Ocean Engineering 142 (2017) 348–362

Contents lists available at ScienceDirect

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

Numerical investigation of rolling response of a 2D rectangular hold, MARK

partially filled with moist bulk cargo
⁎
Lei Jua, Yanzhuo Xuea, , Dracos Vassalosb, Yang Liuc, Baoyu Nia
a
Harbin Engineering University, College of Shipbuilding Engineering, 150001 Harbin, China
b
Maritime Safety Research Centre, Department of Naval Architecture, Ocean & Marine Engineering, University of Strathclyde, 100 Montrose Street,
Glasgow G4 0LZ, Scotland, UK
c
Department of Civil Engineering, University of Science and Technology Beijing, 100083 Beijing, China

A R T I C L E I N F O A BS T RAC T

Keywords: Liquefaction of mineral ores, such as iron ore and nickel ore fines, resulting in cargo liquefaction and hence
Solid bulk cargo displacement (‘cargo shift’) leading to loss of ship stability, has been a major cause of marine casualties. In the
Cargo shift paper, rolling response in regular waves for a 2D rectangular hold, partially filled with moist bulk cargo is
Ship stability investigated, by applying a simulation approach based on an adapter that couples a CFD code and a DEM
CFD-DEM adapter
(‘Discrete Element Method’) code. A parametric investigation is undertaken to identify the key parameters that
trigger cargo and ship instabilities.

1. Introduction intended to ensure that, to avoid liquefaction, only cargoes with

sufficiently low inherent moisture content are loaded. Despite the fact
Solid bulk cargoes can shift by sliding failure or liquefaction that the IMSBC Code prescribes the testing and certification require-
consequence, and whilst the factors involved in each of these processes ments designed to ensure that cargoes are loaded only if the moisture
are different, the potentially disastrous consequences are the same, content is sufficiently low to avoid liquefaction occurring during the
capsizing and structural damage (Gard, 2014). The most widely known voyage, inaccurate declarations and certificates from shippers still
solid bulk cargoes with this hazard are nickel ore and iron ore which appear to be at the heart of the problem.
are widely distributed in all over the world like Indonesia, Philippines With the stimulus, compaction and vibration of cargoes during a
and New Caledonia (Ju et al., 2016). Liquefaction problems involved in voyage, cargoes like iron ore and nickel ore may become fluid. Ship
the carriage of bulk cargoes were recognized by the delegates to the motion may cause liquefied cargoes to shift such that the vessel capsizes.
1960 International Conference on Safety of Life at Sea. The first edition Cargo shift can be divided into two types, including sliding failure or
of the Code of Safe Practice for Solid Bulk Cargoes (BC Code) had been liquefaction consequence as summarized in the IMSBC code. In actual
published in 1965 by Sub-Committee of International Maritime fact, the sliding failure and liquefaction consequence exist simulta-
Organization (IMO). Nickel ore is newly being categorized as Group neously. The sliding failure may occur when liquefaction accumulates
A cargo in IMSBC code. Group A Cargoes mean cargoes which may to a certain degree, which further intensifies the onset of liquefaction.
liquefy if shipped at moisture content (MC) in excess of the transpor- The moisture content of the cargo is a key parameter which causes
table moisture limit (TML). The lowest moisture content at which liquefaction. If the moisture is below a certain level, the interstitial spaces
liquefaction can occur is called the Flow Moisture Point (commonly between the particles could accommodate all of the moisture such that
abbreviated FMP) which can be determined in the laboratory. TML is a the increase of pore water pressure is inhibited during compaction.
parameter that is calculated as 0.9 times the FMP. The numerical value Liquefaction does not occur when the cargo contains very small particles;
of FMP must be determined by laboratory testing. If the moisture the cargo consists of large particles or lumps; the cargo contains low
content of cargoes exceeds the FMP, liquefaction may occur at any time moisture content. In these three cases, cargo movement can be restricted
during the voyage. Even the moisture content is below the FMP, and water will pass through the large porosity. When moisture content
liquefaction may also occur (TWG report, 2013). At the time of loading exceeds the Transportable Moisture Limit (TML), cargo may liquefy and
cargoes, compaction may occur, resulting in liquefaction or cargo shift. generate a dangerous wet base (IMSBC code). Therefore, TML is the
It is for these reasons that the IMSBC Code incorporates provisions estimation criterion for the carriage of solid bulk cargoes.

⁎
Corresponding author.
E-mail address: [email protected] (Y. Xue).

