Heave motion suppression of a Spar with a heave

plate

Longbin Tao1 ∗ and Shunqing Cai2

a School of Engineering, Griffith University, PMB50 Gold Coast Mail Centre, Qld 9726,

Australia

b School of Civil and Resource Engineering, The University of Western Australia, 35

Stirling Hwy, Crawley, WA6009, Australia

Abstract

Vortex shedding flow of an oscillating vertical cylinder with a disk attached at its keel is

considered. This configuration is of interest for the offshore oil and gas industry. A finite

difference method is employed to solve the incompressible Navier-Stokes equations in the

primitive variables formulation. Test cases were used to guide selection of the size of flow

domain, numerical parameters and verify that the resultant method was both convergent

and accurate. Numerical simulations have shown that the geometry configurations of the

cylinder and disk, such as aspect ratio of the disk td /Dd and diameter ratio, Dd /Dc have

significant influence on the vortex shedding modes and associated hydrodynamic proper-

ties, e.g. hydrodynamic damping and added mass coefficients. Which in turn affect the

performance in heave motion control of the structures.

Key words: finite difference method, vortex shedding, heave damping

∗ correspondence to Longbin Tao, Fax: 61 − 7 − 55528065, E-mail:

[email protected]

Preprint submitted to Elsevier Science 6 January 2003

1 Introduction

Newly developed floating structure systems for offshore oil and gas industry op-

erating in deep water, e.g. Tension Leg Platform (TLP) and Spar platform, have

distinct advantage in relatively small magnitude of motion in vertical planes. This

resulted in the wide application of the rigid risers. However, resonant heave oscil-

lations may still occur due to nonlinear wave effects (sum-frequency) on TLPs and

linear wave excitations of wave spectrum with long wave periods on Spars. These

heave resonances are often excessive to cause damage in both risers and mooring

systems.

In order to enhance the damping mechanism during the resonant oscillation of a

TLP or Spar in the vertical direction, additional mechanical damping devices or

other active damping systems are introduced externally. A typical example of these

damping devices is heave plate which is attached to the keel of a TLP column

or Spar (Fig. 1). The additional disk at the bottom of the column enhances the

vortex shedding process, consequently changes the hydrodynamic properties by

introducing extra damping and increasing the added mass of the system.

The two non-dimensional characteristic parameters representing the amplitude and

frequency of the oscillation for cylinder and disk configuration are the Keulegan-

Carpenter number (KC) and the frequency parameter (β). They are defined as:

2πa
KC = (1)
Dd
(Dd )2 f
β= (2)
ν

where Dd is the diameter of the disk, a and f are the amplitude and frequency of

the oscillation respectively and ν is the kinematic viscosity of the fluid. One can

2
obtain a Reynolds number based on the flow oscillation velocity as:

aω(Dd )
Re = (KC)β = (3)
ν

Thiagarajan & Troesch (1998) observed the flow of cylinder+disk configuration

using the Particle Image Velocimetry (PIV) technique. The vortex shedding pattern

was found to be dependent on both flow parameters (KC, β) and the geometry of

the structure. For a disk with two edges oscillating at small amplitude, the flow

was found to be symmetric about the mean position of oscillation. 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 due to low KC, until flow reversal

causes a rapid cancellation of vorticity. The measurements showed that the disk was

found to increase the pressure drag coefficients (Cd ) two-fold. In the case that the

thickness of the disk is very small, the two edges of the disk become virtually one

single sharp edge, a well-established and reasonably stable vortex shedding pattern

was generated by the single sharp edge. Vortex shedding occurred at a large angle

of either positive or negative direction depending on starting condition. Thiagarajan

(1993) noticed a difference in strengths of two consecutive vortices forming a pair,

and interpreted that as due to a longer time period available for one of the vortices

to develop depending on the starting direction of the oscillation.

Investigation of Tao (2002) showed that the three distinct vortex shedding modes,

namely independent, interactive and uni-directional vortex shedding, occur due to

different KC and the thickness of the disk. Diffusion or convection was found to

be predominant in each of the different vortex shedding regimes respectively. Con-

sequently distinct hydrodynamic heave damping behaviours were found associated

with different vortex shedding regimes, i.e. a lower, weakly KC-dependent damp-

ing was found in the independent vortex shedding regime, while the damping was

3
found to be nonlinearly dependent on KC in the later two vortex shedding regimes

resulted from strong convection effect.

This paper is to conduct an investigation of a systematic variation of disk geome-

try, especially the influence of the diameter ratio, Dd /Dc , on the vortex shedding

pattern and associated hydrodynamic heave damping arising from the flow sepa-

ration and vortex shedding around the cylinder+disk. A finite difference solution

of the viscous fluid structure interaction problem is obtained by solving Navier-

Stokes equations for incompressible viscous flow. The present numerical method is

validated against several benchmark problems and available experimental results.

Recommendations on the design of heave plate for floating offshore structures are

made based on the numerical calculation.

2 Theoretical formulation

2.1 Governing equations

The flow induced by a vertical cylinder with a disk attached at its keel oscillating at

low KC along its axis can be idealized as axi-symmetric by neglecting some three-

dimensional effects (see Fig. 1). The governing equations for the time dependent

unsteady flow are the continuity and Navier-Stokes equations for incompressible

viscous fluid. The axisymmetric coordinate system shown in Fig. 1 is fixed to the

bottom open boundary with positive axes pointing right and upwards. The non-

dimensional form of the governing equations in cylindrical polar coordinate system

(r, y) are written as:

1 ∂(ru) ∂v
+ = 0, (4)
r ∂r ∂y

4
 ∂u ∂p 1 u
+ (V · ∇)u = − + 0 (∇2 u − 2 ), (5)
∂t ∂r Re r

∂v ∂p 1
+ (V · ∇)v = − + 0 (∇2 v), (6)
∂t ∂y Re

here V = (u, v) is the velocity vector, (u, v) denote radial and axial components of

velocity respectively, t time, and p the dynamic pressure. In this paper, the govern-

ing equations and all physical quantities are presented in a non-dimensional form.
√
Re0 = (R Rg)/ν in Eqs. (5) and (6) arises due to the fundamental variables used

in non-dimensionalization, i.e. density of fluid %, acceleration of gravity g, and ra-

dius of the cylinder R. ∇ and ∇2 are the gradient operator and Laplacian operator

respectively.

