Ocean Engineering 187 (2019) 106183

Contents lists available at ScienceDirect

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

Liquid sloshing in a two-dimensional rectangular tank: A numerical

investigation with a T-shaped baffle
�ur Oral Ünal *, Gürbüz Bilici, Hakan Akyıldız
Ug
Istanbul Technical University, Istanbul,
_ Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, the liquid sloshing in a closed, partially filled, T-shaped baffled and unbaffled two-dimensional
Sloshing rectangular tank was numerically investigated. The tank was rotated with two different rotation angles of 4�
Baffle and 8� , and a fixed angular velocity of 3.3 rad/s, which was determined by taking the natural frequency of the
CFD
tank into consideration, for the filling depths of 50% and 75%. The baffle heights as well as the rotation angles
Rectangular tank
and filling depths were systematically varied and the effect of these parameters on the hydrodynamic loads on
the tank wall and free surface elevations was examined. The calculations were carried out in two different ways,
by means of laminar and turbulent viscous flow solvers, which use the finite differences and finite volume
discretization techniques in conjunction with the “volume of fluid” technique, respectively. The general flow
topology, free surface deformations, and vorticity distributions are provided and discussed in detail. It was found
that the baffle was fully effective in pressure and wave damping when its height was greater than 80% of the
liquid level.

1. Introduction roofs (Hasheminej and Mohammadi, 2016). Alternatively, it has also

been devised proper control methods for effective slosh suppression. The
Liquid sloshing, in partially filled containers under external excita­ flexible floating panel is a particularly viable option that requires no
tions, has been a crucial engineering issue, which involves the perfor­ support from the tank side walls and is functional for an arbitrary di­
mance, stability, and structural integrity problems in many disciplines rection of external excitation. Application of the active control methods
such as aerospace vehicles, road tankers, liquefied natural gas carriers, may be an effective solution for addressing this challenging problem
elevated water towers and petroleum cylindrical tanks, etc. (Hasheminej (Zang et al., 2015; Pridgen et al., 2012).
and Mohammadi, 2016; Ibrahim, 2005). Hydrodynamic forces acting on Fluid motion in partially filled tanks can cause large structural loads
tank walls as a result of the liquid sloshing may damage the container, and unexpected instability of engineering structural system if the period
thus the sloshing dynamic loads should be restricted in order to avoid of tank motion is close to the natural period of fluid inside the tank.
structural failure because of undesirable dynamic behaviours. The Furthermore, the caused failure may be a tremendous loss of human,
inherent liquid viscosity is not sufficient to reduce the sloshing forces on economic, and environmental resources. The amplitude of the slosh, in
the dynamic characteristics of liquid storage tanks. Therefore, other general, depends on the amplitude and frequency of the tank motion,
methods should be introduced to suppress the sloshing dynamic loads. liquid-fill depth, liquid properties and tank geometry. These parameters
Among them, in order to maintain safe operations, baffles have been have direct effects on the dynamic stability and performance of moving
devised as effective internal components to increase the hydrodynamic tanks. Vertical baffles and perforated plates are useful for suppressing
damping ratio and consequently decrease the slosh forces in most of the resonant sloshing by changing the lowest resonant frequency to a higher
practical engineering problems. Many researchers have proposed, frequency range.
simulated, and tested a wide variety of passive solutions to dissipate There has been a considerable amount of work on investigating the
slosh energy and/or limit the liquid movement by using various physical effects of baffles on liquid sloshing by using analytical, experimental and
devices such as partitions (Mitra et al., 2010), baffles (Wang et al., numerical methods. But, still, further research is of great importance for
2016), tuned, flexible absorbers (Gradinscak et al., 2011), and floating the understanding of the complex sloshing dynamics and the techniques

* Corresponding author. Istanbul Technical University, Faculty of Naval Architecture and Ocean Engineering, Ayaza�
ga Campus, Maslak, Istanbul, 34469, Turkey.
E-mail address: [email protected] (U.O. Ünal).

https://doi.org/10.1016/j.oceaneng.2019.106183
Received 12 November 2018; Received in revised form 31 May 2019; Accepted 3 July 2019
Available online 23 July 2019
0029-8018/© 2019 Elsevier Ltd. All rights reserved.
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 1. The tank configuration used in the computational study.

Table 1
Computational cases investigated.
Case Rotation Angle ( ) Liquid Depth (%) Baffle Height Ratio (hB/h) Case Rotation Angle ( ) Liquid Depth (%) Baffle Height Ratio (hB/h)
� �

A4-D50-U 4 50 Unbaffled A4-D75-U 4 75 Unbaffled

A4-D50-B5 4 50 0.5 A4-D75-B5 4 75 0.5
A4-D50-B7 4 50 0.75 A4-D75-B7 4 75 0.75
A4-D50-B8 4 50 0.8 A4-D75-B8 4 75 0.8
A4-D50-B9 4 50 0.9 A4-D75-B9 4 75 0.9
A4-D50-B1 4 50 1.0 A4-D75-B1 4 75 1.0
A8-D50-U 8 50 Unbaffled A8-D75-U 8 75 Unbaffled
A8-D50-B5 8 50 0.5 A8-D75-B5 8 75 0.5
A8-D50-B7 8 50 0.75 A8-D75-B7 8 75 0.75
A8-D50-B8 8 50 0.8 A8-D75-B8 8 75 0.8
A8-D50-B9 8 50 0.9 A8-D75-B9 8 75 0.9
A8-D50-B1 8 50 1.0 A8-D75-B1 8 75 1.0

Fig. 3. A view of the mesh structure (nominally 2 mm) around the T-sha­
ped baffle.

powerful conformal mapping technique, etc. Developing an analytical

Fig. 2. Pressure signals at T1 obtained from different mesh resolutions. for the
model with acceptable accuracy for the determination of baffle damping
unbaffled A8-D75-U case.
could provide useful means for the design of the baffle geometric
characteristics and arrangements. On the other hand, analytical
for sloshing damping using baffles. The analytical and semi-analytical methods have not been useful anymore with the complexity of the tank
mathematical models have been used to study the liquid sloshing geometry.
characteristics in a baffled or unbaffled half-full horizontal cylindrical Experimental investigations are necessary for evaluating the actual
container of elliptical or circular cross section, rectangular containers, dynamic characteristics of the liquid sloshing with different baffles. On
vertical circular cylindrical tanks subjected to arbitrary external forces. the other hand, there are not enough information on the experimental
Some of these studies, e.g. Gavrilyuk et al. (2006), Maleki and Ziyaeifar investigation of the suppression of sloshing behaviour using baffles in
(2008), Hasheminejad et al. (2014), and Wang et al. (2013), are carried published literature and some problems in the extension of their results
out to approximate the linearized problem on fluid sloshing by using to the full-scale real-world problems due to scaling effects. In recent
some techniques such as the appropriate eigen-function expansions, a years, Goudarzi and Sabbagh-Yazdi (2012) carried out the experimental
weakly nonlinear modal theory, an asymptotic modal method, a measurements to evaluate the efficiency of three types of baffles (upper

