Anda di halaman 1dari 18

Applied Ocean Research 51 (2015) 153–170

Contents lists available at ScienceDirect

Applied Ocean Research

journal homepage:

Direct time domain analysis of floating structures with linear and

nonlinear mooring stiffness in a 3D numerical wave tank
Shivaji Ganesan T a,∗ , Debabrata Sen b
Indian Register of Shipping, 52-A, Adi Shankaracharya Marg, Opp Powai Lake, Powai, Mumbai 400072, India
Department of Ocean Engineering and Naval Architecture, IIT Kharagpur, India

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, motion response of a moored floating structure interacting with a large amplitude and
Received 30 August 2014 steep incident wave field is studied using a coupled time domain solution scheme. Solution of the hydro-
Received in revised form 13 January 2015 dynamic boundary value problem is achieved using a three-dimensional numerical wave tank (3D NWT)
Accepted 6 April 2015
approach based upon a form of Mixed-Eulerian–Lagrangian (MEL) scheme. In the developed method,
nonlinearity arising due to incident wave as well as nonlinear hydrostatics is completely captured while
the hydrodynamic interactions of radiation and diffraction are determined at every time step based on
certain simplifying approximations. Mooring lines are modelled as linear as well as nonlinear springs.
Numerical wave tank
Spread mooring
The horizontal tension for each individual mooring line is obtained from the nonlinear load-excursion
Time domain plot of the lines computed using catenary theory, from which the linear and nonlinear line stiffness are
Nonlinear stiffness determined. Motions of three realistic floating structures with different mooring systems are analyzed
considering various combinations of linear and approximate nonlinear hydrodynamic load computations
and linear/nonlinear mooring line stiffness. Results are discussed to bring out the influence and need for
consideration of nonlinearities in the hydrodynamics and hydrostatics as well as the nonlinear modelling
of the line stiffness.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction structures where the mooring line dynamics was solved by finite
difference scheme and hydrodynamic interaction forces were
For evaluating the global performance of a floating body with computed based on linear radiation/diffraction theory. Coupled
station keeping system, an integrated coupled dynamic analysis in and uncoupled time domain analysis of a turret moored FPSO in a
the time domain is preferred over frequency domain approaches. harsh environment was studied by Heurtier et al. [5] where hydro-
Wichers [1] carried out a coupled time domain simulation of a dynamics loads acting on the FPSO are calculated from linearized
tanker moored to a single point mooring system using impulse potential theory and mooring lines and production risers were
response functions and a constant frequency domain hydrody- represented as cable, bar and slender beam elements, and solved
namics coefficient model. Coupled analysis for a turret moored in the framework of FEM. Garrett [6] compared fully coupled
FPSO with mooring and risers operating in different water depths analysis in both time and frequency domain for a semisubmersible
were studied by Ormberg and Larsen [2]. In their study frequency floating production platform in which the rigid body motions were
dependent added mass and damping coefficients were computed computed using the radiation–diffraction programme WAMIT and
in the frequency domain based on linear hydrodynamics and moor- the mooring lines were represented by elastic rods in FEM. Tahar
ing line dynamics was modelled by Finite Element Method (FEM). et al. [7] made a comparative study of time and frequency domain
A comparative study of quasi-static and dynamic approaches for coupled analysis using a computer programme WINPOST with
a spar platform with mooring lines was carried out by Chen et al. full-scale data of a Horn Mountain spar during Hurricane Isidore.
[3] where the wave force on the spar platform are calculated by Low and Langley [8] proposed a hybrid method (time/frequency
Morison equation and the nonlinear FEM code CABLE3D was used domain) for efficient computation of coupled analysis of ves-
to calculate the dynamics of mooring lines. Kim and Sclavounos sels/mooring/risers. Tahar and Kim [9] studied dynamic analysis
[4] studied large amplitude coupled motions of spar and TLP of polyester mooring lines in deepwater in the time domain
and assessed the effect of large elongation of polyester line and
nonlinear stress–strain relations. Zhang et al. [10] compared the
∗ Corresponding author. Tel.: +91 22 30519504; fax: +91 22 25703611. results of frequency domain, semi-coupled and fully coupled time
E-mail addresses: (S. Ganesan T), domain analyses with experimental data for a cell–truss spar
(D. Sen). and concluded that the frequency domain and semi coupled time

0141-1187/$ – see front matter © 2015 Elsevier Ltd. All rights reserved.
154 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Mixed Eulerian–Lagrangian (MEL) based NWT approach. After the

Nomenclature pioneering work of Longuet-Higgins and Cokelet [19] on devel-
opment of the MEL method for simulating extreme and breaking
d water depth [m] waves, many works have been reported describing developments
H incident wave height [m] of a MEL based NWT simulation for wave-structure interactions
Kxx roll radius of gyration in 2D as well as 3D. A detail review of the available NWT based
Kyy pitch radius of gyration nonlinear methods may be seen in Ma and Yan [20] containing a
Kzz yaw radius of gyration large number of references therein. Although many of the diffi-

