Engineering Applications of Computational Fluid

Mechanics

ISSN: 1994-2060 (Print) 1997-003X (Online) Journal homepage: https://www.tandfonline.com/loi/tcfm20

Crosswind Effect Studies on Road Vehicle Passing

by Bridge Tower using Computational Fluid
Dynamics

Bin Wang, You-Lin Xu, Le-Dong Zhu & Yong-Le Li

To cite this article: Bin Wang, You-Lin Xu, Le-Dong Zhu & Yong-Le Li (2014) Crosswind
Effect Studies on Road Vehicle Passing by Bridge Tower using Computational Fluid
Dynamics, Engineering Applications of Computational Fluid Mechanics, 8:3, 330-344, DOI:
10.1080/19942060.2014.11015519

To link to this article: https://doi.org/10.1080/19942060.2014.11015519

Copyright 2014 Taylor and Francis Group Published online: 19 Nov 2014.
LLC

Submit your article to this journal Article views: 934

View related articles View Crossmark data

Citing articles: 5 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tcfm20
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3, pp. 330–344 (2014)

CROSSWIND EFFECT STUDIES ON ROAD VEHICLE PASSING BY

BRIDGE TOWER USING COMPUTATIONAL FLUID DYNAMICS

Bin Wang*#‡, You-Lin Xu#, Le-Dong Zhu† and Yong-Le Li‡

#
Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University,
Hong Kong, China
†
State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai, China
‡
Department of Bridge Engineering, Southwest Jiaotong University, Chengdu, China
*
E-Mail: wangbin.hkpolyu@gmail.com (Corresponding Author)

ABSTRACT: When a road vehicle moving on a bridge deck is passing by a bridge tower, the vehicle will be
shielded briefly from the crosswind by the bridge tower. This study aimed to apply computational fluid dynamics to
explore this special engineering problem for the first time. A stationary vehicle immerged in the wake of a bridge
tower was first simulated and compared with wind tunnel test results. By using the dynamic mesh method, the
moving of the vehicle passing by a bridge tower was then simulated. The results show that the computed
aerodynamic coefficients of the stationary vehicle are generally larger than wind tunnel results. The side force and
yawing moment coefficients of the moving vehicle are lower than those of the stationary vehicle. The results also
show that the suddenly changing time of aerodynamic forces caused by the tower may be shorter than the average
driver reaction time.

Keywords: road vehicle, bridge tower, aerodynamics, shielding effect, computational fluid dynamics

1. INTRODUCTION have a reliable assessment of the running safety of

In extremely windy environments, vehicle
accidents such as overturning and sideslipping
may occur, and therefore the running safety of
vehicles in crosswinds has been investigated by
some researchers (e.g. Baker 1986; Baker 1987;
Xu and Guo, 2003a; Baker et al, 2009; Chen and
Chen, 2010). When road vehicles run on a bridge
deck, a coupled wind-vehicle-bridge dynamic
system is formed, and the running safety of
vehicles becomes complicated. Xu and Guo
(2003b) extended the safety analysis method of
vehicles on the ground to a situation in which the
vehicles passed through a long-span cable-stayed
bridge. Such a problem was also investigated by
Cai and Chen (2004). However, in these studies
the aerodynamic forces on a vehicle were
assumed to be ideally sudden loads without
considering the real wake environment of a bridge
tower as well as the interference between the
vehicle and the bridge tower. Actually, when a
road vehicle passes by a bridge tower under
crosswinds, the vehicle will be shielded briefly
from the crosswind by the bridge tower within a
very short time period, but when it moves out of
the shelter it enters a sharp-edged crosswind gust
with the obvious danger that it can be turned over.
The aerodynamic forces on a vehicle in the wake
of a bridge tower should be studied in order to
the vehicle. Charuvisit et al. (2004) carried out
wind tunnel tests to measure the side forces and
yawing moments of both a stationary vehicle and
a moving one in the wake of a bridge tower
model. The length scale they used was 1:30 and
the ratio of the vehicle length to the tower width
was 1.39. The test results showed that, compared
with the stationary cases, the peak side-force on
the moving vehicle when it exited the wake
region was higher than that when the vehicle
entered the wake region. Argentini et al. (2011)
also performed wind tunnel tests to measure the
aerodynamic forces and surface pressures on a
stationary vehicle at different locations in the
wake of a bridge tower. Smoke visualization was
used to observe the flow patterns around the
vehicle. The side, rotating moment and yawing
moment coefficients were presented. The length
scale they used was 1:40 and the ratio of the
vehicle length to the tower width was 1.66. Their
results showed that the aerodynamic forces on the
stationary vehicle actually increased when the
vehicle was behind the bridge tower because the
vehicle was longer than the tower width, and that
the end parts of the vehicles were blown over by
the flow that was accelerated by the interaction
with the tower. The smoke visualization also
showed that suction appeared between the vehicle
and the bridge tower with an axial flow moving

