contour, turbulence kinetic energy, bottom view.. Total kinetic energy and total turbulent kinetic energy. xii 139 140 141 144 145 146 148, 149 157 158 162 166 24 22 23 24 25 2.6 27 3.1 32 41 42 43 44 4s 46 47 48 LIST OF TABLES. Experimental parameters.. Equipment resolution. Disk models... Higher harmonic components... Parameters of the six sub-division categories.. Test condition for Lake et al. (2000) experiment... Linear regression curve calculation for total force. Experimental parameters (disk thickness-to-diameter ratio= 1/87.5).. Experimental parameters (disk thickness-to-diameter ratio= 1/37). Experimental conditions. Sample position coordinates. Parameters for global mean study... Sample points coordinates for phase-averaged analysis Sample points coordinates for spectral analysi Sensitivity analysis of position 1 (averaged Reynolds stress).. Four methods comparison for position | (averaged Reynolds stress)...... Four methods comparison for position 4 (averaged Reynolds stress). xiii 25 28 33 34 36 37 68 B 108 4 + 120 128 132 150 151 151 CHAPTERI INTRODUCTION L.1 Motivation In the world energy market, demand is increasing rapidly, but previously explored energy reserves are very limited. Exploring new reserves in hostile environments such as deepwater regions becomes critical for future energy supplies. The offshore oil industry recognizes that these waters hold the key to its future survival and success. Worldwide there are already twenty-eight fields in production in waters deeper than 500 meters. Smith Rea Energy Analysts (SREA), UK, forecasts about seventy additional deepwater fields will be onstream by 2004 (Thomas 1999). The challenging deepwater environment demands cutting-edge technology and innovative platforms, among which Tension Leg Platforms (TLP) and spar platforms (Figure 1.1) are competitive candidates. TLPs have become an attractive option in deepwater production because of their deepwater capability and low cost compared to bottom-fixed platforms. Spar platforms are attractive because of their excellent motion characteristics when in deep draft mode and their large crude oil storage capacity. Both TLPs and spar platforms may experience resonant oscillation in heave due to the large range of water wave frequencies and second order effects. As a result, the damping of the system becomes critical. The underwater structure of a TLP consists of vertical columns and horizontal pontoons. The structures are taut-moored to a foundation on the sea floor by tendons, which are pre-tensioned to increase the stiffness in heave. The natural period of the heave motion is typically two to four seconds. Although the natural period is less than most ocean wave periods, small non-linearities in the incident waves will cause sum frequency resonant oscillation. Figure 1.1 TLP and spar platform illustration (U.S. Minerals Management Service 2000). The underwater structure of a deep-water spar structure is comprised of a long cylindrical hull supporting the risers, and is held in place by a semi-taut spread mooring system. From a design and operational viewpoint, a spread mooring system is preferred over a catenary mooring system due to capital cost reduction and a smaller required watch circle during operation. However, a short mooring can lead to resonance in heave motion. The value of the heave natural period of spar platforms is 25 to 30 seconds, which is higher than most wave periods. However, small non-linearities in the incident waves will cause difference frequency resonant oscillation. Damping becomes critical at resonance for both types of platform. For a typical dynamic system at off-resonant conditions, damping is unimportant and the response is mainly dependent on the system inertia, system stiffness, and external excitation. However, at or near resonant conditions, the stiffness and inertia forces cancel and the response is governed solely by the ratio of excitation to damping. Unfortunately, both types of platforms are lightly damped systems. While the magnitude of the exciting force may be small, the response of the system may not be negligible due to very low damping in the system. As a result, the estimate of the hydrodynamic inertia and damping of the system is essential to ascertain the response of the structure in this resonant condition. Furthermore, drag-augmenting devices may be required to limit the response amplitude to a safe range. Earlier research (see Section 1.3 Literature Review) has studied the damping of ‘TLP columns, appendages, and circular or square plates. Much of the previous research is focused on force measurements and some of these efforts used questionable testing techniques. The geometry dependence (such as thickness-to-diameter ratio) of a circular plate was acknowledged, but not studied. The goal of this research is to study the damping characteristics of circular plates. In this thesis, the problem is investigated from both a force measurement and an underlying flow physics viewpoint. The research covers geometry dependence, which includes thickness-to-diameter ratio, reinforcing structures, and edge radius. In addition, parametric dependence, such as oscillation amplitude and frequency, is studied. The results of this work contribute to the fundamental understanding of the physics of time dependent flow around sharp edges and provide quantitative guidance for engineering practice. The methodology developed here can also be applied to other types of damping, devices and unsteady quasi-periodic flows. 