You are on page 1of 10

CFD Simulation of a Moored Floating

Wave Energy Converter

Abstract—The paper presents incompressible Navier-Stokes is clear that the mooring configuration need to be considered
simulations of the dynamics of a floating wave energy converter on an early stage in the product development and integrated
(WEC) coupled to a high-order finite element solver for cable in the dynamic analysis of the WEC operation. Both [5] and
dynamics. The coupled model has very few limiting assumptions [7] stress the importance of a coupled motion analysis of the
and is capable of capturing the effects of breaking waves, green floating body and its mooring system, i.e. that the motion of
water loads on the WEC as well as non-linear mooring forces
the floater depends on the time- and position dependent cable
and snap loads, all of which are crucial for correct estimates
of the extreme loads acting on the system in violent seas. The tension.
cable dynamics model has been developed as a stand-alone library
that can be coupled to any body motion solver. In this study the There are two main aspects of wave power mooring system
open-source CFD package OpenFOAM R has been employed. design which have to be treated with extra care due to the large
Preliminary test cases using incident regular Stoke’s 5th order motions in relatively shallow water. One is the peak load in
waves are presented, both for wave heights corresponding to the mooring cables in survival sea states and the other is the
operational conditions of the WEC as for a more severe condition life time prediction of the mooring components in the system.
in survival mode. It is illustrated that the coupled model is able to In combination they form the basis of a design procedure.
capture the complicated force propagation in the mooring cables. Connected to both of these problems is the fast propagation of
Keywords—wave energy converter, mooring cable dynamics, tension loads along the axial direction of the cable. For extreme
catenary, Navier-Stokes equations, coupled analysis, time-domain loads, occurring in harsh weather conditions where the velocity
analysis, survival of the mooring attachment point to the WEC is large, there is
an increased risk of slack in the cable. The ensuing snap load
shock wave may then be of considerable magnitude. For the
I. I NTRODUCTION purpose of fatigue prediction it is also important to correctly
The accurate modelling of extreme loads acting on floating estimate the number of load cycles acting on the material.
wave energy converters (WECs) and on their mooring system These fast, transient loads are in this case an important factor
in violent seas is very important for the development of the to consider.
wave power industry. As the cost of the mooring system can
be substantial [6], there is a need for accurate extreme load As the loads on the mooring system are directly coupled
prediction tools so that a more cost effective mooring design to the resulting motion of the WEC, being governed by the
can be made. However, the need for accurate methods of motion of the cable attachment point, an accurate model for
solving the hydrodynamic problem of floating WECs is also cable dynamics is not sufficient to form a solid base for
great for operational conditions. In marine structural design mooring cable design. It is also of great importance to have
the motion of the floating body and the forces acting on it an equally accurate model for computing the motion of the
is the key result of the design calculations. These are then WEC as the combined response due to loads from waves,
compared against some design rules for a safe installation and tides, currents, winds, moorings and power take off (PTO).
management of the structure. In this respect, a conservative In survival sea states the linearisation needed for a frequency
calculation approach for the loads is inherently considered to domain methodology are missing too much of the dynamics
be to some extent necessary since they introduce additional in the WEC – wave interaction and also in the responsive
safety margins for the investment. This is not the case for mooring force. Using potential theory to solve the radiation-
WECs in operational conditions as the conservative approach diffraction problem gives good results for large waves, but is
renders non-conservative estimated values of the extracted unable to capture the effect of green water loads and breaking
power. For this reason it is very important that the simulation waves. A more accurate method, still neglecting viscosity, is to
models used for the predicted power output are accurate and use fully nonlinear potential theory for modelling the motions
reliable. of floating bodies, see e.g. [14]. The use of computational
fluid dynamics (CFD) is becoming more and more used in
The special role of the moorings within floating WECs sea-keeping, see e.g. the discussions in [11], [19]. Needless
has been recognised by several authors, e.g. [9] and [6], and it to say, viscous CFD is the most accurate – and also the
most computationally expensive – approach to model the mooring system dynamics – coupled at each time step of
environmental loads acting on the floater [12], [13]. CFD the hydrodynamics solver through an Automated Program
captures impact loads from breaking waves and green water Interface (API).
loads due to waves overtopping the floater. It should be stressed
that CFD has good overall accuracy including around the
resonance frequency, and typically floating WECs are designed A. Free-Surface Navier-Stokes Equations
to be in resonance. There are a few CFD studies of moored The incompressible Navier-Stokes equations are solved
floating bodies, and often the mooring is modelled by means of using the finite volume method with a dynamic Euler mesh
a spring analogy, as in [21]. An example of a coupled mooring which is fixed to and moves with, the moving WEC. Using
CFD study is found in [2]. +WEC_019 the single fluid (mixture) assumption the two-phase Navier-
The aim of the paper is to present a methodology to solve Stokes equations can be written as
for the coupled motion of a generic moored WEC in any kind
of sea state. A key point is to be able to provide accurate ∇ · u = 0, (1)
predictions of the loads in the mooring cables in order to ∂