Received: 15 May 2013; Revised: 28 Jan. 2014; Accepted: 4 Mar. 2014

330
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)
upwards along the tower axis. The pressures were
negative on all of the vehicle’s surfaces.
Apart from wind tunnel tests, the computational
fluid dynamics (CFD) simulation is now applied
widely to solve flow-structure interaction
problems in engineering (Murakami, 1997; Wu
and Chau, 2006; Cochran and Derickson, 2011;
Andersson et al., 2013). To solve the complicated
flow around the vehicle-tower-deck system, Large
Eddy Simulation (LES) is the ideal method to use
in terms of computation accuracy. However, a
high level of mesh resolution and computational
effort should be applied, especially for a
multiscale system with components of various
sizes. Since the main focus of this study at the
present stage was to explore the turbulence-
averaged aerodynamic forces of a road vehicle
passing by a bridge tower for engineering
application, the Reynolds-Averaged Navier-
Stokes (RANS) simulation method, which
requires less computational effort, was used in
this study instead of LES.
RANS has been adopted widely for wall-bounded
flow. In the aerodynamics study of vehicles,
Hargreaves and Morvan (2008), Guilmineau and
Chometon (2009) and Gunes (2010) simulated the
aerodynamics of a vehicle using RANS. Bettle et
al. (2003) simulated a truck on a bridge using the
steady Reynolds Averaged Navier-Stokes
(RANS) method. Sterling et al. (2010) simulated
the crosswind effects on a high-sided lorry on the
ground. RANS has also been used successfully
Fig. 1 Elevation of Xiangshan Port Bridge (unit: cm).
Protection rail
Hand rail
Wind
Side fairing
7800
Maintenance trace
Fig. 2 Dimensions of bridge tower (unit: cm).
331
for studying the aerodynamics of stationary
bridge deck sections (Bruno and Khris, 2003; Yoo
et al., 2009; Fransos and Bruno, 2010; Mannimi
et al., 2010) and the aerodynamics of bridge deck
sections with motion (Zhu et al., 2007; Huang and
Liao, 2011). In this regard, closure is of critical
importance in industrial applications of RANS
(Corson et al. 2009). The SST k- turbulence
model, a combination of the traditional k- and k-
 models, is one of the successful closure models
for wall-bounded flow (Menter, 2009; Zingg and
Godin, 2009).
The study reported in this paper used the RANS
method, supplemented by the SST k- turbulence
model, to investigate crosswind effects on road
vehicles passing by a bridge tower. The
aerodynamic forces on a stationary vehicle in the
wake of a bridge tower were measured first in a
wind tunnel. The aerodynamic forces
(coefficients) on, flow field around, and surface
pressure distribution over the stationary vehicle in
the wake of the bridge tower were then computed,
and the results were compared with the test
results. The motion of the vehicle was considered
in the CFD simulation using the dynamic mesh
method, and the aerodynamic coefficients of the
moving vehicle passing by a bridge tower were
computed in addition to the flow field around and
the surface pressure distribution over the vehicle.
The computed results of the moving vehicle were
finally compared with those of the stationary
vehicle.
Protection rail 34000 Protection rail Protection rail
3000 2×3750=7500 2×3750=7500 3000
Lane 1 Lane 2 Lane 3 Lane 4 Hand rail
3500
Side fairing
6000 6000 7800
3200 3200
Maintenance trace
Fig. 3 Cross section of bridge deck (unit: mm).
 Engineering Applications
Engineering of Computational
Applications Fluid Mechanics
of Computational Vol. 8,Vol.
Fluid Mechanics No. 8,
3, No.3
pp. 330–344
(2014) (2014)

