International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 2
Number 2
June 2008
137
Numerical Simulation of Nonlinear Water
Wave Problems
D.C. Lo
Institute of Navigation Science and Technology, National Kaohsiung Marine
University, Kaohsiung, Taiwan
Jia-Shen Hu
Department of Shipping Technology, National Kaohsiung Marine University,
Kaohsiung, Taiwan
I-Fu Lin
Institute of Navigation Science and Technology, National Kaohsiung Marine
University, Kaohsiung, Taiwan
ABSTRACT: The main purpose of present paper aims at the establishment of a numerical model for solving
the nonlinear water wave problems. The model is based on the Navier-Stokes equations with the consideration
of a free-surface through the streamfunction-vorticity formulation. The main advantage of the streamfunction-
vorticity formulation is that pressure field can be eliminated from the Navier-Stokes equations. To
demonstrate the model feasibility, the present studies are first concentrated on problems including the
collision of two solitary waves with different amplitudes, and the overtaking collision of two solitary waves.
Then, the model is also applied to a solitary wave passes over the submerged obstacle in a viscous fluid.
Finally, the application of present study is also to simulate the generation of solitary waves by underwater
moving object. All examples give very promising results, those applications reveal that present formulation is
a very powerful approach to simulate the fully nonlinear water wave problems.
1 INTRODUCTION
The applications of free-surface flows find wide
range of navigation science in the design of ship-
waves interaction problems. The last one decade
have seen growing importance placed on research in
free-surface flows simulating. Many analytical
methods have been developed to solve traveling and
ship-wave problems associated with inviscid fluids
as well as for the solution of linear and low order
nonlinear problems. Cao and Beck [1] has employed
a numerical method to calculate solitary waves
generated in an inviscid fluid by a submerged cylinder
using the Laplace equation and fully nonlinear free
surface boundary conditions. A Taylor-Galerkin
method was proposed by Ambrosi and Quartapelle
[2] to compute the evolution of water waves with
moderate curvature of the free-surface shallow water
equations. However, their proposed scheme was
extended to study 2D solitary wave passing over a
circular cylinder in an inviscid fluid. Lo and Young
[3] developed a numerical mode based on velocity-
vorticity formulation to solve several free-surface
flow problems. The aim of present study is to
develop a more general approach to handle inviscid-
viscous free-surface flows. When the inviscid model
is used, the Poisson-type streamfunction equation
and fully-nonlinear free surface boundary conditions
are governed forthe solution of the free-surface
flow. Moreover, it is necessary to deal with viscous
free-surface model in order to comprehend the
complexities of the phenomenon including the
viscous effects. Thus the vorticity transport equation,
Poisson-type streamfunction equation and fully-
nonlinear free surface boundary conditions
should to be solved. The accuracy of the solution
for the inviscid-viscous model depends on the
138
representation of actual free surface unknown
values.
The numerical treatments of free-surface
problems involve the tracking of the moving free-
surface boundary during the flow transients. The
arbitrary Lagrangian-Eulerian (ALE) coordinate
system [4] is used to track the free-surface position
on the moving free-surface and the interior
computational domain at each time step. It is
important to keep the computation of a free-surface
position in perspective. In this study, the nodal points
can be arbitrarily controlled in order to get finer
mesh distribution during computations.
The main difficulty for the solution of the
viscous, incompressible free-surface flows indicate
that the free-surface boundary conditions are not
known priori. In this study, we delivered an accurate
and effective scheme based on streamfunction-
vorticity formulation and its application to the
simulation of wave-wave and wave-structure
interaction. In the computational fluid dynamics, the
fractional step method is an accurate and efficient
scheme for the solution of Naver-Stokes eqations. In
the present study, the fractional step is used for
the discretization of the governing equations and
free-surface boundary conditions (FSKBC, FSDBC).
The above mentioned of governing flow equations,
those equations are coupled and nonlinear, hence an
iterative numerical scheme is adopted in order that
we could obtain a significant saving in computer
memory.
We present an inviscid-viscous free-surface
model using the finite element discretization for the
interior of the domain and finite difference
discretization for a free-surface. The detailed
contents were addressed in the following.
