Ocean Engineering 263 (2022) 112206

Contents lists available at ScienceDirect

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

Numerical coupling strategy using HOS-OpenFOAM-MoorDyn for OC3

Hywind SPAR type platform✩
Sithik Aliyar a,b ,∗, Guillaume Ducrozet b , Benjamin Bouscasse b , Félicien Bonnefoy b , V. Sriram a ,
Pierre Ferrant b
a
Department of Ocean Engineering, IIT Madras, Chennai, 600036, Tamil Nadu, India
b Nantes Université, École Centrale de Nantes, CNRS, LHEEA, UMR 6598, Nantes, F-44000, France

ARTICLE INFO ABSTRACT

Keywords: The coupled (Potential theory and Navier–Stokes) solver is applied to simulate the interaction of sea waves
Numerical modeling with substructure of floating offshore wind turbines (FOWT) platform, notably similar to the OC3 Hywind
CFD SPAR structure. The intention is to develop a numerical tool that allows the study of the survivability of
OpenFOAM
floating structures in extreme sea states. In this study, the moorings are modeled in two ways. One is by
Floating bodies
considering the mooring lines as a linear spring with defined spring stiffness, and another is by coupling the
SPAR
Moorings
solver (foamStar) with a lumped-mass mooring dynamics model (MoorDyn). MoorDyn represents mooring line
FOWT behavior subject to axial elasticity, hydrodynamic forces, and vertical contact forces with the seabed. The
MoorDyn coupled model has been validated against the experiments carried out as part of the SOFTWIND project. The
Nonlinear wave structure interactions numerical model results of free surface elevation, floating body motions and mooring tensions are compared
with the experiments. For wave cases with mild and moderate amplitudes, mooring in the form of a stiffness
matrix is sufficient. However, dynamic mooring simulation (MoorDyn) is required for the extreme sea state
conditions.

1. Introduction techniques without scaling limitations, on the other hand, may improve
model reliability by lowering uncertainties. Based on whether the fluid
Floating Offshore Wind Turbines (FOWT) technology is a new trend viscosity is considered or not, the numerical wave tank (NWT) simu-
that allows wind turbines to be installed in larger, deeper offshore lations can be grouped into potential theory-based NWT (PNWT) and
locations for increased wind potential. But the accurate prediction of Navier–Stokes (NS) equation based CFD-NWT (CNWT). Historically,
floater behavior in such a dynamic environment is still a key chal- wave-structure interaction was modeled using linear potential flow
lenge (Pinguet, 2021; Huang et al., 2021; Zhang and Kim, 2018). theory with the boundary element method (BEM), when a smooth
Wave-structure interaction problems can be assessed by empirical for- geometry and negligible viscous effects prevail (Black et al., 1971;
mulations, model tests in wave tanks, and numerical methods. The most Sun, 2007).However, when flow separation occurs around the body
accurate method is model testing, which provides a direct comparison and nonlinear behavior of a floating structure is the point of interest,
to physical reality within the limitations of experimental error and assumptions of irrotational and inviscid flow in PNWT will not rep-
wave tank fidelity. However, it is rarely possible to employ a real-scale resent a realistic scenario for reproducing model testing. Multiphase
test model and accurately calculate critical loads for the complex multi-
CNWT, together with specific wave generation and absorption tech-
physics problem. There is a restriction in choosing the scaling factors
niques (Jacobsen et al., 2012; Higuera et al., 2013), can reproduce
based on the experimental facilities available in model testing (Aliyar
open sea and experimental conditions (Elhanafi et al., 2017; Gu et al.,
et al., 2021). These scaling effects cannot be ignored in predicting the
2018; Devolder et al., 2018; Guilcher et al., 2007; Miquel et al., 2018).
elemental forces acting over the wind turbine models (Make, 2014).
A large computational domain is required to develop nonlinear wave
Also, moored FOWT physical models add multiple uncertainties like the
evolution to avoid boundary effects resulting from waves interacting
shifts in the equilibrium position, mooring dynamics, motion effects,
with any fixed or floating structures in the domain. Also, high temporal
etc., which are difficult to control during the experiment. Numerical

✩ This document is the results of the research project funded by the Joint doctoral fellowship by IITM-ECN.
∗ Corresponding author at: Department of Ocean Engineering, IIT Madras, Chennai, 600036, Tamil Nadu, India.
E-mail addresses: [email protected] (S. Aliyar), [email protected] (G. Ducrozet), [email protected] (B. Bouscasse),
[email protected] (F. Bonnefoy), [email protected] (V. Sriram), [email protected] (P. Ferrant).

https://doi.org/10.1016/j.oceaneng.2022.112206
Received 15 April 2022; Received in revised form 19 July 2022; Accepted 4 August 2022
Available online 26 August 2022
0029-8018/© 2022 Elsevier Ltd. All rights reserved.
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

and spatial discretization is mandatory in CFD solvers to reproduce the DoF dynamics of rigid and flexible floating/fixed bodies in waves. In-
exact events and avoid excessive numerical dissipation during wave build modules can support the simulation for the decay test, regular
propagation. These factors will significantly increase the computational and irregular waves interaction with the structure, following one way
costs, resulting in the conventional CFD approaches being impractical. DD approach. The solver has been developed and validated for regular
In recent years, many hybrid coupling methods have appeared and and irregular wave propagation (Choi et al., 2020; Li et al., 2021) and
grown in popularity to overcome the weakness of both PNWT and other applications which can be referred at (Kim, 2021; Choi, 2019; Li,
CNWT in the name of domain decomposition (DD). Generally, the com- 2018).
plex interaction (viscous effects and violent free-surface deformation) This paper aims to implement a static and dynamic mooring model
only occurs near the structure, and in the far field, these effects can of- (MoorDyn (Hall, 2015)) into foamStar to conduct a fully coupled anal-
ten be neglected. For this reason, it is viable to divide the computational ysis of moored floating structures replicating the model test in NWT.
domain into a viscous inner sub-domain and an inviscid outer sub- MoorDyn is a lumped-mass mooring model that has been associated
domain. Potential flow models solve the flows in the outer subdomain, with a potential flow solver and successfully validated against measure-
and CFD models solve the inner subdomain. Information between the ments in a wave tank (Hall and Goupee, 2015). The equation governing
models is shared in a common region, either in one-way or two-way dynamics of mooring lines is coupled to the six DoF nonlinear rigid
interaction. Many papers have published the latest advancements in body motions equations in foamStar. MoorDyn calculates the tensions of
domain decomposition, devoted to one way (Hildebrandt et al., 2013; the mooring lines, which are added as an external force to compute the
Lachaume et al., 2003; Biausser et al., 2004; Paulsen et al., 2014; Choi resulting response and motion of the moored floating structures. Based
et al., 2020; Agarwal et al., 2021b)and two-way coupling the (Sriram on our knowledge, this is the first example of coupling between a DD-
et al., 2014; Lu et al., 2017; Kemper et al., 2019). For more details based solver (here foamStar), floating body dynamics, and a dynamic
about the different hydrodynamic coupling, Di Paolo et al. (2020) mooring model with a systematic validation of the structural motions
detailed review can be referred to. In addition, some articles portray and mooring tensions with the experiment.
the robustness of DD for coastal engineering applications (Lachaume Validation of the two-way coupling between the foamStar and Mo-
et al., 2003; Biausser et al., 2004; Sriram et al., 2014) and bottom orDyn is carried out with the experimental results on the floating SPAR
fixed support structures (Hildebrandt et al., 2013; Paulsen et al., 2014; with a delta mooring system. The moored decay tests are examined to
Li et al., 2021; Sriram et al., 2021; Agarwal et al., 2021a), but its ensure the computational model performance in comparison with the
application to the floating structures is the recent research interest experimental model in terms of its natural period, mooring tension,
and damping properties. The SPAR is then subjected to regular and
among the researchers (Xu et al., 2021; Li and Bachynski, 2021).
irregular waves in the following study, and simulations are compared to
Several studies have used CFD to model the floating structures in
experimental data, verifying the developed coupling model’s ability to
single or multiple degrees of freedom (DoF) (Elhanafi et al., 2017;
effectively forecast both body motions and mooring line tensions. The
Simonsen et al., 2013; Dunbar et al., 2015). The assessment of moored
paper has been organized in the following manner. Section 2 discusses
floating structures needs interdisciplinary expertise in hydrodynamics,
the governing equations underpinning the foamStar and MoorDyn mod-
floating body dynamics, mooring, and their interaction. Three types of
els, as well as the framework of coupling between the solvers. Section 3
mooring analysis have been commonly adopted, static, quasi-static, and
shows the general setup and parametric studies used in this study.
dynamic analysis. Static analysis is often used in preliminary design
Section 4 shows moored floating structures validation with both regular
to compute the static response of moored vessels (Karnoski and Palo,
and irregular waves, and the paper concludes with a summary of the
1988). The offshore sector has a tradition of designing the mooring
hybrid model’s capability and the performance.
system through quasi-static analysis, justified by the low responsiveness
of the large masses and corresponding low velocities. In operational
2. Numerical methods
conditions, this assumption might be sufficient (Thomsen et al., 2017).
As the quasi-static model bypasses hydrodynamic drag and inertial
In the present work, the one-way domain decomposition approach
forces on mooring lines, it underpredicts line tensions under extreme (Di Paolo et al., 2020) using foamStar (Choi et al., 2020) as the CFD
conditions. Also, in deep waters, a mooring line’s response may be solver and CN-Stream, HOS-NWT as the potential solvers has been
delayed, causing it to violate the catenary profile assumptions of a adopted as shown in Fig. 1. In simple terms, the potential solver’s
quasi-static model (Vickers, 2012). Thus, employing a dynamic moor- velocity, pressure, and surface elevation will be transferred into the
ing model to reliably predict loads acting on mooring lines and the viscous solver through the inlet relaxation zone, and all parameters
floating structure may be advantageous. MDD (Dewey, 1999), SEA- from the viscous flow domain will be blended to zero in the outlet
WAY (Journee and Adegeest, 2003), OrcaFlex (Randolph and Quiggin, relaxation zone. This section provides the governing equations behind
2009), Moody (Ferri and Palm, 2015), and MoorDyn (Hall, 2015) are both solvers, followed by the adopted coupling approach. Next, the
some of the mooring dynamics models developed in recent years. But mooring models adopted in the solver are presented, and its coupling
there are only a few papers in which CNWT is used in conjunction with methodology has been briefly discussed.
the above library to solve mooring line tensions (Nematbakhsh et al.,
2015; Benitz et al., 2015; Palm et al., 2016; Jiang et al., 2021; Lee 2.1. foamStar
et al., 2021). In addition, most of the research so far has used a floating
box, buoy, or cylinder as a floating structure for validation, and there The open-source solver interDymFoam (of OpenFOAM) is used to
is a lack of validation for real-time floating systems like SPAR, TLP, or solve the multiphase problem by coupling the Navier–Stokes equations
Semi-submersible for industrial applicability. with a Volume of Fluid (VoF) method (Hirt and Nichols, 1981). The
In this study, we have used foamStar (Choi et al., 2020), an in-house computational domain is subdivided into a finite number of control
code co-developed by Bureau Veritas and Ecole Centrale de Nantes volumes(CV), which can be unstructured. The integrals over each CV
in the OpenFOAM framework for solving wave-structure interaction are numerically approximated using the midpoint rule. The pressure
problems. It is based on the standard multi-phase solver interDymFoam Poisson equation is solved based on a PIMPLE algorithm, which com-
in the OpenFOAM package with two special modules. One is wave bines the Semi-Implicit Pressure Linked Equations (SIMPLE) algorithm
generation, similar to waves2Foam (Jacobsen et al., 2012) but it is with the Pressure Implicit with Splitting of Operators (PISO) algorithm.
further developed with a faster implementation and to allow the cou- The unknown variables at the center of a cell face are determined
pling with the nonlinear potential open-source solvers HOS-Ocean and by combining a central differencing scheme (CDS) with an upwind
HOS-NWT. Another is the mechanical solver designed to solve the 6 differencing scheme (UDS). The 𝑘 − 𝑜𝑚𝑒𝑔𝑎 SST turbulence model was

2
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 1. General description of the Numerical Wave Tank (NWT) under one way Domain decomposition strategy between potential and viscous model.

