GPU-SPH simulation of Tsunami-like wave interaction with a...
Transcript of GPU-SPH simulation of Tsunami-like wave interaction with a...
43
GPU-SPH simulation of Tsunami-like wave interaction
with a seawall associated with underwater
Safiyari, Omid Reza1; Akbarpour Jannat, Mahmood Reza
1*; Banijamali, Babak
2
1- Iranian National Institute for Oceanography and Atmospheric Science (INIOAS), Tehran, IR. Iran
2- Darya Bandar Engineering Consultant Company (DBC), Tehran, IR. Iran
Received: January 2016 Accepted: May 2016
© 2016 Journal of the Persian Gulf. All rights reserved.
Abstract Investigation of the waves generated by underwater disturbances gives precious insight into the
effect of man-made underwater explosions as well as natural phenomena, such as underwater
volcanoes or oceanic meteor impact. On the other hand, prediction of the effects of such waves
on the coastal installations and structures is required for preparation worthwhile criteria for
coastal engineers to prepare a reliable design. This study aimed to investigate the interactional
effects of water waves generated by underwater disturbances on sea walls through numerical
modeling using the Smoothed Particle Hydrodynamics (SPH) method. The simulation was
performed using the Dual-SPHysics numerical code. Comparison of the numerical results with
the experimental data extracted from case studies demonstrated the good capability of the SPH
algorithm used in the numerical code in the simulation of initial wave generation by the
underwater disturbance and its propagation over the body of water. To examine the wave force
exerted on the walls, the results of laboratory experiences on the effect of tsunami waves on
coastal structures were used to verify the numerical model. The study found that the phenomena
with such nonlinear behavior can be very well simulated with a calibrated SPH model. We also
explored the effects of this type of wave and temporal changes of its resultant force on the wall.
In this article, the explosion-generated water wave produced much stronger fluctuations in the
vicinity of the wall than did the solitary wave, thus naturally, it can be more destructive.
Keywords: Wave generation, Seawall, Numerical modeling, Wave force, Underwater explosion, Smoothed Particle Hydrodynamics (SPH), Graphics Processing Unit (GPU)
1. Introduction
The research on the likelihood and effects of
underwater explosions is of particular importance for
engineers in different fields of design or construction
because of: (1) the development and presence of
variety of marine installations and structures located
in coastal regions, (2) the potential of manmade
action near critical oil and gas assets in territorial
waters of countries or explosions for construction
purposes in shallow water, and (3) the possibility of
explosion of offshore hydrocarbon facilities and
pipelines. The review of research and experiences in
the field of underwater disturbance demonstrates that
the human quest to apply the waves generated by
manmade explosions as an underwater disturbance
generator since the nineteenth century. In the past
century, the occurrence of unintentional and
unexpected strong explosions in the oceanic area and
Journal of the Persian Gulf
(Marine Science)/Vol. 7/No. 24/ June 2016/23/43-66
* Email: [email protected]
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
44
related deep water tests and field experiments draw
the attention of famous scientists and engineers
working on marine research and wave theory to a
new field of research focused on direct and indirect
consequences of underwater explosions, and
especially wave generation and propagation during
this phenomenon (LeMe’haute’, 1971, 1976, 1993).
In 1917, the explosion of Mont Blanc warship in the
Bay of Halifax, Canada produced huge tidal surges
that affected coastal areas and ports near the area. The destructive water waves generated during this
accident had major implications for military and
marine engineering. The number of deaths and the
extent of damage to buildings and facilities following
this accident were so high that it was labeled a
human and national catastrophe (Greenberg and et
al.., 1993). In 1944 and 1945, Professor Lich at Auckland
University directed a series of field tests aimed at
producing tsunami waves by underwater
disturbances, as part of a common research operation
called “Project Seal” involving New Zealand, US
and British forces. The results of these surveys
indicate that man-made explosions, produced in
various depths of water, can generate waves of long
wavelengths capable of destroying coastal areas. In
the mid-decades of the twentieth century, many
researchers, including Van Dorn (1961-1968),
Whalin (1962-1970), and LeMehaute (1967-1998),
used the results of underwater explosion field tests
conducted by US Army to introduce and develop
some theories in this regard. Considering the difficulty and limitations of
laboratory modeling and the hazards of field
experiments on this subject, numerical simulation is
the preferred method of marine engineers and
researchers investigating the reaction of marine
environments to underwater explosions. Other
researchers, such as Crawford and Mader (1998) and
Simskin and Fiske (2003) have employed the results
and theoretical models of this field to develop a
series of concepts for studying the effects of similar
phenomena such as tsunami waves and waves
generated by explosion of submarine volcanoes and
for calculating the characteristics of waves generated
by the collision of celestial objects and meteorites. From the 1960s until the early 21st century, the
scholars working in this field used to employ the
Eulerian numerical methods to arrange governing
equations into discrete form and solve them
accordingly. But, because of reasons, such as the
high velocity of the shock wave, emergence of large
deformations, and separation of water particles from
each other, solving the explosion-generated wave
equations is a computationally complex process. With the development of meshless computational
methods, improvement of numerical calculation
techniques, and progress of hardware technology
since the 1990s, the Lagrangian-based techniques
have become the analysis method of choice in
several fields. The most common application of such
methods is in the simulation of shock propagation
problems wherein variables change drastically with
time. One of the notable Lagrangian methods for
solving fluid mechanics problems is the smoothed
particle hydrodynamics (SPH) method. This method
was developed by Monaghan and Gingold in 1977 to
simulate problems in various physical aspects. The
original application of this method was to analyze
the problems concerning astronomy and movements
of cosmic devices, but given the variety and
complexity of particular problems of computational
fluid mechanics, such as free surface problems in
coastal engineering, and also because of the extreme
difficulty of developing Eulerian models for these
problems, SPH is being increasingly used for the
analysis of marine phenomena.
Gómez-Gesteira et al. (2004) and Gómez-Gesteira
et al. (2012) studied wave impact on wall structures
by 3D SPH method. Rogers et al.. (2010) simulated
the movement of caisson breakwaters due to the
impact of waves using a 2D SPH model. Robertson
et al.. (2011) tried to understand the response of
coastal walls under tsunami forces by laboratory
investigations. Pu and Shao (2012) investigated
overtopping characteristics of wave impact on
coastal structures by SPH technique. El-Solh (2012)
studied the impact of solitary wave and its
hydrodynamics effects on sea walls. Barreiro et al.
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
45
(2013) tried to solve wave behavior and interaction
by SPH method. Altomare et al. (2014) analyzed
interaction of waves on armor type breakwaters,
Altomare et al. (2015) investigated wave impact on
coastal structures by SPH.
Using parallel processing methods in combination
with properties of graphical processing units of
computers, a new generation of high performance
computing techniques was developed to apply SPH
method in CFD and oceanic computations.
