CFD Simulation of Multiple Moored Floating Structures Using Open Foam
CFD Simulation of Multiple Moored Floating Structures Using Open Foam
net/publication/375182968
CITATIONS READS
0 71
3 authors, including:
All content following this page was uploaded by Haifei Chen on 01 November 2023.
restraints library
a
Interdisciplinary Science Cooperative, University of New Mexico, Albuquerque, NM 87131, USA
b
Universidade da Coruña, Campus Industrial de Ferrol, CITENI, Spain
c
Coremarine Solutions SL, Madrid, Spain
Abstract:
It is not uncommon to observe multiple floating structures in close vicinity in the coastal and ocean
these structures can be constructed to form an offshore floating wind farm, a wave energy converter array,
or an aquaculture farm. This paper presents the development and application of a Computational Fluid
Dynamics model coupled with mooring analysis codes to simulate the dynamic response of multiple
moored floating structures. Both the six degrees of freedom rigid body motion library and the multibody
dynamics library in the open-source finite volume CFD toolbox, OpenFOAM, are extended with a quasi-
static mooring model and two dynamic mooring models. The overset grid method is adopted to account
for the mesh motions of multiple rigid bodies. The coupled CFD-mooring model is validated against
experimental measurements for a single box-type floating breakwater and a twin floating breakwater
moored in regular waves, as well as the steady-state posture of an underwater towed system. The coupled
model is also verified by comparing results from the two rigid body motion solvers in OpenFOAM and a
Smoothed Particle Hydrodynamics solver. The mooring restraints library developed in this work is open-
source and can be applied to simulate the multibody systems involved in emerging offshore wind and
Correspondence: hfchen@unm.edu
Keywords:
Wave-structure interaction; rigid body dynamics; OpenFOAM; overset; mooring dynamics; underwater
towed system.
Highlights:
• There are two libraries in OpenFOAM that can handle floating body motions: rigidBodyMotion
and sixDoFRigidBodyMotion.
• Both libraries are enhanced by coupling with three mooring analysis codes, enabling simulation
• The multi-floater CFD models are validated against experimental data for a moored twin floating
breakwater.
• OpenFOAM simulation results are also compared with those of DualSPHysics model.
1. Introduction
A plethora of floating structure systems have been designed to serve the purposes of various offshore
industries, which must operate in harsh environments and thus withstand the extreme loads of strong
winds, waves, and currents. A robust mooring system is typically an indispensable component of these
systems to ensure the integrity of the structures and the safety of life at sea. To find a sustainable and
without inter-body connections, may be a prospective option to reduce the project’s total cost of
construction, installation, operation, and maintenance. Typical application scenarios of these floating
systems include Wave Energy Converter (WEC) arrays (da Fonseca et al., 2016; Gomes et al., 2020),
offshore wind farms (Lopez-Olocco et al., 2023; Lozon and Hall, 2023), aquaculture farms (Fredriksson
et al., 2004; Ma et al., 2022), and even multi-use platforms. Li et al. (2020) proposed a framework to
assess the technical feasibility of a multi-purpose platform which is comprised of a wind turbine, an array
Unlike conventional offshore oil & gas applications, mooring systems may account for a considerable
portion of the project cost for the emerging wind and wave energy technologies. The design of mooring
systems for WECs has even a unique requirement of maximizing the motion while minimizing loads and
costs (Xu et al., 2019). Compact WEC arrays and wind farms with shared moorings can potentially
reduce the levelized cost of energy. Gomes et al. (2020) presented an experimental study on the dynamic
response and mooring tensions of different mooring arrangements of a five-device array of spar-buoy
Oscillating Water Column (OWC) devices. Compared to the independently moored array, arrays with
inter-body connections presented larger motion and smaller line tension under moderate sea states. But
under extreme sea states, arrays with inter-body connections presented both larger motion and a
significantly larger tension in the inter-body lines. It was suggested that longer inter-body lines should be
used in order to avoid high load peaks observed in the tests. Despite this, the interconnected array
configurations were found to yield 75% more annual energy extracted compared to the baseline (non-
interconnected) arrangement (Howey et al., 2021). To assess the effect of shared mooring on the dynamic
response of floating wind turbines (FWTs), Lopez-Olocco et al. (2023) conducted a wave basin test for
dual-spar FWTs with a shared mooring configuration and for a single spar FWT with catenary mooring. It
was observed that in the extreme wave condition, the shared mooring configuration led to a 40% increase
in the platform surge motion compared with the single spar FWT, and that the shared line
showed extreme tension peak events that were seldom present in the anchor lines. These experimental
campaigns highlight the potential challenges faced by the shared mooring system design for emerging
technologies.
Design and analysis of these floating structure systems is normally conducted using experimentally
validated numerical models. Currently the dominant models are based on the potential flow theory, with
the hydrodynamics solved by a diffraction-radiation code and the system response solved in either the
frequency domain which captures only the linear steady state solution or the time domain which captures
the nonlinear and transient dynamics (Jin et al., 2020; Jiang et al., 2023; Jin et al., 2023; Zheng et al.,
2015). For individually moored multiple bodies without internal connections, the system’s equations of
motion are straightforward extensions of the equation of motion of a single body. Each body’s six-
degree-of-freedom (six-DoF) motions can be solved independently. For multiple bodies interconnected
with shared moorings, it is also relatively easy to solve the equation of motion for each body, except that
the body response and the shared moorings need to be solved implicitly together or explicitly by iteration.
Oikonomou et al. (2020) presented a frequency domain analysis of a triangular array of spar-buoy OWC
devices. Their results showed that the performance of the independently moored array and the array with
inter-body mooring connections were similar, suggesting a positive park effect under a realistic wave
climate. Lozon and Hall (2023) explored the dynamics of a 10-turbine shared-mooring floating wind
array using wind farm simulator FAST.Farm. The results showed that the shared-mooring design has
smaller fluctuations in surge when compared to the baseline array of individually moored turbines. It was
concluded that the shared moorings did not overall introduce any dynamic response concerns in their
For multiple bodies connected with kinematic constraints such as hinges/joints, their dynamics is
described by the equations of motion of the bodies and the constraint equations at the connections. Many
numerical models have been developed by combining a potential flow hydrodynamics solver with
Lagrange multiplier or mode expansion technique to account for hydrodynamic interactions among bodies
and the mechanical coupling effect of the hinges (Jin et al., 2023; Zheng et al., 2015). Zhang et al. (2023)
extended an in-house potential flow code with the Lagrange multiplier technique to investigate the motion
responses of a large array of modularized hinged bodies. The frequency domain method was validated
through a 5-box array physical test, which was then applied to examine the effects of the hinge
constraints, the number of longitudinal modules, and the external stiffness constraints. It was revealed that
the motion responses of a hinged array were highly sensitive to the external constraints such as mooring
lines. Instead of using conventional in-house code, Jin et al. (2023) assessed the application of open-
source tool WEC-Sim (Wave Energy Converter SIMulator) to model a two-body hinged raft WEC.
Developed in MATLAB/Simulink using the multi-body dynamics solver Simscape Multibody, WEC-Sim
can model devices that are comprised of hydrodynamic bodies, joints and constraints, power take-off
systems, and mooring systems (WEC-Sim). Two corrections concerning wave excitation force and body-
body radiation force were made, and a physically validated viscous term was added to the updated time-
Compared to the potential flow theory-based models, it is rare to see Computational Fluid Dynamics
(CFD) models applied to simulate multiple floating structures. CFD models solves the Navier-Stokes
equations directly. Combined with a rigid body dynamics solver, they can account for fluid viscosity,
turbulence, wave breaking, green water, slamming/impact, and overtopping that are all elusive in the
potential flow models. Computationally expensive, these high-fidelity models can provide empirical
damping coefficients that are usually required in potential flow models (Han et al., 2023; Jin et al., 2023).
