Computers & Fluids 38 (2009) 1050–1058

Contents lists available at ScienceDirect

Computers & Fluids

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

CFD modeling of scale effects on turbulence flow and scour around bridge piers
Wenrui Huang a,*, Qiping Yang b, Hong Xiao b
a
Department of Civil and Environmental Engineering, Florida A&M University, Florida State University College of Engineering, 2525 Pottsdamer Street, Room A129,
Tallahassee, FL 32310, United States
b
Department of Civil Engineering, FAMU-FSU College of Engineering, Tallahassee, FL, United States

a r t i c l e i n f o a b s t r a c t

Article history: Sediment scour near bridge piers is a problem of nationwide concern because it has resulted in more bridge
Received 20 August 2007 failures than all other causes in recent years. The existing bridge scour equation from HEC-18 was
Received in revised form 8 December 2007 developed from laboratory experiments in relatively small scale. Field studies by Mueller [Mueller D,
Accepted 18 January 2008
Wagner Chad R. Analysis of pier scour predictions and real-time field measurements. In: Proceedings of
Available online 15 April 2008
ICSF-1 first international conference on scour of foundations, Texas A&M University, College Station, Texas,
USA; 2002] indicate that it is difficult to verify the scour equation with field data obtained from large
bridge piers. In this study, computational model simulations using a 3D CFD model were conducted to
examine scale effects on turbulent flow and sediment scour. For the large-scale model, the physical scale
and boundary velocity were set up from the small scale model based on the Froude similarity law. Results
of flow and sediment scour were obtained from two different approaches: (a) Froude similarity which is
commonly used in physical modeling and (b) full scale 3D CFD modeling. Unlike physical modeling in
which the effect of turbulent Reynolds number is ignored, the CFD model employs a 2nd order turbulent
model to calculate turbulent velocity and sediment scour. Effects of scale on turbulence flow and sediment
scour were investigated by comparing different results obtained from a full scale numerical model to those
derived from the Froude similarity method.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction found that the Shen et al. [18] model was based on a narrow range
of flow and sediment conditions and therefore is probably not very
The threat of local scour around bridge piers has been known well suited to practical applications. The complete absence of sed-
for many years. According to Richardson et al. [16], the local scour iment size as a factor in the model makes it difficult to any appli-
around bridge piers is one of the most common causes of bridge cation of the model to conditions other than those in his
failures. It has been observed that in the free surface flow around experiments. Breusers [3] carried out experiments using piles with
a bridge pier, downwash motions, horseshoe vortices and vortex D = 5 and 11 cm, water depths of 15, 25, and 50 cm and sand par-
shading are formed and the turbulence is intensified in front of, ticles with d = 2 mm. Experiments were carried out for fixed values
around and behind the piers. In addition, a uniquely shaped scour of U0/U0c (U0c = critical water velocity for the initiation of bed
hole on the loose bed around a pier is observed. Experimental stud- movement). Totapally et al. [19] examined the temporal variations
ies have found that both the flow and the sediment transport pro- of local scour under steady flow and using stepped hydrographs.
cesses during the scour hole development are highly complex. They concluded that a logarithmic equation represented the varia-
The study of local scour around a bridge pier started with labo- tion of scour with time better than a power equation and ques-
ratory experiments. Laursen [10] investigated the relationship of tioned the existence of an equilibrium depth, maintaining that
clear water scour in a long contraction as a function of geometry, scour will continue with time though at a greatly reduced rate.
flow, and sediment. He developed an equation for the equilibrium Graf et al. [8] investigated the flow patterns in planes upstream
depth of scour for a pile or abutment. Shen et al. [17,18] conducted and downstream of a cylinder and vertically in the scour hole using
21 experiments using a single cylinder diameter and sediment size, an acoustic-Doppler velocity-profiler (ADVP). They found that the
but varying the hydraulic conditions (water depth and the depth shear stress was reduced in the scour hole as compared to the ap-
averaged flow velocity) to include both clear water and live bed proach flow but that the turbulent kinetic energy was very strong
conditions. They developed the empirical equation for scour depth at the foot of the cylinder on the upstream side. The turbulent ki-
as a function of time for a pile of diameter, D, in a flow with a depth netic energy was also very strong in the wake behind the cylinder.
averaged velocity, V, and an upstream water depth, y0. Cunha [5] In addition to experiments, numerical simulations based on
computational fluid dynamics (CFD) have been widely used to
* Corresponding author. Fax: +1 850 410 6236. study the turbulent flow and sediment transport around a pier.
E-mail address: [email protected] (W. Huang). Most models for predicting sediment transport are based on a

