Ocean Engineering 108 (2015) 738–748

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

CFD investigations of wave interaction with a pair of large tandem

cylinders
Arun Kamath n, Mayilvahanan Alagan Chella, Hans Bihs, Øivind A. Arntsen
Department of Civil and Transport Engineering, Norwegian University of Science and Technology, 7491 Trondheim, Norway

art ic l e i nf o a b s t r a c t

Article history: Wave forces and the flow field around cylinders placed in a periodic wave field are investigated with a
Received 6 January 2015 numerical model using the Reynolds-averaged Navier–Stokes equations. The numerical model is vali-
Accepted 25 August 2015 dated by simulating the wave interaction with a single cylinder and comparing the numerical results
Available online 20 September 2015
with experimental data from a large scale experiment. Then, the wave interaction with a single large
Keywords: cylinder and a pair of large cylinders placed in tandem for different incident wave steepnesses is studied.
Wave forces The numerically calculated forces are compared with predictions from potential theory. The numerical
Wave interaction results are seen to match the predictions at low incident wave steepness but differ at higher incident
Vertical cylinders wave steepnesses. The wave diffraction pattern around the tandem cylinders for waves of low and high
Numerical wave tank
steepness is investigated and the evolution of a strong diffraction pattern is seen in the case of high
CFD
steepness waves, which results in the difference between the wave forces predicted by potential theory
and the numerical model at higher steepnesses.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction the incident wave is steep. The importance of non-linear interac-

tions arising from diffracted waves and the viscous effects in an
Circular cylindrical structures are commonly used in the sup- unseparated flow regime have to be investigated by accounting for
port structures of offshore wind turbines, oil and gas platforms, these phenomena and comparing the results with predictions
offshore mooring dolphins in deep and intermediate waters and from potential theory.
nearshore coastal structures. Understanding the interaction of MacCamy and Fuchs (1954) derived an equation using linear
waves with these structures is important for the accurate predic- potential theory to obtain the first-order wave force on a single
tion of the hydrodynamic loads on them. Moreover, the interaction large cylinder using the wave diffraction potential. This equation is
of waves with large cylindrical structures always modifies the commonly used to determine wave forces on a single large
characteristics of the incident wave field and influences the wave cylinder exposed to regular waves. Chakrabarti and Tam (1973)
induced processes of wave radiation and diffraction. The modified carried out experimental studies on large cylinders exposed to
kinematics of the flow field changes the flow processes such as the small amplitude waves and found good agreement with predic-
wave run-up, reflection and transmission. In the case of a circular tions from linear potential theory. Some studies proposed certain
cylinder, the contribution of drag and inertia forces to the total methods to evaluate higher order forces using potential theory
forces is determined by the KC number and the diffraction para- (Lighthill, 1979; Molin, 1979), but had difficulties in obtaining
convergent solutions.
meter. When the diffraction parameter, which is the ratio of the
In a diffraction regime, the incident wave train is affected by its
cylinder diameter (D) to the incident wavelength (L), is greater
interaction with the cylinder and its effects are seen even outside
than 0.2 ( D/L > 0.2) and the KC number is smaller than 2, the flow
the immediate vicinity of the cylinder. This results in a complex
is inertia dominated and wave diffraction effects are important
hydrodynamic problem when groups of large cylinders are placed
(Isaacson, 1979). Lower-order solutions can be obtained with
in a wave field. Ohkusu (1974) proposed an iterative method to
analytical formulations based on potential theory by assuming that
evaluate successive water wave scattering by floating bodies,
the fluid is inviscid, the flow irrotational and the wave amplitude
based on the work by Twersky (1952) for electromagnetic and
small compared to the diameter of the cylinder. The methods
acoustic waves. The velocity potential functions used in this
based on potential theory are limited by these assumptions, when approach become harder to work with as the number of cylinders
is increased. Spring and Monkmeyer (1974) proposed a method
n
Corresponding author. Tel.: þ 47 73 59 46 40; fax: þ47 73 59 70 21. where all the boundary conditions are enforced at once and the
E-mail address: [email protected] (A. Kamath). wave forces are determined by solving a set of linear equations.

http://dx.doi.org/10.1016/j.oceaneng.2015.08.049
0029-8018/& 2015 Elsevier Ltd. All rights reserved.
 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748 739