2
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Akyıldız (2012) and Jung et al. (2012) numerically examined the effect
of the vertical baffle height relative to the initial liquid depth. The
critical baffle height to reach the roof of the tank have been investigated.
On the other hand, Goudarzi et al. (2012) indicated that an up-mounted
vertical baffle is more effective than a low-mounted one and horizontal
plates have significant damping effects in slender tanks, whereas vertical
plates are more effective in broader tanks. Vertical baffles may reduce
the sloshing amplitudes and dynamic impact loads as well as the natural
frequency of the tank (Wu et al., 2012; Xue et al., 2012). Additionally,
alternative baffle systems have been analysed in tanks, such as annular
baffles and flexible baffles in cylindrical tanks (Biswal et al., 2006),
horizontal and vertical baffles in rectangular tanks (Akyıldız and Ünal,
2005, 2006; Liu and Lin, 2009) and annular baffles in rectangular tanks
(Panigrahy et al., 2009). They concluded that, in an increased fill depth,
the rolling amplitude and frequency of the tank with or without baffle
configurations directly affect the degrees of non-linearity of the sloshing
phenomena. Pal and Bhattacharyya (2010) carried out the numerical
and experimental studies of liquid sloshing for a 2-D problem. The
resulting slosh heights for various excitation frequencies and amplitudes
Fig. 4. Comparative pressure signals at T1 for A8-D75-U case. are compared with the data obtained numerically. Numerical simulation
of liquid sloshing with or without baffles is also examined by Eswaran
et al. (2009) and Chen et al. (2009).
Based on FLUENT software, Modaressi-Tehrani et al. (2007) built a
three-dimensional (3D) nonlinear model of a partly filled horizontal
cylindrical tank with annular curved baffles and investigated the forces
and moments caused by the transient fluid slosh and the effects of baf­
fles. The results clearly showed that the presence of baffles helps to
suppress the peak as well as mean slosh forces and moments signifi­
cantly. The optimisation of baffle installation height and baffle width to
reduce the sloshing of a rectangular storage tank under translational
motion was conducted by using a newly proposed evolutionary opti­
misation method (Kim and Lee, 2008). Based on the numerical tech­
nique named virtual boundary force (VBF) method used to model the
internal baffles with complex geometries, Xue and Lin (2011) developed
a 3D numerical model NEWTANK to study viscous liquid sloshing in a
tank with internal baffles of different shapes and arrangements. Shao
et al. (2015) presented a comparative study of different baffles on
mitigation effects of liquid sloshing in a rectangular tank due to a hor­
izontal excitation using the smoothed particle hydrodynamics (SPH)
method. Using the improved Consistent Particle Method (CPM), water
sloshing with a constrained floating baffle (CFB) was successfully
Fig. 5. Comparative pressure signals at T1 for A8-D50-U case.
simulated. The effect of CFB in sloshing mitigation was investigated
(Koh et al., 2013). Zhou et al. (2014) investigated the nonlinear sloshing
and lower mounted vertical baffles as well as horizontal baffles) on the of liquid in a partially filled rigid cylindrical container with a rigid
hydrodynamic damping of the liquid motion. Akyıldız et al. (2013) annular baffle by using the multi-modal method of Lukovsky-Miles
conducted severalexperiments to analyse liquid sloshing in a cylindrical variational. A time-independent finite difference scheme with the ficti­
tank at a model scale with various fill levels and ring baffles under the tious cell technique was used to study viscous fluid sloshing in 2D tanks
excitation of roll motion. The experimental results of Xue et al. (2013) with baffles, and an experiment setup was also made to validate the
for the liquid sloshing in a rectangular liquid tank with a perforated present numerical sloshing results in a tank with baffles (Wu et al.,
baffle provide an effective baffle arrangement in tanks on reducing the 2012). Wu et al. (2015) provided a finite difference method (FDM) with
sloshing amplitude. The accuracy of the experimental systems was coordinate transformation and fictitious cell approach to analyse the
validated against the numerical results from an in-house robust CFD vortex generation and shedding phenomenon for sloshing liquid in 2D
code. An experimental rig was developed by Zheng et al. (2013) to study tanks with baffles.
non-linear sloshing in a baffled and unbaffled tank of rectangular The above review clearly indicates that there exists a relatively large
dimension. Nayak and Biswal (2015) experimentally investigated the body of literature on liquid sloshing in baffled rectangular, circular,
hydrodynamic damping potential of three different configurations of elliptical and axisymmetric containers with different fill levels. Mean­
centrally installed internal baffles perpendicular to the direction of while, most studies are of concern on determining the natural fre­
lateral excitation in a rectangular tank partially filled with water. It was quencies and mode shapes of liquid sloshing for the baffled problems.
found that the damping coefficient increases with relative baffle height, The review also shows that the studies investigating the detailed flow
and baffles can be effectively used to damp the liquid sloshing near properties around the T-shaped baffles in terms of the flow topology and
resonance conditions. the behaviour of the tip vortices, are highly limited. Accordingly, it was
There are also lots of numerical studies of liquid sloshing with decided to focus on the liquid sloshing in a rotating, partially filled,
different baffles in the containers with complex geometries, such as baffled and unbaffled two-dimensional (2D) rectangular tank to
finite element methods (FEM), finite difference methods (FDM), numerically assess how the height of a fixed-width T-shaped baffle
boundary element methods (BEM), volume of fluid (VOF) technique, relative to the initial liquid (filling) depth affects liquid sloshing and to
virtual boundary force (VBF) method, and meshless method, etc. present detailed flow analysis clearly providing the flow structure in the

3
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 6. Non-dimensional pressure signals obtained from the laminar flow calculations.

tank. Numerical algorithms based on the finite difference and finite In order to address to the tank conditions with relatively high pressure
volume discretization approaches were used in the study. The first nu­ levels, two liquid depths (h) which corresponded to 50% and 75% of the
merical method is based on an in-house code written in FORTRAN, tank height were considered. The ratio of the baffle heights (hB) to the
which performs laminar flow calculations on rather coarse grid systems. liquid depth, h, was chosen as 0.5, 0.75, 0.8, 0.9 and 1.0, respectively.
The turbulent flow computations, on the other hand, were carried out by The time-dependent pressure signals were obtained from a sensor placed
the software package FLUENT 19.2. The first approach is a relatively on the right wall of the tank for each computational case (Fig. 1).
simple method which may be used for the initial design stages to A rotational motion around the axis that is perpendicular to the XY
calculate global impact loads due to the liquid sloshing whilst the latter plane and passing through the bottom centre of the geometry was
is indispensable to perform detailed analyses and to verify the geometry applied to the rectangular tank. The fixed angular frequency of 3.3 rad/s
of the final design. chosen corresponded to roughly 60% of the natural frequency of the
A single-degree-of-freedom motion was considered, and the tank was tank-liquid system for the liquid depths considered. The motion was
rotated with a fixed angular frequency about the axis passing through its represented by a sinusoidal function θ ¼ θ0 sinðωtÞ where θ0 , ω and t
bottom wall centre. The effects of the T-shaped baffle height as well as imply rotation angle (amplitude), angular velocity and time, respec­
the filling depth and rotation angle on the liquid sloshing motion were tively. Two different rotation angles of 4� and 8� were used in the study.
systematically investigated. The time variations of the pressure acting In this way, the effect of the baffle heights as well as the rotation angles
on the tank wall computed by the two methods are presented. The and the liquid depths on the hydrodynamic loads and the wave heights
general flow topology, free surface deformations, and tip vortices was investigated by systematically varying these parameters.
generated by the baffle are also provided and discussed in detail. The cases associated with the liquid depths of 50% and 75% were
coded as D50 and D75, respectively, in the paper. The rotation angles
2. Computational study mentioned above were shown as A4 and A8 throughout the text. On the
other hand, whilst the unbaffled cases were abbreviated as U, the baffled
2.1. Computational models cases were coded as B5, B8 etc. according to baffle height ratio, hB/h. For
example, the case with a liquid depth of 50%, a rotation angle of 4� and a
The 2D rectangular tank geometry, seen in Fig. 1, was used in the baffle height ratio of 0.5 is named as A4-D50-B5. The computational
study. This geometry was formed from the cross-section of the tank with cases investigated in the study can be seen in Table 1.
a length of 46 cm that was used in the experimental work of Akyıldız and
Ünal (2005). The rigid T-shaped baffles mounted to the bottom of the
2.2. Mathematical formulation and numerical approach
tank had a thickness of 10 mm. The preliminary studies and previous
experiences of the authors indicated that a particular width was required
2.2.1. Laminar flow calculations
to provide effective shallow water and blockage effects which helps to
The fluid is assumed to be homogenous, isotropic, viscous and
reduce the sloshing motion. Accordingly, a fixed baffle width of 46 cm,
Newtonian. Tank and fluid motions are assumed to be two-dimensional.
corresponding to the half breadth of the tank was selected in the study.
The domain considered here is a rigid rectangular container partially

4
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 7. Non-dimensional pressure signals obtained from the turbulent flow calculations.