0045-7930/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compfluid.2008.01.029
 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058 1051

single-phase flow approach. Bakker [2] developed a numerical Continuity equation

model to calculate suspended sediment concentration. Hagatun 
et al. [9] presented a turbulence model to simulate the instanta- ou ov ow

þ þ  ¼0 ð1Þ
neous sediment concentration and the turbulent boundary layer ox oy oz
in the sheet flow regime over a flat bed. Ahilan et al. [1] investi- Momentum equations
gated the motion of sediment in oscillatory flow over a flat bed  
ou ou ou ou
both theoretically and experimentally. Nadaoka et al. [12] devel- ðStÞ  þ u  þ v  þ w 
ot ox oy oz
oped a mobile bed model considering the mass and momentum   2  !
transport of a single-phase flow. Olsen et al. [13] predicted local op 
1 o u o2 u o2 u
¼ ðEuÞ  þ þ þ ð2Þ
scour developing processes using a three-dimensional flow and ox Re ox2 oy2 oz2
sediment transport model. Ribberink et al. [14]) conducted time- 
  

ov ov ov ov
dependent measurements of flow velocities and sediment concen- ðStÞ  þ u  þ v  þ w 
ot ox oy oz
trations in a large oscillating water tunnel. Recently, several two-   2  !
op 
1 o v o2 v o2 v
phase flow-modeling techniques have been developed. Richardson ¼ ðEuÞ  þ þ þ ð3Þ
et al. [15]) simulated the flow structures around a bridge pier with oy Re ox2 oy2 oz2

 

and without the scour hole. Comparing the simulated with the ow ow ow ow
ðStÞ  þ u  þ v  þ w 
experimental results, they found that the FLOW3D hydrodynamic ot ox oy oz
    2  !
model well simulates the complex flow patterns around the bridge op 1 1 o w o2 w o2 w
pier. Wang et al. [21] simulated an evolution of the scour hole ¼ ðEuÞ   þ þ þ ð4Þ
oz Fr 2 Re ox2 oy2 oz2
developing around the bridge pier by using CCHE3D. They exam-
ined the importance of including various flow effects on sediment where dimensionless (starred) variables can be expressed as
transport. Chang et al. [4] used a large-eddy simulation (LES) mod-
el to solve the flow equations around a bridge pier with a fixed bed u ¼ u=V; v ¼ v=V; w ¼ w=V ð5Þ
and no scour. Tseng et al. [20] conducted the numerical simulation 
x ¼ x=L; 
y ¼ y=L; 
z ¼ z=L ð6Þ
with the square and the circular piers by the LES. They found that t ¼ t=T; p ¼ p=P ð7Þ
the down flow is made at the front face of the pier and this affects
the creation of the horseshoe vortex. They also compared turbulent These equations show that if two systems are geometrically similar
structures; lift coefficient, and drag coefficient with the experi- and governed by these equations, then the solutions (in terms of u*,
mental results. Good agreements were obtained. v*, p*, x*, y*, and t*) will be the same if the four parameters Re, Fr, St,
The scour prediction methods and the scour equations based on and Eu are equal and the initial conditions expressed in dimension-
laboratory data did not always produce reasonable results for field less form are also equal for the two systems. The two systems will
conditions [11]. Although Froude similarity generally play more be dynamically similar.
important role in gravity surface water flow, effects of turbulence In lab experiments, two systems satisfy dynamic similarity, in
may not be negligible for flows near bridge pier. In this study, addition to kinematical similarity, if corresponding forces are in
numerical simulations were conducted to investigate the scale the same ratio in both. The force ratio is
effect on turbulence flow and sediment scour near bridge pier.
The software FLUENT was used to predict the 3D flow patterns Fp
Fr ¼ ð8Þ
around a cylindrical pier. The turbulence and sediment modeling Fm
by FLUENT was satisfactorily tested against experimental data of
The Reynolds number (Re = qLU/l), and the Froude number ðFr ¼
scour around a small scale bridge pier. For the large-scale model,
pUffiffiffiffiÞ are the dimensionless parameters most commonly encoun-
the physical scale and boundary velocity was set up from the small gL

scale model based on the Froude similarity law. Unlike physical tered in fluid mechanics. In general, it is very difficult to meet both
modeling in which the effect of turbulent Reynolds number is ig- Reynolds and Froude similarities in lab experiments. For free sur-
nored, the CFD model employs a 2nd order turbulent model to face gravity flow around bridge piers, Froude similitude is often se-
calculate turbulent velocity and sediment scour. Effects of scale lected in scaled lab experimental modeling. As a result, this means
on turbulence flow and sediment scour were investigated by that the terms related to Reynolds number in the momentum equa-
comparing different results obtained from full scale numerical tion (Eqs. (2)–(4)) are removed from the study, which would pro-
model to those derived from Froude similarity method. duce truncation errors. Scale ratios for Reynolds and Froude
similarities can be found in Franzini [7]).

