Ocean Engineering 164 (2018) 350–366

Contents lists available at ScienceDirect

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

Nonlinear analysis of sloshing and floating body coupled motion in the time- T
domain
Shuo Huanga,b,c,∗, Wenyang Duand, Xuliang Hane, Ryan Nicollf, Yage Youa,b,c, Songwei Shenga,b,c
a
Guangzhou Institute of Energy Conversion, Chinese Academy of Sciences, Guangzhou, China
b
CAS Key Laboratory of Renewable Energy, Guangzhou, China
c
Guangdong Provincial Key Laboratory of New and Renewable Energy Research and Development, Guangzhou, China
d
College of Shipbuilding Engineering, Harbin Engineering University, Harbin, China
e
CNOOC Research Institute Co., Ltd., Beijing, China
f
Dynamic Systems Analysis Ltd., Victoria, BC, Canada

A R T I C LE I N FO A B S T R A C T

Keywords: The aim of this paper is to develop a coupled nonlinear time-domain simulation scheme for nonlinear interac-
Nonlinear time-domain simulation tions among sloshing flows and floating body motion for both regular and irregular wave excitation. The con-
Sloshing tributions of a variety of nonlinear factors, outside waves and inside sloshing induced forces, as well as their
ISITIMFB influences on body coupled sway and roll motion were investigated. The induced forces are due to the changes in
Energy dissipation
the transient wet surface of the floating body and full nonlinear sloshing. The effects of tank fill ratio and
excitation wave height on the nonlinear coupled motion, as well as the relationship between sloshing and
floating body nonlinear coupled motion under large wave amplitudes and severe sea conditions were also in-
vestigated and the results are presented. Finally, the numerical solutions are compared with existing experi-
mental result. The fully nonlinear sloshing and floating body coupled motion are simulated based on the po-
tential flow theory, with the transient position hydrodynamic assumption. The boundary value problem is solved
by the B-spline higher-order panel method. The ISITIMFB (iterative semi-implicit time integration method for
floating bodies) is applied to solve for the body's velocity and displacements. The sloshing energy dissipation is
modeled by changing the boundary condition on the tank's solid boundaries. An extended principle to determine
the energy dissipation coefficient for both regular and irregular cases is extracted. Then, the sloshing and floating
body nonlinear coupled motion under large wave amplitudes and severe sea conditions are investigated, and the
numerical solutions are compared with existing experimental results. The effects of tank fill ratio and excitation
wave height on the nonlinear coupled motion is also investigated.

1. Introduction
Loading and offloading operations of vessels such as FPSOs and LNG
carriers near shore or in the open sea can cause liquid sloshing in the
carrier's insulated tanks, which are very large and partially filled with
fluid. Violent sloshing waves can generate large loads which can da-
mage the structure or influence stability (vessel sway and roll motion).
In this coupled system, both the global motions of the floating body and
the tank sloshing flows will influence each other.
The literature review revealed several interesting studies on this
topic. Generally, the studies can be categorized into two approaches:
(1) the linear frequency-domain approach based on potential theory
(Molin et al., 2002; Malenica et al., 2003) and (2) the time-domain
approach. According to previous studies, the assumption of linear
sloshing and ship motion appears adequate in small amplitude coupled
motion analysis. However, even if the incident wave is small and in the
linear region, violent nonlinear sloshing can occur due to resonance.
Sloshing-induced impact pressure can lead to a significant change in the
transient wet surface and balance position of the hull in a certain fre-
quency range. The hydrodynamic load characteristics on the hull are
nonlinear, which inherently cannot be accounted for in linear theory.
Some conclusions that can be drawn from previous studies include
the following: sloshing may be violent which can be modeled by Navier-
Stokes equation and nonlinear potential theory model; the motion of
floating structures can be linear or nonlinear, but at the outside of the
wave it is linear; hydrodynamic forces can be calculated in the fre-
quency-domain or by strip theory; Froude–Krylov forces can be non-
linear. The current study focuses on the influencing characteristics of

∗
Corresponding author. Guangzhou Institute of Energy Conversion, Chinese Academy of Sciences, Guangzhou, China.
E-mail address: huangshuo@ms.giec.ac.cn (S. Huang).

https://doi.org/10.1016/j.oceaneng.2018.06.003
Received 20 March 2017; Received in revised form 29 May 2018; Accepted 4 June 2018
Available online 28 June 2018
0029-8018/ © 2018 Elsevier Ltd. All rights reserved.
S. Huang et al.
sloshing and coupled effects on ship motions. To model these, the be-
haviors of both the external and the internal flow are needed. For the
external flow, models based on linear potential theory are used and are
accurate enough in most cases except for viscous effects such as roll
damping (Faltinsen and Timokha, 2009). For the internal flow, models
based on linear potential theory and Computational Fluid Dynamics
(CFD) are used. Sloshing is a nonlinear phenomenon, especially in re-
sonance conditions. Although in some cases it is possible to accurately
predict the coupled sloshing and ship motions by using linear theory
(Rognebakke and Faltinsen, 2003), CFD methods account for non-
linearity in sloshing that cannot be ignored in some cases. Kim (2002,
2003) presented a series of numerical results of the anti-rolling tank and
ship coupled motion by using a finite-difference method and the com-
mercial sea-keeping program LAMP. Kim et al. (2007), Nam et al.
(2009) and Lee et al. (2007) studied the coupled effects between the
ship motion and the sloshing flows by using the impulse response
function (IRF) formulation for linear ship motion and a CFD simulation
for nonlinear sloshing flows. Li et al. (2012) investigated the coupled
motion of sloshing and ship motion using the STF & IRF method for ship
motion, and the CFD model OpenFOAM for the nonlinear sloshing
flows. The coupled model is established for the motion response of the
ship in waves coupled with induced internal sloshing and their effects
on sloshing-induced impact loads, which was analyzed by Jiang (2014)
and Xue et al. (2017). The linear ship motion is solved with an impulse
IRF method combined with OpenFOAM to address the viscous sloshing
flow. However, generally speaking, CFD methods calculation time is
prohibitively long and a more computationally efficient solution for the
problem is needed.
A study of global ship motion coupled with tank sloshing using
potential flow theory for flow inside and outside the vessel was con-
ducted by Hong et al. (2012) and Ning et al. (2012). The research shows
that the wave elevation and pressure can be obtained exactly if a highly
nonlinear phenomenon does not occur. A 2-D coupled numerical code
was developed considering nonlinear sloshing and linear ship motions
based on a boundary element method, and later the results were ver-
ified by a series of corresponding model tests by Zhao et al. (2014).
Zhang (2016) extended the modal decomposition method for floating
vessel simulations and developed an explicit seakeeping-sloshing de-
coupling approach. In this method, the exact acceleration of a vessel at
any instant is calculated explicitly in one step without iterations. Three-
dimensional cases of a floating vessel with four nonlinear sloshing tanks
were also simulated. This method was further used by Wang et al.
(2017) in their 2D simulations. Su and Liu (2017) also investigated the
coupling effects of barge motion and sloshing, using a nonlinear
Boussinesq-type approach in terms of velocity potential for the sloshing
simulation. In the preliminary investigation on the coupling mechanism
between ship motions and nonlinear sloshing, roll motion is most af-
fected compared to other degrees of freedom.
Faltinsen and Timokha (2001, 2005) found that the damping due to
viscous energy dissipation in the form of sloshing wave spilling and
breaking in a certain frequency range plays an important role in de-
termining the coupled motion and it must be carefully considered. In
the past, such problems have been analyzed by using the Navier-Stokes
equations (Su et al., 1982 and Hamlin et al., 1986). The calculation
time of this method is prohibitively long, so a more computationally
efficient solution based on potential theory is considered. With poten-
tial theory, viscous effects are neglected and so development is needed
to resolve an appropriate dissipation effect. Kim and Shin (2008) pro-
posed to add a damping term to the free surface condition. Un-
fortunately, this approach introduces two damping coefficients, which
are not easy to define for modeling the coupled motions of ships with
sloshing in internal tanks. Malenica et al. (2003) proposed a model that
modifies the tank wall condition instead of the free surface condition,
but it is only suitable for frequency-domain analysis. This method was
improved by Huang et al. (2011), who reported a new energy dissipa-
tion condition which is applicable to time-domain simulation.
351
Ocean Engineering 164 (2018) 350–366
The nonlinear body boundary conditions of a floating body and
tank-sloshing are complex. The accurate calculation of the velocity
potential for the time derivative is critical to obtain the correct pressure
and force on the body surface at each time step. To decouple this mu-
tual dependence, some methods have been suggested in the literature,
such as the indirect method (Wu and Eatock, 1994), the mode-de-
composition method (Koo and Kim, 2004) and the iterative method
(Yan and Ma, 2007). Each method has its own advantages and dis-
advantages. In this paper, the iterative method is applied due to the
importance of nonlinearities for sloshing and to obtain the probability
density functions of response variables caused by sloshing.
There are still many uncertainties in the methods described in the
literature. The method based on the linear assumption fails to in-
corporate the strong nonlinearities of internal sloshing and the transient
position hydrodynamic nonlinearities of external ship motion, while the
CFD method can be prohibitively time consuming. Thus, establishing a
fast, efficient time-domain prediction technique, which is capable of
considering the strong nonlinear sloshing flows and nonlinear ship
motion to simulate the real motion of vessels like an LNG carrier or
FPSO in regular and irregular waves is necessary. For vessel design, a
good understanding of the ship's performance and reliability is re-
quired. However, to date, there has been a lack of systematic in-
vestigation of the nonlinear effects analysis of sloshing and ship coupled
motion, especially applied to sloshing energy dissipation.
In the present study, a nonlinear numerical model is developed to
examine the violent sloshing coupled with the floating body motion.
The fully nonlinear model, the linear free surface condition and the
transient nonlinear body condition satisfy the Green function, adopted
to solve for the inner tank flow and the ship's outer domain. In the
numerical implementation, the ISITIMFB (iterative semi-implicit time
integration method for floating bodies) is applied to solve for the body's
velocity and displacements with high efficiency with acceptable accu-
racy. The sloshing coupled with nonlinear floating body motion is si-
mulated for large amplitudes of regular and irregular waves and the
numerical solutions are compared with the existing experimental re-
sults. The change in wave and sloshing-induced forces caused by the
change in hull wet surface and the full nonlinear free surface of the
internal fluid on sway and roll are analyzed in detail, with considera-
tion given to a variety of non-linear contributions and mechanisms.
2. Numerical modeling method
2.1. Coordinate system
Three Cartesian co-ordinate systems are defined: in one, OYZ is fixed
in space and in the others, oyz andoT yT zT , are fixed on the floating body
and on the tank, respectively. When the tank and floating body are at
rest, theOYZ andoyz systems coincide and their origins are at the center
of the undisturbed free surface along the vertical line through the center
of gravity of the floating bodyG0 . The transient wetted surface of the
body under the free surfaceSF is defined asSH . The positive direction of
the normal vector⇀ n points to the outside of the fluid domainSΩ . The
free surface of the tank liquid is defined asSTF and tank wall wetted
surface asSTH . The directions of the axes are shown in Fig. 1. The re-
lationships between the coordinate frames are:
⎧∇YZ = ∇yT zT
⎪ ∂ ⇀
( ) ( ∂
⎪ ∂t YZ = ∂t − U ∇ y z ) T T [T ] = ⎡
cos α − sin α ⎤
⎨ ⎢ sin α cos α ⎥
y y ⎣ ⎦
⎪⎜⎛Y ⎟⎞ = ⎜⎛ o ⎟⎞ + [T ] ⎜⎛ T ⎟⎞
⎪ Z zo zT
⎩⎝ ⎠ ⎝ ⎠ ⎝ ⎠ (1)
where [T ] is the transformation matrix andα is the angle between frames
OYZ and oyz.
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 1. Coordinate systems.

