Proceedings of the ASME 2013 32nd International Conference on Ocean, Offshore and Arctic Engineering
OMAE2013
PREDICTION OF WATER WAVE PROPAGATION USING COMPUTATIONAL FLUID
DYNAMICS
Mohammadreza Javanmardi
Australian Maritime College, University of
Tasmania
Launceston, TAS, Australia
Giles Thomas
Australian Maritime College, University of
Tasmania
Launceston, TAS, Australia
ABSTRACT
In this study the influence of grid effects on free surface
deformations behind a ship-like moving vessel were
investigated. To determine the influence of grid effects on the
water wave propagation, various grid domains with different
quality parameters were produced. Simulations were conducted
for a moving pressure source and the free surface around the
moving body captured. Then three-dimensional numerical
results for different grids, in both the near and far field, were
compared with experimental data over a range of speeds. The
experimental data were obtained using tank tests on a pressure
source model at the Australian Maritime College. Wave probes
at different lateral distances captured the generated wave
parameters. The study revealed that the results of numerical
simulation of water wave propagation depend on the grid
parameters and geometrical mesh quality.
Keywords: CFD, wave propagation, mesh quality
INTRODUCTION
Simulation of fluid flows with a free surface is of interest in
many engineering applications. In maritime hydrodynamics,
prediction of the characteristics of generated waves by moving
vessels in open or restricted waters can be critical, since such
waves can cause many problems for both other vessels and the
June 9-14, 2013, Nantes, France
OMAE2013-10160
Jonathan Binns
Australian Maritime College, University of
Tasmania
Launceston, TAS, Australia
Martin R. Renilson
Australian Maritime College, University of
Tasmania
Launceston, TAS, Australia and
Higher Colleges of Technology, UAE
environment [1]. Generating surge waves, disturbing other
vessels in ports and harbors, shoreline erosion and its impact on
marine life are some of the most important issues in this field
[2]. Therefore, tools for predicting the near-field ship generated
waves and their propagation to far field with accuracy are
needed. Most research has been conducted using theoretical [3]
or experimental [4, 5] approaches.
Computational Fluid Dynamics (CFD) is a key method that
is used to predict free-surface parameters and wave
propagation. Typically, the finite volume method has been used
with significant far-field damping for floating structures,
reducing significantly the accuracy of predicted wave heights
[6]. Regardless of the variety of CFD approaches, e.g.
discretization methodologies and solvers, the quality of the
grids used has a large effect on the final results. To predict flow
parameters, the density of cells in a computational grid needs to
be fine enough, while the extreme number of cells require much
more simulation time, memory requirement and computational
cost, therefore there is a need to trade off between accuracy and
computational effort. The proper geometrical mesh quality is
significant to reduce discretization and computational errors.
Mesh orthogonality, expansion and aspect ratio are major
measures of mesh quality. Improper geometrical mesh quality
may have considerable influence on the propagation of
numerical errors and result in an imprecise solution [7].
In this work, the influence of grid quality effects on free
surface deformations behind a ship-like moving vessel were
1 Copyright © 2013 by ASME
investigated. This was performed to simulate scenarios
generated in the search for producing artificial surfing waves [8,
9].
Other numerical techniques have previously been used to
study the creation of artificial surfing waves. Schipper [10] and
Vries [11] conducted preliminary numerical modeling of a
pressure source moving along a linear track, using panel method
modeling by DELKELV and 3D modeling by Navier Stokes
solver FINLAB in 2007. Then Essen [12] used a non-linear
potential flow code (RAPID) to predict the wave height for a
pressure source. Unfortunately these numerical approaches
were not able to reproduce the experiment results.
Subsequently, Javanmardi et al. [9] conducted an initial
numerical study with Fluent CFD software and showed that
Figure 1: Wavedozer model attached to the towing tank carriage
finite volume can accurately predict the wave parameters and
propagation characteristics.
EXPERIMENTAL TESTS
The wavedozer was tested in the AMC towing tank, which
NUMERICAL SIMULATIONS
is 100 m in length and 3.5 m in width. The water depth was set
In this study, the CFD software FLUENT version 12.1 was
at 1.5m for all of the tests. The model was attached to the
used as the flow solver. The governing equations are three-
carriage using a two post towing system, with the model fixed
dimensional Reynolds Averaged Navier-Stokes equations for
and therefore no sinkage or trim was permitted during a test run.
incompressible flows. The volume of fluid approach was used
Three resistance wave probes were positioned at 0.75, 1.0 and
with a time-dependent and explicit time discretization scheme
1.25 m from the centre-line of the model to record the elevation
employed to solve the equations. The SIMPLE algorithm was
of the vessel-generated waves with respect to time (Figure 2). A
used for the pressure-velocity coupling and the PRESTO
load cell was installed at the connection between each towing
scheme for the pressure interpolation. The k-epsilon model with
post and the model to measure the vertical and longitudinal
the standard wall function was utilised for turbulence
horizontal forces. The model was tested for varying Froude
modelling. The 2nd order upwind scheme was used for solving
depth numbers from 0.26 to 0.99. Figures 3 to 5 present the
the momentum equations and the High Resolution Interface
experimental wave height data for varying Froude depth
Capturing scheme (HRIC) for the solution of the volume
numbers. Each point is for a single run, while at some Frh the
fraction equations. As with the experiments, the heave and trim
tests were repeated to study uncertainly.
of the model were fixed