Most CFD investigations of floating structures are currently focusing on a single body (Chen and Hall,
2022; Huang et al., 2022), and those simulating multiple structures are scarce and may be restricted to
specific degrees of motion only (Devolder et al., 2018; Han et al., 2023; Jiang and el Moctar, 2023). He et
al. (2022) investigated multi-body hydrodynamic interaction between an octagonal platform and absorber-
type WECs utilizing StarCCM+. Li et al. (2022) developed a numerical modeling tool coupling
OpenFOAM and Multi-body Dynamics code (MBDyn) to resolve the mechanical interaction among sub-
structures in a WEC array. Wei et al. (2024) integrated a quasi-static mooring module into a similar
model to investigate the coupling effects between the mooring system and the structural hydroelasticity of
A couple of Smoothed Particle Hydrodynamics (SPH) solvers have also been developed to simulate
multibody hydrodynamics. Chen et al. (2023) studied the hydrodynamics of a twin individually moored
floating breakwater (FB) using 𝛿-SPH coupled with a catenary theory based mooring line model. He et al.
(2023) established an SPH model to simulate a hinged multi-float structure, which is restrained by linear
springs and has rubber bumpers equipped to absorb the collision energy between adjacent modules. Both
of the SPH studies are validated against experimental measurements, which could be used to validate
other numerical models. Domínguez et al. (2019) is the first of its kinds to enable SPH flow solvers to
simulate floating structures with mooring dynamics. A lumped-mass mooring model (Hall and Goupee,
2015) is especially adapted to simulate multiple floating structures with individual moorings or shared
moorings.
The aim of the present paper is to develop and validate a coupled CFD-mooring model to simulate the
dynamic response of multiple moored floating structures. The coupled model is implemented in the
leading open-source CFD toolbox, OpenFOAM, the modular structure of which makes it easy to extend
the code’s functionalities (Weller et al., 1998). There are two libraries in OpenFOAM that can handle
floating body motions: sixDoFRigidBodyMotion and rigidBodyMotion. The former solves directly the
six-DoF motion of a rigid body, while the latter is specially implemented to solve the multibody dynamics
involved in WEC applications. In this paper, both libraries are extended with an open-source mooring
restraint library, which includes a lumped-mass mooring model MoorDyn (Hall and Goupee, 2015), a
quasi-static mooring model MAP++ (Masciola et al., 2013), and a finite element model Moody (Palm et
al., 2017). Thanks to recent developments of MoorDyn, the coupled CFD-mooring model developed in
Chen and Hall (2022), which adopted the sixDoFRigidBodyMotion library, can now be applied to
It should be noted that there are already OpenFOAM applications to simulate multiple floating structures.
Chen et al. (2022) used a coupled OpenFOAM and dynamic mooring model to study the wave dissipation
performance of a single FB and a twin FB under long-period waves. The numerical model (presumably
using rigidBodyMotion library) was validated using experimental data for a single FB in terms of the
wave transmission coefficient, motion and tension amplitude; results for the twin FB were presented
without experimental validation. Eskilsson and Palm (2022) extended the six-DoF rigid body motion
solver with a finite element mooring model to simulate interconnected multiple bodies, specifically
mooring components such as submerged buoys and clump weights. Jiang and el Moctar (2023) extended
the rigidBodyMotion library with a mooring model to simulate hinged multiple bodies. Both latter studies
adopted the mesh morphing technique to accommodate the body motions, which may not be adequate
when the body motions are significant (Chen and Hall, 2022).
Taking FB as an example of multiple moored floating structures, the present study develops and validates
a coupled CFD-mooring model against time series of experimental measurements for a twin FB moored
in regular waves. To encourage reproducible open research and further application to more complex
floating structure systems, the mooring restraint library is released open source. The case studies
presented in this paper may serve as a valuable verification and validation exercise for OpenFOAM and
other CFD models. The remainder of the paper is structured as follows. Section 2 describes the various
components of the coupled model, including free surface flow solver, rigid body motion solver, mesh
motion solver, mooring lines model, and the coupling procedure between the flow solver, the body
motion solver, and the mooring models. Section 3 presents three case studies validating the coupled CFD-
mooring model: a single FB (Section 3.1), a twin FB (Section 3.2), and an underwater towed system
(Section 3.3). The overset grid method is used as the mesh motion solver for multiple bodies. The
numerical results from the two rigid body motion solvers in OpenFOAM are also compared with those by
a meshless SPH solver, DualSPHysics. A simulation of a twin FB with shared moorings is demonstrated
2. Methodology
2.1 Two-phase flow solver
The air-water mixture based two-phase flow solver in OpenFOAM solves the Reynolds-Averaged Navier-
Stokes (RANS) equations for two incompressible phases using a finite volume discretization and the
Volume of Fluid (VOF) surface capturing method (Jasak, 1996; Rusche, 2002). The governing equations
describing mass continuity and conservation of momentum for an incompressible fluid are given by
∇∙𝐔=𝟎 (1)
𝜕𝜌𝐔
+ ∇ ∙ (𝜌𝐔𝐔) − ∇ ∙ (𝜇𝑒𝑓𝑓 ∇𝐔) = −∇𝑝∗ − 𝐠 ∙ 𝐗∇𝜌 + ∇𝐔 ∙ ∇𝜇𝑒𝑓𝑓 (2)
𝜕𝑡
where 𝐔 is fluid velocity vector in Cartesian coordinates, ρ density of the mixed fluid, 𝑝∗ = 𝑝 − 𝜌𝐠 ∙ 𝐗
pseudo-dynamic pressure, 𝑝 total pressure, 𝐠 acceleration due to gravity, 𝐗 position vector of the
computational cells, and 𝜇𝑒𝑓𝑓 = 𝜇 + 𝜇𝑡 effective dynamic viscosity, which is the sum of molecular
dynamic viscosity 𝜇 and turbulent dynamic viscosity 𝜇𝑡 . A variety of turbulence models are available in
OpenFOAM. But for the case studies presented in this study which have simple body geometries, no
turbulence modelling is switched on. Activating turbulence modelling is not expected to alter significantly
the numerical results, as demonstrated in Chen and Hall (2022) and other many other similar mesh-based
CFD studies of wave-structure interactions, as well as the apparent absence of turbulence models in SPH
simulations of floating structures (e.g., Domínguez et al., 2019; Chen et al., 2023; He et al., 2023).
𝜕𝛼
𝜕𝑡
+ ∇ ∙ (𝐔𝛼) + ∇ ∙ (𝐔r 𝛼(1 − 𝛼)) = 0 (3)
in which 𝛼 is volume fraction of water in a computational cell. Note that an extra term is added to limit
the smearing of air-water interface (Hirt and Nichols, 1981). This artificial convective term is active only
in the vicinity of the interface 0 < 𝛼 < 1, where the cells are occupied by a mixture of the two fluids. The
smaller the cell size, the thinner the air-water interface. The density 𝜌 and the viscosity 𝜇 of the mixture
in each cell are 𝛼-weighted average of those of air and water. In this algebraic VOF, the location of the
air-water interface can be approximated by taking an iso-surface of 𝛼 = 0.5 in the interface cells. More
rigidBodyMotion, are used to solve the motions of multiple floating structures. For multiple bodies
without internal connections, each body’s six-DoF motions can be treated separately based on the
where the subscript f denotes the quantities for the floating body. 𝐯𝑓 and 𝛚𝑓 are the linear and angular
velocity of the body, and 𝑚𝑓 and 𝐈𝑓 are the mass and moment of inertia of the body. 𝐅𝑓 and 𝐌𝑓 are the
total external forces and moments acting on the body, which includes the gravity, the fluid force, and the
mooring restraint force. The fluid forces are calculated by integrating the normal pressure and the
tangential shear stress over all the patches enclosing the body (Gatin et al., 2017). The mooring force is
The native rigid body motion library in OpenFOAM, i.e. sixDoFRigidBodyMotion, is applied to solve the
six-DoF motions for each of the floating bodies. The hydrodynamic forces/moments acting on the body
are first calculated, which gives an initial acceleration for each new iteration or time step. All the
order to update the body’s acceleration. Once the linear and angular accelerations are obtained, the
Newmark−𝛽 integration scheme is applied to update the velocity, position, and orientation of the floating
body. Note that the linear and angular accelerations are solved separately using 3D vector equations.