used for the turbulence requirement, however, it was discovered after where ℎ is water depth. In the present work, spectral methods are
the case study that it had the least impact on the cases picked, hence used to solve the potential and surface elevation. The dynamics of non-
it is not included in this study. breaking gravity waves are accurately modeled using the potential flow
The governing equations are based on mass and momentum conser- theory. The numerical methods at use in the present study are based
vation equations written as follows, on the stream function theory for regular waves and the Higher Order
Spectral (HOS) method for irregular waves. Both methods are briefly
∇⋅𝑢 = 0 (1)
presented in this section followed by coupling with foamStar.
𝜕(𝜌𝑢𝑢)
+ ∇ ⋅ (𝜌𝑢𝑢𝑢 𝑇 ) − ∇.(𝜇(∇𝑢𝑢 + ∇𝑢𝑢𝑇 )) = −∇𝑝𝑑 − (𝑔𝑔 ⋅ 𝑥 )∇𝜌 (2) 2.2.1. Regular Waves: Stream function wave theory
𝜕𝑡
The open-source CN-Stream model (the library is published with
where 𝑢 , 𝑥 and g= [0, 0, −𝑔]𝑇 are the fluid velocity, position vector and
GNU General public license and can be downloaded from the URL
gravitational acceleration vector, respectively. The dynamic pressure
https://github.com/LHEEA/CN-Stream) is used to describe nonlinear
𝑝𝑑 = 𝑝 − 𝜌𝑔𝑔 ⋅ 𝑥 is as introduced in Rusche (2002). The free surface is
regular waves propagating over a flat bottom. It is based on the stream
identified using the volume of fluid (VoF) method, where the two-phase
function theory (Rienecker and Fenton, 1981) that allows an efficient
problem is treated as a single fluid with a volume-fraction parameter
and accurate solution of regular wave trains up to the breaking wave
𝛼 ∈ [0, 1] subjected to transport equation,
limit. The model assumes periodic boundary conditions and solves the
𝜕𝛼 propagation in a moving frame, making the problem steady. Spatial
+ ∇ ⋅ (𝛼𝑢𝑢) + 𝑐𝛼 ∇ ⋅ (𝑢𝑢𝑛 𝛼(1 − 𝛼)) = 0 (3)
𝜕𝑡 discretization is achieved thanks to a spectral method that decomposes
where the third term on the left-hand-side is the artificial compression the unknowns (free surface elevation and velocity potential) on a set
to keep the interface sharp. 𝑢 𝑛 is the fluid velocity normal to the of basis functions that satisfy intrinsically Laplace equation (Eq. (6))
interface and 𝑐𝛼 is the compression coefficient. In each cell, then fluid as well as the bottom boundary condition (Eq. (9)). The modal am-
properties (density and viscosity) are computed as a mixture between plitudes are numerically evaluated to satisfy the free surface boundary
air (𝛼 = 0) and water (𝛼 = 1): conditions. They can be used to reconstruct the wave field (velocity and
pressure through Bernoulli’s equation (Eq. (8)) at required locations.
𝜌 = 𝛼𝜌𝑤 + (1 − 𝛼)𝜌𝑎 𝜇 = 𝛼𝜇𝑤 + (1 − 𝛼)𝜇𝑎 (4)
The reader can find more information in Rienecker and Fenton (1981),
where index 𝑤 and 𝑎 indicate water and air respectively. Ducrozet et al. (2019).

2.2. Wave models and coupling 2.2.2. Irregular Waves: Higher order spectral method
Stream function methodology is limited to regular waves. In the
The incident wave generation is based on the potential theory present work, we are addressing another type of unidirectional wave,
approach. With the assumption of an ideal fluid and irrotational flow, irregular sea states (refer Section 3.2.5). With the objective of an ac-
it is possible to define a velocity potential (𝜙) as, curate and computationally efficient numerical model, the High-Order
Spectral (HOS) method (Dommermuth et al., 1988; West et al., 1987)
𝑢 𝐼 = ∇𝜙 (5) has been adopted to describe the complex sea states. This numerical
Then the incident wave potential satisfies Laplace’s equation in the fluid approach relies on a pseudo-spectral discretization scheme that exhibits
domain (𝛺) a high (exponential) convergence rate and that allows the use of Fast
Fourier Transforms, which ensures high efficiency.
𝛥𝜙 = 0 𝑥 ∈ 𝛺 (6) The original HOS method (Dommermuth et al., 1988; West et al.,
1987) was dedicated to the time evolution of an initially prescribed
The nonlinear kinematic and dynamic free surface boundary conditions
wave field in an open domain with periodic boundary conditions. The
for incident waves are given as,
present study intends to replicate experiments carried out in a wave
𝜕𝜂𝐼 𝜕𝜙 𝜕𝜂𝐼 𝜕𝜙 𝜕𝜂𝐼 𝜕𝜙 tank where the waves are generated thanks to a wavemaker in a
+ + + =0 𝑎𝑡 𝑧 = 𝜂𝐼 (𝑥, 𝑡) (7)
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 finite-size domain. Then, in this work, a dedicated Numerical Wave
𝜕𝜙 1 Tank (NWT) based on HOS is used (Ducrozet et al., 2012). This open-
+ (∇𝜙2 ) + 𝑔𝑧 = 0 𝑎𝑡 𝑧 = 𝜂𝐼 (𝑥, 𝑡) (8)
𝜕𝑡 2 source solver named HOS-NWT has been developed at LHEEA Lab,
where 𝑥 and 𝑦 are horizontal coordinate, 𝑡 is time. The flat bottom (Ecole Centrale Nantes (ECN)) and now constitutes a Digital Twin of
condition is given as, the ECN wave tank facility. The geometry is the exact physical one,
including perfectly reflective sidewalls. The initial conditions are at
𝜕𝜙
= 0 𝑎𝑡 𝑧 = −ℎ (9) rest and the waves are generated thanks to a wavemaker. For direct
𝜕𝑧

3
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

comparison with experiments, the exact same wavemaker motion is the center of the domain coinciding with the body, where the body-
used in the numerical and the physical wave tank. In addition, a fixed frame origin is set at the body’s Center of gravity. A positive
numerical absorbing zone mimics the physical beach located close to surge is defined forward, and a positive heave is defined upwards. The
the wall opposite the wavemaker to prevent wave reflection. This NWT rotation angles need to be defined carefully since the transformation
has been applied and validated on different wave configurations of between the global reference frame and the body-fixed frame is not
various levels of complexity (Ducrozet et al., 2012, 2016; Michel et al., unique. The foamStar library supports the Euler angle and quaternion
2020; Canard et al., 2020). representations among different methods of transformation.
The present work follows the definition of Euler intrinsic angles
2.2.3. Relaxation Zone - coupling between solvers with zyx convention. The translation and rotational equations of motion
Because of the unsteady nature of the water waves and the moving are expressed in the body-fixed frame as follows:
free surface, CFD solvers must have appropriate temporal and spatial
discretization to accurately simulate wave-related problems. Further- [𝑚] 𝑞̈ = 𝑄 𝑒 + 𝑄 𝑣 + 𝑄 𝑐 − [𝐾]𝑞𝑞 − [𝐶]𝑞̇ (12)
more, practical marine and ocean engineering applications require a
sufficient simulation time (to eliminate the transient flow regime at [𝐼] 𝜃̈ = 𝑀 𝑒 + 𝑀 𝑣 + 𝑀 𝑐 − [𝐾]𝑟𝑜𝑡𝜃 − [𝐶]𝑟𝑜𝑡𝜃̇ (13)
the start of the simulations and achieve a periodic condition). Also,
the NS solver is known for its energy loss in wave propagation, which Where 𝑚, 𝐼, 𝐾, 𝐶 are the generalized mass, Inertia, stiffness, and
dampens the waves in long duration and its prohibitive computational damping matrix respectively; 𝑞 and 𝜃 represents the translation and
cost. Hence the objective is to couple the viscous solver with the po- rotational motions; 𝑄 𝑒 and 𝑀 𝑒 are the external forces and moments
tential solvers to eliminate its shortcomings, and this coupling happens acting on the body; 𝑄 𝑣 and 𝑀 𝑣 are the quadratic velocity forces
in the relaxation zone (Fig. 1). These relaxation zones are used to (Coriolis/centrifugal forces) and moments; 𝑄 𝑐 and 𝑀 𝑐 represents the
prevent waves reflected from outlet boundaries from interfering with constraint forces and moments. Gravity, mooring, propulsion, and fluid
wavemaker boundaries, as well as to attenuate the waves reflected are some of the components of external forces and moments. Among
internally in the computational domain (Jacobsen et al., 2012). them, fluid force is evaluated as the surface integration of pressure and
A simple way of generating a wave based on one-way DD (foamStar) viscous stress over the body surface. Here, 𝑛 is a unit surface normal
is to impose wave velocity and free surface elevation from potential vector pointing inside the body and 𝜏 indicates the viscous stress tensor.
wave models at the inlet of the CFD domain. In foamStar, the wave
generation and absorption method are based on an explicit scheme that
𝑄 𝑒,𝑓 𝑙𝑢𝑖𝑑 = (𝑝𝑛𝑛 − 𝜏 ⋅ 𝑛 ) 𝑑𝑆 (14)
relaxes the computed solution towards a given target flow field (Ja- ∫𝑏𝑜𝑑𝑦
cobsen et al., 2012; Seng, 2012; Mayer et al., 1998; Engsig-Karup,
2006; Choi et al., 2020)), and the scheme requires a weight function 2.4. Mesh morphing
that varies between 0 and 1 in the relaxation zone. It is achieved as
follows: In general, any problem in computational fluid dynamics (CFD)
implies the use of a grid or mesh to replace a continuous problem
𝜁 = 𝑤𝜁𝑡𝑎𝑟𝑔𝑒𝑡 + (1 − 𝑤)𝜁𝐶𝐹 𝐷 (10)
domain with a finite field. This model works flawlessly when the
where 𝜁 is either the velocity field(u) or the VOF field (𝛼), w is the problem does not involve any solid body motion in the computational
weight function, and 𝜁𝑡𝑎𝑟𝑔𝑒𝑡 will be the given target flow field usually domain. However, many offshore engineering applications require solid
set to be either the incident wavefield or calm-water (to use relaxation body motion in the computational domain, which has a consequence
as an absorbing zone). Two common choices of weight function are the on the fluid flow. Sliding interface, Overset mesh adaptation, and
polynomial weight function (Mayer et al., 1998; Engsig-Karup, 2006; Mesh morphing are the three different ways dynamic mesh motion
Choi et al., 2020) and the exponential weight function (Mayer et al., operates in OpenFOAM. A sliding interface allows one or more internal
1998; Engsig-Karup, 2006; Jacobsen et al., 2012; Seng, 2012; Choi subdomains to translate or rotate relative to a static background mesh
et al., 2020). The present study adopts the explicit relaxation scheme domain, with the interface between the domains sliding relative to each
using the polynomial weight function: other. The main disadvantage of sliding interfaces is that they are only
appropriate for single DoF motions. A complex motion can be simulated
𝑤(𝑥) = −2𝑥3 + 3𝑥2 (11)
using a mesh overset technique, but it may prove very computationally
Where 𝑥 ∈ [0, 1], is the normalized coordinate in the relaxation expensive mainly due to the mesh refinement and adaptation, the
zone. As mentioned, the above technique can be applied in the wave remapping, and the re-computation. This is not necessary if the mesh
absorption zone as if target values (𝜁𝑡𝑎𝑟𝑔𝑒𝑡 ) are set to no waves, then is only deformed and the mesh’s topology remains unchanged. Mesh
the entire wavefield would be damped, and if the target values are morphing is the most commonly used method in CFD floating body
set to incident waves, then only the scattered waves are damped. Choi simulations. The method deforms meshes into new meshes by updating
et al. (2020) can be referred for some examples of 3D wave genera- the node positions near the floating body while preserving the shape of
tion of realistic sea spectrum. To summarize, In the present case, the the target floating body. Because only the nodal positions are updated,
relaxation zone explicitly blends the wave properties from CN-Stream the necessary computation time is shorter than that of the sliding
or HOS-NWT within the foamStar domain. In the case of irregular interface and overset mesh methods. However, in the case of large
waves (HOS-NWT) wave properties are blended using the Grid2Grid deformations, solutions may diverge due to the skewness of the mesh,
wrapper program, which uses inverse FFTs and interpolation technique and meshes that have been optimized for a generated wave signal
to reconstruct the wave amplitude and the velocity field at the inlet for may also get impacted. Assuming that the amplitude of the floating
required time and spatial locations (Choi et al., 2017). body motion is relatively small in the present case, the mesh morphing
deformation method is adopted.
2.3. Rigid body motions Different morphing techniques can be applied within the finite
volume framework (Jasak and Tuković, 2010) like Spring analogy,
There are two reference frames used in the present study, the Pseudo-solid, Laplacian smoothing, Algebraic interpolation, solid body
Inertial (𝑅0 ) and the body reference frame(𝑅𝑏 ) (refer Fig. 5). All the motion, etc. The mesh update strategy is based on the Laplacian mesh
frames are based on Cartesian systems and defined according to the smoothing technique in the present work and the mesh motion of the
right-hand rule. The Inertial reference frame’s origin is positioned at computational domain is calculated by solving the cell center Laplace

4
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 2. Mooring line discretization in MoorDyn with a representation of internal and external forces in inertial reference frame 𝑅0 (Hall and Goupee, 2015).