n unit normal vector culties associated with this method for a full nonlinear simulation
r distance between field and source point of extreme waves with 2D floating bodies have been resolved, a
T wave period [s] full nonlinear simulation of extreme waves with 3D floating bodies
ˇ wave angle [◦ ] still remains difficult particularly for generating long duration sim-
 free surface elevation ulation results for practical geometric configurations like offshore
 fluid density [kg/m3 ] platforms and ships. One of the constraints in this context is the
g gravitational constant [m/s2 ] prohibitively large computational time the method demands, see
 solid angle e.g. the recent work of You and Faltinsen [21].
XF fairlead x-coordinate in earth axis In the present method, a simplified form of the NWT based
YF fairlead y-coordinate in earth axis method has been developed, following the approach outlined in
XAC anchor x-coordinate in earth axis [22,23] but with modifications in the numerical implementation
YAC anchor y-coordinate in earth axis that enable generation of long duration converged steady-state
→ rG time derivative of vector → rG results for practical 3D structures. In this scheme, a circular tank
→ time derivative of vector →  is considered in which there exists an incident wave. The moored
structure is then introduced at the centre of the tank, and the
subsequent interaction of the incident wave with the structure is
domain analysis were computationally more efficient for initial simulated. The interaction hydrodynamics of radiation and diffrac-
conceptual design. Kim et al. [11] carried out experimental studies tion are determined at each time step as the solution progresses
of a spread moored twin pontoon four column semisubmersible through a discretized integral equation solution based on Rank-
and compared with numerical simulation computed by higher ine panel method. Certain simplifying approximations are however
order panel method for hydrodynamic forces and FEM method for introduced in the solution method which allows the formation of
mooring system dynamics in the time domain. Yang et al. [12,13] the integral equation over a time-invariant surface, which in turn
carried out time domain coupled analysis of floating bodies based enables the computation count to be brought to a level suitable
on Stokes perturbation procedure using suitable time integration for routine use and comparable to the available schemes based on
scheme. Hydrodynamic wave loads were computed in the time convolution integral formulations. The incident wave forces and
domain rather than transforming from frequency domain and the the hydrostatic forces which consist of significant nonlinearities
rigid body dynamics was coupled with mooring line dynamics are determined on the exact wetted surface under the incident
computed based on rod theory of Garrett [14]. Kim et al. [15] made wave profile. This is often referred to as body-exact formulation for
a comparative study in the time domain for linear and nonlinear Froude-Krylov (FK) and hydrostatic forces. The incident wave can
mooring systems for a turret moored FPSO and spread moored be steep and defined by a nonlinear numerical model. As the radi-
semi-submersible and concluded that the effect of mooring ation and diffraction are computed at every time step along with
dynamics were significant only for deep water whereas in shallow the solution of the equations of motion, it is in this spirit that the
water the influence of mooring dynamics on the overall restoring term direct time-domain solution has been used above. The moor-
forces and thus to the overall motions was negligible. ing lines are treated as springs but nonlinear stiffness is considered,
Nearly all of the time-domain simulation studies in the above determined from pre-computed load excursion characteristics of
except [12] appears to follow a similar approach in which the individual lines. Thus the formulation enables consideration of cer-
frequency-dependent hydrodynamic coefficients (added mass, and tain important nonlinearities of the hydrodynamic forces, and also
damping) are usually computed by linear radiation–diffraction nonlinear stiffness of the mooring lines. A consistent linear solu-
code, and the frequency-domain equations of motion written in tion can be recovered from this by simply taking a linear incident
the time domain through impulse response and convolution inte- wave, determining all parameters on the mean surface and taking
gral originally suggested by Cummins [16] and further developed a linear spring model for the mooring lines. Numerical computa-
by Ogilvie [17]. In a fully coupled solution method, mooring system tions are performed for three realistic moored floating structures
dynamics is formulated using either FEM or lumped mass method and comparative studies made for various combinations of linear
given by [18], and the combined dynamic equation of motion of the and nonlinear hydrostatics and incident-wave hydrodynamics with
floating body and the mooring line are solved in the time-domain. linear and nonlinear mooring line stiffness, to bring out the relative
Almost all studies conclude that FEM modelling or lumped mass importance of nonlinearities.
method to capture the dynamics of mooring lines are effective for
deep water but their influence on floating body motions in water
depth below 500 m are less important. For the relatively shallow
water cases, therefore, a simpler spring-model for the mooring line 2. Floating body equation of motion
may be adequate for studying the body-motions in place of using a
far more time-consuming modelling of the mooring line dynamics The problem considered is the motion of a moored floating body
through FEM or lumped mass model. in an incident wave field, a schematic diagram of which is shown in
The main objective of the present study is to carry out a coupled Fig. 1. At the outset, two right-handed Cartesian coordinate systems
analysis of moored floating structures in a direct time domain solu- are introduced: an earth fixed coordinated system OXYZ with origin
tion method following a form of the three dimensional numerical O at the undisturbed water surface z = 0 with z directed vertically
wave tank (3D NWT) approach. For full nonlinear wave-structure upwards, and a body-fixed system Gx y z with origin G coincident
interactions in potential flow, one of the available methods is the with the body centre of gravity.
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 155

Fig. 1. Earth and body coordinate systems.

The equations of motion in Newtonian form but defined with 2.1. The hydrodynamic problem
respect to the body-system are given as:
. The hydrodynamic problem is based upon potential flow the-
[M]{U}  × U}
 + [M]{˝  = {Fex } ory assumptions as is usual in studies of wave interaction with
. (1) large structures. Accordingly a velocity potential ϕT exists whose
[I]{˝}  × {[I] ˝}
 + [˝]  = {M
 ex } gradient gives fluid velocity.
Consider a circular fluid domain with the object placed at the
Here U  = {p, q, r} are respectively the three trans-
 = {u, v, w} and ˝ centre as shown in Fig. 2. The domain is bounded on top by the free-
. .
lational and rotational velocities in the body system, U  and ˝  surface SF (t), interior by the instantaneous wetted surface SB (t),
denotes the time derivatives of velocity vector, [M] and [I] are the below by a rigid flat bottom surface SD and on the exterior by SC .
rigid-body mass and moments of inertia matrices. The generalized The boundary value problem for ϕT is specified by the following:
external forces FE = {Fex , M
 ex } can be written as:
∇ 2 ϕT (x, y, z; t) = 0 in the fluid domain (4)
FE = FH + Fm (2) ∂ϕT
= 0 on the bottom surfaceSD (5)
where FH is the total hydrodynamic load on the body including ∂n
the hydrostatic component, and Fm is the restoring forces from ∂ϕT
= V b · n
 on the instantaneous wetted body surfaceSB (t) (6)
the mooring lines. All these forces are defined in the body system ∂n
Gx y z . ⎫
∂ ∂ϕT ∂ ∂ϕT ∂ ∂ϕT
The motions of the body can be determined by integrating Eq. + + − = 0⎪

(1), noting that these are written in the body-system and there- ∂t ∂x ∂x ∂y ∂y ∂z
on free − surfaceSF (t) (7)
fore certain transformations are necessary in extracting the desired ∂ϕT 1  ϕT = 0

+ g + ∇ ϕT .∇
body motions. Briefly, if rG = {xG , yG , zG } represent the coordinates ∂t 2
of G in the earth-fixed OXYZ system, and   = {, , } are the Euler
In addition, a suitable condition on exterior surface SC needs to
angles representing the angular orientation of the body-system be imposed to close the problem specification.
with respect to the earth system, then we have: Here  is the free-surface elevation which means SF (t)is defined
. . by (x, y ; t) − z = 0, V b is the velocity of the body surface at the

rG = [Tx ]U;  = [T ]˝
 considered point where the condition applies, and n  is the normal
vector to the surface, defined exterior to the fluid boundary (and
The transformation matrices are given in terms of the angles
thus interior to the body surface).
{, , } which are the usual roll, pitch, and yaw angles. Similarly
The nonlinearities in the above boundary value problem lies in
{xG , yG , zG } is the usual surge, sway and heave motions. The desired
the body and free-surface boundary conditions Eqs. (6) and (7).
motion displacements can thus be determined by first integrating
 ˝ and then Eq. (3) to get → r , .
 Additionally, a proper exterior condition needs to be imposed on
Eq. (1) to get U, G
SC compatible with the nonlinear problem specification.
For carrying out the integration of equations of motion, the
The numerical solution of the stated boundary value problem in
most important parameters that need to be determined are the
the framework of NWT approach is typically formed by expressing
external forces. These forces are however coupled with the body’s
ϕT as a continuous distribution of fundamental sources or Rankine
motion responses. The procedure for determination of these forces
source and dipoles over the entire fluid boundary using Green’s
along with integration of the equations of motion is based upon
a NWT approach with certain simplifying approximations. In this
approach, we consider a circular numerical tank of a finite extent ∂ 1 1 ∂ϕT (Q )
in which an incident wave is assumed to exist. The object along (P)ϕT (P) = ϕT (Q ) − dS (8)
S(t) ∂n r r ∂n
with its station-keeping constraints is then placed at the centre of
the tank, and the object’s motion responses in the time domain are In the above, S(t) = SF (t) ∪ SB (t) ∪ SD ∪ SC is the total boundary sur-
determined. This approach differs from the conventional NWT in face of the fluid domain, P and Q are the field and source points,
the sense that incident waves are not generated by physical move- and r is the distance between points P and Q respectively. Usually
ment of a wave-maker but introduced in the tank defined by some the flat bottom surface SD is removed from the integrand in Eq. (8)
suitable wave theory. The solution method is described below. by considering image of Q about this surface. There are also many
156 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 2. Computational domain.