2.2. Modeling of fully nonlinear sloshing by using:

2.2.1. The boundary value problem for tank sloshing ∂ϕ ⇀ 1

P = −ρ ⎡ − U ∇ϕ + ∇ϕ 2 + gz oT + g (T21 yT + T22 ςb) ⎤
An ideal fluid is assumed so that the fluid velocity can be described ⎣ ∂t 2 ⎦ (5)
by the gradient of the velocity potentialϕ . The governing equations of
the velocity potential in OYZ system is given by
2
⎧∇ ϕ = 0 in SΩ 2.2.2. An approximation for sloshing energy dissipation
⎪ ∂ς + ∂ϕ ∂ς − ∂ϕ = 0 on Z = ς (y, t ) The theoretical wave elevation given by Eq. (2) does not account for
⎪ ∂t ∂Y ∂Y ∂Z
⎪ ∂ϕ 1 2 + gς = 0
viscous effects and can diverge at the resonance frequency. The energy
+ ∇ ϕ on Z = ς (y, t )
∂t 2 dissipation from viscosity prevents divergence in experimental tests. To
⎨ ⎯⇀
⎯ ⎯⇀
⎯
⎪ ∂ϕ = U · n on STH obtain a theoretical solution in good agreement with the experimental
⎪ ∂n results, the viscous effect should be considered.
⎪ ϕ = 0(t < 0), ∂ϕ = 0 (t = 0, Z = 0)
⎩ ∂t (2) To consider the viscous effect in the velocity potential problem
defined in Eq. (2), Huang et al. (2011) assumed that energy dissipation
It can be seen that the free surface condition is more easily de- mainly occurs in the boundary layer near the tank wall, while solving
scribed in the oT yT zT moving coordinate system of the liquid tank. the sloshing problem in the time-domain. To approximate the dissipa-
Substituting Eq. (1) into the free surface condition Eq. (2), produces the ⇀
tion, the wall boundary condition is modified from ∂n = U ⋅⇀
∂ϕ
n to:
governing equation of the velocity potential in oT yT zT coordinate
system as given by Wu et al. (1998A). ∂ϕ ⇀ μ ∂ϕ
= U ⋅⇀
n − nY , ⇀
n = (nY , nZ ) on STH
2 ∂n g ∂t (6)
⎧∇ ϕ = 0 in SΩ
⎪ ∂ςb
⎪
( ∂ϕ
)∂ςb ∂ϕ
(
⎪ ∂t + ∂yT − U ∂yT − ∂zT − W = 0 ) on zT = ςb (yT , t ) wherenY andnZ are the respective components of ⇀ n in the spatial sys-
temOYZ and μ is the energy dissipation coefficient. In Eq. (6), the term
∂ϕ ⇀ 1
− U ∇ϕ + 2 ∇ϕ 2 + gz oT + g (T21 yT + T22 ςb) = 0 on zT = ςb (yT , t ) μ ∂ϕ
⎨ ∂t n depends on the wetted tank surface shape and only becomes
g ∂t Y
⎪ ∂ϕ = → ⎯⇀ ⇀ ⇀ effective on a wall with a nonzero normal vector horizontal compo-
⎪ ∂n (UoT + Ω × rb )· n on STH
⎪ ∂ϕ nents.
ϕ = 0(t < 0) , ∂t = 0 (t = 0, zT = 0)
⎩ Substituting Eq. (1) into Eq. (6), the modified wall boundary con-
(3) dition in oT yT zT moving system is governed by
whereSTF is defined byzT = ςb (yT , t ) and ςb is the free surface elevation ∂ϕ μ ∂ϕ
⇀
→ = U ⋅⇀
n − T11 n yT + T12 n zT
measured in the moving systemoT yT zT , U is the velocity of the tank, ⇀ rb is ∂n g ∂t
a vector from oT to the point in consideration, g is the gravitational ac- μ ⇀
→ + ( U ∇ϕ) T11 n yT + T12 n zT on STH
celeration, and U and W are the components of U in the yT andzT direc- g (7)
tions, respectively. Tij are entries in the matrix [T ]. The incident wave is
generated by the motion of the tank. The outgoing wave is dissipated by Note that the time derivative of the velocity potential ∂ϕ/ ∂t is re-
an energy dissipation condition as described below in Eq. (6). quired for accurately solving for the boundary condition.
At each time step, the velocity potential and the corresponding free
surface profiles are obtained by solving the discretized form of the
following integral equation: 2.2.3. Velocity potential time derivative calculation in a moving system
The simplest way to calculate the velocity potential time derivative
∂φ ⎞
− c (→
p ) φ (→
p)+ ∫ ⎛φ ∂∂Gn −G
∂n ⎠
dl = 0 ∂ϕ/ ∂t is to use a backward finite difference scheme. However, the
SΩ ⎝ (4) scheme suffers from the problem of saw-tooth instabilities (Sen, 1993).
An alternative approach (Wu, 1998B) is to find ∂ϕ/ ∂t by solving a si-
where G is the Green function satisfying the Laplace equation and c (→ p)
milar boundary value problem to that forϕ . The advantage of this
is the solid angle. For two-dimensional problems, the simple source G is
method is that not only ∂ϕ/ ∂t can be obtained on the free surface but
given by G = ln rpq , whererpq is the distance between the source q (ξ , η)
also the energy dissipation items.
and the field points p (y, z ) (Duan, 1995).
⇀ In the spatial systemOYZ , the boundary value problem for∂ϕ/ ∂t is
The sloshing-induced force Fsloshing can be calculated by integrating
defined by
Bernoulli's pressure over the tank instantaneous wetted surface. The
pressure in the fluid can be similarly evaluated in oT yT zT moving system