1.2 Flow physics background. The problem of interest can be modeled either as circular plates oscillating in an infinite fluid or fixed plates in oscillating flows. Two fundamental non-dimensional parameters for such flow are the Keulegan-Carpenter number (KC) and the frequency parameter (8 ). The KC number is defined as: 27a KC =—— Ld D ay) where D is a characteristic length (here the diameter) of the circular plate and a is the amplitude of disk motion. The KC number, then, is the amplitude of oscillation relative to the diameter of the disk. The magnitude of the KC number indicates the relative importance of drag to inertia forces (Sarpkaya and Isaacson, 1981). The # number (frequency parameter), as defined by Sarpkaya (1975), is given by: B as (1.2) v where v is the kinematic viscosity of the fluid and f is the oscillation cyclic frequency. Considering laminar boundary theory, represents the ratio of the rate of momentum diffusion through the boundary layer to the rate of momentum diffusion across the diameter (Sarpkaya and Isaacson, 1981). Two parameters, # and KC, are related by the Reynolds number in the following manner: R, = BxKC (3) 1.3 Literature review Previous research studied the damping characteristics of TLP columns, appendages, and circular plates. The following is a review of the existing work. 1.3.1 Hydrodynamic damping of TLP columns ‘As mentioned previously, estimates of the hydrodynamic inertia and damping of the system are essential to ascertain the response of the structure at resonant conditions. Earlier research focused on the determination of damping coefficients due to the underwater portion of floating structures. Previous work on TLPs assumed that the total system hydrodynamic damping could be expressed as the net sum of the individual members, i.e. vertical columns and horizontal pontoons. Horizontal pontoons can be modeled as oscillating horizontal cylinders, which have been well researched due to their wide application in offshore engineering. Vertical columns can be modeled as cylinders oscillating parallel to their axes (axially oscillating cylinders). Several research projects have been conducted to study the hydrodynamic behaviors of axial oscillating cylinders. Huse (1990) tested a cylinder with KC ranging from 0.0005 to 0.01 and # approximately 5x10°. His results showed that the drag force varied linearly with velocity. Chakrabarti and Hanna (1990, 1991) reached a similar conclusion from their tests with KC=0.126 and f ranging from 0.25x10° to 1x10*. Huse and Utnes (1994) also placed a TLP column in a current and the results showed that the current increased the damping over the range of KC being tested. Thiagarajan and Troesch (1994) reported a nonlinear trend between the drag force and velocity for axial oscillating cylinders conflicting with Huse’s earlier results. The tested KC numbers ranged from 0.1 to 1.0 and f equaled 0.89%10*. To understand the reasons for the different behaviors in the drag force model, recall that the cylinder's hydrodynamic drag force can be decomposed into its friction and form (pressure) drag components (Thiagarajan and Troesch, 1994 and 1998). The friction drag is due to the viscous tangential stress acting along the walls of the cylinder. The form drag is mainly due to the separation at cylinder edges. At very low KC, the drag is due primarily to the effect of the surface area of the cylinder wall. The friction drag varies linearly with velocity, while the form drag is nonlinear with velocity. Alternatively, in relation to KC, the friction drag is KC-independent, while the form drag is linear with KC. Huse, Chakrabarti, and Hanna's experiments were tested at KCs from 0.0005 to 0.126, where friction drag is dominant, Conversely, Thiagarajan and Troesch’s data covered KC from 0.1 to 1.0, where form drag becomes dominant. The transition from one drag model to the other apparently occurs in this KC range. O'Kane et al. (2002) studied the hydrodynamic interaction between the individual components of a TLP. A direct comparison between the summation of individual hull component coefficients and the coefficients evaluated directly from complete hull configuration models was conducted. The comparison was based on the experimental tests of the hydrodynamic coefficients for individual hull components as well as partial and complete platform models. The results indicate that the hydrodynamic interaction effects between components are small for TLP heave in low current environments. Although most of the research originated from TLPs, the results of the axial- oscillating cylinder research are also applicable to spar platforms. The underwater part of a spar platform is comprised of a long cylindrical hull, which can also be modeled as an axially oscillating cylinder. 1.3.2 Damping-augmenting devices As a further step to the above research, damping-augmenting devices designed to actively reduce the heave motion were studied. Different forms of devices, such as tubes, appendages, and plates, have been proposed and researched. Srinivasan et al. (1991) showed through experiments that an array of small diameter diamond-shaped tubes increased the inline drag coefficient for a cylinder by as much as five times at low KC numbers. Thiagarajan and Troesch (1998) examined the effect of adding an appendage in the form of a disk to TLP columns. The model tests conducted in heave on a cylinder plus disk configuration showed that the heave damping induced by the disk is linear with the amplitude of oscillation. The disk was found to increase the form drag coefficient by twofold. The test was conducted at # = 0.89%10* and KC range of 0.1~1.0. Lake et al. (2000) investigated three possible configurations of TLP/spar platforms and the results show the addition of a disk to the base of the column can enhance the damping but does little to increase the added mass. Separating the disk and cylinder nearly doubles the added mass and increases the damping ratio by 58 percent over the attached cylinder plus disk platform and an impressive 344 percent over the single column. Prislin (1998) experimentally studied the variation of added mass and damping of both the single plate and a multi-plate arrangement for a spar platform. He did not include the effect of the vertical column. The tested Reynolds number ranged from about 4.510? to 1.8x10° and the KC number ranged from about 0.1 to 1.0. His results show that the drag coefficient of an oscillating plate is dependent on both Reynolds number and KC number for Reynolds number less than 1.010°. At higher Reynolds number the drag coefficient does not vary significantly with the Reynolds number and it is significantly lower than the drag coefficient measured at low Reynolds number. Prislin’s results were based on decrement tests, which produce larger experimental error at