2. Dynamic similarity in fluid dynamic equations 3. Numerical model validation

Physical modeling based on similitude theory is commonly used Yanmaz et al. [23]) conducted a series of experiments to study
in research of turbulence flow and sediment scour around bridge the development of scour around bridge pier models. The experi-
piers. In the HEC-18 manual developed by FHWA [6]), a sediment ments were conducted in glass flume which is an L-shaped hori-
scour equation was derived from small scale laboratory experi- zontal rectangular open channel with a depth of 90 cm and a
ments based on Froude similarity, where effects of Reynolds num- width of 67 cm. Single bridge pier structure was set at the center
ber and turbulence were ignored due to the difficulties of meeting of the flume. Only clear water conditions with a flat bed were stud-
both Froude and Reynolds similarities. Although Froude similarity ies. No sediment inflow was allowed into the scour hole from up-
generally plays a more important role in gravity surface water stream. Experiments were stopped at the end of different test
flow, effects of turbulence may not be negligible for flows near durations to determine the contours of the scour hole around
bridge piers. In this study, numerical simulations were conducted bridge piers, namely 5 min, 20 min, 60 min, 100 min and 150
to examine errors in applying Froude similitude in physical model- min. The experiment data for the pier with a diameter of 6.7 cm
ing of turbulence flow and sediment scour near bridge piers. Refer- were selected to compare with the numerical results. The materials
ring to White [22], non-dimensional fluid dynamic equations can used for the bed are quartz-sand with specific weights of 26.4 kN/
be expressed as follows: m3, mean particle sizes of 1.07 mm. Bed materials were placed as a
1052 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058

15 cm thick layer in the flume bed. The numerical simulation was 3

set up according to the experimental conditions.

31
32
The geometric setup for the numerical simulation is given in

30
34
2
Fig. 1 and the grids of the computational domain generated by 28
34
GAMBIT are shown in Fig. 2. The grids near the cylinder are gener-
1
ated more densely because the flow pattern in the region is more

y-direction (y/R)
complex. The inlet boundary was placed at the left side of the pier, 35 28

36
10R (R is the radius of the pier) from the pier center. The type of the

37
0

boundary was set as velocity-inlet. A uniform velocity of 0.067 m/s 35

in accordance with the experiment was applied on the inlet bound- -1
ary. The outlet boundary was placed at 10R right to the center of 28

the pier. The type of boundary was set as outflow where zero-gra- 34 29
-2
dient boundary conditions are used. The velocities are set equal to

33

30
32
the values in the elements closest to the outflow. At the down-
stream outlet, the normal gradients of all dependent variables -3
-4 -3 -2 -1 0 1 2
are set to zero, i.e., variables at the downstream end are extrapo- x-direction (x/R)
lated from the interior domain. The wall boundary was set at the
Fig. 3. Contour of local scour around a cylinder pier at T = 5 min.
top of the water surface in order to simplify the simulation. The
two-phase (water and sand) Eulerian model was used in order to
simulate the local scour and simplify the problem. The top layer
is phase water and the bottom layer is phase sand (Granular). indicates that the computed scouring pattern is in good agreement
It was observed that the maximum scour depths occurred at the with physical observations.
midpoint of the upstream face of cylindrical pier. The grains at the
upstream side of the piers were observed to be dislocated due to 4. Scale effects on modeling turbulence flow in solid bottom
horseshoe vortices. As the scour hole enlarged with respect to time, condition
the strength of the horseshoe vortices weakened, causing a smaller
rate of scour development, and the scour depth approached an In order to study the scale effects on turbulence flow around a
equilibrium value asymptotically. bridge pier, two different sized models were set up for numerical
The contours around the cylindrical pier obtained at the end of simulations. The small-sized model stands for the physical model
5 min of test duration are shown in Fig. 3. The results of numerical and the large-sized model stands for the prototype. They are geo-
model simulations of local scour for the bridge pier at T = 5 min are metrically similar. The condition boundaries were set up according
given in Fig. 5.5. the maximum scour depth and the deepest scour to the Froude number similarity. Hence, they meet the require-
hole in front of the cylinder are reasonable in good agreement with ments of the geometric similarity and the Froude number similar-
those from the experiments conducted by Yanmaz et al. [23]). It ity. However, this model is distorted because they do not meet the
requirements of the Reynolds number similarity.
The aim of the numerical simulation is to check the difference of
flow field around a bridge pier when the simulation results of the
small-sized model are used to predict the situations of the large-
sized model according to the similitude theory. To simplify the
problem and avoid the feedback of the sediment move, a one-
phase (water only) model was employed in the group of simula-
tions. The 3D computational fluid dynamic model package, FLUENT
CFD package, was utilized to test the scale effects on turbulence
flow around a bridge pier in the simulations.