The Poisson equation for pressure can be derived by taking divergence of the mo-

mentum Eqs. (5) and (6),

∂ D̃ ∂u ∂v ∂v ∂u u2
∇2 p = − − [( )2 + ( )2 + 2 ]− 2 (7)
∂t ∂r ∂y ∂r ∂y r

where D̃ is divergence, for axisymmetric flow

1 ∂(ru) ∂v
D̃ ≡ ∇ · V ≡ + (8)
r ∂r ∂y

2.2 Boundary conditions

Since the cylinder with a disk is forced to oscillate sinusoidally along its longitudi-

nal axis (see Fig. 2) as

y(t) = y(0) + a sin(ωt + θ) for t > 0, (9)

5
where a, ω are the amplitude and circular frequency of the oscillation respectively.

No-flux and no-slip velocity boundary conditions are imposed along the oscillatory

cylinder and disk surfaces:

u = 0 and v = ẏ(t) = aω cos(ωt), (10)

Symmetric boundary condition is applied to the axis of symmetry based on the

axisymmetric flow assumption.

According to Thiagarajan & Troesch (1998) based on their experimental observa-

tion, disturbance of the free surface is very small due to small amplitude and high

frequency heave oscillation of the floating structures. Therefore, non-linear free

surface effects are not included in the present numerical modelling, instead a free

stream flow boundary condition is assumed, and pressure on the free surface is a

constant. Free stream velocity conditions are also applied to the entire open bound-

ary, and static pressure is assumed at far open boundaries. The Neumann boundary

condition for the pressure on the cylinder and disk surfaces is obtained by applying

the momentum equations in the direction normal to the body boundary.

2.3 Hydrodynamic force coefficients

The hydrodynamic heave force acting on the cylinder+disk is calculated by in-

tegrating the vertical component of the stress vector along the cylinder and disk

surfaces.

Z ( )
1 ∂u ∂v 2 ∂v
F =− −pny + ( + )nr + ny ds, (11)
Re ∂y ∂r Re ∂y
B

6
where ds denotes the differential cylinder+disk surface area, p the dynamic pres-

sure, and (ny , nr ) the axial and radial components of the unit-normal vectors along

the cylinder+disk surfaces. The three terms of the right hand side of the above ex-

pression represent the dynamic pressure component, the viscous shear-stress com-

ponent and the normal viscous-stress component respectively. The viscous drag

force on the body is first represented by an expression similar to the formulae of

Morison’s equation in which the drag is proportional to the square of the velocity.

Since the cylinder and disk is subjected to a sinusoidal oscillation as shown in Eq.

(9), the force can be linearised in a Fourier averaged sense,

∂v 4
F (t) = −Ca ρ∀ − Cd ρSVm2 cos(ωt) (12)
∂t 3π

where Vm = aω is the amplitude of the velocity of the oscillation, and T is the

period of the oscillation. S is the cross-section area of the disk, and ∀ is the im-

mersed volume of the cylinder and disk. The drag and added mass coefficients are

then obtained from a force time history by Fourier analysis, which gives:

T
3ω Z
Cd = − F (t) cos(ωt)dt, (13)
4%SVm2
0

T
1 Z
Ca = F (t) sin(ωt)dt, (14)
π%∀Vm
0

The equivalence between the Morison’s equation-like formulae and the alternative

expression (Fd (t) = Bv(t)) of the drag force by using the equivalent linear damp-

ing coefficient B results in

1
B = µβDd (KC)Cd , (15)
3

7
where µ is the dynamic viscosity. In the analysis of the results, the equivalent linear

damping coefficient B is then normalised to give a non-dimensional damping ratio

B
Z= , (16)
Cc

where Cc = 2(m + ma )ω is the critical damping of the system, m and ma represent

the mass and added mass respectively. By inserting expressions for the mass m +

ma = Cm %∀ and B, one can obtain the damping ratio

1 Dd2 Dc Cd
Z= 2 2 2
(KC) (17)
3π (Dc Td + Dd td ) Cm

where Dc is the diameter of the cylinder while Td and td represent the draft of

cylinder and the thickness of the disk respectively.

Viscous shear stress acting on the cylinder+disk side wall in phase with velocity

is the main source of friction damping force, form damping component is mainly

attributed to the flow separation and vortex shedding at the sharp edges of the disk.

Substituting the viscous shear stress term and the dynamic pressure term of Eq.

(11) instead of total force F (t) into Eq. (13), one can derive the two components of

damping coefficient and damping ratio respectively.

3 Numerical implementation

The initial boundary value problem outlined above has fully non-linear field equa-

tions with boundary conditions. The theoretical formulation is implemented by us-

ing finite difference method based on curvilinear coordinates. The method is to

solve the non-linear governing equations together with the given boundary condi-

tions in primitive variable form. A non-staggered grid for the velocity and pressure

8
is used in the present work.