(ρu) + ∇ · (ρ (u − ug ) u) = −∇p + ∇ · S + ρfb , (2)
evaluate the required ultimate and fatigue limit states. In order ∂t
to handle breaking waves and green-water loads the hydro-
where eq. (1) is the continuity equation for incompressible
dynamics is simulated by means of a free surface capturing
flow and eq. (2) is the momentum equation. Here u is the
incompressible Navier-Stokes model including viscous effects.
fluid velocity, ug the grid velocity, p is the pressure, ρ is the
The mooring cable dynamics is modelled assuming the cable
density of the mixture. The viscous stress tensor is denoted
to be perfectly elastic while neglecting bending and torsion of
S = 2µD in which µ is the viscosity of the mixture and D
the cable. This is a frequently used approximation for cable
the rate-of-strain tensor. The body force made up of gravity
dynamics [1], [15], but it should be noted that the absence of
and surface tension is denoted by fb .
bending stiffness make the cable dynamics problem ill-posed
in slack condition [20]. +WEC_022 The two-phase problem is treated using the volume of fluid
The paper is organized as follows. Section II contains (VoF) formulation. In the VoF approach the volume-fraction
an overall description of the methods used, focusing on α ∈ [0, 1], where 0 represents a pure air phase and 1 represents
the governing equations rather than the numerical solution pure water phase, is used used to track the free-surface. The
techniques, and is divided into four sub parts. The first two mixture properties are computed based on the volume-fraction:
sub parts give a brief description of the techniques used for ρ = αρw + (1 + α) ρa , (3)
modelling the hydrodynamics of the fluid and the rigid body
six-degrees of freedom (6DoF) dynamics. These parts are µ = αµw + (1 + α) µa , (4)
both handled by the existing two-phase Navier-Stokes solver
where index w indicate water and a indicate air. The volume
with mesh motion, interDyMFoam, available in the open-
fraction is updated by solving the transport equation
source CFD toolbox OpenFOAM R [16], [22]. OpenFOAM
uses the finite volume technique to discretize the equations. +OP_001 ∂α
In addition, wave generation and absorption is performed + ∇ · (α (u − ug )) = 0 . (5)
∂t
through the waves2Foam package [8]. The third part describes
the in-house mooring cable solver MOODY [17]. MOODY The VoF method is known to diffuse the wave interface, but
solves the resulting second-order non-linear wave equation for in interDyMFoam there is a possibility to keep the wave sharp
the cable position using a high-order discontinuous Galerkin by means of artificial compression of the interface, see [18].
method. The fourth part is devoted to how the coupling and
In the interDyMFoam solver the viscous terms can be
interaction between the fluid/6DoF solver and MOODY is
modelled using either the Reynolds-averaged Navier-Stokes
achieved. The coupled solvers are then applied to test cases
(RANS) or the large eddy simulation (LES) approach. In this
consisting of a spread moored cylindrical WEC subjected to
study, the RANS approach with the standard k −  turbulence
regular waves. The spread mooring system is made up of four
model has been used. +OP_002
symmetrically distributed catenary chains. In section III the
WEC geometry, mooring system set-up, PTO as well as the The governing equations are solved using the cell-centred
spatial and temporal discretization of the fluid domain and finite volume method on unstructured polyhedral meshes [22].
cable lines, are discussed. Results are presented in section IV There are a large number of discretization schemes that can
for mild wave conditions and in section V for a harsh wave be chosen at run-time. Here a second-order linear scheme for
condition. Please note that the results presented are preliminary the gradient and laplacian operators, and a second-order TVD
and are only to be viewed as illustrations of the abilities of scheme for the divergence operators in order to add robustness
the coupled method. In depth simulations are still required to to the simulations are used. Eqs. (1) and (2) are solved using
fully validate the proposed methodology. Finally, in section VI the PIMPLE algorithm.
the study is summarised and discussed.
Waves are generated using the waves2Foam package [8]
II. M ETHOD using separate relaxation zones for wave generation and for
wave absorption. Waves2Foam supports a number of regular
The motions of the moored WEC are calculated using wave theories including Stoke’s 5th order and cnoidal waves as
two solvers – the interDyMFoam solver in OpenFOAM for well as irregular wave generation based on PM or JONSWAP
the hydrodynamics and the MOODY cable solver for the spectra.
B. Rigid body dynamics stiffness of the cable is denoted by EA0 . The right hand side,
f = f1 + f2 + f3 + f4 , contains the external forces acting on the
The rigid body dynamics of the WEC are solved using the
cable segment. Here f1 is the sum of gravity and buoyancy, f2
rigid body displacement solver implemented in OpenFOAM
is the added mass force, and the tangential and normal drag
for motions in six degrees of freedom. It solves for the fluid
forces are given as f3 and f4 . Introducing the unit tangential
force Fp and torque Mp acting on the body by integrating the
vector t, defined as
fluid pressure, p, over its entire surface area S.  
ZZ ∂r ∂r ∂r
t= = (1 + ) , (13)
Fp = pn̂ dS (6) ∂s ∂s ∂s
ZSZ and the decomposition notations xt = (x · t) t and xn = x−xt
Mp = rCS × pn̂ dS (7) for the tangential and normal components of a vector x, the
expressions for the external forces on the cable segment read
S
γe t2c
where rCS represents the position vector from the center of f1 = − g, (14)
mass of the WEC to each point at its surface. The pressure γ0 Lc
is integrated in the global coordinate system and the resulting Aρw
f2 = CM arel,n (1 + ) , (15)
moment is then transformed to the the local body frame before γ0
it is added to the total torque together with the contribution 1 ρw dLc
from the restraints, FR . The total force and moment, Ftot and f3 = CDt |vrel,t | vrel,t (1 + ) , (16)
2 γ0
Mtot acting on the body is then expressed as 1 ρw dLc
X (j) f4 = CDn |vrel,n | vrel,n (1 + ) . (17)
Ftot = Fp + FR , (8) 2 γ0
j Here γe = ((ρc − ρw ) /ρc ) γ0 is the effective mass per unit
 