4.1. Geometric setup

The geometries of the small-sized and large-sized models were

used for numerical simulations. Their dimensions were setup as in
Table 1. They are geometrically similar and the scale ratio of length
Fig. 1. The geometric setup for bridge pier scour simulation. is Lr = 25.

Fig. 2. Computational grid for 3D numerical simulation of local scour around a bridge pier.
 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058 1053

Table 1 At the top surface, the wall boundary conditions are used. There
Geometric setups for the small-sized and large-sized simulations for solid bottom are neither convective fluxes nor diffusion fluxes across the top
condition
surface. This implies that the normal velocity component, as well
Small-size Large-size as the normal gradients of all flow variables, is zero. The water sur-
Length of the flume 0.64 m 16 m face elevation in the channel is fixed. The water elevation at the
Width of the flume 0.64 m 16 m outlet is specified. The wall boundary conditions were set at the
Depth of water of the flume 0.32 m 8m bottoms.
Diameter of cylinder 0.04 m 1m

4.4. Results and analysis

4.2. Model grid setup

The results of numerical simulations of the flow field around a
small-size and large-size bridge pier are given in Fig. 5a and b,
The grid of the simulation generated by GAMBIT is shown in
respectively. For both piers, the velocity vector field clearly shows
Fig. 4. A multi-block grid generation techniques was used here to
there is a separation zone behind the cylinder. The location of the
generate the grids. The grids near the pier are denser than in other
separation point extends further downstream at the surface than
regions since the flow is more complex in this area.
close to the bed. The patterns of these two simulations are almost
the same. The velocity is made dimensionless by the mean
4.3. Boundary conditions setup
approaching velocity (u0) in the figure.
A down flow is developed due to the downward negative stag-
At the inlet boundaries, the types of the boundaries were set as
nation pressure gradient of the non-uniform approaching flow
velocity-inlet. The uniform distributions of water velocities are gi-
close to the upstream side of the pier. The interaction between this
ven (small-size: 0.2 m/s and large-size: 1.0 m/s). The scale ratio of
downward flow and the horizontal boundary layer separation close
velocity is five. It is equal to the square root of the scale ratio of
to the riverbed results in the formation of a vortex system. The two
length (25). They satisfied the requirement of Froude number
ends of this vortex system are swept downstream by the flow as
similarity.
they warp around the pier in the shape of a horseshoe in top view.
At the downstream outlets, the types of boundaries were set as
This vortex significantly affects the scouring process at the base of
outflow. The normal gradients of all dependent variables were set
the bridge piers. Fig. 6a and b shows the contours of x-direction
to zero. (i.e., variables at the downstream end are extrapolated
velocity of the small-sized and large-sized models respectively.
from the interior domain).
Their patterns are very similar. However, the values near the pier
are very different.
The contours of x-direction velocity difference between the
large-sized simulation model and the physical model derived from
small-sized model according to Froude law are shown in Fig. 7. The
current speed drops as it approaches the cylinder and the water
piles up in front of it. The flow stops at the face of the pier and
accelerates around the structures. The associated stagnation pres-
sures are highest near the surface, where the deceleration is the
greatest, and decrease downwards. In response to the downward
pressure gradient at the pier face, the flow reaches a maximum just
below the bed level. The water surface increases due to the stagna-
tion pressure in front of the cylindrical pier and a bow wave was
formed as the surface roller at the water surface. It can be seen
from Fig. 7 that large errors of velocity are located near two sides
of the cylinder where currents are strong. These two models are
set in according with Froude Similitude. It can be concluded from
the figures that Froude similarity generally results in big errors
Fig. 4. Grid used in simulations. at the neighboring area of the pier if it is used to predicate the sed-

2 2

1.5 1.5

1 1
y-direction (y/R)

0.5 0.5
y-direction (y/R)

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5

-2 -2
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)
(a) (R=0.04 m, u0=0.2 m/s) (b) (R=1 m, u0=1 m/s)

Fig. 5. Velocity vector near water surface for one-phase simulation.