smoothing equation (Jasak and Tukovic, 2006). The Laplacian equation where 𝑚𝑖 is the mass matrix for node 𝑖. The mass of the mooring cable
is given as, is discretized into point masses at each node by allotting each node half
1 the combined mass of the two neighboring cable segments.
∇ ⋅ (𝛾(𝑟)∇𝑥
𝑥) = 0; 𝛾(𝑟) = (15)
𝑟2 The mooring line is represented as a cylinder of diameter 𝑑, un-
where 𝑥 is the point displacement field and 𝛾 is the diffusivity coeffi- stretched length L, and density 𝜌. For a mooring line with 𝑁 segments
cient. 𝑟 is the distance to the moving boundary. Several choices for 𝛾 connecting 𝑁 + 1 nodes, the unstretched length of each line segment is
are presented in Löhner and Yang (1996), Jasak and Tuković (2010). 𝑙0 = 𝐿∕𝑁. The forces acting on the nodes are divided into internal and
In the present work, the inverse quadratic distance relation has been external forces. The internal forces include net buoyancy (𝑊𝑖 ), tension
chosen as the coefficient will be significant, and the gradient will be (𝑇𝑖 ) and internal damping (𝐶𝑖 ) forces. The external forces contain
small, yielding a small relative displacement for the points close to transverse and tangential hydrodynamic forces (𝐷𝑖 ) based on Morison’s
the moving boundary, preserving the cells’ quality near the moving equation, added mass force (𝑎𝑝𝑖 and 𝑎𝑞𝑖 ) and the seabed contact force
boundaries (Jasak and Tukovic, 2006). (𝐵𝑖 ). The tension, 𝑇𝑖+1∕2 in any line segment is calculated as,

2.5. Mooring system ⎧ 𝜋 (𝑙 )

⎪𝐸 4 𝑑 2 𝑙 − 1 if 𝑙 > 𝑙0
𝑇𝑖+1∕2 =⎨ 0 (18)
Mooring systems are widely used in offshore engineering for station ⎪0 if 𝑙 ≤ 𝑙0
⎩
keeping of floating bodies in various sea states. In the present study,
two different mooring systems are adopted. One is based on the static Where 𝐸 is Young’s modulus. The calculations assume that line tensions
mooring condition, and another approach is based on the dynamic are always positive 𝑙 > 𝑙0 (compression line forces are not allowed).
mooring condition. Both are discussed briefly. By multiplying a unit vector directed along a line segment, the above
equation can be rewritten in vector form as
2.5.1. Static mooring model: Stiffness Matrix
⎧ 𝜋 ( ‖𝑟𝑟𝑖+1 −𝑟𝑟𝑖 ‖ )(
𝑟 −𝑟𝑟
)
The external mooring used to control the motions in six DoF is ⎪𝐸 𝑑 2 − 1 ‖𝑟𝑟𝑖+1 −𝑟𝑟𝑖 ‖ if ‖𝑟𝑟𝑖+1 − 𝑟 𝑖 ‖ > 𝑙0
simplified as the linear spring model with stiffness matrix [𝐾]. The 𝑇 𝑖+(1∕2) = ⎨ 4 𝑙0 𝑖+1 𝑖 (19)
⎪0 if ‖𝑟𝑟𝑖+1 − 𝑟 𝑖 ‖ ≤ 𝑙0
restoring force is given as, ⎩
𝑄𝑒,𝑀 ] = [𝐾𝑀 ][𝛿𝑞𝑞 ]
[𝑄 (16) The net buoyancy (𝑊𝑖 ) at node 𝑖 assuming the force is evenly divided
among two connecting nodes is described in vector form as follows:
where 𝑄 𝑒,𝑀 represents external force and moment from mooring, and
𝛿𝑞 represents the change in motion (in six DoF). This force is added 𝑊𝑖− 1 + 𝑊𝑖+ 1
2 2
to the right-hand side of the equation of motion (Eq. (13)) that can be 𝑊 𝑖 = 𝑒̂ 𝑧 (20)
2
interpreted as attaching a spring to the body’s center of gravity. The
Where
working procedure of the stiffness matrix method inside the foamStar
algorithm is depicted in the general flowchart shown in Fig. 3. 𝜋 2
𝑊𝑖 = 𝑑 𝑙(𝜌𝑤 − 𝜌)𝑔 (21)
4
2.5.2. Dynamic mooring model: MoorDyn Where 𝜌𝑤 is the density of water and 𝜌 is the density of the mooring
The dynamic mooring solver MoorDyn (Hall, 2015) is a lumped- line. In order to improve numerically stability, a numerical internal
mass mooring model that lumps all the effects of mass, external forces, damping force is included:
and inertial reactions at a finite number of nodes, N, along the mooring ( 𝑟 −𝑟 )
𝜋 𝑖+1 𝑖
line connected by massless spring as shown in Fig. 2. It follows the 𝐶 𝑖+(1∕2) = 𝐶𝑖𝑛𝑡𝑒𝑟 𝑑 2 𝜖̇ 𝑖+(1∕2) (22)
4 ‖𝑟𝑟𝑖+1 − 𝑟 𝑖 ‖
right-handed inertial reference frame defined with the 𝑧-axis being
measured positive up from the water plane. The location of each Where 𝐶𝑖𝑛𝑡𝑒𝑟 is the numerical internal damping coefficient and 𝜖̇ 𝑖+1∕2
node 𝑖 is defined by the position vector 𝑟 𝑖 which includes [𝑋𝑖 , 𝑌𝑖 , 𝑍𝑖 ] is the strain rate of segment. The linear spring damper model is acti-
components of the vector. The equation of motion for each node 𝑖 is vated only when the node interacts with the sea bottom (𝑧𝑏𝑜𝑡 ) and it is
written as follows: given as:

𝑚𝑖 + 𝑎𝑝𝑖 + 𝑎𝑞𝑖 )̈𝑟𝑖 = 𝑇 𝑖 + 𝐶 𝑖 + 𝑊 𝑖 + 𝐵 𝑖 + 𝐷 𝑝𝑖 + 𝐷 𝑞𝑖

(𝑚 (17) 𝐵 𝑖 = 𝑑𝑙[(𝑧𝑏𝑜𝑡 − 𝑧𝑖 )𝑘𝑏 − 𝑧̇ 𝑖 𝑐𝑏 ]̂𝑒𝑧 (23)

5
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 3. Computational algorithm of coupling between the potential theory wave models, foamStar, and MoorDyn solver.

where 𝑘𝑏 is the seabed stiffness coefficient, and 𝑐𝑏 is the seabed damp- present work has adopted the weakly coupling between the foamStar
ing coefficient. and MoorDyn that it implies that mooring lines will influence the plat-
The drag and added mass are calculated at each node 𝑖 based form’s motion but not the flow field. The coupling procedure follows
on Morison’s equation. The direction of a line passing between two three phases: initialization, information sharing, and closing.
adjacent nodes is approximated by the tangential direction 𝑞̂ 𝑖 (Fig. 2) : Initialization phase: Before the time loop begins, the solvers are
being set up, and the mooring system is created. An initialization
𝑟 𝑖+1 − 𝑟 𝑖 function is then called calculating its initial conditions based on body
𝑞̂ 𝑖 = (24)
‖𝑟𝑟𝑖+1 − 𝑟 𝑖 ‖ position specified by the equation of the motion.
Transverse and tangential drag forces are calculated using the trans- Information sharing phase: As time progress, foamStar transfers
verse drag coefficient 𝐶𝑑𝑛 , and the tangential drag coefficient, 𝐶𝑑𝑡 the six degrees of freedom displacement (𝑥), velocities (𝑥𝑣), and current
respectively. Similarly, transverse and tangential added mass forces are timestep size into the MoorDyn. MoorDyn calculates the fairleads’
calculated using the transverse added mass coefficient 𝐶𝑎𝑛 , and the position, velocities, and accelerations based on the input. Along with
tangential added mass coefficient, 𝐶𝑎𝑡 respectively. Both the forces are that, the mooring system tensile load and restoring forces are calculated
computed as follows: and shared with the foamStar as one of the external forces [𝑄𝑒,𝑀 ]. Then
the foamStar calculates the platform motions, accelerations etc., and the
1
𝐷 𝑝𝑖 = 𝜌 𝐶 𝑑𝑙 ‖ (̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 − 𝑟̈ ‖ [(̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 − 𝑟̈ ] (25) loop continues till the last time step.
2 𝑤 𝑑𝑛
Closing phase: The last stage releases the variables and memories
1 used to calculate the tensile load of the mooring system after the end
𝐷 𝑞𝑖 = 𝜌 𝐶 𝑑𝑙 ‖ (−̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 ‖ [(−̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 ] (26)
2 𝑤 𝑑𝑡 of calculations.
𝜋 It should be noted that foamStar uses a body reference system (𝑅𝑏 )
𝑎 𝑝𝑖𝑟̈ 𝑖 = 𝜌𝑤 𝐶𝑎𝑛 𝑑 2 𝑙[(̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 − 𝑟̇ 𝑖 ] (27)
4 for its governing equation, and MoorDyn uses an inertial frame of
𝜋 reference (𝑅0 ) for its governing equation. Hence, body displacements
𝑎 𝑞𝑖𝑟̈ 𝑖 = 𝜌𝑤 𝐶𝑎𝑡 𝑑 2 𝑙(−̈𝑟 ⋅ 𝑞̂ 𝑖 )̂𝑞 𝑖 (28) and velocities are transferred to the inertial reference frame at each
4
The second-order system of ordinary differential equations pre- timestep and mooring tension to the body reference frame. The detailed
sented in Eq. (17) can be reduced to a system of first-order differ- framework of the coupling procedure is shown in Fig. 3.
ential equations, and then solved using a second-order Runge–Kutta
integration scheme with a constant time step 3. General Setup and parametric study

2.6. foamStar -MoorDyn coupling 3.1. Physical model

MoorDyn is coupled to the foamStar solver in two way coupling Arnal (2020), experimentally tested a 1:40 scale 10 MW SPAR
approach where the information between the solvers is passed at each floating wind turbine (FOWT) in the ECN wave basin. The basin (50 m
time step, and each solver will influence the solutions of the other. The 𝑥 30 m 𝑥 5 m) is fitted with a 48-hinged flap-wavemaker on its shorter
core functions in MoorDyn are compiled into a shared library to use side for wave generation. A beach at the opposite end of the basin
the dynamic loading of the functions inside the foamStar. Generally, absorbs the incoming wave energy and reduces the reflections. The
there are two ways of coupling between the solvers that can be adopted, basin can generate regular waves up to 1 meter height and irregular
strong and weak coupling. Each solver is solved separately in the weak waves up to 0.8 meter height. The experiment’s goal is to see how
coupling, and some variables are exchanged between the solvers. Also, the wind turbine controllers affect the global motions of the FOWT.
it can adopt different time step sizes for each solver. However, in the The experiment’s general layout is depicted in Fig. 4. Eight wave
strong coupling, both the solvers save all the variables in the common probes were used to measure wave heights. Two wave probes, WG1
solver, and each solver only composes its governing equation. The and WG2, are set 15 and 17 meters from the paddle, respectively,

6
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 4. The general layout of the experiment in side and top view with the mooring cable configuration, Arnal (2020). Figures are not drawn to scale.