Domínguez et al. (2013), studied and developed new
computational techniques for applying GPU to
numerical modeling by SPH. Crespo et al. (2015)
developed traditional SPH model on CPU into new
form by applying GPU. Altomare et al. (2015)
studied applicability of smoothed particle
hydrodynamics for estimation of sea wave impact on
coastal structures. Altomare et al. (2017),
investigated modeling of long-crested wave
generation and absorption by SPH on GPU.
Previous studies have shown that the use of this
new methodology on GPU, provides better
computational speed and precision to study complex
marine phenomena that require long analysis time.
Besides, for problems with complicated and large
deformation on free water surface, a Lagrangian
point of view is more practical and relevant.
In this study, we use the SPH method to simulate
the waves generated by the underwater disturbance,
such as explosion and examine its capability in this
application. One of the main goals of this research is
to study the impacting forces on the coastal wall in
front of an explosion-generated water wave. Having
the pressure distribution, the size of force applied to
the wall could be easily determined. Naturally,
knowing the load induced by explosion-generated
water waves, engineers will be able to design the
walls and coastal installations with better measures
to counteract such effects. Another issue implicitly
addressed in this study is the difference between the
time histories of the force produced by the tsunami
wave and the explosion-generated water wave on the
vertical coastal walls.
2. Materials and Methods
The smoothed-particle hydrodynamics (SPH)
method is a Lagrangian approach based on the
simulation of the path of particle motion in an
analytical domain. The major advantages of the SPH
method are the independence of its simulations from
the computational grid. This feature allows the SPH to
bypass the limitation of the finite element or other
mesh-based methods for computational problems and
avoid the complexity imposed by re-meshing in the
problems that require the simulation of particle
segregation. Besides, with this numerical analysis
method, problems of varying complexities can be
numerically solved with significantly higher speed and
reduced simulation run time.
Given the high complexity of calculations required
for simulation of explosion-generated water waves,
utilization of high-performance computing systems for
this purpose is essential. For this purpose, we used a
GPU-based parallel processing tool called
DualSPHysics. The DualSPHysics is an open source
numerical code capable of using GPU’s processing
power, which has been developed in a joint research
project involving several European universities. In this
study, DualSPHysics and its capabilities in the
Lagrangian reference frame were used for modeling,
and the simulation results were compared with field
and experimental data to ensure the validity of the
numerical code. The hardware specifications of the
simulation system are listed in Table (1).
The uncertainties that exist in field and laboratory
observations of this phenomenon have caused
numerous and complicated errors in the results.
LeMehaute (1971) has said that the amount of
laboratory errors is between 30% and 100%. To
simulate this problem, a numerical model should be
used to minimize the effects on the accuracy and results
of such errors.
The SPH model has the ability to provide conditions
for separating water particles from each other and take
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
46
Table 1: General Specification of the applied hardware and analytical parameters of various simulations
Executive
Parameters
Hardware Specifications
Machine No. 1
Hardware Specifications
Machine No. 2
Particle size (m) 10-3 ~ 5×10-1 CPU type AMD A8-
3870 APU CPU type
Intel®
Core™ i7-
4700MQ
Total
Particle 105 ~ 4×106 GPU type
GeForce
GTX 680 GPU type
GeForce
GTX 780M
GPU Run Time (sec) 1200 ~ 3400 CUDA
Core No. 1536
CUDA
Core No. 1536
CPU Run Time (sec) 3×103 ~
40×103
into account the high displacement of the particles in
the disturbance origin. Also, when waves impact on a
coastal wall, it is possible to examine the collision of
water particles with structures more precisely. These
advantages make it possible to simulate the generation
and propagation of waves from underwater explosion
with adequate reality in comparison with the Eulerian
type models that previously used by researchers (Mader
1998, Mader 2003, LeMehaute et al. 1986, LeMehaute
et al. 1998). Also, by applying the parallel processing
technique on GPUs, huge analyses could be performed
with cheaper prices in a shorter computational time.
2.1. Hydrodynamics of explosion-generated water
waves - variations of the initial conditions of the free
surface
The governing equations for an underwater explosion
consist of a set of conservation equations (mass
continuity, momentum conservation, and energy
conservation), the equation of state of detonation
products, the equation of state of water, and the
combustion equation, which is used to analyze the main
process of an explosion in the water environment. For
an explosion-generated surface wave to form, the gas
contained bubble produced at the moment of explosion
migrates toward water surface, and then, collides with
the free surface, rupturing its wall. This is followed by
the release of gas and fume constituents, which push
the water body outward while throwing the water
particles upward, thus generating the primary wave. Hence, the wave generation process can be divided into
two phases. In the first phase, first, the nucleus of the
bubble is formed, then, starts its evolution from a small
high-pressure dome at the depth of water to a large
bubble pulsating toward the free surface. The second
phase starts when the bubble reaches the water surface,
where it forms an initial deformation.
The initial deformation equation expressed by
Eq.(1), which presents a second order parabolic curve,
is based on the theoretical proposals of Poisson (1816)
and Kranzer and Keller (1959):
𝜂(𝑟) = {𝜂0 [2 (
r
R0)
2− 1] ; r ≤ R0
0 ; r > R0
(1)
Where 𝜂0 represents the maximum initial depth
and the initial height of the edges of the explosion
crater, R is the maximum span, "𝑟" represents the
distance to the center of the explosion location at
the water level and 𝜂(𝑟) is the function of water
surface variations at "𝑟" . Normally, the
characteristic parameters of crater shape, 𝜂0 and
R0, can be evaluated by empirical equations or
experimental results as a function of charge weight,
water depth and depth of explosion (Van Dorn,
1961, LeMehaute et al.. 1989). The initial geometry
of the free surface as characterized by Eq.(1) is
shown in Figure 1:
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
47
Fig.1: Initial Parabolic with lip deformation of water level
The field studies conducted by Penney (1945 and
1946) and Jordan (1965) have proven that only 40% of
the energy released by an underwater explosion through
the release of chemical energy stored in an explosive
charge may generate water waves at free surface, as the
rest of the explosion energy will be spent in generation
of pressure waves in the water body and increase of
water temperature. The effective part of the released
chemical energy will be rapidly transformed into the
potential energy stored in the form of the initial
deformation of the water surface around the site of the
explosion. This potential energy is the generator of base
water wave.
