1. Introduction
The ocean waves may not affect the sea bed due to energy dissipation owing to the
distance between the free surface and the sea bed. However, when the water depth is
shallow enough, or the energy of the waves is large enough (i.e. with a tsunami),
the sea bed can be disturbed. The behavior of the sea bed may affect environmental
and ocean structures, and thus, accurate estimation is important.
The breaking wave problems have been investigated theoretically, experimentally, and
numerically (1,2) considering soil movement induced by breaking wave energy. Large deformation of the
sea bed induced by tsunami waves was also studied (3). Despite numerous efforts to simulate sea bed motion due to waves, its complexity
has dramatically increased, and has thus been restricted to the conventional numerical
method.
With the conventional numerical method in the field of computational fluid dynamics,
the grid system following the Eulerian approach is the most representative method.
The Eulerian approach uses a grid as a control volume in fluid dynamics, and then
expects physical quantities by measuring the in-and-out flow through the control surface.
It is, indeed, a very efficient method to estimate fluid dynamics; however, its original
concept of a grid system may cause numerical errors, which is a concern for accuracy.
The particle method follows the Lagrangian approach, and thus, all physical quantities
can be calculated by tracing each particle. For the particle method, the Smoothed
Particle Hydrodynamics (SPH) and Moving Particle Semi-implicit (MPS) methods are well-known.
The SPH method was first announced in (4). At that time, the pressure on each particle was predicted by using the equation-of-state,
and thus, pressure fluctuation was the most common issue with SPH. In (5), the authors used a Poisson pressure equation (PPE) for pressure calculation following
the MPS method to suppress pressure fluctuation. The MPS method was originally proposed
for incompressible flow (6). At that time, the MPS method focused on the impact load from the sloshing problem
and the broken dam problem, and the PPE was adopted to predict loads induced by fluids.
However, there was still pressure disturbance, and thus, a multi-source term for the
PPE was investigated (7). Furthermore, the particle method can be achieved by tracking each particle, and
the free surface searching model was very important. The free-surface searching model
in (6) was modified in (8) by using multiple criteria to judge the free-surface particle. Another investigation
suppressed the pressure perturbation by using a corrective matrix on the PPE equation
(9).
The particle method has the advantage in a multi-phase flow. In (10), the authors studied multi-phase interaction models for the MPS method to reproduce
the three-phase sloshing problem. In (11), the authors extended the MPS method to solid particles with a newly adopted solid
particle interaction model to simulate the gravel dumping problem. In that investigation,
the angle of repose on dumped solid particles was observed. Furthermore, the MPS method
was coupled with the Direct Element Method (DEM) to simulate the sediment transport
problem (12).
In this study, the newly developed MPS method simulated the mobile sea bed–behavior
problem with large waves induced by a broken dam. The numerical results were compared
with corresponding experimental results, and then, various densities of soil particles
were investigated.
2. Moving Particle Semi-implicit Method
The MPS method is a computational fluid dynamics method following the fully Lagrangian
approach. Its governing equations are continuity and Navier-Stokes’ equations, as
follows:
where $\rho $ is density of the fluid, $t$ is time, $\overset{\rightarrow }{u}$ is
velocity of a particle, $P$ is pressure on the particle, $\nu $ is a dynamic viscosity
coefficient, and $\overset{\rightarrow }{g}$ is gravitational acceleration. Since
the MPS method follows the particle method, all differential equations should be replaced
by forms of the particle method, as follows:
where $\nabla $ is gradient, $\phi $ is arbitrary physical quantities, $d$ is
the number of dimensions, $n_{0}$ is the particle number density in the initial arrangement,
$w$ is a kernel function, $\lambda $ is an adjustment parameter, subscript $i$ denotes
the i-th particle, and $\phi _{ij}$ denotes $\phi _{j}-\phi _{i}$. The adjustment
parameter makes a value equal to an analytical solution, which can be calculated as
follows:
The particle number density and kernel function are key parameters of the MPS method
to keep the continuum of the system to avoid violating the laws of physics. The particle
number density uses fluid density, and the kernel function measures the effect of
neighboring particles on center particles. The particle number density and kernel
function can be expressed as follows:
According to Eq. (8), when the distance between center and neighboring particles exceeds the effective
range ($r_{e}$), the kernel function has a value of zero, which means neighboring
particles beyond the range cannot affect center particles. The effective range is
set to $2.1l_{0}$ from a numerical convergence test, and $l_{0}$ is the particle distance
in the initial arrangement.
For the solid particle problem, the viscosity term, which is the second term on the
right-hand side of Eq. (2), needs to be eliminated. And then, the friction and drag terms will be replaced by
a viscosity term. Thus, the governing equation of the MPS method for solid particles
will be expressed as follows:
where $\overset{\rightarrow }{F}_{f}$ is friction force and $\overset{\rightarrow
}{F}_{d}$ is drag force.