near the sidewall. Wave probes WG3 to WG8 are spaced 15 cm apart 3.2.1. Computational domain
at a distance of 7.5 m from the SPAR. The motions were measured In the present coupling, two frames of reference are used, as indi-
using a MOCAP system based on a Qualisys environment that included cated in Fig. 5, the Inertial frame(𝑅0 ) and the body reference frame
multiple cameras, reflecting markers, and QTM software for real-time (𝑅𝑏 ). In 𝑅0 , HOS-NWT, has its origins in the NWT left corner at the
processing. The processing software takes the 3D positions of various still water level (SWL). In 𝑅𝑏 foamStar (CFD), which has its origin at
markers as input, smooths them out, and then computes the body’s 6 the body’s center of gravity. Finally, the MoorDyn (Mooring library) has
DoF positions. Wave elevation data were collected at a sample rate of its origins in an inertial frame of reference (𝑅0 ) at the SWL in the center
100 HZ, and 6-DoF motions are sampled at the rate of 200 HZ. of the foamStar domain. The generation of the regular incident waves is
The experiment model’s components include the floater, mooring achieved by using the stream function wave theory using an analytical
system, and the tower. SPAR was chosen as a floater with proportions approach, with the far-field boundary condition. The computational
based on the existing OC3 Hywind SPAR platform (Jonkman, 2010) domain for the regular wave study is similar to the foamStar domain
and Hywind Scotland as a reference system. The SPAR is made up of in Fig. 5. The HOS computational domain reproduces the experimental
two parts: an aluminum shell that gives the buoy its outward shape, set-up but without the structure. Only through the inlet relaxation zone,
and a ballast, battery, controller, power units, suspended wires, and the wave generated has been transferred into the foamStar domain and
other components that make up the buoy’s core. A vertically truncated in general, all relaxation zones can suppress the waves generated from
cylinder with the taper and transition piece to the tower forms the inside the viscous domain. The foamStar domain is solely confined near
SPAR floater. The primary dimensions and mass attributes have been the SPAR region, which is 6 times the diameter of the SPAR (excluding
summarized in Table 1. The three mooring cables were catenary chains RZ) and was determined based on the expected motions of the SPAR
positioned symmetrically 120◦ apart for the mooring mechanism. The in 6 DoF from the experiment. The length of the relaxation zones is
main mooring line was fastened to the bottom of the wave tank and one wavelength for inlet and outlet and half of the wavelength for
connected to two ‘‘bridles’’, which are part of the delta connection side zones decided based on the numerical tests. The numerical domain
on each cable. The attachment points for moorings were 0.335 m water depth of 5 m has been maintained in HOS and foamStar as
similar to that of the experiment. This research aims to evaluate the
below the waterline. Fig. 4 depicts the various pieces that make up
moored floating SPAR’s hydrodynamics. Therefore, RNA assembly in
the mooring model. The underwater unidirectional load cells (LS1, LS2,
the upper tower is not considered in modeling or analysis. However,
and LS3) connect the bottom chains to the bridles and are used to
while considering the total mass characteristics of FOWT, the mass,
measure mooring tension. Studless stainless steel chains were chosen
inertia, and COG of the ignored parts are also taken into account. The
in order to preserve the chain’s mass properties and geometry. Model
tower’s freeboard has been kept at 1.2 m to appropriately capture the
scale mooring properties are shown in Table 1. Because the goal is
wave run-up during the wave interaction.
to understand the system’s hydrodynamics, the details of the upper
part of the model, which includes aerodynamics components (actuator
3.2.2. Boundary conditions
assembly, nacelle etc.), are not considered in this study. More details
Boundary conditions must be set at the boundaries of the CFD
related to experiment can be found in Arnal (2020).
domain in order to guarantee the uniqueness of the governing equa-
tions. Table 2 lists the boundary conditions employed in the present
3.2. Numerical setup investigation and most of them are standard OpenFOAM boundary
conditions. The boundary conditions, Wavevelocity and waveAlpha
Regular and irregular wave simulations are used to validate the (free surface elevation in the form of VOF parameter 𝛼) sets the values
coupling in the domain decomposition. The decay test is also used from respective wave models at the inlet of the CFD domain. At other
to verify the natural period of SPAR and to estimate the damping RZ, the target velocity or VOF parameter is blended to zero with
coefficient along with the mooring tension. In the NWT, simulations the weight function by these boundary conditions. Thus, represent-
were run with and without the floating structure. The mesh conver- ing the simplest one-way domain decomposition approach for wave
gence for wave propagation was investigated using simulations without generation. The pressureInletOutletVelocity boundary condition is a
the structure. The following section describes the computational setup, blend of pressureInletVelocity and inletOutlet boundary conditions.
mesh convergence analysis and finally, the validation of the mooring In the pressureInletVelocity, velocity is computed from the difference
study is reported to demonstrate the model’s capability. between total and static pressure where the direction is normal to

7
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Table 1
Mass and dimensional properties of the SPAR with its mooring arrangement parameters.
Parameter Units Value
SPAR Mass (including floater, tower and RNA) kg 329.4
Center of Gravity with respect to Inertial Frame m −0.004, 0, −1.535
Mass Moment of Inertia with respect of CoG FOWT (𝐼𝑥𝑥 , 𝐼𝑦𝑦 , 𝐼𝑧𝑧 ) kg/m2 490, 490, 10
Diameter of floater (SPAR) m 0.45
Diameter at MSL (after tapering) m 0.28
Taper height m 0.2
Draft of the SPAR m 2.285
Mooring (Delta or Crow foot mooring)
Anchors depth m 5
chain diameter mm 3.7
Mass of chain in air kg/m 0.275
Fairlead depth m −0.335
Bridle chain length m 1.2
Bottom chain length (Line 1, Line 2 , Line 3) m 15.69, 15.74, 15.72
Mass of load cell in water (Clumped mass) kg 0.43

Fig. 5. Schematics of the computational domain within the HOS domain with Relaxation zones (RZ) are represented in terms of wavelength (𝜆). 𝑅0 represents the inertial reference
frame, and 𝑅𝑏 represents the body reference frame. HOS is based on 𝑅0 , but its origin is different.

Table 2
foamStar Boundary conditions.
Boundary Velocity Pressure Alpha
Inlet waveVelocity fixedFluxPressure waveAlpha
Outlet waveVelocity fixedFluxPressure waveAlpha
Side waveVelocity fixedFluxPressure waveAlpha
Atmosphere pressureInletOutletVelocity totalPressure inletOutlet
SPAR-Buoy movingWallAndFrameVelocity fixedFluxPressure zeroGradient
Bottom noslip fixedFluxPressure zeroGradient

the patch faces. The inletOutlet boundary condition is normally the information (U) travels across a computational grid cell during the unit
same as zeroGradient, but it switches to fixedValue if the velocity timestep. If the Courant number is more than one, the information
vector next to the boundary aims inside the domain (reverse flow). propagates through more than one grid cell at each time step, causing
The movingWallAndFrameVelocity boundary condition was created in- the solution to be inaccurate and potentially resulting in nonphysical
house and is utilized for translation and rotational DoF motions of the results. For adequate resolution and capture of the transient flow field
body. This boundary condition corrects the flux due to mesh motion in
in ocean wave modeling, relatively small Co numbers are necessary.
such a way that the total flux through the moving wall is always zero.
The key advantage of the interFoam solver’s PIMPLE algorithm is the
The fixedFluxPressure boundary condition sets the pressure to ensure
that the flux on the boundary is as defined in the velocity boundary possibility to use large Co numbers. When smaller Co numbers are
condition and the totalPressure boundary condition sets the pressure to necessary to accurately capture minor variations in the fluid, then the
zero at the atmospheric boundary. More detailed descriptions of the benefit becomes less meaningful. In the present study, the maximum
boundary conditions can be referred in OpenFOAM (2017). Co number of 0.5 was adopted along with 10 outer (SIMPLE) pressure-
momentum corrections, 3 inner (PISO) pressure correction loops have
3.2.3. Solver settings been adopted. The adopted residuals limit in the present study was 10−8
The Courant–Friedrich–Levy number (Co) is often used in CFD for all simulations. Table 3 summarizes the methods of discretization
solvers for temporal discretization and it indicates that the how much schemes in space and time used in the present study.

8
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Table 3 Table 4
Numerical schemes. Progressive wave test case: wave parameters with LC 1 represents regular waves and
Term Scheme Order LC2 represents irregular waves.
Parameters Value
Gradient Gauss linear Second
Divergence Gauss vanLeer Second LC 1.1 LC 1.2 LC 2.1 LC 2.2
Laplacian Gauss linear corrected Second
Wave period (𝑇 or 𝑇𝑝 ) [s] 1.9 1.58 1.72 2.37
Time CrankNicolson Second
Wave height (𝐻 or 𝐻𝑠 ) [m] 0.187 0.132 0.145 0.273

Table 5
Progressive waves test case: Test matrix for convergence study.
Mesh level Mesh size (𝛥x, 𝛥z)
Very Coarse 𝜆/32, H/2
Coarse 𝜆/64, H/5
Medium 𝜆/100, H/8
Fine 𝜆/150, H/10
Very Fine 𝜆/200, H/15

3.2.5. Mesh validation: Wave propagation

The fundamental prerequisite for wave-structure interaction simu-
lations is that the incident waves have to be of accurate quality. This
section validates the incident waves for various spatial and temporal
discretization combinations. In total, four-wave test cases are investi-
gated for SPAR interaction, as indicated in Table 4. Regular waves are
represented by LC 1, while irregular waves are represented by LC 2
(generated based on the Bretschneider spectrum). The chosen cases in
Table 4 are based on the number of cases attempted in the experiment,
which is likewise investigated with and without wind factors. In the
present cases, LC 1.1 represents a severe situation in regular waves,
Fig. 6. Mesh improvement using different types of refinement over the SPAR body.
while LC 1.2 represents a normal sea state. For irregular sea states,
LC 2.1 and LC 2.2 can be portrayed as normal and severe sea states,
respectively.
3.2.4. SPAR body mesh validation: Hydrostatics The mesh and temporal convergence studies for all four test cases
The mesh was created with OpenFoam’s SnappyHexMesh tool. Near for wave propagation in a 2D NWT are discussed in this subsection.
the SPAR, where boundary layer separation occurs, cells were refined This 2D study’s computational domain is depicted in Fig. 8. The domain
into much smaller ones, and refinement boxes were formed at varying is a two-dimensional plane that is an exact reproduction of the three-
distances from the SPAR to minimize a rapid shift from very fine grids dimensional domain illustrated in Fig. 5, but without the structure. The
to coarser grids. This is critical as the inadequate meshing will lead to regular waves (LC 1.1 and LC 1.2) are generated utilizing the foamStar’s
loose body volume, causing the change in the draft and subsequently built-in Stream function wave theory and periodic boundary conditions.
affecting the hydrodynamic characteristics. Thus, the accuracy of the HOS-NWT is used for irregular waves (LC 2.1 and LC 2.2). The HOS-
meshed SPAR volume (𝑉𝑂𝐹 ) is estimated and then compared to the NWT exactly reproduces the experimental conditions paddle motion. As
𝑉 −𝑉
physical volume of the SPAR (𝑉𝑃 ) for the error ( 𝑃 𝑉 𝑂𝐹 ). The SPAR can previously mentioned, in the 𝑅0 reference system, the origin is different
𝑃
be refined in several ways to enhance the accuracy, including refining for the HOS domain and CFD domain. The HOS domain (𝐿𝐻𝑂𝑆 ) has
the background mesh (BlockMesh), surface refinement, and tapering a full-scale length of 50 meters and a water depth of 5 m. The full-
edge refinement as shown in Fig. 6. Figure depicts a typical five-step re- scale length of the CFD domain is 1𝜆 + 6𝐷𝑆𝑃 𝐴𝑅 + 1𝜆, and its origin is
finement for various background mesh cell configurations (expressed in positioned from 𝑥𝐻𝑂𝑆 = 17𝑚 at the inertial frame origin. The relaxation
terms of SPAR diameter). It can be seen that as expected finer refining zones length is set to 1𝜆 for both inlet and outlet, respectively. To
reduces the error to zero. To further comprehend the refinement aspect, be noted, for irregular wave generation cases, the HOS domain must
the new parameter PPSD (Points Per SPAR Diameter) is introduced. entirely incorporate the CFD computational domain into it.
The fine mesh configuration has been chosen (it has accuracy of more Upper zone, free surface zone, and lower zone are the three zones
than 99% for even coarser background cells). Based on these results, that make up the computational domain in the vertical direction
the number of points per SPAR diameter is increased, and the result (Fig. 8). Between 𝑧 = −𝐻𝑠 and 𝑧 = 𝐻𝑠 is the free-surface refinement
is illustrated in Fig. 7(b). These PPSD results are also compared in region. According to Kim (2021), the cell aspect ratio (𝛥𝑧/𝛥𝑥) in the
terms of the number of cells that are created, in addition to the error free-surface refinement region is maintained as 0.5, and the outer
comparison. As demonstrated in Fig. 7(b), the PPSD of 35 has been zones are kept at ratio four. To understand the solver’s convergence,
found to be cost-effective (in terms of the number of cells) with a several discretizations are used to investigate the computing grid and
volume loss of less than 0.1% and hence it is adopted in the present time steps. Table 5 shows the cell length 𝛥𝑥 and height 𝛥𝑧 that were
study. examined. Based on the wave period (𝑇 or 𝑇𝑝 ) considered, different
Hydrostatic characteristics are used to validate the new mass (which temporal resolutions for wave propagation are performed for each cell
must be revised to account for the little volume lost while maintaining arrangement ( Table 5), such as T/100, T/200, T/400, and T/800. The
the draft measured experimentally) and other mass attributes. The grid and time steps are configured to modify the cell size while keeping
SPAR model has been placed in the NWT, allowing it to settle and an the time step constant, and the grid is fixed while the time step is
equilibrium position has been established in all six degrees of freedom. only varied in the second set of investigations. From Mesh32TH2T100
A slight mass adjustment has been made to bring all motions (especially through Mesh200H15T400, eighteen combinations of mesh and tem-
heave) within 10−6 m, and the investigation was done for 50 s, proving poral resolutions are employed. Mesh32TH2T100 indicates that there
the optimal configuration for further investigations. are 32 cells per wavelength in the wave propagation direction, 2 cells

9
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 7. Validation of SPAR underwater volume for different mesh refinements.