For multiple bodies loosely connected by shared moorings, towing cables, or those constrained with
kinematic joints, one has to resort to the rigid body dynamics library, rigidBodyMotion, to solve their
motions. Following Featherstone (2014), 6D vectors are used to combine the linear and angular aspects of
rigid-body motions and forces. For example, linear and angular velocity are combined to form a spatial
velocity vector; force and moment are combined to form a spatial force vector. This provides a compact
notation for studying rigid-body dynamics. Using an angular-before-linear spatial notation, the two 3D
vector equations of motion, Eqs. (4) and (5), for a rigid body can be written as
𝐟 = 𝐈𝐚 + 𝐯 ×∗ 𝐈𝐯 (6)
𝐟 = (𝐌f , 𝐅f ) (7)
𝐯 = (𝛚f , 𝐯f ) (8)
𝐚 = (𝛚̇f , 𝐯̇ f − 𝛚f × 𝐯f ) (9)
𝐈f 𝟎
𝐈=[ ] (10)
𝟎 𝐦f 𝐈3
where 𝐟 is the spatial force acting on the body, 𝐯 and 𝐚 are the body’s spatial velocity and acceleration, 𝐈3
is identity matrix of size 3, and 𝐈 is the body’s spatial inertia tensor. The symbol ×∗ in Eq. (6) denotes a
For a rigid-body system containing multiple rigid bodies and joints, the equations of motion expressed in
where 𝐪, 𝐪̇ and 𝐪 are vectors of generalized (joint) position, velocity, and acceleration variables,
respectively, 𝐇 is generalized inertia matrix, 𝐂 is generalized bias force, and 𝛕 is a vector of generalized
forces. The bias force is simply the value of 𝛕 that produces zero acceleration. It accounts for the Coriolis
and centrifugal forces, gravity, and any other forces (hydrodynamic and mooring force in this study)
The rigid body dynamics library in OpenFOAM implemented the articulated-body algorithm for forward
dynamics problems
𝐪̈ = FD(𝑚𝑜𝑑𝑒𝑙, 𝐪, 𝐪̇ , 𝛕) (12)
which aims to calculate the acceleration response of a rigid-body system to any applied forces. The
symbol model in Eq. (12) refers to a collection of data that describes a particular rigid-body system: the
number of bodies and joints, their connectivity, the values of every parameter associated with each
component such as inertia and geometric parameters (Featherstone, 2014). While the built-in six-DoF
motions solver calculates the body motion one by one, this multibody dynamics solver calculates the
motion responses of all bodies together. After the hydrodynamic forces/moments acting on all the bodies
are calculated, the reacting forces/moments by all restraints are added up to the body each of the restraint
is attached to. The forward dynamics algorithm then calculates the joint acceleration from the joints state
and forces. The joints velocity and position are then integrated using the Newmark−𝛽 scheme. The
bodies state is finally updated to correspond to the current joints state. This algorithm is inherently
amplitude body motions and multiple moving bodies (Chen et al., 2019). In the overset mesh method, two
sets of grids are defined: a relatively large background grid and a set of local overset grids each enclosing
one of the moving bodies. A composite computational domain is then generated via cell-to-cell mappings
between the two sets of disconnected grids, which overlap each other. The background grid is mostly
stationary, while the overset grid moves following the body motion, prescribed in advance or calculated
using the two rigid body motion libraries in OpenFOAM (see Fig. 1). All the computational cells in the
CFD domain are classified into three categories: calculated, interpolated, and holes. The flow governing
equations are solved for calculated cells. The interpolated cells are used to interpolate flow variables
between the two mesh regions. The holes cells, which represent the moving body, are blocked out during
the calculation. Because of the interpolation between different mesh regions, the overset mesh method
may be more computationally demanding than the mesh deformation method (Chen and Hall, 2022). It is
recommended to use similar background and overset grid cell sizes in the vicinity of the overset patch in
order to reduce the interpolation error and speed up the overset grid computation.
Fig. 1. The mooring restraints package, foamMooring, in the context of a flowchart of CFD simulation of
mass mooring model MoorDyn (Hall and Goupee, 2015), and a finite element model Moody (Palm et al.,
2017), are coupled with OpenFOAM’s rigid body motion solvers by developing three corresponding rigid
body motion restraints. Herein only an overview of the three mooring models is provided. Readers are
referred to the original references for detailed implementations of the models. Table 1 in Chen and Hall
(2022) listed the main features of the three mooring analysis codes. Both MAP++ and MoorDyn are open-
source, with the latter under continuous development. Moody is an in-house code but a precompiled
library is released for coupling with other codes. MAP++ is a quasi-static mooring model based on an
extension of conventional single line catenary solutions, allowing multi-line mooring systems with
arbitrary connections to be analyzed. MoorDyn is a lumped-mass mooring analysis library. Each mooring
line is discretized into a number of equal-length segments, each segment connecting two neighboring
nodes. All the line mass and external forces acting on the line, including buoyancy, seabed contact force,
and hydrodynamic drag and inertia forces based on Morison’s equation, are lumped at a finite number of
nodes. Massless springs are conceived to connect the adjacent nodes representing the internal tension and
damping forces. The lumped-mass formulation solves the equations of motion for each node i
where 𝑚𝑖 is the point mass which is half the combined mass of all line segments connected with the node,
𝐫𝑖 , 𝐫̈𝑖 the global coordinate position vector and acceleration of the node, 𝐓𝑖 , 𝐂𝑖 the internal tension and
damping resultant force from the node’s connecting segments, 𝐖𝑖 , 𝐁𝑖 the submerged weight and seabed
contact force, 𝐃𝑛𝑖 , 𝐃𝑡𝑖 the transverse and tangential components of the drag force, and 𝐚𝑛𝑖 , 𝐚𝑡𝑖 are the
transverse and tangential components of the added mass matrix. A new version of MoorDyn (v2) enables
its use for an array of floating structures with and without shared mooring lines (Lozon and Hall, 2023).
Moody is a finite element mooring dynamics model developed with the special intent of capturing and
resolving snap loads (Palm et al., 2017). Unlike the lumped-mass formulation, this model is derived first
and foremost from the governing equations of motion for an elastic cable
𝜕
𝛾0 𝐫̈ = (𝑇𝐭̂) +𝐟 (14)
𝜕𝑠
𝜕𝐫 𝜕𝐫 −1
𝐭̂ = 𝜕𝑠 |𝜕𝑠| (15)
𝜕𝐫
𝑇 = 𝐸𝐴 (|𝜕𝑠| − 1) (16)
where 𝐫 and 𝐫̈ are global coordinate position vector and acceleration of the cable, s curvilinear coordinate
along the cable, 𝛾0 mass per unit length, 𝐭̂ tangent unit vector along the arclength s, 𝑇 tension force
magnitude, 𝐸𝐴 axial stiffness assuming a linear elastic cable material, and 𝐟 all the external forces
consisting of the added mass and Froude-Krylov force, the viscous drag force, the net force of gravity and
buoyancy, and the seabed contact force. The governing equations are then discretized using an hp-
adaptive discontinuous Galerkin method. The high-order formulation ensures engineering accuracy can be
loop implemented in OpenFOAM’s flow solvers, which is a combination of SIMPLE and PISO
algorithms. For each PIMPLE iteration (Fig. 1), the floating body motion is solved first and the mesh
motion updated. The VOF transport equation is then solved to advect the air-water interface. The
pressure-velocity coupling is resolved with a PISO loop, which consists of one optional momentum
predictor and several pressure corrections. The PIMPLE loop within each time step may alleviate the time
lag between and thus converge the flow and the body motion. More tightly coupled CFD and body
motion solvers may be needed to enhance this partitioned approach (Gatin et al., 2017; Bruinsma et al.,
2018).