http://dx.doi.org/10.1016/j.oceaneng.2017.07.032
Received 29 March 2017; Received in revised form 16 June 2017; Accepted 10 July 2017
Available online 13 July 2017
0029-8018/ © 2017 Elsevier Ltd. All rights reserved.
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Nickel ore with a diameter below 10 mm with various moisture a discrete character along with the occurrence of liquefaction. Micro-
contents is shown in Fig. 1 (ClassNK, 2011). As the moisture content scale granular method based molecular dynamics can solve these
increases, the samples become increasingly muddy in appearance. In problems better (Spandonidis and Spyrou, 2013). The micro-scale
the Guidelines for the Safe Carriage of Nickel Ore published by modelling approach was developed to simulate the dynamic behavior of
ClassNK, it focused on how to test the TML and take precautions granular materials having physical properties and conforming to bulky
during the carriage of nickel ore. As introduced in the IMSBC code, ship cargos. This method predicted cargo shift and related phenomena.
cargoes with moisture content in excess of TML should not be loaded Container dimensions, the number of particles, and filling ratio
on the ship. Even cargoes with moisture content less than the TML variations are considered to study the surface profile and centre of
have the risk of liquefaction because of measurement error and mass. When experiencing the sway, heave and roll motion, the shift of a
atmospheric moisture migration. The work conducted by ClassNK cargo's centre of mass changes with time (Spandonidis and Spyrou,
has combined laboratory measurements and numerical simulations 2016). The Discrete Element Method (DEM) has been applied to solve
aimed at developing a set of operational guidelines pertaining to the ship related problems in recent years. For example, the interaction
stability and strength of carrier ships. between drifting pancake ice floes and moored ships have been
The Iron Ore Technical Working Group (TWG, 2013) has expended simulated with the DEM. The pancake ice floes are modelled with
considerable effort on the evaluation and verification of liquefaction three-dimensional discrete elements by considering the buoyancy, drag
and considered the adequacy of current methods for determining TML. force and added mass of current. The ship hull is constructed with
TWG, comprising one Brazilian and two Australian mining companies, combined discrete elements via certain overlaps. The interaction
has undertaken a program of research on representative Australian and between ice floe and ship hull is determined through element contact
Brazilian Iron Ore Fines (IOF) materials to assess their behaviors when detection and impact force calculation (Ji, 2013).
seaborne transported in bulk, with emphasis on liquefaction aspects. Two simulation methods are discussed below to investigate the
However, the research appears to focus on cargo liquefaction based on behavior of granular particles.
the traditional criteria of earthquake liquefaction, which may not apply
to the transportation problem of ship bulk cargo. More importantly, the
1.1. Soft sphere molecular dynamics (MD = DEM)
research separated cargo shift from ship stability, with all the ensuing
limitations, i.e., emphasis correctly put on avoiding the hazard but
The DEM was originally developed by Cundall in 1971 to address
without due consideration to the impact of its occurrence (not a “fail-
problems in rock mechanics. Essentially, MD and DEM are identical in
safe” system), which is an uncharacteristic approach when applied to
terms of particle shape and both methods are based on Newton's
safety-critical systems (loss of human life at stake).
second law. MD studies the physical movement of atoms and mole-
Cargo shift will influence the position of the centre of gravity of the
cules, which is applied today mostly in chemical physics, material
ship. As the Fig. 2(a) shows, the weight (W) of the ship acts downward
science and the modelling of biomolecules. MD can simulate a wide
through the centre of gravity of the ship (G). This force is resisted by an
variety of granular flow in static and dynamic conditions. However, the
upward buoyant force equal to W through the centre of buoyancy (B)
maximum number of particles is limited by computation ability. DEM
which is the geometric centre of the submerged volume displaced by
considers the rotation movement, particle shape, and distribution of
the ship. At stable state, B and G keep in line along the vertical
particles. Three codes based on MD (DEM) will be introduced as below.
direction. With an external force on the ship (from wind or waves), the
Code developed at National Technical University of Athens, pre-
ship will result in a large heel. This will result in B to shift while G will
sented some distinctive features: 1) The code can solve the coupled
not move. This variation will cause a righting moment (W*GZ) and the
roll-granular cargo problem formulation to the ship's body. Two
ship will oscillate from one side to the other side. If transverse cargo
problems have been tackled simultaneously: prediction of the wave-
shift occurs, it is possible for B and G to reverse position, as shown in
induced ship motion and prediction of the motion of a granular particle
Fig. 2(b) which can result in the ship capsizing quickly (Committee on
in the hold; 2) Two interaction forces have been analysed: dry granular
Ship's Ballast Operations, 1996). A ship's stability is often expressed by
solids and granular solids with humidity. Particularly for the particles
the ‘stability curve’ (Fig. 3), which provides information about the
with humidity, it brings dry granular solids to wet granular solids,
righting lever (GZ) for any angle of heel (based on an assumed centre of
which are much closer to the reality of unsaturated cargoes
gravity). Some information can be obtained by this curve: 1) the range
(Spandonidis and Spyrou, 2013; Spandonidis, 2016).
of stability; 2) the angle of vanishing stability; 3) the maximum GZ, and
LAMMPS: LAMMPS is a classical molecular dynamics code which
4) the initial GM. It also compares the decrease of the righting level
can be used to model atoms. However, due to the limitation of no pair-
when the cargo liquefies (Koromila et al., 2013), which shows that
style capability between particles, LAMMPS is not suitable to model the
when cargo liquefies, the ship stability will decrease greatly.
cargo shift or its interaction with ship movement. A recently distributed
Cargoes at risk of liquefaction include those with moisture, air, and
package, called LIGGGHTS, extents LAMMPS in order to handle
solid skeletons. Before loading, the cargoes may look dry, however, the
granular material and allows CFD-DEM coupling, which is an open
ship motion will compact the cargoes and rearrange, resulting in the
source high-performance scientific computing in fluid mechanics and
loss of water holding ability. The pore water pressure between particles
particle science by CFDEM research GmbH.1 The motion of the
will increase and the shear strength will decrease (see Fig. 4). The water
particles is resolved with DEM, and the CFD method is used to
climbs to the top of the free surface resulting to combined sloshing.
calculate the fluid flow. The granular phase which occupies a certain
Sometimes, the cargo sits on top of a liquid layer causing major
volume in each computational cell is accounted for by introducing a
displacement during ship motion, which decreases the ship stability
"volume fraction" into the Navier-Stokes equations. Furthermore, the
(see Fig. 5).
granular and fluid phase can exchange momentum, heat, and mass.
Liquefied cargo will behave as a non-Newtonian fluid which flows to
Some typical applications for the CFD-DEM method are the simulation
one side with roll but cannot flow to the other side in a timely manner
such as fluidized beds, metallurgical processes (e.g. blast furnace),
(Ju et al., 2013). The accurate predictions of highly nonlinear free-
pneumatic conveying chemical and pharamceutical reactors any kind of
surface flows and the impact by fluid are very important, especially in
granular flow influenced by the interstitial fluid phase (Goniva et al.,
waves. Many researchers have conducted work on the interaction
2012). However, the CFD-DEM method at present is based on a
between water flow and ship motion in waves (Kim, 2001; Toumi
saturated state, while ship cargoes contain three phases: water, air and
et al., 2009; Lee et al., 2010; Zhao and Hu, 2012; Andrei and Hanzu,
2013; Luo et al., 2016). However, there are some limitations based on
1
the continuum methods. Cargo shift with nickel ore or iron ore reflects http://www.cfdem.com/.