Table 2
Extreme non-dimensional pressure levels and relative reduction in da computed with RANS approach for the first two sloshing periods (up to ωt ¼ 14).
Case Min. Pres. Max. Pres. % red. in da Case Min. Pres. Max. Pres. % red. in da

A4-D50-U 0.60 1.01 N/A A8-D50-U 0.42 1.20 N/A

A4-D50-B5 0.54 1.10 36.6 A8-D50-B5 0.48 1.29 3.8
A4-D50-B7 0.69 1.08 4.9 A8-D50-B7 0.62 1.20 25.6
A4-D50-B8 0.70 1.01 24.4 A8-D50-B8 0.64 1.17 32.1
A4-D50-B9 0.72 0.89 58.5 A8-D50-B9 0.67 1.02 55.1
A4-D50-B1 0.75 0.85 75.6 A8-D50-B1 0.71 0.93 71.8
A4-D75-U 0.77 0.99 N/A A8-D75-U 0.71 1.16 N/A
A4-D75-B5 0.66 1.10 100.0 A8-D75-B5 0.60 1.29 53.3
A4-D75-B7 0.73 1.06 50.0 A8-D75-B7 0.74 1.32 28.9
A4-D75-B8 0.77 1.06 31.8 A8-D75-B8 0.61 1.27 46.7
A4-D75-B9 0.80 0.98 18.2 A8-D75-B9 0.81 1.11 33.3
A4-D75-B1 0.83 0.90 68.2 A8-D75-B1 0.84 1.01 62.2

A4-D25-U 0.24 1.01 N/A A8-D25-U 0 1.34 N/A

A4-D25-B5 0.25 1.06 5.2 A8-D25-B5 0.15 1.49 0.0
A4-D25-B7 0.43 0.94 33.8 A8-D25-B7 0.31 1.10 41.0

filled with liquid, as shown in Fig. 1. conditions of moving walls, the moving coordinate system is used. The
The governing equations are solved simultaneously with the corre­ origin of the coordinate system is at the position of the centre plane of
sponding boundary conditions and free surface kinematics and dynamic the tank and on the undisturbed free surface. The moving coordinate is
boundary conditions in the fluid domain. translating and rotating relative to an inertial system which can be used
to represent general roll or pitch of the tank.
rUðu; vÞ ¼ 0 (1)
The external force per mass consists of gravitational forces, the
∂U 1 translational and rotational inertia forces, which can be written as,
þ U⋅rU ¼¼ ⋅rP þ F þ νr2 U (2)
∂t ρ ∂V ∂Ω
F¼g 2Ω � V �r Ω � ðΩ � rÞ (3)
∂t ∂t
where U ðu; vÞ is the velocity vector defined in the tank fixed coordinate
and, ρ, P, ν and F are the liquid density, pressure, kinematic viscosity and where g; V and Ω are the gravitational vector, the translational velocity
external forces per mass respectively. and the rotational velocity vector, respectively. In addition, r is the
In order to include the non-linearity and avoid the complex boundary position vector of the considered point relative to O. On the free surface,

5
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 8. Free surface deformations at different sloshing phases. (50% filling depth, θ0 ¼ 8� , unbaffled, A8-D50-U).

both the kinematics and dynamic conditions should be satisfied: local effects, such as turbulence and wave breaking have been ignored.
In some cases, these local effects are important, but the simulation of
∂η
þ U⋅rðη yÞ ¼ 0 (4) global flow plays a more critical role in many sloshing problems, due to
∂t
the slosh-induced moment in the ship cargo (Kim, 2001).
P ¼ Patm (5) Volume of Fluid (VOF) technique was used (Hirt and Nichols, 1981)
to track the interface between the fluids. The scheme adopted in this
where η and y represent the disturbed and undisturbed free surface study is the SOLA method of Hirt and Nichols (1981). Tank volume was
levels, respectively. Patm is the air pressure or ullage pressure inside the discretised into Cartesian staggered grid cells. A single layer of fictitious
tank. The surface tension is ignored in this study. Therefore, a no-shear is cells (or boundary cells) surrounded the fluid region. The fictitious cells
needed on the free surface. However, proper wall conditions are were used to set the boundary conditions so that the same difference
necessary on the tank walls and the internal members. equation can be used in the interior of the mesh (Liu and Lin, 2009;
Eswaran et al., 2009).
2.2.1.1. Numerical computation. For the analysis of the sloshing flow Fluid velocities were located at the centres of the cell boundaries and
inside a partial filled tank, a finite difference method (FDM) was applied pressure (P) and the volume of fluid function (F) were computed at the
to the governing equations. FDM is useful when there are internal centre of the cell. The solution algorithm works as a time cycle or “movie
structures inside the tank, or the fluid contacts the tank ceiling frame”. The results of the time cycle act as initial conditions for the next
frequently. As the internal structures exist, the viscous effects may be one. At each step, suitable boundary conditions must be imposed at all
dominant. This method concentrates on the global fluid motion, so some boundaries.

6
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 9. Free surface deformations at different sloshing phases. (75% filling depth, θ0 ¼ 8� , unbaffled, A8-D75-U).

There are two alternatives for the wall conditions; when the viscosity observed in defining the mesh increments Δxi and Δyj, the time incre­
effect on the tank boundary is significant, the no-slip condition should ment Δt and the upstream differencing parameter α.
be imposed. However, in most sloshing problems, the viscous effect is For accuracy, the mesh increments must be chosen small enough to
not significant, and the boundary layer thickness is much less than the resolve the expected spatial variations in all dependent variables.
cell size. Therefore, the free slip condition was applied for the tank walls Generally, the choice of Δt is of extreme importance. In explicit schemes,
and the baffle in the present study. For example, if the left boundary of it will govern the stability and accuracy of the computation, while in
the computing mesh is to be a rigid free slip wall, the normal velocity implicit schemes it will affect the accuracy. Once a mesh has been
will be zero and the tangential velocity should have no normal gradients, chosen, the choice of the time increment necessary for stability is gov­
i.e. erned by two restrictions. First, material cannot move through more
than one cell in one time step, because the difference equations assume
u1;j ¼ 0 v1;j ¼ v2;j for all j (6)
fluxes only between adjacent cells. Therefore, the time increment must
satisfy the inequality,
where j implies the value of the indices in the vertical direction and is
related to the grid number. Δðx; yÞ
Numerical calculations can have quantities that develop large, high Δt < (7)
ΔUi;j
frequency oscillations in space or time or both of them. This behaviour is
usually referred to as a numerical instability. To prevent this type of or,
numerical instability or inaccuracy, certain restrictions must be

7
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 10. Free surface deformations at different sloshing phases. (50% filling depth, θ0 ¼ 8� , hB/h ¼ 0.5, A8-D50-B5).