The coupling of the mooring models with floating body motion solver follows a loose-coupling approach.
Several mooring restraints are developed in foamMooring package to complement the two rigid body
motion libraries sixDoFRigidBodyMotion and rigidBodyMotion in OpenFOAM (Fig. 1). Two coupling
modes, i.e., body coupling and point coupling, are proposed to indicate whether the body-mooring
coupling is achieved via either the body’s center of gravity (rotation) or the mooring lines’ attachments
points (fairleads) on the body. For all three mooring models, the mooring system is initialized the first
time the mooring forces are requested via the body motion solver.
For body coupling, the position and velocity of the floating body are passed from the body motion solver
to the mooring restraint, which updates the fairlead kinematics and the mooring system states such as the
mooring line nodes’ position, velocity, and tension. The mooring model then calculates the total
restraining forces and moments acting on the floating body by summing all mooring line contributions.
The summed forces and moments are returned to the body motion solver to update the body acceleration.
For point/fairlead coupling, the mooring attachments positions (calculated from the instantaneous
orientation of the floating body) are passed to the mooring models to update the mooring system states.
Correspondingly, instead of the total mooring forces/moments exerting on the body, the fairlead tensions
from all the mooring lines are returned from the mooring model to the rigid body restraint. The restraint
then sums up the total mooring forces/moments and passes them to the motion solver to update the body
motion.
The new mooring restraints are compiled into two dynamic libraries (libsixDoFMooring.so and
librigiBodyMooring.so) that can be simply loaded at runtime into the built-in OpenFOAM solvers
interFoam and overInterDyMFoam, or into other variants (such as waveFoam and olaFlow) developed by
the community. For the sixDoFRigidBodyMotion restraint, both point and body coupling should work
equally. However, only the MoorDyn mooring restraint is equipped with both coupling modes at the
moment. MAP++ by itself does not calculate the total mooring forces/moments acting on the body.
Moody’s in-house code was designed to support both coupling modes. But the pre-compiled library does
not support debugging the issues encountered with the body coupling approach. Thanks to MoorDyn’s
continuous developments, more than one mooring system instance can now be created (Lozon and Hall,
2023). This makes it possible to simulate an array of individually moored floating structures using the
significant rewrite of the library is deemed necessary (Eskilsson and Palm, 2022). For the
rigidBodyMotion restraints, only the point coupling mode is valid as the mooring restraining moments
requested by the multibody dynamics formulation should be in global coordinate system. It should be
noted that the point coupling mode is more universal and can be applied to simulations of interconnected
floating bodies.
3. Results
Three case studies are presented to validate the coupled CFD-mooring model: a single FB, a twin FB, and
an underwater towed system. In Section 3.1, the coupled model is first validated against experimental
tests for a single box-type FB. Predictions by the two rigid body motion solvers in OpenFOAM compare
well with the measurements of surface elevation, body motion, and mooring line tension. In Section 3.2,
the coupled model is validated against experimental tests for a twin FB, which are individually moored by
four catenary lines. This case study also verifies the two rigid body motion solvers in OpenFOAM and a
meshless DualSPHysics solver to simulate the coupled dynamics of multiple bodies. A simulation of a
twin FB with shared moorings is demonstrated as well. In Section 3, the coupled model is validated
against the steady-state position and posture of an underwater towed system under four towing speeds.
The new mooring restraint library developed for the rigidBodyMotion library is also tested using the
floating box experiment studied in Chen and Hall (2022), in which the single box motion was solved
using the sixDoFRigidBodyMotion library. No differences are observed between the surface elevations,
body motions, and mooring line tensions predicted by the two methods. These results are omitted here for
the sake of brevity. Interested readers could refer to the tutorial (link provided in acknowledgement) to
2022).
0.3 m high (Liang et al., 2022). The mass and moment of inertia are 58.09 kg and 2.441 kg·m2,
respectively. The model draft was adjusted to 0.16 m through steel ballast fixed at the bottom of the
breakwater. A four-line catenary mooring system was used to restrain the breakwater. Each mooring line
was made of stainless steel with a line density of 0.177 kg/m. The upper end of the mooring line is
connected to the bottom of the floating breakwater, and a spring was inserted near the top end to simulate
the stiffness of the mooring chain with a coefficient of 2.36 kN/m. The lower end was anchored to the
flume bottom. The horizontal distance between the two attachment points is 0.639 m. The total length of
Fig. 2 shows the OpenFOAM model setup for the box-type floating breakwater experiment (Liang et al.,
2022). The 2D numerical flume is 13 m long and 0.9 m high. The still water depth is 0.6 m. As in the
physical experiments, four wave gauges (denoted G1, G2, G3, and G4) are used to record the surface
elevations. A global coordinate system O-xyz is defined such that x = 0 is located at the center of the
rectangular floating breakwater and z = 0 at the still water level. The draft of the breakwater is 0.16 m and
the center of gravity is located at z = -0.0948 m. The overset mesh is adopted to accommodate the body
motion.
An active generating-absorbing wavemaker is applied to the inlet and outlet boundaries of the numerical
wave flume (Higuera et al., 2013). Second-order wave theory is used to specify the velocities and VOF
along the stationary inlet boundary. Note that cell stretching is also applied to the rightmost 3 m of the
flume to help damp the waves. The simulation is intended to run for up to 20 s and the time step is
adjusted at runtime with a maximum Courant number limit of 0.5. The Euler scheme is used for time
marching. Other discretization schemes follow the standard practices of wave simulations using
OpenFOAM’s interFoam series of solvers (Chen and Hall, 2022). Both six-DoF motion and rigid body
dynamics library are used to solve the breakwater motion and move the mesh. Only the MoorDyn
mooring restraint is tested. For each time step, 3 outer correctors (PIMPLE iterations) are used along with
two pressure correctors per PISO loop. To avoid numerical instabilities, the floating body acceleration is
relaxed by a factor of 0.8. According to the grid convergence study in Appendix A, it is sufficient to use a
experiment. There is barely difference in the two sets of numerical results (‘6dof’ and ‘rgb’). As the
numerical simulation is conducted for 20 s only, surface elevation at the leeward wave gauges (G3 & G4)
may not be fully established yet in the first several wave periods (10 s ~ 15 s). But later on, the wave
shape changes and there appears better agreement between the numerical predictions and the
experimental measurements. As a matter of fact, there is excellent match in the first few waves
experienced by the leeward wave gauges if we compare the numerical result with the first initial waves of
the experimental data (not shown here). Overall, the numerical model is able to capture the wave shape
changes, predicting better surface elevations at G1 & G2 than those at the leeward waves.
Fig. 3. Numerical and experimental results of surface elevations for the single FB test (wave height 0.1 m
and period 1.4 s). OF 6dof: OpenFOAM simulation using six-DoF motion library; OF rgb: OpenFOAM
simulation using rigid body dynamics library; Exp: experimental data (Liang et al., 2022).
Fig. 4. Numerical and experimental results of the single FB motion.
Figs. 4 and 5 show the comparison of numerical predictions and experimental measurements of the
breakwater motion and the mooring line fairlead tension, respectively. Good agreements are achieved
among the two sets of numerical results (‘6dof’ and ‘rgb’) and the experimental data. For some unknown
reasons, the heave motion is consistently underestimated. While all the motions (surge, heave, and pitch)
seem regular in time, the mooring tension exhibits slight irregularity due to some disturbances in both the
numerical and the physical flume. The mooring code MoorDyn captures the sudden pick up and release of
the tension in both mooring lines. The tension in the seaward line is in general larger than that in the
leeward line.