Linton and Evans (1990) improved the method by Spring and 2. Numerical model
Monkmeyer (1974) and proposed a method with a simplified
expression for the velocity potential to obtain the maximum first- 2.1. Governing equations
order force, the mean second-order force on the cylinder and to
calculate the free surface amplitudes for equally spaced identical REEF3D uses the incompressible Reynolds-averaged Navier–
cylinders. Using this analytical method, it is possible to evaluate Stokes (RANS) equations together with the continuity equation to
the amplitude of the wave forces on cylinders placed in a group solve the fluid flow problem:
and to determine the maximum variation of the free surface ∂ui
around the cylinders. =0
∂xi (1)
The limitation of analytical formulae based on potential theory
is that they have to be modified to deal with different scenarios,
for example, to study structures of different geometries, to study ⎡
∂ui ∂u 1 ∂p ∂ ⎢ ⎛ ∂u ∂uj ⎞⎤
non-linear wave–wave and wave–body interactions due to waves + uj i = − + (ν + νt )⎜⎜ i + ⎟⎟⎥ + gi
of high steepnesses. Numerical modeling based on boundary ∂t ∂xj ρ ∂xi ∂xj ⎢⎣ ⎝ ∂xj ∂xi ⎠⎥⎦ (2)
integral equations (Ferrant, 1995; Boo, 2002; Song et al., 2010)
have the same limitations as potential theory, on which they are where u is the time averaged velocity, ρ is the density of the fluid,
based. On the other hand, Computational Fluid Dynamics (CFD) p is the pressure, ν is the kinematic viscosity, νt is the eddy visc-
modeling provides an immense amount of detail regarding the osity and g is the acceleration due to gravity. The pressure is
wave hydrodynamics by representing most of the wave physics determined using the projection method (Chorin, 1968) and the
with few assumptions. CFD modeling of wave interaction with resulting Poisson equation is solved with a preconditioned BiCG-
large cylinders placed close to each other can provide more insight Stab solver (van der Vorst, 1992). Turbulence modeling is carried
into the physical processes, such as the effect of wave diffraction out using the two equation k–ω model proposed by Wilcox (1994).
on neighboring objects including the wave elevation, wave forces, The strain in the flow due to the waves leads to unphysical over-
water particle velocities, the influence of the center-to-center production of turbulence in the wave tank. To avoid this, eddy
viscosity limiters are used as shown by Durbin (2009). Also, the
distance and the incident wave steepness. The scale and geome-
strain due to the large difference in density at the interface
tries considered in studies using a CFD model may not be directly
between air and water causes an overproduction of turbulence at
applicable to determining the hydrodynamic loads on an offshore
the interface. This is avoided by free surface turbulence damping
structure, but the validation of such a model provides the first step
around the interface as shown by Naot and Rodi (1982). The
towards establishing such methods to an eventual application to
damping is carried out only around the interface using the Dirac
larger and more complicated problems, with realistic geometries
delta function. REEF3D is fully parallelized using the domain
and scales in the future, since full scale data and field observations
decomposition strategy and MPI (Message Passing Interface).
are generally lacking. Another application is to extend the studies
to random wave forces (Boccotti et al., 2012) after establishing the
numerical model for regular waves in this study. The validated 2.2. Free surface
numerical model can be used to gain further insight into the
The free surface is determined with the level set method. The
applicability of the Morison equation in the case of random waves →
and build upon the knowledge gained from the field experiments zero level set of a signed distance function ϕ( x , t ) is used to
in recent literature (Boccotti et al., 2013). represent the interface between air and water (Osher and Sethian,
In this study, the open source CFD model, REEF3D (Alagan 1988). Moving away from the interface, the level set function gives
Chella et al., 2015) is used to analyse wave interaction with bot- the shortest distance from the interface. The sign of the function
tom-fixed vertical cylinders in a 3D numerical wave tank. The distinguishes between the two fluids across the interface as shown
paper presents studies with a large number of simulations inves- in the following equation:
tigating the changes in the wave hydrodynamics with small ⎧ →
⎪ >0 if x is in phase 1
incremental changes in parameters using CFD simulations. The → ⎪
model is validated by comparing the computed wave forces on a ϕ( x , t )⎨ =0 if →
x is at the interface
⎪
single cylinder, free surface elevations around the cylinder and the ⎪ →
⎩ <0 if x is in phase 2 (3)
water particle velocities with the experimental data from the
large-scale experiments carried out at the Large Wave Flume The level set function is moved under the influence of an external
(GWK) in Hannover, Germany by Mo et al. (2007). Then, the wave velocity field uj with the convection equation:
forces on a single cylinder and on a pair of tandem cylinders for ∂ϕ ∂ϕ
different wave steepnesses and center-to-center distances are + uj =0
∂t ∂xj (4)
calculated in 108 numerical simulations. The wave forces on a
single cylinder due to waves of different steepnesses are studied, The level set function loses its signed distance property on con-
along with the wave elevation around the cylinder. The wave vection and is reinitialized after every iteration using a partial
forces experienced by a pair of tandem cylinders with different differential equation based reinitialization procedure by Peng et al.
center-to-center distances and different incident wave steepnesses (1999) to regain its signed distance property.
are evaluated. A total of 96 simulations are carried out to inves-
tigate the change in the wave forces with respect to the center-to- 2.3. Discretization schemes
center distance and the wave steepness. The wave elevation in the
vicinity of the cylinders is studied to gain more knowledge about The fifth-order conservative finite difference Weighted Essen-
the wave propagation and the evolution of wave diffraction pat- tially Non-Oscillatory (WENO) scheme proposed by Jiang and Shu
terns between the neighboring cylinders. In addition, the analy- (1996) is applied for the discretization of the convective terms of
tical formula proposed by Linton and Evans (1990) is used to the RANS equation. The level set function, turbulent kinetic energy
compare the wave forces on the tandem cylinders for low wave and the specific turbulent dissipation rate are discretized using the
steepnesses where linear potential theory is valid. Hamilton–Jacobi formulation of the WENO scheme by Jiang and
740 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748

Peng (2000). The WENO scheme is a minimum third-order accu- also absorbs reflections from structures in the wave tank and
rate and numerically stable even in the presence of large gradients. prevents them from affecting wave generation. At the end of the
Time advancement for the momentum and level set equations is tank, the wave enters the numerical beach. Here, the computa-
carried out using a Total Variation Diminishing (TVD) third-order tional values of velocity and free surface are reduced to zero in a
Runge–Kutta explicit time scheme proposed by Shu and Osher smooth manner. This simulates the effect of a beach where the
(1988). Adaptive time stepping is employed to satisfy the CFL wave energy is removed from the wave tank.
criterion based on the maximum velocity in the domain. This
ensures numerical stability throughout the simulation with an
optimal value of time step size. A first-order scheme is utilized for 3. Calculation of wave forces
the time advancement of the turbulent kinetic energy and the
specific turbulent dissipation, as these variables are mostly source 3.1. Numerical evaluation of wave forces
term driven with a low influence of the convective terms. Diffusion
terms of the velocities are also subjected to implicit treatment in The numerical model evaluates the wave force F on an object as
order to remove the diffusion terms from the CFL criterion. The the integral of the pressure p and the surface normal component
convergence studies for the simulations are then just carried out of the viscous shear stress tensor τ on the object according to the
for the grid size to determine the accuracy of the results, since the following equation:
adaptive time stepping approach determines the optimal time step
required to maintain the numerical stability. As an example, in the F= ∫Ω ( − np + n·τ ) dΩ (7)
case of non-breaking wave interaction with a vertical cylinder
presented in this study, time steps are smaller, about 0.002 s where n is the unit normal vector pointing into the fluid and Ω is
during the first few seconds of the simulation as the waves are the surface of the object. This is readily accomplished by the
introduced into the wave tank and then increase to about 0.004 s numerical model as the values for pressure and shear stress are
as the periodic waves are established in the tank. In this way, the available at every point in the domain at any given time of the
adaptive time stepping approach determines the optimal time simulation.
step, reducing the cost of the simulation and avoiding numerical
instability in a simulation which could occur with a fixed time step 3.2. Analytical formulae for wave forces
approach.
The numerical model uses a uniform Cartesian grid for the Potential theory is used to obtain the wave diffraction potential
spatial discretization together with the Immersed Boundary and calculate the force on a single cylinder using the equation
Method (IBM) to represent the irregular boundaries in the domain. presented by MacCamy and Fuchs (1954), shown in the follwing
Berthelsen and Faltinsen (2008) developed the local directional equation:
ghost cell IBM to extend the solution smoothly in the same
4ρgia tanh(kd)
direction as the discretization, which is adapted to three dimen- |F | =
sions in the current model. k 2H1′(kr ) (8)

