Mohammad Saud Afzal1

Three-Dimensional Numerical
Department of Marine Technology,
Norwegian University of Science and Technology,
Trondheim 7491, Norway
e-mail: mohammad.s.afzal@ntnu.no
Hans Bihs
Department of Civil and Transport Engineering,
Norwegian University of Science and Technology,
Trondheim 7491, Norway
Arun Kamath
Department of Civil and Transport Engineering,
Norwegian University of Science and Technology,
Trondheim 7491, Norway
Øivind A. Arntsen
Department of Civil and Transport Engineering,
Norwegian University of Science and Technology,
Trondheim 7491, Norway
Modeling of Pier Scour Under
Current and Waves Using
Level-Set Method
A three-dimensional (3D) computational fluid dynamics (CFD) model is used to calculate
the scour and the deposition pattern around a pier for two different boundary conditions:
constant discharge and regular waves. The CFD model solves Reynolds-Averaged
Navier–Stokes (RANS) equations in all three dimensions. The location of the free-surface
is represented using the level-set method (LSM), which calculates the complex motion of
the free-surface in a very realistic manner. For the implementation of waves, the CFD
code is used as a numerical wave tank. For the geometric representation of the moveable
sediment bed, the LSM is used. The numerical results for the local scour prediction are
compared with physical experiments. The decoupling of the hydrodynamic and the mor-
phodynamic time step is tested and found to be a reasonable assumption. For the two sit-
uations of local pier scour under current and wave conditions, the numerical model
predicts the general evolution (geometry, location, and maximum scour depth) and time
development of the scour hole accurately. [DOI: 10.1115/1.4029999]

Introduction bridge pier both experimentally and numerically using a finite vol-
ume scheme and calculated the turbulence with the k–x model.
Piers find application in large marine structures as platform legs
The model used by Ref. [4] did not have a free-surface algorithm.
and bridge piers which are generally exposed to waves and cur-
Later, Bihs and Olsen [5] performed numerical simulations of pier
rents. Observations have shown that the stability of these struc-
scour. Good agreement was found with the physical model stud-
tures is in constant danger due to scouring near the structure at the
ies. They later focused on the influence of the sloping bed in scour
ocean bed. In the event of a high flow discharge in a river or
holes on the incipient motion of sediment particles and the corre-
extreme weather in oceans where wave heights are large, a scour
sponding sand avalanche mechanism, Bihs [6]. All the studies dis-
hole appears around bridge piers. The local scour depth develops
cussed above for pier scour studies consider steady currents only.
under these conditions and if it is not predicted correctly the bot-
To investigate scour process under waves, a 3D numerical model-
tom level of local scour hole will exceed the original level of the
ing of local pier scour was done by Liu and Garcia [7] in which
pier foundation. In this case, the safety and stability of the bridge
the volume-of-fluid method was used for the calculation of the
will be threatened. The flow field around a pier is extremely com-
free-surface.
plex, 3D and turbulent due to separation and generation of multi-
The current study uses a 3D numerical model to calculate pier
ple vortices [1]. The complexity of the flow field is further
scour under current and waves using a LSM based numerical
magnified due to the dynamic interaction between the flow and
wave tank. Apart from being more accurate, the current numerical
the moveable boundary during the development of a scour hole in
model is unique in the use of LSM for the representation of both
the presence of waves. The solution to this technique is the provi-
the free-surface and the bed topology. This approach maintains
sion of scour protection, a design which is very much dependent
the smooth and continuous solution at the air–water and
upon a clear understanding of the physics behind the scour process
water–sediment interface and avoids numerical instability as the
and its correct prediction. Hence, for the past two decades, the
level-set is smooth across the interface making it differentiable.
study of pier scour under current and wave conditions has been a
Other numerical models [7] use Lagrangian approach for repre-
research topic among scientists.
sentation of bed topology, which requires regridding after every
The publication by Olsen and Melaaen [2] on 3D numerical
sediment time step that smears out the solution and introduce
modeling of local pier scour was the earliest of its kind followed
numerical instabilities.
by Olsen and Kjellesvig [3], who calculated the entire time de-
In the current paper, the numerical investigations of pier scour
pendent process for the same configuration as in Ref. [2]. Both
are validated against the pier scour experiments conducted in the
these studies were done under the assumption of stationary flow.