Fig. 5. Numerical and experimental results of mooring line tensions for the single FB test. Top: seaward
Fig. 6. Numerical model setup for the physical experiment of a twin floating breakwater (Chen et al.,
flume is 17 m long and 0.9 m high. The still water level is fixed at 0.514 m. As in the physical
experiments, four wave gauges (denoted G1, G3, G4 and G5) are used to record the surface elevations.
Wave gauge G2 is not placed for relatively long waves. The two floating breakwaters, spaced 1 m apart,
are identical in size and mass properties (Table 1). Each of the floating breakwaters is 0.745 m long, 0.5
m wide and 0.3 m high. The mass and moment of inertia are 55.875 kg and 2.20795 kg·m2, respectively.
Each breakwater is restrained by four identical stainless-steel mooring lines, each having a total length of
0.62 m and an inserted spring with a stiffness of 2.6 N/mm. The upper end of the mooring line is
connected to the bottom of the floating breakwater and the lower end is anchored to the flume bottom.
A global coordinate system O-xyz is defined such that z = 0 is located at the still water level and x = 0 at
the seaward edge of the first floating breakwater. The draft of the floating breakwaters is 0.15 m and the
center of gravity is located at z = -0.0771 m. The overset mesh is adopted to accommodate the body
motion. Second-order wave theory is used to specify the velocities and VOF along the stationary inlet
boundary. Both six-DoF motion and rigid body dynamics library are used to solve the breakwater motion.
The boundary conditions, discretization schemes, and rigid body and mesh motion settings are the same
Although the OpenFOAM model is set up as a 2D simulation, a single cell with a width equal to 0.745 m
is specified in the y-direction. Same as the physical experiments, four identical mooring lines are
specified to restrain each of the twin rectangular FBs. Each mooring line is discretized into 10 equal
segments. The normal and tangential drag coefficient are 2.4 and 0.05, respectively. No efforts are made
to fine-tune these coefficients, as there are no measurements available for the mooring tensions. Three
cases of regular waves are tested to validate the model, all having the same wave height 0.05 m but
The base grid size of the background mesh is 0.01 m, which is refined to 0.005 in the vicinity of the free
surface and the breakwaters. The overset grid for the two floating breakwaters is uniformly 0.005 m.
There are 10 grids per wave height and 518 grids per wave length for the shortest waves considered in
this study. According to the convergence study in Appendix B, finer meshes may be required to achieve
convergence for the surge motion. The heave and pitch motions are less sensitive to the grid size, and are
FB setup as its OpenFOAM counterpart. An active absorbing piston-type wavemaker is applied to the
inlet, and the rightmost 4.5 m of the flume is designated as damping area for the transmitted waves
(Altomare et al., 2017). Different from the OpenFOAM model setup which includes both air and water,
the SPH model resolves only the water phase. The wavemaker at inlet, the two FBs, and the flume bottom
To simulate the twin FB in a 2D setup, the FB length is scaled up from 0.745 m to 1.0 m. The same is
done with FB mass, moment of inertia, and mooring spring stiffness (He et al., 2023). Two mooring lines,
each equivalent to two lines used in the physical test and OpenFOAM model, are used to restrain each of
the FB’s motion. The mooring lines are solved by MoorDyn+ library, which is adapted from MoorDyn to
couple specifically with the SPH flow solver (Domínguez et al., 2019). The DualSPHysics simulation is
run using Tesla K40m GPU at UNM’s Xena cluster, which has 2880 CUDA cores and a computing
capacity of 3.5. According to the convergence study in Appendix B, the particle size is chosen to be
height 0.05 m and periods 1.4 s (left), 1.7 s (middle), 2.0 s (right). Exp: experimental data (Chen et al.,
2023); OF 6dof: OpenFOAM simulation using six-DoF motion library; OF rgb: OpenFOAM simulation
height 0.05 m and periods 1.4 s, 1.7 s, and 2.0 s. Negligible discrepancies are seen between the two sets of
OpenFOAM simulations using six-DoF (‘OF 6dof’) motion library and rigid body dynamics (‘OF rgb’)
library. The only noticeable difference appears at wave gauge G5 for case h05t17. There are overall good
agreements between the numerical results and the experimental data for all the three cases. The same can
also be said for the DualSPHysics simulation, except that the SPH model underestimates the surface
Fig. 8. Numerical and experimental results of breakwater motions for the twin FB test h05t14 (wave
height 0.05 m and period 1.4 s). Refer to Fig. 7 for legend clarifications.
Figs. 8-10 show the comparison of numerical and experimental results of breakwater motions for the twin
FB tests in regular waves for cases h05t14, h05t17, and h05t20, respectively. There are overall good
agreements between OpenFOAM and DualSPHysics simulation results and the measurements for the
latter two cases (Figs. 9 and 10). There are significant discrepancies in surge motion of case h05t14; the
experimental data show both FBs have drifted about 0.1m downwave from their initial equilibrium
positions (Fig. 8). The OpenFOAM results momentarily catches up the surge motion of FB1 and nearly
catches up that of FB2 near the end of simulation. The DualSPHysics results, however, never obtain the
experimental drifts of both FBs. Instead, it is observed that both FBs’ surge motions oscillate steadily
around an equilibrium position not far from their initial positions. The DualSPHysics model employs a
piston-type wavemaker to generate the waves. Both 1st order and 2nd order wave theories are tested, and
no improvement is observed for the surge predictions. The same is true for the relaxation zone technique,
which, instead of mimicking a physical wavemaker, specifies a transition region to smoothly generate the
targeted waves. The unsatisfactory prediction of surge motion may arise from possible differences in the
present 2D model setup and the physical wave flume, which makes it challenging for the model to
accurately calculate the mean drift force for short waves. The single FB test in Section 3.1 seems not to
suffer much from this issue, although the crest amplitude of the surge response is indeed underpredicted
(Fig. 4). These discrepancies in surge predictions were also noticed by Palm et al. (2016). For waves with
the same height, their Fig. 8 demonstrated larger discrepancies for short waves than for long waves.
The OpenFOAM and DualSPHysics results are in phase with each other. But they have inexplainable
phase shift relative to the surge measurements of FB2 for case h05t14 (Fig. 8). Certain time shift may
have accidentally occurred with the recordings. For cases h05t17 and h05t20, both numerical results
generally follow the patterns of the measured surge motion, albeit with some small discrepancies in the
oscillation positions and amplitudes. The best comparisons between the numerical and experimental
results are achieved for the heave motion, which are dominated by the hydrostatic restoring stiffness.
There are also satisfactory agreements for the pitch motions. While the OpenFOAM results match slightly
better with the measurements, the SPH results underestimate the pitch motions of both FBs for case
Fig. 11 shows the comparison of instantaneous horizontal velocity contours and mooring system states
obtained by OpenFOAM and DualSPHysics for case h05t17. The OpenFOAM model is a two-phase flow
solver, encompassing both the water and the air (clipped to show the water wave motion only). The
DualSPHysics model resolves only the water phase. The top boundary of the computational domain is
bound by the free surface and the two FBs. Despite the difference in the domain setup, the two models
actually have about the same number of computational cells/particles, 236k cells for the OpenFOAM
Table 2 lists the computing resources and simulation wall-clock time for case h05t17. It should be noted
that the DualSPHysics model solves the mooring dynamics only once per time step, while the
OpenFOAM model performs three iterations per time step in order to better converge both the motion and
mooring response. It is observed that the OpenFOAM runtime (using six-DoF motion library) is in
general noticeably longer for short waves than for long waves, while the DualSPHysics runtime for
difference cases are about the same. For the same number of cells/particles, the OpenFOAM model runs
significantly slower than the DualSPHysics model, for the fine grids/particles in particular. While the
former may increase the number of CPUs used to hundreds, these computing units, however, are not
generally available on a typical HPC cluster. Each cluster may have specific policies limiting the number
of CPUs accessible to each user and the maximum wall-time allowed for each job. In comparison, a
single GPU, such as the Nvidia Tesla K40m used in this study, is more readily available. It is logical to
speculate that a high-end GPU with higher computing capability may further increase the computational
efficiency. In this perspective, the DualSPHysics model is an attractive alternative to mesh-based CFD
models, which can significantly speed up the runtime required for common engineering simulations.