Efficiency of an explosion as a wave producing
mechanism is one of the important issues of the
physical phenomenon. Amount of energy related to
water waves produced by underwater disturbances was
proposed by Whalin (1967) for different types of initial
water surface deformation on the basis of Sakurai
(1965) studies. The potential energy of deformed water
elevation can be represented by Eq.(2):
𝐸 = 2𝜋𝜌𝑔 ∫𝜂0
2(𝑟)
2𝑟𝑑𝑟
∞
0 (2)
By insertion of water level formula such as shown
by Eq.(1) the energy equation for quadratic surface
deformation with lip is determined as following by
Eq.(3):
𝐸 =𝜋𝜌𝑔𝜂0
2𝑅02
6 (3)
According to Figure 1, Van Dorn, LeMehaute and
Hwang (1968) by performing physical laboratory tests
show that for a parabolic crater, the radius of crater at
edge of lip is related to the yield of the explosive charge
by an empirical formula as per Eq.(4):
𝑅0 = 9.6𝑊0.3 (4)
Where the charge yield "𝑊" has units of pounds of
TNT, 𝑅0 and 𝜂0 have units of feet. 𝜂0 can be
determined for the parabolic crater by Eq.(5)
𝜂0 × 𝑅0 = 0.81𝐻𝑚𝑎𝑥 × 𝑟 (5)
Which was resulted from experiments. 𝐻𝑚𝑎𝑥 is the
maximum wave height in a wave group at a distance of
𝑟 from the location of the explosion. The value of
𝐻𝑚𝑎𝑥 × 𝑟 is dependent upon the depth of submergence
of the charge and can be determined for any depth of
submergence that was presented by Whalin, Pace and
Lane (1970) using curve fitting of field data. At upper
critical depth they proposed Eq.(6):
𝐻𝑚𝑎𝑥×𝑟
𝑊0.54 ≅ 34 (6)
Also, LeMehaute (1980) presents slightly greater
value by Eq.(7) on the basis of test results carried out at
Waterways Experimental Station:
𝜼𝟎
𝜼𝟎
𝑹𝟎
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
48
𝐻𝑚𝑎𝑥×𝑟
𝑊0.54 ≅ 36 (7)
In this study, equations 3 to 7 are used to calculate
and align the energy of waves generated by an
explosion with a solitary wave in the further studies.
2.2. Smoothed Particle Hydrodynamics (SPH)
method
SPH is a Lagrangian method whereby the
continuous medium can be discretized into a set of
disordered points. SPH allows a function to be
expressed in terms of its values at a set of particles by
interpolation without needing any grid to calculate
spatial derivatives. In this way, physical properties of
each particle, such as acceleration, density, etc. are
quantified as an interpolation of the same values in
neighbor nodes. The SPH technique approximates a
scalar function 𝐴(𝒓) at any point with r vector of
position, as follows:
𝐴(𝒓) ≅ ∫ 𝐴(𝒓′). 𝑊(𝒓 − 𝒓′, ℎ)𝑑𝒓′ (8)
In Eq.(8), ℎ is called the smoothing length and
represents the influence of the nearest particles in a
neighboring domain (see Figure 2), and it is therefore
weighted in proportion to the distance between
particles. Kernel function 𝑊 is used to estimate the
amount of participation through the parameter ℎ.
Fig.2: Affected area and neighbor particles of ith
particle
In Eq.(9), the summation is extended to all particles
within the neighboring distance of particle "a", and the
volume associated with the particle "b" is 𝑚b
𝜌b, where
𝑚b and 𝜌b denote respectively the mass and the density
of this neighbor particle. The kernel functions shall
have several properties, including: positivity inside the
area of interaction, compact support, normalization, and
monotonic decrease with distance. Among notable
options for the kernel is the quintic kernel proposed by
Wendland (1995), where the weighting function
vanishes for inter-particle distances greater than 2ℎ.
This kernel has been defined as (Altomare et al., 2014,
2015) by Eq.(10):
𝑊(𝑞) = 𝛼𝐷 (1 −𝑞
2)
4(2𝑞 + 1) ; 0 ≤ 𝑞 ≤ 2 (10)
Where 𝑞 =|𝒓|
ℎ is the ratio of particle distance to
smoothing length, and 𝛼D is a normalization constant,
which is equal to 7
14𝜋.ℎ2 in two dimensions and 21
16𝜋.ℎ3
in three dimensions.
The kernel functions have been used to transform the
conservation laws of continuum fluid dynamics from
their differential form into particle forms. The
momentum equation proposed by Monaghan (1992)
has been used to determine the acceleration of particle
"a" as a function of its interaction with a neighbor
particle, such as "b":
𝑑𝒗𝑎
𝑑𝑡= − ∑ 𝑚𝑏 (
𝑃𝑏
𝜌𝑏2 +
𝑃𝑎
𝜌𝑎2 + 𝛱𝑎𝑏) 𝜵𝑎𝑊𝑎𝑏 + 𝒈𝑏 (11)
In Eq.(11), 𝒗 is the particle velocity, 𝑃 is the particle
pressure, 𝒈 is the gravitational acceleration vector, and
𝑊𝑎𝑏 is the kernel function, which depends on the
distance between particles "a" and "b". 𝛱𝑎𝑏 is the
viscous term according to the artificial viscosity
definition proposed by Monaghan (1992):
𝛱𝑎𝑏 = {−
𝛼𝑣 . 𝑐�̅�𝑏 . 𝜇𝑎𝑏
�̅�𝑎𝑏 ; 𝒗𝑎𝑏 . 𝒓𝑎𝑏 < 0
0 ; 𝒗𝑎𝑏 . 𝒓𝑎𝑏 ≥ 0 (12)
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
49
In Eq.(12),𝒓ab = 𝒓a − 𝒓b and 𝒗ab = 𝒗a − 𝒗b are
the particle position and velocity, respectively. 𝛼𝑣 is a
coefficient that needs to be tuned to introduce the
proper attenuation. In this study, we have used
𝛼𝑣 = 0.01 , as it is the minimum value that prevents
instability and spurious oscillations in the numerical
scheme. 𝑐a̅b = 𝑐a + 𝑐b
2 is the mean speed of sound of
particles, �̅�ab = 𝜌a + 𝜌b
2 is the mean density of particles
and 𝜇ab =ℎ .𝒗ab . 𝒓ab
𝒓ab . 𝒓ab+ 𝜂2 where 𝜂2 = 0.01 × ℎ2.
The equation of states for ideal gases are:
𝑃 = 𝐵 [(𝜌
𝜌0)
𝛾− 1] (13)
𝐵 =𝑐0
2 . 𝜌0
𝛾 (14)
𝑐0 = 𝑐(𝜌0) = √𝜕𝑃
𝜕𝜌|
𝜌0
(15)
Where 𝐵 is a constant related to the fluid
compressibility modulus, 𝜌0 = 1000 kg/m3 is the
reference density, chosen as the density at the free
surface, 𝛾 is a constant, normally between 1 and 7, and
𝑐0 is the speed of sound at the reference density.
3. Results
3.1 Simulation of Tsunami-like wave interaction
with a seawall
Most of the previous researches about the effects of
water waves generated by underwater explosion were
focused on wave propagation and inundation effects.
Besides, some scientists studied the near field or
offshore effects of these waves on ships or underwater
infrastructures. So far, very limited field or laboratory
investigations have provided observational and
qualitative information about the effects and magnitude
of the impact of explosion-generated water waves on
coastal structures. The lack of quantitative recorded
data of interaction forces or pressure distribution on
coastal structures due to underwater explosion induced
wave, cause limitation for verification of numerical
models of this phenomenon. Thus, in this article, before
estimating the effects of these waves on coastal walls
by numerical SPH analysis, the validity of the
DualSPHysics code was checked with the help of
variously available field datasets in different stages
such as wave generation, propagation, and interaction.