3.1 Spatial scheme

The discretisation of the convection terms of the momentum equations is of pri-

mary concern due to its non-linear nature. This is particularly important for high

Reynolds number flows since the convection is expected to be the dominant fea-

ture. In the present study, a modified second-order four-point upwind scheme (K-K

scheme) proposed by Kawamura & Kuwahara (1984) is used to discretise the non-

linear convection terms of the momentum equations,

∂u fi,j
(f )i,j = (−ui+2,j + 8ui+1,j − 8ui−1,j + ui−2,j )
∂ξ 124ξ
|fi,j |
+ (ui+2,j − 4ui+1,j + 6ui,j − 4ui−1,j + ui−2,j ). (18)
44ξ

The second-order central differencing is used for all the other terms.

∂f fi+1,j − fi−1,j
( )i,j = , (19)
∂ξ 2

∂ 2f
( )i,j = fi+1,j + fi−1,j − 2fi,j . (20)
∂ξ 2

All the spatial terms in the transformed pressure Poisson equation are discretised

by using second-order central differencing except those on the boundaries. The

spatial derivatives in the boundaries are discretised by a second-order one-sided

differencing, i.e. for left boundary, we have,

∂f −3fi,j + 4fi+1,j − fi+2,j

( )i,j = (21)
∂ξ 2

9
 ∂ 2f
( )i,j = fi+2,j − 2fi+1,j + fi,j . (22)
∂ξ 2

Similar spatical derivatives were derived for all the other boundaries.

3.2 Time integration

The momentum Eqs (5) and (6) are marched forward in time by using a second-

order fully implicit scheme (Beaudan, 1994):

3Vn+1 − 4Vn + Vn−1 1

+ Vn+1 · 5Vn+1 = − 5 pn + 0 52 Vn+1 , (23)
2∆t Re

The primitive variables of velocity and pressure are obtained by solving momentum

equations and pressure Poisson equation alternatively using an iterative method.

The discrete form of nonlinear Eq. (23) is first solved by using the Gauss-Seidel

iterative method with pressure at time level (n). A solution for pressure at time

level (n + 1) is then obtained by solving the discrete form of pressure Poisson

equation using Successive Over-Relaxation (SOR) scheme. The correction term

( ∂∂tD̃ ) in the pressure Poisson equation is discretised with the second-order time-

accurate scheme.

The governing equations together with the given boundary conditions are solved

as an initial boundary value problem. It is assumed that the flow induced by the

heaving cylinder starts from a quiescent state, i.e. velocity is zero at time t = 0.

10
3.3 Mesh generation and numerical parameters

Due to the axisymmetric flow assumption, meshes for the simulation of the flow

induced by the oscillating cylinder and disk are generated in a rectangular region

simply by an algebraic method. Although interference of the shear layer developed

along cylinder surface to the vortex shedding is expected as amplitude of the oscil-

lation increases, since the amplitudes of oscillations are relatively small, the vortex

patterns around the cylinder+disk mainly depend on the local flow around the sharp

edges of the disk. Based on these considerations, a mesh is then constructed in such

a way that the node points are concentrated near the cylinder and disk surfaces and

stretched out gradually. As can be seen from Fig. 3, the grid is fully orthogonal, and

of high quality from the mesh generation point of view. The small regions around

the sharp edges of the disk, where the vortices are generated and shed during the os-

cillation, are emphasised with clustering the grid points to the regions. Such a grid

distribution ensures both high quality computation in the vicinity of sharp edges

and global efficiency.

The size of flow domain and the numerical parameters ∆r, ∆y, ∆t as well as the

number of nodes necessary for the simulation of the flow generated by the oscillat-

ing cylinder and disk, are evaluated in order to ensure the stability of the compu-

tations and to minimize the numerical error for each value of the flow parameters.

Based on the extensive numerical tests of (Tao et al., 2000; Tao, 2002), the radial

and bottom open boundaries of the flow domain are located at 10R from the cylin-

der wall and bottom surface respectively. A mesh of (134 × 120) with a minimum

grid spacing of 0.0005R around the sharp edges of the disk in both radial and verti-

cal directions are used in the numerical simulations. All the calculations are carried

out with a time step ∆t/T = 1/5000. The mesh is regenerated at each time step

11
according to the forced oscillation of cylinder+disk.

4 Validation of numerical method

The first principal step taken to validate the present numerical procedure is the

convergence tests with varying flow domains and mesh parameters (Tao, 2002).

In this section, the present numerical procedure is further validated against two-

dimensional benchmark problem: lid-driven flow in a square cavity, and compar-

isons with the experimental results of Thiagarajan & Troesch (1994) for the flow of

an oscillating truncated vertical cylinder.

4.1 Lid-driven flow in a square cavity

The lid-driven flow in a square cavity is a typical steady separated flow which has

been examined experimentally and numerically in the past (Burggraf, 1966; Boze-

man & Dalton, 1973; Schreiber & Keller, 1983)). Accurate numerical results are

available for this case over a certain range of Reynolds number which is defined in

terms of the driven velocity at the top and the width of the cavity. Therefore, it has

been frequently used to test the numerical solutions of the Navier-Stokes equations.

The problem definition and the flow pattern in the cavity are illustrated schemat-

ically in Fig. 4. The flow is characterised by a strong primary vortex generated

by the top boundary (lid) in the upper portion of the cavity. In addition, there is

a pair of small counter-rotating secondary vortices of very low strength located in

the lower corners of the cavity. The location of the primary vortex and the velocity

distributions in the cavity depend on Reynolds number.