To demonstrate the OpenFOAM model’s capability to simulate floating structures with shared moorings,
the mooring system of the twin FB (8 mooring lines in total, Fig. 6) is revised so that the twin FB is
interconnected with two shared mooring lines, each of which is1.06 m long. There are now a total number
of 6 mooring lines restraining the floating bodies’ motion. All lines share the same properties as in the
individually moored system. Figs. 12 and 13 show the time history of the surge motion and mooring line
tensions for this twin FB with shared mooring lines in regular waves (h05t17), predicted by a coupled
OpenFOAM (rgb) and mooring model. It is clearly observed that the surge motions of FB1 and FB2 are
tightly coupled, with the wave frequency surge component superimposed with a relatively long cycle of
second-order drift. Correspondingly, the mooring tensions exhibit large spikes whenever the external
mooring lines (lines 1 & 5, see the first panel plot in Fig. 14) reach the positive/negative extremities of the
surge displacements. The shared mooring line (line 3) experiences only mild tension in this case, as its
length is sufficiently long to accommodate the relative surge motion while still maintaining a catenary
shape throughout the simulation. A shorter length of 1.02 m is observed to cause straightening of the
shared line and thus tension spikes as seen in the external lines.
Fig. 12. Surge motion of a twin FB with shared moorings in regular waves (h05t17).
Fig. 13. Mooring line tensions of a twin FB with shared moorings in regular waves (h05t17).
Fig. 14 presents the instantaneous horizontal velocity contours and mooring configurations predicted by
the coupled OpenFOAM and mooring model (Moody). The left column corresponds to instants, 13.1 s ~
13.9 s, when the windward lines (line 1) of twin FB experience snap loads. The right column corresponds
to instants, 18.3 s ~ 19.1 s, when the leeward lines (line 5) experience snap loads. To avoid these costly
tension spikes, it is advisable to carefully configure the shared mooring system with intermediate
weights/floats in order to decouple the body motions and the extreme mooring tensions.
Fig. 14. Snapshots of instantaneous horizontal velocity contours and mooring configurations of a twin FB
connected to a subsurface object through a towing cable. The steady-state position and posture of a simple
towed system consisting of a subsurface cylinder and a towing cable (Guan et al., 2021) are
simulated using the coupled OpenFOAM-mooring model. The subsurface cylinder has a diameter of 8.4
cm and a height of 6.6 cm. The gravity of the object is 9.95 N, and the buoyancy is 3.43 N. The towing
cable is made of nylon, which is 3 m long, 1 cm in diameter, and 1050 kg/m3 in density. The positions
and postures of the towing cable at four towing speeds of 0.3 m/s, 0.5 m/s, 1.0 m/s, and 1.5 m/s are
studied.
The computational fluid domain contains 1.6 million cells in a rectangular box defined by the dimensions
-3.0 ≤ x ≤ 0.2 m, -0.3 ≤ y ≤ 0.3 m and -3.5 ≤ z ≤ 0.5 m. A global coordinate system, O-xyz, is established,
with the origin x = 0 and y = 0 located at the center of the subsurface object, and z = 0 at the still water
level. The top end of the towing cable is held stationary, which is located 0.03 m above the still water
level. The lower end of the towing cable is set at (0, 0, -2.97) m. The initial center of gravity of the heavy
object is thus positioned at (0, 0, -3.003) m. Instead of mimicking the towing operation, a background
current is imposed. The six-DoF dynamic library is used to solve for the motion of the subsurface object,
although the rigid body dynamics library is equally applicable. The overset mesh approach is adopted to
Fig. 15 illustrates the OpenFOAM model setup for the towed subsurface object experiment (Guan et al.,
2021), including views of the background mesh and the overset mesh. The base grid size of the
background mesh is 0.0283 m, which is refined to 0.0071 m along the body’s prospective trajectory. The
overset grid for the subsurface object is maintained at a uniform size of 0.0071 m. This size selection
ensures that there are a minimum of 10 cells placed between the body patches and the overset patch.
Additionally, the mesh is subject to refinement near the body’s wall to enhance the representation of the
structure and capture the flow patterns in the immediate vicinity. Grid convergence study has been
conducted to confirm the cell sizes selection. The towing cable is discretized into 60 equal length
segments. The normal and tangential drag coefficients of the towing cable are set as 1.2 and 0.025,
respectively. The simulations are expected to run for a maximum of 20 s, with the time step adjusted at
runtime to maintain a maximum Courant number of 0.9. Body acceleration is relaxed by 0.4 at each time
experimental measurements (Guan et al., 2021) and numerical simulations obtained by AQWA, Lumped-
Mass Method (LMM), and STAR-CCM+ (Yang et al., 2022). There are no significant differences between
the OpenFOAM simulations (‘OF’) and the experimental measurements. It is worth noting that at a towing
speed of V=0.5 m/s, while there is generally good agreement in the posture predicted by OpenFOAM, the
other numerical methods (AQWA, LMM, and STAR-CCM+) present significant differences. At a towing
speed of V=1.0 m/s and V=1.5 m/s, the results of the LMM method differ significantly from those of OF,
AQWA, and STAR-CCM+. It is in the case of V=1.5 m/s where OpenFOAM exhibits more noticeable
differences form the experimental data for the lower-middle section of the towing cable. However, the tip
position of the towing cable shows negligible difference from the experimental data.
Fig. 16. Numerical and experimental results of the towing cable posture with four towing speeds.
Table 3. Results of towing cable tip positions and deviations with experimental data (Guan et al., 2021).
Table 3 lists the results of the submerged body’s coupling position with the towing cable and deviations
with experimental results (Guan et al., 2021). The largest deviation occurs at a towing speed of 0.3 m/s
(6.3%) for the horizontal component and at a towing speed of 1.5 m/s (2.2%) for the vertical component.
Fig. 17 shows the simulation snapshots of horizontal velocity contours and towing cable tension values at
steady-state (20 s) predicted by the coupled OpenFOAM and MoorDyn model. The posture of the
subsurface object and the towing cable change significantly with the increase of towing speeds. It should
be noted that a single (subsurface) body is simulated for this case study. Using the mooring restraints
developed for the rigidBodyMotion library, it is straightforward to accommodate a second (surface) floating
Fig. 17. Snapshots of instantaneous horizontal velocity contours and towing cable posture and tension
predicted by OpenFOAM for a towing speed of (a) 0.3, (b) 0.5, (c) 1.0, and (d) 1.5 m/s.