large KC compared to forced oscillation testing at the same KC values. Magee et al. (2000) discussed the application of square plates to truss spars. His experimental results and numerical prediction showed that square plates help to significantly reduce the heave response of a truss spar. He also argued that the loads on the square plate can be computed accurately using Morison’s equation. Molin (2001) used a matched eigen-function expansion method to solve the porous disk problem. He compared his calculation result of a porous disk with Lake et al.’s (2000) solid disk results and reached following results: a .9SKC* (1.4) tt where Brora is the linear damping coefficient for porous disks and B,,, is the counterpart for solid disks. The result shows that when the KC number is larger than one, no extra damping can be gained by making the disk porous, but as KC becomes smaller we may see an increase in damping. As pointed out by Molin, the results need validation through dedicated model testing. 1.3.3 Flow physics of oscillating edges Much research has been done to study the flow physics of oscillating cylinders. ‘Sumer and Fredsoe (1997) gave a comprehensive review of this topic. Only limited studies have been conducted on axial oscillating plates. We start with early research on flow physics of oscillating cylinders with sharp edge cross-sections (Graham 1980). Graham (1980) studied the damping coefficients for different cylinder cross- sections. His results showed that the pattern of vortex shedding from a single isolated edge consists of one vortex pair for each cycle, i.e. each half cycle generates one vortex which then form a pair. The first vortex is generated at the first part of the cycle due to separation of the stream passing the edge. As the stream slows down and reverses its course, a second vortex starts to grow on the other side of the edge. The two vortices appear to convect away rapidly once the second vortex has reached an approximately equal and opposite strength to the first. He also pointed out that the pairing process might split the second vortex sheet since the equalization of vorticity does not necessarily occur at the end of the second vortex period of growth. In that case a smaller amount of residual vorticity is left and engulfed by the next strong vortex. De Bemardinis (1981) extended Graham’s discrete vortices model to axisymmetric flows and studied the oscillatory flow around a disk both theoretically and experimentally. The formation of vortex pairs and their unidirectional convection were observed in both his theoretical prediction and flow visualization experiments (disk vD=1/25 and KC=3.5). During the first half-cycle the shed vorticity rolls into a single vortex ring. As the free stream slows, the flow at the edge of the disk reverses and a second vortex with opposite sign starts to form and is shed during the second half cycle. 10 Part of this second vortex interacts with the first vortex to form a vortex pair and convects away from the disk. The residual of the second vortex sheet is neither as strong nor as organized as the first vortex and it does not interact strongly with the third vortex when the flow again reverses at the start of the second cycle, The third vortex sheet develops into an organized vortex ring very much like the first one and the process is repeated in succeeding cycles. It was also noticed by De Bernardinis that a stable asymmetric wake is established and its direction depended only on the starting direction of the flow. Thiagarajan (1993) conducted flow visualization studies on both a disk with uniform thickness and a disk with sharp edge, which had infinitesimal thickness at the edge. His results showed that the flow was anti-symmetric about the mean position of oscillation at large KC number. Vorticity was shed from the disk edges and rolled into vortex rings, which were convected around the disk edges and diffused over time. Similar to De Berardinis (1981), it was also observed that the shedding direction of vortex pairs ‘was dependent on the initial start-up condition. Basically, initial conditions defined the stronger vortices, which determined the direction of vortex shedding. At small KC number, the disk flow was symmetric about the mean position of oscillation and the gross features of the flow were periodic and repeatable. At higher Reynolds numbers, the flow became quite complicated and dissimilarities were noticed between the flows during successive half-cycles. Lake et al. (2000) completed a series of flow visualization experiments on both a disk and a cylinder plus disk configuration. The flow parameters were f of 1.28x10*, disk thickness-to-diameter ratio of 1/24, disk diameter to cylinder diameter ratio of 1.32, and KC of 0.34. For the cylinder plus disk configuration, a symmetric vortex pattem was u observed. The presence of a vertical cylinder had little influence on the comer flows of the disk in the KC studied. It is expected that at larger KC numbers, the cylinder would start to influence the vortex ring. For the disk-only configuration, similar symmetric vortex patterns were observed. Tao (2001) examined the flow patterns around the disk edges of a cylinder-disk configuration numerically using a finite difference method based on boundary fitted coordinates. The flow parameters were 8 of 0.89x10°, disk thickness-to-diameter ratio range of 1/1330~1/2.23, disk diameter to cylinder diameter ratio of 1.33, and KC range of 0.00075~0.75. Three distinct vortex patterns, i.e. an independent vortex shedding regime, an interactive vortex shedding regime, and a unidirectional vortex shedding regime, were observed near the edge of disks and each of them related to certain ranges of KC values. 