352
S. Huang et al. Ocean Engineering 164 (2018) 350–366

∂ϕ ∂
⎧∇2 ∂t = 0 in SΩ ⎧ σ (q, τ ) = ∂nq [ϕp (q, τ ) − ϕpi (q, τ )]
⎪ ∂ϕ ⎪
⎪
⎪ = − 1 ∇ϕ 2 − gς on Z = ς (y, t ) rpq ∼
∂t 2 G (p , t; q, τ ) = ln r δ (t − τ ) − H (t − τ ) G (p , t; q, τ )
⎨ ⎨ pq′

⎪ ∂ ∂ϕ = ⎡U→˙ → ∂ → ⎪G∼ ∞ g k (z + ζ )
rb − Ω × UT ⎤ ⇀
+ Ω˙ × ⇀ n − ∂n (U ∇ϕ) on STH ⎪ (p , t; q, τ ) = 2∫0 e cos k (Y − ξ )sin gk (t − τ ) dk
⎪ ∂n ∂t ⎢ T ⎥ ⎩ k (13)
⎩ ⎣ ⎦ (8)
A andB are the points of intersection between the body surface and the
Applying ∂ϕ/ ∂t to Eq. (1) shows free surface, given as

∂ ⎛ ∂ϕ ⎞ ∂ ⎛ ∂ϕ → ∂ ⎯⇀ ⎯⇀
= − U ∇ϕ ⎞ V˙ (τ ) A−B = [ X (A, τ ) − X (B, τ )]
∂n ⎝ ∂t ⎠YZ ∂n ⎝ ∂t ∂τ (14)
⎠ yT zT (9)

Then substituting Eq. (9) into Eq. (8), the boundary value problem
2.3.2. Equations of coupled motion for the 2D case
for ∂ϕ/ ∂t inoT yT zT moving system is governed by
The coupled motion of a floating body is governed by
∂ϕ ⇀
⎧∇2 ∂t = 0 in SΩ ⎧[M ] U˙ c = ⇀
F ext
⎪ ∂ϕ
⎪ =⇀ 1
U ∇ϕ − 2 ∇ϕ 2 − gz oT − g (T21 yT + T22 ςb) on zT = ςb (yT , t ) ⎯⇀
⎯
⎨[I ] Ω˙ = ⎯⇀
∂t ⎩ N ext (15)
⎨
→˙ → ⇀
( )
⎪ ∂ ∂ϕ ⎡ ˙ ⇀ ⎤
⎪ ∂n ∂t yT zT = ⎣UoT + Ω × rb − Ω × UoT ⎦ n
⎩
on STH ⇀ ⎯⇀
where F ext and N ext are the total forces and moments acting on the
⇀ ⎯⇀
⎯
floating body, U̇c and Ω̇ are the accelerations of the gravitational cen-
(10)
ters, and [M ] and [I ]are the mass and inertia matrices, respectively.
There is also a difficulty with solving for the time derivative ∂ϕ/ ∂t of Combining the gravitational and the external hydrostatic forces to a
the velocity potential in Eq. (10). The accelerations of the body should restoring force, the total force can be calculated as follows:
be known when solving the boundary value problem for ∂ϕ/ ∂t . In turn, ⇀ ⇀ ⇀ ⇀ ⇀ ⇀
to compute the force, ∂ϕ/ ∂t is needed. It should be noted here that the ⎧ F ext = F + Fsloshing − Kη2 j − Cη˙ 2 j − mg k
μ ∂ϕ
energy dissipation term g ∂t nY in Eq. (6) is not considered in the ⎨ ⎯⇀
N
⎯⇀ ⎯⇀ ⎯⇀
= N + Nsloshing − Nc
⎩ ext (16)
boundary value problem for∂ϕ/ ∂t ; its effects are only computed in the
iterative method. The scheme to overcome these difficulties is discussed where K is the spring constant, C is the damping coefficient and m is
in detail in Section 2.5. the mass of the floating body. To overcome the divergent phenomena
caused by integration of ∂ϕ/ ∂t after differentiating t in Bernoulli's
pressure over the instantaneous wetted surface at each time step, which
2.3. Nonlinear floating body hydrodynamics modeling requires the hydrodynamic forces are resolved by integrating first and
differentiating in a second stage (Duan, 1995). The wave-induced force
⇀
2.3.1. Body nonlinear boundary value problem F and the moment⎯⇀ N can be evaluated by
Waves with large amplitudes can induce large ship motions. The
nonlinear body case is developed to identify the nonlinearity of the ⎧⇀ ⎡d ⎡
⎪ F = −ρ ⎢ dt ∫S ϕ⇀ n ds − ρ∫S ⎢Un ∇ϕ
hydrodynamic force of the freely floating body. The incident wave ⎪ H (t ) H (t )
⎢
⎣ ⎣
velocity potential is given asϕ = ϕ0 + ϕp . ϕ0 is a prescribed incident ⎪
wave velocity potential. The boundary conditions of the disturbance ⎪
⎤ ⇀⎤
− 2 (∇ϕ)2⇀ n ⎥ ds +g∫S z⇀ n ds + ρϕ ⎛⎜A, 0, t ⎞⎟ V˙ (t )
⎪ 1
velocity potentialsϕp are described by k⎥
⎪ H (t )
⎝ ⎠
⎪ ⎦ A−B ⎦
2 ⎪
⎧∇ ϕp = 0 in SΩ ⎯⇀ ⎡ d ⎛⇀ ⇀⎞ ⇀
⎪ ∂2ϕ
p ∂ϕp ⎨ N = −ρ ⎢ dt ∫SH (t ) ϕ ⎜ rb × n ⎟ ds − ∫SH (t ) rb × [Un ∇ϕ
⎪ ⎝ ⎠
⎪ ∂2t + g ∂Z = 0 on Z =0 ⎪ ⎣
∂ϕp
⎪
⎨ ⇀ ∂ϕ
= U ⋅⇀
n − ∂n0 on SH ⎪ − 2 (∇ϕ)2⇀
1
n )] ds +g∫S z (⇀ n ) ds + ρϕ ⎛⎜A, 0, t ⎞⎟ V˙ (t ) rbY
rb × ⇀
⎪ ∂n ⎪ H (t )
⎪ ∂ϕp ⎪ ⎝ ⎠ A−B
⎪ ϕp = 0(t < 0) , ∂t = 0 (t = 0, Z = 0) (11) ⎪
⎩
⎪ ⇀⎤
i ⎥
⎪
where the Green function employed satisfies the free surface boundary ⎦
⎩
condition, the wave radiation condition at the infinite field boundary
(17)
and the fluid domain bottom condition, and the velocity potential in the
fluid domain can be described by the following integral equation The viscous damping moment in roll motion is expressed by (Molin
(Duan, 1995 and Huang and Duan, 2012): et al., 2002)
⎯⇀ 1
rpq t ∼ ∂ϕp Nc = − ρCd B 4L η˙4 η˙4
⎧ 2πϕp (p , t ) + ∫ σ (q, t )ln r ds = ∫0 dτ∫S (τ ) σ (q, t ) G ∂τ ds 2 (18)
⎪ SH (t ) pq H
⎪ 1 t ∼ where Cd is a non-dimensional coefficient, B is the beam of the floating
⎪ + g ∫0 σGnY V˙ (τ )2 dτ
⎪ body and L is its length.
A−B
⎪ ∂ rpq ∂ϕp
− πσ (p , t ) + ∫S (t ) σ (q, t ) ∂n ln r dsq=−2π ∂n
⎨ H P pq p 2.4. A high-order panel method based on the combination B-spline in the
⎪ t
∼
∂G time-domain
⎪ + ∫0 dτ∫S (τ ) σ (q, τ ) ∂n dsq
H P
⎪
⎪ ∼
1 t ∂G A B-spline higher-order panel method is developed to solve the
⎪ + g ∫0 σ ∂n nY V˙ (τ ) 2 dτ
⎩
p
A−B (12) boundary value problem for a freely floating bod. In this approach, the
⎯⇀
body geometry X (p) and the unknown potential ϕ are defined by the
where B-spline, respectively. At any instantt , the unknownϕ andϕn can be