where i = −1 , a is the incident wave amplitude, k = 2π /L is the

2.4. Numerical wave tank wave number, d is the water depth and H1′ is the first derivative of
the Hankel function of the first kind and r is the radius of the
The numerical wave tank uses the relaxation method (Larsen cylinder. An extension of the diffraction theory proposed by Linton
and Dancy, 1983) for wave generation and absorption. This method and Evans (1990) to calculate wave forces on multiple cylinders
requires a certain length of the wave tank to be reserved as wave placed in proximity is presented in the following equation:
generation and absorption zones. Relaxation functions are used to
N M
moderate the velocity and the free surface using a wave theory in l
Am + ∑ ∑ Anj Z njei(n − m)αjl Hn − m(kRjl ) = − Ileim(π /2 − β)
the relaxation zones with j = 1 n =−M
≠l
urelaxed = Γ (x)uanalytical + (1 − Γ (x))ucomputational
l = 1 ,..., N , m = − M ,..., M . (9)
ϕrelaxed = Γ (x)ϕanalytical + (1 − Γ (x))ϕcomputational (5) where M is the order of the solution, N is the number of cylinders, I
where Γ(x ) is the relaxation function and x ∈ [0, 1] is the x-coor- is the incident wave potential, β is the angle of wave propagation
dinate scaled to the length of the relaxation zone. The relaxation with respect to the x-axis, H is the Hankel function of the first
function proposed by Jacobsen et al. (2012), shown in (6), is used kind, Rjl is the length of the line joining the centers of the jth and
in the numerical model: the lth cylinder, αjk is the angle between the x-axis and the line
joining the centers of the cylinders and Z = J′(krj ) /H′(krj ), where J is
3.5
ex − 1 the Bessel function of the first kind. The unknown coefficients A
Γ(x) = 1 −
e−1 (6) are to be evaluated. This results in a set of N (2M + 1) equations.
Linton and Evans (1990) suggest that a value of M ¼6 provides
The wave theory for moderating the numerical values is chosen
sufficiently accurate solutions. So, M ¼6 is used in the equations to
according to the wave steepness and the water depth in the
obtain the analytical prediction of wave forces in this study. The
simulation. Typically, the wave generation zone is one wavelength
unknown coefficients A are evaluated by solving (9) and the wave
long and the absorption zone is two wavelengths long. In the wave
forces are obtained using the following equation:
generation zone, the computational values of velocity and free
surface are raised to the analytical values prescribed by wave Fj 1
theory. The generation zone releases waves into the working zone = |A−j 1 ± A1j |
F 2 (10)
of the tank. The objects to be studied are placed in the working
zone of the tank. The relaxation function in the generation zone The subtraction of the coefficients on the right hand side gives the
 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748 741