1.3.4 Limitations of previous work The above research provides rich guidance for engineering design and lays the foundation for further research. However, the following issues remain regarding circular plates: ‘© Thickness-to-diameter ratio dependence: Most of the existing research did not address the effect of plate thickness. Those that did only acknowledged the effect but did not study it. The results from the present study show that thickness affects the damping coefficient dramatically, especially for KC range of 0.1 to 1.1. © Vortex formation pattem: Two different vortex formation pattems were acknowledged. However, the relationship between vortex formation patterns 12 and damping coefficient has not been clearly identified, partly due to limited available experiment data. In this thesis, a systematic flow visualization study on the KC number and thickness-to-diameter ratio effects on the vortex Pattern is presented helping to understand the flow physics and its contribution to damping force. * Quantitative analysis: The existing flow physics studies are all based on qualitative flow visualization experiments. No quantitative analysis has been attempted. A quantitative study is presented here helping to decompose the coherent vortical structures and turbulent fluctuations. © Azimuthal effect: All the existing research assumes the flow is axisymmetric. ‘As demonstrated in this thesis, the generation, development, and dissipation of time varying vortex rings are three-dimensional phenomena. ‘A thorough understanding of flow physics can not only help explain the behavior of damping coefficients, but can also provide the basis for an analytical or numerical analysis. Furthermore, it may lead to the development of methods for the design of new devices that optimize damping. 1.4 Thesis outline The objectives of the present research © To understand the hydrodynamic force coefficients’ dependence on disk KC numbers, * To understand the hydrodynamic force coefficients’ dependence on disk thickness-to-diameter ratios, 13 * To understand the hydrodynamic force coefficients’ dependence on disk reinforcing structure and comer radius, * To understand the hydrodynamic force coefficients’ dependence on disk frequency, ‘* To understand the vortex formation modes’ dependence on disk KC numbers, ‘* To understand the vortex formation modes’ dependence on disk thickness-to- diameter ratios, * To understand the coherent structure, cycle-to-cycle variation, and turbulent Reynolds stress of the quasi-periodic unsteady flow around oscillating plates, and * To understand the generation, development, and dissipation of three- dimensional vortex rings associated with oscillating disks. In this chapter, Chapter 1, the background of the present research is first discussed. Then a review of the literature is given and the limitations of the existing work are discussed. Finally, the objectives of the present research and an outline for the thesis, are presented. In Chapter 2, the hydrodynamic force coefficients resulting from forced oscillation experiments are presented. The experimental models and experimental method are discussed. The quality of experimental data is verified by repeatability tests and comparison with existing published data. The results of damping and added mass coefficients are presented in groups based on the parameters being studied. The parametric/geometric dependences of the force coefficients, including KC number, thickness-to-diameter ratio, frequency, reinforcing structures, and comer radius, are 14 investigated. In Chapter 3, a qualitative flow visualization study on the vortex flows around oscillating circular plates is presented. The flow is assumed to be axisymmetric. The vortex formation modes as a function of KC number are investigated. Four vortex modes are identified and their relationship to damping coefficients is studied. The thickness-to- diameter ratio effect on vortex modes is also addressed. In addition, effects of initial startup conditions and oscillation frequency are included. Lastly, scaling laws for damping coefficients are derived based on the results from force measurements (Chapter 2) and flow visualization (Chapter 3). In Chapter 4, a quantitative analysis of the unidirectional vortex shedding mode is presented. The flow is identified as quasi-periodic unsteady flow. The coherent vortical structures and cycle-to-cycle variations are discussed. Different decomposition methods are investigated and turbulent Reynolds stresses are calculated. Chapter 5 summarizes the thesis and suggests future work. CHAPTER I FORCE MEASUREMENTS: DAMPING AND ADDED MASS COEFFICIENTS In this chapter, the in-line force coefficients (ie. added mass and damping) experienced by axially oscillating thin plates are studied. The KC number dependence, thickness-to-diameter dependence, frequency dependence, and reinforcing structure effects are studied through a series of forced oscillation tests. Flow visualization studies are discussed in Chapter 3. 2.1 Introduction The problem of interest can be modeled as a circular plate oscillating in its axial direction in a quiescent flow. Although flow around oscillating objects is not a new topic, most of the existing work is focused on cylinders due to their broad engineering applications. Relatively fewer studies have been conducted using circular plates. In a comprehensive survey, Dalzell (1978) reviewed earlier research on forces experienced by oscillating plates. He included analyses of two sets of experiments on circular plates with the following parameters: thickness-to-diameter ratio of 1/32 and 1/40, KC rage of 0.07~0.63, and f range of 6x10°~4.8x10" . These experiments were conducted using the decrement test method, which is questionable when the damping is highly KC dependent (Faltinsen 1990). 15 16 De Bemardinis (1981) calculated the unbounded oscillatory flow around a circular plate, He extended the modeling of 2-D free shear layers by applying a discrete vortices model to axisymmetric flows; his calculation assumes the flow is inviscid and the shed vortex sheets can be represented by sequences of discrete vortex rings. The flow around the disk edge is characterized by the formation of vortex pairs and their unidirectional convection from the plate. His calculation results for the drag coefficient, Cd, follow the KC~' behavior to a value of KC about 1.5; above this value, the results diverge rapidly from KC“ behavior. Lake et al. (2000) tested a circular plate with thickness-to-diameter ratio equaling 1/24, The experiment was conducted using the forced oscillation at resonance method initiated by Kim and Troesch (1991). An impressive damping increase was observed ‘when compared to axially oscillating cylinders with the same diameter as the thin disk. As is evident, the existing research regarding axial oscillating circular plates is limited and preliminary. Important parametric/geometric dependences, such as KC number and thickness-to-diameter ratio, need to be examined. Flow visualization studies around the disk edges are needed to help understand the flow physics and their contribution to the damping force. This chapter studies the in-line force coefficients through a series of forced oscillation tests. The flow visualization study is discussed in Chapter 3 and the quantitative flow physics study is discussed in Chapter 4. 