353
S. Huang et al. Ocean Engineering 164 (2018) 350–366

expressed by: ⇀n (0) ⇀n − 1 Δt ⎛ ⇀ ˙ n (0) ⇀˙ n−1 ⇀˙ n − 2⎞

U =U + ⎜5 U + 8U −U ⎟
n1 n1 12 ⎝ ⎠ (22)
⎧ ϕn (p , t ) = ∑i = 0 ϕni (t ) Ni,3 (u) , X (p , t ) = ∑i = 0 di (t ) Ni,3 (u) p ∈ SF (t )
⎨ ϕ (p , t ) = ∑n2 ϕj (t ) Nj,3 (u), X (p , t ) = ∑n2 dj (t ) Nj,3 (u) p ∈ SB (t ) ⇀n (0) ⇀˙ n (0)
⎩ j=0 j=0 where U and U represent the predicted body velocity and ac-
celeration values, respectively, at the current time step, which are used
(19)
⇀ ⇀ as the initial values of the iteration
where Ni,3 (u) is the third order B-spline basis function, di (t ) and dj (t )
represent the control points on SF (t ) and SB (t ) , and n1 and n2 are the ⇀n (0)
(2) Solve the boundary value problem forϕ andϕp by using U in Eqs.
number of control points. At the intersection point P between the free
(3) and (11) for the boundary condition on the tank and body
surface and the walls, multiple knots and end-point conditions of the B-
surfaces.
spline are used. Thus, the intersections of boundary and continuity
(3) Calculate the fluid velocity and the time derivative of the velocity
conditions are written as:
potential on the tank internal fluid free surface.
∂ϕ n (k ) ⇀n (k − 1)
⎧
∂ϕ (p)
∂n SF
= ϕni , ϕ (p) SB = ϕj (4) Solve the boundary value problem for ∂t ( )
using U and
⎨ϕ = ϕn2 , ϕn1 SF = ϕ0 ⇀˙ n (k − 1)
⎩ 0 SF SB SB (20) U in its boundary condition, where the subscript n (k − 1)
represents the variables at time tn , but at the k th iteration (k = 1,2,3
Details can be found in Datta and Sen (2007) and Huang et al. ⇀n (k )
…). Calculate the total forces F and the acceleration by using:
(2014).
⇀˙ n (k ) ⇀n (k ) ⇀n (k − 1) ⎤
2.5. A numerical scheme for coupled simulation U = [M ]−1 ⎡α nF + (1 − α n ) F
⎢ ⎥ (23)
⎣ ⎦
A ramp-function at the input boundary is applied at the start of the ⇀n (k )
Estimate the new body velocity U using the same method by
simulation to reduce the effect of undesired transient waves, which solving Eq. (22).
increases simulation stability and brings the onset of steady state ear-
lier. The ramp function is: ⇀n (0)
(5) Solve the boundary value problem for ϕ using U in Eqs. (3) and
for t > 2T (11), but now the tank body surface condition is replaced by Eq.
⎧1
r (t ) = (7). Check whether the relative error of the accelerations (or forces)
⎩⎣
t
⎨⎡1 − cos π 2T ⎤/2
⎦ ( ) for t ≤ 2T
(21) is small enough; if not, return to the beginning and continue the
same procedure until the relative error of the accelerations (or
The ramp function is applied in 2T or 4T (wave period) depending
forces) is small enough; otherwise, go to step (6).
on the input wave heightH below.
(6) Update the position of the body using the final body velocity and
An iterative semi-implicit time integration method for floating
acceleration in the above loop by using the third-order Taylor ex-
bodies developed by Yan and Ma (2007) is applied to calculate the
pansion
velocities of the floating body and the time derivative of the velocity
potential∂ϕ/ ∂t . Suppose all calculations before t = tn − 1 have been fin- ⇀˙ n (u)
⇀n + 1 ⇀n ⇀n (u) Δt 2 ⇀˙ n (u) Δt 3 dU
ished and the velocity potential, its time derivative on the free surface, S = S +U Δt + U +
2 6 dt (24)
and the positions of all boundaries including the free surface and the
body surface have been obtained through updating; the only difference ⇀n + 1
where S is the displacement of the body to be used for the calcu-
in this work is the effect of ∂ϕ/ ∂t on the sloshing energy dissipation ⇀n (u) ⇀˙ n (u)
lation of the next time step; U and U represent the final values of
term defined by Eq. (7). The modified method is summarized below.
the body velocity and acceleration in the above loop, respectively. An
Note all calculations before t = tn − 1 are completed and the velocity
under-relaxation coefficient α n is used in Eq. (23) to improve the con-
potential, its time derivative on the free surface, and the positions of all
vergent efficiency obtained from Yan and Ma (2007).
boundaries including the free surface and the body surface have been
obtained through updating.
3. Experimental and time domain simulation parameters
⇀˙ n (0)
(1) Predict the body acceleration U at time tn by curve fitting the The focus of this section is on the coupled nonlinear effects of
accelerations at previous time steps using a least square method and
sloshing and body sway and roll motion. The present nonlinear results
estimating the corresponding body velocity by using the are compared with the linear frequency-domain and experimental re-
Adams–Moulton method (William, 1971) as follows:
sults of Rognebakke and Faltinsen (2003) and Molin et al. (2002, 2008).
The experimental set-ups and dimensions of Model I and Model II are
given in Fig. 2 and Fig. 3.

Fig. 2. The test facility and sway motion of the model of the barge containing tanks, Model I (Rognebakke and Faltinsen, 2003).

354
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 3. The test facility and the motions of the model of the barge containing tanks, Model II (Molin et al., 2002, 2008).

Table 1 IRR3) at a scale of 1:50.

Experiment conditions of Model I (h is the water depth andL is the tank width). Fig. 4 indicates that the present numerical results are not very
Tank NUM None One Two sensitive to the time step size. The time-marching scheme used in the
present study works well for time steps as large as T/30 . From this
Fill ratioh/ L 0 0.25 0.5 0.77 0.5 convergence study, the time step T/40 was used for most simulations.
Empty model mass m (kg) 47.01 43.51 37.01 31.51 26.51

3.1. Calibration of the energy dissipation coefficient

Table 2
Experiment conditions of Model II.
The sloshing energy dissipation coefficient μ scales the effect of
Tank number None Two energy dissipation from sloshing and is a critical facet of the potential
flow sloshing model. Thus, the nonlinear coupled motion is compared
Fill ratio h/ L 0.0 0.2375/ 0.3625/0.3625
0.2375
with a set of experimental data to show the effectiveness of the present
Empty model mass (kg) 285 209 181 method and to calibrate the energy dissipation coefficient. For this
Position of gravity Center (below still 0.1483 0.1397 0.182 purpose, the experiment Model I (Fig. 2) is considered. The sloshing
water level) (m) effect depends on its internal energy dissipation for regular waves.
Moment of inertia (kg•m2 ) 48.84 35.82 31.02
Malenica et al. (2003) suggests that this dissipation is not dependent on
the excitation frequency and only refers to the coupled motion RAO
peak in the frequency domain. However, in some cases this is not
During the experiments, the tank liquid filling ratio varies and the
clearly defined and can introduce significant uncertainty. For different
barge draft is fixed. The location of center of gravity, mass and roll
frequencies, the energy dissipation is also different due to differences in
moments of inertia of the empty model are shown in Table 1 and
sloshing intensity.
Table 2. The barges were moored by a horizontal cabling spring system
To fully reflect the sloshing characteristics, a computation duration
and motion outside of the degrees of freedom of heave, sway, and roll
of 60T is considered. The coupled motion RAO in the time-domain
were not significant in beam waves and therefore well suited for vali-
should be calculated from the steady-state portion of respective time
dation for the theoretical method described in this work. The resulting
histories through FFT due to the long-term transient effects. The sen-
natural period in sway was 10.8 s. Model I and Model II were run with
sitivity to the sway coupled motion RAO from a range of μ from 0.0 to
beam direction regular and irregular waves. The irregular waves used a
1.1 is compared with the modeled tests in Fig. 5.
JONSWAP spectrum with γ = 2 , and a peak period TP of 1.6 s. The
The most notable point is that the sway RAO curves of barges with
specified HS values were 6 cm, 9 cm and 12 cm (sea states IRR1, IRR2,
tanks and without tanks have obvious differences in the intersection

Fig. 4. Comparison of different time-step results for hull sway motion η2 without. coupling effects. ω = 5.65rad/s; ςa = 0.015m (wave amplitude).