First, the available experimental data about the
generation of the initial wave by the underwater
explosion and its propagation in the surrounding
environment were used to verify the ability of the SPH
method to simulate this part of the phenomenon.
In the second step, the ability of the numerical code
to simulate the effects of the wave collision with the
coastal structures was measured with the help of
existing experimental data about the impulse applied by
recorded laboratory data of artificially generated
tsunami (solitary) wave on a vertical coastal wall.
In the final step, the water waves generated by the
underwater explosion were simulated to estimate the
force applied to the wall. On the other words, an
explosion generated a wave with the equal energy of
solitary waves of the second step was reproduced and
simulated in the same numerical flume. The results of
these waves with different types of origin are compared
with each other.
In order to obtain the appropriate results in each of
the models, the number of water particles was increased
to the extent that the error results were less than 5% in
two successive sequences of the numerical model (see
tables 3,4,5). Subsequent simulations were executed in
each case using the number of particles fixed for each
example to assure the results.
3.1.1 Calibration and validation of the SPH Model
for generation and propagation of explosion-
generated water wave
We used two laboratory test cases to validate the
DualSPHysics code and the SPH method incorporated
therein. For generation of primary waves at the region
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
50
of explosion, we first calculate the amount of radius of
the crater lip, 𝑅0, by insertion of the charge weight "𝑊"
in the Eq.(4). Then, with the help of equations 6 and 7,
the range of expression 𝐻𝑚𝑎𝑥 × 𝑟 could be evaluated.
At last, with the help of the Eq.(5) the amount of
𝜂0 × 𝑅0 is calculated, and then 𝜂0 may be estimated. In
this way, the main parameters of Eq.(1) are determined
and the deformed surface of water level as the initial
boundary condition is applicable to the numerical
model.
In order to measure the results of the numerical
model in comparison with the experimental results,
using the RMSE formula as Eq. (16), the output values
of the numerical results were evaluated.
𝑅𝑀𝑆𝐸 = √∑ (𝑋𝑚−𝑋𝑒)2𝑛
𝑖=1
𝑛 (16)
where 𝑋𝑚 is numerical model output, 𝑋𝑒 is recorded
data from experiments and 𝑛 is number of data. In the following, we first describe experimental
cases and then compare the results of numerical
modeling with the experimental data:
A. The first set of experimental data is the
result of studies conducted by Charlesworth (1945)
at the London Road Research Laboratory. In this
test, an underwater explosion was created in a 55 ft
(16.76 m) wide and 70 ft (21.34 m) long rectangular
gulf surrounded by the coast in the south and the
west. The test area had an almost uniform water
depth ranging from 15 to 18 feet (4.57 to 5.49
meters). The wave amplitude was recorded at a
point 56.5 ft (17.22 m) away from the explosion at
different times. A sample of the surface waves
produced following the explosion of a 32-pound
(14.5 kg) charge at a depth of 8 ft (2.44 m) was
recorded. The initial condition of this case is
considered on the basis of Eq.(1) as a parabolic
crater with a lip. The characteristics values of this
initial condition are: 𝜂0 = 2.07 m, R0 = 8.28 m.
This initial crater is defined in the midpoint of the
simulated basin. In Figure 3, the result of this test
case is shown with the red curve. The results of sthe
imulation with the SPH method were obtained by
adjustment of different parameters for this sample.
Fig. 3: Comparison of SPH model results with Charlesworth test results (1945)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0 1 2 3 4 5 6 7 8 9
Wav
e A
mp
litu
de
(m)
Time (sec)
Charlesworth's experimental curve
SPH 3D,vis=0.00,D=50m
SPH 3D,vis=0.25,D=50m
SPH 3D,vis=0.10,D=50m
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
51
In Figure 3, these results are plotted in the form of
dash-line and dot-line curves. The range of variation of
the parameters examined in this verification is
presented in Table (2). The plotted results demonstrate
the ability of SPH method to accurately simulate the
initial waves generated by the explosion at the desired
point. The maximum root mean square error (deviation
of simulation results from the corresponding
experimental values) was 11.4% and the minimum
coefficient of correlation between simulation results
and experimental data was estimated to 0.93.
Table 2. Characteristics of the main input parameters in the 3D SPH model used for simulation
of the Charlesworth’s test (1945)
Model Executive
Parameters Simulation
Parameters
Step Scheme Symplectic dP (m) 0.2 ~ 0.5
No. Particles 1383984
Kernel Wendland Model Size (m) 100 × 100
Viscosity
Treatment Artificial CFL
number 0.2
Viscosity 0.01 ~ 0.25 Coefficient
of Sound 10 ~ 15
Shepard Steps 20 ~ 30 Smooting
Coefficient 0.8 ~ 1.0
The second set of experimental data that was used to
measure the efficiency of the SPH model is the data
from the Prins' experiments (1956 and 1957) in the
Wave Research Laboratory at the UC Berkeley’s
Engineering and Technology Research Institute. These
experiments were carried out in a flume with a width of
one foot (0.3048 m) and length of 60 feet (18.288 m).
To produce an initial shock at water level (similar to an
underwater explosion event), an air blowing-suction
system was used to lower and raise the water level at
the flume’s upstream. The initial water wave was
released with a positive or negative height (compared to
the flume’s normal water level) to generate a shock in
the flume’s longitudinal direction. In this study, we
used the test results reported for the uniform water
depth of 2.3 ft (0.7 m), positive initial wave height of
0.4 feet (0.122 m), and length of 2 feet (0.61 m) at the
left side of the numerical flume. The initial wave is
considered at the center of the basin model. The
changes in the water elevation were measured at a point
15 ft (4.57 m) away from the shock initiation point. In
Figure 4, these measurements are displayed by the blue
curve. In Figure 4, the results of the simulation with the
SPH method following the adjustment of different
parameters for this test are shown in the form of dash-
line and dot-line curves. The range of variation of the
parameters examined in this verification is presented in
Table (3). As the presented results indicate, the SPH
method can accurately simulate the initial waves
generated by the explosion at the desired point.
In this case, the maximum root mean square error
(deviation of simulation results from the corresponding
experimental values) was 12.4% and the minimum
coefficient of correlation between simulation and
experimental results was 0.89.
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
52
Fig. 4: Comparison of Two-Dimensional SPH model results with the Prins’ Test Results (1956)
Table 3. Characteristics of the main input parameters in the 2D SPH model used
for simulation of the Prins’ test (1945)
Model Executive
Parameters
Simulation
Parameters
Step Scheme Symplectic dP (m) 6.096×10
-3 ~ 1.016×10-2
No. Particles 141075
Kernel Wendland Model Size
(m) 20
Viscosity
Treatment
Laminar
SPS Artificial
CFL
number 0.2
Viscosity 10-6 0.01 ~ 0.25 Coefficient
of Sound 10 ~ 20
Shepard
Steps 20
Smoothing
Coefficient 0.87
3.1.2 Calibration and validation of the SPH Model
for estimation of laboratory generated tsunami
(solitary) wave induced force on the coastal wall
A large number of experiments have been carried
out to measure the force exerted on the sea walls by
tsunami waves. In this part of the study, we use two
laboratory experiences to assess the capability of the
SPH model for estimating the wave force on the coastal
walls.