The comparison of the centre locations (Xc , Yc ) of the primary vortex obtained

12
Table 1

Locations of the primary vortex centre

Re Schreiber & Keller (1983 Present

Mesh size Xc Yc Mesh size Xc Yc

40 121 x 121 0.56667 0.75833 81 x 81 0.55898 0.74900

100 121 x 121 0.61667 0.74167 81 x 81 0.62393 0.74413

400 141 x 141 0.55714 0.60714 81 x 81 0.55478 0.60265

1000 141 x 141 0.52857 0.56429 81 x 81 0.53307 0.56744

from the present numerical calculation and the results of Schreiber & Keller (1983)

in Table 1 shows very good agreement with maximum difference of 1.36%. Fig.

5 presents the distributions of the horizontal velocity component along the vertical

lines at x = 0.1, 0.5 and 0.9, obtained from the present numerical solution and

Schreiber & Keller (1983) at Reynolds number Re = 1000. As can be seen in Fig.

5, although a relatively coarse mesh is applied in the present calculations, excellent

agreement has been achieved.

4.2 Flow generated by an oscillating truncated vertical cylinder

Several researchers investigated the flow generated by a truncated vertical cylin-

der oscillating at low Keulegan-Carpenter numbers, corresponding to the hydrody-

namic phenomenon ‘springing’ of deep water offshore structure TLPs (Chakrabarti,

1991; Huse, 1990; Huse & Utnes, 1994; Thiagarajan & Troesch, 1994). Experi-

mental data available from Thiagarajan & Troesch (1994) is the main reference

for validating the present numerical investigation. Forced oscillation experiments

13
were performed with a circular cylinder of 0.457 m diameter and 1.219 m draft.

The cylinder was oscillated corresponding to KC from 0.1 to 1.0. The frequency

parameter β is 89236, corresponding to the natural frequency of the cylinder of

0.41 Hz. Since the total force is inertia dominated, Thiagarajan & Troesch (1994)

conducted damping measurements on cylinders in forced oscillation at or near res-

onance, in witch the restoring forces nearly offset the system inertia forces. The

resulting force measurements give a good estimate of the drag coefficients.

A rectangular flow domain of 11R × 15R with a mesh 88 × 93 is used in the

calculations. The damping ratio for a single cylinder can be derived as

1 D Cd
Zc = (KC) (24)
3π 2 Td Cm

Fig. 6 is a plot of the damping ratio Zc calculated by using Eq. (24) and the exper-

imental results of Thiagarajan & Troesch (1994). The friction drag component of

the damping ratio of the present numerical estimation Zf riction and that calculated

based on the laminar boundary theory are also plotted in Fig. 6. As can be observed

from the figure, the numerical solutions agree reasonably well with the experimen-

tal results. Damping ratio of both numerical solution and experimental data showed

approximately linear trends in the range of KC = 0.1 ∼ 1.0.

Applying laminar boundary layer theory to a vertical cylinder of radius R and draft

Td in heave, the friction drag component, i.e. the shear force in phase with velocity,

of the damping coefficient and damping ratio are obtained as:

Bf riction = π 1.5 Td µβ 0.5 , (25)

2 1 0.5
Zf riction = ( ) . (26)
Cm 4πβ

Eqs. (25) and (26) show that Bf riction and Zf riction are independent of KC based

14
on laminar boundary layer theory.

The following observations were made regarding the friction drag of the present nu-

merical results and that calculated using Eq. (25) shown in Fig. 6. Firstly, Zf riction

of the present numerical solution is approximately independent of KC, and about

7% higher than the value calculated from laminar boundary layer theory. This

agreement can be regarded as acceptable. The additional friction drag may be at-

tributed to the turbulence effect although KC is very small. Secondly, Zf riction

tends to decrease slightly with increasing KC due to the leading edge effect of the

truncated cylinder, while Eqs. (25) and (26) are based on the infinite length cylin-

der. Tao et al. (2000) revealed that the effect of the leading edge depends on the

aspect ratio (diameter/draft) of the cylinder. In the present case of the aspect ra-

tio about 0.35, numerical computation shows that Zf riction drops about 8% as KC

increases from 0.001 to 1.0 at β = 89236.

As shown in Fig. 6, numerical estimation of the damping ratio Z is approximately

linear with KC in the range of 0.045 ∼ 1.0 and appears to be a constant as KC →

0 at β = 89236. By subtracting the KC independent friction component, the form

drag component of damping ratio which contributes to the slope of the curves also

shows approximately linear trend with KC in the range of 0.045 ∼ 1.0, and tends

to be constant as KC → 0.

Fig. 6 also shows that the form drag component of the damping ratio of the numer-

ical results appears to be in general larger than the experimental data. In the present

numerical investigation, cylinder bottom edge is at 90 degrees, but this may be hard

to obtain in an experimental setup. Tao & Thiagarajan (2000) demonstrated that a

very small corner radius of the bottom edge can result in a significant reduction of

the form drag.

15
5 Vortex shedding and hydrodynamics of oscillating cylinder+disk configu-

ration

Tao (2002) have conducted numerical study on the effect of the oscillation param-

eter (KC) and geometry, primarily the thickness of the disk for a cylinder+disk

configuration on the behaviour of the vortex shedding. As a result of his compre-

hensive study, three vortex shedding modes were observed for a parameter range

of KC = 0.00075 ∼ 0.75 and aspect ratio td /Dd = 0.00075 ∼ 1.642. The en-

hancement on vortex shedding due to reducing thickness of the disk was identified.