2.2 Hydrodynamic force coefficients preliminaries The geometry of damping plates is idealized to a thin disk with disk diameter D and disk thickness . The amplitude of vertical oscillation is a, and the frequency (in 7 rad/s) is @. Cyclical frequency (in Hz) is given as: f = 2x. Two primary non- dimensional parameters, KC number and number, are defined by Eqns. (1.1) and (1.2) respectively. For an object oscillating in otherwise quiescent fluid or a fixed object in oscillating flows, the force acting on it can be decomposed into three parts: ‘© Inertia force due to the inertia of the accelerated/accelerating outer flow. © Friction force due to the influence of viscous boundary layers. © Form drag force due to the separation of boundary layers leading to the shedding of vortices. ‘The combination of the latter two parts contributes to the hydrodynamic damping force. The hydrodynamic damping force, Fyy, is typically represented in one of two forms: an equivalent linear representation as given in Eqn. (2.1): Fya(t)= BU() (2.1) or a quadratic representation (Morison’s equation): Fug) = proc veut) (2.2) Here the linear damping coefficient is B, the disk vertical velocity is U() = awcos(ax) (assuming displacement is {=asin(at)), and Cy is the drag coefficient. The two coefficients, B and Cy, can be related to each other by the equivalent linearization of the quadratic damping using Fourier decomposition (Sarpkaya and Isaacson, 1981). This relation can be written as: B HADC,KC (23) 18, Maull and Milliner (1978) have suggested that there is no basis for using the Morison’s equation at low KC, since a quasi-steady drag-generating wake does not occur. This implies that the limiting behavior of Cy as KC -» 0 is not well defined. Graham (1980) showed the different behaviors of drag coefficients at very low KC for different cross-sectional areas. The two extreme cases are the normal flat plates for which C, — and the infinite circular cylinder which C, > 0 as KC—+ 0. He concluded that the differences appear to be due to the different behavior and strength of the vortices shed by the various bodies. Estimates of drag coefficients for disks from Lake et al. (2000) support Graham’s claim by showing C, —>e at small KC number. In this research, experimental damping results are presented and discussed in the form of linear damping coefficients. The drag coefficient, Ca, is graphed for reference. 2.3 Description of experiment For lightly damped systems, the inertia force dominates the total force experienced by the system, thus making it difficult to identify the damping force in the total force measurement. In this section, a description of the methods for measuring the hydrodynamic damping force of lightly damped systems will be given and a mathematical model for the experimental method will be descrited. Based on this model, the calculation equations of added mass and damping coefficients are derived. 2.3.1 Measurement techniques, mathematical model, and data reduction To measure the hydrodynamic damping forces, model tests can be conducted by either oscillating the flow over a stationary model or by oscillating a model in a still fluid. 19 For the first case, the inertia force is increased by an equivalent Froude-Krylov force, a buoyancy force caused by the pressure gradient imposed on the fluid to generate the oscillating flow (Bearman et al. 1985). Uniform fluid oscillation can be produced by the means of a U-tube as described by Sarpkaya (1975). He used the device to conduct an extensive study of circular cylinders under oscillating flows. A second possibility is to oscillate the model in a still fluid. Two commonly used techniques are forced oscillation and decrement tests: Forced oscillation: The system is forced to oscillate at a certain frequency, and the damping is estimated from results using Fourier analysis. © Decrement tests: The system is given initial conditions and is allowed to oscillate freely, and the equivalent linear damping at the system natural frequency is estimated from decrement curves. Difficulties exist with both methods when the system is lightly damped, which is common in offshore systems. The decrement test is easier to setup, but its application to the case presented here is questionable. Faltinsen (1990) pointed out that if the drag coefficient has a large KC number dependence, the decrement test cannot predict the damping correctly. More than 10 oscillation periods with no significant variation in KC number may be needed to give a good estimate of the damping. In the case presented here, the damping coefficient is highly KC dependent. As will be shown later, flow visualization results illustrate different vortex generation pattems with variation of the KC number. Consequently, decrement tests are questionable for this particular application. 