355
S. Huang et al.
Fig. 5. Steady-state sway motion.η2
point of the two curves defined as the crossover frequency, ωd . For this
case, the crossover frequency is ωd = 9.0 rad/s. When ω < ωd , sloshing
weakens the barge sway motion, and the optimum value of μ is 0.3,
selected on the basis of the minimum error between the numerical and
experimental results at the sloshing natural frequency. However, when
ω > ωd , the sloshing is exacerbated by the barge sway motion, and the
optimum value of μ is 0.9, selected on the basis of the minimum error at
the frequency according to the RAO curve at the peak. The amplitude of
the sway RAO is only slightly affected by the energy dissipation coef-
ficient μ at ωd . The numerical results are in good agreement with the
experimental results using the present method.
This defines the sloshing energy dissipation coefficient μ in regular
waves. In the following section, this rule is applied in different models
and cases and validated by comparing with experimental results.
3.2. Comparison of experimental and numerical sway and roll motion of
barge with tanks in regular waves
To validate the above rule for the selection of μ and identify the role
of nonlinearity and coupling effects, a numerical comparison to ex-
perimental data is completed. Frequencies of minimum and maximum
sway of Model II were investigated primarily to see how the nonlinear
model could improve on the linear model results.
Overall, the present numerical results are in good agreement with
the experimental and linear frequency domain results. Due to the
coupling effect, the peak of the sway motion appears near the natural
frequency of the sway mode and the sway motion decreases at fre-
quencies less than the first natural frequency of sloshing, ω1. When the
frequencies are higher than but not close to the combinedω1, the sway
motion increases due to the fluid motion in the tanks.
For the case of two tanks with h/ L = 0.5 (Fig. 6a), at ω = ω1, the
sway motion is almost zero. The linear results are similar to the general
trend of the experimental results. However, the experimental observa-
tion is underestimated at the peak compared with the linear results.
This is due to the phase shifts of the sloshing force based on linear
theory. Detailed literature on sloshing phase can be found in Faltinsen
and Timokha (2009).
However, the difference between linear and nonlinear results is
found to be greater near ω1 for the case of one tank with neither low nor
high fill ratio (Fig. 6a and c). In addition, the wave frequencies are
almost equal to the resonance frequencies of the fluid motion in the
tanks. For example, when ω = 8.48rad/s, the sway motion is almost
zero for the case of one tank with h/ L = 0.5 (Fig. 6a). This frequency is
known as the cancellation frequencyωσ .
A time series for sway motion of the case of one tank with fill ratio
356
Ocean Engineering 164 (2018) 350–366
h/ L = 0.5 is shown in Fig. 7. When the incident wave front is at the
barge position, a significant sway motion is initiated. This in turn ex-
cites sloshing in the tanks, and thus a sloshing force starts to counteract
the excitation force from the waves. The sway motion decreases until an
equilibrium is reached. At this stage, the sway-induced sloshing force is
almost equal and opposite to the wave excitation force as shown in
Fig. 8. However, ω = 8.48rad/s is not the first-order natural frequency,
which is ω1 =
π⋅g
L
⋅tanh (π⋅ ) based on linear theory. This is because
h
L
ω1 decreases in a nonlinear fashion as the sloshing amplitude increases.
In other words, a very small sway motion may be present with a violent
sloshing response in the nonlinear model, which does not occur atω1.
At ω = 8.65 and ω = 8.7 rad/s, the wave frequency is slightly larger,
and the sway motion amplitude of steady state is much higher (Fig. 7).
The sloshing response is very sensitive to the change in sway frequency.
Fig. 9 illustrates how the sloshing-induced force is larger and also op-
posing the wave excitation force. The complex shape of the sloshing-
induced force time series indicates a strong nonlinearity that is also
affecting the shape of the wave excitation force. This implies that, al-
though the wave slope is small, the nonlinear sloshing flow can result in
a nonlinear wave excitation force if the present body nonlinear model is
used. However, the steady-state motion shows almost no higher-order
harmonics. This indicates that the higher-order component of the
sloshing force and the external wave loads are filtered out by the
system. Rognebakke and Faltinsen (2003) experienced a similar trend
in their experiments.
The linear theory does not accurately predict the frequencies of
minimum and maximum sway motion, because the assumption that a
constant natural frequency for the fluid motion in the tank is not sui-
table due to the effect of large amplitude sloshing. When the fill ratio
h/ L is higher than the critical value of 0.3374, ω1 decreases in the
nonlinear model as the free surface elevation amplitude of the internal
fluid increases. The effect of the fill ratio has been further discussed by
Faltinsen and Timokha (2001). This is also reinforced in the case of two
tanks withh/ L = 0.5, shown in Fig. 8a, where the linear and nonlinear
models give acceptable results near ω1 because the amplitude of the
sloshing motion is relatively small and does not changeω1.
The nonlinear simulation scheme presented shows satisfactory
agreement when compared with experimental results. In addition, the
nonlinear and experimental results do not converge monotonically to
the linear results when ω ≈ ω1 or ω > ω1. This indicates there are large
errors in the linear sloshing model in predicting sloshing and floating
body coupled motion.
The nonlinear effects in roll coupled motion is also assessed by
numerical comparison to Model II experiment results. The experimen-
tally measured roll damping coefficient Cd = 0.2 (Molin et al., 2008) was
added to the numerical model.
The comparison of roll RAOs between the present nonlinear time-
domain, the linear frequency-domain results and the experimental re-
sults for regular waves are shown in Fig. 10. The positions of the first-
order natural frequency of sloshing ω1 and the cancellation frequency
ωσ are indicated. The trends of the present nonlinear frequency-domain
and the experimental test results are consistent, with small differences.
At the second peak, the nonlinear solution is larger than the linear
frequency domain solution. The peak position of the nonlinear solution
moves to the low frequency field and the trend more closely approx-
imates the test results. In a similar manner to the results in comparison
of sway motion, the difference between the nonlinear and linear solu-
tions occurs in the vicinity of ω1.
ωσ calculated by linear theory is equal to ω1, but ωσ calculated by
nonlinear theory changes to a high frequency range. The response
amplitude calculated by nonlinear theory is larger than the linear so-
lution near areas with a frequency less than the natural frequency of
sloshing such as ω = 4.5rad/s and ω = 5.0 rad/s. The second peak po-
sition of roll response is likely to move to a lower frequency
(ω = 6.0 − 6.25rad/s).
S. Huang et al.
Fig. 6. Comparisons between experiments and calculations of steady-state sway motion η2 ; ςa = 0.015m.
Fig. 7. Time series of sway motion η2 ; one tank; h/ L = 0.5; ζa=0.015 m.
To analyze the reasons for differences between the linear and non-
linear solutions, the comparison between the time history curve of
sloshing-induced and wave-induced moment under different excitation
357
Ocean Engineering 164 (2018) 350–366
frequency incident waves is given in Fig. 11. In a barge roll resonance
interval such as ω = 4.5rad/s, the phase difference between the sloshing
and the wave moment is approximately 180°, and the amplitude is al-
most equal. It is important to reduce barge roll motion amplitudes after
superposition.
At the cancellation frequency ωσ = 5.5rad/s, which is far from the
barge roll resonance frequency, the wave-induced moment is very small
and the sloshing-induced moment is more than an order of magnitude
larger. The total moment is almost equal to the sloshing-induced mo-
ment after superposition. Due to the increase in natural frequency of the
sloshing, the sloshing-induced moment is increased, and the total mo-
ment increases. Thus, the roll motion amplitude calculated by the
nonlinear model is larger than the amplitude calculated by the linear
theory. In the frequency band of the second peak of the roll RAO curve
(ω = 6.0 − 6.25rad/s), the sloshing-induced moment is increases when
accounting for nonlinearities, and further nonlinearity is introduced in
wave excitation by accounting for the change in body surface in the
large-amplitude barge motion model.
The second peak in the roll RAO is caused by liquid sloshing. The
linear model does not account for nonlinear wave excitation nor for
nonlinear sloshing loads which are two significant differences that ac-
count for the improved comparison to experimental data, especially at
higher frequencies. Since the second peak in the RAO is caused by
sloshing resonance, improved accuracy in the sloshing loads in the
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 8. Time series of sway force Fy acting on one tank of the barge. Forces are normalized by ρgdςa (d is the draft of the barge). ω = 8.48Rad/s; ζa=0.015 m.