349
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 1. Appearance of nickel ore samples with various moisture contents (Indonesian Product).

Fig. 2. (a) Stable ship inclining and (b) GZ become negative, causing capsize (schematic diagram) (Rose, 2014).

Fig. 4. The granular particles: Left: before liquefaction; Right: after liquefaction (Rose,
2014).

soil, which reflect an unsaturated state. Thus, when considering the

unsaturated state, the soil suction should be included into the govern-
ing equations.
Fig. 3. Key features of a GZ stability curve. Before liquefaction (solid line) and after PFC: Particle Flow Code (PFC) is advanced multi-physics simula-
liquefaction (dashed line) (Koromila et al., 2013). tion software based on the DEM. The particles are rigid bodies using

350
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 5. Illustration of liquefaction phenomenon and ensuing effects (Rose, 2014).

Table 1
Constants of friction coefficient VS MC.

Moisture Friction Rolling Moisture Friction Rolling

Content Coef. Friction Content Coef. Friction
Coef. Coef.

0.49 0.005 0.02 0.37 0.3 0.4

0.44 0.01 0.02 0.36 0.7 0.5
0.42 0.08 0.02 0.35 0.7 0.6
0.4 0.1 0.02 0.34 0.8 0.6
0.39 0.2 0.05 0.29 0.8 0.6
0.38 0.3 0.1 0.24 0.9 0.6

Fig. 6. Algorithm for solution of coupled calculation.

Fig. 7. Appearance of nickel ore sample (ClassNK, 2011) and granular particles inclined at 50° (Blue color reflects that displacement of the particle is small during incling) (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

the soft contact approach, where the particles are allowed to overlap at solution to Newton's laws of motion. One of the assumptions is that
contact points. Particles interact with each other at pair-wish contacts particles are treated as rigid bodies. However, elastic-plastic strain of
by means of an internal force and moment, which provides a dynamic the cargo occurs during compaction.

351
L. Ju et al. Ocean Engineering 142 (2017) 348–362

do not completely return the other way. The ship progressively reaches
a dangerous heel. The particles interact with each other and also, with
the ship hold in waves in a comprehensive manner, which is a cargo-
sea-ship problem and cannot be investigated separately. With this in
mind, a simulation approach based on an adapter that couples a DEM
code and a CFD code is applied to identify the key parameters that
trigger cargo and ship instabilities. This approach will provide a better
Fig. 8. 2D Numerical tank model.
understanding of the coupled interaction between solid bulk cargoes
and ship vessels, which could possibly support a suitable regulatory
1.2. Event-driven (ED) simulation framework for assessing the ship stability due to cargo shift so that
decision making during the design, loading and transport could be
ED is conceptually different from DEM since collisions are dealt effective to avoid the ship capsize and casualties.
with using a collision matrix that determines the momentum change on
physical grounds. ED is also called hard sphere modelling. Particles are 2. Numerical models
assumed to be rigid and follow a disturbed motion until a collision
occurs (Lubachevsky, 1991; Miller and Luding, 2004a, 2004b). It is Prediction for ship response in waves due to cargo shift needs to
purely algebraic calculations. However, the instantaneous nature of address two problems simultaneously including 1) Particle movement
hard sphere collisions is artificial, and it is a valid limit in many inside the hold induced by ship motion; 2) Ship response in waves.
circumstances. Even though details of the contact or collision behavior Particle Flow Code (PFC) is advanced and multi-physics simulation
of two particles are ignored, the hard sphere model is valid when binary software based on DEM. Particles interact with each other at pair-wish
collisions dominate and multi-particle contacts are rare. The lack of contacts by means of an internal force and moment, which provides a
physical information in the model allows a much simpler treatment of dynamic solution to Newton's laws of motion. The particle-interaction
collisions (Luding, 2008). laws are referred to as contact models. Rolling Resistance Linear Model
As a matter of fact, solid bulk cargoes can flow like fluids, but they in PFC code incorporates the rolling torque to control the roll motion of
cannot behave like water and be described by Navier-Stokes equation. the particle, and the internal moment is incremented linearly with the
For the research on the cargo shift of solid bulk cargoes considering the accumulated relative rotation of the contacting pieces at the contact
interaction with ship response that concerns us, the behavior of point, which can effectively simulate the angle of repose commonly
granular cargoes transported by sea has not been sufficiently investi- found in granular materials. Rolling Resistance Linear Model provides
gated yet. Solid bulk cargoes in nature contain certain moisture more a force-displacement law relating the generalized internal force to the
or less, although the moisture in solid bulk cargoes is not visibly wet in relative motion at the contact. The internal force and moment (Fc and
appearance, but nevertheless is the main factor causing cargo shift. Mc ) act at the contact location in an equal and opposite sense on the
Under the requirement of International Maritime Bulk Solid Cargoes two pieces. The relative displacement increment of each particle during
(IMSBC) code, the moisture content of the solid bulk cargoes shall be a time step is given according to the force-displacement law, which
kept less than its Transportable Moisture Limit (TML) in advance of consists of the following steps (for details see PFC manual):
loading. As moisture content increases, the solid bulk cargoes show
muddy characteristics progressively, which is more suspected to cause 1) Update the linear normal force and the linear shear force;
cargo shift. Cargo shift is subjected to regular and irregular wave loads 2) Update the dashpot normal force and the dashpot shear force;
incurring on significant transverse and rotational accelerations. Such a 3) Update the rolling resistance moment.
movement during ocean carriage can cause a sudden loss of stability of
the carrying vessel. In the ClassNK magazine 64th edition, it read that Fc = F l +F d , Mc = Mr
from the end of 2010 to the end of 2011, for a period of approximately Both forces are resolved into normal and shear forces:
a year, four ships capsized and sank due to a loss of hull stability
attributed to cargo shift. The International Association of Dry Cargo F l = −Fnl+Fsl F d = −Fnd +Fsd Fsl = −μFnl
Shipowners (INTERCARGO) has recently published the “Guide for the
The rolling resistance moment is expressed by:
Safe Loading of Nickel Ore”, which aims to explain, through use of a
flowchart, how nickel ore can be safely shipped, within limitations, Mr = μr R Fnl
whilst raising awareness of the serious problem of cargo liquefaction. It
is targeted at the widest possible audience within the industry,
F l : the linear force
including shippers, shipowners and masters (INTERCARGO, 2012).
F d : the dashpot force
Liquefaction is a phenomenon in which a soil-like material is abruptly
Mr : the rolling resistance moment
transformed from a solid dry state to an almost fluid state. If
R : the contact effective radius
liquefaction occurs on board a vessel, the stability will be reduced
μ: the friction coefficient
due to the free surface effect and cargo shift, possibly resulting in
μr : the rolling resistance coefficient
capsizing of the vessel. The ship structure may also be damaged due to
increased cargo pressures (INTERCARGO, 2017).
To predict the ship response in waves due to cargo shift, it is a key
In particular, the solid bulk cargoes flow to one side of the ship but
factor to build a numerical tank to determine the wave generation and