1054 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058

2 2

0
1.02

1.07

1.20

1.2
1.5 0.87 0.87 1.5 0.93 7
0.73 1.0 0.93
1.16 1.33 0.80
1 0.58 1 0.67
0.53
0. 44 0.40
0.5

y-direction (y/R)
0.5
y-direction (y/R)

0.80

0.67
0.53
0.73

0.14
0.58

0.29
0.44

0.27
0.13
0.40
0 0

3
0.1
-0.5 -0.5 0.40
0.44
-1 -1 0.67
0.58 1.33
1.16 0.80
0.73 0.93
1.0

1.2
-1.5 -1.5

1.20
0.93

1.07
0.87 0. 87 7

0
1.02
-2 -2
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)
(a) (R=0.04 m, u0=0.2 m/s) (b) (R=1 m, u0=1 m/s)

Fig. 6. Contour of X-direction relative velocity (u/u0) for one-phase simulation.

8
0.131
0.0 0. 098
6 0. 65
03
2 2

0.165
0.1
0. 26 0.2

8
31

2
4 4

98
19
1.5

0.13
0. 099

0.
0.330
0.098 6
06

2 1
31 0.099
y-direction (y/R)

0.

0.098
0.1
0.032 0.298
y-direction (y/R)
0.065
0.032

0.03

0.5 0.033
0.065

0.033
0.066
2

0.032

0.132
0
0.098 0.198
-2
0.131
0.098 -0.5 0.033
0.298
-4 0.099
-1
0.

0. 099
0.132
06
6

0. 065 0.330

4
2

-6

0. 1
03

0.1

0. 26
-1.5

98
31
0.

0.098 0.2

65
98

0.1
0.131
-8 -2
-8 -6 -4 -2 0 2 4 6 8 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)

(a) (b)
(a) Result of Large-Sized Simulation
(b) Result derived from the Small-Sized Simulation according to Froude Law

Fig. 7. Contours of x-direction relative velocity (u/u0) errors between the large-sized simulation and the physical model derived from the small-sized simulation according to
Froude law.

8
0. 029 0.059
6
2
0.0
3
4 0.059 1.5 0. 03 0.06
0.02

0.03
9

12

1
0.

2
y-direction (y/R)

0.029
0.03
0.029

0.029

y-direction (y/R)

0.5
0.03

0
0.0

0
29

0.029
-2
0. 03

-0.5
0.03

-4 0.059
0.02

-1
0.
12
9

-6 0. 03
-1.5 0.03
0.029 0.059 0.06
3
-8 -2 0.0
-8 -6 -4 -2 0 2 4 6 8 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)

Fig. 8. Contours of y-direction relative velocity (v/u0) errors between the large-sized simulation and the physical model derived from the small-sized simulation according to
Froude law.

iment scour. The y- and z-direction velocity differences between be seen that they are very small compared to x-direction velocity.
these two models are shown in Figs. 8 and 9 respectively. It can Nevertheless, the large differences still existed around the pier. It
 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058 1055

0.041
6 0.013 0.083 2
0.041

0.0

0.02
4 1.5

14

8
0.041

0.027

0. 05
1
2

0.055

6
y-direction (y/R)

y-direction (y/R)
0.5
0.013
0

0.085
0.028

0.042

0. 056
0

0.055
-2
-0.5
0.041

7
02
-4

0.
-1
0.041

4
01

6
0.05
0.083

0.
0.02

-6

8
0.013 -1.5

0.02
7

-8 -2
-8 -6 -4 -2 0 2 4 6 8 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)

Fig. 9. Contours of z-direction relative velocity (w/u0) errors between the large-sized simulation and the physical model derived from the small-sized simulation according to
Froude law.

3 3

3.73
1.5

3.31
2 2
7

41

1.41 0
1.1
1.