Fig. 8. Sketch of the HOS and CFD domains for wave propagation validation in 2D NWT and the position 𝑥𝐻𝑂𝑆 = 17 m is identical to 𝑥0 = 0.

per wave height in the vertical direction, and 100 time steps per where 𝜂𝑥𝑦 (𝑡) represents two real continuous surface elevation of HOS
wave period. With all of the above factors mentioned, the outcomes and any of the foamStar wave case. The maximum cross-correlation
or validation of the regular wave is described initially followed by the coefficient of +1 indicates a perfect match between the foamStar and
irregular wave. the HOS wave, but a difference in the coefficient other than +1 reveals
Regular wave validation: For the regular wave validation, foam- the quantification of inaccuracy between the signals. Typical cross-
Star computational domain was considered with periodic boundary correlation results for LC 2.1 and LC 2.2 is shown in Fig. 10. Although
conditions. In this study, the duration of the simulation is kept at the experiment and HOS were conducted for a total of 600 s, to avoid
35 wave periods for investigation. Figs. 9(a) and 9(b), presents the the large computational cost, the comparisons were carried out for 100s
results of the simulation in the case of LC1.1. Similar observations (i.e. around 50𝑇𝑝 ). Eighteen different configurations were carried out
are noticed for LC 1.2 and hence results are not reproduced here. for both LC 2.1 and LC 2.2, as indicated by the circles and squares
Fig. 9(a) displays the peak wave amplitude (𝜂𝑚𝑎𝑥 ) for each combination in Fig. 10. The lines connect meshes that are of the same type but
of mesh and time step measured at nearly 20𝑇𝑝 and Fig. 9(b) represents with varying temporal discretization. The findings demonstrate that
the first harmonic amplitude variation for the whole duration of the even the coarser combinations have an accuracy of approximately 0.97
wave and its damping behavior. In both the figures, the parameters and that refining the discretizations will bring the accuracy closer to
clearly demonstrates convergence when modifying the mesh and time 1. For the wave structure interaction study, the mesh combination of
step. Thus, a cell size of 150 cells per wavelength, 10 cells per wave Mesh150H12T400 which is nearly on and above 0.99 in both load cases
height, and 400 time steps per wave period seem to be computationally is proven to be an affordable combination and hence this combination
efficient and considered for further analysis. will be used in the present study.
Irregular wave validation:
3.2.6. Irregular wave Validation: HOS wave Vs Experiment
The irregular waves for various combinations of spatial and tem-
It is important to remember that the HOS-NWT formulation was
poral discretization (as in regular waves) measured at WG5 were
effectively validated with studies on various cases during the last
used in the following analysis. The results from the foamStar are
decade (Ducrozet et al., 2012, 2006; Canard et al., 2020; Hasan et al.,
cross-correlated with the HOS simulations to investigate the spatial
2019). LC 2.1 and LC 2.2 imply a wide variety of wavemaker motions
and temporal convergence ( Table 5) . The suggested technique has
with large nonlinearities in order to validate our current concept and
the advantage of being able to assess the time shift between signals
implementation. The experimental paddle motion has been used in the
as well as the similarity between them (Fernández et al., 2014). In
HOS-NWT to simulate the wave characteristics and compared with the
the present case, the cross-correlated time series is normalized with
experimental measurements. For this simulation, 325 modes in the flow
auto-correlation coefficient at zero lag such as,
direction (𝑁𝑥 ) and 33 modes in the vertical direction (𝑁𝑧 ) have been
𝜂𝑥𝑦 (𝑡) used. Throughout the simulation, the HOS order is kept at 3. Figs. 11(a)
𝜂𝑥𝑦 (𝑡) = √ (29)
𝜂𝑥𝑥 (0)𝜂𝑥𝑥 (0) and 11(b) shows the comparison of the HOS-NWT wave generation and

10
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 9. Progressive waves test case: Harmonic amplitudes of the free surface elevation at the WG5 for different mesh and time configurations.

Table 6
Parameters of the mooring lines in prototype scale.
Parameters Units Prototype Values
(Scale 40)
Diameter of chain [m] 0.266
Mass density [kg/m] 440
EA [N] 6E9
Added mass coefficient 1.0,0.0
Drag coefficient 1.8, 0.25
Clumped mass (sensor) [kg] 2.75e4
UnstrLen (Major and Bridle) [m] 638.2, 48

are symmetric, the figure shows a single bar representing both lines. For
Line 2∕3, the difference between the models is quite small, but for Line
Fig. 10. Cross-Correlation Coefficient for LC 2.1 and LC 2.2 for different mesh and
time combination. 1, the difference is nearly 2% of the pretension between the models. It
is realized that the original iteration of the connection node position
displaces into a new position, resulting in the Line tension difference.
the experiment at the WG2 probe location, respectively for both the LC. The dynamics assumed by MoorDyn could also be a concern. However,
In both figures, we can see that the agreement between the experiment because the error is so minimal, the MoorDyn model is agreed upon for
and numerical models is identical. After 50 s, however, slight deviations its robustness.
in the peak magnitude of the crest or trough were noted. This error is For the final validation, the mooring setup used in the present
clearly connected with wave amplitude, indicating high nonlinearity study, the crowfoot model having 3 main lines connecting 6 bridle
but the cross-correlation coefficient for these simulations is around lines with the load sensor (LS1 to LS3) as shown in Fig. 4 are tested.
0.994, which makes the discrepancy quite small, hence disregarded. This configuration is validated in a standalone MoorDyn model for a
It should also be noted that these deviations in the incident wave will total of 9 mooring lines and 9 nodal locations, and the typical setup
have a minor impact on the motion responses in wave-floating body is as illustrated in Fig. 13. It is important to validate input character-
simulations. istics such as connecting nodes, unstretched length, mass, and stiffness
properties from the experiment. The primary mooring properties from
3.2.7. MoorDyn Validation the experiment are listed in the Table 1. In order to obtain proper
MoorDyn is a dynamic mooring solver based on lumped mass ap- tension, in the numerical model symmetry adjustment of the anchor
proach. The version of MoorDyn used in this paper was v1.01.03C. and fairlead positions (120◦ to each main line) are considered, because
The solution for the mooring solver is based on a three-step process. of the fact that achieving symmetry while doing physical tests was
First, it generates a Quasi Static model based on the input conditions. difficult. The axial stiffness (AE) and equivalent diameter are obtained
Second, to allow the system to achieve the initial equilibrium, dynamic from ORCINA recommendations (ORCINA, 2014). In the transverse
relaxation will be carried out, which includes dynamic drag and accel- direction(Eq. (25)), quadratic damping coefficients are set to 1.8, and in
eration forces. Finally, tension at each fairlead as well as the motion the axial direction (Eq. (26)), they are set to 0.25 based on the numeri-
(position, velocity, and acceleration) of each node at each time step cal analysis. The added mass coefficient in the transverse direction are
will be evaluated. set to the value of 1.0. The convergence study for a number of segments
In the present study, the standalone version of the MoorDyn model in the mooring line was studied, and it was determined that from the
was first validated for a static problem, and compared with other quasi- 50 segments in the main line and 4 segments in the bridle were the
static models such as MoorPy (Hall et al., 2021), MAP (Masciola, 2018), most efficient and converged solution begins, anyhow with a factor of
and AQWA. The single delta line is the subject of the first validation, as safety of 1.5, 75-6 combination was chosen.
it is one part of the mooring configuration used in the SPAR interaction Although MoorDyn is simple in concept, its practical implementa-
study. Masciola et al. (2013) can be used to look up the nomenclature of tion can be difficult, as it is sensitive to the input parameters chosen,
the single delta line considered. This line’s properties and dimensions causing numerical instability (Paduano et al., 2020). Similar instability
are taken from the MAP examples section. This delta line introduces is observed in the present study when working with the mooring
the connecting node between the lines, the position of which will be properties on experimental model values. However, using the prototype
iterated by the solvers, and the accuracy of the pretension is dependent values shows that numerical instability can be treated. Hence the
on the final node’s position. The static pretension comparison between mooring properties used in the experiment are modified to prototype
the solvers for the case is as shown in Fig. 12(a). Since Lines 2 and 3 values (1:40 scaled model of FOWT) using Froude and Cauchy scaling,

11
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 11. Comparison between experimental and HOS -NWT numerical time series of wave elevation for the two Irregular wave test cases - LC2.1 and LC 2.2.

Fig. 12. Comparison of line tensions along single delta line (left) and complete delta type mooring(right) for different mooring models.

to note that the restoring force stated here refers to the overall restoring
of all 9 lines, not just the catenary line.

3.2.8. Stiffness matrix coefficients

The stiffness coefficient as determined by the pull-out test for surge
(𝐾11 ) mooring lines is 75𝑁𝑚−1 . Similarly, the sway (𝐾22 ) and yaw
coefficients (𝐾66 ) were calculated using similar conditions and also
modeling the mooring lines in ORCAFLEX (ORCINA, 2014) for valida-
tion. The values are 80𝑁𝑚−1 and 127𝑁𝑚 𝑟𝑎𝑑 −1 , respectively for surge
and yaw. Non-diagonal stiffness in terms of restoring stiffness in heave,
roll, or pitch from the mooring appeared to be extremely minimal in
this investigation. In the case of solving wave structure interaction with
stiffness matrix, the mooring line mass of 11.4 kg (±1 kg) is added to
Fig. 13. Typical delta mooring model adopted in the present study with its top the SPAR’s total mass, which is then fine-tuned to maintain the same
view(left) and isometric view (right). draft.

4. Results and discussion

and the updated mooring properties are shown in Table 6. Similar
setups were also built in various mooring models, and pretension at This section discusses the validation of the developed numerical
the static conditions for major lines Line 1 to 3 (Fig. 4) are compared approach with experimental measurements. The first moored decay test
as shown in Fig. 12(b). The pretension aligns well with MAP, and only is discussed, followed by the interactions of regular and irregular waves
minor variations were noticed compared to MoorPy and AQWA proving with SPAR for mild and severe sea state conditions.
the efficiency of the MoorDyn and input conditions considered. The
final system of coupling considered is as shown in Fig. 14, wherein 4.1. Moored Decay
the mooring analysis will be carried out in prototype conditions for
motions and velocities and the computed forces are scaled-down and In the coupled foamStar-MoorDyn solver, two independent DoF
coupled with the CFD. decay tests are carried out with delta mooring lines. The SPAR is
Finally, the MoorDyn results are validated from the experimental perturbed along the DoF of interest, then released to move freely from
pull-out test for the estimation of stiffness. To characterize the hori- that point. Surge and heave are the decay tests performed. The mooring
zontal mooring stiffness for the platform surge DoF, various pull-out tensions and their accompanying modeling results are discussed.
tests were done in the experiment. One such test comparison with
experiments is shown in Fig. 15, wherein the mooring stiffness is nearly 4.1.1. Surge decay
75𝑁𝑚−1 based on the slope of the force displacement relationship. The The initial displacement for the Surge in foamStar is accomplished
mooring restoring force can be quite similar to a linear force, which is using the moveDynamicMesh utility, which only conducts the dynamic
interesting for a catenary system since the total restoring force in Deep- mesh motion without solving the flow field while monitoring the mesh
CWind campaigns and the related OC4 numerical model (Robertson quality aspects. The SPAR floater with delta mooring lines is moved
et al., 2014) tends to be nonlinear at larger offset values. It is important 0.11 m and then released to move freely from its initial position. The

12
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 14. Flow chart of the short algorithm between the foamStar (in model scale) and MoorDyn library (in prototype scale).