� �
Δxi Δyj unstable. The instability is related to the instability of the fluid motion,
Δt < Min �� ��; �� �� (8)
ui;j vi;j such as the occurrence of turbulence, wave breaking and the transition
from homogeneous flow to a two-phase flow. For these situations, the
where the minimum is with respect to every cell in the mesh. When a present numerical model is limited to the period prior to the inception of
non-zero value of kinematic viscosity, momentum must not diffuse more these flow perturbations.
than one cell in one time step. A linear stability analysis shows that this
limitation implies, 2.2.2. Turbulent flow calculations
The incompressible, unsteady Reynolds-Averaged-Navier-Stokes
1
υΔt < ! (9) (RANS) equations (Wilcox, 2006) were used in the simulations. In Car­
2⋅ 1
þ 1 tesian tensor notation these equations can be expressed as follows:
Δxi 2 Δyj 2
∂Ui
¼0 (11)
In this study, Δt is automatically chosen to satisfy the above in­ ∂xi
equalities. In order to ensure the numerical stability, the parameter α is, �
� � ∂Ui ∂ Uj Ui ∂P ∂ � ———— �
ui;j ⋅Δt vi;j ⋅Δt ρ þρ ¼ þ μSij ρU ’i U ’j (12)
1 � α � Max ; (10) ∂t xj ∂xi ∂xj
Δx Δy
When the period and amplitude of excitation are large, the liquid where ρ, μ, U, P and U’ imply fluid density, molecular viscosity, aver­
responds violently and causes the numerical solution to become aged velocity, averaged static pressure and fluctuating velocity,

8
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 11. Free surface deformations at different sloshing phases. (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.5, A8-D75-B5).

respectively. Sij term, in Equation (12), represents the mean strain rate pressure-based segregated algorithm was applied for the solution. The
————
velocity-pressure coupling was achieved with the PISO method (Issa,
tensor while the term ρU’i U’j expresses the stresses which are pro­
1986) which a pressure-correction technique providing relatively high
duced by turbulent fluctuations, namely Reynolds stresses, where an convergence speed for time-dependent simulations. The
overbar signifies a shorthand for average. The calculation of the Rey­ pressure-staggering approach of Patankar and Spalding (1972) were
nolds stresses and the turbulence field were realised by means of the SST employed for the calculation of the pressure field where the pressure and
k-ω turbulence model of Menter (1994), which is based on the Boussi­ velocity values are stored at different grid nodes. Whilst a second-order
nesq hypothesis (Tennekes and Lumley, 1972) which, for incompressible central differencing scheme was used for the viscous terms, the spatial
flow, can be represented as follows: discretization of the convective terms of the RANS and turbulent
————
2 transport equations was achieved with a second-order-upwind scheme
ρ U ’i U ’j ¼ 2μt Sij ρkδij (13)
3 (Pletcher et al., 2013). The three-time-level implicit time-discretization
(Ferziger and Peric, 2002) was applied for unconditionally stable and
where μt and k represent turbulent viscosity and turbulent kinetic energy accurate solutions.
respectively while the symbol δij denotes the Kronecker delta. The two-phase flow problem was solved based on the Volume of
Fluid (VOF) technique of Hirt and Nichols (1981). The tracking of the
2.2.2.1. Numerical computation. A fine-volume-discretization-based free surface was performed with a high-accuracy second-order scheme.
solution algorithm was employed (Blazek, 2001; Versteeg and Malala­ In order to increase the accuracy of the simulations, a dynamically
sekera, 2007) to solve the RANS and turbulence transport equations. A varying time step was chosen such that the maximum local Courant

9
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 12. Free surface deformations at different sloshing phases. (50% filling depth, θ0 ¼ 8� , hB/h ¼ 0.9, A8-D50-B9).

number (Courant et al., 1928) in the flow field was kept around 1. This it was found that the normalized time step was somewhat independent
approach also avoided extremely long solution periods due to the un­ of the forcing period. Therefore, about 200 time steps were required per
necessarily small time-step use. The convergence of the solution at each forcing period in this program. This appeared to hold regardless of
time step was checked by the examination of the variation of the discretization.
primitive variables, such as velocity and pressure, in addition to the Reasonable agreement of the time variations of the pressure for
scaled residuals of the equation systems. different grid systems at T1 for A8-D75-U was obtained for all grid
The tank walls and baffle faces were modelled as no-slip walls in the systems while the difference between the results of the medium and fine
turbulent flow calculations. mesh systems was simply indistinguishable. A relatively fine dis­
cretization 46 � 31 with a finer enforcement of the velocity divergence
2.3. Grid dependency requires 3 s of computer time per time step comparing a coarse dis­
cretization 16 � 31. Therefore, 23 � 31 grid system is chosen to obtain
2.3.1. Laminar flow calculations results in a reasonable time considering stability and accuracy.
In order to testify and verify the discretization of the numerical
model for the laminar flow calculations, three different grid systems 2.3.2. Turbulent flow calculations
were chosen (16 � 31, 23 � 31 and 46 � 31) for the unbaffled A8-D75-U A similar analysis with the same computational case was carried out
case. In this study, the value of Δt can be automatically calculated by the for the RANS simulations. The pressure signals obtained from the T1
program and dynamically modified to ensure the stability and also point by using three different mesh resolutions with nominal grid
optimize the pressure solution. When the excitation is harmonic rolling, spacing of 1, 2 and 4 mm can be comparatively examined for A8-D75-U

10
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 13. Free surface deformations at different sloshing phases. (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.9, A8-D75-B9).

in Fig. 2. In order to place a few cells in the boundary layer, the grid size of the data collection in the tests the computational and experimental
in these regions were kept around 1 mm and gradually increased, if signals were time-aligned such that the first peaks of the signals overlap.
necessary, depending on the nominal grid resolution. As seen in the As shown in Figs. 4 and 5, the magnitude of the non-dimensionalised
figure, the agreement between the three plots is high. Since the results pressure signals are in reasonable agreement with the experimental
obtained from the grid spacings of 1 mm and 2 mm were almost iden­ data for both laminar and RANS calculations. The concordance between
tical, subsequent simulations were carried out with the former grid the turbulent and laminar calculations applied is also remarkable.
resolution. A view of the mesh structure used for the RANS simulations Whilst the periodicity of the sloshing motion is also well captured with
can also be seen in Fig. 3. the numerical methods used for both cases, the first case investigated,
A8-D75-U, produced a slightly lower sloshing frequency compared to
2.4. Validation the experiments (Fig. 4). The main sloshing motion frequency is calcu­
lated as 0.525 Hz whilst this value was 0.560 Hz in the experiments. The
In order to assess the accuracy of the methods used, the results of the frequency obtained from the second computational case, A8-D50-U, on
pressure at T1 obtained from the unbaffled conditions by using the the other hand, is in good agreement with the experimental data (Fig. 5).
above mentioned two numerical methods were compared with those of The computational and experimental values of the main motion fre­
the experimental work of Akyıldız and Ünal (2005, 2006) for the two quency was obtained as 0.525 Hz and 0.515 Hz, respectively.
computational cases, A8-D75-U and A8-D50-U. Due to the The validation study showed that the simulations involving T-shaped
low-frequency data collection, the experimental signals have sharp baffles can be confidently performed by using the two numerical
edges. It should be noted that due to the uncertainty of the initialisation methods adopted.

11
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 14. Streamlines at different sloshing phases. (75% filling depth, θ0 ¼ 8� , unbaffled, A8-D75-U, coloured by velocity magnitude in m/s).