Department of Hydrodynamics and Water Resources (ISVA),
These publications employed a finite volume scheme to discretize
Technical University of Denmark [8] by Sumer and Fredsøe for
the RANS equations and the equations of the k– turbulence
waves and the pier scour experiments conducted in the Hydraulic
model. Semi-implicit method for pressure-linked equations (SIM-
Laboratory of the Technical University Darmstadt [9] for the
PLE) method was used to calculate the nonhydrostatic pressure
current case.
with provisions of bed shear stress reduction and implementation
of sandslide algorithm in the model. They compared empirical
formulas with their results for the maximum scour depth. Later, Numerical Model
Roulund et al. [4] investigated the flow and the scour around a The CFD code uses the continuity and the incompressible RANS
equations as the governing equations for mass and momentum
1
Corresponding author. conservation
Contributed by the Ocean, Offshore, and Arctic Engineering Division of ASME
for publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING. @Ui
Manuscript received March 10, 2014; final manuscript received March 4, 2015; ¼0 (1)
published online April 6, 2015. Assoc. Editor: Dong S. Jeng. @xi

Journal of Offshore Mechanics and Arctic Engineering JUNE 2015, Vol. 137 / 032001-1
C 2015 by ASME
Copyright V

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

    @U
@Ui @Ui 1 @P @ @Ui @Uj s ¼ qðt þ Þ (3)
þ Uj ¼ þ ð þ t Þ þ þ gi @z
@t @xj q @xi @xj @xj @xi
(2) The sediment transport rates in the bed cells are calculated with
where U is the velocity averaged over time t, q is the fluid density, Engelund and Fredsøe bed load formula [19]
P is the pressure,  is the kinematic viscosity,  t is the eddy vis-
cosity, and g is the gravity term. The turbulence modeling qb;i ¼ 0; s < sc;i
approach in this study makes use of the RANS equations where qb;i ¼ 18:74ðs  sc;i Þðs0:5  0:7s0:5 s > sc;i
c;i Þ; (4)
the eddy viscosity  t in the RANS equations is determined sc;i
through the two-equation, the k–x model [10]. The pressure gradi- sc;i ¼
ent term in the RANS equations is modeled using the projection ðqs  qÞgdi
method, which requires explicit time treatment of the
Navier–Stokes equations as proposed by Chorin [11] for incom- Here, qs is the density of the sediment, q is the density of the
pressible flow. In this method, an intermediate velocity field is water, g is the gravity, and di the sediment particle diameter. qb;i
first obtained by ignoring the pressure gradient. The intermediate is the dimensionless bed load transport rate also called Einstein
velocity Ui is computed using the transient RANS equation which number [20], and sc;i is dimensionless critical shear stress.
does not satisfy the continuity equation. In the second step, the Van Rijn formula for the suspended load [21] is used to account
projection step, the pressure is used to determine the velocity at for the suspended load transport. The governing equation for sus-
the next time step, n þ 1 by convecting it and making it diver- pended load is a standard convection–diffusion equation. The
gence free. The Poisson equation is solved using the fully parallel- numerical treatment for this transport equation is done in a similar
ized Jacobi-preconditioned BiCGStab solver [12]. fashion as the momentum equations
The fifth-order weighted essentially nonoscillatory (WENO)  
@c @c @c @ @c
scheme as proposed by Jiang and Shu [13] is used to discretize the þ Uj þ wS ¼ C (5)
convective terms of the RANS equations in conservative finite dif- @t @xj @z @xj @xj
ference framework. The conservative WENO scheme is used for The fall velocity of the sediment particles is denoted wS. The
the treatment of the convective terms for the velocities Ui, while value of the diffusion coefficient C is assumed to be equal to the
the Jacobi–Hamilton version is used for the variables of the free- eddy viscosity. The values of the suspended load concentration in
surface and turbulence algorithms. The time treatment is dealt the bed cell are used as a boundary condition for the
with using a third-order accurate total variation diminishing convection–diffusion equation, which is applied in the rest of the
(TVD) Runge–Kutta scheme consisting of three Euler steps [14]. computational domain. The deposition rate D at the bed is
This scheme provides a high-order of temporal accuracy, and for