However, a systematic variation of disk geometry, especially the diameter of the

disk and the corresponding influence on damping and added mass was not fully

explored. Further, current practice in offshore engineering design of heave plate for

Spar or truss spar platform is based on model tests which were conducted at low

β and the recommended drag coefficient for plates was between 2 and 3. However,

there could be a significant difference in drag coefficient between model test re-

sults and prototypes (Prislin et al., 1998). The focus of this study is on the effect of

the diameter ratio of the cylinder+disk configuration on heave damping and added

mass. Flow of oscillating cylinder with a disk attached to its base is calculated for

a broad range of diameter ratio, Dd /Dc , from 1.0 to 2.0. The vortex shedding pat-

terns are presented for various values of Dd /Dc . Finally, the optimum diameter

ratio in terms of heave motion suppression is determined based on the heave damp-

ing calculation. The influence of Reynolds number (and consequently β) on the

hydrodynamic damping is also investigated by examining the damping behaviour

in realistic β in a case study.

In order to illustrate the development of the vortex shedding pattern, the vorticity

in a meridional plane is computed by taking the curl of the velocity field. In the

16
axi-symmetrical case,

∂u ∂v
ζ= − . (27)
∂y ∂r

5.1 Influence of diameter ratio on vortex shedding pattern

Fig. 7 show vorticity contours of the flow around the disk edges due to cylin-

der+disk oscillating at KC = 1.0 computed from Eq. (27). The instantaneous vor-

ticity contours presented in Fig. 7 are equally spaced over one cycle. The vortices

formed with a smaller disk diameter (Fig. 7(a-d)) appear compressed showing the

influence of the presence of the cylinder wall. On the other hand, vortices at larger

disk diameter (Fig. 7(e-h)) are more rounded and appear to move around without

the cylinder’s hindrance. From the point of view of energy conservation, vortices

leaving a control volume surrounding the cylinder and disk represents effective

means of energy dissipation and therefore imply increased damping forces due to

stronger vortex shedding processes at larger disk diameter. According to Thiagara-

jan et al. (2002), the core diameter is of the same order of the oscillation stroke.

Therefore, the disk extension should be at least of the same order being sufficient

to accommodate the core of the vortex ring shed during an oscillation process.

5.2 Influence of geometry on the hydrodynamic coefficients

In order to compare the influence of the disk geometry on the hydrodynamic coef-

ficients of oscillating cylinder+disk configuration at same amplitude of oscillation,

a new Keulegan-Carpenter number and frequency parameter based on the diameter

17
of cylinder are defined as

2πa
KCc = . (28)
Dc

(Dc )2 f
βc = (29)
ν

5.2.1 Effects of disk diameter on added mass coefficient

The added mass of a disk oscillating along its axis approximately equal to the mass

of a sphere of water enclosing the disk (Sarpkaya & Isaacson, 1981)

1
ma = ρDd3 . (30)
3

For the configuration of a cylinder with a disk attached to its base, if the diameter

of the disk is greater than that of the cylinder, there is only a part of the disk on the

cylinder side producing added mass effect since the presence of the cylinder (see

Fig. 8). Thus, the added mass of a cylinder+disk becomes

1 πρ q
ma = ρDd3 − [ Dc2 (Dd − Dd2 − Dc2 )
3 q8 q
πρ
+ (Dd − Dd2 − Dc2 )2 (2Dd + Dd2 − Dc2 )]. (31)
24

The added mass coefficient, Ca0 , of a cylinder+disk is then obtained as the ratio of

the added mass to the displaced mass (m + ma )

1 4 3 1 2 q
Ca0 = 2 { D − [ Dc (Dd − Dd2 − Dc2 )
Dd td + Dc2 Td 3π d 2
1 q q
+ (Dd − Dd2 − Dc2 )2 (2Dd + Dd2 − Dc2 )]}. (32)
6

Fig. 9 shows the results for added mass coefficient versus the diameter ratio Dd /Dc .

Excellent agreement between the results of the present numerical computation and

18
theoretical expression demonstrated the applicability of Eq. (32) for calculating

added mass coefficient for a cylinder+disk configuration. The little variation among

the results from numerical calculation at three different KC numbers shown in Fig.

9 indicates a weak influence of KC on added mass.

Typically a classic Spar platform has a natural heave period around 25 ∼ 30 sec-

onds, which in most circumstances is sufficiently outside the prevailing wave fre-

quency range and thus heave motion is generally insignificant. However, a long

swell condition, with peak period lying in the 23 ∼ 25 second range may persist a

considerable portion of the year in West African offshore. Possible resonance with

long period swell may result in large amplitude heave oscillations, and causes fur-

ther damage to risers and mooring systems, as demonstrated in the model tests by

Thiagarajan et al. (2002). For a floating vertical cylinder, the heave natural period

of the cylinder can be derived from dynamic analysis as

s
m + ma
T = 2π (33)
ρgA + km

where A denotes the area of water plane, and km is the mooring stiffness. For a

freely floating cylinder of uniform cross section area without mooring effects, Eq.

(33) is simplified to
s
Td (1 + Ca )
T = 2π (34)
g

where Td is the draft of the cylinder. Eqs (33) and (34) indicate that the natural heave

period increases on increasing added mass. Thus, the longer natural heave period

of a Spar by increasing the diameter of the disk is favourable in preventing serious

occurrence of heave resonance. However, extension of this approach to another

type of deep water offshore structure TLPs should be done with caution, since

19
increasing the natural heave period by a disk of increasing diameter, from very low

value, typically 2 ∼ 5 seconds, may result in closer to the peak period of a wave

spectrum, and further amplifying springing vibration.