Fig. 9. The same as Fig. 8, but ω = 8.65 rad/s.

nonlinear model explains the improved accuracy when compared to
experimental results.
3.3. Investigation of the effect of tank fill ratio on barge motion
The steady responses of the barge roll coupled motion of Model II at
different fill rates are shown in Fig. 12. Even under the action of the
small amplitude wave excitation, nonlinear sloshing can introduce large
changes in roll response when compared to the linear model as seen in
Fig. 12a. Because the second peak of the roll motion RAO is caused by
the liquid sloshing, and the wave-induced moment is small, the non-
linear effect mainly comes from the liquid sloshing. Thus, the difference
between the linear and the nonlinear solutions is large near the sloshing
natural frequency and the second peak. With increasing liquid fill ratio,
this difference is gradually reduced. When the tank is full of water, the
model degenerates into that of a single barge, with effectively no dif-
ference between them.
Near the first-order natural frequency of sloshing ω1, the steady state
amplitude predicted by nonlinear theory is larger than the amplitude
predicted by linear theory; the former is closer to the experimental
value as shown in Fig. 12b–c. For example, the fill ratioh/ L = 0.3625 is
greater than the critical value h/ L = 0.3347 studied by Rognebakke and
Faltinsen (2003) in detail. The sloshing natural frequency is shifted to a
lower frequency. Therefore, at or below ω1, the increase in sloshing-
induced moment leads to the increase in roll motion amplitude.
Based on the analysis of sloshing-induced and wave-induced mo-
ments at ω = 4.9rad/s, the roll motion amplitude calculated by linear
theory is zero and ω = ω1 = 5.598 rad/s as shown in Fig. 13. With de-
creasing incident wave frequency, which is closer to the barge roll re-
sonance frequency, the wave-induced moment increases. However, due
to ω1 shifting to a lower frequency, the liquid sloshing moment does not
decrease significantly as it moves away fromω1. Due to these factors, the
linear and nonlinear solutions are significantly different.
3.4. The effect of wave height on barge motion
In the linear theory assumption, the coupled motion RAO does not
change with the variation in wave height. However, it is predicted that
nonlinear theory is very sensitive to changes in wave height. Fig. 14
illustrates the comparison of the nonlinear roll motion RAO of Model II

358
S. Huang et al.
Fig. 10. Comparison of the barge model roll RAOs; h/ L = 0.2375; H = 0.03m.
at different incident wave heights; experimental results are for sea state
IRR1 and IRR3.
The two peaks of the roll coupled motion RAO curve are sig-
nificantly reduced with increasing wave height; most notably the
second peak changes to a low frequency field when h/ L = 0.3625.
Another significant change is the anti-rolling frequency band between
the two peaks increased for a low filling condition h/ L = 0.2375. The
anti-rolling effect of sloshing on a floating body is enhanced under large
amplitude wave excitation, which is does not occur with the linear
theory assumption. One reason for this is the wave excitation moment
effect on the floating body roll motion increases and the growth of the
sloshing-induced moment is slower than the growth of the roll motion
under large amplitude wave excitation. In addition, another reason for
this is the growth of wave height does not scale linearly with the growth
of the wave-induced moment. As shown in Fig. 14a, at ω = 4.0 rad/s,
with increasing wave height, the roll coupled motion amplitude is
significantly reduced. By analyzing the sloshing-induced and the wave-
induced moments, the relationship between the increase in the
sloshing-induced and the wave-induced moments and the wave height
growth is not linear. Additionally, there is phase difference either in the
sloshing-induced moments or in the wave-induced moments at different
wave heights (see Fig. 15). The nonlinear characteristics of both the
wave-induced and the sloshing-induced moments are more obvious as
their curves become less sinusoidal. Thus, the total moment amplitude
varies greatly, and produce larger roll response in smaller waves, and
smaller roll response in larger waves. This leads to an increase in the
wave height, and the two peaks of the roll coupled motion RAO curve
are significantly reduced and the liquid sloshing anti-roll effect is in-
creased.
3.5. An analysis of different nonlinear factors on coupled motions
The previous sections present analysis of the nonlinear effects of the
floating body surface and sloshing on the coupled motion. To in-
vestigate what type of nonlinear effect is most important in coupled
motion, two partly reduced numerical models, N1 and N2, are used and
compared with Model I experimental data. Further comparison is done
with the nonlinear time-domain model. In the numerical model N1, the
hydrodynamic calculation of the floating body flow field is based on the
linear theory, and the sloshing in the liquid tank is based on the fully
nonlinear theory. In the numerical model N2, uses a nonlinear body
flow field with a linear sloshing model.
As shown in Fig. 16, in the case of a small amplitude wave,
H = 0.03m, the RAO calculated by the nonlinear time-domain model
359
Ocean Engineering 164 (2018) 350–366
and N1 are close to each other with only with a slight difference near
the peak of the RAO curve (ω=9.0 rad/s). However, the difference is
more obvious when the incident wave height increases. This is because
the effects of external wave force on the coupled motion are strength-
ened. For example, for ω=9.0 rad/s and H = 0.07m, the wave force
amplitude calculated by N1 is only the 50% of that calculated by the
nonlinear time-domain model. The second-order component of the
wave force is very small, which has degenerated into a linear curve
relative to significant high-order nonlinear characteristics of the results
of the nonlinear time-domain model shown in Fig. 17.
By analyzing the second-order wave force shown in Fig. 18, the
overall hydrodynamic force acting on the floating body increases, ap-
parently because the second-order component of the wave force is
greater near ω=9.0 rad/s. Such nonlinear effects result in the increase of
the coupled motion amplitude. The coupled motion also shows dis-
tinctly nonlinear characteristics due to the influence of nonlinear
sloshing.
As shown in Fig. 19, in the case of a large amplitude wave
H = 0.07 m, a difference appears between the RAO calculated by the
nonlinear time-domain model and N2, when the incident wave fre-
quency ω is greater than the cancelation frequency ωσ . The coupled
motion response calculated by N2 is similar to that calculated by the
linear frequency-domain model. The ωσ calculated by N2 is the same as
the sloshing natural frequency ω1, because ω1 is based on the linear
assumption in the linear sloshing model and does not increase with an
increase in sloshing amplitude. Another big difference is the position of
the RAO curve's peak, which when calculated by the nonlinear time-
domain model, the response amplitude is too large when ω is greater
than ω1. Through the analysis of hydrodynamic forces shown in Fig. 17,
the wave force amplitudes calculated by the two numerical models are
similar, and the higher-order nonlinear characteristics are all notice-
able. However, the difference in the sloshing force calculation is large,
and the linear sloshing force amplitude calculated by N2 is almost twice
that of the nonlinear time-domain model and a phase difference exists
between them.
From the above comparison, the sloshing nonlinear effect has the
greatest influence on coupled motion. For small amplitude waves, the
floating body surface nonlinear effect has less of an effect on the cou-
pled motion amplitude than the floating body hydrodynamics force. For
large amplitude waves, the floating body surface nonlinear effect is
particularly important. It is significant that the floating body surface
nonlinear effect has a great influence on the steady state response
amplitude of the coupled motion. The first-order and second-order
components of the floating body external wave force are different, and
the second-order component is relatively close to the first-order com-
ponent near ω1.
3.6. A comparison of the motions for a barge with tanks in irregular waves
The analysis performed in the previous section showed that, for a
regular wave, the number of motion peaks of the floating body is in-
creased due to sloshing effects. In the vicinity of the sloshing natural
frequency, it has an obvious anti-rolling effect on the floating body's
motion. To investigate the response in more representative conditions
of the ocean, a comparison of response in irregular waves between the
numerical and experimental Model II was completed. The roll motion
RAOs obtained by cross-spectrum analysis of the nonlinear time history
curves are compared with the frequency-domain and the irregular wave
experimental results (Fig. 20 and Fig. 21). The roll damping coefficient,
Cd and the sloshing energy dissipation coefficient, μ are applied by 0.0,
0.2, and 0.0, 0.3, 0.5, respectively, in the calculation. The optimal va-
lues are obtained by comparison with the experimental results.
As discussed earlier, for an irregular wave, the anti-rolling effect of
sloshing is noticeable near its first-order natural frequencyω1. However,
in some frequency ranges, the barge motions are increased. Although
the increase of both μ and Cd are effective for anti-rolling, these effects
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 11. Comparison of the horizontal force and roll moment components; h/ L = 0.2375; H = 0.03m.