Courant–Friedrichs–Lewy (CFL) numbers below 1 it shows very D ¼ wS cb (6)
good numerical stability through its TVD properties. Adaptive time
stepping is used in order to control the CFL number. The third-order Here, cb is the sediment concentration near the bed calculated
accurate TVD Runge–Kutta is used for all transport equations in the using Eq. 5. In the current model, the concentration at the nearest
numerical wave tank with the exception of the turbulence model. cell center is used in Eq. 6. The entrainment rate E is calculated
The LSM is employed to model the free-surface. This method with van Rijn’s suspended load formula [21]
was proposed by Osher and Sethian [15] for computing and ana-
lyzing the motion of an interface between two phases in two or E
three dimensions. This method is employed in the current study to cbed;susp:load;i ¼
wS
model interface between water–air and water–sediments. The  1:5
location of the interface is represented implicitly by the zero s  sc;i
level-set of the smooth signed distance function /ð~ x; tÞ. The level- di sc;i (7)
cbed;susp:load;i ¼ 0:015    !0:3
set function gives the closest distance to the interface in every a qS =ðqW  1Þ 1=3
point of the modeling domain. The phases are distinguished by di
the change of the sign. The main advantage of using the LSM to 2
calculate the interface between the fluids is that /ð~ x; tÞ is smooth
across the interface which makes it differentiable at the interface The LSM is employed in the current numerical model to track
and avoids numerical instabilities. the movable sediment surface. The evolving bed is represented
The current study uses a numerical wave tank [16] for the gen- implicitly by the zero level-set of the smooth signed distance
eration of waves by the use of relaxation method with two relaxa- function /ð~x; tÞ ¼ 0. This is essentially an Eulerian approach and
tion zones. Relaxation functions are used to generate the waves at regridding at the water–sediment interface is not necessary. The
the beginning, absorb the waves at the end, and prevent reflected evolution of /ð~ x; tÞ corresponding to the motion of the interface is
waves from affecting the wave generation. Each zone has its own adjusted by simply moving the level-set up and down depending
relaxation function associated with it. Zone 1 takes care of wave upon the erosion or accretion of the bed cells.
generation and prevents reflected waves from affecting the wave Here, zb is the local bed surface elevation. The velocity of the
generation and the third zone is the numerical beach, which sediment surface F is determined by the mechanism of sedimenta-
absorbs the waves at the end of the tank. Zone 2 is a working tion and erosion due to bed load and suspended load transport,
zone. Absence of zone 3 leads to wave reflection and hence can which particularly is dependent of the local wall shear stress. This
be used to simulate the formation of standing waves. The relaxa- is known as Exner equation
tion function presented by Jacobsen et al. [17] is used. The wave
generation zone for this study has the length of one wavelength @qB;x @qB;y
and the numerical beach also extends over one wavelength. The ð1  p0 ÞF ¼   EþD (8)
@x @y
numerical wave tank zones can be visualized as in Fig. 6.
F is the propagation velocity of the interface along its vertical
Sediment Transport Modeling
direction, which is given by
Transport Equations and Bed Morphology. Turbulent vis-
cosity based bed shear stress formulation [18] is used for calcula- @zb
F¼ (9)
tion of bed shear stress @t

032001-2 / Vol. 137, JUNE 2015 Transactions of the ASME

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Here, p0 is the porosity of the bed layer, E is the erosion rate
caused due to external actions, and D is the corresponding deposi-
tion rate. The terms on RHS of Eq. (8) are evaluated in the direc-
tion normal to the sediment surface. In the model, the flow field is
first calculated by solving Navier–Stokes equation. Then, the
obtained flow field and the turbulent eddy viscosity are employed
to solve for the bed load and suspended load. The position of the
sediment surface, indicated by the zero level-set /ð~ x; tÞ ¼ 0 is
updated for each morphological time step.
Bed Shear Stress Reduction on a Sloping Bed. The threshold
of sediment motion is one of the most important mechanisms that
determine the bed load transport rate occurring under stream flow.