2 MATHEMATICAL FORMULATION
2.1 Incompressible viscous flow
The partial differential equations of the viscous
incompressible fluid are governed by the Navier-
Stokes equations. The corresponding non-
dimensional streamfunction-vorticity form of
Navier-Stokes equations in ALE form can be
expressed as follows
Streamfunction Poisson equation
2
ψω
∇=
(1)
Vorticity transport equation
2
1
ˆ ˆ
( )( )
Re
uv
ty x x y
ωψωψω
ω
∂∂∂
+−+−=
∂∂
(2)
where
ψ
is the streamfunction,
ω
is the vorticity,
Re
is the Reynolds number,
ˆ ˆ
(,)uv
is the mesh
velocity in the
direction, respectively.
Equations (1) and (2) are known as the Navier-
Stokes equations in streamfunction-vorticity form.
We seek a solution in the domain
, which satisfies
the initial conditions, with no-slip or slip velocity
boundary conditions depending on the need on the
solid boundary
Γ
of
.
2.2 Boundary conditions
Consider a two-dimensional, viscous and
incompressible fluid. In the present study, the free-
surface flow is considered as a two-phase flow. It is
assumed that the adjacent fluids are impermeable at
the interface with constant density and viscosity. The
main difficulty of the free-surface flow is that the
position of the free-surface boundary is not known a
priori. The boundary-fitted coordinates system is
used to solve the equations for the free-surface
boundary conditions in streamfunction-vorticity
form. The relationships among velocity,
streamfunction and vorticity can be defined as
(,) ( , )
yx
uv
ψψ
=
,
yx
vu
ω
=
. The arbitrary
physical space
(, )xy
can be general mapped to a
normalized computational domain
(,)
ξη
with the
help of the boundary-fitted coordinates system.
According to the statement of a material surface,
a particle on the surface must remain on the surface
itself. By satisfying the above kinematic condition in
an arbitrary frame of reference to be
ˆ
()
tx
h v u uh=−−
(3)
In the above equation, the frame of reference
moves with a free surface in the vertical direction
and
ˆ
u
is either zero or equal to
u
.
The free surface dynamic boundary condition
(FSDBC) represents the continuity of the stresses on
the free-surface and is obtained by force balance
equations. Neglecting the surface tension effect on
the free-surface, the FSDBC in an arbitrary frame of
reference can be obtained by combining momentum
equations to yield
2
1
ˆ ˆ
( ( ) ( ) ( ))
Re
11
ˆ
ˆ
( ( ) ( ) ( )) 0
Re
t x y xx yy
t x y xx yy
r
u u uu v vu u u x
v u uv v vv v v y
F
ξ
ξ
+ +− + +
+ +− + + =
(4)
As far as initial condition is concerned, an initial
solitary wave is imposed as present in Gtimshaws
third order formulas [5] in the free-surface profile
and streamfunction.
139
3 NUMERICAL FORMULATION
For the solution of 2D free-surface flow using
streamfunction-vorticity formulation, the
quadrilateral finite element is used to discretize the
Poisson-type streamfunction equation and the
vorticity transport equation. The free-surface
boundary conditions are solved by the FDM.
Table 1. Comparison of present and Su and Mirie results [6] of
the maximum amplitude run-up during the collision
,AB
0.2,0.2
0.3,0.2
0.4,0.2
0.5,0.2
0.6,0.2
Present
0.4298
0.5436
0.6604
0.7866
0.9420
Su and
Mirie
0.4240
0.5375
0.6520
0.7675
0.8840
Error
(%)
1.37
1.13
1.29
2.49
6.56
4 RESULTS AND DISCUSSIONS
4.1 Validation of the numerical model for two
solitary waves head-on collision
It is interesting to note that the solitary waves which
retain their identity upon collision are called
solitons. The topics of solitons had a large impact on
various branches of modern physics, and this topic
attracts many investigators to study. In this section,
the results for collisions of two opposite solitary
waves using the present numerical scheme were
verified. Su & Mirie [6] study the head-on collisions
of two solitary waves generated using a perturbation
method. The flow is an inviscid and incompressible
fluid. In a similar work, Mirie & Su [7] depicted the
collisions of two solitary waves to confirm the
existence of a secondary wave as predicted in their
previous work.