the heave direction) for the present heave decay. The motion recorded
in both mooring scenarios is shown in Fig. 17(a). The figure has a
few noteworthy observations as follows. The mooring in the form of
a stiffness matrix component can be thought of as having no resistance
in the heave direction (𝐾33 ). As a result, the motion characteristics
only depict the body motions for a given mass and displacement. In
this situation, the natural period appears to be 4.91 s, and the damping
ratio appears to be 1.5%. Incorporating the mooring into the problem
changes the damping of the problem by adding 0.5%–0.6% to total
damping, resulting in a natural period of 4.8 s. The natural period
in the experiments was determined to be 4.8–4.9 s, and the damping
was found to be similar to that of the moored simulation. Fig. 17(b)
compares the numerical tension of Line 1 (at LS1) with the experiment,
Fig. 15. Comparison of the results of the Pullout test in the MoorDyn solver with the to illustrate the effect of mooring tension variation. The vertical axis is
findings of the experiment. kept the same as in the surge free decay tension figure to show the
reader that the influence in the heave is too small when compared to
the surge but cannot be ignored.
COG is set to a displaced position, and the mooring lines are given
4.2. Wave Interaction with SPAR
initial displacement and zero initial velocity and are free to move with
the body motions. The surge had a strong influence on the pitch and a
In this section, the performance of the foamStar and MoorDyn
weak influence on the heave, making it difficult to estimate the proper
coupling in reproducing the experiments on wave interactions with
surge natural period. Hence only for this testing, only Surge DoF taken SPAR are reported. The wave interaction’s computational domain and
into account. The findings of the motion shown in comparison to the relaxation zones are the same as those described in Section 3. A
experiment (Figs. 16(a) and 16(b)), as well as the mooring tension in common Mesh150H12T400 is chosen in order to make a similar dis-
Line 1(ref Fig. 4 for Line nomenclature), appear to be quite good. All cretization for regular (Mesh150H10T400) and irregular waves studies
of the line tensions were in good agreement, but 𝐿𝑖𝑛𝑒1 was the most (Mesh150H12T400), and it will be represented in the results with the
critical (as it is in every test scenario), therefore it is only shown. When prefix ‘fs’ denoting foamStar. The Fig. 18 depicts a typical compu-
the SPAR is shifted towards 𝐿𝑖𝑛𝑒1 for initial displacement, there is a loss tational domain with mooring lines. The fine refinement in the free
of up to 8𝑁 in pretension at first, and the variation approaches ±10𝑁. surface and near the SPAR can be seen in this figure, along with the
Surge’s natural period is found to be 18.82 s, while the experiment’s typical contour of the free surface. The SPAR body is fine-tuned based
natural period is 18.6 − 18.7 s. The overall damping ratio is calculated on the previous estimates in order to maintain PPSD 35 and a little
using the logarithmic decrement method, and it is found that the mass modification to ensure the same drafts. Based on the wave being
percentage changes between 2.2% to 2.3% for the experiment, but from considered, the wavelength will vary for each load case, influencing
the numerical simulation it is between 2.35% to 2.45%. In the instance the number of cells. In the present study of LC 1.1 to LC 2.2, it varies
of the surge decay test, the overall difference is found to be less than from 1 to 2.5 million cells, and the code was executed concurrently
1%, indicating that the coupling model performs well for both motion using MPI on the LIGER supercomputer at Ecole Centrale Nantes. The
and tension estimation in the surge decay problems. duration of the simulation as a whole varies from case to case. Only 25𝑇
is kept as the simulation duration for regular wave cases (LC 1) because
the steady state is reached within 4 to 5 peaks of the simulation, but
4.1.2. Heave decay 50𝑇𝑝 has been kept as the simulation duration for irregular wave cases
The heave decay follows a similar procedure as that of the surge (LC 2). One CPU per 30000 cells is chosen for parallel processing and
simulation, with the SPAR being moved in the heave direction for an computational time varies from 10 h in LC 1 to 36 h in LC 2. Each
initial displacement (in this case 0.045 m) and then released in calm LC involves two methods of coupling to handle the mooring line, one
water. The decay simulation in heave is carried out with all 6 DoF in the form of a stiffness matrix and the other with dynamic mooring
motions active since the heave motion has only slight influences from (MoorDyn). As previously indicated, two regular wave test cases (LC
all other motions. Also note that the mooring in the form of stiffness 1.1 and LC 1.2) and two irregular wave test cases (LC 2.1 and LC 2.2)
matrix was examined as well (although it did not have any stiffness in are being carried out and discussed in the following sections.

13
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 16. Comparison between experimental and numerical time series of surge decay test with its corresponding DoF and mooring tension.

Fig. 17. Comparison between experimental and numerical time series of heave decay test with its corresponding DoF and mooring tension.

behavior will be similar to Line 2. When the trough reaches the SPAR
at the final instant (10𝑇 ), it gently returns to its previous motion, which
is similar to 9.25𝑇 , lessening the tension in Lines 2 and 3 and achieving
equilibrium before the next cycle begins.
The comparison between experiments and numerical simulation
for wave probe WG4 is shown in Fig. 20. Out of all the probes, the
probe WG4 which is at the same location along the SPAR is shown
for the comparison. The initial ramping in the experiment is different
from the numerical ramping leading to a small difference up to 10 s.
After achieving a steady-state, the wave appears to achieve suitable
amplitudes and is in phase with the experiment. For both the LC,
the total duration of simulation recorded in the experiment is 30 s.
Both the LC 1.1 and 1.2 were initially evaluated for 30 s, however, to
Fig. 18. Typical representation of computational domain in the wave structure
interaction study with mooring. quantify the surge effects (Surge natural period roughly 19 s) the LC 1.1
simulations were extended for additional 15 seconds. The numerical
and experimental time series for surge, heave, and pitch angles (LC 1.1
and LC 1.2) are shown in Fig. 21. To demonstrate the influence of LC
4.2.1. Regular wave interaction
1.1 over LC 1.2, the vertical axis limit is kept constant, indicating that
The hydrodynamic responses of the SPAR under regular waves are
investigated in this section. All the waves propagate in a positive LC 1.1 has larger amplitudes than LC 1.2. Sway, roll, and yaw motions
surge direction. The waves with moderate steepness (𝐻∕𝜆 = 0.03) were not shown since they were negligible, both in the numerical
are simulated and the wave characteristics areas listed in Table 4. At simulations as well as in the experiments. The surge motions shown
various time instants, a typical interaction between the wave, structure, in Figs. 21(a) and 21(b), appears to oscillate with increased amplitudes
and mooring is demonstrated in Fig. 19. The wave LC 1.1 (𝑇 = 1.9 s) has at the ramping time itself before the steady-state due to the difference
an initial ramp of 10 s, hence the steady-state time instant of 17.5s to in the ramping. The rate of change of oscillation amplitude compares
19 s was chosen for the presenting outcome. The magnitude of velocity well to the surge experimental period. In the case of LC 1.2, the
in the 𝑥-direction as the wave progresses is shown in red and blue on surge motion follows a similar trend as LC 1.1, albeit with a much
the graph. For clarification, the free surface contour is also displayed smaller amplitude. The behavior of the stiffness matrix component
(𝛼 = 0.5). The mooring is represented by segment tension, which varies and MoorDyn is virtually identical, although MoorDyn has a very
from 18N to 45N and is colored blue to red. The crest of the wave slightly higher amplitude. This is due to the fact that MoorDyn includes
approaches before SPAR(already has its own motion during transient coupling of all DoF, whereas, in the stiffness matrix approach, we have
and few wave periods before) at 9.25𝑇 . Line 1 has the highest mooring considered only three DoFs. The heave and pitch motions of the LCs
tension and Lines 2 and 3 have slightly lower tension. As the wave are depicted in Figs. 21(c) and 21(d), and Figs. 21(e) and 21(f). The
progresses, the crest reaches the SPAR (9.5T) with an impact, (9.75𝑇 ) motions were well simulated by both the stiffness matrix approach
then the SPAR moves along with the wave orbital, reducing the tension and MoorDyn and both the motion amplitudes are quite small, with
in Line 1 to 30-35N and causing significant tension in Line 2 and Line heave amplitudes of nearly 2 cm and 1 cm and pitch amplitudes of
3. Line 3 is on the opposite side of the figure, making it difficult to 1.25◦ and 0.8◦ . From the results , it proves that the accurate stiffness
visualize in color. Because Line 3 is symmetric, it can be said that the matrix components are sufficient to handle the problem for this small

14
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 19. Typical representation of different time instants of the interaction of regular waves with the moored floating SPAR using foamStar-MoorDyn coupling for case LC 1.1.

Fig. 20. Comparison between experimental and numerical time series of wave elevation at WG4 of the Floating SPAR for the two regular wave test cases — LC1.1 and LC 1.2.

amplitude of motions (despite LC 1.1 having higher amplitude regular Lines 1, 2, and 3 are considered, and since the bridle lines had no ten-
waves). Thus, dynamic mooring analysis may not be required in such sion measurements, they were ignored. In the MoorDyn, the top most
cases. However, MoorDyn has an advantage if the requirement is for segment tensions in the major lines were used for comparison, which
the estimation of restoring force or the mooring tension assessment. was near to LS in the experiment. The experimental data (sampled at
Fig. 21(g) and Fig. 21(h) represent the wave excitation forces over the 200 Hz) was subjected to an average filter, whereas numerical results
floating SPAR for the load cases LC 1.1 and LC 1.2 respectively. As are saved at a sampling rate (as determined by the CFD solver) with
stated in Section 2.3, the total force acting on the SPAR is influenced no filter. The figure shows good comparisons and it is worth noting
by a number of factors like mass, buoyancy, gravity, mooring, fluids that Line 1 is at the receiving end of the wave, at the back of the
(including pressure and viscosity), and so on. Using fluid forces, we SPAR. As a result, the tension oscillations are up to four times larger
can derive solely the wave excitation forces operating over the SPAR
than the oscillations of Lines 2 and 3 at the SPAR’s front junction.
by taking into account the normal pressure acting over the SPAR and
Another point to note is that LS2 and LS3 are symmetric, therefore the
its tangential viscous component.
tension and variation are identical in both the lines. Despite the input
Following Huseby and Grue (2000), the wave excitation forces
conditions, LC 1.2 tension amplitudes are small when compared to LC
(removing the hydrostatic forces), are normalized by 𝜌𝑔𝑆 𝐻2 , where 𝑆
1.1, it demonstrates that foamStar-MoorDyn coupling could solve for
represents the cross-sectional area of the SPAR. Because the motions
both the smaller and higher amplitude regular wave interactions with
closely match the experiment, the forces calculated by the solver should
be fairly accurate, and also the numerical cases (stiffness matrix form the floating body. Similar observations for motions and tensions were
and MoorDyn form) have similar motions, only results from one form also seen with the quantitative analysis as presented in Section 4.2.3.
(MoorDyn) is presented. The obtained force trend is proportional to Hence concluding that the regular wave generated inside the foamStar
each case’s wave elevation, and the 𝑥 component of the force (𝐹𝑥 ) am- by the CN-Stream function wave model has been evaluated for floating
plitude appears to dominate the vertical force (𝐹𝑧 ) by closely 1.5 times. SPAR interaction with mooring models, and it appears that the coupling
Otherwise, the normalized amplitude has only a 5%–7% difference is effective in accurately predicting the motions and forces. Despite the
between LC 1.1 and LC 1.2. accuracy, validation of coupling with HOS-NWT is the next requirement
Fig. 22 shows the line tensions from both LC. It should be noted that as this way multiple types of waves can be generated, making the model
the experiment uses three load sensors at the LS position, therefore only more robust.

15
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 21. Comparison between experimental and numerical time series of the surge, heave, and pitch of the floating SPAR and wave excitation forces for the two regular wave
test cases — LC1.1 and LC 1.2.

4.2.2. Irregular wave structure interaction wavelength and 12 cells per wave height, 400 timesteps per peak wave
Now the SPAR response was investigated for the uni-directional period (fsM150H12T400). The time history comparison between the
irregular sea state generated based on the Bretschneider spectrum foamStar and experimental measurements for WG4 is shown in Fig. 23.
(Table 4) in HOS-NWT. Two different irregular sea state was analyzed, Unlike regular waves, here the experimental wave paddle motion has
Case LC 2.1 is a relatively mild sea state, and Case LC 2.2 corresponds been used, so there is no ramping issue in both the test cases. However,
to severe sea state conditions. HOS-NWT is used to simulate the in-
it should be noted that the HOS has few discrepancies in comparison
cidence wave field in the whole wave tank based on the motion of
with the experimentally measured wave for some amplitudes and that
the wavemaker from the experiment. The simulation has been carried
these discrepancies are also seen in the foamStar produced waves. Also,
out for 500–600s in the experiment and HOS, however in the CFD,
it is arranged to keep 50 𝑇𝑝 to restrict the computational cost. The as noted in the parametric study, discrepancies created owing to mesh
computational domain is comparable to that of the domain used in and time discretization in foamStar could also be added to the total
regular wave interaction study, with the exception that 𝜆𝑝 instead of 𝜆 discrepancy. However, the difference is very small in the case of a small
has been adopted. Based on the peak wavelength and significant wave amplitude case (LC 2.1) and a little higher for a higher amplitude case
height (𝐻𝑠 ), the domain zones were updated to maintain 150 cells per (LC 2.2). It should also be emphasized that the waves recorded are with

16
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 22. Comparison between experimental and numerical time series of tensions of the front, line 1(LS1), and rear mooring lines, line 2 and line 3 (LS2 and LS3), for the two
regular wave test cases — LC1.1 and LC 1.2.