This threshold of motion is defined by Shield’s diagram [22]
which uses a semi-empirical approach, combining the equilibrium
of forces on a sediment particle with his experimental observa-
tions. This diagram does not take into account the effect of bed
slope. On a sloping bed, the incipient motion of sediment particles
is not only a function of hydrodynamic forces, but also additional
gravity and tractive components have to be included, in order to
correct the critical shear stress obtained from the Shields diagram.
One such example is scour hole. To calculate the reduced shear
stress (Eq. 10), initially the critical shear stress s0 is calculated for
a flat bed. Then, a reduction factor r, based on formulations given
by Dey [23], is calculated which takes the slope of the bed into
account. The reduction factor r is multiplied with s0 to give the
final critical shear stress sc
sc ¼ r  s0 (10)
Sandslide. The onset of erosion induces a tilting effect in the
individual bed cells, which leads to reduction of critical bed shear
stress in the next time step. This causes more erosion in the cell
leading to even steeper slope of the bed cell resulting in even
smaller r in the following time steps. A special algorithm (Sand-
slide) is hence required otherwise the erosion in the bed cells
would never stop. The sandslide algorithm thus acts as limiter for
the bed shear stress reduction in erosion process. Some of the pre-
vious studies [3,4] implemented the sand slide algorithm by cor-
recting the slope of the bed when it became larger than the
constant angle-of-repose under air u. Some studies [3] made use
of a constant u, where others [4,6] made use of the sand slide
algorithm that gives all bed cells the same slope, namely,
u  2 deg, giving the scour hole a uniform look.
The current study utilizes the sand slide algorithm as presented
in Ref. [24]. When the slope of the bed becomes larger than u, the
slope of the concerned bed cell is readjusted by redistributing the
volume of sediment in all the eight neighboring cells of the plane.
This process is continued until the bed angle becomes equal to the
angle-of-repose. This process gives a uniform look to the scour
hole.
Time Step Decoupling. Numerical modeling of erosion and
deposition is a very complex process. It involves the correct pre-
diction of the flow conditions coupled with the sediment transport
modeling and the bed evolution. The movement, erosion, and dep-
osition of the sediments are extremely sensitive to slight changes
in the flow conditions and lead to evolution of the bed. As these
processes are interdependent, the numerical models should ideally
use the same time step for the bed evolution as for the hydrody-
namic solver. This fully coupled approach for the bed evolution is
computationally very expensive [18]. As a result most of the exist-
ing numerical models make use of either semicoupled or fully
decoupled approach. The numerical model utilizes a fully
decoupled approach which assumes no interaction of flow with
suspended load, the bed load, or the bed level changes. The mor-
phological time step (tsed), which is much larger than the hydrody-
namic time step (thyd), is used for the computation of the
Journal of Offshore Mechanics and Arctic Engineering
Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
suspended sediment transport, the bed load transport, and evolu-
tion of the bed. These two time steps are related with the relaxa-
tion factor (rel) for sediment time step as
tsed ¼ thyd =rel (11)
This assumption of fully decoupled approach is valid in cases
where suspended sediment concentration is too low to alter the
flow conditions forming a feedback process. This approach for
current numerical model is further strengthened by the fact that in
the numerical model the morphology and the bed load transport
processes are calculated in terms of bed load layer and not
individual sediment particles.
Results
Pier Scour Under Current. The experimental results, utilized
to compare with the numerical results, were conducted in the
hydraulic laboratory of the Technical University Darmstadt [9].
The flume used for the experiments had a rectangular cross sec-
tion with a length of 37 m, width of 2 m, and depth of 1 m. The
side walls were confined by plexiglass walls. The pier, made of
plexiglass, was placed in the flume center, 16 m away from the
inflow. The pier diameter was 0.2 m. The stretches 1.5 m before
and after the pier were filled with sediment. The rest of the flume
consisted of concrete. The boundary between sediment bed and
concrete floor was sealed with aluminum plates in order to prevent
horizontal water flow through the sediments. The pier scour
geometry was measured with a laser distance sensor, which was
placed inside the pier. The constraint of this method is that the
bed topography outside the actual scour hole is not considered.