PRESSURE SOURCE Tank centre-line

Wavedozer
A wavedozer was used as a pressure source to generate the 1.0m
0.75m

wave. The wavedozer is a wedge shape body with constant WP1

beam[13]. Table 1 shows the main particulars. 1.75 m 1.25m

WP2
Table 1: Wavedozer Principal Particulars
Length (m) 1.5 WP3
Beam (m) 0.3 Tank side wall
Draft (m) 0.1 Figure 2: Layout of probes relative to wavedozer pressure
Angle of attack (deg) 14 source
Where the angle of attack is the relative angle between the
entry surface and the water surface. Figure 1 shows the
wavedozer attached to the carriage in the Australian Maritime
College (AMC) towing Tank prior to a test run.

2 Copyright © 2013 by ASME

Figure 3: The experimental wave height data for varying Froude Figure 5: The experimental wave height data for varying Froude
depth numbers at 0.75m lateral distance depth numbers at 1.25m lateral distance

COMPUTATIONAL DOMAIN
The computational domain was constructed to be 6m
upstream and 13.5m downstream of the model. The domain
length is long enough to have no significant effect on the first
wave propagation. As the flow has a plane of symmetry about
the centre plane, to reduce the processing time, half of the
domain was used. The origin of the coordinate system was
located at the middle of the model. The open channel boundary
condition was used to specify the inlet and outlet boundary
condition. A symmetry plane was used along the centre plane,
and the remaining boundary surfaces along the exterior of the
domain were set to no-slip wall conditions. Figure 6 shows the
exterior of the domain. In addition, the width of the domain was
the same as the towing tank described in the previous section.

Figure 4: The experimental wave height data for varying Froude 13.5 m
depth numbers at 1.0m lateral distance part II
0.3 m
6.0 m
1.75 m
part I

y
x
Figure 6: Computational grid domain

To study the effects of mesh quality on the wave

propagation, the exterior computational grid domain dimensions
were kept constant. The cells' parameters for the rear domain
were changed (part II as shown in Figure 6) while the rest of the

3 Copyright © 2013 by ASME

domain (part I) was kept constant in all of the computational
domains. The changing area starts from 0.3m aft of the transom
to the termination of the domain. The reason for changing only
the rear portion was to maintain the same accuracy to predict
pressure around the pressure source. These simulations allowed
a comparison of the results in terms of mesh quality with the
same wave generating pressure for all conditions. The cells'
dimensions in longitudinal (x) and vertical (z) directions were
changed, but there was no change in lateral (y) direction. It
should be mentioned that the neighbouring cells of Part I and
Part II in the computational domain had the same size in all of
the mesh domains. In addition, all cells in Part II were
hexahedral cells and were all orthogonal. Therefore the cell
skewness for all cells in Part II was zero.