8
0.79 2. 4
1 1
0.48
y-direction (y/R)

y-direction (y/R)

1. 65
0.17 0.8
4.14

3.31

2. 48
0.95

0.63

1.65
1
0.32
1.26

1.10

0 0
3.73

0.32 1.23

-1 0. 63 -1 2.06
4.56
0. 9
5 4.14

3.31
1.
1.41

2. 8
-2 26 -2
7
1.5

9
3. 73
-3 -3
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
x-direction (x/R) x-direction (x/R)
(a) (R=0.04 m, u0=0.2 m/s) (b) (R=1 m, u0=1 m/s)

Fig. 10. Shear stress (Pa) at the bottom for small-sized (a) and large-sized (b) model.

can be concluded that the distorted model with only geometric The aim of the simulations is also to check the errors when the
similarity and Froude number similarity can result in the large er- simulation results of the small-sized model are used to predict the
rors if it is used to predict the flow situations of prototypes because situations of the large-sized model.
it ignores the Reynolds number similarity.
Fig. 10a and b illustrates the contours of shear stress of small- 5.1. Geometric setup
sized and large-sized simulations, respectively. It can be seen from
these two figures that at the neighboring area of the pier, the pre- Both a small-sized and a large-sized model were used for
dicted bed shear stress values are the largest. If it exceeded the numerical simulations. Their dimensions were setup as in Table
critical shear stress for the particles, sediment transport may be ex- 2. They are geometrically similar and the scale ratio of length is
pected and local scour may be happened. The uniform distribu- Lr = 25.
tions of water velocities are given (small-size: 0.2 m/s and large-
size: 1.0 m/s). The scale ratio of velocity is five. It is equal to the 5.2. Model grid setup
square root of the scale ratio of length (25). They satisfied the
requirement of Froude number similarity. The grid of the simulation generated by GAMBIT is shown in
Fig. 11. The grids near the pier were generated more densely than
5. Scale effects on modeling sediment scour around bridge pier

In this section, two different scale models were set up for Table 2
Geometric setups for the small-sized and large-sized simulations for solid scour
numerical simulation. One model stands for the prototype and an-
other model stands for the physical model. In order to study the Small-size Large-size
scale effects on local scour around a bridge pier, the two models Length 2 50
are set up to satisfy the requirements of both geometric similarity Width 0.67 16.75
and Froude number similarity. However, they do not meet Depth of water 0.15 3.75
Thickness of sand 0.15 3.75
Reynolds number similarity. Therefore, they are also distorted
Diameter of cylinder 0.067 1.675
models.
1056 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058

Fig. 11. Grid for two-phase simulation.

in other area because the flow near the pier is more complex than will follow the flow pattern and are carried away downstream. The
in other area. transport rate of sediment out of the base region is greater than the
rate into the region. Consequently, a horseshoe-shaped scour hole
5.3. Boundary conditions setup is developed. As the depth of the scour increases, the strength of
the horseshoe vortex is reduced. The vortex extends downstream,
At the inlet boundaries, the types of boundary were set as veloc- past sides of the pier for a short distance before losing its identity
ity-inlet. The uniform distributions of water velocities are given and becoming part of the general turbulence. The results of numer-
(0.26 m/s for small-sized model and 1.30 m/s for large-sized mod- ical model simulations of flow field for the small-size and large-
el). The scale ratio of velocity is five (1.30/0.26 = 5), and is equal to size bridge piers are shown in Fig. 12. The patterns of these two
the square root of the scale ratio of length (25). They satisfied Fro- simulations are very similar.
ude number similarity. Velocities along the sides of the scour hole were weak and a
At the downstream outlet, the type of boundary was set as out- grove or lip exists at the base of the cylinder where velocities are
flow. The normal gradients of all dependent variables were set to the highest, indicating this is the most important area of sediment
zero. (i.e., variables at the downstream end are extrapolated from removal. The contours of x-direction relative velocities of simula-
the interior domain). tions for small-sized and large-sized are shown in Fig. 13. The
At the top surface, wall boundary conditions were used. There velocity is made dimensionless by the mean approach velocity
are neither convective fluxes nor diffusion fluxes across the top (u0). The contours of x, y, z-direction relative velocity difference be-
surfaces. This implies that the normal velocity component and tween the large-sized simulation and the physical model derived
the normal gradients of all flow variables are zero, and the water from the small-sized simulation are shown in Fig. 14. It can be seen
surface elevations in the channel are fixed. The water elevations from Fig. 13 that larger errors of velocities are located at some loca-
at the outlet are specified. The wall boundary condition was set tions, especially at the neighboring area of the pier where currents
at the bottoms. are strong. However, the absolute values in y and z directions are
very small compared to x-direction velocities.
5.4. Results and analysis Eventually, for riverbed local scour, equilibrium is established
between bed material flow into and out of the scour hole. For clear
The horseshoe vortex around the bridge pier results from the pi- water scour, scouring ceases when the shear stress caused by the
leup of water on the upstream face of the cylinder and subsequent horseshoe vortex equals the critical shear stress of the sediment
acceleration of the flow around the pier. When the flow reaches a particles at the bottom of the scour hole. These two different scale
certain velocity in the channel, the sediment particles close to the models were set up in according to geometric similarity and Fro-
cylinder begin to move, and scour is initiated. The eroded particles ude number similarity. As shown in Fig. 15, results of sediment