The bed material consisted of natural sand with a d50 of
0.00097 m, density of 2650 kg/m3, and an angle-of-repose of
29 deg in air. From a total of eleven runs, the one with a discharge
of 0.18 m3 =s and a water level of 0.3 m was chosen for the com-
parison in the present study. The duration of the run was 21 hr
(75,600 s). The numerical model setup for pier scour case is as
shown in Fig. 1.
Numerical setup is used which consists of a 3D channel which
is 4 m long, 2.0 m wide, and 0.8 m deep with 0.30 m water depth.
The location of center of the pier is 2.50 m and 1.0 m in x and y
directions, respectively. The pier diameter is 0.20 m. A 0.30 m
thick sediment layer has been used in the numerical model setup.
The test has been performed with a mesh width of 0.025 m. CFL
criterian of 0.3 is used along with the adaptive time stepping for
this case. Figure 2 shows the measured scour geometry in the
form of a contour plot from 21 hr (75,600 s) after the start of the
experiment. The maximum scour depth after 75,600 s is 0.15 m
and is located on the upstream side of the pier.
Figure 3 shows the modeled scour geometry with Dey’s empiri-
cal formula for the correction of the critical shields stress and the
k–x model after 75,600 s. Here, the maximum depth is 15 cm
(0.15 m). This corresponds well with the experimental results.
Some difference in scouring pattern downstream of the pier is
Fig. 1 Pier numerical setup under steady current
JUNE 2015, Vol. 137 / 032001-3

Fig. 2 Contour plot of experimental result (as in Ref. [6])
Fig. 3 Model result with k–x model and Dey empirical formula
for bed shear stress reduction
noticed in the numerical model which shows an arrow shaped
scour hole on the downstream side compared to the experimental
result which shows a uniform circular scour hole. This can be
attributed to the measurement procedure adopted during the
experimental investigation, which is unable to capture the topog-
raphy outside the actual scour hole.
The development of the maximum scour depth over time is
shown in Fig. 4. The overall tendency is strong erosion in the
beginning and less in the end. In the first 7000 s, the erosion pre-
dicted by the numerical model is similar to the experiments. For
next 15,000 s, the simulated scour depth grows more rapidly than
the experiment until both lines meet up near the 25,000 s mark
after which there is a very close match of the numerical simulation
with the experiment. The shape of both curves indicates that the
equilibrium scour depth has not yet been reached.
032001-4 / Vol. 137, JUNE 2015
Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Fig. 4 Time development of scour, numerical model (REEF3D)
with k–x model and Dey empirical formula for bed shear stress
reduction
Fig. 5 3D model result under constant discharge showing
free-surface and topography
The numerical model is also able to represent the complex free-
surface using the LSM. Figure 5 shows the bed elevation changes
in combination with free-surface under constant discharge condi-
tions in a very realistic manner using a 3D plot of the numerical
model result after 75,600 s or 21 hr.
Pier Scour Under Waves. The experimental results used to
compare with the numerical results were conducted in the Depart-
ment of Hydrodynamics and Water Resources (ISVA), Technical
University of Denmark [8] by Sumer and Fredsøe. These experi-
ments were carried out in a wave flume that was 12.5 m long and
23.5 m wide. The water depth used for the experiments was
0.40 m. The scour experiments used a rectangular shaped sand pit
in the basin. The pit was 5.6 m long, 3.6 m wide, and 0.14 m deep.
Three cylinders with diameters of 0.54 m, 1.0 m, and 1.53 m were
used for the scour experiments. The same cylinders were also
used for rigid bed experiments. The cylinder was placed in the
sediment bed extending down to the bottom of the sand pit. The
bed material consisted of natural sand with a d50 of 0.00020 m
and a density of 2650 kg/m3. Several tests with varying cylinder
diameter and varying wave properties were conducted. For the
current study, the test chosen for comparison with the numerical
results had a wave period of 3.5 s, a wave height of 0.12 m, and
wavelength of 6.79 m. The cylinder diameter was kept at 1.0 m.