NUMERICAL MODELING
The first simulations were conducted using a mesh domain Figure 8: Comparison of wave height of numerical, with mesh
(M1), which had high aspect ratio cells in Part II. This mesh domain M2, and experimental results at varying lateral
was not able to accurately predict wave height (Figure 7). distances for varying Froude depth

The results for M2 were acceptable for all range of Froude

depth numbers. In the next change, the number of cells in x
direction has been increased again from 80 to 120 (M3), which
caused a decrease in the average aspect ratio to less than 5, but
there was no significant improvement in the wave height
predictions (Figure 9). The reason for the mismatch between the
experimental and numerical data for the middle wave probe is
currently unknown, however due to the trends inherent in the
experimental data this point can be considered as an outlier.

Figure 7: Comparison of wave height of numerical, using high

aspect ratio mesh domain (M1), and experimental results at
varying lateral distances

The average aspect ratio of cells for different Froude depth

numbers 0.9 and 0.99 were more than 15 at different lateral
distance. Therefore in the second mesh (M2), the number of
cells in the x direction was increased and, as a result, the
average aspect ratio was decreased. It means that the average
number of cells per wave length was increased from about 20 to
80 and the aspect ratio was decreased to less than 7. The wave
height prediction was improved as shown in Figure 8. Hence Figure 9: Comparison of wave height of numerical, with mesh
further Froude depth numbers, 0.7 and 0.75, were modeled domain M3, and experimental results at varying lateral
using this mesh. distances for varying Froude depth

By increasing the number of cells in the x direction, two

parameters: aspect ratio and the number of cells per wave length
were changed. Next it was investigated whether the
improvement in the prediction of the wave height was due to the
aspect ratio modification or the change in the number of cells
per wave length. A new mesh was generated (M4), wherein the