Fig. 9. Total mesh layout and Enlarged view of mesh around paddle.

352
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Table 2 future work. In fact, the rotation axis changes dynamically due to the
Parameters for making linear wave. cargo shift. Also, the moment of inertia varies continuously with cargo
shift since the mass distribution alters. However, ANSYS Fluent
Wave length Wave height Depth of paddle Depth of numerical
(unit: m) (unit: m) under water (unit: tank (unit: m) software cannot change the rotation axis in the calculations.
m) Therefore, as the simplified approach with fixing the rolling axis to
the centre of mass, the moment of inertia I is the moment of the
8m 0.4 m 1.6 10 ⎯→
⎯
container itself and the moment M is deriving from the sum of all
corresponding particles at a fixed rolling centre, which is considered as
the external load. The vessel's roll motion can be described as follows:
→
dω ⎯→
⎯
I +→
ω × (I ⋅→
ω) = →
τ + M
dt

→
ω : the angular velocity of the body
→
τ : the resultant torque acting on the body
Fig. 10. Appearance of wave surface. I : the moment of inertia
⎯→
⎯
M : the external torque due to cargo shift
predict the response of bodies in waves. The regular wave is generated
using the paddle wave generating method based on ANSYS Fluent
software. Navier-Stokes equation, Volume of Fluid (VOF) method and 3. Contact properties
dynamic meshing method are solved to establish a two-dimensional
wave tank with a wave-absorbing method employed to absorb the wave The moisture content of a cargo is vitally important to transporta-
energy. An adapter coupling ANSYS Fluent software and PFC software tion safety. Solid bulk cargo, such as iron ore or nickel ore, can behave
is developed to simulate the interaction between cargo movement and like liquids if subject to moisture is at a certain level in the cargo. As a
ship motion, which allows transferring data during the coupled matter of fact, the cargo in nature contains certain moisture more or
calculations without any changes in the source of the simulation code. less and it is the moisture that leads to cargo shift, which results in a
The algorithm of the adapter is sketched in Fig. 6. Running a co- large angle of heel, although the moisture in the cargo is not visibly wet
simulation requires the following steps: in appearance. The cargo with certain moisture behaved like non-
Newtonian fluid (Ju, 2013) flows to one side with a roll but does not
1) Preparation of Model Files: Before starting a co-simulation, each completely return with a roll the other way, the ship progressively
domain must be prepared separately. For ANSYS Fluent software, reaches a dangerous heel. Coupled continuum-discrete model is widely
the mesh, turbulence model, materials, boundary condition and proposed and developed to simulate the mesoscale pore water flow and
patch domain for water phase must be prepared first. For PFC solid matrix deformation of saturated soil, by solving the Navier-Stokes
software, the preparation process includes the distribution of equation for fluid and linear and angular momentum for each particle
cargoes and the initialization of gravity and properties of each (Usama and Mourad, 2005). However, coupled continuum-discrete
particle. methods to date are based on the theory of saturated materials. Ship
2) Coupling Process: The ANSYS Fluent software calculates firstly to cargoes contain three phases: water, air and solid skeleton. ClassNK
obtain an angular velocity of the container or ship in waves, and
then sends it to the PFC software code. Thus, the corresponding Table 3
moment on the container due to cargo shift will be obtained. This Summary on particle size, time-step size and computing time (CFD).
moment will transfer to ANSYS Fluent, and so on until the
Mesh Overall cells Calculation time Time-step Simulation time (1
appearance of wanted data. size processor)

In the paper, only the roll motion of the 2D rectangular hold is Coarse 47,144 40 s 0.001 s 5.4 h
Medium 72,694 0.001 s 6.6 h
considered as ship capsize happens frequently along with a large angle
Fine 111,788 0.001 s 9.3 h
of heel of the ship vessels significantly. In addition, the 6DOF motion Coarse 47,144 0.0008 s 6.1 h
also will influence the compaction and the moisture content of bulk Coarse 47,144 0.0005 s 14.6 h
cargo, further the bulk cargo shift, which will be investigated in the

Fig. 11. Time history curves of wave free surface (x = 16 m, 24 m).