length of the submerged cable and ρc and ρw are the cable
(j) (j) (j)
X
Mtot = Q Mp + rCR × FR + MR  , (9) and fluid densities. The terms CM , CDt and CDn denote
j the hydrodynamic coefficients of added mass, tangential drag
and normal drag forces respectively. The last three forces are
where Q is the rotation matrix from global to local frame functions of the relative velocity and relative acceleration of
(j) (j)
of reference, FR and MR is the force and moment from the water with respect to the mooring cable, vrel and arel ,
(j)
restraint condition j and rCR is the global frame position given by
vector of the restraint point of action on the body relative ∂r
to the centre of gravity. In this paper, there are two types of vrel = vw − , (18)
restraints implemented. The linear PTO-restraint, implemented ∂t
as FPTO = Bii vi for i = [1, 2, 3] and MPTO = Bii Qvi for ∂2r
arel = aw − 2 . (19)
i = [4, 5, 6], both acting in the centre of gravity. Here vi is the ∂t
velocity in direction i, index 1 to 3 applies to translational
motion and 4 to 6 applies to rotational motion. The PTO A ground model consisting of a bilinear spring and damper
coefficients Bii used in the simulations are given in section combination with tangential friction is implemented in this
III-D. A mooring cable restraint, described in section II-D is paper. It is a simplified approach, and it is seen in the results
used to couple the motion of the WEC to the mooring forces. that this approach gives rise to numerical disturbances in the
element containing the point of contact between the cable and
the ground. ++Mooring_002
C. Cable Dynamics
The mooring cable dynamics problem is solved using
Assuming the mooring cables to be completely flexible, the
the in-house code MOODY [17]. MOODY is based on a
equation of motion for the cable position r, expressed along
modal high-order finite element model [10] using a modified
the curvilinear abscissa s of the unstretched cable, reduces to
version of the local discontinuous Galerkin (LDG) method [4].
a coupled set of one-dimensional, non-linear wave equations.
MOODY has been shown to exhibit exponential convergence,
Given below, in its non-dimensional form, is the equation of
giving that engineering accuracy can be obtained with just a
motion for a linear elastic cable material [1], [15]:
few elements of high order. The hallmark of the discontinuous
∂2r
 
∂ 2 ∂r Galerkin method is that the solution is allowed to be discon-
− c =f, (10) tinuous over element boundaries – thus the discrete space can
∂t2 ∂s ∂s
be claimed to be better suited for shocks such as snap loads.
t2 EA0 
c2 = c2 , (11) On the other hand, it is well known that high-order scheme
L γ 1+ yield oscillatory results in the case of discontinuities in the
c 0
∂r solution. Nevertheless, most of the problems can be overcome
 = − 1 . (12)
∂s by careful filtering etc. In MOODY this problem is not yet
fully solved, as will be evident from the results.
Here r = (rx , ry , rz ) is the cable position vector, γ0 is the
mass per meter cable, c represents the non-linear celerity of the
D. Coupling Scheme
cable and  is the cable tangential strain. Further, notations Lc
and tc are characteristic length and time scales of the problem The coupling between the two codes is made through
used to non-dimensionalise r, s and time t. The linear, axial the 6DoF rigid body motion solver within OpenFOAM and
Fig. 2. The initial set-up of the floating body and the mooring configuration.
Note the numbering of the mooring cables and that this is the starting position
with a draught of T = 5 m.