0.00 10.00 20.00 30.00 40.00
5
10
15
20
t
x
(a)
0.00 10.00 20.00 30.00 40.00
5
10
15
20
t
x
(b)
Fig. 1. Evolution of free surface position of different
amplitudes. (a)
0.3, 0.2AB= =
(b)
0.6, 0.2AB= =
We consider the collision of two solitary waves
with different amplitudes. Calculations were carried
out for initial heights of a solitary wave at the left-
hand side given by the
expression,
0
A
=0.2,0.3,0.4,0.5,0.6. The initial
amplitude of a solitary wave at the right-hand side
was assumed to be equal to 0.2. The grid size was
0.1, 0.02xt= ∆=
and the range consisted of 8000
quadrilateral elements. The Reynolds number is
Re 60,000=
for this case. In this moving grid, the
interval of
x
is fixed but
y
is changed by the
FSKBC. The maximum amplitudes of the head-on
collision of two solitary waves were shown in Table
1. Good agreement with the solution obtained by Su
& Mirie [6] was depicted in above table. Table 1
provides the list of 5 values that were computed for
verifying the model in this study. The present
numerical model could predict the maximum run-up
with less than 2% error in the cases of
0.2,0.3,0.4A =
and
0.2B =
. However, the error
turned out to be a 6.56% at
0.6A =
,
0.2B =
. A
partial explanation for this may lie in the fact that the
computation of present case the fully nonlinear
boundary conditions are considered. However, the
result obtained by Su & Mirie [6] was to assume the
weakly nonlinear effect on the free surface for this
case. Figure 1 shows the evolution of free surface
profiles of the two solitary waves during the
collisions. The high nonlinear interaction has been
observed during the collisions between the two
solitary waves. We also found that after the
collision, the solitary waves recuperate almost all of
their original amplitudes in the time range of the
transient. Also, the solitary waves have experienced
phase change and were trailed by a dispersive wave
train when the two solitary waves head-on collide in
our computation.
140
0.00 40.00 80.00 120.00 160.00 200.00
t
120
1
x
Fig. 2. Evolution of free surface position of different
amplitudes. (
0.6, 0.2AB= =
)
4.2 The overtaking collision of two solitary waves
As mentioned before of head-on collision of two
solitary waves, now we consider the simulation of
the 2D overtaking of two solitary waves of
amplitudes 0.6, 0.2. The initial amplitudes of the two
solitary waves were assumed to be equal to 0.6 on
the left-hand side and 0.2 on the right-hand side. The
positions of the two solitary waves were initially
fixed at
10x =
and
20x =
in the same direction.
The present simulation of free-surface flow was
tested for inviscid model. In the present
computations, the non-dimensional parameter with
finite depth, D equal to 1 and length in the
x
-
direction equal to 200 were assumed, respectively. A
total number of 16,000 elements and 17,017 nodes
(1001, 17 divisions in the
x
,
y
directions,
respectively) with a time step
t
=0.02 were used for
computation in the present case. Figure 2 depict the
time evolution of the interaction of two solitary
waves. The highly nonlinear behavior during the
overtaking process between two opposite solitary
waves is presented in Figure 2. However, both
solitary waves emerge completely unscathed.
0.00 10.00 20.00 30.00 40.00
10
15
20
5
x
Fig. 3. Evolution of free surface elevation at different time for
Re 60,000=
4.3 Solitary wave passing over a submerged
structure
The computation for a fluid-structure interaction
with a free-surface was carried out for a solitary
wave passing through the submerged structure. We
assume the dimensionless length of the mean water
level as 1 and length as 42 as well as initial
amplitude of the solitary wave as 0.4. The initial
condition of a solitary wave was fixed in
10x =
of
the water channel. The flow was considered to be
viscous. For the present case of viscous model, the
fluid is assumed to have small viscosity (Re=60,000,
Fr=1), our goal is the observation of the separated
flow pattern for the viscous model. Figure 3 lists the
time evolution of a solitary wave passing through the
submerged structure in present simulation.
In order to plot the vortex motion generated
during wave-structure interactive process. The
streamline were plotted and were shown in Figure 4.