353
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 12. Time history curves of wave free surface (x = 24 m). (a) Mesh number. (b) Time-step size.

agreement with nickel ore flow. Table 1 shows the correspondent

friction coefficient between particles at various MC. From Table 1, it
can be seen that when the moisture content is close to 0.36–0.37, the
friction coefficient has the abrupt change from 0.7 to 0.3. The
possible reason is that when the friction coefficient is larger than
0.3, the friction can resist the flow of the particles at the inclined
angle of 50 degrees. If the friction coefficient reaches 0.3, the
particles begins to flow because of the angle of repose. The angle
of repose of granular material is the steepest angle of descent to the
horizontal plane to which a material can be piled without slumping.
It is a basic property of non-cohesive granular materials: the
maximum slope angle at which the material is at rest. On account
of the assumption above, DEM can effectively simulate the flow
characteristic of the moist bulk cargo.

4. Numerical validation
Fig. 13. Initial appearance of cargoes for sway motion (The color blue reflects that the
displacement is small) (For interpretation of the references to color in this figure legend,
Numerical validation includes two parts: CFD numerical validation
the reader is referred to the web version of this article.).
and DEM numerical validation. CFD numerical validation focuses on
the sensitivities of grid size and time-step size based on a making wave
Table 4
test. DEM numerical validation emphasizes the sensitivities of particle
Summary on particle size, time-step size and computing time (DEM).
numbers and time-step size based on a sway oscillation test. In
Radius Particle Calculation time Time-step Simulation time (1 addition, coupled motion in waves based on the CFD-DEM adapter is
number size processor) investigated to identify the sensitivities of particle numbers.
0.01 m 256 40 s 0.0001 s 8 min
0.005 m 1024 40 s 0.0001 s 20 min 4.1. CFD Numerical validation
0.003 m 2844 40 s 0.0001 s 68 min
0.005 m 1024 40 s 0.0005 s 18 min
As shown in Fig. 8, the numerical tank model includes three zones:
0.005 m 1024 40 s 0.001 s 15 min
the wave making zone, working zone, and absorption zone. The length
of the wave absorption zone should be at least one times the desired
conducted the laboratory experiment that the behavior of the nickel ore wave length in order to eliminate the effect from a reflected wave. The
stored in the container was observed by inclining the container boundary AB is set as a ‘pressure outlet’ and the others are set as
statically up to the maximum inclined angle of 50° (Fig. 7). At low ‘walls’, including the paddle OP. Fig. 9 shows the total mesh layout of
moisture content (MC), nickel ore shows great inertia. Along with the the numerical tank and enlarged view of the mesh around the paddle.
increasing of MC, nickel ore begins to show the characters of fluid. The dynamic mesh technique is used around the paddle to adapt the
Therefore, MC is the key factor which cannot be neglected in the deformation of the mesh with triangle grids.
simulations. According to the plane standing wave theory and the parameters
Since methods to date have the limitation to address the unsatu- shown in Table 2, the motion of paddle in this simulation is as follows:
rated materials, an assumption in the paper is specified that, MC of the θ(t) = 0.1451sin (2.7743t )
cargo is described by correspondent friction coefficient. The static
incline test is simulated by PFC2d software to identify the contact θ̇ (t) = 2.7743*0.1451cos (2.7743t )
properties of nickel ore at various MC (Fig. 7). With the quantitative
This equation can be read by using the User Defined Function in
comparison of numerical results with experimental data, by changing
Fluent software to control the motion of paddle based on dynamic
the friction coefficient of the particle, the state of the cargo inside the
mesh method. The regular wave is presented in Fig. 10, and it can be
container is compared with the experimental data from ClassNK to
seen clearly that the wave length is 8 m. Based on the wave absorption
identify the relation between the friction coefficient and MC.
theory, from x = 50 m to x = 58 m, the damping effect is obvious and
It can be seen that the granular particle flow shows a good
the wave energy is clearly removed. As we can see from Fig. 11, the

354
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 14. Fluctuation force and gravity centre of cargoes varying from particle number (Force_X and Force_Y means the horizontal and vertical direction force on the container).

wave surface at the probe is similar to the analytical results. The 4.2. DEM numerical validation
numerical waves are in phase with the analytical wave profiles. When
waves propagate gradually away from the wave paddle, the wave A prescribed sway oscillation test is conducted to identify the
elevation demonstrates a higher order effect. Compared with the first sensitivities of the particle size and time-step size in PFC software.
order analytical results, the CFD simulated wave has a higher wave Sway oscillations of a 2D rectangular structure (50 cm*50 cm) with
crest and a lower wave trough. Based on this method, the regular wave amplitude of 0.3 m and frequency of 0.5HZ are investigated to identify
is generated to provide the wave condition for the coupled calculation the interaction between cargo shift and container motion (see Fig. 13).
of the response of a container due to cargo shift. Details of overall particle size and time-step size are listed in
Based on the validation studies of making waves above, additional Table 4. The same mass of cargo with 201.06 kg is applied in the
validation work with special emphasis on the sensitivities of grid size simulations and the density of the particle is 2500 kg/ m3. The time
and time-step size is conducted. For the grid dependence study, a time history curves of the fluctuation force on the container and the gravity
step of 0.001 s and three meshes are used. Special attention to the centre in Figs. 14 and 15 indicates that the results on different particle
mesh near the free surface in the working zone is considered. It takes numbers and different time-step sizes show little difference, which
5.4, 6.6 and 9.3 CPU hours to finish 40 s simulation time on coarse, demonstrates that PFC software shows good computational stability.
medium and fine grids, respectively (Table 3). The comparisons of The measuring process of the fluctuation force is to identify the
numerical results are shown in Fig. 12a. It is apparent that the results, contact particles with container wall firstly, then calculate the global
as calculated using different grids, are in good agreement. For the time- contact force as shown above according to the contact models as stated
step size dependence study, three steps with coarse mesh are analysed. in PFC help manual. During the rolling motion, the porosity of the
It takes 5.4, 6.1 and 14.6 CPU hours to finish 40 s simulation time with cargo decreases, resulting in a slight motion in vertical direction. The
large, medium and small time-step, respectively. The comparison of fluctuation force on the container and the gravity centre of cargoes with
computed results is shown in Fig. 12b. It is apparent that the results of the particle numbers of 2844 and the time-step size of 0.0001 s during
the time history curves of wave free surface (x = 24), computed with the the oscillation test almost coincide with those with the particle
three time steps, are nearly identical. As to the special emphasis on the numbers of 256 and the time-step size of 0.001 s. Considering an
mesh near to the free surface, the mesh close to the paddle needs to be increased computing time, as list in Table 4, the time-step size of
adjusted and refined, resulting in an increase in simulation time from 0.0001 s is used in the following simulations and the particle size is
coarse mesh to fine mesh. As to the decrease in time-step size, User dependent on the situation.
Defined File needs more time to update the motion of the paddle,
resulting in an increased simulation time between the time-step of 4.3. Coupled motion test in regular waves
0.0008 s and the time-step of 0.0005 s. Considering an increased
computing time, as listed in Table 3, the coarse grid and the time- A coupled motion test is conducted to identify the sensitivities of
step size of 0.001 s is used in the following simulations. particle size in PFC software. The rectangular structure with dimen-