4. Concluding remarks
To help analyze multiple floating structures operating in close vicinity in the coastal and ocean
environments, this paper presents the development and validation of a high-fidelity coupled CFD-
mooring model to simulate the dynamic response of these structures using the leading open-source CFD
toolbox, OpenFOAM. The two rigid body motion libraries that solve body motions respectively by
Newton’s second law and articulated body algorithm are extended with an open-access mooring restraint
library, which includes a quasi-static mooring model MAP++, a lumped-mass mooring model MoorDyn,
and a finite element model Moody. Thanks to recent developments of MoorDyn, the coupled CFD-
mooring model using the built-in sixDoFRigidBodyMotion library can now be applied to simulate
multiple individually moored bodies. Because of its inherent multibody dynamics formulation, the
interconnected via hinges or shared moorings. The coupled CFD-mooring model is validated against
experimental measurements for a single box-type floating breakwater and a twin floating breakwater
moored in regular waves. The overset grid method is adopted to solve the mesh motions of (multiple)
rigid bodies. The coupled model is also verified by comparing results from the two rigid body motion
solvers in OpenFOAM and a meshless SPH solver DualSPHysics. Overall, good agreements are achieved
for the heave and pitch motions, as well the surge motion in relatively long waves. Results for the short
waves show that both OpenFOAM and DualSPHysics models may be unable to reproduce the physical
wave flume so that the surge response can be accurately predicted. Results from the three mooring models
(Appendix C) are comparable for the mild wave cases studied in the paper. Numerical instability may
arise when applying these mooring models to simulate laboratory-scale physical tests. Because of its
open-access, MoorDyn affords users more freedom in controlling the mooring system output and
generating visualization files at runtime. The coupled CFD-mooring model has also been validated using
experimental measurements for a subsurface object with towing cables. In general, it demonstrates good
agreement in terms of the position and posture of the towing cable across four different towing speeds.
Using the mooring restraints developed for the multibody dynamics library, the applicability of the model
can be extended by incorporating the dynamics of a floating surface object towing in the presence of
waves.
Three-dimensional CFD simulations of multiple floating structures can be computationally intensive. But
the development of coupled potential flow and viscous flow solvers makes it computationally amenable to
(Aliyar et al., 2022; Lu et al., 2022). The open-access mooring restraints library developed in this work
can be compiled as a dynamic library and loaded at runtime into such coupled flow solvers. Apart from
the floating array of wind and wave farms, it is straightforward to apply the coupled CFD-mooring model
to study gap resonance and vessel stability which respectively occur during offloading and
lifting/installation operations. Using the articulated body algorithm, it can also be applied to other floating
structure systems, such as floating treatment islands, floating solar farms, open ocean aquaculture farms,
The authors declare that they have no known competing financial interests or personal relationships that
Writing - original draft, Writing - review & editing; Tanausú Almeida Medina: Methodology, Validation,
Data curation, Visualization, Writing - original draft, Writing - review & editing; Jose Luis Cercos-Pita:
Software, Validation, Visualization, Writing - original draft, Writing - review & editing.
Acknowledgement
The authors would like to thank the UNM Center for Advanced Research Computing (CARC), supported
in part by the National Science Foundation, and Centro de Supercomputación de Galicia (CESGA) for
providing the computing resources used in this work. Part of this research work was made possible
through the financial support provided by Universidade da Coruña, Campus Industrial de Ferrol (Galicia,
Spain) and Coremarine Solutions SL (Madrid, Spain). The authors would also like to thank Dr. Yong Liu
for providing the experimental data of the floating breakwaters. The mooring restraints package
Fig. A.1. Single FB motion predictions by two grids using OpenFOAM coupled with MoorDyn. Left
column: six-DoF motion library; right column: rigid body dynamics library.
Grid convergence tests for the single FB were conducted using both six-DoF motion and rigid body
dynamics library in OpenFOAM. All runs were performed using 32 CPUs on UNM’s CARC Hopper
cluster. Fig. A.1 shows the breakwater motions predicted by two grids using the MoorDyn mooring
restraint. Apart from the slight difference in the surge predictions by the two grids, the heave and pitch
motions clearly converged. The medium grid size of 1 cm is the same in both x- and z-directions and
uniform across the entire computational domain. The total number of background and overset mesh cells
is 95.8k. The wall-clock time for a 20 s simulation is about 0.9 h for one outer corrector per time step, and
1.6 h for three outer correctors per time step. It was found that solving the mooring system and thus the
body motion only once per time step resulted in slightly larger tension in the mooring lines. The motion
predictions were largely unaffected though. More than one outer corrector ensures more tight coupling
between the fluid motion and the rigid body motion in a single time step. Thus, three outer correctors are
used for results presented in this paper. A uniformly fine grid of 0.5 cm across the entire computational
domain was used for six-DoF motion simulation, resulting in about 383k cells. The wall-clock time for a
20 s simulation using 32 CPUs is about 32.5 h. For rigid body dynamics simulation, a base grid size of 1
cm was refined only in the vicinity of free surface and the breakwater, generating about 175k cells. It took
Grid convergence test was conducted to determine the cell size for the twin FB simulations using
OpenFOAM. Three grids with minimum cell sizes of 1 cm, 5 mm, and 2.5 mm were used to simulate case
h05t17 using the 6dof motion library, resulting in a total number of 133k, 236k, 533k cells, respectively.
The wall-clock time for a 22 s run using 64 CPUs on UNM’s Hopper cluster was about 1.9 h for the
coarse grid and 6.6 h for the medium grid. The runtime for the fine grid was terminated at 19.4 s when it
reached the resource time limit of 48 h. Fig. B.1 shows that coarse grid tends to underestimate the heave
motion. Clear convergence was obtained between the heave/pitch motions predicted by the medium and
fine grids. However, it seems more difficult to achieve convergence for the surge motion. The predicted
surges by the medium and fine grids are initially the same but later on they tend to deviate slightly as the
simulation progresses.
Fig. B.1. Breakwater motion predictions for the twin FB test using OpenFOAM with three resolutions
Grid convergence test was also conducted to determine the particle size using DualSPHysics. Uniform
size particles were generated across the computational domain. A particle size of 10 mm, 6 mm, and 3.6
mm resulted in a total of about 94k, 264k, and 731k particles, including both the fluid and the boundary
(fixed/moving/floating) particles. The wall-clock time for a 30 s run on UNM’s Xena cluster was about
0.8 h, 2.4 h, and 9.6 h, respectively. It was observed that particle size 10 mm is not fine enough to capture
the transmitted waves (G4 & G5 in Fig. B.2) leeward of the second FB and converge the surge motions of
both FBs (Fig. B.3). There are virtually no differences between the surface elevation and heave/pitch
motions predicted by the two finer particle sizes (6 mm and 3.6 mm). It is more difficult to achieve
convergence for the surge motion of the second FB, in particular. The same observations were true for
case h05t14.
Fig. B.3. Breakwater motion predictions for the twin FB test using DualSPHysics with three particle
Fig. C.1-C.3 show the comparison of breakwater motion predictions by three mooring codes MAP++,
MoorDyn, and Moody for the twin FB test using OpenFOAM’s rigid body dynamics library. There were
virtually no differences between the heave and pitch motions among the predictions by three mooring
codes. The only noticeable difference was observed for the surge motion. The shorter the waves, the
larger the discrepancy. For case h05t14, both breakwaters experience large pitch motions, which makes
the mooring system highly dynamic. Both MAP++ and MoorDyn were able to simulate this case, while
Moody failed to converge at around 11.8 s. Specifically, the mooring lines in leeward of FB1 returned
nan, which in turn crashed the simulation. It should be noted that MoorDyn may also experience
numerical instability for similar laboratory-scale physical tests (Aliyar et al., 2022; Paduano et a., 2020).
Fig. C.1. Comparison of breakwater motion predictions by three mooring codes MAP++ (dotted) and
MoorDyn (dashed) for the twin FB test h05t14 (wave height 0.05 m and period 1.4 s) using
MoorDyn (dashed), and Moody (dashed-dot) for the twin FB test h05t17 (wave height 0.05 m and period
References
Aliyar, S., Ducrozet, G., Bouscasse, B., Bonnefoy, F., Sriram, V., & Ferrant, P. (2022). Numerical
coupling strategy using HOS-OpenFOAM-MoorDyn for OC3 Hywind SPAR type platform. Ocean
Altomare, C., Domínguez, J. M., Crespo, A. J. C., González-Cao, J., Suzuki, T., Gómez-Gesteira, M., &
Troch, P. (2017). Long-crested wave generation and absorption for SPH-based DualSPHysics model.
numerical wave tank for simulating floating offshore wind turbines. Ocean Engineering, 147, 647-658.