It should be born in mind that in order to achieve 2D flow conditions, above the bottom for all cases. The absolute pressure values which are
the length-to-breadth ratio of the tank should be as small as possible to equal to the sum of the aero-hydrostatic and static pressure values were
avoid any 3D flow such as swirling, diagonal waves etc. for the rolling non-dimensionalised with respect to the hydrostatic pressure level
excitation of a tank (Faltinsen and Timokha, 2009). Hence, the results of acting on the bottom surface. P, ρ, g, seen in Fig. 6, imply absolute
the present 2D simulations should be essentially interpreted with the pressure (Pa), liquid density (kg/m3) and gravitational acceleration (m/
“slender tank” assumption. However, the agreement with the results of s2), respectively. As is seen, some of the signals cannot reach to the
the 3D tank experiments with a tank length of 46 cm also implied targeted time point, ωt ¼ 14, since the computer code used in the
negligible 3D flow influences for the unbaffled cases. Nonetheless, the calculation was not designed to model the complex sloshing motions,
application of a baffle may increase these effects if it does not signifi­ such as liquid spray and wave breaking, which occurs after the impact of
cantly affect the global impact loads and main flow structure. the liquid on the top surface. The pressure oscillations mainly display a
sinusoidal structure for the unbaffled cases. The amplitude of the os­
3. Results and discussion cillations is higher for the D50-U cases with respect to the D75-U cases
for both rotational angles adopted. The increase of the rotational angle
3.1. Laminar flow calculations also raised the amplitude of the oscillations, as expected.
As far as the baffled cases are concerned, it can be clearly observed
Fig. 6 displays the non-dimensional pressure signals obtained from that the baffle can effectively dampen the amplitude of the pressure
the T1 probe which was placed on the right wall at a height of 6 cm oscillations for D50-B cases when hB/h � 0.8. For the cases with the same

12
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 15. Vorticity magnitudes at different sloshing phases. (%75 filling depth, θ0 ¼ 8� , unbaffled, A8-D75-U, in 1/s).

liquid depth and hB/h < 0.8, the amplitude of the oscillations and/or the which could be captured with the very high time resolution of the
maximum level of the pressure further increased. This is also evidenced simulations (also see Figs. 10–13). It can also be seen that some of the
by the figures (Figs. 10–13) presented in the following section. A similar unbaffled cases displayed multiple peaks during the second period of the
situation applies to the D75-B cases. This time, however, the results sloshing motion. This is due to the interference of the free surface waves
indicate that the baffle was incapable of damping the oscillations when and fluid flow heading to the side walls of the tank and reflecting from
hB/h ¼ 0.8. For the cases with hB/h > 0.8, whilst the signal amplitudes them within this interval. The time resolution of the computations
significantly diminished. allowed the capture of the phenomenon. The maximum and minimum
pressure levels along with the differences between the extremes of the
3.2. Turbulent flow calculations pressure signal amplitudes (da ) obtained from the computations are
collected in Table 2. Fundamentally, the results pointed out that the
Much more detailed analyses were performed for the turbulent flow baffles were fully effective with the baffle height ratios of hB/h > 0.8 for
calculations. Shown in Fig. 7 are the non-dimensional pressure signals the two liquid depths and rotation angles considered in the study.
obtained from the T1 sensor. This numerical technique, which allowed a Although the general character of the pressure signals is different, in
higher precision in resolving the complex sloshing motion, indicated a terms of determination of the effective height ratios of the baffles, the
non-sinusoidal and oscillating curve structure for the unbaffled cases in results of the turbulent flow simulations are in close agreement with the
contrast to the results of the laminar flow calculations. The instant rises observations previously presented for the laminar calculations. Table 2
and drops, particularly observed in the baffled cases, are due to the demonstrates that the application of baffle with an adequate height can
liquid sprays and wave breakings occurring during the sloshing motion significantly damp the liquid sloshing motion and reduce da with respect

13
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 16. Streamlines at different sloshing phases (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.5, A8-D75-B5, coloured by velocity magnitude in m/s; The figure shares the
same legend given in Fig. 14).

to those of the unbaffled cases. Considering D50 cases, this effect along other hand, the motion and impact load damping effect of the baffles
with the reduction in the maximum pressure level can be observed when appeared to be relatively diminished for the higher filling depth. This
hB/h � 0.8. For instance, whilst the non-dimensional pressure values can be clearly evidenced by the examination of the reduction in da for
were within ~0.4–1.2 range for the unbaffled A8-D50-U case, the values the B9 and B1 cases for both rotation angles (Table 2).
of the signal varied within a range of ~0.7–0.9 with the application of For the two liquid depths and rotation angles adopted, the period of
the baffle with a height ratio of hB/h ¼ 1.0 (A8-D50-B1). This indicates a the pressure signals was also increased with increasing baffle heights.
serious decrease of about 72% in da . However, for the D50-B8 cases (hB/ For the case of hB/h ¼ 1.0, however, the oscillation period value
h ¼ 0.8), during the second period of the sloshing motion, the maximum returned to the value of the unbaffled case’s period.
levels of the pressure signals were increased in comparison with those of Although this study is focused on the sloshing motion in the tank
the unbaffled cases, due to the complex flow structure generated by the considering two filling depths of 50% and 75%, additional simulations
baffles. A similar trend can be observed for D75. Only the cases with hB/ with a filling depth of 25% were also performed to observe the influence
h > 0.8 were successful in diminishing both da and maximum pressure of the baffles for relatively lower liquid levels and to make a clearer
level. In A8-D75, the decrease of da was computed as around 62% within conclusion about their effectiveness. The ratio of the angular frequency
the first two sloshing periods with the application of the largest baffle. of the simulations to the natural frequency of the tank was kept at 0.6.
The table indicates a significant rise in da for the rest of the computa­ Table 2 also covers the results of these simulations. It can be observed
tional cases. For these cases, the application of baffle negatively affected that the motion damping effect of the baffles was increased. For the two
sloshing motion and increased the amplitude of the impact loads. On the rotation angles considered, the baffle with a height ratio of hB/h ¼ 0.75

14
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 17. Vorticity magnitudes at different sloshing phases (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.5, A8-D75-B5, in 1/s; The figure shares the same legend given
in Fig. 15).

was sufficient to keep the oscillation amplitudes and maximum pressure structure of the free liquid surface at the different phases within the first
levels below those of the unbaffled tank. two sloshing periods (2T) for different cases.
The reason of the rise of the pressure signal amplitudes for the cases As seen from Figs. 8 and 9, the motion has a simple and organised
of hB/h � 0.8 in comparison with the unbaffled cases can be perceived by form with a minimal free surface deformation for the unbaffled cases.
the investigation of the graphs in Figs. 8–13. The phase numbers seen in The maximum liquid level, which was approximately 0.5 m high for A8-
Fig. 8 correspond to the motion times of t¼T/4, t¼T/2, t¼3T/4, t¼T, D50-U, reached to the ceiling of the tank for A8-D75-U within the first
t¼5T/4, t¼3T/2, t¼7T/4, t¼2T, respectively. The same phases were used sloshing period. In the second period, due to the wave reflections created
in all subsequent figures. The X and Y axes in the figures show the dis­ by the wall and mutual cancellation of the waves, the free surface levels
tance from the bottom centre of the tank in meters. In this section, six remained lower than those of the first period.
specific cases were selected for the demonstration of the effect of the On the other hand, Figs. 10 and 11 indicated that the structure of the
baffle on the sloshing motion for a clearer and concise presentation. The free surface became highly complex and violent with the application of
reason of the selection of the rotation angle of 8� is that this configu­ the T-shaped baffle with a ratio of hB/h ¼ 0.5, which was amongst the
ration provides a violent free surface motion that can easily reach to the cases that made the wall pressures and the amplitudes of the oscillations
top wall of the tank. B5s are clear examples of the strong negative impact increase with respect to those of the unbaffled cases (see Figs. 6 and 7
of the introduction of the baffle and B9s are the cases where the damping and Table 2). Even within the first period of the motion, the levels of the
influence of the baffle is explicitly apparent. The figures show the liquid were higher than those of the unbaffled case and the fluid strongly

15
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 18. Streamlines at different sloshing phases (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.9, A8-D75-B9, coloured by velocity magnitude in m/s; The figure shares the
same legend given in Fig. 14).