For this part of the study, the amount of energy for
solitary waves is also calculated, so that in the next
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Wav
e am
pli
tud
e (m
)
Time (sec)
Prins' experimental curve
SPH 2D, dp=0.01016,Vis.=0.01,dt=0.0005
SPH 2D, dp=0.01016,Vis.=0.05,dt=0.001
SPH 2D, dp=0.01016,Vis.=0.15,dt=0.001
𝑅𝑀𝑆𝐸𝑚𝑎𝑥 = 12.4% 𝑅𝑚in = 0.89
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
53
section of the paper, the surface waves due to
underwater explosion with their equivalent energy are
generated.
Longuet-Higgins (1974), reviewed a number of
successful models for prediction of solitary waves. His
studies showed that using the Eq.(17) may represent
excellent results to predict the waveform.
{𝜂 = 𝐴𝑒−𝜇|𝑥| + 𝐵𝑒−𝜈|𝑥| 𝐴 = 1.5389 ; 𝐵 = −0.7093𝜇 = 1.0495 ; 𝜈 = 1.4630
(17)
He also assumed the equations 18 and 19 to calculate
the kinetic "𝑇" and potential "𝑉" energy of the solitary
wave. Then total energy "𝐸" of wave could be
calculated by 𝐸 = 𝑇 + 𝑉.
𝑇 = ∫ ∫1
2(𝑢2 + 𝑣2)
𝜂
−ℎ𝑑𝑦 𝑑𝑥
∞
−∞ (18)
𝑉 = ∫ 1
2𝑔𝜂2𝑑𝑥
∞
−∞ (19)
Where 𝑢 and 𝑣 are velocity component of water
particle in horizontal and vertical directions,
respectively. Also 𝜂 characterizes water elevation and ℎ
is water depth.
By applying equation 17 into equations 18 and 19,
he was able to provide a very good approximation for
estimating the energy values of the solitary wave,
which is considered in equations 20 and 21, properly.
𝑇 = 0.527 𝑔ℎ3 (20)
𝑉 = 0.431 𝑔ℎ3 (21)
Finally, the total energy of a solitary wave can be
approximated by Eq.(22) as follows:
𝐸 = 0.958 𝑔ℎ3 (22)
Besides, in this section, the corresponding simulation
results obtained with the DualSPHysics numerical code
are presented correspondingly for each case.
A. The results of laboratory-based physical
modeling performed by Robertson et al.
(2011). This test set was constructed in a
laboratory flume with a length of 83.13 m,
the width of 3.66 m and depth of 4.57 m in
the laboratory of Oregon State University
(see Figs. 5 and 6).
Fig. 5: Side view of the Laboratory Flume of Robertson et al. (2011) at the University of Oregon State Laboratory
Fig. 6: SPH environment of Robertson et al. Test case flume
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
54
To reproduce the experimental data in simulation, a
solitary wave (representing the tsunami wave) with a
wave height of 1.064 meters and wavelength of 30
meters was generated in the water depth of 2.66 m. The
propagation of the generated solitary wave in the
numerical simulation of experiments in deep water
region is displayed in Figure 7. Time history variations
of pressure distribution on a vertical wall are shown in
Figure 8. In Figure 9, time history variations of the
force applied to the wall are compared with the
corresponding experimental data. Also, variations of
impulse with time are plotted in Figure 10. The
maximum sum of squares of relative errors of
simulation is 5.03% for force values and 1.41% for
impulse values. The minimum coefficient of correlation
between simulation results and experimental data is
98.16 for force values and 99.96 for impulse. The total
energy of this wave is estimated to be 𝐸 ≅ 177 kJouls.
Fig. 7: Snapshots of solitary wave generation (by paddle)
and propagation along the numerical SPH flume for Robertson’s experiments
t=2.0 sec
t=4.0 sec
t=6.0 sec
t=8.0 sec
t=10.0 sec
t=12.0 sec
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
55
Fig. 8: Time history of pressure distribution on the vertical wall
Fig. 9: Time history of the force on the wall during the wave impact to the wall, (a) laboratory (circle) and (b) simulation of
the SPH (star)
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20
Hei
gth
(m
)
Pressure (kPa)
t=21 sec t=24 sec
t= 28 sec t= 30 sec
0
5
10
15
20
25
30
35
40
0 2 4 6 8 10 12
Fo
rce
(kN
)
Time (sec)
Experiment data SPH results
𝑅𝑀𝑆𝐸𝑚𝑎𝑥 = 5.03% 𝑅𝑚in = 98.16
20 22 24 26 28 30
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
56
Fig. 10: Time history of the impact on the wall by Robertson (2011), laboratory results (dash) and SPH results (line)
B. Results of laboratory experiments performed
by Yu Yao et al. (2017)
This case study was carried out in a laboratory flume
with a length of 36 m, the width of 0.55 m and depth of
0.60 m in Nanyang Technological University,
Singapore. A 1:5.7 steep fore slope starting from
16.33m away from the wave paddle was constructed
and a rectangular seawall model with 6 cm long, 5.5 cm
wide, and 5 cm high, was placed on the top point of the
slope to represent a vertical caisson type seawall (see
Figure 11).
The recorded data by G1 and G2 wave gauges,
which were located 5.24 m and 0.14 m upstream from
the seaside of the wall, respectively. Besides, induced
dynamic pressure by wave impinging on the wall was
recorded by a pressure sensor installed on the center of
the seawall (see Figure 12).
To reproduce the experimental data in the
simulation, a solitary wave with a wave height of 0.05
m and wavelength of 2.05 m was generated in the water
depth of 0.36 m. The numerical model is considered
symmetrical with sufficient far sloped beach from point
of initial wave generation in the middle of the
numerical flume for canceling of reflecting waves.
Geometry and initial generated solitary wave in the
numerical flume are presented in Figures 13 and 14.
Fig.11 Sketch of the experimental setup (reported by Yu Yao et al., 2017)
0
20
40
60
80
100
120
140
160
0 2 4 6 8 10 12
Imp
uls
e (k
N-m
)
Time (sec)
Experiment data
SPH results
𝑅𝑀𝑆𝐸𝑚𝑎𝑥 = 1.41% 𝑅𝑚in = 99.96
20 22 24 26 28 30
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
57
Fig.12 Seawall model and pressure gauge location on the wall (reported by Yu Yao et al., 2017)
Fig. 13: SPH environment of Yu Yao et al. Test case flume (up)
and initial solitary wave generated in the model (down)
Fig. 14: Enlargement of solitary wave generated in SPH model
(height of generated wave is approximately 5 cm)
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
58
Comparing of measured free surface elevations with
the numerical model at different wave sampling
locations (by G1 and G2 sensors) are presented in
Figures 15 and 16. In addition time history of pressure
at the midpoint of the seawall is plotted in Figure 17.