Fig. 12. Velocity vector of two-phase simulation.

 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058 1057

3 3

1. 21
1. 21
1. 0 1.1
6 2 12
1.
2 1. 06 2
1.01
1.37 1.01
0.91 0.90
1 0.91 1

y-direction (y/R)

y-direction (y/R)

0. 79
0.12
0.68

0.68
0. 45
0.90

0.45
0.12
0.61
0.31
0.30

0.76
0.76
0.61

7
0 0

-1 0. 91 -1
0.91 1.01 0.90
1.37

1.06 1.01
-2 -2
6 1.
1.0 1. 21

1. 21
12
1. 12
-3 -3
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
x-direction (x/R) x-direction (x/R)

(a) (R=0.0335 m, u0= 0.26 m/s) (b) (R=0.8375 m, u0= 1.3 0m/s)

Fig. 13. Contour of X-direction relative velocity (u/u0).

Fig. 14. Contour of X, Y, Z-direction relative velocity (u/u0) differences between the large-sized simulation model and the physical model derived from the small-sized
simulation according to Froude law.

4 4
1.60
1.6

1.
1.59
0.93

1.6
0.96

65
0

3
0.9

0.9

87 3
1. 5
2
4

0.
1.5
9

1.6 4
7
2 0.84 2 1.5
1. 02

1
y-direction (y/R)

5 1
y-direction (y/R)

1.0
1.68
1.08

0.82
1. 11

0 0
0.8

1.68
-1 -1
1.03
1.5
1.68 3
-2 0.8 -2
5
0.94

1.
0

0.
0.97

0.9
1.0

6
55
88

1.6
1.5
1

-3 -3
3
64

1.58
1.60

6
1.6

1.61
1.

-4 -4
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
x-direction (x/R) x-direction (x/R)

Fig. 15. (a) Contours of relative scour (ds/R) for small-sized simulation (R = 0.0335 m, u0 = 0.26 m/s), and (b) contours of relative scour (ds/R) for large-sized simulation
(R = 0.8375 m, u0 = 1.3 m/s).

scour in the direct simulation of large bridge pier is different in 6. Conclusion

scour patterns and maximum scour depth from those derived from
Froude law based on simulations from small scale experiment. It Ignoring Reynolds similarity in physical modeling approach
can be concluded from above results that the physical model with- may result in errors for scour in large bridge piers. The degree of
out Reynolds number similarity might result in large errors at the the turbulence intensity, especially in an obstructed flow field, also
neighboring area of the pier if it is used to predicate flow situations contributed significantly in the scouring process. In a physical
of the prototype. model, not all relevant quantities could be furnished. Some physi-
1058 W. Huang et al. / Computers & Fluids 38 (2009) 1050–1058