hit the side and top walls of the tank. A large amount of liquid appeared are coloured according to the velocity magnitudes and were obtained for
to be clearly free to move with a high momentum intensity. For the the aforementioned three cases of D75. The thick solid line represents
second period of the motion the free surface became highly disorganised the free surface. The vorticity distributions at different sloshing phases
and very rough. It was also exposed to a strong liquid spray and wave computed for the same cases can also be examined in Figs. 15, 17 and 19.
breaking phenomenon (Figs. 10 and 11). One can deduce that, if the The figures more clearly reveal the structure of the sloshing motion.
height of the baffle is insufficient, the sloshing motion and hence the For A8-D75-U, the flow primarily included two well-defined co-
pressure levels acting on the tank walls are notably amplified. rotating structure. The odd numbered phases of this unbaffled case
As far as the cases with hB/h ¼ 0.9 are concerned (see Figs. 12 and retained this organised counterclockwise rotational motion for the first
13), it can be clearly seen that the structure of the free surface is still two periods examined. Two additional clockwise rotating motions were
complex. Although the liquid level could still reach to the top wall of the also visible at the bottom corners of the tank for the 7th phase. The
tank, the major portion of the liquid motion was effectively blocked by organised structure of the rotation was somewhat broken down for the
the baffle. The amount of liquid hitting to the walls was low and the free even numbered phases, in which rightward and leftward motions were
surface roughly remained horizontal. This flow structure clarifies the dominant. The 6th case, however, included a saddle point around which
drops of the pressure signals for the cases hB/h � 0.8. weak rotations were existent. The motion, in general, was remarkably
The flow topology of the sloshing motion can be deeply investigated slow compared to those seen in the baffled cases. The area-weighted-
benefiting from the plots presented in Figs. 14–19. Shown in Figs. 14, 16 average velocity magnitudes at each sloshing phase for the three cases
and 18 are the streamlines of the flow at different sloshing phases, which in question can be examined in Table 3. The area-weighted-average can

16
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Fig. 19. Vorticity magnitudes at different sloshing phases (75% filling depth, θ0 ¼ 8� , hB/h ¼ 0.9, A8-D75-B9, in 1/s; The figure shares the same legend given
in Fig. 15).

Table 3
Area-weighted-average velocity magnitude at constant phases.
Case Magnitude (m/s) Global
Average
1 2 3 4 5 6 7 8

A8-D75-U 0.036 0.138 0.065 0.143 0.085 0.106 0.041 0.118 0.092
A8-D75-B5 0.050 0.132 0.149 0.213 0.090 0.165 0.067 0.203 0.134
A8-D75-B9 0.043 0.135 0.085 0.176 0.083 0.162 0.098 0.174 0.120
A8-D75-B1 0.059 0.144 0.055 0.141 0.053 0.146 0.054 0.136 0.099

be represented as follows: obtained from A8-D75-B1, which will be discussed later. The values of
Pn the odd numbered phases were substantially small compared to those of
i¼1 ai ∅i
∅ ¼ P n (14) the even numbered phases probably due to the effect of the mutual
i¼1 ai cancellation of the free surface waves and momentum. The maximum of
these values occurred in 4th phase where the maximum velocity
where ∅ implies an arbitrary variable value to be averaged and ai is the
magnitude was around 0.95 m/s within a highly confined zone
area of each computational grid cell. Table 3 also includes the values

17
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Table 4
Area-weighted-average vorticity magnitude at constant phases.
Case Magnitude (1/s) Global
Average
1 2 3 4 5 6 7 8

A8-D75-U 0.486 0.788 0.632 0.367 0.513 0.792 0.571 0.307 0.557
A8-D75-B5 0.671 1.282 0.982 0.811 1.116 1.600 1.187 1.088 1.092
A8-D75-B9 0.358 0.659 0.501 0.908 0.672 0.901 0.594 0.947 0.693
A8-D75-B1 0.777 1.372 1.100 0.699 0.905 1.371 0.977 0.666 0.983

appearing at the rightmost free surface boundary for the unbaffled case. the 4th phase. Eventually, the vortices mixed with the vorticity in the
Also, it appeared that there existed almost no shear stress for the most boundary layer and the other vortices generated due to the boundary
part of the flow indicating a similar character to that of a rigid body layer or free surface motion. The interpretation of these results along
motion. For this reason, the associated vorticity distributions shown in with the streamlines shown in Fig. 16 reveals that these systematic
Fig. 15 imply that the vorticity in the liquid flow was simply generated vortex structures and the blockage of the baffle increase the momentum
due to the existence of the boundary layer zones. The area-weighted- of the flow above the T-shaped baffle and contribute to the violent
average vorticity levels at each sloshing motion phase are collected in motion of the free surface. Consequently, while one could expect a
Table 4. The average value of the vorticity magnitude was lower than diminishment in the sloshing motion and attenuation in the pressure
those computed for the baffled cases and almost 50% lower compared to levels acting on the tank walls due to the motion blockage of the baffle,
the average vorticity level of A8-D75-B5. The wall shear effect which despite the energy loss due to the generation of the vortices, blockage
generated the vorticity, indeed, directly contributed to the liquid and additional wall shear, the opposite effect occurred.
sloshing as a motion damping factor. The liquid motion in the vicinity of The results associated with the case A8-D75-B9 (hB/h ¼ 0.9) are
the side walls also created well defined small vorticity areas around the presented in Figs. 18 and 19. It can be seen that the flow velocity
intersection of the free surface and side walls (see 4th and 5th phases of magnitudes significantly reduced for the most part of the liquid zone.
Fig. 15). Table 3 indicates a reduction of 10% with respect to the average velocity
Considering the case of A8-D75-B5 (hB/h ¼ 0.5), whilst various levels of A8-D75-B5, however they were still significantly higher (30%
centre of rotations can be observed in Fig. 16, the velocity magnitude in global average) than those observed in the unbaffled case. Relatively
was significantly and progressively risen, particularly in the zone high velocities occurred in the confined area between the top of the
confined between the free surface and the top surface of the baffle. baffle and the free surface and in the local zones where high vorticity
Mainly, due to the limitation of the flow motion zone by the T-shaped levels existed due to the shallow water effect. Although a reduction in
baffle, the momentum of the flow above the baffle considerably the global averaged velocity magnitudes was obtained, they reached to
increased and intensified the sloshing motion of the liquid. The area- relatively very high values of up to 1.75 m/s in specific regions.
weighted-average velocity magnitudes calculated are given in Table 3. The odd numbered phases displayed a flow structure basically
Almost at each phase the magnitudes were higher than those observed in formed by two large rotational zones at both sides of the baffle, which
the unbaffled case. also included several smaller rotational structures. The tip vortices were
The global average velocity magnitude was around 46% larger for still generated by the ends of the baffle; however, they were prematurely
this baffled case. It should be noted here that the average includes the shed as they rapidly broke down by the free surface flow. The vortex
contribution of the slower motion zones below the baffle which more shedding frequency was substantially high. The first vortex shedding
clearly reveals the higher momentum levels occurring above. The higher took place before the first sloshing phase. According to Table 4, a
values of the velocity magnitudes took place near the tips of the baffle notable reduction of 36% in the global average vorticity magnitudes
due to the vortices generated and around the free surface. As a com­ created, due to the absence of the well organised tip vortices, was ob­
parison, for the most accelerated sloshing phase (4th phase), whilst the tained. As the free surface motion occurred very close to the top surface
maximum velocity around the tip was 1.4 m/s, the flow velocity due the of the baffle, the vorticity levels within this area was quite strong. The
free surface motion reached to values higher than 1.8 m/s. This value is organised and diffused vorticity zones apparent after the 5th phase is a
almost 90% larger than that calculated for A8-D75-U. Although the result of the blending of the tip vortices with the significant vorticity
generation of the tip vortices causes a loss in the motion energy, it generated by the boundary layer flow and/or free surface motion due to
appeared that it augmented the momentum of the liquid above the baffle the shallow water effect. The contribution of the wall shear was also
and caused the free surface motion to grow stronger. Well defined higher due to the increased baffle surface area and stronger interaction
rotational motions gradually increased in the odd numbered phases. The of the boundary layer and free surface flow. The latter coupled with a
even numbered phases, on the other hand, contained higher momentum lower velocity/weaker momentum fluid flow effectively dampened the
levels, however, displayed less rotational structures. The associated free surface levels and the pressures acting on the walls.
vorticity distribution presented in Fig. 17 demonstrates the existence of It should be noted here that, although the related plots are not given
the well-organised tip vortices generated by the two free ends of the in an effort to provide a concise presentation, the motion blockage seen
baffle, which were immediately formed with the rotational motion of in A8-D75-B1 was significantly high, and the velocity levels dropped to
the tank and subsequently shed into the liquid zone in a regular fashion the values of the unbaffled cases (Table 3). The systematic vortex
advancing in the right and left directions depending on the clockwise or shedding was completely absent in this case due to the dynamic free
counterclockwise rotation of the tank. By the investigation of Table 4, surface motions contacting the upper wall of the baffle. Still, the
one can see that, due to this vortex generation, the global average vorticity levels were once again amplified since the latter was the
vorticity level doubled in respect to that of the unbaffled case. The levels dominant factor generating the vorticity (Table 4). Due to the stronger
were also significantly larger at each sloshing phase. The strength of influence of the convective blockage and wall shear effect, the pressure
these tip vortices was initially quite high (up to 200 1/s in the 2nd levels and free surface deformations were substantially damped.
phase). They also became stronger before the instant they broke up.
Subsequently, the vorticity area was grown but its strength gradually 4. Conclusion
diminished. For instance, the core strength of the clockwise vortex
generating from the left tip at the 2nd phase of the motion was about 86 Two distinct numerical methods involving laminar and turbulent
1/s. After the shedding phenomenon its core strength dropped by 50% at flow calculations based on the VOF technique were applied for