5.2.2 Effects of disk diameter on damping ratio

As observed in flow visualisation, the diameter of the disk has a significant in-

fluence on vortex shedding patterns. Since form damping dominant, the most basic

and typical diameter ratio effects on vortex shedding behaviour of the cylinder+disk

flow can also be investigated by examining the diameter ratio effects on the hydro-

dynamic damping, and this is certainly of more interests in offshore engineering

design. In Fig. 10, the viscous damping ratio, as a function of diameter ratio Dd /Dc

for three KCc numbers, 0.1, 0.5 and 1.0 are plotted. The curves in Fig. 10 show a

weak nonlinear trend of Z with Dd /Dc at different KCc . The damping ratio appears

to increase as diameter ratio increases for all three KCc numbers. However, consid-

erable flattening of the curves are seen at larger disk diameters. As observed in flow

visualisation, for small diameter ratio, since the small extension of the disk edges

relative to the cylinder wall, interference of the cylinder boundary on the vortex for-

mation and shedding process is evident, even at small KCc number. The vortices

formed with a disk of smaller diameter appear suppressed, resulting in lower damp-

ing. As the diameter of the disk increases to certain stage, the extension is much

larger than the length scale of the vortex formation primarily dependent on KCc ,

the interference of the cylinder boundary does not occur. The vortex shedding pro-

cess around the disk edges becomes independent of diameter ratio, indicating that

any further increase in the disk diameter would not result in appreciable increases

in drag. It is also seen in the figure that the slope of the curve increases on higher

KCc number, and tends to flatten at higher diameter ratio due to the interference

20
from the cylinder to the larger vortices generated. Therefore, the stronger diame-

ter ratio dependence of viscous damping is only significant in combination with

KC number effects. In realistic offshore situation, the amplitude of second-order

“springing” vibration of a TLP is very small, corresponding to KC number in the

order 0.01 or less, while the KC number of possible heave resonance for a Spar

can be as high as 1.0. Therefore, different diameter extensions should be selected

based on the above argument. For the highest KC of 1.0 in Fig. 10, the flatterning

occurs at Dd /Dc ≈ 1.65, corresponding to an extension of the disk approximately

four times the oscillation stroke.

5.2.3 Effects of KC

Fig. 11 shows the variation of damping ratio, Z, against KCc at βc = 89236. Again,

cases considered here include diameter ratio range from 1.0 to 2.0, being represen-

tative for flow generated by a single cylinder and a cylinder with disks of small to

large diameter ratio respectively. Two components of viscous damping ratio, i.e.

friction drag (Zf riction ) and form drag (Zf orm ) for Dd /Dc = 1.33 are also plotted

in Fig. 11. As shown in the figure, over the range of KCc calculated, friction damp-

ing is approximately constant. This indicates that friction damping is independent

of amplitude of the oscillation. However, form damping shows predominant and

increases dramatically as KCc increases except for extremely low KCc , in which

form damping is approaching a constant value as KCc → 0, since separation may

not occur or very weak, resulting in very small form damping. Similar trend of vis-

cous damping was also observed for a single truncated cylinder and cylinder+disk

at Dd /Dc = 2.0. Depending on the diameter ratio, all curves of Z ∼ KCc show

that the non-linear relationship exists between the damping ratio and KCc .

21
5.2.4 Effects of disk thickness td /Dd

As observed in flow visualisation, a fully developed vortex shedding process and

increasing vorticity formed around the disk due to lack of interference from the

cylinder boundary layer at a larger diameter ratio, indicate enhancing energy dissi-

pation. As show in Fig. 10, this further results in the significant increase in heave

damping. Damping variation due to different aspect ratio, td /Dd , is shown in Fig.

12. As can be seen from the figure, the increase in damping ratio is steeper when

the disk is thinner. This is due to the absence of disk thickness significantly en-

hances interaction of the vortices formed during any two successive half cycles

of the flow leading to increased shed vorticity. However, the trend of increasing

heave damping for the thinner disk is reversed for aspect ratio around 0.12 ∼ 0.16,

probably due to the forces of viscosity and convection cancelling each other in the

small range. Therefore, to maximise the damping effects, minimum thickness that

satisfies structural and fatigue strength requirements should be chosen.

6 Case study

As shown in Fig. 11, the heave damping comprises of two major components, i.e.

small and KC independent friction drag and predominant KC dependent form

drag. This is a clear indication that flow separation and vortex shedding are the

main mechanism of energy dissipation. Therefore, one could argue the influence

of β, and consequently Reynolds number to be a minimum. The hydrodynamic

behaviour of a Spar prototype with a heave plate oscillating at realistic β number

of 107 will be examined through the calculation of viscous damping and added

mass using the present numerical method.

22
Table 2

Oscillation parameters and geometry of prototype Spar and heave plate

Configuration 1 2 3 4 5 6

Spar diameter (m) 39.0

Total draft (m) 198.1

Disk thickness (m) no disk 0.475 0.475 0.475 0.475 0.475

Disk diameter (m) no disk 42.1 45.1 51.2 61.0 78.0

Diameter ratio (Dd /Dc ) 1.0 1.08 1.16 1.31 1.56 2.0

Aspect ratio (td /Dd ) no disk 0.0113 0.0105 0.0093 0.0078 0.0061

Amplitude (m) 0.91 2.74 4.57

KC 0.15 0.44 0.74

Period (sec) 28

β 4.57 × 107

Numerical simulations were conducted with a 198.1 m draft and 39.0 m diameter

Spar column and a 0.475 m thick disk and five different disk diameters of 42.1

m, 45.1 m, 51.2 m, 61.0 and 78.0 m. Six configurations comprising of the bare

hull and five different disk diameters, oscillating at three different amplitudes of

oscillation at a constant period of 28 sec were considered. The corresponding KCc

and β numbers were 0.15, 0.44, 0.74 and 4.6 × 107 respectively (see Table 2).

The behaviour of Z as a function of the diameter ratio is shown in Fig. 13. Shown

also in the figure are the experimental values obtained by Thiagarajan et al. (2002)

for low amplitudes and high amplitudes of oscillation at 1 : 75 model scale. It is

23
noted that while the computational problem is for forced heave oscillation, the ex-

perimental method is based on heave decay oscillations, where amplitude changes

with time. In spite of this difference in approaches, the agreement shown in Fig.