20 In forced oscillation tests, the model is oscillated at a predefined frequency and amplitude. The damping and inertia force can be calculated from the results via Fourier analysis. By varying the oscillation amplitude, the KC number dependence can be studied. By varying the frequency of oscillation, the f parameter dependence is determined. In this thesis, the forced oscillation method is used. Figure 2.1 is a schematic of the forced oscillation test where: F,, = The total vertical force measured by the load cells. F, = The total vertical hydrodynamic force acting on the system. ¢ = The displacement of vertical oscillation. If we define the mass of system (excluding hydrodynamic inertia) as my, and oscillation frequency as @ , the equation of motion in the vertical direction can be written Fig ty = tg 4) Fe Total vertical force t | . F,, Hydrodynamic force Figure 2.1 Schematic of forced vertical oscillation method. The excitation motion of the model is periodic and frequency analysis can be done on the periodic output. In the frequency domain, a periodic signal can be separated into its harmonic components, and conversely, it may be synthesized by summing its 21 components (Newland, 1987). Under periodicity assumptions, the analysis of the equation of motion can be accomplished one frequency at a time. As a result, the ‘components of Eqn. (2.4) may be written in the complex form of Eqn. (2.5): =nefSaerrend F,= ad 5 fuerten (2.5) Sacre} F= where: ‘Su i8 the magnitude of the force of the jth component of F,, Jy is the magnitude of the hydrodynamic force of the jth component of F,. The a’s, w’s, and c’s are real-valued amplitudes, frequencies, and phase shifts, respectively. Each harmonic term of Eqn. (2.4) can be written as Eqn. (2.6) by using complex notation, canceling the time dependent term, and solving for the desired hydrodynamic force term: Sye'™ =-fge™ -m,,a,0}e (2.6) The hydrodynamic force is assumed to have three significant components: a velocity dependent term (damping), an inertia term (added mass), and a displacement dependent term (hydrostatic stiffiness). Consequently, the hydrodynamic force can be modeled by the following equati fye"? = Aya,0 ~iB a0, Ca, en : Fourier coefficient of hydrodynamic force in phase with acceleration (added 2 mass coefficient), Bj: Fourier coefficient of hydrodynamic force in phase with velocity (damping coefficient), and Cj: Fourier coefficient of hydrodynamic force in phase with displacement (stiffness coefficient). ‘The stiffness coefficient, C; , is related to the buoyancy force that results as the disk shaft is displaced in the water by the motion of the shaker. Specifically: 2.8) where: : density of water, g: gravitational acceleration, Avp’ area of the shaft water-plane, and 1: shaft radius. By combining the real and imaginary parts of Eqn. (2.7), the added mass and damping coefficients may be written as: -1 C, = Fig, 11 OM Aig 2) May oF 9) (2.10) For harmonic motions, the primary force component occurs at the fundamental excitation frequency, so that in the above equation j= | is a close approximation of the force record. The effects of the higher harmonics are typically less than 5% of the fundamental harmonic in the case presented here and therefore are relatively small, 23 contributing primarily to higher order forces. After discarding the higher order forces, added mass and damping equations can be rewritten as Eqns. (2.11, 2.12): 11) (2.12) The hydrodynamic coefficients in heave, added mass 4 and damping B, are non- dimensionalized as 4’ and B’ , respectively, where (2.13) (2.14) 2m'o Here, m’= ; pD? is the theoretical ideal fluid added mass for a flat disk of diameter D. The mass m’ is selected as the nondimensionalization factor since a flat disk has little displaced mass. Substituting Eqn.(2.14) into (2.3), Morison’s drag coefficient, C,, can be expressed in terms of B’ as: C, = 4n8'/KC (2.15) For a lightly damped system, such as the TLP column in heave, he and the phase difference between the force and the displacement measurement 4g =, —a is usually within the 175 to 180 degree range. From Eqn. (2.12), the damping coefficient is calculated based on sin(@yy), and therefore a small error in obtaining ay will result in a significant error in damping force results. During the 24 experiment, phase shift errors can exist due to mechanical load cells, electronic filters, and data acquisition boards. To overcome this difficulty, Troesch and Kim (1991) introduced springs to the mechanical system and oscillated cylinders at the resonance frequency of a new spring-mass-damper system. By doing so, the inertial forces are nearly balanced by the restoring forces, and thus damping becomes a large part of the net force measured. This method proved to be very effective in measuring damping force in lightly damped systems, such as circular cylinders, at small KC numbers. The above technique increases the precision of damping measurement, but also complicates the experiment by introducing extra spring force measurements. In addition, the testing frequency is limited to the resonance frequency of the system. Different springs are required to test different frequencies, which will increase the complexity of experiment and increase the chance for error. Considering the fact that the damping force of oscillating disks is much larger than cylinders (Lake et al. 2000), a simplified technique (without springs) is used in the current experiment. The model is forced to oscillate without springs and the phase shift due to the measuring equipment is carefully corrected for each testing frequency. 2.3.2 Experimental setup The experiments were conducted in two phases at the MHL (Marine Hydrodynamic Laboratory) and the Small Scale Fluid Dynamics Laboratory located at the University of Michigan. Two experimental setups were used to study different KC ranges. In both setups, a vertically mounted shaker capable of producing periodic sinusoidal motion was used to force the oscillation. The shaker is a Unholtz-Dickie (UD) 25 Model 20 Shaker with a UD Model TALOOA Solid State Power Amplifier, capable of oscillating the disk in the vertical direction at a frequency to 5.0kHz with a maximum one inch peak-to-peak oscillation amplitude and 250 Ibs. peak-to-peak force. The experiments were conducted in a stationary tank sufficiently large so as to avoid the hydrodynamic interaction between the model and the tank walls. A detailed description of the tank is listed in Table 2.1. In the first phase, the model was mounted to the shaker table directly. The schematic of the experimental setup is shown in Figure 2.2. Two load cells, a linear variable displacement transducer (LVDT) and a piezoelectric accelerometer, were used to ‘measure the heave response force, the vertical displacement, and