18
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

simulating the single-degree-of-freedom liquid sloshing in a partially within the first two periods of the motion for the filling depths of
filled, rotating 2D rectangular tank with bottom-mounted, fixed-width 50% and 75%. However, the effectiveness of the baffles was
T-shaped baffle. A systematic approach was presented to reveal the ef­ increased with decreasing filling depths. When the liquid depth was
fect of the vertical baffle height on the sloshing motion for two rotating further reduced to 25%, a height ratio of hB/h ¼ 0.75 was sufficient
angles and different filling depths. The time history of the pressure to keep the oscillation amplitudes and maximum pressure levels
signals acting on the side wall as well as the general topological features below those of the unbaffled tank.
of the flow, free surface deformations and vorticity distributions around � The motion damping property of the baffle is connected with the
the baffle at constant sloshing phases were comparatively presented and optimal balance between the wall shear and blockage effects along
discussed in detail. with the energy loss by the generation of the vorticity. When the
The two numerical methods produced well agreed results in terms of height was not adequate, the liquid convection was intensified, and
the fundamental effect of the T-shaped baffle on the pressure signals the generated vorticity and friction was insufficient to expense
obtained from the tank surface for all cases investigated. As pointed out enough energy to suppress the free surface motion and pressure
previously, the laminar flow calculations were incapable of simulating levels. As the baffle height increased, the blockage and wall shear
the wave breaking and spraying phenomena and this method was not became the dominant effect and a significant reduction in free sur­
appropriate for the situations when the liquid reached the ceiling of the face and pressure levels were achieved.
tank. However, as presented, this did not jeopardise the global impact
load predictions of this approach. As the turbulent simulations It is believed that the present study serves as a valuable tool in the
detailedly modelled wave braking and spraying phenomena, the pres­ estimation of the sloshing characteristics in rectangular tanks and in the
sure signals obtained were rather irregular. The results of both simula­ design of the T-baffles suitable to the user’s requirements and also sheds
tion techniques were also in agreement with the experimental results further light to the understanding the complex fluid motion due to the T-
found in the open literature for the unbaffled cases. The computations shaped baffles. Future works will involve multiple-degree-of-freedom
indicated that the laminar flow approach with its relatively coarse grid analyses to investigate the impact loads due to the sloshing motion in
resolution and rapid calculation performance associated with the global 3D prismatic tanks.
sloshing loads is suitable to be used at early design stages of tank ge­
ometry and structure enabling the consideration of various design al­ References
ternatives. The turbulent flow approach, on the other hand, is
computationally much more expensive; however, it may produce highly Akyıldız, H., 2012. A numerical study of the effects of the vertical baffle on liquid
sloshing in two-dimensional rectangular tank. J. Sound Vib. 331, 41–52.
detailed results in the sense of the turbulent flow features around the Akyıldız, H., Ünal, N.E., Aksoy, H., 2013. An experimental investigation of the effects of
baffle which affect the sloshing motion and impact loads. It is, then, the ring baffles on liquid sloshing in a rigid cylindrical tank. Ocean Eng. 59,
suitable for performing extensive liquid sloshing analyses and it may be 190–197.
Akyıldız, H., Ünal, E., 2005. Experimental investigation of pressure distribution on a
used at later stages of the design process when the detailed geometrical rectangular tank due to the liquid sloshing. Ocean Eng. 32, 1503–1516.
definitions of the tank and baffle are available. Akyıldız, H., Ünal, N.E., 2006. Sloshing in a three dimensional rectangular tank:
The critical findings of the investigation are briefly addressed as numerical simulation and experimental validation. Ocean Eng. 33 (16), 2135–2149.
Biswal, K.C., Bhattacharyya, S.K., Sinha, P.K., 2006. Non-linear sloshing in partially
follows: liquid filled containers with baffles. Int. J. Numer. Methods Eng. 68, 317–337.
Blazek, J., 2001. Computational Fluid Dynamics: Principles and Applications. Elsevier,
� The amplitudes of the pressure signals decreased with increasing Oxford, UK.
Chen, Y.G., Djidjeli, K., Price, W.G., 2009. Numerical simulation of liquid sloshing
filling depth for both rotation angles considered for the unbaffled
phenomena in partially filled containers. Comput. Fluid 38, 830–842.
cases. On the other hand, the increase of the rotation angle resulted Courant, R., Friedrichs, K., Lewy, H., 1928. Über die partiellen differenzengleichungen
in the intensification of the sloshing motion along with the pressure der mathematischen physik. Math. Ann. 100 (1), 32–74.
oscillation amplitudes, as expected. Eswaran, M., Saha, U.K., Maity, D., 2009. Effect of baffles on a partially filled cubic tank:
numerical simulation and experimental validation. Comput. Struct. 87, 198–205.
� The application of the T-shaped baffle amplified the complexity of Faltinsen, O.M., Timokha, A.N., 2009. Sloshing. Cambridge University Press, UK.
the free surface and caused stronger wave breaking, liquid spraying Ferziger, J.H., Peric, M., 2002. Computational Methods for Fluid Dynamics. Springer,
compared to those observed in the unbaffled cases. New York, USA.
Gavrilyuk, I., Lukovsky, I., Trotsenko, Y., Timokha, A., 2006. Sloshing in a vertical
� The baffle caused a substantial increment in the area-weighted- circular cylindrical tank with an annular baffle. Part 1. Linear fundamental solutions.
average velocity and vorticity levels in the tank. Initially, for hB/ J. Eng. Math. 54 (1), 71–88.
h ¼ 0.5, the global average velocity raised by 46% and this value Goudarzi, M.A., Sabbagh-Yazdi, S.R., 2012. Analytical and experimental evaluation on
the effectiveness of upper mounted baffles with respect to commonly used baffles.
gradually decreased as the height of the baffle increased. Still, the Ocean Eng. 42, 205–217.
velocities in the zone between the top wall of the baffle and free Gradinscak, M., Semercigil, S., Turan, O.,€ 2011. A sloshing absorber with a flexible
surface remained high. container. Structural Dynamics 3, 315–322.
Hasheminejad, S.M., Mohammadi, M., 2016. Active sloshing control in a smart flexible
� The vorticity in the sloshing motion was typically created by the tips cylindrical floating roof tank. J. Fluids Struct. 66, 350–381.
of the baffle, free surface and tank boundary layers. When hB/ Hasheminejad, S.M., Mohammadi, M.M., Jarrahi, M., 2014. Liquid sloshing in partly-
h < 0.8, since the free surface weakly affected the flow around the filled laterally-excited circular tanks equipped with baffles. J. Fluids Struct. 44,
97–114.
baffle, the tip vortices generated were well-defined and organised.
Hirt, C.W., Nichols, B.D., 1981. Volume of fluid method for the dynamics of free
However, for hB/h � 0.8, these vortices were prematurely shed as boundaries. J. Comput. Phys. 39, 201–225.
they rapidly broke down by the free surface flow, until the systematic Ibrahim, R.A., 2005. Liquid Sloshing Dynamics: Theory and Applications. Cambridge
shedding is completely disappeared for hB/h ¼ 1.0. University Press, New York.
Issa, R.I., 1986. Solution of the implicitly discretized fluid flow equations by operator
� The vorticity levels were more influenced by the existence of the splitting. J. Comput. Phys. 62 (1), 40–65.
baffle and its height than the velocity levels. For hB/h ¼ 0.5, the Jung, J.H., Yoon, H.S., Lee, C.Y., Shin, S.C., 2012. Effect of the vertical baffle height on
vorticity levels doubled compared to the unbaffled case examined. the liquid sloshing in a three-dimensional rectangular tank. Ocean Eng. 44, 79–89.
Kim, H.S., Lee, Y.S., 2008. Optimization design technique for reduction of sloshing by
The increase of the height of the baffle significantly dropped the evolutionary methods. J. Mech. Sci. Technol. 22, 25–33.
vorticity intensities. However, due to the dominance of the free Kim, Y., 2001. Numerical simulation of sloshing flows with impact load. Appl. Ocean
surface motion contacting the upper wall of the baffle, the vorticity Res. 23, 53–62.
Koh, C.G., Luo, M., Gao, M., Bai, W., 2013. Modelling of liquid sloshing with constrained
levels were once again significantly augmented for hB/h ¼ 1.0. floating baffle. Comput. Struct. 122, 270–279.
� According to the pressure signals and free surface deformations ob­ Liu, D.M., Lin, P.Z., 2009. Three-dimensional liquid sloshing in a tank with baffles. Ocean
tained from the simulations it was revealed that the T-shaped baffle Eng. 36, 202–212.
was ineffective in completely damping the sloshing for hB/h � 0.8