The local velocity distribution near the structure was
given in Figure 5. The gradual development of the
recirculating flow regimes at the downstream of the
structure and was clearly predicted by the present
numerical study. The recirculating flow generated at
the trailing edge of the structure was clearly
observed in Figure 5. The results reflected in Figure
5 indicate that the viscosity plays an important
role to affect the interactive processes between
the solitary wave and the solid structure. In general,
the vortical flow is always grown up on the rear
surface of the structure while the solitary wave
passes through the submerged structure in viscous
fluid.
x
16 18 20 22 24 26
t=8
x
16 18 20 22 24 26
t=12
x
16 18 20 22 24 26
t=14
Fig. 4. Evolution of streamline distribution at different time
141
x
y
t=10
x
y
t=12
x
y
t=16
Fig. 5. Local velocity distribution at different time
4.4 The generation of solitary wave by underwater
moving object
The final test case illustrates the capability of 2D
free surface algorithm for a succession of solitary
waves generated by underwater moving object. The
submerged object is moving with a constant velocity
in the negative x-coordinate direction in shallow
water. The layout of geometry and conditions is well
illustrated in Figure 6. We assume the constant
velocity of underwater object is moving in a channel
[ ] [ ]
0,70 0,1×
with the object as shown in Figure 6,
initially the region of object is
2
max
cos ( 40) , ( 40)
(, )
22
0,
LL
Bx x
bxy
L
otherwise
π


−< <

=




(5)
For the case of present used study, we assume
max
0.15B =
and
1L =
. The computation for a
submerged moving object in a channel is assumed to
be 70 in the
x
direction and with the upstream
dimensionless depth of 1 in the
y
direction.
The computation of the present case is used with
a total number of 9,600 elements and 10,217 nodes
with a time step
0.02t∆=
and
Re 30,000=
.
Figure 7 depicts the local velocity and vorticity
distributions at t = 100 for the flow past the
submerged object. The flow circulation behind
the object is seen in Figure 7, in which the region
of the vortex structure near the object is delineated.
Further, the streamline contours at different time are
also revealed in figure 8. Finally, the evolution
of free surface elevation at different time for
the propagation of solitary waves is shown in
figure 9. It can be observed that the generation and
propagation of the upstream advancing solitary wave
become more dominated as time increases.
u=1
Fig. 6. Layout of an underwater moving object
x
y
40 40.5 41 41.5 42
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
t=100
142
x
y
40 40.5 41 41.5 42
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
40
24.4444
8.88889
-6.66667
-22.2222
-37.7778
-53.3333
-68.8889
-84.4444
-100
t=100
Fig. 7. Local velocity and vorticity distribution at t=100
5 CONCLUSIONS
A numerical model by combing FEM and FDM for
free-surface flow problems was proposed using the
Navier-Stokes equations in streamfunction-vorticity
form. The present study has extended to an inviscid-
viscous model for free surface flow problems.
The accuracy of results obtained we have is
excellent. Also, both inviscid and viscous models
play an important role to free surface flows.
However, the key issue in the viscous model is to
find the accurate solution for separated flow pattern
of a solitary wave passing over the submerged
structure. It shows that the streamfunction-vorticity
formulation gives accurate and correct results for
flow problems with a free surface. We can conclude
with certainty that both inviscid and viscous models
are able to simulate non-linear water waves in
present work.
ACKNOWLEDGMENTS
The work reported in this chapter was supported
by the National Science Council, Taiwan under
the grant no. 95-2221-E-022-020. It is greatly
appreciated.
t=30
t=60
Fig. 8. Streamline distributions at different time
0.00 20.00 40.00 60.00
0
120
T
X
Fig. 9. The evolution of free surface elevation at different time
REFERENCES
[1] Y. Cao and R.F. Beck, Numerical computations of two-
dimensional solitary waves generated by moving
disturbances, Int. J. Numer. Meth. Fluids, 17 (1993) 905-
920.
[2] D. Ambrosi and L. Quartapelle, A Taylor-Galerkin method
for simulating nonlinear dispersive water waves, J. Comput.
Phys., 146 (1998) 546-569.
[3] Lo, D.C., and Young, D. L. (2004). Arbitrary Lagrangian--
Eulerian finite element analysis of free surface flow using
a velocity-vorticity formulation.J. Comput. Phys., 195,
175-201.
[4] C.W. Hirt, A.A. Amsden and J.L. Cook, An arbitrary
Lagrangian Eulerian computing method for all flow speeds,
J. Comput. Phys., 14 (1974) 227-253.
[5] R. Grimshaw, The solitary wave in water of variable depth,
part 1. J. Fluid Mech. 46 (1971) 611-622.
[6] C.H. Su and R.M. Mirie, On head-on collision between two
solitary waves, J. Fluid Mech. 98 (1980) 509-529.
[7] R.M. Mirie and C. H. Su, Collisions between two solitary
waves, Part 2: A Numerical Study, J. Fluid Mech. 115
(1982.