the vertical acceleration, respectively. The accelerometer was used to verify the output from the LVDT. The outputs from these instruments were amplified and subsequently filtered using low-pass Butterworth filters, with a cutoff frequency of 40 Hz. The filtered outputs were sampled and acquired by a National Instrument DAQ board with a sample rate of 1024Hz. Sixty- four cycles of data were recorded for a typical run. This data was then Fourier analyzed to identify the magnitude and phase of the force and displacement. The phase shift due to the filters and the DAQ board was calibrated before each experimental setup and a correction was made for each experimental run. A typical phase correction was 22 degrees at 4 Hz. A Mac IIFX was used to do the data acquisition and analysis. Table 2.1 Experimental parameters. ‘Surface Area of Tank ‘Water Depth Depth of Disk Oscillation Frequency 1.73mx0.97m (5.67RX3.1 78) 1.2m (4.08) 36.25em (14.27in) 2,4, 8 Hz 26 Shaker Shaker Motor LVDT Sleeve Accelerometer. LvpT po Armature Load Cell #1 Load Celt #2 Water Surface Disk 7 Figure 2.2 Schematic of experimental setup one. In the second phase, two experimental configurations were used to cover different KC ranges. For small KC number ranges (0~0.45), the same setup as Phase One was used and repeatability was validated. Since the shaker is capable of a maximum KC of only 0.45, a MAM (Mechanical Amplifier Mechanism) was built to amplify the motion of the shaker by a factor of four. Due to the restriction of shaker loading capability, the maximum KC number tested is 1.1. A displacement transducer utilizing a rotary potentiometer, which is capable of +2 inches, was used to measure the displacement and a LVDT was used to calibrate the transducer. A Dell PC was used for data acquisition and analysis. The schematic of the setup is shown in Figure 2.3. The added mass and 27 damping coefficients were calculated based on the phase-corrected data using Eqns. (2 11, 2-12). ‘with sliding slots ‘Displacement transducer Water Surface Dak Figure 2.3 Schematic of experimental setup two (MAM). ‘The resolution of equipments is listed in Table 2.2. Table 2.2 Equipment resolution. Equipment Load cells LVDT LVDT phase shift Phase shift (filter-DAQ) Resolution 0.05 Ib 0002 inch 0.044? /4Fz 0.02° /4Hz Accelerometer 0.01g's 28 2.3.3 Experimental models Both a single-disk and a two-disk configuration were used. The goal of the two- disk configuration was to examine the role that reinforcing structures play in damping reduction. Two disks were stacked atop of one another, each with a different diameter. This has the effect of eliminating the need for stiffness by using a solid surface. The larger bottom disk is thin while the smaller diameter top disk is significantly thicker, representing the addition of the vertical stiffeners. Figure 2.4 shows the two-disk configuration. The bottom disk is made of aluminum while the top one is made of PVC. VE Be Figure 2.4 Two-disk configuration. ‘The parameters of the disk models are summarized in Table 2.3. Table 2.3(a) Disk models (single-disk models). Model Diameter (inch/mm) | Thickness (inch/mm) | Thickness/Diameter Model 1 TITS 0.0471.01 175 Model 2 TI78 0.08 72.03 1875 Model 3 TTS 0.189748 137 Model 4 THATS 0.278 77.09 1725 Table 2.3(b) Disk models (two-disk models). Model Da2/Ddl Thickness t1+t2 (inch/mm) Model 5 0.95 0.2097 5.31 ‘Model 6 TTS 09 0.20975.31 ‘Model 7 TATE 0.85 0.20975.31 29 During the experiment, the single disk models were tested at 2, 4, 8 Hz, while the two-disk models were tested at 4 Hz. 2.4 Results and discussions To show the quality of the data, Figures 2.5 and 2.6 illustrate the measurement results from LVDT (displacement) and load cells at two different KC numbers. The illustrated measurement is from the single disk with thickness-to-diameter of 1/87.5 and oscillation frequency of 4 Hz (f=1.24x10*). The displacement is found to be sinusoidal; higher order harmonics are very small. On the other hand, the total force shows nonlinear behavior, especially at large KC numbers. The higher harmonics in force time history could be due to the higher order harmonics of displacement or the nonlinear effect of lower harmonics. The corresponding amplitude spectra of the total force and displacement for KC of 0.375 are shown in Figures 2.7 and 2.8. A calculation of the force due to the higher harmonic oscillation (F;) was conducted and the results are shown in Table 2.4. F, is calculated using Eqn. (2.16): F, a,(m,, + 4,) (2.16) a, is the oscillation amplitude of i-th harmonic; 4, is the added mass of i-th harmonic. F, is 98% of the second harmonic of the total force and 99% of the third harmonic. It can be seen that the higher harmonics in total force time history are mainly due to the higher harmonics of shaker oscillation. 30 The test matrix is divided into four primary categories: KC number dependence, thickness-to-diameter dependence, # number dependence and reinforcing structure effect. Also presented are studies of the effect of disk deflection (i.e. hydroelasticity) and edge sharpness (i.e. comer radius). The test parameters of each group are shown in Table 2.5. The complete test results representing the conditions in Table 2.5 are presented as non-dimensional added mass, Eqn. (2.13), damping coefficients, Eqn. (2.14), and Morison’s drag coefficient, Eqns. (2.3) and (2.15), as a function of KC. 31 Load Cell #1 Figure 2.5 Time histories of the force and displacement measurements for the 1/87.5 disk (KC=0.185, f4Hz, 8=1.24x10*). 32, Load Celt #1 Figure 2.6 Time histories of the force and displacement measurements for the 1/87.5 disk (KC=0.375, &4Hz, 8=1.24x10°). B10 — ge _| Bo | f 2 Lo 4 0 4 8 12 16 2 2% 2 32 26 40 frequency (He) Figure 2.7 Amplitude spectrum of force. Dlepiacement amplitude spectrum 04s 04 - F008 | £03 | 025 oz} — os | —] ox 005 ° ——> 4 8 12 16 20 2% 2% 22 36 40 frequency (H2) Figure 2.8 Amplitude spectrum of displacement. Table 2.4 Higher harmonic components Force Displacement Harmonics | “mag (Ibs) | phase(rad.) | mag. (in.) | phase(rad.)