Fig. 1. Showing a visualisation of the structure of the mesh, the orientation


of the global coordinate axes and the relaxation zone layout. Shown in tan III. C ASE D ESCRIPTION
color is the wave generating inlet zone and shown in red is the damping outlet.
The light green in the center is pure computational domain.
The coupled solver has been tested on the same generic
WEC that is used in [5], represented by a floating cylinder
with spread catenary mooring.
the possibility to add restraints therein. For this purpose a
new restraint was developed that invokes a Matlab engine
child process and uses it to launch the mooring solver. For A. Geometry
each time that the restraint is called, the mooring attachment
point positions are sent to Matlab and the mooring solver The WEC is represented by a floating cylinder with height
computes the cable dynamics by advancing in time from the h = 10 m, diameter d = 5 m and draft T = 5 m in water
time of the last call and saved dynamic state, to the present of density ρw = 1025 kg/m3 . Thus the mass of the cylinder
OpenFOAM time. The instantaneous end point forces in the is approximately 100 tonnes. The moment of inertia is that
cables connected to the WEC are then returned to the rigid of a cylindrical shell where half of the mass is located in the
body motion solver’s restraint and the corresponding total net vertical walls and half is located in the bottom plate, to render
force and local body frame moment from the moorings are a more stable roll and pitch response. The lid is assumed to
computed. be massless. Initially the cylinder is at rest in the center of the
time step difference between openFOAM and MOODY computational domain with a center of mass position at rC =
As the order of magnitude of the time step size of Open- [0, 0, −2.42] m as can be seen in Fig. 1. The computational
Foam is ∆tOF ∝ 1E-2 s and the time step needed by the domain is itself contained to x ∈ [−80, 80] m, y ∈ [−40, 40]]
mooring cable solver is in the order of ∆tM ∝ 1e-4 s, there m and z ∈ [−50, 25]m with the still water level being at z = 0.
is a time scale difference that needs to be taken into account. Also seen in Fig. 1 is the layout of the relaxation zones used
The computational time would be significantly increased if for wave generation and absorption.
OpenFOAM was to be limited to the time step restriction of
the mooring system and therefore this option is considered to
be impractical and expensive without any large improvement B. Mooring System +WEC_024
in accuracy. Instead, the software handles the different time
scales of the two problems through a sub-stepping scheme To maintain comparability with [5], a mooring design of
implemented in MOODY. type C studied therein has been used. The chosen mooring
system consists of four catenary chains with mass γ0 = 61
kg/m, unstretched length L = 100 m and axial stiffness EA =
Let tkOF and tk−1
OF be the times of steps k and k − 1 in the 100 MN. Further the hydrodynamic coefficients used for the
OpenFOAM simulation, and let tM ∈ [tk−1 k
OF , tOF ] denote the cable type are CDn = 2.5, CDt = 0.5 and CM = 3.8. In Fig. 2
sub time used by the mooring solver. The mooring attachment it is seen that the moorings are attached symmetrically around
points of all cables attached to the body sent to MOODY the outer rim of the bottom plate so that the cables initially
are used as input to the time dependent Dirichlet boundary are aligned with the x- and y-coordinate axes. Each cable was
conditions, gD , for the cable dynamics problem. For each call anchored at the sea floor a horizontal distance ∆a = 74.8
to the mooring solver from the OpenFoam environment, a m from the attachment point on the device, leaving the total
linear interpolation over the OpenFOAM time step is made footprint of the mooring system to be contained within a circle
to allow for a smooth evolution of the boundary conditions of diameter 154.6 m.
between the last OpenFoam call and the present time:
The mooring system forces were added to the free floating
k−1 body over a 5 s linear ramp time. The equilibrium condition
tk−1
 
gD tkOF − tM + k
gD tM − OF of the moored body is 0.63 m below that of the free floating
gD = (20)
∆tOF case.
D. Power Take-Off
Some trials have been made with a generic, linear Power
Take-Off (PTO). The damping coefficient matrix for the PTO
has been taken to be the same as is used in [5] for heave,
surge and pitch. For the sway and roll directions, the same
damping as for surge and pitch is used, and the yaw direction
is undamped. The values presented in eq. (22) are in kNs/m
and kNms respectively.