schemes that have been devised to remove SC from the integrand typically extending over 10–20 time periods. Additionally, for 3D
through implementation of a suitable radiation condition. Recent floating body simulations, motions are restricted mostly to three
work of [24] provides a survey of such schemes. degree of freedom (dof) instead of the full six dof [20].
Briefly the computational scheme for a fully nonlinear NWT The purpose of the present development is to devise a scheme
approach proceeds as follows. Solution of a linear system of equa- that has the advantages of a 3D NWT approach in terms of its
tions formed through a discretization of Eq. (8) gives ϕT on SB (t) generality and versatility to consider realistic 3D offshore struc-
∂ϕ ∂ϕ tures and yet retaining some of the most important nonlinearities
and T on SF (t) assuming T on SB (t) and ϕT on SF (t) is known
∂n ∂n
prior. The latter two boundary variables are determined through in the wave-structure interactions but practical to apply routinely
integration of body boundary condition and free-surface condi- by industry with the present-day available computing resources.
tions. The relevant quantities of interest, in particular the forces This requires certain approximations in the solution of the hydro-
acting on the body and the body motions, both of which are deter- dynamic problem.
mined in the time-integration procedure for establishing the body The simplification introduced here is that ϕT is assumed to be
and free-surface boundary conditions are extracted as the solution expressible as:
progresses in time.
The above is an overly simplified description of the nonlin- ϕT = ϕI + ϕP (10)
ear wave-tank solution scheme. In reality, there are numerous
details that need to be addressed for a successful solution. Apart where ϕI incident wave potential and ϕP = ϕT − ϕI the total pertur-
from proper consideration of these details such as how to absorb bation or disturbance potential which includes the radiation and
the numerical instability that accompanies these type of nonlin- diffraction of the incident wave by the hull. Additionally, on the
ear free-surface flow time-marching solutions, how to treat the assumption that ϕP small, it can be further decomposed into linear
exterior boundary conditions to ensure no reflection of interior dis- diffraction and radiation potentials ϕD and ϕR respectively. These
turbance from this surface, how to compute the pressures on the potentials therefore will be specified by the boundary value prob-
body which involves a time-gradient of the potential ensuring no lem given by Eq. (4) for the respective potentials along with the
numerical instability arising from this computation, etc., another linearized free-surface and body boundary conditions as defined
important point in this method is the need to form large matrices below:
at every time step. The discretized form of Eq. (8) can be expressed For ϕD :
in the form:
∂D ∂ϕD ∂ϕD
− = 0; + gD = 0 onz = 0 (11)
[A]{ϕ, ϕn } = [B]{ϕn , ϕ} (9) ∂t ∂z ∂t
where elements of the matrices [A] and [B] contain the so-called ∂ϕD ∂ϕI
influence coefficients which are essentially integration of the Rank- − on SB (0), the mean body surface (12)
∂n ∂n
ine source and dipoles over the discrete surface elements. Elements
of [A] and [B] are thus purely a function of the geometry of the sur- For ϕR :
face over which the integration in Eq. (8) applies. Since this surface
is formed by a continuously evolving SF (t) and SB (t), the elements ∂R ∂ϕR ∂ϕR
of [A] and [B] need to be computed at every instant of time. If the − = 0; + gR = 0 on z = 0 (13)
∂t ∂z ∂t
number of surface elements in which the surface is to be discretized
is NE and the number of time steps is NT , then the solution needs ∂ϕR
determination of NE × NE elements of [A] and [B] as well as solu- = V b · n
 onSB (0), the mean body surface (14)
tion of the system of linear equations to be repeated in the order
of NT times. For a realistic 3D offshore structure NE will be of the Here D and R are diffracted and radiated wave elevations
order of a few thousand, and also the simulation need to be contin- respectively.
ued for a few hundred to thousand time steps. Application of a full The most important consequence of this assumption is that the
nonlinear NWT scheme for studying wave-structure interaction of boundary integral equation Eq. (8) formed for the numerical solu-
realistic 3D offshore and ship geometries therefore represents pro- tion of ϕD and ϕR now applies over a time-invariant surface SF (t = 0)
hibitively large computation efforts, at least at the present time, and SB (t = 0) since the free-surface and the body surfaces over which
for this scheme to be applied routinely. This may be one of the rea- the boundary conditions for these potential apply are the respec-
sons why a 3D full nonlinear NWT study did not enjoy its success as tive mean surfaces. The respective matrices [A] and [B] in Eq. (9)
much as its 2D counterpart for complex realistic structures (see e.g. can thus need to be computed only once, and by pre-computing
[20,21]). Most 3D full nonlinear solutions available at present are [C] [A]−1 [B], the solution of unknowns through Eq. (9) becomes
for idealized geometries like cylinders, hemisphere, etc., and also only a simple matrix multiplication with computing count of a
their run times are somewhat limited to a few hundred time steps meagre NE2 at every time step.
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 157

The total fluid forces and moments including the hydrostatic hydrodynamics). If the problem is fully and consistently linear (lin-
component can be obtained by integration of pressure given by ear hydrodynamics) then the distinction between earth and body
Bernoulli’s equation: system disappears. In this case the quadratic pressure term of inci-
⎧ dent potential is not taken, and the hydrostatic force combined

⎪  dS
⎨ − pn with the body weight is expressed using the familiar restoring force

FH = SB (t) (15) coefficients.

⎪ − p(x × n
 )dS In carrying out the numerical solution of ϕD and ϕR , through Eq.

SB (t) (8), proper treatment of the exterior boundary surface SC is needed.
Kim et al. [24] has provided a survey of various schemes avail-
∂ϕT 1
p = −gz −  − (∇ ϕT )2 (16) able for the treatment of the exterior boundary in the context of
∂t 2 full nonlinear numerical wave-tank simulations. Introduction of a
The first term in Eq. (16) is the hydrostatic pressure and the rest wave-absorbing damping layer adjacent to the exterior surface is
terms are dynamic pressures. Because of Eq. (10) and linearization one of the most effective means of removing this surface from the
of ϕD and ϕR , the forces arising from these are from the 2nd term of integrand in Eq. (8). There have been many suggested methods to
Eq. (16). However, for the incident wave part and the hydrostatic introduce such damping layer. In our present work, we introduce
part, following the so-called body-exact formulation (termed here a damping layer in an annular zone near SC following the work of
as FK-nonlinear), the integrations are carried out over the exact Cointe et al. [25] and Ferrant [26]. This is achieved by modifying the
wetted surface under the incident wave profile. This means, we linear free-surface conditions as follows:
= −g −
(Re ) (ϕ − ϕe )
FH = FS + FHI + FHR + FHD (17) ∂t
with ∂ ∂ϕ
⎧ = −
(Re ) ( − e )

⎪  dS
∂t ∂z
⎨ − gz n

FS = SB (t)
(18) Here ϕ and  are the velocity potential and the corresponding

⎪ − gz(x × n
 )dS free-surface elevation of any linear wave system,
(Re ) is a damping