Transactions of the ASME
 The numerical model setup for pier scour under waves is as
shown in Fig. 6. For numerical setup, a 3D numerical wave tank
20 m long, 10 m wide, and 1.1 m deep with 0.40 m water depth is
used. The location of the center of the pier is 10.0 m and 5.0 m in
x and y directions, respectively. The pier diameter is 1.0 m. A
0.30 m thick sediment layer has been used in the numerical model
setup. The test has been performed with grid size of 0.10 m. CFL
criterian of 0.3 is used along with the adaptive time stepping for
this case. A maximum scour depth of 0.047 m (47 mm) was
observed after 28,800 s (8 hr) of experimental testing. Figure 7
shows the contour plot of bed elevation change at the equilibrium
stage in experiment under consideration (after 28,800 s (8 hr) test
time) [8]. The sediments upstream of the cylinder are stirred by
waves and transported by wave-induced steady streaming near the
bed outwards toward the shadow region (downstream). This
results in formation of scour (erosion) hole in this area near the
cylinder. The sediment that is transported will eventually end up
in the shadow area (downstream) causing the observed deposition
pattern. This is quite evident from Fig. 7. Figure 8 shows the con-
tour plot of bed elevation change at the equilibrium stage obtained
from numerical simulation using the numerical model after
28,800 s (8 hr) of simulation. The pattern of erosion and deposi-
tion is similar to that of the experimental result.
There is very slight difference in shape of the contour obtained Fig. 8 Contour plot of bed elevation changes under waves
from numerical solution in comparison to the experiment and its (model result)
looks somewhat mollified. But more importantly the magnitude of
the scour hole, the location of the maximum scour, and the loca-
tion of the deposition are identical to that of the experiment. The
development of the maximum scour depth over time is shown in
Fig. 9. The overall tendency is strong erosion in the beginning and
less in the end. In the first 5000 s, the erosion predicted by the
numerical model is less than the experiments. After that the simu-
lated scour depth grows more rapidly than the experiment until
the end of the simulation. The shape of both curves indicates that
the equilibrium scour depth has been reached at 15,000 s. The
maximum scour depth of the numerical solution is slightly over-
predicted. The difference in time development of scour using

Fig. 9 Time development of scour under waves using numerical

model (REEF3D)

numerical model compared to experimental observations can be

Fig. 6 Pier numerical setup under waves attributed to the fact that the sediment transport phenomenon is a
highly complex process and depends upon many parameters and
empirical formulae. The difference in the equilibrium scour depth
is around 10%. The reason for this is that in reality (experiments),
the porosity, permeability, and even the density are nonhomoge-
neous, whereas in numerical investigation we have to use homo-
geneous values for these. Considering these differences, a
difference of 10% in scour depth seems acceptable. It is mostly
the scour location and magnitude which are of importance to
design engineers. Scour location is predicted correctly and magni-
tude is not underpredicted, which is acceptable for design pur-
poses. With fine-tuning of the parameters used in empirical
formulae, the exact match could be achieved but this is not the
main aim of the study. The study aims to simulate the sediment
transport process using mostly the default variables.
One of the important issues with sediment transport modeling
is whether or not to decouple hydrodynamic and morphodynamic
Fig. 7 Contour plot of bed elevation changes under waves time step. This aspect has been studied using this test case by
(experimental from Ref. [8]) varying the relaxation factor (rel) for time step size for sediment

Journal of Offshore Mechanics and Arctic Engineering JUNE 2015, Vol. 137 / 032001-5

Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 10 Time development of scour under waves by varying
relaxation factors for sediment time step
Fig. 11 3D model result under waves showing free-surface
and topography
transport. The values for relaxation factor (rel) used are 0.05, 0.1,
and 0.3 while keeping the same relaxation factor for hydrody-
namic time step. Figure 10 shows time development of scour
under three different sediment time step relaxation factors. The
time development indicates that there is no difference in scour
development over time using these three values. This means that
the approach to decouple hydrodynamic and morphodynamic time
step is a reasonable assumption. The decoupling of hydrodynamic
and morphodynamic time step is widely used and accepted by the
scientific community [25–28]. Because the local scour both under
current and waves is modeled with a very detailed free-surface
algorithm, the time step resulting from the CFL is prohibitively
small. Using a decoupling strategy for the hydrodynamic and mor-
phodynamic time step is thus necessary. The local scour results
are insensitive to the decoupling, as has been shown through care-
ful testing above. The conclusion is not surprising, since the major
mode of transport in this case is the bed load transport and
the CFD code does not calculate the morphology and bed load
transport processes in terms of individual sediment particle move-
ment. Rather, the bed level changes and bottom roughness redis-
tribution are calculated via an Eulerian approach using a much
larger time step than the one used for the flow and suspended
sediment transport solvers.