 
100 0 0 0 0 0
 0 100 0 0 0 0 
0 0 15 0 0 0
 
BPTO = (22)
 
Fig. 3. The mesh surrounding the cylinder. There is a clear band of fine
cells surrounding the still water level and one can also see the extra cubic and  0 0 0 350 0 0 

more robust cells where the lid or bottom meets the vertical walls.
 0 0 0 0 350 0 
0 0 0 0 0 0

C. Mesh Description and Model Settings The instantaneous and mean power for motion direction i
over a time interval t ∈ [t1 , t2 ] is then, in that order, expressed
The fluid domain mesh is a structured hexahedral mesh
as
with a narrow band of high resolution in the z-direction
surrounding the still water level to better capture the motion Pi (t) = Bii vi (t)2 , (23)
of the generated waves. The structure of the global mesh can R t2
be seen in Fig. 1. The total number of cells in the mesh is Pi (t) dt
P i = t1 . (24)
roughly 2.7M. Around the WEC the resolution is finer in order t2 − t1
to better resolve the boundary layer. Seen in Fig. 3 is the
The peak power absorbed over a time interval is taken directly
mesh surrounding the cylinder. Worth to note here are the
as the maximum occurrence of Pi (t) within that interval.
fine cubic cells surrounding the corner of the body. As the
mesh is attached to the body, these cells are prone to high
skewness during large displacements of the body, particularly IV. R ESULTS IN REGULAR WAVES
during excessive pitch and roll motion. Therefore they have The WEC motions are computed for four different configu-
been given extra care in the mesh generation. rations, denoted as Free, Moored, PTO and MooredPTO. Free
Mesh deformation is spread across the domain using a is a configuration without any mooring cables or PTO restrain-
diffusive Laplacian solver for the mesh motion, with sliding ing the device, Moored and PTO has only mooring and PTO
mesh motion on the top, bottom and side patches of the restraints respectively and finally MooredPTO is the combined
domain to allow for large surge displacements with minimum case of a moored WEC with PTO. The configurations have
mesh skewness. The points on the inlet and outlet are held been tested in regular waves of period Tw = 5 s, for wave
fixed throughout the computations. heights H1 = 1 m and H2 = 2 m. It should be noted that
Tw = 5 is very close to the resonance period of the buoy, see
[5].
The different time scales of the problem described in
section II-D make it important to carefully set the time step A. Motion Response
size of the different solvers. When the maximum Courant
number in the CFD domain is relatively low, the adaptive The surge, pitch and heave motions of the cylinder moving
time stepping in interDyMFoam allows the time step size to in regular waves with height H1 and H2 for the different
be increased. In these cases a maximum allowed time step configurations are seen in Fig. 4. The lack of restoring force
of ∆tOF,max = 0.05 s is used in order to capture the fast from the moorings or damping from the PTO resulted in very
variations of the tension force in the mooring cables, as well large surge and pitch motions of the free floating device. The
as for maintaining stability in the time stepping scheme in results for the Free configuration have therefore for clarity
interDyMFoam. The maximum time step is therefore chosen been removed from the figures showing the surge and the pitch
so that it is smaller than half of the return period of a shock motion.
wave along the cable. This characteristic time in the mooring
cables is computed as tc = L/c = 0.078 s, which is the same It is clear from Fig. 4 that there is a large difference in
tc used in the formulation in section II-C. the response per meter wave height between the wave heights
H1 and H2 . The heave amplitude of the 2 m wave is distinctly
smaller per wave height than the 1 m equivalent for all cases
L studied. This is not predicted by the linear theory used by [5].
∆tOF,max < (21)
c It would be tempting to claim that the difference is caused by
viscous effects, but the influence of viscosity is most likely
The characteristic length, Lc , used is the length of one of small for these simulations of a relatively large buoy in small
the mooring lines, i.e. Lc = 100 m. In the cable solver each waves [3]. Remembering that the same mesh is used for both
of the cables are discretized into 10 elements of order eight. cases, the most probable cause of the small heave amplitude of
Surge motion in regular waves, H=1m Surge motion in regular waves, H=2m
5.0 5.0
PTO Moored MooredPTO PTO Moored MooredPTO
4.5 4.5

4.0 4.0
Surge motion per wave height [−]

Surge motion per wave height [−]


3.5 3.5

3.0 3.0

2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0

−0.5 −0.5
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
time [s] time [s]

(a) (b)

Heave motion in regular waves, H=1m Heave motion in regular waves, H=2m
1.6 1.6
PTO Moored MooredPTO Free PTO Moored MooredPTO Free
1.4 1.4
1.2 1.2
Heave motion per wave height [−]

Heave motion per wave height [−]


1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
−0.2 −0.2
−0.4 −0.4
−0.6 −0.6
−0.8 −0.8
−1.0 −1.0
−1.2 −1.2
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
time [s] time [s]

(c) (d)

Pitch motion in regular waves, H=1m Pitch motion in regular waves, H=2m
10.0 10.0
9.0 PTO Moored MooredPTO 9.0 PTO Moored MooredPTO
8.0 8.0
Pitch motion per wave height [deg/m]

Pitch motion per wave height [deg/m]