4 Copyright © 2013 by ASME

number of cells per wave length was kept same as previous
mesh (M3), about 120, but the cells height was decreased by
changing the expansion ratio which caused the aspect ratio
increase. The results are presented in Figure 10.
Figure10: Comparison of wave height of numerical, with mesh
domain M4, and experimental results at varying lateral
distances for varying Froude depth
The average aspect ratio for this mesh (M4) was less than
10. Since this mesh did not result in a significant change in the
wave height predictions, the cells in z direction were condensed
further to increase the aspect ratio. In the fifth mesh (M5) the
average aspect ratio was about 15. The accuracy of predictions
reduced dramatically (Figure 11).
Figure 11: Comparison of wave height of numerical, with mesh
domain M5, and experimental results at varying lateral
distances for varying Froude depth
As a reminder, the total number of cells in mesh domain
M3, M4 and M5 were similar and the changes in accuracy
resulted from changing the ratio and interval size between cells.
By comparing the results from the last three mesh domains, it
can be concluded that the aspect ratio is a significantly more
important parameter than number of cells per wave length.
To optimse the computation time, the number of cells need
to be reduced as much as possible. In the next stage a new mesh
domain was generated (M6). In M6 the size of cells in the x
direction were the same as M2, which means about 80 cells per
wave length and the number of cells per wave height were
reduced to less than 10 at different lateral distances. Generally
the wave height predicted by mesh domain M6 is higher than
that predicted using M2 in the far field domain (Figure 12).
Figure 12: Wave height for varying Froude depth number for
mesh domains M2 and M6
In mesh domain M7, the number of cells in x direction was
the same as in mesh domain M3 and in the z direction was the
same as in mesh domain M6. The average aspect ratio for this
mesh domain was less than 3. The wave height did not change
significantly compared to M6. Figure 13 presents the results for
M6 and M7.
Figure 13: Wave height for varying Froude depth number with
mesh domains M6 and M7
5 Copyright © 2013 by ASME
CONCLUSION
In this research the effect of mesh quality on prediction of
the water wave propagation generated by a travelling pressure
source has been investigated. Different mesh domains were
generated with varying mesh parameters. The commercial CFD
software, Fluent was used to simulate the generated wave by a
moving body and propagation to the far filed. The results were
compared with those obtained from experimental investigations
conducted in the Australian Maritime College towing tank.
The comparison of the numerical predictions with the
experimental data has confirmed that the mesh quality has a
significant effect on wave height prediction. According to
numerical results and comparison with experimental data, the
aspect ratio is a significantly more important parameter than the
number of cells per wave length. Based on the numerical
results, an average aspect ratio of about 7 is required to predict
wave height. In terms of the number of cells per wave height,
the results show that 10 cells are sufficient to predict wave
height and that increasing this value does not mean that the
predicted wave height will be improved.
In general, the results indicate that the Fluent CFD software
is a viable tool to predict wave height and propagation to the far
field.
ACKNOWLEDGMENTS
This research has been conducted as part of a collaboration
research project between Webber Wave Pools, TU Delft, and
the Australian Maritime College at the University of Tasmania
with funding from the Australian Research Council, project ID:
LP0990307.
[6] Kim, Y., 2003, "Artificial Damping In Water Wave
Problems II: Application to Wave Absorption," The
International Society of Offshore and Polar Engineers, 13(2), p.
5.
[7] Karcz, J., and Kacperski, L., 2012, "An effect of grid quality
on the results of numerical simulations of the fluid flow field in
an agitated vessel," 14th European conference on
mixingWarszawa.
[8] Schmied, S., Binns, J., Renilson, M. R., Thomas, G.,
Macfarlane, G., and Huijsmans, R., 2011, "A novel method for
generating continuously surfable waves- comparison of
predictions with experimental results," 30th International
Conference on Ocean, Offshore and Arctic Engineering, ASME,
Rotterdam, The Netherlands.
[9] Javanmardi, M., Binns, J., Renilson, M. R., Thomas, G.,
Schmied, S., and Huijsmans, R., 2012, "The formation of
surfable waves in a circular wave pool- comparison of
numerical and experimental approaches," 31th International
Conference on Ocean, Offshore and Arctic EngineeringRio de
Janeiro, Brazil.
[10] Schipper, 2007, " On the generation of surfable ship waves
in a circular pool: Part I Physical background & Wave pool
design " Master, Delft University of Technology.
[11] Vries, T. d., 2007, "On the Generation of Surfable Ship
Waves in a Circular Pool, part II."
[12] Essen, S. V., 2011, "RAPID Non-linear Potential Flow
Wave Height Prediction," Australian Maritime College.
[13] Driscoll, A., and Renilson, M. R., 1980, "The wavedozer.
A system of generating stationary waves in a circulating water
channel," AMTE(H) TM80013.

REFERENCES

[1] Varyani, K. S., 2006, "Full scale study of the wash of high
speed craft," Ocean Engineering, 33(5-6), p. 18.
[2] Varnousfaaderani, M. R., and Ketabdari, M. J., 2008,
"Numerical simulation of waves generated by ships in shallow
water," journal of ship technology, 4(2).
[3] Chandraprabha, S., and Molland, A. F., 2004, "A numerical
preiction of wash wave and wave resistance of high speed
displacement ships in deep and shallow water," Faculty of
Mechanical Engineering Network of Thailand 18.Khon Kaen,
Thailand.
[4] Fontaine, E., and Tulin, M. P., 1998, "On the prediction of
Nonlinear free-surface flows past slender hulls using 2D+t
Theory: the evolution of an idea," Fluid dynamics of vehicles
operating near or in the air-sea interfaceAmsterdam, The
Netherland.
[5] Henn, R., Sharma, S. D., and Jiang, T., 2001, "Influence of
canal topography on ship waves in shallow water," 16th
International workshop on water waves and floating bodiesAKI
grand Hotel, Hiroshima.

6 Copyright © 2013 by ASME

 ANNEX A

PUT ANNEX TITLE HERE

Number of cells per wave length= number of cells which cover

the first wave in x direction

Average cells size in x direction=

Number of cells per wave height= number of cells which cover

the first wave in z direction

Average cells size in z direction=

Aspect ratio=

7 Copyright © 2013 by ASME

8 Copyright © 2013 by ASME