13 is remarkably good, for both low and high KC regimes. The significance of the

good agreement is the fact that the computational results were obtained for realis-

tic β numbers typical of prototype values, while the β numbers in model test are

several orders smaller than prototypes. This is a clear indication that β effects in es-

timation of heave damping can be considered negligible, since most of damping is

geometry related form damping due to flow separation and vortex shedding. Simi-

lar to the results obtained from the calculations at model scale, the curves in Fig. 13

show a weak nonlinear trend of Z with Dd /Dc at different KCc . Considerable flat-

tening of the curves is seen when the disk diameter gets larger due to the diminish

of the interference between the vortex shedding and Spar hull. For this particular

Spar prototype, Z is found to be approximately constant beyond Dd /Dc = 1.16 at

KCc = 0.15, corresponding to a disk of diameter 45.2 m, i.e. an extension of 3.1 m

to Spar hull. This value is about three times of the amplitude. However, at seastates

with wave spectrum of long peak period, such as West Africa, a classic Spar can

suffer from severe heaving with KCc up to 0.7 ∼ 1.0. As can be seen from Fig.

13, the flattening trend of the curves for larger KCc , i.e. 0.44 and 0.74, starts at

much larger diameter ratio values, approximately 1.3 and 1.6 respectively. There-

fore, consideration of KC number range of the oscillation must be taken when one

to determine the disk geometry for heave suppression. For instance, the heave os-

cillation of a Spar may be rather large (KCc up to 1.0) while the amplitude of a

TLP “springing” is very small (KCc < 0.01). As a general guideline it is suggested

that the disk extension be at least four times the typical heave amplitude to get the

optimum drag effect.

24
7 Conclusions

An axisymmetric finite difference method has been validated for the prediction of

the flow generated by an oscillating vertical cylinder with a circular disk attach to

its base at low Keulegan-Carpenter numbers. The numerical scheme has been used

to investigate the vortex shedding flow and the associated hydrodynamic behaviour

of the cylinder+disk configuration. The dependence of vortex shedding pattern on

the diameter ratio Dd /Dc and KC number have been accurately predicted.

The strong influence of diameter ratio was found on the vortex shedding and the

viscous damping. However, the effect of changing diameter ratio was found to be

dependent very strongly on KC number. Within the parameter range of KC up to

1.0 calculated in this study, it is found that the form drag due to flow separation

and vortex shedding generated by the sharp edges of the disk are the dominant

mechanism of energy dissipation.

Selection of the disk geometry in terms of the diameter should be considered in

combining with KC number of heave oscillation since the disk geometry effect on

drag is strongly influenced by KC number. A general trend of increasing damping

ratio on increasing diameter ratio is found from the present numerical calculation.

However, any increase beyond a certain diameter would not result in appreciable

increases in drag. Based on the present study, it is suggested that the disk extension

be at least four times the typical heave amplitude to get the optimum drag effect.

Case study further demonstrated that β effects in estimation of heave damping can

be considered negligible, since most of damping is geometry related form damping

due to flow separation and vortex shedding. The results obtained from this study

can be directly applied to the design of heave plate of floating offshore structures

25
such as TLP, Spar and truss spar.

References

Beaudan, P. & Moin, P.1994 Numerical experiments on the flow past a circular

cylinder at sub-critical Reynolds number, Report No. TF-62, Department of Me-

chanical Engineering, Stanford University.

Bozeman, J. D. & Dalton, C. 1973 Numerical study of viscous flow in a cavity.

Journal of Computational Physics 12, 348–363.

Burggraf, O. R. 1966 Analytical and numerical studies of the structure of the steady

separated flows. Journal of Fluid Mechanics 24, 113–151.

Chakrabarti, S. K. & Hanna, S.Y. 1991 High-Frequency Hydrodynamic Damping

of a TLP Leg. Proceedings of 10th International Offshore Mechanics and Arctic

Engineering 1-A, 147–151.

Huse, E. 1990 Resonant Heave Damping of Tension Leg Platforms. Proceedings

22nd Annual OTC Paper 6317, 431–436.

Huse, E. & Utnes, T. 1994 Springing Damping of Tension Leg Platforms. Proceed-

ings 26nd Annual OTC Paper 7446, 259–267.

Kawamura, T. & Kuwahara, K. 1984 Computation of High Reynolds Number Flow

around a Circular Cylinder with Surface Roughness. AIAA Paper No. 84-0340,

1–11.

Prislin, I., Blevins, R.D. & Halkyard, J. 1998 Viscous damping and added mass

of solid square plates. Proceedings of The 17th International Conference on

Offshore Mechanics and Arctic Engineering.

Sarpkaya, T. & Isaacson, M. 1981 Mechanics of Wave Forces on Offshore Struc-

tures., van Nostrand Reinhold, New York.

26
Schreiber, R. & Keller, H. B. 1983 Driven cavity flows by efficient numerical tech-

niques. Journal of Computational Physics 49, 310–333.

Tao, L., Thiagarajan, K. P. & Cheng, L. 2000 On the parametric dependence of

springing damping of TLP and Spar columns. Applied Ocean Research 22,

281–294.

Tao, L. & Thiagarajan, K. P. 2000 The Influence of Edge Sharpness on the Heave

Damping Forces Experienced by a TLP Column. Proceedings of The 10th Inter-

national Offshore and Polar Engineering Conference (ISOPE) I, 277–282.

Tao, L. 2002 Numerical Investigation of Hydrodynamic Heave Damping of Deep

Water Offshore Structures, Ph.D. Thesis, The University of Western Australia.