The numerical model is also able to represent the complex free-
surface and topology using the LSM. Figure 11 shows the bed ele-
vation changes in combination with representation of the complex
032001-6 / Vol. 137, JUNE 2015
Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
free-surface with waves in a very realistic manner using a 3D
plot of the numerical model result after 28,800 s (8 hr) of
simulation.
Conclusions
The sediment transport module is used to calculate the scour,
the deposition pattern, and time development of scour for different
cases of pier scour under current and waves. The numerical results
for both the test cases show good agreement with the experiments,
which confirms that the numerical model is able to predict the
general evolution (geometry, location, and maximum scour depth)
and time development of the scour hole accurately. Very slight
difference in scour pattern is noticed for the pier under constant
discharge compared to the experiments. This is because in the
experiments, the bed topography outside the scour hole is not
measured. The contour plot was slightly mollified in the pier scour
experiments both under waves and currents. Since the location
and magnitude of scour and deposition are the most important
parameters in the field of marine civil engineering, the results
obtained through the numerical simulation are considered satisfac-
tory. The study of pier scour under waves also confirms that the
decoupled approach for the simulation of hydrodynamic and sedi-
ment transport processes is a reasonable assumption. The numeri-
cal model represents the complex free-surface and the bed
topology in both cases satisfactorily.
References
[1] Graf, W., and Istiarto, I., 2002, “Flow Pattern in the Scour Hole Around a Cyl-
inder,” J. Hydraul. Res., 40(1), pp. 13–20.
[2] Olsen, N. R. B., and Melaaen, M. C., 1993, “Three-Dimensional Calculation of
Scour Around Cylinders,” J. Hydraul. Eng., 119(9), pp. 1048–1054.
[3] Olsen, N. R. B., and Kjellesvig, H. M., 1998, “Three-Dimensional Numerical
Flow Modelling for Estimation of Maximum Local Scour Depth,” IAHR J.
Hydraul. Res., 36(4), pp. 579–590.
[4] Roulund, A., Sumer, B. M., Fredsøe, J., and Michelsen, J., 2005, “Numerical
and Experimental Investigation of Flow and Scour Around a Circular Pier,”
J. Fluid Mech., 534, pp. 351–401.
[5] Bihs, H., and Olsen, N. R. B., 2008, “Three-Dimensional Numerical Modeling
of Pier Scour,” Proceedings of the Fourth International Conference on Scour
and Erosion, ICSE 4, Tokyo, pp. 147–151.
[6] Bihs, H., 2011, “Three-Dimensional Numerical Modeling of Local Scouring in
Open Channel Flow,” Ph.D. thesis, Norwegian University of Science and Tech-
nology, Trondheim, Norway.
[7] Liu, X., and Garcia, M., 2008, “Three-Dimensional Numerical Model With
Free Water Surface and Mesh Deformation for Local Sediment Scour,”
J. Waterw., Port, Coastal, Ocean Eng., 134(4), pp. 203–217.
[8] Sumer, B., and Fredsøe, J., 2001, “Wave Scour Around a Large Vertical Circu-
lar Cylinder,” J. Waterw., Port, Coastal, Ocean Eng., 127(3), pp. 125–134.
[9] Link, O., 2006, “Untersuchung der Kolkung an einem schlanken zylindrischen
Pfeiler in sandigem Boden,” Wasser Abwasser GWF, 147(6), pp. 421–422.
[10] Wilcox, D. C., 1994, Turbulence Modeling for CFD, DCW Industries, La Can-
ada, CA.
[11] Chorin, A., 1968, “Numerical Solution of the Navier Stokes Equations,” Math.
Comput., 22(104), pp. 745–762.