SB (t) coefficient and ϕe and e are some reference values. By properly
⎧    selecting the form of
(Re ) as a function of the strength of damping

⎪ ∂ϕI 1
+ (∇ ϕI )2 n as well as the extent of the damping zone, both the potential ϕ and
⎪ −
⎨ 2
SB (t) ∂t free-surface elevation  at the exterior surface R = Re can be reduced
FHI =    (19) to zero or near zero. This then eliminates the need for discretization

⎪ ∂ϕI 1

⎩ − + (∇ ϕI )2 (x × n
 )dS of SC in Eq. (8). Further details with regard to the choice of the
SB (t) ∂t 2 damping coefficient are given in Sen and Srinivasan [22].
⎧  Eq. (22) is applied to both potentials i.e. ϕD and ϕR , Note that

⎨ − n in the present case the outgoing radiation and diffracted waves are
 R SB (0) ∂t linearized and the amplitude of outgoing waves emanating from a
FH = (20)

⎪ ∂ϕR
(x × n
 )dS 3D source attenuate with radial distance. This scheme thus found
⎩ −
SB (0) ∂t to work very well in the present case, as the results below will
⎧  demonstrate.

 dS Another important detail that needs attention in this method is
⎨ − ∂t
in the evaluation of the forces from the pressure terms ∂ϕD /∂t and
S (0)
⎪ ∂ϕD ∂ϕR /∂t in Eqs. (20) and (21). These time derivatives are to be com-

⎩ −  )dS
(x × n puted from the determined of ϕD and ϕR as the solution progress in
SB (0) ∂t
time. This computation is however not trivial. If ∂ϕD /∂t and ∂ϕR /∂t
Certain details with respect to the evaluation of force com- are computed from backward difference rule, it leads to numeri-
ponents and their addition through Eq. (17) are instructive here, cal instability due to force motion coupling, which at many times
arising due to two coordinate systems in which these components cannot be suppressed using a smoothing technique. Consequently,
are determined. Here, x and n  are the radial and normal vector an alternative estimation method for ∂ϕD /∂t and ∂ϕR /∂t is sought.
defined in body coordinate system. Note that transformation of Numerical scheme outlined in Sen [23] is followed in the present
vectors defined in earth and body coordinate system is straight- method, a robust algorithm has been developed for the determi-
forward through the transformation matrix. Force components in nation of radiation and diffracted pressure terms directly from a
Eqs. (18) and (19) are defined in body-system, thus components of discretized integral equation written for the time derivatives of
 in Eqs. (18) and (19) are in the body-system. Note however that
n these potentials.
in Eq. (18), z is the z-coordinate in the earth system. We also need Another feature of the present scheme that deserves a mention
to consider the body weight force which is (0, 0, −W) in the earth is the specification of incident wave field by any chosen wave
system, and thus this force needs to be transformed to the body- theory. If a consistently linear solution is to be recovered, then the
system and added with hydrostatic component. The radiation and incident wave must be defined by a linear (Airy) wave, FK forces
diffraction forces Eqs. (20) and (21) are however determined based determined on the mean wetted hull and hydrostatic restoring
on the earth system (with components of n  defined in the earth sys- forces defined by the well known linear restoring force/moment
tem). In carrying out the summation indicated in Eq. (17) such that coefficients. In the present case, the nonlinear incident wave
the computed FH is in the body-system, FHR and FHD first needed to be is defined by a nonlinear numerical wave based on Fourier
transformed to the body-system and then added to the hydrostatic approximation method given by Rienecker and Fenton [27]. This
and FK forces. description of nonlinear incident wave is preferred here since
The above therefore produces FH where the hydrostatic and not only the scheme is versatile over a wide range of steepness
incident wave forces are nonlinear, and this is what is referred and water depth, the total pressure field including hydrostatic
to as body-exact hydrostatic and FK force computation (nonlinear component under the wave is consistently determined including
158 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 3. Mooring line configuration for catenary solution.

the domain within z = ± . In some of the other theoretical wave mooring lines can be modelled as nonlinear spring with nonlin-
theories, there can be a discontinuity in the total pressure vari- ear stiffness. We have used a catenary mooring line equations to
ation and in the fluid velocity at z = 0 and/or z = −  depending determine the nonlinear stiffness or load-excursion characteristics
on whether there is a crest or a trough. Note that incident wave of the lines. Although mooring lines do provide stiffness against
kinematics is necessary for the diffraction problem, see Eq. (12). A motions in the vertical plane (heave, roll and pitch), the line-
consequence of using a nonlinear incident wave is that in applying restraining forces and moment in these modes are much smaller
the body boundary conditions for the diffraction problem the compared to the hydrostatic restoring forces/moments and thus
 ϕI term will have certain nonlinearity in the wave
∂ϕI /∂n ≡ n · ∇ can be neglected. Here we consider only the horizontal restoring
kinematics, even though it is applied at the mean wetted surface. forces from the lines.
Usually the fluid velocities particularly under the crest are larger The horizontal stiffness of any catenary mooring line as it trans-
from nonlinear wave theories compared to linear wave theory. lates in its own plane can be established by the scope of the line for
Therefore indirectly partial nonlinear effects get embedded in the any horizontal force. Following Barltrop [28] and referring to Fig. 3,
solution of the ϕD problem albeit in an inconsistent manner. for an elastic mooring line we have the relations:

H wL H0 wL0 LT
S − S0 = sinh−1 − sinh−1 + (H − H0 ) − (L − L0 )
w H w H0 AE
2.2. Mooring line load
T 2wh 1

Consideration of the effects of mooring lines on the motion H = AE +1 − − AE; L= T 2 − H2 (23)
response of the moored vessel requires that the mooring line
dynamics are coupled with the body dynamics. In a fully coupled T0 2wh 1

dynamic analysis of the body response with mooring lines, dynamic H0 = AE +1 − − AE; L0 = T0 2 − H02
equation of motion for the mooring lines based on methods such as
finite element method/lumped mass method need to be developed where w is the submerged weight per unit length, h is water depth
and coupled with the body equations of motion. If the objective or vertical span of the line, A and E are the average cross sectional
is only the response of the vessel under the restraining action of area and effective modulus of elasticity of the line, S is the horizontal
the mooring lines, then a simplified method can also be applied scope, L is the length of the line in water (distance between touch-
where the mooring lines are considered as springs and the line down to fairlead), and LT is the total un-stretched length. H, V, and
induced forces are added to the equations of body motion as stiff- T are the horizontal, vertical and inline tension force at the fairlead
ness terms. This latter method may not be capable of capturing as indicated in Fig. 3. Quantities with subscripts ‘0’ are correspond-
the line dynamics fully, but the implementation is simple and can ing quantities related to the initial working pretension, thus H0 is
provide the moored body motions results within acceptable error horizontal pretension force.
limits provided the proper line spring-stiffness values are used. A plot of the horizontal load H–H0 against the horizontal excur-
As shown recently by Kim et al. [15] a simple linear-spring mod- sion or displacement of the fairlead point S–S0 gives the load
elling of the mooring lines can produce motion-response results excursion characteristic of the line. An illustrative plot for load
practically same as the results from more complex finite-element excursion is shown in Fig. 4. This plot gives the horizontal load for
modelling of the mooring lines if the water depth is not too large. any given horizontal displacement of the line along its plane. For
In the present work, the simplified spring-method is used excursions normal to its plane, the load is considered negligible. In
to model the mooring line restoring forces. However, since our case of a linear-spring model, the spring-stiffness will be given by
method is in time domain where at every instant of time the equa- the slope of the curve at origin, and the load excursion line will be
tions of motion given by Eq. (1) are solved, it is possible to model for example as illustrated in Fig. 4. A linear stiffness model of the
the nonlinearities in the line restoring forces depending upon the line will typically under-predict the horizontal force compared to
exact displacement of the contact (fairlead) point. In other words, a nonlinear stiffness model.
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 159