The friction term consists of two forces; the friction force from the relative acceleration
between the center and neighboring particles, and the accumulated weight from the
center and neighboring particles over the center particle. Fig. 1shows a schematic of the friction force calculation.
The friction force due to relative acceleration can be calculated as follows:
Fig. 1. Schematic for friction force.
Fig. 2. Schematic for accumulated weight.
Fig. 3. Schematic of the drag force calculation.
where $\mu $ is a friction coefficient, $\overset{\rightarrow }{a}_{n}$ is a normal
component of acceleration, and $\overset{\rightarrow }{t}$ is the unit tangential
vector of the friction force.
Another term in friction force is accumulated weight friction force, which can be
expressed as follows:
where $\theta _{ij}$ is the angle between the connecting line and the moving direction
of the center particle, and $p_{s,ij}$ is the accumulated weight of a stacked particle
(see Fig. 2). The accumulated stacked particle weight can be calculated as follows:
where
The friction force can be calculated as the summation of the two terms above:
For the calculation of drag force, the following equation was implemented:
where $C_{D}$ is the drag coefficient, and $v_{ij}$ is the relative velocity between
center and neighboring particles. In this case, the center particle should be a solid
particle, and the considered neighboring particles are fluid particles only (see Fig. 3).
3. Simulation Results
Sea-bed behavior under violent waves was simulated for this paper. A model is
shown in Fig. 4. Water was captured by a wall and the soil particles lay over the wall. A total of
144,000 particles were used for this simulation: 70,000 were water particles, and
52,000 were soil particles. The time interval for the simulation was 0.0001s, and
the total simulation time was set to 1.5s in real time. The density of water was $1,000kg/m^{3}$,
and the soil particles were $900kg/m^{3}$, the same as in the corresponding experiment.
When the wall moves upward at 0.5m/s, the captured water collapses spontaneously
from gravitational acceleration. The collapsed water collides with the soil particles
and disturbs them.
In the experiment, a polyethylene oxide (PEO) solution was used for the soil bed.
The reason for using the PEO solution in the experiment was because the soil bed can
be regarded as a liquid due to liquefaction effects. For this reason, fluid-soil and
solid-soil sea beds were both simulated and compared.
Fig. 4. A model of the numerical simulation.
Fig. 5. Comparison of snapshots from the experiment and the numerical simulation.
Fig. 6. Comparison of pressure on the liquid soil particles (left) and on the solid soil particles (right).
Fig. 5shows snapshots from the experiment and both of the numerical simulations of the fluid
soil bed and the solid soil bed over an arbitrary time period. When the gate opened,
the captured water column collapsed spontaneously from gravitational acceleration.
The collapsed water collided with the soil bed beneath the soil layer. The water scooped
up the soil bed, and the wave energy transferred into the soil particles. The disturbed
soil particles had the shape of a wave, and it propagated to the right end of the
tank. The soil bed wave also had a leading edge, the same as the water wave, creating
an overturning wave. The leading edge of the overturning wave made a rolling shape
that scooped up the soil bed layer.
Figs. 6 and 7 represent a detailed comparison between the numerical results from the MPS method
for fluid and solid particle velocity. In Fig. 6, the soil particle waves are compared. The contour of the particles denotes the pressure.
From this comparison, the pressure on the particles and the tendency of the generated
waves were not different. The velocity vector on the soil particles is visualized
in Fig. 7. As mentioned with Fig. 6, the difference was too small to specify; however, a comparison of velocities on
the particles shows a significant difference. The liquid soil particle movement has
the form of a splash, and thus, the velocity tendency looks like hydraulic jumping.
Another point with the liquid model is that the collision model in the algorithm affects
particle movement to add velocity. With the solid soil particles, the soil bed wave
was generated in a way similar to the liquid soil bed; however, the velocity tended
to have a lateral direction due to the pushing effects of the soil bed, and the collision
model did not influence the particles. In this regard, the velocity of solid particles
after the collision is smaller, compared to liquid particles.
Fig. 7. Comparison of velocity on the liquid soil particles (left) and the solid soil particles (right).
The wave-like soil bed movement was also observed in both the numerical simulation
for the liquid MPS method and the solid MPS method. However, a few differences were
observed. The prior point of difference was the shape of the wave edge of the soil
particle. With liquid soil, the soil particle looks like it is being pushed, but the
result from the solid soil bed was a carpet-shaped wave edge, and it was more similar
to the experiment. Another difference was the development of a soil bed wave. When
the overturned soil particle wave stroke the soil bed, the liquid particle was exaggerated,
compared to the solid particle. In the MPS method, there is a collision model that
is made to avoid unexpected particle concentrations and to calculate collision effects.