Both of free surface and pressure graphs are presented
in non-dimensional scale as per following pictures and
show good agreement between experimental and SPH
results. The total energy of this wave is estimated to be
𝐸 ≅ 0.44 kJouls.
Fig. 15 Comparison between SPH and Experimental time history of non-dimensional solitary wave
surface at gauge G1 (Experimental results are reported by Yu Yao et al., 2017)
Fig. 16 Comparison between SPH and Experimental time history of non-dimensional solitary wave
surface at gauge G2 (Experimental results are reported by Yu Yao et al., 2017)
Fig. 17 Comparison between SPH and Experimental time history of non-dimensional dynamic
pressure of solitary wave (Experimental results are reported by Yu Yao et al., 2017)
-1
0
1
2
3
0 5 10 15 20 25 30 35 40 45
η/h
t/(h/g)
WG1 (SPH) WG1 (EXP)
-1
0
1
2
3
0 5 10 15 20 25 30 35 40 45
η/h
t/(h/g)
WG2 (SPH) WG2 (EXP)
-0.1
0
0.1
0.2
0.3
0 5 10 15 20 25 30 35 40 45
P/r
gh
t/(h/g)
SPH EXP
𝑅𝑀𝑆𝐸𝑚𝑎𝑥 = 1.6%
𝑅𝑚in = 0.98
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
59
3.1.3 Simulation of the force applied by interaction
of explosion-generated water wave and wall
As mentioned earlier, in this section, we characterize
the surface wave produced by a fictitious explosion
with equivalent energy to the solitary waves of Section
(3.1.2) in the numerical flume. We compared the
characteristics of wave propagation and interaction
force with previous results in order to measure and
describe variations in the amount of force on the wall
and water level in similar locations. For quantifying the
energy of reproduced tsunami-like waves due to
explosion, by referring to Eq.(3) the value of 𝜂0 × 𝑅0
should be estimated for each case. Then, by use of
Eq.(5) simply 𝐻𝑚𝑎𝑥 × 𝑟 could be evaluated and if this
value be inserted in the equations 6 and 7, the margin of
charge weight "𝑊" may be calculated. At last, by
applying charge weight value into Eq.(4), the radius of
crater at edge of lip 𝑅0 and then 𝜂0 are found easily
(see Table 4). Now the initial boundary condition of
water free surface may be defined by Eq.(1) and
considered in numerical model.
For applying this procedure to the laboratory setup
which was performed by Yu Yao et al., an explosion is
assumed at the middle of the flume that produces a
wave with crater depth and lip height of 𝜂0 = 0.23 𝑚
and crater span of 𝑅0 = 1.3 𝑚 at the maximum height
of the lips, which propagates upstream and downstream
along the flume’s length (see Figures 18 and 19). All
other properties such as the geometry of flume, water
depth, etc. are considered same as previous.
Table 4: Characteristics of initial boundary condition of artificial explosion generated waves
Case Study Energy
(kJouls)
𝜼𝟎 × 𝑹𝟎
(m2)
Eq.(3)
𝑯𝒎𝒂𝒙 × 𝒓
(m2)
Eq.(5)
Average
Charge Weight
(kg)
𝑹𝟎
(m)
Eq.(4)
𝜼𝟎
(m)
A
Robertson et al. (2011) 177 5.9 7.25 3.83 5.4 1.5
B
Yu Yao et al. (2017) 0.44 0.6 0.75 0.03 1.3 0.23
Fig. 18: The Initial geometry of the explosion generated water wave for Yu Yao et al. Test case
Fig. 19: Enlargement of primary surface wave generated by underwater explosion in
SPH model for Yu Yao et al. Test case
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
60
Figure 20 shows the time history of water level
variations at two gauges in deep and shallow regions of
flume. Figures 21 and 22 display the time history of
force on the vertical wall and the snap shots of wave
location during the insertion of peak forces at 25-meter
and 50-meter, respectively. The major modeling and
simulation parameters are listed and described in Table
(5).
Fig. 20: Time history of the water level due to underwater explosion at gauge G1 (line) and gauge G2 (dash line)
Fig. 21: Time history of the wave force on the vertical coastal wall generated
by the explosion (line) and the solitary wave (dash line)
-0.5
-0.25
0
0.25
0.5
0 10 20 30 40 50
η/h
t/(h/g)
gauge G1
gauge G2
-10.00
0.00
10.00
20.00
30.00
40.00
50.00
4.0 5.0 6.0 7.0 8.0 9.0 10.0
Forc
e (N
)
Time (sec)
Force due to underwater explosion
Force due to solitary wave
(a
(b)
(c)
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
61
Fig. 22: Snap shots of wave impact on wall at (a) t=6.74 sec, (b) t=7.5 sec, (c) t= 8.78 sec
Table 5: Characteristics of the main input parameters in the 2D SPH model used for comparison of simulations
Executive Parameters Modeling Parameters
Symplectic Time step
algorithm 0.1
Particle size
(m)
Wendland Kernel type 200 Model overall
size (m)
Explosion
model
Tsunami
model GPU run time
(sec.)
Explosion
model
Tsunami
model Total Number
of Particles 1422 1224 45374 46702
Explosion
model
Tsunami
model CPU run time
(sec.) 0.2 CFL number
3560 3062
According to the plotted water level variation
diagrams, solitary wave resulted in conservation of
depth at the points close to the wave initiation point, but
explosion-generated water wave led to depth reduction
at these points. In other words, by moving away from
the source of the wave generator, the explosion waves
which was simulated by an initial parabolic water level,
in comparison with a solitary wave as a representative
of the tsunami wave, show attenuation of wave
amplitude and force, when they are propagated on a
flatbed.
4. Conclusions
In this research, we used the DualSPHysics
numerical code to simulate and examine the forces
induced by the waves generated by an underwater
explosion on a vertical coastal wall. To study the
behavior and effects of the explosion-generated water
wave, the results were investigated in comparison with
the effects of a solitary wave of equal energy. The
following conclusions are deduced from the results:
(a) (b)
(c)
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
62
(a) The validation results show the ability of the
DualSPHysics code and the SPH method to
accurately simulate the free surface
variations due to an underwater explosion. In
this study, we used two sets of laboratory
data to assess the ability of the SPH method
in the simulation of phenomena for which
recorded field or laboratory data have
variations and inherently these types of
experiments may record with large error
tolerances. As shown in the validation part of
the study, the matching of numerical
simulation results with laboratory values
demonstrated the ability of the numerical
model to capture the phenomenon with ill-
conditioned data.