is the angle the plane of ith line with the +x axis, and nm is the
total number of lines. All the quantities in Eq. (24) are defined in
earth coordinate system. In carrying out the computations, a cer-
tain coordinate transformations become necessary. At every time
step, the exact position and orientation of the body in the earth
system in known from rG and .  Thus the position of the fairlead
points in the earth system are easily determined from which the
horizontal excursions of the fairlead points can be found. The hori-
zontal excursion along the plane of the line can then be determined
by resolving the displacement in the appropriate direction. From
the load excursion curve the horizontal forces are determined from
which the total mooring line loads in earth system is found through
Eq. (24). To get the total force as given in the equation of motion
Eq. (1), these loads are transformed to the body-system. The main
Fig. 4. A typical load-excursion plot for a catenary mooring line. point being made here is that proper care needs to be employed in
determining various quantities like forces, coordinates of the panel
If the body is connected to multiple mooring lines, then the total centroid, etc. since some of these are defined in earth system and
horizontal load exerted on the body can be obtained by combining some in body-system. Switching from one to other is however fairly
the contribution from each individual line given by Faltinsen [29]: easy through the transformation matrix.
For the mooring line load calculation, the load excursion curve

for each line can be pre-computed and stored in a graphical form
Fm,1 = Hi cos i and the desired load for a given excursion determine though
i=1 interpolation. Alternatively, through curve fitting the load can be

nm expressed as a polynomial. Usually a 2nd order polynomial is found
Fm,2 = Hi sin i (24) adequate:

H = k0 + k1 + k2 2 (25)
Fm,6 = Hi [xi sin i − yi cos i]
where k1 and k2 are the linear and quadratic spring-stiffness
Here Fm,1 , Fm,2 , and Fm,6 are respective surge, sway and yaw restor- coefficients, and is the excursion along the plane of the line. Note
ing forces and moment from all mooring lines, (xi , yi ) are the that k0 here is essentially the pretension H0 . Linear spring model
coordinate of the ith fairlead or attachment points on the hull, i emerges by simply taking k2 as zero.

Fig. 5. Body panels and free surface panels of S175 container ship.

Fig. 6. Surge, heave and pitch RAOs for the S175 hull at wave angle (ˇ) = 180◦ .
160 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 7. Six degree of freedom RAOs of S175 hull at wave angle (ˇ) = 135◦ .

3. Verification and validation of present method Table 1

Main particulars of the S175 container ship.

The motion results from the developed numerical method needs Length, breadth, depth (m) 179, 25.4, 14
to be first validated/verified against available experimental data or Draft (m) 9.5
Centre of gravity from midship (m) (−2.65, 0, −1.2)
with other well-validated and established codes. While a detailed
Radius of gyration Kxx , Kyy , Kzz (m) 8.75, 44.45, 44.45
validation results are presented in a separate source [30] consid- Total displacements (T) 24,725
ering different geometries here we produce a few results for the Number of body panels 863
standard S175 hull. Main particulars of S175 hull are detailed in Number of free-surface panels 2426
Table 1, and the discretized panels of hull and the free-surface are
shown in Fig. 5.

Fig. 8. Heave and pitch motion of S175 container ship (H = 2 m, T = 10 s, ˇ = 180◦ ).

S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 161

Fig. 9. Comparison of heave and pitch motions of the S175 hull for T = 12 s, H = 10 m, ˇ = 180◦ between present and SWAN FK-nonlinear computations; only a part of the time
history is shown to bring out the difference more clearly.

Figs. 6 and 7 show the comparison of motion RAOs for head horizontal modes. However, good correlations between the results
and bow quartering waves obtained from the present code, the are observed validating the linearized version of the present code.
Rankine panel code SWAN [31], and 3D Green-function based zero For the above results, the RAOs from the present code are
speed radiation–diffraction code. This last method is well known obtained from the motion time histories that the code generates.
and much standard software is available. Here the software used is Thus it is of interest to compare directly the time-domain results
a code termed SOFORCE (Singh and Dhavilkar [32]) developed at IRS with another code that also produces time histories. Fig. 8 shows
(Indian Register of Shipping, Mumbai) which has been extensively such a comparison for the S175 hull interacting with 10 s wave of
validated against industry standard codes like WAMIT. All results height 2 m in head wave condition with results from SWAN which
are for a consistently linearized hydrodynamics problem. In gen- also produces directly the time histories. A very good agreement
eral, the agreement is very good in all cases. Some differences are between the two results is clearly seen. Note that as these results
found in the horizontal motions (sway and yaw) at low frequen- are for a consistently linear formulation, the smallness of wave
cies in oblique waves between the results of Rankine panel code height for the computation is of no consequence.
(SWAN) with the other two. Note that present time domain results We next consider the results of body-exact approximate nonlin-
compare well with the results from the frequency domain code ear hydrodynamics computations. SWAN has a FK nonlinear option,
even in these cases. The reason for the discrepancies of the sway which has been used for these comparisons. Fig. 9 shows time series
and yaw motions from the present code with the Rankine code at results of S175 hull for a wave of 12 s period and height of 10 m in
low frequencies is not clear at this time, but we note that prediction head wave condition between the present nonlinear computations
of horizontal motions are relatively more difficult than the verti- and SWAN nonlinear computations. Some differences between the
cal mode motions because of lack of hydrostatic restoration in the two results are found. To study this, we have compared the results

Fig. 10. Heave and pitch motions of the S175 hull for T = 12 s, H = 10 m, ˇ = 180◦ ; between present linear and FK nonlinear computations.
162 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 11. Hull and free surface panels and a schematic view of a turret mooring system of weathervane ship.

from linear and FK-nonlinear computations from each method over the hull length) is expected because of the significant geometric
a range of wave parameters (frequency and amplitude). This study nonlinearity (flare) of this hull. The present computations do show
has revealed that SWAN produces in general a relatively less dif- significant difference between the linear and nonlinear computa-
ference between linear and FK nonlinear computations compared tions as expected. Comparatively, the SWAN results have shown
to the differences obtained from the present code. Fig. 10 shows a much less difference between its linear and FK nonlinear versions.
comparison of motions between the linear and FK nonlinear com- While there can be several reasons contributing to the difference
putations from the present code for a 12 s 10 m high wave. Note between the SWAN and present nonlinear results, in absence of
that fairly large difference between linear and FK nonlinear motions knowledge on detail implementation of the SWAN nonlinear com-
for this wave (a reasonably high wave with length about 1.2 times putations we are unable to make further comment except to remark

Fig. 12. Surge, heave, and pitch motions of weathervane ship (H = 10 m, T = 12 s, ˇ = 180◦ ); computations for linear hydrodynamics.
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 163

Fig. 13. Initial Transient Surge motions of weathervane ship (H = 10 m, T = 12 s, ˇ = 180◦ ).