wave force along the x-axis and the addition of the terms gives the the simulations with grid sizes of dx¼0.15m and 0.2 m. The force
wave force along the y-axis. In the current study, the angle of in these cases is compared with the calculated force using a grid
incidence β = 0 and the waves propagate along the x-axis. size of dx¼0.1 m and the experimental result. It is seen that the
numerical result converges to the experimental value at a grid size
of dx¼ 0.1 m in Fig. 2b. Thus, the selected grid size is sufficiently
4. Results small to accurately calculate the force on the cylinder.
The numerically obtained free surface elevation near the wall
4.1. Validation of the numerical model along the front line of the cylinder is compared with the experi-
mental data in Fig. 3a. The amplitude at the first crest is con-
The numerical model is validated by simulating the experi- sidered the maximum amplitude of the wave elevation recorded
ments carried out at the Large Wave Flume (GWK), Hannover, by the gage near the wall, ηmax, wall . The comparisons of the com-
Germany by Mo et al. (2007). The numerically computed values for puted and measured free surface elevation in front, at the side and
the free surface elevation around the cylinder, the water particle behind the cylinder are presented in Fig. 3b, c and d respectively.
velocity in the numerical wave tank and the wave force on the The difference in pressure in front and behind the cylinder is seen
cylinder are compared with the experimental data to confirm that in the free surface elevation around the cylinder. The numerically
the numerical model accurately calculates the wave kinematics obtained free surface elevation data shows a good match with the
and dynamics. The wave flume in the experiments is 309 m long, experimental measurements. The water particle velocity calcu-
5 m wide and 7 m deep. A cylinder of diameter D¼ 0.7 m is placed lated by the numerical model is compared with the experimental
111 m from the wavemaker and strain gages are placed at the top measurements at 0.93 m, 1.53 m and 2.73 m below the still water
and bottom of the cylinder in order to measure wave forces. Wave level at the side wall of the tank along the front line of the cylinder
gages are placed at several locations around the cylinder to mea- in Fig. 4. The numerical results are scaled with the numerically
sure the time histories of the free surface elevation. Four acoustic calculated wave celerity, C ¼5.48 m/s. The water particle velocity is
Doppler velocimeters (ADVs) are placed at the side wall along the expected to reduce with increasing distance from the free surface
front line of the cylinder at various depths to measure the water as seen in Fig. 4 with the amplitude of the velocity being the
particle velocities. lowest in Fig. 4a at 2.73 m from the still water level. The water
The numerical wave tank used in this simulation is 132 m long, particle velocities calculated by the model match the values
5 m wide and 8 m high. Fifth-order Stokes waves with a wave observed in the experiments very well, showing that the numer-
height H¼1.2 m, wave period T¼ 4.0 s, wavelength L ¼21.9 m are ical model is able to represent the wave kinematics correctly.
generated with a water depth d ¼4.76 m on a grid of dx ¼0.1 m.
The grid in the numerical wave tank is 1320 × 50 × 80 cells 4.2. Grid convergence study for wave propagation
resulting in a total number of 5.28 million cells. The cylinder is
placed in the center with respect to the side walls as seen in the Accurate wave generation and propagation in the numerical
numerical setup in Fig. 1. The diffraction parameter D/L = 0.032 wave tank is verified with a grid convergence study. A two-
and KC ¼6.1 in this case. dimensional wave tank with a length of 15 m, height of 1.0 m and
A net inline force acts on the cylinder due a difference in water depth d ¼0.5m is used. Fifth-order Stokes waves are gen-
pressure in front and behind the cylinder. The calculated force on erated with a wave height of H¼ 0.1 m, wavelength of L¼2.0 m
the cylinder is compared with the experimental data and a good and wave period T ¼1.14 s. This setup of the numerical wave tank
agreement is seen in Fig. 2a. Mo et al. (2007) noted that the force is used in the following sections to simulate the wave interaction
measured in the experiments matched the inertial force given by with large cylinders. The grid convergence is carried out for the
the Morison formula with Cm ¼2. So, it appears that the forces are most stringent case with the highest wave steepness used in the
study. The grid size dx in the wave tank is varied from 0.1 m to
inertia dominated, although the KC number is 6.1 in this case. A
0.01 m. The results are presented in Fig. 5. It is seen that the free
grid convergence study for the forces is carried out by repeating
surface elevation η conforms to the required value at a grid size of
dx¼0.025 m. The damping of the wave amplitude at grid sizes of
132m
0.1 m and 0.05 m is seen in the figure. This is reduced as the grid
44m size is reduced to 0.025 m and the improvement in the results on
further reducing the grid size is negligible. Thus, a grid size of
G WG 1
E B dx¼0.025 m is selected for the following simulations in the cur-
N
E WG 3 E rent study.
R
A WG 2 WG 4 A 5m
T C 4.3. Wave interaction with a single large cylinder
I
O H
N
Simulations are carried out with a cylinder of diameter
Fig. 1. Numerical setup used for validation of the model. D¼ 0.5 m in a wave tank 15 m long, 5 m wide and 1 m high with a

Numerical Experiment
0.5 0.5
F/ρgDH2

F/ρgDH2

0 0

−0.5 −0.5
5 6 7 8 5 6 7 8
t/T t/T

Fig. 2. Comparison of experimental and numerical results for the inline wave force on the cylinder: (a) wave force on a single circular cylinder and (b) convergence study for
wave force calculation.
742 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748

Fig. 3. Comparison of experimental and numerical results for free surface elevations around the cylinder: (a) along the frontline near the wall, (b) in front of the cylinder,
(c) behind the cylinder, and (d) beside the cylinder.

Fig. 5. Grid convergence study for wave propagation.

in Fig. 6a. The computed wave force on the cylinder for different
wave steepnesses is compared with the prediction from the
MacCamy–Fuchs equation in Fig. 6b. It is seen that the numerical
results agree with the predictions at lower wave steepnesses but
the numerical results for the higher wave steepnesses are seen to
be lower than the predictions from the equation. According to the
MacCamy–Fuchs equation, the wave force on the cylinder increa-
ses linearly with an increase in the incident wave height H for a
given cylinder diameter D. The variation of the computed force on
the cylinder with increasing steepness suggests that the total force
on the cylinder is reduced due to non-linear interaction of high-
steepness waves with the cylinder and the diffracted waves.
The variation of the free surface elevation η in front, behind and
beside the cylinder for an incident wave of low steepness
H /L = 0.003 shows 1.72 times the incident wave crest height ηc in
i
front of the cylinder in Fig. 7a. The phase difference between the
Fig. 4. Comparison of experimental and numerical results for wave particle velo-
wave elevations in front and behind the cylinder is 0.78π and it is
city in the wave tank: (a) z = − 0.93 m , (b) z = − 1.53 m, and (c) z = − 2.73 m.
0.24π between the elevations in front and beside the cylinder. In
water depth of d ¼0.5 m. Linear waves of height H ¼0.006 m and the case of an incident wave with the high steepness of H /L = 0.1
0.02 m, second-order Stokes waves with H ¼0.06 m and 0.1 m, in Fig. 7b, the evolution of wave asymmetry is apparent with the
fifth-order Stokes waves with H ¼0.11 m, 0.12 m, 0.13 m, 0.14 m, crest height 1.55ηc and the trough 0.95ηc in front of the cylinder.
i i
0.15 m, 0.16 m, 0.18 m and 0.2 m with a wavelength L¼ 2 m are The phase difference between the wave elevations in front and
incident on the cylinders resulting in D/L ¼0.25. The KC numbers behind the cylinder is 0.80π and it is 0.20π for the elevations in
for these simulations are between 0.04 and 1.37. The resulting front and beside the cylinder. Thus, the high steepness waves
wave steepnesses and the incident wave frequency for the differ- move faster around the upstream half of the cylinder but slower
ent cases are listed in Table 1. The linear and 2nd -order Stokes around the downstream half of the cylinder, in comparison to the
waves have the same wave frequency for different incident wave waves of low steepness. This points towards a deceleration of the
heights but in the case of 5th-order Stokes waves the wave height water particles in the region after the upstream half of the cylin-
is included in the dispersion relation and a small decrease in the der. The waveform behind the cylinder is also highly asymmetrical,
wave frequency is seen with increasing wave height. The com- resulting in shallower troughs behind the cylinder, when a crest is
puted inline wave force on the cylinder for H /L = 0.003 is com- incident in front of the cylinder. This increased asymmetry points
pared to the analytically predicted maximum and minimum value towards a different pressure difference regime in the case of the
from the MacCamy–Fuchs equation and a good agreement is seen high-steepness waves. As a result of the deceleration of the water
 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748 743

Table 1
Combination of parameters for simulations with a single large cylinder of diameter D¼ 0.5 m in a water depth of d¼ 0.5 m.