(b) Although the SPH model used in this study
has an accuracy order of 1 ≤ 𝒪(r) ≤ 2,
which is more inclined toward 1, the model
verification phase revealed the ability of this
method to predict a complicated nonlinear
phenomenon of high complexity with proper
precision. On the other hand, this method can
reproduce physical characteristics of
complex CFD problems in a short analysis
time, which is a valuable benefit for
engineers to understand the impacts of such a
dangerous condition for coastal structures.
(c) Due to the behavior of the explosion-
generated water waves, caused by the
lowering of the free surface of the water at
the “ground zero” (i.e. the location of the
center of an explosion), the water elevation
was decreased and then increased by the
progressive negative wave in the vicinity of
the obstacle. These statuses produced
reduction and increase of pressure and force
on the wall respectively.
(d) In fact, this effect is different from the forces
resulting from the collision of the tsunami
waves (solitary) on the coastal walls, in the
opposite way. Thus, in identifying the risk of
surface waves generated by underwater
explosion than the tsunami waves, it can be
inferred that this type of waves initially show
a decrease in water level, resulting in a
reduction of pressure and force on the wall,
and then these values are an uptrend.
(e) Obviously, both time history of forces have
approximately the same peak value, but total
impact time for tsunami waves is longer than
those related to explosion generated waves.
This result shows that the explosion-induced
waves are very dissipative and their effects
are important for near field objects.
(f) The force exerted on the vertical wall by
explosion-generated water waves has a
periodic nature and attenuates with time. In
contrast, the force of a solitary wave with an
equal peak acts on the structure in a short
time period and then settles down.
(g) The explosion-generated water wave has a
stronger perturbation character and remains
imposed on the structure for a longer period
of time. Obviously, repeating the simulation
by taking into account the effect of water
particles released into the surrounding area
and their return to the water surface (which
may produce secondary waves) will result in
longer duration of turbulence in the water
level and more fluctuations in the force
exerted on the wall.
(h) For the 2-D models, utilization of GPU-
based method yields a better computational
speed than the parallel computation in multi-
core CPU. In the case of this study, using the
former approach reduced the computational
time by at least 60%. This feature could be
particularly advantageous for the processing
of more complex models, especially those
involving the effects of sudden and
impulsive loads, such as explosion generated
water waves, and save the engineers precious
time in design as well as damage estimation.
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
63
(i) For the 3-D models, the use of GPU-based
method led to 75-90% reduction in the
execution time, which demonstrates the
greater efficiency of simulation with GPU
and the SPH technique for more complex
problems. This significant reduction of run
time also signifies the nonlinear relationship
of processing time with the model
dimensions and the effect of large scale of
analyses that have a greater number of
particles on the computational time and the
performance of the optimization simulation
algorithm in the SPH method.
Acknowledgements
Funding for this research was provided by the
Iranian National Center for Ocean Hazards (INCOH),
affiliated to Iranian National Institute for
Oceanography and Atmospheric Science (INIOAS).
The authors appreciate this support.
References
Altomare, C., Crespo, A.J.C., Rogers, B.D.,
Dominguez, J.M., Gironella, X. and Gomez-
Gesteira, M., 2014. Numerical modelling of
armour block sea breakwater with smoothed
particle hydrodynamics. Computers & Structures,
130, 34-45.
Altomare C., Crespo A. J. C., Domínguez J. M. et
al., 2015. Applicability of smoothed particle
hydrodynamics for estimation of sea wave impact
on coastal structures. Coastal Engineering, 96(2),
1-12.
Altomare, C., Crespo, A.J.C., Dominguez, J.M.,
Gomez-Gesteira, M., Suzuki, T. and Verwaest, T.,
2015. Applicability of Smoothed Particle
Hydrodynamics for estimation of sea wave impact
on coastal structures. Coastal Engineering Journal,
96, 1-12.
Altomare C., Domínguez J. M., Crespo A. J. C. et
al., 2017, Long-crested wave generation and
absorption for SPHbased DualSPHysics model.
Coastal Engineering, 127(7), 37-54.
Barreiro, A., Crespo, A.J.C., Domínguez, J.M. and
Gómez-Gesteira, M., 2013. Smoothed Particle
Hydrodynamics for coastal engineering problems.
Computers & Structures, 120, 96-106.
Bottin, R.R., 1990. Impulsive waves generated by
falling weights in shallow water. Final Technical
Report CERC-90-9, Coastal Engineering Research
Center, Department of the army, Waterways
Experiment Station, Corps of Engineers, 3909
Halls Ferry Road, Vicksburg, Mississippi 39180-
6199.
Capone, T.; Panizzo, A.; Dalrymple, R., 2007.
Accuracy and Stability of Numerical Schemes in
SPH. Secon SPHERIC Workshop. Madrid.
Chan, E. S., 1994. Mechanics of Deep Water
Plunging-wave Impacts on Vertical Structures.
Coastal Engineering, 22, 115-133.
Chanson, H., 2006. Tsunami Surges on Dry Coastal
Plains: Application of Dam Break Wave
Equations. Coastal Engineering Journal, 48(4),
355-370.
Charlesworth, G., 1945. Surface waves produced by
firing underwater 32-lb charges of Polar Ammon
Gelignite at various depths. Final Report, Road
Research Laboratory, London.
Choi, B.H.; et al, 2003. Simulation of the trans-
oceanic tsunami propagation due to the 1883
Krakatau volcanic eruption. Natural Hazards Earth
System Science, 3, 321–332.
Cole, R. H., 1948. Underwater Explosions, Princeton
University Press, Princeton, N.J.
Crawford, D. A.; Mader, C. L., 1998. Modeling
asteroid impact and tsunami. The science of
Tsunami Hazards, 16, 21–30.
Crespo A. J. C., Domínguez J. M., Rogers B. D. et
al., 2015. DualSPHysics: open-source parallel
CFD solver on SPH. Computer Physics
Communications, 187(2), 204-216.
Crespo, A.J.C., Altomare, C., Dominguez, J.M.,
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
64
Suzuki, T., Verwaest, T. and Gomez-Gesteira, M.,
2015. SPH modeling in coastal engineering, E-
proceedings of the 36th IAHR World Congress,
Hague, The Netherlands.
Domínguez, J.M., Crespo, A.J.C. and Gómez-
Gesteira, M., 2013. Optimization strategies for
CPU and GPU implementations of a smoothed
particle hydrodynamics method. Computer Physics
Communications, 184(3), 617-627.
Director, M. N.; Dabora, E. K., 1977. An
experimental investigation of variable energy blast
waves. Acta Astronautica. 4, 391-407.
El-Solh, S., 2012. SPH Modeling of Solitary Waves
and Resulting Hydrodynamic Forces on Vertical
and Sloping Walls. Thesis for Master of Applied
Science in Civil Engineering, Department of Civil
Engineering, University of Ottawa.
Gómez-Gesteira, M.; Dalrymple, R. A.; 2004.
“Using a 3D SPH method for wave impact on a
tall structure, Journal of the Waterway Port,
Coastal and Ocean Division, 130(2): 63-69.