7.0 7.0
6.0 6.0
5.0 5.0
4.0 4.0
3.0 3.0
2.0 2.0
1.0 1.0
0.0 0.0
−1.0 −1.0
−2.0 −2.0
−3.0 −3.0
−4.0 −4.0
−5.0 −5.0
−6.0 −6.0
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
time [s] time [s]

(e) (f)
Fig. 4. The time history of the WEC surge (a)-(b), heave,(c)-(d), and pitch (e)-(f), response to regular waves of period Tw = 5 s for wave heights H1 = 1
m and H2 = 2 m. The responses are presented per m wave height, and the surge and heave responses are given as the displacement of the center of gravity
relative to its still water equilibrium position.
TABLE I. T HE RESULTING MOORING FORCES FOR THE TESTS WITH
Force in cable 1 [kN] 50 REGULAR WAVES . T HE TABLE IS BASED ON THE INTERVAL t ∈ [60, 100] S .

45
Cable no. 1 2 3 4 1 2 3 4
40
Moored H1 H2
35
Mean [kN] 34.1 33.2 32.3 33.2 36.4 32.9 30.2 32.9
30
Max. [kN] 39.9 37.6 36.5 37.6 48.8 40.2 36.6 40.2
25
Range [kN] 12.2 9.7 8.2 9.7 28.1 17.8 13.1 17.8
20
80 82 84 86 88 90 92 94 96 98 100 MooredPTO H1 H2
time [s] Mean [kN] 34.2 33.1 32.0 33.1 37.4 32.9 29.6 32.9
50 Max. [kN] 38.9 36.5 35.0 36.5 47.1 38.5 34.1 38.5
Force in cable 3 [kN]

45 Range [kN] 10.4 7.8 6.4 7.7 22.3 13.3 9.9 13.3
40
35
30 simulations, which can be seen in the general increase in
25 mean tension force between the two wave height simulations.
20
80 82 84 86 88 90 92 94 96 98 100
time [s]
C. Power Take-Off
Fig. 5. The tension force magnitude at the attachment point to the WEC In order to compare how the absorbed power is affected
in mooring cable no. 1 and 3 between 80 and 100 s of simulation. The time by the moorings and by the change in wave height, the mean
series is from the case Moored, H2 = 2 m.
and peak absorbed power in the interval 60-100 s of simulation
is calculated. These values are compiled in Table II together
the H2 case is simply the mesh being too coarse for resolving with the mean up-crossing heave response height for the same
the H2 problem. interval. Here the presented heave motion values are calculated
as the mean value of the difference between the maximum
The wave frequency oscillations in surge are very similar and minimum vertical position of the WEC, normalised by the
for the two wave heights studied but the slow varying surge incoming wave height. In the case of the free floating device,
response is more pronounced for the larger wave, which is to the heave response at t = 45 s have been used instead.
be expected as the drift force grows with the square of the
velocity for the same frequency. The clear decrease in heave response per m wave height,
as discussed in section IV-A, naturally has a great impact also
on the generated power. The velocity of the device scales
B. Mooring Forces
linearly with the displacement and the instantaneous power
The resulting tension force time history at the point con- is proportional to the square of the velocity, see eq. (23). The
nected to the WEC of cable 1 and 3 is shown in Fig. 5. deviations of the results from the linear theory presented in
The dominating wave period of excitation is clearly visible in [5] are large for all test cases and are more accentuated as the
the end point force time history, and the disturbances arising wave height increases.
from the cable hitting the sea bottom are most visible in the
TABLE II. T HE RESULTING MEAN HEAVE RESPONSE AND THE MEAN
troughs of the motion. In this context the importance of the AND PEAK EXTRACTED POWER IN HEAVE . T HE HEAVE RESPONSE IS GIVEN
choice of ground model parameters should be stressed. The PER M WAVE HEIGHT, AND THE POWER VALUES ARE GIVEN IN KW. A LL
appearance and smoothness of the end point mooring force VALUES ARE BASED ON THE INTERVAL t ∈ [60, 100] S WHEN POSSIBLE .
is very much governed by the modelling of the sea bottom,
PTO Moored MooredPTO Free
its stiffness and damping properties both in tangential friction
H1 H2 H1 H2 H1 H2 H1 H2
and in the normal direction. Here it should also be noted that
Heave /H [-] 1.6 1.1 1.2 0.9 1.0 0.7 2.4 1.6
the mooring cable model has no internal damping in itself and
Mean power [kW] 9.4 14.4 - - 3.0 7.8 - -
very little numerical diffusion from the formulation, rendering
Peak power [kW] 14.7 28.2 - - 5.9 15.2 - -
long-lived transients in the tension force time history.
The resulting mean and maximum mooring forces for
all cases tested are compiled in Table I together with the V. R ESULTS IN L ARGE WAVES
range (max-min) of the oscillating end point force. The
As an illustrative example, a large and steep Stoke’s 5th
values are calculated based on the last 40 s of simulation
order wave of height H = 15 m and period T = 12 s, is
where the solutions are closer to steady state. Looking at
generated at the inlet boundary. The response of the WEC to
table I, the highest loads are occurring in cable 1, the cable
this wave is tested for the moored device with the PTO turned
facing the waves, both in mean and maximum force and in
off as a simplified model of a survival mode of the device.
the range of the load. The range of the force in cable 1 is
Fig. 6 shows a snapshot of the simulation and the mooring
more than doubled between H1 and H2 for both Moored
system positions in the crest of the wave. The figure illustrates
and MooredPTO, which is somewhat surprising as the heave
a failure in the mooring system design as cable 1, facing the
motion increase between the two wave heights in Fig. 4 is
wave direction, is completely lifted from the sea-bed. For drag
much smaller. The pitch motion could contribute to this effect
embedded anchors this is a design criteria that equals to failure.
for the Moored response, however the MooredPTO response
per m wave height is also smaller for H2 than for H1 . One The mooring force response at the attachment point of
explanation is the large increase in surge offset between the cable 1 is shown in Fig. 7. In the figure the clearly visible
Force at attachment to the WEC, Cable no. 1
600