L (m) H /L

Linear waves 2nd -Order Stokes 5th -Order Stokes

2.0 0.003 0.01 0.03 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.09 0.10
f (Hz) 0.846 0.846 0.846 0.846 0.862 0.865 0.868 0.872 0.876 0.880 0.889 0.899

Fig. 6. Comparison of analytical and numerical results for the inline wave force on a single large cylinder: (a) H¼ 0.006 m and (b) wave force for different incident wave
steepnesses.

Fig. 7. Relative free surface elevations around the single cylinder for incident waves of low and high steepness: (a) H /L = 0.003 and (b) H /L = 0.1.

15m simulated. The different combinations of incident wave steepness

7m
2m 4m and the center-to-center distance for the 96 simulations are listed
G in Table 2. The cylinder directly facing the incident waves is
E cylinder 1 and the downstream cylinder is cylinder 2. Previous
N S B
E
E works using analytical methods (Linton and Evans, 1990; McIver
R 5m
A D
A and Evans, 1984; Malenica et al., 1999) have shown that the wave
C
T Cylinder 1 Cylinder 2 H forces on tandem cylinders are influenced by not only the incident
I 2.5m
O wave height and the spacing between the cylinder, but also by the
N
incident wave frequency. In order to maintain the focus on the
Fig. 8. Schematic diagram of the setup used for the simulations with two tandem effect of the incident wave height with small increments in wave
cylinders. steepness for different distances between the cylinder, the effect of
the incident wave frequency is not analysed in this paper.
particles and the asymmetry of the wave, the force acting on the The variation of the computed inline wave force on the cylin-
cylinder due to an incident wave of high steepness is lower than ders with center-to-center distances S for different incident wave
the prediction from MacCamy–Fuchs equation based on linear steepnesses H /L is presented in Fig. 9. The prediction from the
potential theory. formula by Linton and Evans (1990) is also included for obtaining a
baseline comparison. It is clearly seen that the analytical predic-
4.4. Wave interaction with a pair of tandem cylinders tion matches the computed wave force closely at the lowest wave
steepness of H /L = 0.003 for both cylinders, in Fig. 9(a) and (b). The
A set of simulations is carried out to study the wave interaction computed wave forces show a similar form of variation for
with two cylinders placed in tandem in the direction of wave H /L = 0.05 as predicted by the analytical formula but with lower
propagation. Cylinders with a diameter D ¼0.5 m are placed in a magnitudes in Fig. 9(c) and (d). The deviation from the predictions
wave tank that is 15 m long, 5 m wide and 1 m high with a water by the analytical formula is clear in Fig. 9(e) and (f) for the highest
depth d¼ 0.5 m on a grid of dx ¼0.025 m. A schematic diagram wave steepness simulated, H /L = 0.1. In addition to the amplitude
illustrating the numerical setup is given in Fig. 8. The grid is of the force, the form of the variation is also different at longer
600 × 200 × 40 cells resulting in a total of 4.80 million cells in the distances of separation S. Cylinder 1 experiences large changes in
numerical wave tank. Linear waves with a wave height the wave force when the center-to-center distance between the
H ¼0.006 m and 0.02 m, second-order Stokes waves with cylinders is changed. The difference between the largest force at
H ¼0.06 m and 0.1 m, fifth-order Stokes waves with H¼0.11 m, S¼ 0.8 m and the lowest force at S¼ 3.37 m is 35% for H /L = 0.003
0.12 m, 0.13 m, 0.14 m, 0.15 m, 0.16 m, 0.18 m and 0.2 m with a and H /L = 0.05, but about 22% for H /L = 0.1. The change in the
wavelength L ¼2m are incident on the cylinders. The KC numbers center-to-center distance S strongly affects cylinder 2 at small
in these cases range between 0.04 and 1.37. For each of the inci- values of S ¼0.8 m and S¼ 1.2 m, with a change of 17.4% for
dent wave heights, center-to-center distance between the two H /L = 0.003, 18% for H /L = 0.05 and 16% for H /L = 0.1. Whereas, the
cylinders S ¼0.8 m, 1.2 m, 1.6 m, 1.8 m, 2.0 m, 2.3 m and 3.37 m are difference in the forces at S ¼2.0 m and S¼ 3.37 m is 8% for
744 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748

Table 2
Combination of parameters for simulations with two tandem large cylinders with diameter D ¼0.5 m, incident wavelength L ¼ 2.0 m in a water depth d¼ 0.5 m.

S (m) H /L

Linear waves 2nd -Order Stokes 5th -Order Stokes

0.8 0.003 0.01 0.03 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.09 0.10
1.2 0.003 0.01 0.03 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.09 0.10
1.6 ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
1.8 ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
2.0 ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
2.3 ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
2.8 ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
3.37 0.003 0.01 0.03 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.09 0.10

Fig. 9. Variation of the inline wave forces on tandem cylinders with center-to-center distance for different wave steepnesses: (a) cylinder 1 for H /L = 0.003, (b) cylinder 2 for
H /L = 0.003, (c) cylinder 1 for H /L = 0.05, (d) cylinder 2 for H /L = 0.05, (e) cylinder 1 for H /L = 0.1, and (f) cylinder 2 for H /L = 0.1.