355
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 15. Fluctuation force and gravity centre of cargoes varying from time-step size.

actual mass of the structure is 0.236 kg m2. Details of overall particle

size are listed in Table 5. The same mass of the cargo with 14.988 kg is
applied in the simulations. The density of particles is 1642.865 kg/m3
and the MC is 49%. The gravity centre is set at the location of (0 m,
0 m). The structure is free to respond only in the roll direction under an
external excitation owed to regular waves, with a frequency of
5.932 rad/s and wave length of 1.75 m.
The adapter coupling ANSYS Fluent software and PFC software is
used to simulate the interaction between cargo movement and ship
motion, which allows for the transferring of data during the coupled
calculations without any changes in the source of the simulation code.
The time history curves of the roll angle of the container and the gravity
centre of the cargo in Fig. 17 indicates that results on different particle
numbers show little difference, though they have slightly different
cargo height and gravity centre at the initial stage. It takes 14.6, 15 and
23 CPU hours to finish 35 s simulation time on coarse, medium and
Fig. 16. 2D rectangular structure. fine particle numbers, respectively. To maintain the same mass of the
cargo, the initial gravity centre of the cargo is different, resulting in
Table 5 minor variations for the roll angle of the container. The selection of the
Summary on particle size, time-step size and computing time (CFD-DEM). particle size depends on the specific situation.
Radius Particle Calculation time Time-step Simulation time (1
number size processors) 5. Cargo shift

0.0025 m 464 35 s 0.0001 s 14.6 h 5.1. Static incline test

0.002 m 726 35 s 0.0001 s 15 h
0.0015 1291 35 s 0.0001 s 23 h
A 2D rectangular structure (100 cm*50 cm) is used for the static incline
test. At the initial state, the cargo has uniform moisture content of MC =
24%, 29%, and 39%, separately (Fig. 18a) and mixed moisture content
sions of 0.3 m long and 0.3 m high, partially filled with moist bulk which the MC at the bottom is set to a value of 49%, while the MC above is
cargo is used in the simulations (Fig. 16). The structure is floated at a set to a value of 24%, 29% and 39%, separately (Fig. 18b). The container
draft that equals one-sixth of its height. The moment of inertia of the statically inclines at 50° and then keeps stable to investigate the cargo shift.

356
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 17. Roll angle and gravity centre of the cargo varying from particle number.

Fig. 18. Cargoes with uniform MC (left) and mixed MC (right). Blue color reflects the small displacement (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.).

It is observed from Fig. 19 that the sample behavior shows various While cargoes mineral ores loaded on board a vessel, the granular
flow characteristics by monitoring the coordinates of the gravity centre materiel is exposed to mechanical agitation and energy input in the form
at various initial MCs. What is worth mentioning is that MCs of the of engine vibrations, vessel movement, and wave impact, resulting in a
sample are expressed in the form of correspondent friction coefficient gradual settling and compaction of the cargo. When some cargoes lost
based on the assumption above. However, the assumption neglects the water holding ability, the wet base at the bottom will be generated. Then,
transfer of moisture content during compaction. The cases only consider the water will flow up to shape the free surface. During this process,
the influence on the ship vessel motion due to cargo flow with a given some cargoes keep dry which is not subjected to compaction, and some
moisture content, which may decrease the accuracy of the numerical become muddy state like non-Newtonian fluid. This will be a compli-
results. It is beneficial to bring in the concept of matric suction in soil cated state, which is difficult to simulated based on current numerical
mechanics to predict the transfer of moisture content precisely. General methods. The likelihood of formation of a ‘wet base’ is a significant factor
speaking, with moisture content increasing, the sample comes close to to be considered within the overall assessment and understanding of
liquid and tends to flow. However, from the results, the x coordinate of liquefaction and cargo stability. Some cargoes are susceptible to
the gravity centre of the cargo with a MC of 29% is larger than that with a moisture migration and may develop a dangerous wet base even if the
MC of 39%, which shows that the particle flow shows complicated average moisture content is less than the TML. Therefore, cargoes at the
nonlinear transition due to the interaction between particles. bottom could be regarded as the wet base with given friction coefficient

357
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 19. Coordinates of gravity centre of cargoes with uniform MC in static incline test.

Fig. 20. Coordinates of gravity centre of cargoes with mixed MC in static incline test (Bottom: MC = 49%, Top: MC = 24%, 29%, 39%).

Fig. 21. Iron ore piles before and after a sea state test (TWG report, 2013).