550

500

450

400

Tension force [kN]


350

300

250
Fig. 6. The solution state after 50 s of simulation. Note the design failure
200
due to completely lifted cable in mooring cable 1.
150

100
spikes are caused by an initial snap load that is subsequently 50
propagating back and forth in the cable. The period of the 0
spikes matches exactly the return period of the shock wave, 46 46.5 47 47.5 48 48.5 49 49.5 50 50.5
time [s]
being 2tc from eq. (21). It is important to stress that these
spikes do originate from a snap load and not from the cable
interacting with the ground, as the high frequency oscillations Fig. 7. Tension force in attachment point to the WEC of mooring cable 1.
visible in Fig. 5.

VI. D ISCUSSION AND F UTURE W ORK numerically smoother ground modelling is an area of ongoing
work with the cable code, as is an accurate handling of cable
A methodology for performing coupled analysis of floating slack. Figure 7 shows significant numerical noise from the
WECs was presented: to pair the free-surface Navier-Stokes slack period which is due to the complete loss of stiffness in
solver interDyMFoam – part of the OpenFOAM toolbox – with the system, rendering the mathematical problem ill-posed [20].
an in-house dynamic mooring cable solver MOODY through This can be avoided by the inclusion of bending and torsional
a simple API. In this approach all arising non-linearities, both stiffness [7].
in the hydrodynamic part as well as in the mooring dynamics
part, are included. The main advantage of performing coupled Future work will concentrate on a thorough validation
analysis using CFD is the completeness of the method and study looking specifically into spatial and temporal resolution
the very few underlying assumptions. The main disadvantage needed for different sea states, and later to investigate different
is the excessive computational effort needed. For the test mooring configurations for one device and the interaction
cases presented the wall clock time was in the order of 5 effects between several devices.
h per wave period using roughly 80 000 cells per core for Finally, it should be noted that MOODY does not rely
the moored device in with wave height H1 . For H2 , having on interDyMFoam for coupled analysis. MOODY has been
higher velocities and thereby also a smaller average time step developed as a stand-alone library so it can be coupled to any
size, the corresponding value was roughly 10 h. By increasing hydrodynamics 6DoF solver as long as the solver can provide
the mesh resolution, the computational time cost would grow the variables requested by the API.
accordingly. Thus coupled analysis using CFD is most suitable
for use in survival conditions where the time span of simulation
needed is limited and the non-linear effects such as green water ACKNOWLEDGMENT
and wave breaking are more likely to occur.
This work was funded by the Swedish collaboration plat-
Albeit no conclusion should be drawn from the test cases form Ocean Energy Centre (hosted by Chalmers University
presented – a proper validation and in-depth sensitivity analysis of Technology and supported by a grant from Region Västra
is required before that – they served to illustrate that the Götaland, the regional development agency of Västra Götaland
coupled approach can capture the complex interaction between in Western Sweden) and by the Portuguese Foundation for
the mooring system and the floating device. Science and Technology (FCT – Fundação para a Ciência e
Focusing on the mooring forces, the results both for the Tecnologia) through research grant SFRH/BD/62040/2009.
regular waves and the survival case indicate a need for high The simulations were performed on resources at Chalmers
resolution in both time and space to correctly capture the Centre for Computational Science and Engineering (C3SE)
perturbations in the tension load. In mooring models of low provided by the Swedish National Infrastructure for Comput-
resolution these fast transients tend to dissipate and may not ing (SNIC).