H /L = 0.003, 4% for H /L = 0.05 and 2.5% for H /L = 0.1. It is observed and it does not influence the wave forces on the cylinders
that the Bessel wave-like variation of the wave forces with the anymore.
center-to-center distance is damped out with increasing incident Wave gages are placed in front ( F1, F2), behind ( B1, B2), beside
wave steepness for both cylinders. Even though, the analytically each of the cylinders (C1, C2) and at the midpoint between the two
predicted wave force on cylinder 1 matches the computed wave cylinders ( C0) at locations shown in Fig. 11 for H /L = 0.003 and
force at S ¼3.37 m for H /L = 0.05 in Fig. 9c and S¼2.3 m, S¼ 3.37 m H /L = 0.1 with S¼ 0.8 m. In the case of low steepness incident
for H /L = 0.1 in Fig. 9e, the wave force variation with S is clearly waves of H /L = 0.003, the variation of the free surface elevation is
different. sinusoidal around both the cylinders in Fig. 12(a) and (b). It is
The variation of the wave forces on the two cylinders for dif- observed that the crest height is increased in front of the cylinders
ferent center-to-center distances S at various incident wave due to the incident wave interaction with the cylinders ( F1, F2)
steepnesses H /L is presented in Fig. 10. It is seen that the wave and due to the superposing of the incident waves and the reflected
forces on both cylinders match the analytical prediction at lower waves behind the cylinder ( B1). The computed free surface ele-
H /L = 0.003 and 0.01. On increasing the wave steepness, the vations at B1, F2 and C0 have the same amplitude and phase,
computed wave forces gradually deviate from the analytical pre- implying uniform heave motion of the water along the line joining
diction. The computed forces are lower than the predictions from the centers of the two cylinders.
the analytical formula. The computed wave force on cylinder 1 at In the case of high steepness incident waves of H /L = 0.1, the
S ¼0.8 m for H /L = 0.1 is 30% lower than the analytical prediction incident waveform is asymmetrical with shallow troughs and
and 35% lower on cylinder 2 (Fig. 10a). It is also observed that at a sharp crests in Fig.12(c) and (d), characteristic of fifth-order Stokes
center-to-center distance of S = 3.37 m (Fig. 10h), the wave forces waves. The waveform computed at C1 shows increased asymmetry
on both the cylinders are almost equal. At this point, the effect of compared to the incident waves. This is attributed to the interac-
diffraction in between the two cylinders is reduced significantly tion of the incident waves with the out of phase reflected waves
 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748 745

Fig. 10. Variation of the inline wave forces on tandem cylinders with wave steepness for different center-to-center distances: (a)S ¼0.8 m, (b) S¼ 1.2 m, (c) S ¼1.6 m, (d)
S ¼1.8 m, (e) S ¼ 2.0 m, (f) S ¼2.3 m, (g) S¼ 2.8 m, and (h) S¼ 3.37 m.

S front of the cylinders is similar for both high and low steepness
waves. This is clearly seen in the case of the downstream cylinder
C1 C2 2, where the relative crest height in front of the cylinder looks
similar in Fig. 12(b) and (d) but the waveform is highly asymme-
F1 B1 C0 F2 B2 trical for H /L = 0.1. Also, the free surface elevation is seen to
continuously increase as the wave propagates away from cylinder
Cylinder 1 Cylinder 2 1 and towards cylinder 2. This large variation is not seen for the
low steepness waves, where the free surface elevation behind
Fig. 11. Schematic diagram of the domain around the two tandem cylinders cylinder 1, in front of cylinder 2 and at the midpoint between the
showing wave gage locations.
two cylinders is seen to be the same. A uniform heave motion of
the water is observed along the line joining the centers of the
from the cylinder. Wage gages B1, C0 and F2 show a continuously cylinders for low steepness waves and this is absent in the case of
increasing crest elevation as the wave propagates away from high steepness waves. These changes seen in the wave interaction
cylinder 1 and towards cylinder 2, due to the strong diffraction with a pair of tandem cylinders for incident waves of low and high
regime between the two cylinders. The crest elevation then steepness result in different flow regimes in the two cases. This
reduces at C2 and B2, as the wave propagates around cylinder 2. justifies the large deviation observed in the calculated wave force
Also, the free surface elevations at B1, C0 and F2 are slightly out of compared to the analytical predictions for high wave steepnesses.
phase and have different amplitudes signifying a complex wave In order to obtain further clarity on the wave field around the
diffraction regime in the region between the cylinders. two tandem cylinders with S ¼0.8 m, the diffraction patterns
Several differences are observed between the interaction of low around the cylinders for H /L = 0.003 and H /L = 0.1 are studied. The
and high steepness waves with a pair of tandem cylinders. The free surface elevation around the cylinders in the numerical wave
incident high steepness fifth-order waves are asymmetrical by tank for H /L = 0.003 over one wave period is presented in Fig. 13.
nature with a shallow trough and a sharp crest. This characteristic The increase in the free surface elevation when the crest is inci-
of the waves is magnified as it interacts with the large cylinders dent on cylinder 1 is seen in Fig. 13(a) and Fig. 13(b) shows the
and the waveform becomes more asymmetrical. This is in contrast change in the wavefront due to wave diffraction around cylinder 1.
to the interaction of the low steepness linear waves, where the The decrease in the free surface elevation as the wave travels
waveforms remain sinusoidal. The relative crest height η/ηc in around the upstream half of cylinder 1 is seen in Fig. 13(c).
i
746 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748

Fig. 12. Relative free surface elevations around two cylinders placed in tandem with S = 0.8 m for incident waves of low and high steepnesses: (a) cylinder 1: H/L ¼ 0.003,
(b) cylinder 2: H/L ¼ 0.003, (c) cylinder 1: H/L ¼0.1, and (d) cylinder 2: H/L ¼ 0.1.