19
U.O. Ünal et al. Ocean Engineering 187 (2019) 106183

Maleki, A., Ziyaeifar, M., 2008. Sloshing damping in cylindrical liquid storage tanks with Versteeg, H.K., Malalasekera, W., 2007. An Introduction to Computational Fluid
baffles. J. Sound Vib. 311, 372–385. Dynamics: the Finite Volume Method. Longman Group Ltd., Harlow, England.
Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineering Wang, W., Peng, Y., Zhou, Y., Zhang, Q., 2016. Liquid sloshing in partly-filled laterally-
applications. AIAA J. 32 (8), 1598–1605. excited cylindrical tanks equipped with multi baffles. Appl. Ocean Res. 59, 543–563.
Mitra, S., Khoo, B., Kumar, A., Sinhamahapatra, K., 2010. Influence of various internal Wang, J.D., Lo, S.H., Zhou, D., 2013. Sloshing of liquid in rigid cylindrical container with
components on slosh dynamics: a study using finite element method. Int. J. Fluid multiple rigid annular baffles: lateral excitations. J. Fluids Struct. 42, 421–436.
Mech. Res. 37 (4). Wilcox, D.C., 2006. Turbulence Modelling for CFD. DCW Industries, California, USA.
Modaressi-Tehrani, K., Rakheja, S., Stiharu, I., 2007. Three-dimensional analysis of Wu, C.H., Faltinsen, O.M., Chena, B.F., 2015. Dynamics of vortex evolution in a 2D
transient slosh within a partly-filled tank equipped with baffles. Veh. Syst. Dyn. 45, baffled tank. Comput. Math. Appl. https://doi.org/10.1016/j.camwa.2015.10.007.
525–548. Wu, C.H., Faltinsen, O.M., Chen, B.F., 2012. Numerical study of sloshing liquid in tanks
Nayak, S.K., Biswal, K.C., 2015. Fluid damping in rectangular tank fitted with various with baffles by time-independent finite difference and fictitious cell method.
internal objects – an experimental investigation. Ocean Eng. 108, 552–562. Comput. Fluid 63, 9–26.
Pal, P., Bhattacharyya, S.K., 2010. Sloshing in partially filled liquid containers – Xue, M.A., Lin, P.Z., Zheng, J.H., Ma, Y.X., Yuan, X.L., Nguyen, V.T., 2013. Effects of
numerical and experimental study for 2-D problems. J. Sound Vib. 329, 4466–4485. perforated baffle on reducing sloshing in rectangular tank: experimental and
Panigrahy, P.K., Saha, U.K., Maity, D., 2009. Experimental studies on sloshing behavior numerical study. China Ocean Eng. 27 (5), 615–628.
due to horizontal movement of liquids in baffled tanks. Ocean Eng. 36 (3–4), Xue, M.A., Lin, P., 2011. Numerical study of ring baffle effects on reducing violent liquid
213–222. sloshing. Comput. Fluids 52 (1), 116–129.
Patankar, S.V., Spalding, D.B., 1972. A calculation procedure for heat, mass and Xue, M.A., Zheng, J., Lin, P., 2012. Numerical simulation of sloshing phenomena in cubic
momentum transfer in three-dimensional parabolic flows. Int. J. Heat Mass Transf. tank with multiple baffles. J. Appl. Math. 203–222. https://doi.org/10.1155/2012/
15, 1787–1806. 245702.
Pletcher, R.H., Tannehill, J.C., Anderson, D.A., 2013. Computational Fluid Mechanics Zang, Q., Huang, J., Liang, Z., 2015. Slosh suppression for infinite modes in a moving
and Heat Transfer, third ed. CRC Press, Florida, USA. liquid container. IEEE ASME Trans. Mechatron. 20 (1), 217–225.
Pridgen, B., Bai, K., Singhose, W., 2012. Shaping container motion for multimode and Zheng, J.H., Kargbo, O., Yuan, X.L., Xue, M.A., Lin, P.Z., 2013. An experimental study of
robust slosh suppression. J. Spacecr. Rocket. 50 (2), 440–448. nonlinear liquid sloshing in a rectangular tank. In: Proceedings of the 23rd
Shao, J.R., Li, Z.R., Liu, M.B., 2015. A comparative study of different baffles on International Offshore and Polar Engineering Conference, pp. 171–177.
mitigating liquid sloshing in a rectangular tank due to a horizontal excitation. Eng. Zhou, D., Wang, J.D., Liu, W.Q., 2014. Nonlinear sloshing of liquid in rigid cylindrical
Comput. 32 (4), 1172–1190. container with a rigid annular baffle: free vibration. Nonlinear Dynam. 78,
Tennekes, H., Lumley, J.L., 1972. A First Course in Turbulence. MIT Press, Cambridge, 2557–2576.
UK.

20