360
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 12. The comparison of steady state roll RAO with different fill ratios; H = 0.03m.

Fig. 13. The comparison of roll moment components; h/ L = 0.3625; H = 0.03m.

361
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 14. Steady state roll RAO with different incident wave heights.

are different due to differences in the mechanisms. Since the first peak
of the roll motion RAO is caused by the barge's inertia, a change in Cd
has a significant effect. However, since the second peak is caused by
liquid sloshing, a change in Cd has almost no effect on the peak, as
shown in Fig. 20. This shows that it is not enough to modify roll motion
by simply adding roll damping to counteract sloshing and ship coupled
motions. The sloshing energy dissipation effect is very important and
must be considered in ship motion equations.
As can be observed from Fig. 21, μ = 0.3, and Cd = 0.2 values are
more suitable for this barge's roll coupled motion in these conditions.
The numerical results agree well with the experimental results, proving
the reliability of the prediction of liquid violent sloshing and floating
body nonlinear coupled motion in irregular waves. When no sloshing
energy dissipation is considered ( μ = 0.0 ), the second peak amplitude
of the roll motion RAO is far greater than the experimental value. In-
stead, when sloshing energy dissipation is considered, whether μ is lar-
ge(μ = 0.5) or small ( μ = 0.3), the second peak amplitude of roll motion
RAO is significantly reduced. This shows that μ also plays an important
factor in the anti-rolling effect. For an irregular wave, the value of μ is Fig. 16. A comparison of the steady state sway RAO using the nonlinear time-
single and relatively stable, according to a relationship between the ωd domain model and N1.
and ω in regular waves. It should be noted the fill ratio has an important
influence on the time domain simulation as it drives the magnitude of the experimental results and from the present nonlinear results in the
the sloshing effect. second peak of the roll motion RAO curve. The amplitude of the barge
For a regular wave, the frequency-domain results are different from roll motion calculated by the frequency-domain method is smaller than

Fig. 15. A comparison of roll moment with different wave elevations; h/ L = 0.2375; ω = 4.0 rad/s.

362
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 17. A comparison of the hydrodynamic force using different nonlinear models; ω = 9.0 rad/s.

Fig. 18. A comparison of hydrodynamic force components; H = 0.07 m.

Fig. 19. A comparison of sway RAO using the nonlinear time-domain model Fig. 20. Roll RAO of a barge with tanks with different Cd ; h/ L = 0.2375; IRR1.
and N2.

363
S. Huang et al.
Fig. 21. Roll RAO of a barge with tanks with different μ ; IRR1.
two others. This difference occurs in the two states of liquid fill ratio.
One reason is that there are limitations of the frequency-domain
method for viscous corrections of roll motion and sloshing. Another
reason is that the excitation force components of non-monochromatic
incident waves are superimposed on the floating body motion due to
multiple frequency oscillation of the liquid sloshing in tanks, which
leads to the floating body motion that does not satisfy the linear single
frequency assumption. Thus, there is an error introduced in certain
conditions in sloshing and ship coupled motion using the linear fre-
quency-domain theory.
Under the action of the high sea states, a floating body motion can
lead to a large change in the wet surfaces of the body and tank. The
nonlinearity caused by the change in transient wet surface conditions
relative to the average wet surface conditions cannot be ignored. When
the tank is subjected to large amplitude motion, nonlinear sloshing is
apparently enhanced, and the frequency band of resonance is wider,
and the sloshing-induced moment is larger around the natural fre-
quency of sloshing. Therefore, for these conditions, strong nonlinear
sloshing and floating body nonlinear coupled motion cannot be ig-
nored.
The violent sloshing and floating body nonlinear roll coupled mo-
tion RAOs in different sea states are shown in Fig. 22. The two peaks of
Fig. 22. A comparison of roll RAO of a barge with tanks in different sea states.
364
Ocean Engineering 164 (2018) 350–366
roll coupled motion RAO decreased as the sea state increased and the
second peak became smoother. When h/ L = 0.3625, the second peak
moved to a low frequency. Another significant change is that, at the low
fill ratio h/ L = 0.2375, the anti-rolling frequency band between the two
peaks increased. This suggests that, for the bad sea state, the anti-roll
effect of sloshing to roll motion is enhanced, which is similar to the role
of roll RAO variation for regular waves.
Based on a comparison of each component of the floating body roll
moment, there is a phase difference between the sloshing-induced and
the wave-induced moments, for both low or high sea states, as shown in
Fig. 23. In the high sea state, both nonlinear characteristics of the
moments are more obvious. Based on the study of the roll moment RAO
as shown in Fig. 24, in the high sea state, the growth of wave and
sloshing excitation moment does not scale linearly with the growth of
the waves and are slow to adapt to the growth of roll motion. This leads
to a significant reduction in the two peaks of the roll coupled motion
RAO for the high sea state. The shape of the roll motion RAO curve at
the second peak becomes gentle, mostly due to faster wave excitation
moment growth and slower sloshing excitation moment at high fre-
quencies(ω ≥ 6.5rad/s). Thus, the superimposed total moment and roll
motion are both increased.
S. Huang et al. Ocean Engineering 164 (2018) 350–366

Fig. 23. A comparison of roll moment components.h/ L = 0.2375

4. Conclusions amplitude regular and irregular wave excitation. The effects of the
nonlinearities introduced in sloshing and wave excited are investigated
A time-domain method based on nonlinear potential flow theory is through sensitivity studies and comparison with experimental data.
established to predict sloshing and ship coupled motion for large- Nonlinear sloshing affects the coupled sway and roll motion, followed