(Fig. 18b). As shown in Fig. 20, it can be concluded that when the wet decrease the ship stability greatly. The area of the wet base is determined
base is generated, flow characteristic of the cargoes is more evident and by a range of parameters including the cargo holding water ability, cargo
the cargo flows to one side of the container with high probability. deformation model, the void ratio, cargo stow formation, initial bulk
Comparison with Fig. 19, the coordinate of the gravity centre of the density and initial moisture content. The Hexapod tests conducted by
cargoes is 0.22 m along x-axis and −0.1 m along y-axis if no wet base. consulting firm MARIN of Australia (TWG, 2013) as shown in Fig. 21,
Whereas the coordinate of the gravity centre of cargoes changes to which showed that at high moisture contents, cracks will occur in the
0.37 m along the x-axis and −0.15 m along the y-axis after static incline cargo surface. Free surface apparent in the corners of the ship hold can
test if the wet base happened. That is, the presence of a wet base in the be shaped after shaking and a considerable compaction has occurred
ship hold may result in a high flowability of the cargoes. In fact, the free from the initial state. The water flows out to the bottom to form the wet
surface can be observed in the bottom of holds of vessels during a base. The peak pile of cargoes turns to the flat pile, which is the typical
voyage. The presence of the wet base in the corners of the hold does process during liquefaction as described above.

358
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Table 6
Parametric investigation.

ω (rad/s) H (m) MC

Case 1 6.75 0.02 37%

40%
49%
Case 2 7.85 0.02 49%
6.75
5.93

in this region decreases as to the friction coefficient decreases, causing

cargo collapse (Fig. 23). When the container returns to the initial position
again at the 8.5 s, the coordinate of the centre of gravity migrates to one
side and the position of the gravity centre of the container has been
greatly shifted. It seems to be not obvious for the migration of gravity
centre in the simulation of container 1 (50 cm*35 cm). When the
Fig. 22. Motion of the container. container, such as container 2 (100 cm*35 cm) with the same number
of particles, has enough space to flow, particles tend to flow to one side
5.2. Cargo collapse and stay stable during the collapse. Consequently, the gravity centre of
the container will be shifted greatly which reduce ship stability (Fig. 24).
Cargo collapse is investigated in this section to discuss a probability
causing ship capsize. The angular velocity of the container is shown in 6. Parametric investigation
Fig. 22 in one cycle. When the container rotates to a maximum degree at
the 6 s, cargo collapse is happened by decrease the friction coefficient to A parametric investigation is undertaken to identify the key
investigate a possibility of the cargo shift. The shear strength of particles parameters that trigger cargo and ship instabilities. On the XY plane

Fig. 23. Comparison before and after cargo collapse (container 1: left; container 2: right. particles: 1024).

Fig. 24. X, Y coordinate of gravity centre of the cargoes during collapse: container 1(left); container 2 (right).

359
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 25. State of the cargo inside the hold during the coupled motion.

appears 726 particles with the density of 1642.865 kg/m3 and the low moisture content (Fig. 26a). From the X, Y coordinate of the
radius of 0.002 m. Two cases are carried out for the purpose of mass centre of the cargo in Fig. 27a, b, the cargo shows no cargo
determining the effect of shifting cargo on the structure. The first is shifting for the MC of 37% and 40%. While for the MC of 49%, due
to change the moisture content of the cargo with the wave amplitude to the low friction coefficient between particles, the cargo shows
fixed at 0.02 m and the frequency at 6.75 rad/s (Case 1). The second is large flowability. The moment due to cargo shift on the structure
to change the wave frequency with moisture content fixed at 49% and affects the roll response of the structure.
amplitude at 0.02 m (Case 2), as shown in Table 6. b) Case 2: At the wave frequency of 6.75 rad/s, the peak roll angle is
Fig. 25 shows the state of the cargo inside the structure in waves. 17.5°. While at the wave frequency of 7.85 rad/s and 5.93 rad/s, the
The key findings are as follows: peak roll angles are 5° and 17.6°. It can be explained that at the
wave frequency of 7.85 rad/s, the moving cargo in the hold opposes
a) Case 1: The maximum roll angle of the structure is approximately the external excitation.
18° at the moisture content of 49%, which shows that the cargo has c) Therefore, for the wave frequency above or below the resonant
a high probability tending to flow with higher initial MC of the frequency, the response of the structure is not obvious (Fig. 26b).
cargoes. The development of the roll angle at the MC of 37% and the Especially for the wave frequency values of 5.93 rad/s, the peak roll
MC of 40% is almost the same. It means that the roll motion of the angle is no more than 5°. From Fig. 27c, d, at the wave frequency of
structure is not affected by the displacement of the cargo containing 7.85 rad/s, the phase of cargo's displacement opposes that of the

360
L. Ju et al. Ocean Engineering 142 (2017) 348–362

Fig. 26. Comparizon of the response of the structure in waves. The wave apmlitude is fixed at 0.02 m, the frequency is fixed at 6.75 rad/s and the MC is varied as 37%, 40% and 49%
(left). The MC is fixed at 49%, the wave amplitude is fixed at 0.02 m and the wave frequency is varied as 7.85 rad/s, 6.75 rad/s and 5.93 rad/s (right).

Fig. 27. Displacement of mass centre of the cargo. The wave apmlitude is fixed at 0.02 m, the frequency is fixed at 6.753 rad/s and the MC is varied as 37%, 40% and 49% (a, b). The MC
is fixed at 49%, the wave amplitude is fixed at 0.02 m and the wave frequency is varied as 7.85 rad/s, 6.75 rad/s and 5.93 rad/s (c, d).

structure, and thus the overall motion is reduced. e) The roll angle of the structure in waves in Fig. 26 shows some
d) For the MC of 49% and the resonant frequency, the cargo has symmetry. However, we would expect some asymmetry to show the
shown the shifting phenomenon. The X coordinate of the mass interaction between cargo shift and ship vessel response. The result
centre of the cargo fluctuates around the balance position. could be explained as: 1) The cargo keeps the same moisture
However, due to the limited space of the structure along the content and flowing capacity that the horizontal gravity centre does
transverse direction and the small roll angle, the displacement of not change a lot, even with the cargo collapse; 2) The 2D structure
the cargo has been restricted slightly. limits the flow of the cargo.