Fig. 23. Comparison between experimental and numerical time series of Wave elevation WG4 of the Floating SPAR for the two Irregular wave test cases — LC2.1 and LC 2.2.

the SPAR in place and hence the influence of scattered waves from the stiffness matrix. However, for the case of LC 2.2., the comparisons were
SPAR will also contribute to the wave elevation recorded. good at the beginning, but when the wave amplitudes are increased, the
The SPAR response in the surge, heave, and pitch for the cases of stiffness matrix approach leads to more deviation modifying the SPAR
LC 2.1 and LC 2.2 are shown in the in Fig. 24 and mooring tensions response. However, the fsM150H12[MoorDyn] prediction compared
from the corresponding cases are shown in Fig. 25. The influence well with the experimental surge motion. For the heave (Fig. 24(c),
of the two mooring approaches is discussed similar to regular wave Fig. 24(d)) and pitch (Fig. 24(e), Fig. 24(f)) motions also, mild case (LC
interaction with SPAR. In the case of LC 2.1, there were very few to no 2.1) compared well with the experiment but for the severe case (LC 2.2)
discrepancies between the motions predicted in surge, heave, and pitch stiffness matrix approach has certain discrepancies. The differences are
with the experiment. Fig. 24(a) and Fig. 24(b) represent surge motion also identical to the wave amplitude variations reported. The MoorDyn
recorded for LC 2.1 and LC 2.2. For LC 2.1, fsM150H12[MoorDyn] overpredicts the pitch by a few decimals in the amplitudes and this little
has predicted comparable motions using the mooring as similar to the over prediction may be seen throughout the LC, although it is of modest

17
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 24. Comparison between experimental and numerical time series of the surge, heave and pitch of the floating SPAR and wave excitation forces for the two irregular wave
test cases — LC2.1 and LC 2.2.

size. The cause for this might be that the mooring configuration in As noted in the regular wave cases, the force trend corresponds very
the experiment is not perfectly symmetric, resulting in small unsettled well to the wave elevation, and the amplitude ranges are confined
moments in the pitch direction. Also, with all of the nacelle RNA between ±0.5 for both the load cases. Overall, the range of forces with
supports, it is possible that the 𝐼𝑦𝑦 provided by the experiment is not respect to wave height demonstrates that regular waves develop nearly
quite accurate. Overall, the findings showed that the coupled model’s 1.5 times higher force range than the irregular waves, whereas irregular
motion prediction is precise for both the sea state conditions. The waves display peak amplitudes at certain locations exceeding 0.9–1,
stiffness matrix approach prediction is very good for mild sea states, but indicating the strength of its impact.
for severe sea states, a dynamic mooring model is required for accurate The abrupt change of tension (Fig. 25(d) to Fig. 25(f)) happening
motion prediction similar to that of the experiment. A similar conclu- in very short interval of time for the LC 2.2 required robust mooring
sion is quantitatively obtained in Section 4.2.3. The Fig. 24(g) and model to capture tensions accurately. In the case of LC 2.1, the tension
Fig. 24(h) represents the normalized hydrodynamic force components forces match quite well in all three lines. In large surge drift offset
obtained over the SPAR for the cases of LC 2.1 and LC 2.2, respectively. circumstances, the most substantial tension is evident in the seaward

18
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 25. Comparison between experimental and numerical time series of tensions of the front, line 1 (LS1), and rear mooring lines, line 2 and line 3 (LS2 and LS3), for the two
irregular wave test cases — LC2.1 and LC 2.2.

cable. The force amplitudes of the leeward and seaward mooring lines and Figs. 26(a) and 26(b) illustrate this discrepancy between the signals
are identical in amplitude when there is little or no drift. For LC 2.2, the and the experiment for both the stiffness matrix and dynamic mooring
predicted mooring tension appears to be in phase with the experiment, cases. In the stiffness matrix cases, the coefficient are 0.89 and 0.93
however, the amplitudes are slightly higher than expected, the reason for LC 1.1 and 1.2, respectively. However, a minor improvement is
has been addressed earlier. The amplitude variation of tension between seen with MoorDyn (coefficient of 0.94 in both cases), even though the
15 to 25𝑁 in a short time frame of less than a second (LC 2.2) is 3 ramp effect is still present. The stiffness matrix approach achieved a
to 4 times higher than any variations reported earlier (particularly LC good comparison of 0.98 for the mild sea condition LC 2.1 (Fig. 26(c))
2.1). The coupled solver captures all of these tension variations with but lost accuracy to 0.86 for LC 2.2 (Fig. 26(d)). With values higher
little to no inconsistencies, showing the efficiency of the coupling. In than 0.96 in both LC 2.1 and LC 2.2, MoorDyn appeared to perform
comparison to all the cases reported, the LS2 and LS3 lines experience well, supporting the validity of the discussions from earlier sections.
more impact tension variations only in LC 2.2 which is also accurately In LC 1.1, LC 1.2, and LC 2.1, heave and pitch motions are well
captured by the coupled solver. reproduced by the numerical model, with coefficient ranges of 0.99
and higher by both the approaches. While MoorDyn appears to perform
4.2.3. Quantitative analysis well in LC 2.2, reaching a coefficient up to 0.96 and above, the stiffness
The qualitative comparison presented in the previous sections is matrix cases are less accurate, and the coefficient stays at 0.9 to 0.94
quantitatively compared in this section using a cross-correlation analy- (Fig. 26(d)). The mooring tensions appear to match well in all the cases
sis. The coefficient is computed with Eq. (29), but the motions and the with the coefficients in the range of 0.99 to 0.999 (Fig. 26), proving the
mooring tensions of the numerical model are compared with the exper- efficiency of the mooring model and its coupling.
iment instead of the surface elevation. Recalling from Section 3.2.5, the
maximum cross-correlation coefficient of +1 indicates a perfect match 5. Conclusion
between the foamStar and the experimental results, and the distance
to 1 quantifies the difference between the signals. The surge motions The paper presents the development of a coupled numerical model
in the regular wave were sensitive to the ramp in the wave generation, that can simulate the station keeping of real-time floating offshore

19
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

Fig. 26. Relative analysis using Cross Correlation between foamStar cases and Experiment where [K] represents stiffness matrix approach, MoorDyn represents dynamic mooring
coupled approach.

structures in both normal and severe sea state conditions. The efficiency wave cases with mild and moderate amplitudes. However, to address
and safety of the floating offshore structures have direct implications mild, moderate, and even severe sea state conditions with both regular
from the structure’s elasticity, structural dynamics, hydrodynamics and and irregular waves, dynamic mooring simulation (MoorDyn) is proven
aerodynamics interaction. Hydrodynamics is among the most critical to be vital.
fields since it affects all other disciplines directly or indirectly. In the Hence, the developed coupled modeling concept can be used to
present study, the hydrodynamic interaction of fully nonlinear regular simulate the more complex wave-floating structure interactions. These
and irregular waves over a floating moored SPAR was simulated using capabilities will aid in the design and analysis of the floating structures
an OpenFOAM-based foamStar solver with the domain decomposition and the mooring to efficiently survive in extreme wave conditions.
technique. The CN-Stream wave model was used for regular wave Apart from hydrodynamics with rigid body motions, hydroelasticity
generation, while for irregular wave generation, the HOS-NWT was of the structure, aerodynamics for wind will be incorporated in the
used. The floating structure’s station keeping was characterized by a developed solver to extend its capability for practical aero-hydro-elastic
static (as a stiffness matrix) and dynamic mooring model (two-way applications
coupling with MoorDyn). The numerical model results were compared
with the experiment carried by Arnal (2020) in the ECN wave basin. CRediT authorship contribution statement
The physical model consists of the SPAR structure moored to the tank
bottom using delta type mooring with nine mooring lines. Sithik Aliyar: Conceptualization, Methodology, Software, Writing –
original draft. Guillaume Ducrozet: Data curation, Writing – original
Meshing a floating SPAR was a critical task since poor meshing
draft. Benjamin Bouscasse: Methodology, Writing – original draft.
would modify the volume of the body, altering the draft and pushing
Félicien Bonnefoy: Data curation, Writing – original draft. V. Sriram:
the motions to be irrelevant. Hence, a new parameter, PPSD (points
Writing – original draft. Pierre Ferrant: Writing – original draft.
per SPAR diameter) is introduced, and a value of 35 is recommended
for an accurate meshing of the floating structure to avoid volume
Declaration of competing interest
loss. Different types of mooring setups were investigated to see the
efficiency of the dynamic mooring model MoorDyn, and the results
The authors declare that they have no known competing finan-
were compared with other mooring models (MAP, MoorPy, AQWA)
cial interests or personal relationships that could have appeared to
along with experimental data. Since working in the model scale in influence the work reported in this paper.
the MoorDyn caused numerical instability in the delta type mooring
case, the MoorDyn solver was made to operate in the prototype scale. Data availability
Thus, forces from the mooring line is down-scaled and implemented
in the foamStar solver. Surge and heave decay tests were numerically The authors are unable or have chosen not to specify which data
simulated, and it is interesting to find that in the heave free decay, has been used.
mooring driven damping accounts for nearly 25% of overall damping.
In the case of surge decay, the natural period, damping, and moor- Acknowledgments
ing tension from the coupled model provided acceptable results in
the experiments. Further, good agreement between the numerical and This work has been performed in the framework of the Chaire
experimental motions (surge, pitch, and heave) was observed for the Hydrodynamique et Structure Marines CENTRALE NANTES - BUREAU
regular and irregular wave interaction over the SPAR. The tensions in VERITAS. The first author acknowledges IITM and LHEEA, ECN for the
the mooring lines also compared well with the measurements. Mooring financial support for this Ph.D. study. This work is partially supported
in the form of a stiffness matrix provides an accurate comparison for by Centre For Large scale Ocean Research (CFLOR), IIT Madras.