Fig. 13. Free surface elevation in a part of the domain around the cylinders with S = 0.8 m for H/L ¼0.003: (a) t /T = 32.2, (b) t /T = 32.4 , (c) t /T = 32.6 , (c) t /T = 32.8,
(d) t /T = 33.0 , and (e) t /T = 33.2.

Fig. 13(d) shows the increase in the free surface elevation as the contours in all the figures is the region with the uniform heave
crest is incident on cylinder 2 and reduced free surface elevations motion of the free surface.
are seen in behind cylinder 2 in Fig. (13) (e) and (f). The region Fig. 14 shows the variation of the free surface elevation around
between the two cylinders with equal free surface elevation the two tandem cylinders with S ¼0.8 m for H /L = 0.1 over one
wave period. The increase in the free surface elevation in front of
 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748 747

Fig. 14. Free surface elevation in a part of the domain around the cylinders with S = 0.8 m for H /L = 0.1. (a) t/T = 32.2, (b) t/T = 32.4, (c) t/T = 32.6, (d) t/T = 32.8, (e) t/T = 33.0
and (f) t/T = 33.2.

the cylinder and the formation of distinct reflected waves are seen results that the interaction of high steepness waves is different
in Fig. 14a and b. The incident and reflected waves meet behind from low steepness waves due to the strong diffraction pattern
cylinder 1 in Fig. 14c and the intersection of two semi-circular and the transformation of the high steepness waves. The non-
waves is seen. The constructive interference of the two semi-cir- linear wave interaction in the case of high steepness waves is not
cular waves in the region between the two cylinders leads to a accounted for in the analytical formulae based on potential theory.
continuous increase in the free surface elevation around the line This results in the difference between the computed wave forces
joining the centers of the cylinders in Fig. 14d. The resulting large on the cylinders compared to those predicted by the analytical
free surface elevation in front of cylinder 2 is also seen in the formulae.
figure. Fig. 14e shows the reflected waves in between the cylinders
over the trough of the incident wave. The circular diffracted waves
formed in the wave tank around the two cylinders are seen in 5. Conclusions
Fig. 14f.
The free surface elevation contours around the tandem cylin- The calculation of wave forces on a single cylinder using the
ders in the simulations with a low wave steepness of H /L = 0.003 open source CFD model REEF3D is validated by comparison of
and a high wave steepness of H /L = 0.1 show that the wave regime experimental data for wave forces, wave elevation around the
is different in the two cases. The incident straight wavefronts cylinder and water particle velocity with the computed results
transform to a bent wavefront due to diffraction in the case of low from the numerical wave tank. Simulations are carried out to
steepness waves. In the case of the high steepness waves, forma- study the wave interaction with a large cylinder for different wave
tion of several semi-circular diffracted wavefronts are seen in steepnesses. The numerically calculated wave forces match the
addition to the bending of the incident wavefront. A uniform predictions by MacCamy–Fuchs equation for low wave steep-
heave motion of the free surface is seen for the waves of low nesses. Whereas for higher wave steepnesses, the computed wave
steepness in the region between the two cylinders. In the case of forces are lower than the predictions by the equation. The wave
the high steepness waves, distinct semi-circular diffracted waves elevation around the cylinder is investigated and the evolution of
interfere constructively in the region between the two cylinders. an asymmetrical waveform is seen in the case of high steepness
The large free surface elevation is concentrated around the line waves, whereas low steepness waves maintain their symmetrical
joining the centers of the two cylinders. It is seen in the numerical sinusoidal form. The difference in the wave phase in front, beside
748 A. Kamath et al. / Ocean Engineering 108 (2015) 738–748