[12] van der Vorst, H., 1992, “BI-GStab: A Fast and Smoothly Converging Variant
of BI-CG for the Solution of Nonsymmetric Linear Systems,” SIAM J. Sci.
Stat. Comput., 13(2), pp. 631–644.
[13] Jiang, G. S., and Shu, C. W., 1996, “Efficient Implementation of Weighted Eno
Schemes,” J. Comput. Phys., 126(1), pp. 202–228.
[14] Shu, C. W., and Gottlieb, S., 1998, “Total Variation Diminishing Runge–Kutta
Schemes,” Math. Comput., 67(221), pp. 73–85.
[15] Osher, S., and Sethian, J. A., 1988, “Fronts Propagating With Curvature-
Dependent Speed: Algorithms Based on Hamilton-Jacobi Formulations,”
J. Comput. Phys., 79(1), pp. 12–49.
[16] Bihs, H., Ong, M., Kamath, A., and Arntsen, Ø. A., 2013, “A Level Set Method
Based Numerical Wave Tank for Calculation of Wave Forces on Horizontal
and Vertical Cylinders,” Proceedings of the 7th National Conference on Com-
putation Mechanics MekIT‘13, Trondheim, Norway, pp. 59–70.
[17] Jacobsen, N. G., Fuhrman, D. R., and Fredsøe, J., 2011, “A Wave Generation
Toolbox for the Open-Source CFD Library: Openfoam,” Int. J. Numer. Meth-
ods Fluids, 70(9), pp. 1073–1088.
[18] Zeng, J., Constantinescu, G., and Weber, L., 2005, “A Fully 3D Non-Hydrostatic
Model for Prediction of Flow, Sediment Transport and Bed Morphology in Open
Channels,” Proceedings of the 31st IAHR Congress, pp. 1327–1338.
[19] Engelund, F., and Fredsøe, J., 1976, “A Sediment Transport Model for Straight
Alluvial Channels,” Nord. Hydrol., 7(5), pp. 293–306.
[20] Einstein, H. A., 1950, The Bed-Load Function for Sediment Transportation in
Open Channel Flows, U.S. Department of Agriculture, Washington, DC.
Transactions of the ASME
 [21] van Rijn, L. C., 1984, “Sediment Transport—Part II: Suspended Load Trans-
port,” J. Hydraul. Eng., 110(11), pp. 1613–1641.
€
[22] Shields, A., 1936, “Anwendungen der Ahnlichkeitsmechanik und der turbulenz-
forschung auf die geschiebebewegung,” Mitteilungen der Preuischen Versuch-
sanstalt f€ur Wasser-, Erdund Schiffbau, Berlin.
[23] Dey, S., 2001, “Experimental Studies on Incipient Motion of Sediment
Particles,” J. Sediment. Res., 16(3), pp. 391–398.
[24] Burkow, M., 2010, “Numerische simulation stromungsbedingten sedimenttran-
sports und der entstehenden gerinnebettformen,” Master’s thesis,
Mathematisch-Naturwissenschaftlichen Fakultat der Rheinischen Friedrich-
Wilhelms-Universitat Bonn, Bonn, Germany.
Journal of Offshore Mechanics and Arctic Engineering
Downloaded From: http://offshoremechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
[25] Wu, W., Rodi, W., and Wenka, T., 2000, “3D Numerical Modeling of Flow
and Sediment Transport in Open Channels,” J. Hydraul. Eng., 126(1), pp. 4–15.
[26] Jacobsen, N. G., Fredsoe, J., and Jensen, J. H., 2014, “Formation and Develop-
ment of a Breaker Bar Under Regular Waves—Part 1: Model Description and
Hydrodynamics,” Coastal Eng., 88, pp. 182–193.
[27] Liang, D., and Cheng, L., 2005, “Numerical Modeling of Flow and Scour
Below a Pipeline in Currents: Part I: Flow Simulation,” Coastal Eng., 52(1), pp.
25–42.
[28] Liang, D., Cheng, L., and Li, F., 2005, “Numerical Modeling of Flow and Scour
Below a Pipeline in Currents—Part II: Scour Simulation,” Coastal Eng., 52(1),
pp. 43–62.
JUNE 2015, Vol. 137 / 032001-7