20
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

References Hall, M., Housner, S., Sirnivas, S., Wilson, S., et al., 2021. MoorPy (Quasi-Static Mooring
Analysis in Python). Technical Report, National Renewable Energy Lab.(NREL),
Agarwal, S., Saincher, S., Sriram, V., Yan, S., Xie, Z., Schlurmann, T., Ma, Q., Yang, X., Golden, CO (United States).
Wan, D., Gong, J., et al., 2021a. A comparative study on the nonlinear interaction Hasan, S., Sriram, V., Selvam, R.P., 2019. Evaluation of an eddy viscosity type wave
between a focusing wave and cylinder using state-of-the-art solvers: part B. Int. J. breaking model for intermediate water depths. Eur. J. Mech. B Fluids 78, 115–138.
Offshore Polar Eng. 31 (01), 11–18. Higuera, P., Lara, J.L., Losada, I.J., 2013. Realistic wave generation and active wave
absorption for Navier–Stokes models: Application to OpenFOAM® . Coast. Eng. 71,
Agarwal, S., Sriram, V., Yan, S., Murali, K., 2021b. Improvements in MLPG formulation
102–118.
for 3D wave interaction with fixed structures. Comput. & Fluids 218, 104826.
Hildebrandt, A., Sriram, V., Schlurmann, T., 2013. Simulation of focusing waves and
Aliyar, S., Meyer, J., Sriram, V., Hildebrandt, A., 2021. Experimental investigation of
local line forces due to wave impacts on a tripod structure. In: The Twenty-Third
offshore crane load during installation of a wind turbine jacket substructure in
International Offshore and Polar Engineering Conference. OnePetro.
regular waves. Ocean Eng. 241, 109979.
Hirt, C.W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of free
Arnal, V., 2020. Experimental Modelling of a Floating Wind Turbine Using A
boundaries. J. Comput. Phys. 39 (1), 201–225.
«Software-In-The-Loop» Approach (Ph.D. thesis). École Centrale de Nantes.
Huang, Y., Zhuang, Y., Wan, D., 2021. Hydrodynamic study and performance analysis
Benitz, M.A., Schmidt, D.P., Lackner, M.A., Stewart, G.M., Jonkman, J., Robertson, A.,
of the OC4-DeepCWind platform by CFD method. Int. J. Comput. Methods 18 (04),
2015. Validation of hydrodynamic load models using CFD for the OC4-DeepCwind
2050020.
semisubmersible. In: International Conference on Offshore Mechanics and Arctic
Huseby, M., Grue, J., 2000. An experimental investigation of higher-harmonic wave
Engineering. vol. 56574, American Society of Mechanical Engineers, V009T09A037.
forces on a vertical cylinder. J. Fluid Mech. 414, 75–103.
Biausser, B., Fraunié, P., Grilli, S.T., Marcer, R., 2004. Numerical analysis of the internal Jacobsen, N.G., Fuhrman, D.R., Fredsøe, J., 2012. A wave generation toolbox for the
kinematics and dynamics of 3-D breaking waves on slopes. Int. J. Offshore Polar open-source CFD library: OpenFoam® . Internat. J. Numer. Methods Fluids 70 (9),
Eng. 14 (04). 1073–1088. http://dx.doi.org/10.1002/fld.2726.
Black, J.L., Mei, C.C., Bray, M.C.G., 1971. Radiation and scattering of water waves by Jasak, H., Tukovic, Z., 2006. Automatic mesh motion for the unstructured finite volume
rigid bodies. J. Fluid Mech. 46 (1), 151–164. method. Trans. FAMENA 30 (2), 1–20.
Canard, M., Ducrozet, G., Bouscasse, B., 2020. Generation of 3-h long-crested waves of Jasak, H., Tuković, Ž., 2010. Dynamic mesh handling in openfoam applied to fluid-
extreme sea states with HOS-NWT solver. In: International Conference on Offshore structure interaction simulations. In: Proceedings of the V European Conference on
Mechanics and Arctic Engineering. vol. 84386, American Society of Mechanical Computational Fluid Dynamics ECCOMAS CFD 2010.
Engineers, V06BT06A064. Jiang, C., el Moctar, O., Schellin, T.E., Paredes, G.M., 2021. Comparative study of
Choi, Y., 2019. Two-Way Coupling Between Potential and Viscous Flows for a Marine mathematical models for mooring systems coupled with CFD. Ships Offshore Struct.
Application (Ph.D. thesis). École centrale de Nantes. 16 (9), 942–954.
Choi, Y., Gouin, M., Ducrozet, G., Bouscasse, B., Ferrant, P., 2017. Grid2Grid: HOS Jonkman, J., 2010. Definition of the Floating System for Phase IV of OC3. Technical
wrapper program for CFD solvers. arXiv preprint arXiv:1801.00026. Report, National Renewable Energy Lab.(NREL), Golden, CO (United States).
Choi, Y.M., Kim, Y.J., Bouscasse, B., Seng, S., Gentaz, L., Ferrant, P., 2020. Performance Journee, J.M., Adegeest, L.J., 2003. Theoretical Manual of Strip Theory Program ‘‘
of different techniques of generation and absorption of free-surface waves in SEAWAY for Windows’’. Report 1370, TU Delft.
Computational Fluid Dynamics. Ocean Eng. 214, 107575. Karnoski, S., Palo, P., 1988. Validation of a static mooring analysis model with full-scale
Devolder, B., Stratigaki, V., Troch, P., Rauwoens, P., 2018. CFD simulations of floating data. In: Offshore Technology Conference. OnePetro.
point absorber wave energy converter arrays subjected to regular waves. Energies Kemper, J., Windt, C., Graf, K., Ringwood, J., 2019. Development towards a nested
11 (3), 641. hydrodynamic model for the numerical analysis of ocean wave energy systems. In:
Dewey, R.K., 1999. Mooring Design & Dynamics, a Matlab® package for designing and European Tidal and Wave Energy Conference Proceedings. (1414), EWTEC.
analyzing oceanographic moorings. Mar. Model. 1 (1–4), 103–157. Kim, Y.J., 2021. Numerical Improvement and Validation of a Naval Hydrodynamics CFD
Di Paolo, B., Lara, J.L., Barajas, G., Losada, Í.J., 2020. Wave and structure interaction Solver in View of Performing Fast and Accurate Simulation of Complex Ship-Wave
using multi-domain couplings for Navier-Stokes solvers in OpenFOAM® . Part I: Interaction (Ph.D. thesis). École centrale de Nantes.
Implementation and validation. Coast. Eng. 103799. http://dx.doi.org/10.1016/j. Lachaume, C., Biausser, B., Fraunié, P., Grilli, S.T., Guignard, S., 2003. Modeling of
coastaleng.2020.103799. breaking and post-breakingwaves on slopes by coupling of BEM and VOF methods.
Dommermuth, D.G., Yue, D.K., Lin, W., Rapp, R., Chan, E., Melville, W., 1988. Deep- In: The Thirteenth International Offshore and Polar Engineering Conference.
water plunging breakers: a comparison between potential theory and experiments. OnePetro.
J. Fluid Mech. 189, 423–442. Lee, S.C., Song, S., Park, S., 2021. Platform motions and mooring system coupled solver
Ducrozet, G., Bonnefoy, F., Le Touzé, D., Ferrant, P., 2006. Implementation and for a moored floating platform in a wave. Processes 9 (8), 1393.
validation of nonlinear wavemaker models in a HOS numerical wave tank. Int. Li, Z., 2018. Two-Phase Spectral Wave Explicit Navier-Stokes Equations Method for
J. Offshore Polar Eng. 16 (03). Wave-Structure Interactions (Ph.D. thesis). École centrale de Nantes.
Ducrozet, G., Bonnefoy, F., Le Touzé, D., Ferrant, P., 2012. A modified high-order Li, H., Bachynski, E.E., 2021. Experimental and numerical investigation of nonlinear
spectral method for wavemaker modeling in a numerical wave tank. Eur. J. Mech. diffraction wave loads on a semi-submersible wind turbine. Renew. Energy 171,
B Fluids 34, 19–34. 709–727.
Li, Z., Bouscasse, B., Ducrozet, G., Gentaz, L., Le Touzé, D., Ferrant, P., 2021.
Ducrozet, G., Bouscasse, B., Gouin, M., Ferrant, P., Bonnefoy, F., 2019. CN-stream:
Spectral wave explicit navier-stokes equations for wave-structure interactions using
Open-source library for nonlinear regular waves using stream function theory. arXiv
two-phase computational fluid dynamics solvers. Ocean Eng. 221, 108513.
preprint arXiv:1901.10577.
Löhner, R., Yang, C., 1996. Improved ALE mesh velocities for moving bodies. Commun.
Ducrozet, G., Fink, M., Chabchoub, A., 2016. Time-reversal of nonlinear waves:
Numer. Methods. Eng. 12 (10), 599–608.
Applicability and limitations. Phys. Rev. Fluids 1 (5), 054302.
Lu, X., Chandar, D.D.J., Chen, Y., Lou, J., 2017. An overlapping domain decomposition
Dunbar, A.J., Craven, B.A., Paterson, E.G., 2015. Development and validation of a
based near-far field coupling method for wave structure interaction simulations.
tightly coupled CFD/6-DOF solver for simulating floating offshore wind turbine
Coast. Eng. 126, 37–50.
platforms. Ocean Eng. 110, 98–105.
Make, M., 2014. Predicting scale effects on floating offshore wind turbines.
Elhanafi, A., Macfarlane, G., Fleming, A., Leong, Z., 2017. Experimental and numerical
Masciola, M., 2018. MAP++ Documentation. NREL, Golden, CO, USA.
investigations on the hydrodynamic performance of a floating–moored oscillating Masciola, M., Jonkman, J., Robertson, A., 2013. Implementation of a multisegmented,
water column wave energy converter. Appl. Energy 205, 369–390. quasi-static cable model. In: The Twenty-Third International Offshore and Polar
Engsig-Karup, A., 2006. Unstructured Nodal DG-FEM Solution of High-Order Engineering Conference. OnePetro.
Boussinesq-Type Equations (Ph.D. thesis). Mayer, S., Garapon, A., Sørensen, L., 1998. A fractional step method for unsteady free-
Fernández, H., Sriram, V., Schimmels, S., Oumeraci, H., 2014. Extreme wave generation surface flow with applications to non-linear wave dynamics. Int. J. Numer. Methods
using self correcting method—Revisited. Coast. Eng. 93, 15–31. Fluids 28, 293–315. http://dx.doi.org/10.1002/(SICI)1097-0363(19980815)28:23.
Ferri, F., Palm, J., 2015. Implementation of a Dynamic Mooring Solver (MOODY) into 0.CO;2-1.
a wave to wire model of a simple WEC. Department of Civil Engineering, Aalbor Michel, G., Bonnefoy, F., Ducrozet, G., Prabhudesai, G., Cazaubiel, A., Copie, F.,
G University. Tikan, A., Suret, P., Randoux, S., Falcon, E., 2020. Emergence of Peregrine solitons
Gu, H., Stansby, P., Stallard, T., Moreno, E.C., 2018. Drag, added mass and radiation in integrable turbulence of deep water gravity waves. Phys. Rev. Fluids 5 (8),
damping of oscillating vertical cylindrical bodies in heave and surge in still water. 082801.
J. Fluids Struct. 82, 343–356. Miquel, A.M., Kamath, A., Alagan Chella, M., Archetti, R., Bihs, H., 2018. Analysis of
Guilcher, P., Ducorzet, G., Alessandrini, B., Ferrant, P., 2007. Water wave propagation different methods for wave generation and absorption in a CFD-based numerical
using SPH models. In: Proceedings 2nd International Spheric Workshop. pp. wave tank. J. Mar. Sci. Eng. 6 (2), 73.
119–122. Nematbakhsh, A., Bachynski, E.E., Gao, Z., Moan, T., 2015. Comparison of wave load
Hall, M., 2015. MoorDyn User’s Guide. Department of Mechanical Engineering, effects on a TLP wind turbine by using computational fluid dynamics and potential
University of Maine, Orono, ME. flow theory approaches. Appl. Ocean Res. 53, 142–154.
Hall, M., Goupee, A., 2015. Validation of a lumped-mass mooring line model with OpenFOAM, 2017. OpenFOAM v5 User Guide. URL: https://cfd.direct/openfoam/user-
DeepCwind semisubmersible model test data. Ocean Eng. 104, 590–603. guide.

21
S. Aliyar et al. Ocean Engineering 263 (2022) 112206

ORCINA, 2014. OrcaFlex—Documentation. URL: https://www.orcina.com/releases/ Seng, S., 2012. Slamming and Whipping Analysis of Ships. DTU Mechanical
orcaflex-101/. Engineering.
Paduano, B., Giorgi, G., Gomes, R.P., Pasta, E., Henriques, J.C., Gato, L., Mattiazzo, G., Simonsen, C.D., Otzen, J.F., Joncquez, S., Stern, F., 2013. EFD and CFD for KCS heaving
2020. Experimental validation and comparison of numerical models for the mooring and pitching in regular head waves. J. Mar. Sci. Technol. 18 (4), 435–459.
system of a floating wave energy converter. J. Mar. Sci. Eng. 8 (8), 565. Sriram, V., Agarwal, S., Yan, S., Xie, Z., Saincher, S., Schlurmann, T., Ma, Q.,
Palm, J., Eskilsson, C., Paredes, G.M., Bergdahl, L., 2016. Coupled mooring analysis Stoesser, T., Zhuang, Y., Han, B., et al., 2021. A comparative study on the nonlinear
for floating wave energy converters using CFD: Formulation and validation. Int. J. interaction between a focusing wave and cylinder using state-of-the-art solvers: Part
Mar. Energy 16, 83–99. A. Int. J. Offshore Polar Eng. 31 (01), 1–10.
Paulsen, B.T., Bredmose, H., Bingham, H.B., 2014. An efficient domain decomposition Sriram, V., Ma, Q., Schlurmann, T., 2014. A hybrid method for modelling two
strategy for wave loads on surface piercing circular cylinders. Coast. Eng. 86, dimensional non-breaking and breaking waves. J. Comput. Phys. 272, 429–454.
57–76. Sun, H., 2007. A boundary element method applied to strongly nonlinear wave-body
Pinguet, R., 2021. Hydrodynamics of Semi-Submersible Floater for Offshore Wind interaction problems.
Turbines in Highly Nonlinear Waves Using Computational Fluid Dynamics (CFD), Thomsen, J.B., Eskilsson, C., Ferri, F., 2017. Assessment of available numerical tools
and Validation of Overset Meshing Technique in a Numerical Wave Tank (Ph.D. for dynamic mooring analysis: WP1. 2 & M1.
thesis). Ecole Centrale Marseille. Vickers, A., 2012. Improve the Understanding of Uncertainties in Numerical Analysis
Randolph, M., Quiggin, P., 2009. Non-linear hysteretic seabed model for catenary of Moored Floating Wave Energy Converters (Ph.D. thesis). University of Exeter.
pipeline contact. In: International Conference on Offshore Mechanics and Arctic West, B.J., Brueckner, K.A., Janda, R.S., Milder, D.M., Milton, R.L., 1987. A new
Engineering. vol. 43437, pp. 145–154. numerical method for surface hydrodynamics. J. Geophys. Res. Oceans 92 (C11),
Rienecker, M.M., Fenton, J.D., 1981. A Fourier approximation method for steady water 11803–11824.
waves. J. Fluid Mech. 104, 119–137. Xu, H., Zhang, Y., Santo, H., Chua, K.H., Law, Y.Z., Chan, E.S., 2021. Coupling of
Robertson, A., Jonkman, J., Masciola, M., Song, H., Goupee, A., Coulling, A., Luan, C., potential flow and CFD model for fluid and structure interactions. In: International
2014. Definition of the Semisubmersible Floating System for Phase II of OC4. Conference on Offshore Mechanics and Arctic Engineering. vol. 85185, American
Technical Report, National Renewable Energy Lab.(NREL), Golden, CO (United Society of Mechanical Engineers, V008T08A009.
States). Zhang, Y., Kim, B., 2018. A fully coupled computational fluid dynamics method for
Rusche, H., 2002. Computational Fluid Dynamics of Dispersed Two-Phase Flows at High analysis of semi-submersible floating offshore wind turbines under wind-wave
Phase Fractions (Ph.D. thesis). excitation conditions based on OC5 data. Appl. Sci. 8 (11), 2314.

22