| F;(Ibs) | Percentage 1 @Hz) [11.62 4669 0.422 1.653 10.88 | 93.6% 2™(8Hz) | 0.323 3.458_| 0.00314 | 1.166 0318 98% 37(12Hz) | 1.535 2.963 | 0.00698 | 0.0309 1.52 99% 34 Table 2.5 Parameters of the six sub-division categories. Thickness-to-diameter | 1/87.5 Frequency Hz Results Figure 2.13 Thickness-to-diameter 1/87.5, 137, V/25 Frequency 4Hz Results Figure 2.14 Thickness-to-diameter | 1/87.5 Frequency 2,4, 8 Hz Results Figure 2.15 0.85, 0.9, 0.95, 1.0 ‘Thickness-to-diameter Frequency Thickness-to-diameter Figure 2.17, 2.18 Frequency Results 35 2.4.1 The quality of experimental data ‘An experimental error analysis was conducted based on the 1/87.5 disk by considering the largest uncertainty of equipments. The low and upper bounds of the error are calculated using Eqns. (2.17) and (2.18). The errors in percentage are graphed in Figure (2.9). As can be seen, the error decreases as the KC number increases. For KC>0.1, the measurement error in damping coefficient is less than 3%. Bane =—— Senin ity - 2g 17 (2.18) 2.19) Figure 2.9 Estimated maximum error in damping coefficient measurement (percentage). To show the quality of measured damping and added mass coefficients, a comparison with the results from the spring-based system (Lake et al. 2000) is presented. 36 Furthermore, the repeatability of the experiment is demonstrated through two experiments conducted at different locations and times. Lake et al. (2000) did an extended study on the effect of attaching a damping plate to the bottom of the cylinder, where they examined cylinder-disk combinations and then disks in isolation. In that study, springs were introduced to minimize the measurement error. The test was conducted at the resonant frequency of the system. Disk parameters and test conditions are shown in Table 2.6. Table 2.6 Test conditions for Lake et al. (2000) experiment. meter of disk 7.34 inch / 1864 mm Thickness of disk 0.305 inch /7.74 mm “Thickness-to-diameter ratio _| 1/24 “Testing frequency ——=—=—=«d; SL iz ‘KC range 0. B 1.28x10* 0 Figure 2.10 shows the comparison between current results (disk YD=1/25) and Lake et al.’s (disk VD=1/24). Lake et al.’s test was performed at resonance frequency 3.91 Hz, From the damping curve it can be seen that the 1/24 disk is a little below the 1/25 disk. From later thickness-to-diameter ratio dependence results, it is shown that the thinner disk generates more damping at the same KC number. Overall, both the added mass and damping curves agree very well. While these two results are from two different testing methods, the comparison demonstrates that the direct forced oscillation method ccan generate high quality data. 37 (© VD=1726 Lake ot x 3 i wis Bos —— (2000) 2 0.00 0.10 020 030 040 KC Number 0.20 7 por = 8 os tt 3 (lovers isto © 20} 21/2000) 3 goss ——" 010 : 2 F 0s 3 00 0.10 020 030 040 Figure 2.10 Added mass and damping coefficient comparison (fe4Hz and 6 =1.24x10" for 1/25 disk, f3.91Hz and $=1.28%10? for 1/24 disk). ‘As mentioned before, the experiment was conducted in two phases. Phase One covers KC numbers from 0 to 0.45, while Phase Two extends the KC to 1.1. The two phases were conducted at different times (one-year difference) and locations. To show the quality of data, repeatability tests were conducted. Figure 2.11 shows the repeatability of the 1/87.5 disk at 4 Hz. As shown, the difference in the damping curves is within 3%, while the added mass results, showing slightly larger differences, are within 5%. 38 «< 2 i os = Phase 2, Jos Ne j a 5 rs Kc Number | i Be: go. i 7 oo 01 ~*~CSSSC Ke Number Figure 2.11 Repeatability test results taken one year apart (VD=1/87.5, f4Hz, B=1.24x10*). The original shaker is capable of generating KC numbers from 0 to 0.45 for the current model size. By using the MAM, the maximum XC is extended to 1.1. The repeatability and quality of the MAM data is shown in Figure 2.12, where the added mass and damping curves for the two different experimental setups are shown. The test was conducted at #=1.24x10°. The “Small KC” result is based on the original shaker while the “Large KC” curve is from the MAM. The difference is within 3% in both damping coefficient and added mass indicating the reliability of the experimental method used. ‘Added Mass Ratio A" a5 fe SretKC | - [= Lerge Kc| os 5 3 - Zoe we Bos es os f : B00 Figure 2.12 Quality of the MAM system. The above comparison and repeatability tests show the quality of the experimental data and demonstrate that the current direct forced oscillating method is capable of generating high quality damping coefficients data. 40 2.4.2 Effect of KC numbers The KC number is the amplitude of fluid motion, or amplitude of disk motion in the present case, relative to the diameter of the disk. The magnitude of the KC number indicates the relative importance of drag and inertia forces. At low KC numbers, the inertia force dominates the flow. With the increase of KC, the drag force due to flow separation becomes more important. With increasing KC numbers, the flow around blunt-shaped structures with no sharp comers, such as cylinders, experience no separation, separation with no vortex shedding, and separation with vortex shedding respectively (Faltinsen 1990, Bearman et al, 1981). At very low KC numbers, no separation occurs in oscillatory fluid motion. As a result, the drag force is mainly from friction drag due to the viscous shear stress on the surface. As the KC number increases, flow separation begins to dominate and as a result, form drag dominates the drag force. However, the flow separation does not always indicate vortices or vortex shedding (Faltinsen 1990). Bearman et al. (1981) did a study on the flow around a circular cylinder at KC= 4. A pair of vortices was noticed during one half cycle. During the next half, the flow reversed and the vortices were convected ‘back to the cylinder, their circulation appeared to be cancelled by the vorticity of opposite sign and no vortex shedding occurred. Vortex shedding occurs as the KC number is increased further (Williamson 1985). Unlike blunt-shaped structures, for bodies with sharp edges, e.g. the current damping plate case, flow separation will always be expected to occur at the edges. Like blunt-shaped structures, flow separation does not always lead to vortices or vortex

You might also like