Table 2 that our nonlinear results appear to capture at least qualitatively

Main particulars for the tanker.
the expected trend.
Length, breadth, depth (m) 325, 58, 30
Draft (m) 20.5 4. Results and discussion
Centre of gravity (m) (11.0, 0, −6.25)
Radius of gyration Kxx , Kyy , Kzz (m) 20.3, 81.25, 81.25
Total Displacements (T) 314,486 We now consider motions of moored bodies. The first example
Number of body panels 854 taken is a turret moored tanker hull. The geometric particulars of
Number of free-surface panels 1862 the hull are given in Table 2. The turret mooring system consists of
8 lines, and the operating water depth is 200 m. The details of the
mooring lines are given in Table 3.
Fig. 11 provides diagrams of the hull panels, free-surface
panels and the mooring lines. As stated previously, the hydro-
dynamic computations can be linear (Lin-FK) and FK nonlinear

Table 3
Mooring line details for the tanker.

Mooring line number Angle [◦ ] with +ve x axis Fairlead coordinates (m) Anchor coordinates (m)


1 0 145 0 745 0
2 45◦ 137 17 561 441
3 90◦ 120 25 120 625
4 135◦ 102 17 −321 441
5 180◦ 95 0 −505 0
6 225◦ 102 −17 −321 −441
7 270◦ 120 −25 120 −625
8 315◦ 137 −17 561 −441

Fig. 14. Steady state surge, heave, and pitch motions of weather vane ship (H = 10 m, T = 12 s, ˇ = 180◦ ).
164 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 15. Incident FK force and moment acting on weather vane ship (H = 10 m, T = 12 s, ˇ = 180◦ ).

(Nonlin-FK). The mooring line stiffness can also be modelled as of the mooring lines does not appear to result in much differ-
either a linear spring (Lin-Moor) or a nonlinear spring (Nonlin- ence between the motions except in the initial transient surge
Moor). Thus there are four possible combinations of hydrodynamic motion. Careful observation reveals that the difference is owing
and mooring line models for which the computations can be to a lower natural surge period from the nonlinear modelling of
performed. mooring stiffness compared to a linear mooring-stiffness model.
Fig. 12 shows the time histories of surge, heave, and pitch for This is as expected since nonlinear mooring stiffness is larger than
both linear and nonlinear mooring line stiffness model but with linear mooring stiffness.
linear hydrodynamic computations. Results are for head waves of Fig. 13 shows a comparison of the initial transient surge motion
height 10 m and period 12 s. As can be observed, the transients in between the four computations for this wave case: linear and
vertical mode of motions (heave and pitch) decay quickly, within nonlinear mooring stiffness with linear and FK nonlinear hydro-
1–2 periods, and very long-duration steady-state converged time dynamic model. Large differences between the linear and nonlinear
histories could be achieved. As regards surge, the transients take a hydrodynamic model results are found. The lower natural surge
long time to decay; clearly displaying a low-frequency oscillation period in nonlinear mooring-line model compared to the lin-
corresponding to the natural period is surge. Stiffness modelling ear mooring-line model can be observed in the FK nonlinear

Fig. 16. Body and free surface panels (a) rectangular barge (b) six column semisubmersible.
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 165

Table 4
Main particulars of barge and six column twin pontoon semi-submersible.

Structure Rectangular barge Semi-submersible

Length, breadth, depth (m) 150, 50, 20 100, 76, 30

Draft (m) 10 20
Centre of gravity (m) (0, 0, 0) (0, 0, −12.08)
Radius of gyration Kxx , Kyy , Kzz (m) 24.2, 39, 39 30.55, 30.89, 41.74
Total displacements (T) 75,655 35,000
Number of body panels 413 480
Number free-surface panels 1768 2248

hydrodynamic computations. Additionally, the initial transients are

considerably larger for this case. The other observation of interest
is the steady or mean offset in surge from the nonlinear hydro-
dynamic computations, which cannot be predicted from the linear
hydrodynamic computations. This offset is more clearly demon-
strated in Fig. 14 where the steady-state time histories of all the
three motions of surge, heave and pitch are shown. The plots clearly
show that the converged steady-state results are far less influenced
by mooring line modelling (linear vs. nonlinear) compared to the
hydrodynamic model for this case. A steady offset in surge dis- Fig. 17. Spread mooring system for rectangular barge.
placement or the mean drift is clearly visible. The heave and pitch
motions show considerable difference between linear and non-
of 10 m from all the four computations. The plots display similar
linear hydrodynamic computations. While heave amplitude from
features as in the case of the tanker hull: the influence of nonlinear-
nonlinear hydrodynamic model appear to have increased by about
ities from mooring line stiffness is relatively less pronounced than
20%, the increase in pitch amplitudes is almost 50%. Addition-
the influence of nonlinearities in FK hydrodynamic forces. The rela-
ally, the phase difference between heave and pitch from the two
tively shorter periods of the low frequency transient motions from
hydrodynamic models are quite different. The important practical
the nonlinear mooring line model are once again consistent with
consequence of this is that the computed vertical motions at differ-
the relatively larger stiffness of the lines in the nonlinear mooring
ent hull locations from the two computations will be considerably
model compared to the linear model. The steady-state results for
different. In order to study further the differences between linear
these computations are shown in Fig. 20. The differences from
and nonlinear hydrodynamic models, the incident wave hydro-
the linear and nonlinear hydrodynamic computations are clearly
dynamic forces from linear and FK nonlinear computations are
visible: a steady drift is seen in all the three horizontal motions in
plotted in Fig. 15 as only this component of the total hydrodynamic
the FK nonlinear computations which cannot be predicted from the
force is different in the two computations. The difference between
linear hydrodynamic force model. The nonlinear characteristics of
the forces and moments from the two approaches is visible. The
the FK incident forces and moments are evident in Fig. 21 where
nonlinear computations clearly display the presence of higher har-
these are plotted. Another observation of interest is that mooring
monics in surge. Careful observation will also reveal that the mean
values for the nonlinear computations are slightly offset from zero
while for linear computations the mean is zero as expected. Com-
putations have also been performed for several other combinations
of wave parameters (heading, period, height) for this geometry and
similar features have been observed.
The next considered computations are for a rectangular barge
and a six column twin pontoon semisubmersible. Fig. 16 shows
the discretized panels of hull and the computational free-surface
domain. The geometric details of these two structures are given in
Table 4. The barge is spread moored with 6 lines, as shown in Fig. 17.
Table 5 provides the details of the mooring line parameters oper-
ating in water depth of 200 m. Fig. 18 displays the load-excursion
characteristic of the lines.
First we discuss the barge results. Fig. 19 shows the initial tran-
sient part of the horizontal plane motions (surge, sway and yaw)
for a bow quartering wave of heading 135◦ , period 11 s and height Fig. 18. Load-excursion characteristics of the mooring lines used for the barge.

Table 5
Spread mooring system parameters of a barge.