361
L. Ju et al. Ocean Engineering 142 (2017) 348–362

7. Conclusions ClassNK, 2011. Guidelines for the Safe Carriage of Nickel Ore.
Committee on Ship’s Ballast Operations, 1996. ClassNK magazine 64th edition.
Gard, 2014. Liquefaction of solid bulk cargoes.
Rolling response of a 2D rectangular hold, partially filled with moist Goniva, C., Kloss, C., Deen, N.G., Kuipers, J.A.M., Pirker, S., 2012. Influence of rolling
bulk cargo in regular wave was described to address a solid bulk cargo- friction modelling on single spout fluidized bed simulations. Particuology 10,
582–591.
sea-ship problem. Due to the limitations to simulate the unsaturated INTERCAEGO, 2017. report.
materials based on presented methods, an assumption in the paper was INTERCARGO, 2012. report.
presented. That is, the moisture content of the cargo is described by International Maritime Organization.International Maritime Solid Bulk Cargoes
(IMSBC) Code.
correspondent friction coefficient and rolling friction coefficient. With Ji, S., 2013. Analysis of interaction between ice floe and ship hull with discrete element
the quantitative comparison of numerical results with experimental method in broken-ice field. Chin. J. Appl. Mech. 30 (4), 520–526.
data from ClassNK, this assumption was evaluated and justified. Ju L., Vassalos D., Jin W., 2013. Rolling motion of Newtonian and non-Newtonian fluids
in a ship hold. 5th International Maritime Conference on Design for Safety.
An adapter was developed to couple the DEM code and CFD code to
Ju, L., Vassalos, D., Boulougouris, E., 2016. Numerical assessment of cargo liquefaction
study the rolling response of the 2D rectangular hold in regular wave. potential. Ocean Eng. 120, 383–388.
Numerical validation, including CFD numerical validation and DEM Kim, Y., 2001. Coupled Analysis of Ship Motions and Sloshing Flows. 16th International
numerical validation was conducted to identify the sensitivities of grid Workshop on Water Waves and Floating Bodies, 21, 81–84.
Koromila, I.A., Spandonidis, C.C., Spyrou, K.J., 2013. Experimental investigation of
size, time-step size and particle numbers, which demonstrates that PFC cargo liquefaction and impact on the stability of a bulk – carrier. Proc. 13th Int. Ship
software shows good computational stability. Different moisture con- Stability Workshop.
tents of the cargo and different wave conditions were applied on the 2D Lee B.H., Park J.C., Kim M.H., 2010. Two-dimensional vessel-motion/liquid-sloshing
interactions and impact loads by using a particle method. ASME 2010, 29th
rectangular hold to investigate the cargo shift problem. Special atten- International Conference on Ocean, Offshore and Arctic Engineering, 6, 679–686.
tion was paid to the comparison of the roll angle of the hold and the Lubachevsky, B.D., 1991. How to simulate Billards and similar systems. J. Comp. Phys.
coordinate of the mass centre of the cargo in the hold. The results 94 (2), 255.
Luding, S., 2008. Introduction to discrete element methods -basics of contact force
indicated that, the cargo containing high moisture content has low models and how to perform the micro-macro transition to continuum theory.
friction coefficient, leading to the high risk of capsizing. Therefore, Discret. Model. Geomater., 785–826.
reducing the initial moisture content will enormously increase the Luo, M., Koh, C.G., Bai, W., 2016. A three-dimensional particle method for violent
sloshing under regular and irregular excitations. Ocean Eng. 120, 52–63.
stability of the ship. Based on the adapter developed, the assessment of
Miller, S., Luding, S., 2004a. Cluster growth in two- and three-dimensional granular
the ship stability can be investigated to identify the key parameters that gases. Phys. Rev. E 69 (031305), 1–8.
trigger ship instabilities, including the cargo properties, wave energy Miller, S., Luding, S., 2004b. Event driven simulations in parallel. J. Comp. Phys. 193 (1),
306–316.
and the structure, distribution of ship holds. The method will provide a
PFC help manual.
better understanding of the shifting phenomenon of solid bulk cargo Rose, T.P., 2014. Solid Bulk Shipping: Cargo Shift, Liquefaction and the Transportable
that interacts with ship vessel motion, identification, and quantification Moisture Limit (Thesis of Master of Science). University of Oxford.
of the controlling parameters. Spandonidis, C.C., 2016. Modelling and Numerical/experimental Investigation of
Granular Cargo Shift in Maritime Transportation (Ph.D. Thesis). National Technical
University of Athens.
Acknowledgements Spandonidis, C.C., Spyrou, K.J., 2013. Micro-scale modeling of excited granular ship
cargos: a numerical approach. Ocean Eng. 74, 22–36.
Spandonidis, C.C., Spyrou, K.J., 2016. Coupled vessel-dry-granular-cargo roll dynamics
This work was supported by the National Natural Science in regular beam seas. Ocean Eng. 120, 238–245.
Foundation of China (NSFC) [Grant Nos. 51639004, 51579054]. Technical Working Group (TWG). Reference tests, 2013.
Toumi, M., Bouazara, M., Richard, M.J., 2009. Impact of liquid sloshing on the behavior
of vehicles carrying liquid cargo. Eur. J. Mech. A/Solids 28, 1026–1034.
References Usama, E.S., Mourad, Z., 2005. Coupled continuum-discrete model for saturated
granular soils. J. Eng. Mech. 131, 413–426.
Andrei, C., Hanzu, P.R., 2013. The impact of bulk cargoes liquefaction on ship's intact Zhao, X.Z., Hu, C.H., 2012. Numerical and experimental study on a 2-D floating body
stability. UPB Sci. Bull. Ser. B: Chem. Mater. Sci. 75, 47–58. under extreme wave conditions. Appl. Ocean Res. 35, 1–13.

362