Chen, H., & Hall, M. (2022). CFD simulation of floating body motion with mooring dynamics: Coupling
Chen, H., Qian, L., Ma, Z., Bai, W., Li, Y., Causon, D., & Mingham, C. (2019). Application of an overset
mesh based numerical wave tank for modelling realistic free-surface hydrodynamic problems. Ocean
Chen, J., Zhang, J., Wang, G., Zhang, Q., Guo, J., & Sun, X. (2022). Numerical simulation of the wave
dissipation performance of floating box-type breakwaters under long-period waves. Ocean Engineering,
266, 113091.
Chen, Y. K., Liu, Y., Meringolo, D. D., & Hu, J. M. (2023). Study on the hydrodynamics of a twin
da Fonseca, F. C., Gomes, R. P. F., Henriques, J. C. C., Gato, L. M. C., & Falcão, A. F. O. (2016). Model
testing of an oscillating water column spar-buoy wave energy converter isolated and in array: Motions
Devolder, B., Stratigaki, V., Troch, P., & Rauwoens, P. (2018). CFD simulations of floating point
absorber wave energy converter arrays subjected to regular waves. Energies, 11(3), 641.
Domínguez, J. M., Crespo, A. J., Hall, M., Altomare, C., Wu, M., Stratigaki, V., ... & Gómez-Gesteira,
M. (2019). SPH simulation of floating structures with moorings. Coastal Engineering, 153, 103560.
Deng, R., Luo, F., Ren, H., Mo, X., & Wu, T. (2022). Experimental research for the influence of stern
towing force on a planing hull in the towed system. Ocean Engineering, 259, 111849.
Eskilsson, C., & Palm, J. (2022). High-fidelity modelling of moored marine structures: multi-component
simulations and fluid-mooring coupling. Journal of Ocean Engineering and Marine Energy, 8(4), 513-
526.
Fredriksson, D. W., DeCew, J., Swift, M. R., Tsukrov, I., Chambers, M. D., & Celikkol, B. (2004). The
design and analysis of a four-cage grid mooring for open ocean aquaculture. Aquacultural Engineering,
32(1), 77-94.
Gatin, I., Vukčević, V., Jasak, H., & Rusche, H. (2017). Enhanced coupling of solid body motion and
Gomes, R. P., Gato, L. M., Henriques, J. C., Portillo, J. C., Howey, B. D., Collins, K. M., ... & Greaves,
D. M. (2020). Compact floating wave energy converters arrays: Mooring loads and survivability through
Guan G., Zhang X., Wang Y., Yang Q. (2021). Analytical and numerical study on underwater towing
cable dynamics under different flow velocities based on experimental corrections. Applied Ocean
Hall, M., & Goupee, A. (2015). Validation of a lumped-mass mooring line model
with DeepCwind semisubmersible model test data. Ocean Engineering, 104, 590-603.
Han, M., Cao, F., Shi, H., Kou, H., Gong, H., & Wang, C. (2023). Parametrical study on an array of point
He, G., Luan, Z., Jin, R., Zhang, W., Wang, W., Zhang, Z., ... & Liu, P. (2022). Numerical and
Higuera, P., Lara, J. L., & Losada, I. J. (2013). Realistic wave generation and active wave absorption for
Hirt, C. W., & Nichols, B. D. (1981). Volume of fluid (VOF) method for the dynamics of free
Howey, B., Collins, K. M., Hann, M., Iglesias, G., Gomes, R. P., Henriques, J. C., ... & Greaves, D.
(2021). Compact floating wave energy converter arrays: Inter-device mooring connectivity and
Huang, H., Gu, H., & Chen, H. C. (2022). A new method to couple FEM mooring program with CFD to
Jasak, H (1996). Error analysis and estimation for the finite volume method with applications to fluid
Jiang, C., & el Moctar, O. (2023). Extension of a coupled mooring–viscous flow solver to account for
mooring–joint–multibody interaction in waves. Journal of Ocean Engineering and Marine Energy, 9(1),
93-111.
Jiang, C., Xu, P., Bai, X., Zhao, Z., el Moctar, O., & Zhang, G. (2023). A review of advances in modeling
hydrodynamics and hydroelasticity for very large floating structures. Ocean Engineering, 285, 115319.
Jin, C., Bakti, F. P., & Kim, M. (2020). Multi-floater-mooring coupled time-domain hydro-elastic
analysis in regular and irregular waves. Applied Ocean Research, 101, 102276.
Jin, S., Wang, D., Hann, M., Collins, K., Conley, D., & Greaves, D. (2023). A designed two-body hinged
raft wave energy converter: from experimental study to annual power prediction for the EMEC site using
response of an offshore floating multi-purpose platform for the Blue Economy. Ocean Engineering, 217,
107943.
Li, X., Xiao, Q., Zhou, Y., Ning, D., Incecik, A., Nicoll, R., ... & Campbell, D. (2022). Coupled CFD-
MBD numerical modeling of a mechanically coupled WEC array. Ocean Engineering, 256, 111541.
Liang, J. M., Liu, Y., Chen, Y. K., & Li, A. J. (2022). Experimental study on hydrodynamic
characteristics of the box-type floating breakwater with different mooring configurations. Ocean
Lopez-Olocco, T., Liang, G., Medina-Manuel, A., Ynocente, L. S., Jiang, Z., & Souto-Iglesias, A. (2023).
Experimental comparison of a dual-spar floating wind farm with shared mooring against a single floating
Lozon, E., & Hall, M. (2023). Coupled loads analysis of a novel shared-mooring floating wind farm.
Lu, X., Dao, M. H., & Le, Q. T. (2022). A GPU-accelerated domain decomposition method for numerical
Ma, C., Bi, C. W., Xu, Z., & Zhao, Y. P. (2022). Dynamic behaviors of a hinged multi-body floating
Masciola, M., Jonkman, J., & Robertson, A. (2013, June). Implementation of a multisegmented, quasi-
static cable model. In The Twenty-third International Offshore and Polar Engineering Conference.
Oikonomou, C. L. G., Gomes, R. P. F., Gato, L. M. C., & Falcão, A. F. O. (2020). On the dynamics of an
array of spar-buoy oscillating water column devices with inter-body mooring connections. Renewable
Experimental validation and comparison of numerical models for the mooring system of a floating wave
Palm, J., Eskilsson, C., Paredes, G. M., & Bergdahl, L. (2016). Coupled mooring analysis for floating
wave energy converters using CFD: Formulation and validation. International Journal of Marine
Palm, J., Eskilsson, C., & Bergdahl, L. (2017). An hp-adaptive discontinuous Galerkin method for
Rusche, H. (2002). Computational fluid dynamics of dispersed two-phase flows at high phase fractions.
Wei, Y., Yu, S., Jin, P., Huang, L., Elsherbiny, K., & Tezdogan, T. (2024). Coupled analysis between
catenary mooring and VLFS with structural hydroelasticity in waves. Marine Structures, 93, 103516.
Weller, H. G., Tabor, G., Jasak, H., & Fureby, C. (1998). A tensorial approach to computational
Xu, S., Wang, S., & Soares, C. G. (2019). Review of mooring design for floating wave energy
Yang X., Wu J., Xu S. (2022). Dynamic analysis of underwater towed system under undulating motion
Zhang, D., Du, J., Yuan, Z., Yu, S., & Li, H. (2023). Motion characteristics of large arrays of modularized
two-raft wave energy conversion device. Journal of Fluids and Structures, 58, 271-290.