Fig. 24. A comparison of roll moment RAO components.h/ L = 0.2375

365
S. Huang et al.
by nonlinear changes to the floating body wetted surface. A modified
wall boundary condition that is driven by an energy dissipation coef-
ficient is used. Based on the investigations completed, the following
conclusions are made:
• The present model successfully approximates sloshing energy dis-
sipation and thus achieves the steady state of sloshing motion in a
computationally efficient simulation and is a significant improve-
ment in accuracy compared to linear models. The value of energy
dissipation coefficient μ in regular waves, if ω < ωd , is 0.3, and if
ω > ωd , it is 0.9. However, for irregular wave excitations, a value of
μ = 0.3 is relatively stable. It is found that μ is independent of the
wetted tank surface, fill ratio and model scale and is an important
aspect in the time-domain simulation. Although the specific value
may need to be modified for other tanks, such as those with baffles,
this rule may be considered as a good reference when no other data
is available.
• In comparison to the results of the linear theory, there are some
factors that need attention. First, for small amplitude waves, the
corresponding frequency at peak and near zero of floating body
sway motion RAO showed a shift phenomenon under some condi-
tions. This phenomenon is associated with the changes in the
sloshing natural frequency, at which large amplitude sloshing oc-
curs. For large amplitude waves, the floating body roll motion RAO
curves decrease near their second peaks, induced by the sloshing
peaks' positions moving outwards. The role of sloshing in reducing
the roll is particularly evident because the anti-roll frequency bands
are wider. This is the result of involvement between nonlinear ef-
fects of the floating body outside wave loads and the tank inside
liquid sloshing loads. Near the sloshing natural frequency, the first-
order and second-order components of the outside wave forces are
close because body motions are small while sloshing effects are
large. The offset of the coupled motion balances the position and the
amplitude of the steady state coupled motion response.
• To explore the impact of different nonlinear effect has in coupled
motion, two partly reduced numerical models, N1 and N2, were
applied to a calculation of the same barge with tanks for comparison
with experimental and the nonlinear time-domain model results.
Under small amplitude waves, the nonlinear changes of the floating
body surface has little influence on the steady-state amplitude of the
coupled motion. However, it has more a distinct effect on the hy-
drodynamic forces and equilibrium position of the floating body. For
large amplitude waves or high sea states, the nonlinear changes of a
floating body surface are particularly important.
By using the present nonlinear model, the ship-sloshing coupled
system by the potential theory can be reasonably represented in a wide
range of conditions. This is advantageous because it is much more
computationally efficient than CFD models based on viscous fluid
equations, such as VOF or SPH methods. As the research of foundation
theories, the investigations are limited to two-dimensional problems.
However, the sloshing-induced total force is the key factor, and the
influence of the sloshing phenomena on the ship roll and sway motion is
most significant in the coupled analysis. The main nonlinear effects are
reflected. Thus, further research into the three-dimensional problem
can be developed under the guidance of the present conclusion.
Acknowledgements
This work is supported by the National Natural Science Foundation
of China (Grant No. 51609232 and Grant No. 50779008), the Strategic
Priority Research Program of the Chinese Academy of Sciences (Grant
No. XDA13040203), the Floating support platform project, High-tech
ship research projects sponsored by MIT (No. 201622) and Marine
Renewable Energy Funds Projects (GHME2017SF01 and
GHME2017YY02), to which the authors are most grateful.
366
Ocean Engineering 164 (2018) 350–366
References
Duan, W.Y., 1995. Nonlinear Hydrodynamic Forces Acting on a Ship Undergoing Large
Amplitude Motions. Ph.D. Thesis. Harbin engineering university (in Chinese).
Datta, R., Sen, D., 2007. The simulation of ship motions using a B-Spline-Based panel
method in time domain. J. Ship Res. 51 (3), 267–284.
Faltinsen, O.M., Timokha, A.N., 2001. Adaptive multimodal approach to nonlinear
sloshing in a rectangular tank. J. Fluid Mech. 432, 167–200.
Faltinsen, O.M., Timokha, A.N., 2009. Sloshing. Cambridge University Press.
Faltinsen, O.M., Rognebakkeb, O.F., 2005. Timothy, A.N. Classification of three-dimen-
sional nonlinear sloshing in a square-base tank with finite depth. J. Fluid Struct. 20,
81 103P.
Hamlin, N.A., Lou, Y.K., Maclean, W.M., Seibold, F., Chandras, L.M., 1986. Liquid
sloshing in slack ship tanks-theory, observations, and experiments. Trans. SNAME 94,
159–195.
Hong, L., Zhu, R.C., Miao, G.P., Fan, J., Li, Y.L., 2012. Numerical calculation of ship
motions coupled with tank sloshing in time domain based on potential flow theory. J.
Harbin Eng. Univ. 33 (5), 750–758.
Huang, S., Duan, W.Y., Ma, Q.W., 2011. An approximation to energy dissipation in time
domain simulation of sloshing waves based on linear potential theory. China Ocean
Eng. 25 (2), 189–200.
Huang, S., Duan, W.Y., 2012. Nonlinear coupling motion numerical simulation of freely
floating body and sloshing. In: Proceedings of the 22th International Offshore and
Polar Engineering Conference Rhodes, Greece, pp. 341–347.
Huang, S., Duan, W.Y., You, Y.G., Jiang, J.H., 2014. Wang Wensheng. Nonlinear time
domain simulation of sloshing and coupled ship motion. J. Harbin Eng. Univ. 35 (9),
1045–1052 (in Chinese).
Wang, J., Sun, S.L., Hu, J., 2017. The coupling analysis of tank motion and sloshing by a
fully nonlinear decoupling method. Nonlinear Dynam. 89, 971–985.
Jiang, S.C., 2014. Numerical Simulation on Coupled Effect between Ship Motion and
Liquid Sloshing under Wave Action. Ph.D. Thesis. Dalian university of technology.
Kim, Y., 2002. A numerical study on sloshing flows coupled with ship motion—the anti-
rolling tank problem. J. Ship Res. 46 (1), 52–62.
Kim, Y., Shin, Y.S., Lin, W.M., Yue, D.K.P., 2003. Study on sloshing problem coupled with
ship motion in waves. In: 8th International Conference on Numerical Ship
Hydrodynamics, Busan, Korea, September.
Kim, Y., Nam, B.W., Kim, D.W., Kim, Y.S., 2007. Study on coupling effects of ship motion
and sloshing. Ocean Eng. 34, 2176–2187.
Kim, B.i, Shin, Y., 2008. Coupled seakeeping with liquid sloshing in ship tanks. In:
Proceedings of the ASME 27th International Conference on Offshore Mechanics and
Arctic Engineering. Estoril, Portugal, June 15–20.
Koo, W., Kim, M., 2004. Freely floating body simulation by a 2D fully nonlinear numerical
wave tank. Ocean Eng. 9, 2011–2046.
Lee, S.J., Kim, M.H., Lee, D.H., Kim, J.W., Kim, Y.H., 2007. The effects of LNG-tank
sloshing on the global motions of LNG carriers. Ocean Eng. 34, 10–20.
Li, Y.L., Zhu, R.C., Miao, G.P., Fan, J., Lu, Z.M., 2012. Simulation of ship motions coupled
with tank sloshing in time domain based on OpenFOAM. J. Ship Mech. 16 (8),
635–647.
Molin, B., Remy, F., Ledoux, A., Ruiz, N., 2008. Effect of roof impacts on coupling be-
tween wave response and sloshing in tanks of LNG-carriers. In: Proceedings of the
ASME 27th International Conference on Offshore Mechanics and Arctic Engineering,
Estoril, Portugal, June 15–20.
Molin, B., Remy, F., Rigaud, S., de Jouette, Ch, 2002. LNG-FPSO’s: frequency domain
coupled analysis of support and liquid cargo motions. In: Proceeding of INAM
Conference, Rethymnon, Greece.
Malenica, Š., Zalar, M., Chen, X.B., 2003. Dynamic coupling of seakeeping and sloshing.
International Offshore and Polar Engineering, ISOPE 3, 486–492.
Nam, B.W., Kim, Y., Kim, D.W., Kim, Y.S., 2009. Experimental and numerical studies on
ship motion responses coupled with sloshing in waves. J. Ship Res. 53 (2), 68–82.
Ning, D.Z., Song, W.H., Liu, Y.L., Teng, B., 2012. A boundary element investigation of
liquid sloshing in coupled horizontal and vertical excitation. J. Appl. Math. http://dx.
doi.org/10.1155/2012/340640. 20 pages.
Rognebakke, O.F., Faltinsen, O.M., 2003. Coupling of sloshing and ship motions. J. Ship
Res. 47 (3), 208–221.
Sen, D., 1993. Numerical simulation of motions of two-dimensional floating bodies. J.
Ship Res. 37 (4), 307–330.
Su, Y., Liu, Z.Y., 2017. Coupling effects of barge motion and sloshing. Ocean Eng. 140,
352–360.
Su, T.C., Lou, J.K., Flipse, J.E., Bridges, T.J., 1982. In: A Numerical Analysis of Large
Amplitude Liquid Sloshing in Baffled Containers, Report MA-RD-940–82046 of the
Texas a and M University College Station.
William, G.C., 1971. Numerical Initial Value Problems in Ordinary Differential Equations.
Prentice-Hall, Englewood Cliffs, NJ, pp. 102–115.
Wu, G.X., Eatock, T.R., 1994. Finite element analysis of two dimensional non-linear
transient water waves. Appl. Ocean Res. 16, 363–372.
Wu, G.X., Ma, Q.W., Eatock, T.R., 1998. Numerical simulation of sloshing waves in a 3D
tank based on a finite element method. Appl. Ocean Res. 20, 337–355.
Wu, G.X., 1998. Hydrodynamic force on a rigid body during impact with liquid. J. Fluid
Struct. 12, 549–559.
Xue, M.A., Zheng, J.H., Lin, P., Xiao, Z., 2017. Violent transient sloshing-wave interaction
with a baffle in a three-dimensional numerical tank. J. Ocean Univ. China 16 (4),
661–673.
Yan, S., Ma, Q.W., 2007. Numerical simulation of fully nonlinear interaction between
steep waves and 2D floating bodies using the QALE-FEM method. J. Comput. Phys.
221, 666–692.
Zhao, W.H., Yang, J.M., Hu, Z.Q., Xiao, L.F., Tao, L.B., 2014. Hydrodynamics of a 2D
vessel including internal sloshing flows. Ocean Eng. 84, 45–53.
Zhang, C.W., 2016. Numerical Study of Nonlinear Sloshing and its Coupling with Vessel
Motions. PhD thesis. University College London.