and behind the cylinder suggests a deceleration of water particles Science and Technology (NTNU) provided by The Norwegian
around the downstream half of the cylinder in the case of high Metacenter for Computational Science (NOTUR, Project no.
steepness waves. NN2620K), http://www.notur.no
Further, simulations with a pair of large tandem cylinders are
carried out with different incident wave steepnesses and center-
to-center distances between the two cylinders. The computed References
wave forces are compared with the predictions from an analytical
formula based on potential theory. It is observed that the com- Alagan Chella, M., Bihs, H., Myrhaug, D., Muskulus, M., 2015. Breaking character-
puted wave forces match the predicted wave forces for lower wave istics and geometric properties of spilling breakers over slopes. Coast. Eng. 95,
4–19.
steepnesses. The computed wave forces are lower than the ana- Berthelsen, P.A., Faltinsen, O.M., 2008. A local directional ghost cell approach for
lytically predicted wave forces for higher wave steepness, with incompressible viscous flow problems with irregular boundaries. J. Comput.
about a 35% lower force for the highest wave steepness simulated Phys. 227, 4354–4397.
Boccotti, P., Arena, F., Fiamma, V., Barbaro, G., 2012. Field experiment on random
in the study. The analytical formulae predict a linear increase in wave forces acting on vertical cylinders. Probab. Eng. Mech. 28, 39–51.
the wave force with an increase in the incident wave height, for a Boccotti, P., Arena, F., Fiamma, V., Romolo, A., 2013. Two small-scale field experi-
given cylinder diameter and incident wavelength. The numerical ments on the effectiveness of Morison's equation. Ocean Eng. 57, 141–149.
Boo, S.Y., 2002. Linear and nonlinear irregular waves and forces in a numerical
results show that due to the wave transformation and the result-
wave tank. Ocean Eng. 29, 475–493.
ing asymmetrical nature of the higher steepness waves, the Chakrabarti, S.K., Tam, W.A., 1973. Gross and local wave loads on a large vertical
computed wave forces on the cylinders from these waves are cylinder – theory and experiment. In: Proceedings of Offshore Technology
lower than the predictions based on potential theory. The pre- Conference, Dallas, USA.
Chorin, A., 1968. Numerical solution of the Navier–Stokes equations. Math. Comput.
dictions from the CFD model at the scales considered in these 22, 745–762.
studies are good and provide insight into the interaction between Durbin, P.A., 2009. Limiters and wall treatments in applied turbulence modeling.
two relatively closely spaced cylinders. In the case of longer arrays Fluid Dyn. Res. 41, 1–18.
Ferrant, P., 1995. Time domain computation of nonlinear diffraction loads upon
of cylinders additional resonant effects such as wave near-trapping three dimensional floating bodies. In: Proceedings of 5th International Offshore
can occur, which have not been studied in this paper. and Polar Engineering Conference, The Hague, The Netherlands.
The diffraction patterns around tandem cylinders at different Isaacson, M., 1979. Wave Induced Forces in the Diffraction Regime. Pitman
Advanced Publishing Program, London, England.
wave steepnesses and the wave elevation around the tandem Jacobsen, N.G., Fuhrman, D.R., Fredsøe, J., 2012. A wave generation toolbox for the
cylinders are also studied. The evolution of semi-circular diffracted open-source CFD library: OpenFOAM. Int. J. Numer. Methods Fluids 70,
waves are seen in the case of high steepness waves, which meet on 1073–1088.
Jiang, G.S., Peng, D., 2000. Weighted ENO schemes for Hamilton–Jacobi equations.
the downstream side of the first cylinder. Whereas, in the case of SIAM J. Sci. Comput. 21, 2126–2143.
low steepness waves, the wavefront is only bent as a result of wave Jiang, G.S., Shu, C.W., 1996. Efficient implementation of weighted ENO schemes. J.
diffraction. A uniform heave motion of the free surface elevation is Comput. Phys. 126, 202–228.
Larsen, J., Dancy, H., 1983. Open boundaries in short wave simulations—a new
observed in the region in between the cylinders in the case of low
approach. Coast. Eng. 7, 285–297.
steepness waves. The complex diffraction regime in the case of Lighthill, J., 1979. Waves and hydrodynamic loading. In: Proceedings of 2nd Inter-
high steepness with clearly formed semi-circular diffracted waves national Conference on Behaviour of Offshore Structures, London, England.
results in an increasing free surface elevation as the wave crest Linton, C.M., Evans, D.V., 1990. The interaction of waves with arrays of vertical
circular cylinders. J. Fluid Mech. 215, 549–569.
propagates away from the upstream cylinder and towards the MacCamy, R., Fuchs, R., 1954. Wave Forces on Piles: A Diffraction Theory. U.S. Army
downstream cylinder. Corps of Engineers Beach Erosion Board, Technical Memorandum no. 69,
Thus, clear differences are seen between the interaction of low Washington DC, USA.
Malenica, S., Taylor, R.E., Huang, J.B., 1999. Second-order water wave diffraction by
and high steepness waves with large cylinders. In the case of a an array of vertical cylinders. J. Fluid Mech. 390, 349–373.
single large cylinder, the asymmetry of the steep incident waves McIver, P., Evans, D.V., 1984. Approximation of wave forces on cylinder arrays. Appl.
results in a different diffraction regime, which results in lower Ocean Res. 6, 101–107.
Mo, W., Irschik, K., Oumeraci, H., Liu, P., 2007. A 3D numerical model for computing
forces on the cylinders than predicted by linear potential theory. non-breaking wave forces on slender piles. J. Eng. Math. 58, 19–30.
For a pair of tandem cylinders, the center-to-center-distance Molin, B., 1979. Second order diffraction loads upon three-dimensional bodies.
between the cylinders contributes to further change the diffraction Appl. Ocean Res. 1, 197–202.
Naot, D., Rodi, W., 1982. Calculation of secondary currents in channel flow. J.
regime, in addition to the effects due to wave asymmetry. The
Hydraul. Div., ASCE 108, 948–968.
evolution of distinct semi-circular reflected waves around the Ohkusu, M., 1974. Hydrodynamic forces on multiple cylinders in waves. In: Pro-
cylinders in the case of high incident wave steepness has a con- ceedings of International Symposium on Dynamics of Marine Vehicles and
Structures in Waves, London, England, pp. 107–112.
sequence on objects close to the cylinders. The current results
Osher, S., Sethian, J.A., 1988. Fronts propagating with curvature-dependent speed:
show a smooth deviation from the linear results as the incident algorithms based on Hamilton–Jacobi formulations. J. Comput. Phys. 79, 12–49.
wave steepness is increased. Further work is needed to determine Peng, D., Merriman, B., Osher, S., Zhao, H., Kang, M., 1999. A PDE-based fast local
the transition of the wave force regime from non-breaking wave level set method. J. Comput. Phys. 155, 410–438.
Shu, C.W., Osher, S., 1988. Efficient implementation of essentially non-oscillatory
forces where the wave forces vary at a frequency similar to the shock capturing schemes. J. Comput. Phys. 77, 439–471.
incident wave to breaking wave forces which are impulsive in Song, H., Tao, L., Chakrabarti, S., 2010. Modelling of water wave interaction with
nature with a sharp peak over a period much shorter than the multiple cylinders of arbitrary shape. J. Comput. Phys. 229, 1498–1513.
Spring, B., Monkmeyer, P.L., 1974. Interaction of plane waves with vertical cylinders.
incident wave period. Application of the numerical model to In: Proceedings of International Conference on Coastal Engineering, ASCE,
determine random wave forces can also be explored. Copenhagen, pp. 1828–1847.
Twersky, V., 1952. Multiple scattering of radiation by an arbitrary configuration of
parallel cylinders. J. Acoust. Soc. Am. 24, 42–46.
van der Vorst, H., 1992. BiCGStab: a fast and smoothly converging variant of Bi-CG
Acknowledgement for the solution of nonsymmetric linear systems. SIAM J. Sci. Stat. Comput. 13,
631–644.
Wilcox, D.C., 1994. Turbulence Modeling for CFD. DCW Industries Inc., La Canada,
This study has been carried out under the OWCBW project (No. California.
217622/E20) and the authors are grateful to the grants provided by
the Research Council of Norway. This research was supported in
part with computational resources at the Norwegian University of