cal quantities in a turbulence flow are difficult to measure. One of [6] Federal Highway Administration (FHWA). Evaluating scour at bridges. 4th ed.
Hydraulic Engineering Circular No. 18; 2001.
these quantities is the vortex, which was known to be the major
[7] Franzini JB, Finnemore John. Fluid mechanics. 9th ed. New York: McGraw-Hill;
factor responsible for base scouring. In these experiments, how- 1997.
ever, a very crude technique was employed. This did not provide [8] Graf WH, Istiarto I. Flow pattern in the scour hole around a cylinder. J Hydraul
adequate information for direct comparison with the results ob- Res 2001;40(1):13–20.
[9] Hagatun K, Eidsvik KJ. Oscillatory turbulent boundary layers with suspended
tained from the numerical study. Results from this study indicate sediments. J Geophys Res 1986;91(C11):13,045–55.
that the physical modeling approach without Reynolds number [10] Laursen EM. An analysis of relief bridge scour. J Hydraul Div, ASCE
similarity may result in considerable errors in modeling turbulence 1963;89(HY3):93–118.
[11] Mueller D, Wagner Chad R. Analysis of pier scour predictions and real-time
flow and sediment scour around large bridge pier. For velocity, lar- field measurements. In: Proceedings of ICSF-1 first international conference on
ger errors are significant near the two sides of the cylinder where scour of foundations, Texas A&M University, College Station, Texas, USA;
currents are strong. 2002.
[12] Nadaoka K, Yagi H. Single-phase fluid modeling of sheet-flow toward the
The numerical simulation using 3D CFD model that consists of tur- development of numerical mobile bed. In: Proceedings of the
bulence models include both Froude and Reynolds number effects. In 22th international conference coastal engineering ASCE; 1990. p.
comparison, in physical modeling, either Reynolds or Froude similar- 2346–59.
[13] Olsen NRB, Melaaen MC. Three-dimensional calculation of scour around
ity has to be ignored due to difficulties in meeting both similarity cylinders. J Hydraul Eng 1993;119(9):1048–54.
laws. Therefore, numerical modeling may provide better predictions [14] Ribberink JS, Al-Salem AA. Sheet flow and suspension of sand in oscillatory
of turbulence flow and sediment scour around a bridge pier. On the boundary layers. Coastal Eng 1995;25:205–25.
[15] Richardson JE, Pancheng VG. Three dimensional simulation of scour inducing
other hand, it should also be expected that ‘perfect’ results are diffi-
flow at bridge piers. J Hydraul Eng ASCE 1998;124(5).
cult to obtain in view of the many factors involved, which cannot [16] Richardson EV, Davis SR. Evaluating scour at bridges. 4th ed. Federal Highway
be modeled directly using a numerical simulation. Despite all the Administration Hydraulic Engineering Circular No. 18, FHWA NHI 01-001;
limitations, FLUENT generally exhibits good qualitative predicted 2001.
[17] Shen HW, Schneider VR, Karaki SS. Mechanics of local scour, Colorado State
flow patterns around the piers. Further detailed and rigorous study University, Civil Engineering Department, Fort Collins, Colorado, Pub. No.
needs to be carried out to validate the results quantitatively. CER66-HWS22; 1966.
[18] Shen HW, Schneider VR, Karaki SS. Local scour around bridge piers. Proc ASCE,
J Hydraul Div 1969;95(HY6):1919–40.
References [19] Totapally HGS, Aziz NM, Kulne RA, Alonso CV. Temporal variation of local
scour at abutments under steady flow conditions and stepped hydrographs. In:
Proceedings of the 1999 international water resources engineering conference,
[1] Ahilan RV, Sleath JFA. Sediment transport in oscillatory flow over flat bed. J August, Session BS-05, Water Resources Publications, LLC, Highlands Ranch,
Hydraul Eng 1987;113(3):308–22. CO.; 1999.
[2] Bakker WT. Sand concentration in an oscillatory flow. In: Proceedings of the [20] Tseng M, Yen CL, Song CCS. Computation of three-dimensional flow
14th conference coastal engineering ASCE, Copenhagen; 1974. p. 1129–48. around square and circular piers. Int J Numer Meth Fluids 2000;34:
[3] Breusers HNC, Nicollet G, Shien HW. Local scour around cylindrical piers. J 207–27.
Hydraulic Res ASCE 1977;15(3):211–52. [21] Wang SSY, Jia Y. Computational simulations of local scour at bridge crossings –
[4] Chang WY, Lai JS, Yen CL. Simulation of scour depth evolution at pier nose. In: capabilities and limitations. In: Proceedings of the 1999 international water
Proceedings of the 1999 international water resources engineering conference, resources engineering conference, August, Session BS-06, Water Resources
August, Session BS-05, Water Resources Publications, LLC, Highlands Ranch, Publications, LLC, Highlands Ranch, CO.; 1999.
CO.; 1999. [22] White FM. Fluid mechanics. 4th ed. New York: McGraw-Hill; 1999.
[5] Cunha LV. Time evolution of local scour. In: Proceedings 16th international [23] Yanmaz AM, Altinbilek HD. Study of time-dependent local scour around bridge
association for hydraulic research congress, Sao Paulo, Brazil, 27 July to 1 piers. J Hydraul Eng 1991;117(10):1247–68.
August, vol. 2; 1975. p. 285–99.