Gómez-Gesteira, M., Rogers, B.D., Crespo, A.J.C.,
Dalrymple, R.A., Narayanaswamy, M. and
Dominguez, J.M., 2012. SPHysics - development
of a free-surface fluid solver – Part 2: Efficiency
and test cases. Computers & Geosciences, 48,
300-307.
Greenberg, D. A.; et al., 1993. Modeling the tsunami
from the 1917 Halifax harbor explosion. The
International Journal of the Tsunami Society,
11(2), 67-80.
Hartmann, G. K.; Laboratory, N. S. W. C. W. O.,
1976. Wave making by an underwater explosion.
Naval Surface Weapons Center, White Oak
Laboratory.
Houston, J. R.; Chou, L. W., 1983. Numerical
modeling of explosion waves. Final Technical
Report HL-83-1, Waterways Experiment Station,
Hydraulics Laboratory, Vicksburg, Mississippi.
Hsiao, S.C.; Lin, T.C., 2010. Tsunami-like solitary
waves impinging and overtopping an impermeable
seawall: Experiment and RANS modeling. Coastal
Engineering, 57(1): 1-18.
Jiang, C.; Liu, X.; Yao, Y.; Deng, B.; Chen, J. 2017.
Numerical Investigation of Tsunami-Like Solitary
Wave Interaction with a Seawall. Journal of
Earthquake and Tsunami 11(1), 1-18.
Jordaan, J. M., 1966. Model studies of impulsively
generated water waves. Proceedings of 10th
Conference on Coastal Engineering, Tokyo, Japan,
1(10), 434-454.
Kedrinskii, V. K., 2004. Hydrodynamics of
Explosion: Experiments and Models. Springer,
Berlin.
Kajiura, K., 1963. The leading waves of tsunami.
Bulletin of the Earthquake Research Institute, 41,
535-571.
Kranzer, H. C.; Keller, J. B., 1959. Water Waves
Produced by Explosions. Journal of Applied
Physics, 30 (3), 398-407.
LeMehaute, B., 1971. Theory of Explosion-
Generated Water Waves. Advances in
Hydroscience, Academic Press, New York, 7, 1-
79.
LeMehaute, B., 1976. An introduction to
hydrodynamics and water waves. Springer
Science, Business Media, New York.
LeMehaute, B.; Wang, S.; Lu C. C., 1987, Cavities,
domes and spikes. Journal of Hydraulic Research,
25 (5), 583-602.
LeMehaute, B. and Wang, S., 1996, Water waves
generated by underwater explosion, Advanced
Series on Ocean Engineering - Volume 10, World
Scientific, River Edge.
Lin, T.C.; Hwang, K.S.; Hsiao, S.C.; Yang, R.Y.,
2012. An experimental observation of a solitary
wave impingement, run-up and overtopping on a
seawall. Journal of Hydrodynamics, Seri B, 24(1):
76- 85.
Liu, M. B.; et al, 2003. Smoothed particle
hydrodynamics for numerical simulation of
underwater explosion. Computational Mechanics,
30(2), 106–118.
Mader, C. L.; Gittings, M., 2003. Dynamics of water
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
65
cavity generation. The science of Tsunami
Hazards, 21(2), 91-102.
Monaghan, J.J., and Gingold, R.A., 1983, Shock
simulation by the particle method SPH, Journal of
Computational Physics, 52, 374-389.
Noda, E.K., 1971, Water waves generated by a local
surface disturbance, Journal of Geophysical
Research, 76, 7389-7400.
Penney, W. G., 1945. Gravity waves produced by
surface and underwater explosions. Underwater
Explosion Research: The Gas Globe, U.S. Office
of Naval Research, Washington, D. C, 2, 679-693.
Prins, J. E., 1958. Characteristics of waves generated
by a local disturbance. Transactions American
Geophysical Union, 39(5), 865-874.
Pu, J. H.; Shao, S.; 2012. “Smoothed Particle
Hydrodynamics Simulation of Wave Overtopping
Characteristics for Different Coastal Structures”.
The Scientific World Journal, Volume 2012.
Robertson, I.N.; Paczkowski, K.; Riggs, H.R.;
Mohamed, A.; 2011. Tsunami bore forces on
walls. Proceedings of the ASME 2011 30th
International Conference on Ocean, Offshore and
Arctic Engineering, OMAE2011-49487, June 19-
24, 2011, Rotterdam, the Netherlands.
Rogers, B.D.; Dalrymple R. A.; Stansby, P. k; 2010.
“Simulation of caisson breakwater movement
using 2-D SPH”. Journal of Hydraulic Research,
48(Extra Issue), 135-141.
Sakurai, A., 1965. Blast Wave Theory. Included in
Basic Developments in Fluid Dynamics (Edited by
M. Holt), Academic Press, 1, 309-374.
Simskin, T.; Fiske, R. S., 1983. Krakatau 1883 – the
volcanic eruption and its effects. Smithsonian
Institution Press, Washington.
Van Dorn, W. G., 1964. Explosion-generated waves
in water of variable depth. Journal of Marine
Research, 22(2), 123-141.
Van Dorn, W. G., 1961. Some Characteristics of
Surface Gravity Waves in the Sea Produced by
Nuclear Explosions. Journal of Geophysical
Research, 66(2), 3845-3862.
Van Dorn, W. G.; LeMehaute, B.; Hwang, L. S.,
1968. Handbook of Explosion-Generated Water
Waves. Volume I: State-of-the-Art. Tetra Tech
Final Report TC-130, Pasadena, California.
Whalin, R. W., 1965. Water waves produced by
underwater explosions: Propagation theory for
regions near the explosion. Journal of Geophysical
Research, 70, 5541–5549.
Whalin, R. W., 1967. Analysis of data on water
waves, Volume I: Theoretical developments
related to water waves generated by explosions in
deep water. Tetra Tech Final Report No. 105-1,
Waterways Experiment Station, Vicksburg,
Mississippi.
Wang, S. B.; LeMehaute, B.; Lu , C. C., 1988. Effect
of Dispersion on Impulsive Waves. Journal of
Marine Geophysics Research, 9 (1), 95-111.
Wang, Q. X.; Yeo, K. S.; Khoo, B. C.; Lam K. Y.,
1996. Nonlinear interaction between gas bubble
and free surface. Computational Fluids 25(607).
www.dual.SPHysics.org, Official site of
DualSPHysics software. Xie, W. F.; et al, 2007. The Simulation of Cavitating
Flows Induced by Underwater Shock and Free
Surface Interaction. Applied Numerical
Mathematics, 57, 34-745.
Safiyari et al. / GPU-SPH Simulation of Tsunami-like wave intraction…
Journal of the Persian Gulf (Marine Science)/Vol .7/No .24/June 2016/23/43-66
Journal of the Persian Gulf
(Marine Science)/Vol. 7/No. 24/ June 2016/23/43-66
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020
Dow
nloa
ded
from
jpg.
inio
.ac.
ir at
17:
53 IR
DT
on
Sun
day
Aug
ust 3
0th
2020