Mooring line number Angle [◦ ] with +ve x axis Fairlead coordinates (m) Anchor coordinates (m)


1 45◦ 75 25 499 449

2 90◦ 0 25 0 625
3 135◦ −75 25 −499 450
4 225◦ −75 −25 −500 −448
5 270◦ 0 −25 0 −625
6 315◦ 75 −25 498 −450
166 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 19. Initial transient surge, sway and yaw motions of barge (H = 10 m, T = 11 s, ˇ = 135◦ ).

Fig. 20. Steady state surge, sway and yaw motions of barge (H = 10 m, T = 11 s, ˇ = 135◦ ).

Fig. 21. Steady state surge, sway and yaw motions of barge (H = 10 m, T = 18 s, ˇ = 135◦ ).
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 167

Fig. 22. Steady state surge, sway and yaw motions of barge (H = 6 m, T = 8 s, ˇ = 135◦ ).

line stiffness model also appears to have some influence on the load excursion plot for the lines are presented in Table 6 and Fig. 24
motions. For example, the nonlinear mooring line stiffness model respectively.
results in a relatively lower mean offset compared to only linear The horizontal plane motions of this structure in a wave of
stiffness model. This is also an expected behaviour since larger period 10 s and height 8 m for 135◦ wave heading is shown in
stiffness should results in a lower mean offset. Note that a similar Fig. 25. In this plots, the yaw results are shown for a considerably
feature was also found in case of the tanker hull (Fig. 14). At larger longer duration to capture the low-frequency yaw oscillation
wave-length, the mean offset in surge and sway from the nonlinear at its natural frequency. Once again the qualitative features are
hydrodynamic model tends to reduce and converge towards the similar to previous results, but quantitatively some differences are
results from linear hydrodynamic computations as can be seen observed. For example, the difference between the motions from
in Fig. 21 where plots are displayed for a wave of period 18 s and linear and nonlinear hydrodynamic models is relatively much less
height 10 m. This is most likely because at higher wave lengths compared to the similar differences in the tanker and barge cases.
(lower frequencies) the mean drift force tend to reduce for this The offset in surge and sway is also much less, of the order of only
body. On the other hand, at higher frequencies, for this box-like 0.5–1 m compared to an offset of about 2.5–4 m obtained in case of
body with vertical walls and large water plane area, the mean the barge in a comparable 11 s, 10 m high wave. This is most likely
drift force is expected to be large with a consequent large drift in because of the low water-plane area of the hull combined with
horizontal motions. Fig. 22 shows the horizontal motions of barge the low relative vertical motion of the body, resulting in a small
interacting with wave of period 8 s and height 6 m. Observed mean overall difference between the exact and mean wetted surface area
drift for periods 8, 11, and 18 s are respectively about 3.8, 2.5 and even when the motions are moderate. In contrast, the tanker and
0.5 m in surge and 2.2, 4 and 0.3 m in sway. The trend in mean drift the barge having large water-plane areas will have considerable
can therefore be captured by the nonlinear hydrodynamic model.
The twin pontoon six column semi-submersible is moored with
four lines is shown in Fig. 23. This structure is placed in a deeper
water of depth 1000 m. The mooring line characteristics and the

Fig. 23. Spread mooring system of semisubmersible. Fig. 24. Load excursion plot of the semisubmersible mooring line.
168 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

Fig. 25. Steady state surge, sway and yaw motions of semisubmersible (H = 8 m, T = 10 s, ˇ = 135◦ ).

Fig. 26. Incident FK force acting on semisubmersible (H = 8 m, T = 10 s, ˇ = 135◦ ).

Fig. 27. Steady state surge, sway and yaw motions of semisubmersible (H = 10 m, T = 22 s, ˇ = 135◦ ).
S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170 169

Table 6
Spread mooring system parameters of a six column twin pontoon semi-submersible.

Mooring line number Angle [◦ ] with +ve x axis Fairlead coordinates (m) Anchor coordinates (m)


1 45◦ 50 37 1464 1451

2 135◦ −50 37 −1464 1451
3 225◦ −50 −37 −1464 −1451
4 315◦ 50 −37 1464 −1451

differences between the exact and mean wetted surface areas. stiffness however become important in some cases, particularly
For the semisubmersible geometry therefore the variations in the when motions are computed with nonlinear hydrodynamic loads.
forces and moments between nonlinear and linear hydrodynamic For example, the semi-submersible results show considerable
computations will be small. This can be observed in the incident difference in the horizontal mode motions when nonlinear FK and
FK surge and sway force plots in Fig. 26 which shows that the hydrostatic loads are considered.
forces from linear and nonlinear computations are quite close. The The influence of nonlinearities in wave loads and mooring line
relatively large-period yaw oscillation is due to low yaw-stiffness stiffness on the motion responses of the body in the horizontal
from the mooring lines for this structure with the chosen line plane depends on several parameters: the geometry of the object,
parameters. However, at a higher period of 22 s, the mean drift wave parameters and water depth, mooring line characteristics,
from nonlinear hydrodynamic model is found to be significant, as etc. Depending on the various combinations, there may or may not
shown in Fig. 27. At this period heave, pitch and roll motions are be significant influence of these nonlinearities. As the developed
found to be significant. Large drift force and moment are therefore method can account for the significant associated nonlinearities of
expected due to large relative motions in the vertical plane. hydrostatic and incident wave loads as well as mooring line stiff-
As regards the effect of mooring line stiffness on the motions, ness directly in its solution for the motions of the object and can
this is insignificant when computations are with linear hydrody- generate motion time histories over very long durations with only
namics. However, when computations are performed with nonlin- modest computational effort, it can be considered a versatile and
ear FK hydrodynamic model, differences in the motions between practically useful computational tool for analysis and simulation
linear and nonlinear mooring line stiffness model are of compa- studies of motions of large floating systems under steep incident
rable magnitude with differences in motions resulting from linear waves. Typical computation time for simulation of motions of prac-
and nonlinear hydrodynamic models. For the 22 s wave (Fig. 27), tical structures like a semisubmersible for over 10,000 time steps
the mean drift in surge with linear and nonlinear mooring stiffness spanning over 100 wave periods is in the order of 20 min in a usual
are about 8 and 3 m, and the corresponding values for sway are 3.5 desktop PC of 3.10 GHz and 8GB RAM (for example, the semisub-
and 3 m respectively, indicating a strong influence of nonlineari- mersible results simulated over 140 wave periods in Fig. 25 took
ties in mooring line stiffness on drift. Yaw motions show significant 30 min). This can be contrasted with the time taken from a full non-
difference not only between computations from linear and nonlin- linear MEL simulation, e.g. Ma and Yan [20] in their very efficient
ear hydrodynamic models, but also between linear and nonlinear MEL solver reports a computational time of about half a day to
stiffness models particularly when hydrodynamic model is nonlin- one day for simulating about 15 periods of motions of a Wigley
ear. This signifies the importance of using both FK nonlinearity in hull.
hydrodynamic loads and nonlinearity in the mooring line stiffness. The nonlinear hydrodynamic model of the present method can
predict the mean offset in horizontal modes of motions which is not
5. Conclusion possible from a linear hydrodynamic model. The approach followed
here enables it to be extended for more complex but practically
In this paper, a coupled direct time domain solution is formu- relevant situations such as motion studies for multiple bodies in
lated based on a 3D NWT approach wherein the moored floating close proximity, structures exposed to other environmental loads
body is placed inside a circular numerical tank with an existing like wind and current loads.
incident wave field. Under some simplifying approximations to
the boundary value hydrodynamic problem wave radiation and References
diffraction, termed here as interaction hydrodynamics, are com-
puted using a boundary element technique based on Rankine [1] Wichers JEW. Slowly oscillating mooring forces in single point mooring sys-
panel method. The method can consider the exact FK forces and tems. In: 2nd international conference on behaviour of offshore structure. 1979.
[2] Ormberg H, Larsen K. Coupled analysis of floater motion and mooring dynamics
hydrostatic restoring loads. Additionally, both linear and nonlinear
for a turret-moored ship. Appl Ocean Res 1998;20(1):55–67.
stiffness of the mooring lines can be considered. [3] Chen X, Zhang J, Ma W. On dynamic coupling effects between a spar and its
A verification and validation of developed scheme has been mooring lines. Ocean Eng 2001;28(7):863–87.
[4] Kim S, Sclavounos PD. Fully coupled response simulations of theme offshore
carried out for the standard S175 container ship based on linear
structures in water depths of up to 10,000 feet. In: Proceedings of the 11th
formulation and the approximate nonlinear formulation. Numer- international offshore and polar engineering conference, ISOPE, 2001. 2001.
ical results are compared with results obtained from other linear [5] Heurtier JM, Le Buhan P, Fontaine E, Le Cunff C, Biolley F, Berhault C. Cou-
solvers which are based on zero speed Green function and Rank- pled dynamic response of moored FPSO with risers. In: Proceedings of the 11th
international offshore and polar engineering conference, ISOPE. 2001.
ine panel method. Comparison shows that the present scheme is in [6] Garrett DL. Coupled analysis of floating production systems. Ocean Eng
good agreement with both time and frequency domain results. 2005;32(7):802–16.
The developed method has been applied to study motions of [7] Tahar A, Halkyard J, Irani M. Comparison of time and frequency domain analysis
with full scale data for the Horn Mountain Spar during hurricane Isidore. In:
three moored floating bodies, a weathervane turret moored tanker Proceedings of the 25th offshore mechanics and arctic engineering conference,
hull, a rectangular barge and a semisubmersible. These latter two OMAE. 2006. p. 105–13.
structures are spread-moored. In all cases, results demonstrate [8] Low YM, Langley RS. A hybrid time/frequency domain approach for
efficient coupled analysis of vessel/mooring/riser dynamics. Ocean Eng
that the influence of nonlinearity in the mooring line stiffness 2008;35:433–46.
is relatively small compared to the influence of FK and hydro- [9] Tahar A, Kim MH. Coupled-dynamic analysis of floating structures with
static nonlinearity on the motions. The nonlinearities in the line polyester mooring lines. Ocean Eng 2008;35:1676–85.
170 S. Ganesan T, D. Sen / Applied Ocean Research 51 (2015) 153–170