Thiagarajan, K. P. 1993 Hydrodynamics of Oscillating Cylinder and Disks at Low

Keulegan-Carpenter Numbers, Ph.D. Thesis, Department of Naval Architecture

and Marine Engineering, The University of Michigan.

Thiagarajan, K. P. & Troesch, A. W. 1994 Hydrodynamic heave damping estima-

tion and scaling for tension leg platforms Columns. Journal of Offshore Me-

chanics and Arctic Engineering, ASME 116, 70–76.

Thiagarajan, K. P. & Troesch, A. W. 1998 Effects of Appendages and Small Cur-

rents on the Hydrodynamic Heave Damping of TLP Columns. Journal of Off-

shore Mechanics and Arctic Engineering, ASME 120, 37–42.

Thiagarajan, K. P. Datta, I., Ran, Z., Tao, L. & Halkyard, J. 2002 Influence of

heave plate geometry on the heave response of classic Spars. Proceedings of The

21st International Conference on Offshore Mechanics and Arctic Engineering

OMAE2002-28350.

27
 Cylinder

F
a sin( ωt)

Disk ∑

o r

Fig. 1. Flow configuration for a cylinder with a disk attached on its bottom in heave

∂v
u=0, =0
∂y
5.335 R

y(t)=y(0)+a sin(ω t) for t>0

u=0, v=a ω cos(ω t)
R
Dd/2
∂u
 =0
∂r
10 R

u=0
v=0
∂v
=0
∂r

y
∂v
r u=0,  =0
∂y
11 R

Fig. 2. Flow domain and boundary conditions

28
 10.4

14

10.3
12

10 10.2

8
y

y
10.1

6
10
4

9.9
2

0 9.8
0 5 10 15 1 1.1 1.2 1.3 1.4 1.5 1.6
r r
(a) (b)

Fig. 3. Mesh generation for the simulation of flow induced by an oscillating cylinder: (a)

Physical grid; (b) Typical grid distribution around the disk surface.

u=1, v=0

u=0 xc
u=0
v=0 v=0

yc

x u=0, v=0

Fig. 4. Problem definition and flow pattern of a lid-driven square cavity

29
 x=0.1 Schreiber & Keller (1983)
1 x=0.5 Schreiber & Keller (1983)
x=0.9 Schreiber & Keller (1983)
x=0.1 Present
x=0.5 Present
0.75 x=0.9 Present

0.5
u

0.25

-0.25

-0.5
0 0.25 0.5 0.75 1
y

Fig. 5. Distributions of the horizontal velocity component of a lid-driven square cavity at

Re = 1000

Z - exp (Thiagarajan & Troesch 1994)

Zfriction (Boundary layer theory)
Z - num
0.02
Zfriction - num
Damping ratio - Z

0.01

0
0 0.2 0.4 0.6 0.8 1
KC

Fig. 6. Damping ratio Z versus KC at β = 89236.

30
 (a) (e)
11.5
11.5

11
11

10.5
10.5

10
y

10

y
9.5
9.5

9
9

8.5
8.5

(b) (f)
11.5
11.5

11
11

10.5
10.5

10
y

10

y
9.5
9.5

9
9

8.5
8.5

(c) (g)
11.5
11.5

11
11

10.5
10.5

10
y

10
y

9.5
9.5

9
9

8.5
8.5

(d) (h)
11.5
11.5

11
11

10.5
10.5

10
y

10
y

9.5
9.5

9
9

8.5
8.5

Fig. 7. Vortex formation around the disk at KCc = 1.0, βc = 89236. (a) to (d) represent

flow for Dd /Dc = 1.33 and (e) to (h) represent flow at Dd /Dc = 2.0. Dashed lines indicate

negative vorticity.

31
 Fig. 8. Added mass of a disk attached to a cylinder

1.2 Ca (Theory)
Ca KC c=0.1
Ca KC c=0.5
1 Ca KC c=1.0
Added mass coefficient Ca

0.8

0.6

0.4

0.2

0
1 1.2 1.4 1.6 1.8 2
D d/D c

Fig. 9. Added mass coefficient Ca versus Dd /Dc at different KCc (βc = 89236)

32
 td/D d
0.05 0.04 0.03
0.08

Ztotal KC c=0.1
Ztotal KC c=0.5
Ztotal KC c=1.0
Damping ratio Z 0.06

0.04

0.02

0
1 1.2 1.4 1.6 1.8 2
D d/D c

Fig. 10. Damping ratio Z versus Dd /Dc at different KCc (βc = 89236)

0.08 Z [D d/D c=1]

Z [D d/D c=1.33]
Zfriction [D d/D c=1.33]
Zform [D d/D c=1.33]
Z [D d/D c=2]
0.06
Damping ratio Z

0.04

0.02

0
0 0.2 0.4 0.6 0.8 1
KCc

Fig. 11. Damping ratio Z versus KCc at different Dd /Dc (td /Dd = 0.417, βc = 89236)

33
 0.06

Damping ratio - Zform

0.04

0.02

0
0 0.5 1 1.5
Aspect ratio - td/D d

Fig. 12. Effect of aspect ratio td /Dd on heave damping (KCc = 1.0).

td/D d
0.012 0.011 0.01 0.009 0.008 0.007
0.06
Z(KCc=0.15)
Z(KCc=0.44)
Z(KCc=0.74)
0.05
Z(exp.- lowKC)
Z(exp.- highKC)

0.04
Damping ratio Z

0.03

0.02

0.01

0
1 1.2 1.4 1.6 1.8 2
D d/D c

Fig. 13. Zf orm versus Dd /Dc at different KCc (βc = 4.57 × 107 )

34