be seen at all in the results. The present numerical scheme
implemented in MOODY resolves the tension perturbations,
but not without the creation of some Gibbs-type oscillations. R EFERENCES
The presented results also show some numerical noise caused
[1] O.M. Aamo and T.I. Fossen, “Finite element modelling of mooring
by the ground model. The ground is modelled using a vertical lines,” Math. Comp. Sim., vol. 53, pp. 415–422, 2000.
load only on the nodes in contact with the ground. This will [2] S. Aliabadi, J. Abedi and B. Zellars, “Parallel finite element simulation
introduce a discontinuity in the touchdown element, which in of mooring forces on floating objects,” Int. J. Numer. Meth. Fluids, vol.
turn produces the noise in the simulation. More accurate and 41, pp. 809–822, 2003.
[3] M.A. Bhinder, A. Babarit, L. Gentaz and P. Ferrant, “Assessment of
viscous damping via 3D-CFD modelling of a floating wave energy
device,” in Proc. EWTEC2011, 2011.
[4] B. Cockburn and C.-W. Shu, “Runge-Kutta discontinuous Galerkin
methods for convection-dominated problems,” J. Sci. Comp., vol. 16,
pp. 173–261, 2001.
[5] J. Fitzgerald and L. Bergdahl, “Including moorings in the assessment
of a generic offshore wave energy converter: a frequency domain
approach,” Marine Structures, vol 21, pp. 23–46, 2008.
[6] J. Fitzgerald, “Position Mooring of Wave Energy Converters,” Ph.D.
thesis, Chalmers University of Technology, Gothenburg, Sweden, 2009.
[7] D.L. Garrett, “Coupled analysis of floating production systems,” Ocean
Eng., vol. 32, pp. 802–816, 2005.
[8] N.G. Jacobsen, D.R. Fuhrman and J. Fredsøe, “A wave generation
toolbox for the open-source CFD library: OpenFoam ,” R Int. J. Num.
Meth. Fluids, vol. 70, pp. 1073-1088, 2012.
[9] Johanning, L. et al. “Measurements of static and dynamic mooring line
damping and their importance for floating WEC devices,” Ocean Eng.,
vol. 34, pp. 1918-1934, 2007.
[10] G.Em. Karniadakis and S.J. Sherwin, Spectral/hp element methods for
CFD, 2nd ed., Oxford University Press, 2005.
[11] L. Larsson, F. Stern and M. Visonneau (eds.) “Gothenburg 2010 – A
Workshop on Numerical Ship Hydrodynamics,” Chalmers University of
Technology, Sweden, Department of Shipping and Marine Technology
Report No. R-10:122, 2010.
[12] Y. Li and M. Lin, “Regular and irregular wave impacts on floating
body,” Ocean Eng., vol. 42, pp. 93–101, 2012.
[13] Y. Li and Y.-H. Yu, “A synthesis of numerical methods for modeling
wave energy converter-point absorbers,” Renew. Sust Energy Rev., vol.
16, pp. 4352-4364, 2012.
[14] Q.W. Ma and S. Yan, “QALE-FEM for numerical modelling of non-
linear interaction between 3D moored floating bodies and steep waves,”
Int. J. Numer. Meth. Engng., vol. 78, pp. 713-756, 2009.
[15] A. Montano, M. Restelli and R. Sacco, “Numerical simulation of
tethered buoy dynamics using mixed finite elements,” Comp. Meth.
Appl. Mech. Engng., vol. 196, pp. 4117–4129, 2007.
[16] OpenCFD Ltd (2013). OpenFOAM homepage. [Online]. Available
http://www.openfoam.org
[17] J. Palm, G. M. Paredes, C. Eskilsson, F. Taveira-Pinto and L. Bergdahl,
“Simulation of mooring cable dynamics using a discontinuous Galerkin
method,” in Proc. MARINE2013, 2013.
[18] H. Rusche, “Computational fluid dynamics of dispersed two-phase flows
at high phase fractions,” Ph.D. thesis, Imperial College, London, 2002.
[19] F. Stern, J. Yang, Z. Wang, H. Sadat-Hosseini, M. Mousaviraad, S.
Bhushan and T. Xing, “Computational ship hydrodynamics: Nowadays
and way forward,” in Proc. 29th Symp. Naval Hydrodyn., 2012.
[20] M.S. Triantafyllou and C.T. Howell, “Dynamic response of cables under
negative tension: An ill-posed problem,” J. Sound and Vibration, vol.
173, pp. 433–447, 1994.
[21] Y. Yu and Y. Li, “Preliminary results of a RANS simulation for
a floating point absorber wave energy system under extreme wave
conditions,” in Proc. OMAE2011, 2011.
[22] H.G. Weller, G. Tabor, H. Jasak and C. Fureby, “A tensorial approach
to CFD using object oriented techniques,” Comp. in Physics, vol. 12,
pp. 620–631, 1998.

View publication stats

You might also like