[10] Zhang F, Yang J-M, Li R-P, Chen G. Coupling effects for cell–truss spar plat- [22] Sen D, Srinivasan N. Long duration simulation of wave-structure inter-
form: comparison of frequency- and time-domain analyses with model tests. J actions in a numerical wave tank. In: 10th international symposium on
Hydrodyn B 2008;20:424–32. practical design of ships and other floating structures. Houston: ABS; 2007.
[11] Kim JH, Kim YS, Hong SY, Kim BW, Shin SH. An experimental study on extreme p. 122–31.
responses of a moored semi-submersible rig. In: Proceedings of the 21st inter- [23] Sen D. Time-domain simulation of motions of large structures in nonlinear
national offshore and polar engineering conference, ISOPE, 2011. 2011. waves. In: Proceeding of 21st offshore mechanics and arctic engineering con-
[12] Yang M, Teng B, Ning D, Shi Z. Coupled dynamic analysis for wave interaction ference – OMAE. 2002. p. 7–17.
with a truss spar and its mooring line/riser system in time domain. Ocean Eng [24] Kim MW, Koo W, Hong SY. Numerical analysis of various artificial damp-
2012;39:72–87. ing schemes in a three-dimensional numerical wave tank. Ocean Eng
[13] Yang M, Teng B, Xiao L, Ning D, Shi Z, Qu Y. Full time-domain nonlinear coupled 2014;75:165–73.
dynamic analysis of a truss spar and its mooring/riser system in irregular wave. [25] Cointe R, King GP, Molin B, Tramoni BM. Nonlinear and linear motions of a rect-
Sci China Phys Mech Astron 2014;57(1):152–65. angular barge in a perfect fluid. In: 18th symposium on naval hydrodynamics.
[14] Garrett DL. Dynamic analysis of slender rods. J Energy Resour Technol The National Academies Press; 1990.
1982;104(4):302–6. [26] Ferrant P. Simulation of strongly nonlinear wave generation and wave–body
[15] Kim BW, Sung HG, Kim JH, Hong SY. Comparison of linear spring and nonlinear interactions using a 3-D MEL model. In: 21st symposium on naval hydrody-
FEM methods in dynamic coupled analysis of floating structure and mooring namics. Washington, DC: National Academy Press; 1996. p. 93–109.
system. J Fluids Struct 2013;42:205–27. [27] Rienecker MM, Fenton JD. A Fourier approximation method for steady water
[16] Cummins WE. The impulse response function and ship motions, No. Dtmb- waves. J Fluid Mech 1981;104:119–37.
1661. Washington, DC: David Taylor Model Basin; 1962. [28] Barltrop NDP, editor. Floating structures: a guide for design and analysis, vol. II.
[17] Ogilvie T. Recent progress towards the understanding and prediction of ship The Centre for Marine and Petroleum Technology Publication; 1998 [Chapter
motions. In: Sixth symposium on naval hydrodynamics. 1964. 9].
[18] Van den Boom HJJ. Dynamic behaviour of mooring lines. In: International Con- [29] Faltinsen OM. Sea loads on ships and offshore structures. Cambridge University
ference on Behaviour of Offshore structure. Delft: BOSS; 1985. Press; 1990 [Chapter 8].
[19] Longuet-Higgins MS, Cokelet ED. The deformation of steep surface waves on [30] Ganesan T S, Sen D. Time-domain simulation of large amplitude wave structure
water. I. A numerical method of computation. Proc R Soc Lond A: Math Phys Sci interactions by a 3D numerical tank approach. J Ocean Eng Mar Energy 2014.
1976;350(1660):1–26. Springer [in press].
[20] Ma QW, Yan S. QALE-FEM for numerical modeling of nonlinear interac- [31] Sclavounos PD. Computation of wave ship interactions. Computational
tion between 3D moored floating bodies and steep waves. J Fluids Struct Mechanics Publications; 1996.
2009;78:713–56. [32] Singh SP, Dhavalikar S. Computation on wave loads on floating structures. In:
[21] You J, Faltinsen OM. A 3D fully nonlinear numerical wave tank with a moored 3rd Indian national conference on harbour and ocean engineering. Goa: NIO;
floating body in shallow water. In: 22nd international offshore and polar engi- 2004. p. 266–73.
neering conference. 2012. p. 908–14.