It is one of most important factors in simulating fluid dynamics for the particle
method; however, it might not be useful for a solid particle method. The spreading
shape of the soil particles after the soil bed collision also differed. The solid
particles splashed further than in the liquid model. This is due to the viscosity
effects on the liquid model, which might not be considered.
4. Conclusion
In this study, the MPS method was used to simulate mobile-bed behavior under a
large-wave collision problem. To demonstrate soil particle behavior, the MPS method
for liquid particles was modified and developed. For modification and development,
friction- and drag-force models were implemented instead of a viscosity model.
For validation of the accuracy of a solid particle program, the results from the
MPS method for solid particles was compared to the MPS method for liquid particles.
Both numerical simulation results were compared to corresponding experimental results.
The numerical simulations agreed well with the experimental results; however, differences
were observed between the MPS method for the liquid model and the solid model. For
a macro view of the comparison, the tendency of the soil bed behavior was similar;
however, the details were dissimilar. The MPS for the solid model had a more blunt
leading edge in the soil particle wave, and smoother behavior on soil wave propagation.
The MPS method has been extended in its applicability to not only the fluid dynamics
problem, but also to the solid particle problem, including the fluid\textendash{}solid
interaction problem. It would be useful to study a more complicated problem, such
as the scour and erosion problem, in the future.
ACKNOWLEDGMENTS
This research was supported by the National Research Foundation of Korea (NRF)
funded by the Ministry of Education (NRF-2018R1D1A1B07048254).
REFERENCES
Chenson H., 2009, Application of the method of characteristics to the dam break wave
problem, J. Hydraul. Res., Vol. 47, pp. 41-49
Liao C. B., Wu M. S., Liang S. J., 2009, Numerical simulation of a dam break for an
actual river terrain environment, Hydrol. Process, Vol. 21, pp. 447-460
Janosi I. M., Jan D., Szabo K. G., 2004, Turbulent drag reduction in dam break flows,
Exp. Fluids, Vol. 37, pp. 219-229
Monaghan J.J., 1994, simulating free surface slows with SPH, J. Comput. Phys., Vol.
110, No. 2, pp. 39-406
Xu R., Stansby P., Laurence D. L., 2009, Accuracy and Stability in incompressible
SPH based on the projection method and a new approach, J. Comput. Phys., Vol. 228,
No. 18, pp. 6703-6725
Koshizuka S., Oka Y., 1996, moving particle semi-implicit method for fragmentation
of incompressible fluid, Num. sci. and Eng., Vol. 123, pp. 421-434
Tanaka M., Masunaga T., 2010, Stabilization and smoothing of pressure in MPS method
by quasi-compressibility, J. Comput. Phys., Vol. 229, No. 11, pp. 4279-4290
Lee B.H., Park J.C., Kim M.H., 2010, Numerical simulation of impact loads using a
particle method, Ocean Eng., Vol. 37, pp. 164-173
Duan G., Matsunaga T., Yamaji A., Koshizuka S., Sakai M., 2020, Imposing accurate
wall boundary conditions in corrective-matrix-based moving particle semi-implicit
method for free surface flow, Int. J. Numerical method in fluids
Kim K.S., Kim M.H., 2015, Simulation of Multiliquid-layer sloshing with vessel motion
by using moving particle semi-implicit method, J. offshore Mech. And Arctic Eng.,
Vol. 137, 051602-1
Kim K.S., Kim M.H., Jang H., Cho H.C., 2018, Simulation of solid particle interactions
including segregated lamination by using MPS method, Comp. Modeling of Eng. Sci.,
Vol. 116, No. 1, pp. 11-29
Harada E., Gotoh H., Ikari H., Khayyer A., 2019, Numerical simulation for sediment
transport using MPS-DEM coupling model, Advanced in Water Resource, Vol. 129, pp.
354-364
Author
Kyung Sung Kim is an Associate Professor of Dept. Mechatronics Convergence Engineering
at Changwon National University, Changwon, Korea. He received a bachelor degree and
a master’s degree in Naval Architecture and Ocean Engineering from Pusan National
University in Korea. He then moved to Texas in the U.S.A., and earned a PhD in Ocean
Engineering from Texas A&M University. He was employed at the Pohang University of
Science and Technology as a Research Professor. His major field is hydrodynamics and
computational fluid dynamics, including particle methods. However, he is now in charge
of experimental approaches to offshore experiments for education. He is working to
develop efficient methods of educating young researchers with fourth industrial revolution
technology in both numerical and experimental research.