2. WIND TUNNEL TESTS above the tunnel floor (see Fig. 5). A uniform
upcoming wind was generated perpendicular to
2.1 Models and test arrangements the length of the deck. The aerodynamic forces on
the stationary vehicle model were measured by a
As pointed out by Baker and Reynolds (1992), force balance for three positions of the vehicle on
high-sided vehicles are a most dangerous type of the first upstream deck lane: ZV=0, LV and 2LV (ZV
vehicle, with a tendency to suffer overturning or is the distance from the right bound of the tower
sideslip accidents under crosswinds. For this to the head surface of the vehicle as shown in Fig.
reason, a high-sided lorry passing a long-span 6; and LV is the maximum length of the vehicle).
cable-stayed bridge was taken as the study case. A gap of 1 mm was left between the vehicle
The selected high-sided articulated lorry was model and the upper surface of the deck model to
investigated by Coleman and Baker (1990) in prevent the contact of vehicle wheel sets to the
wind tunnels and later used in the dynamic deck surface during tests. The details of the tests
analysis of wind-vehicle-bridge systems (Xu and can be found in Li (2009).
Guo, 2003b; Cai and Chen, 2004; Chen and Wu,
2008). The Xiangshan Port Bridge in China was
taken as a reference bridge; this is a cable-stayed
bridge of two towers with a main span of 688m
(see Fig. 1). Each bridge tower contains a pair of
legs side-by-side (see Fig. 2). The two legs are
inclined from the bridge deck to both their ends
with a chamfered rectangular cross section. The
bridge deck is a flat box girder with side fairing as
shown in Fig. 3. The cross section of the deck is
34.0m wide and 3.5m high carrying a dual two-
lane highway on the upper surface. Two lines of
handrails, four lines of protection rails and two
lines of I-shape maintenance traces are mounted Fig. 5 Wind tunnel tests of vehicle.
on the bridge deck. For this study, the geometric
scale was selected as 1:25 for both the bridge Wind
deck and the vehicle. Fig.4 shows the geometric
dimensions of the vehicle model. The ratio of the ZV
Tower
vehicle length to the tower width was 1.35. The
Vehicle
deck sectional model with a width of 1.36m and a
length of 5m were manufactured for wind tunnel Deck

tests.
1 Tower

Fig. 6 Illustration of vehicle location.

2 2

2.2 Definitions of aerodynamic coefficients

and measured results
1

Six aerodynamic forces and moments on the

vehicle in the Cartesian coordinate were measured:
lift force FL, drag force FD, side force FS, pitching
moment MP, yawing moment MY and rolling
Fig. 4 Dimensions of vehicle model with length scale moment MR, as shown in Fig. 7. All these
of 1:25 (unit: mm). components referred to the gravity center of the
vehicle. The corresponding non-dimensional
The wind tunnel tests were carried out in the TJ-3 aerodynamic coefficients can be defined by
wind tunnel of the State Key Laboratory for F F FS
Disaster Reduction in Civil Engineering at Tongji C L  L ; C D  D ; CS  (1)
qA qA qA
University in China. This is a closed-circuit wind
MP MY MR
tunnel with a test section 15m wide, 2m high and CP  ; CY  ; CR  (2)
14m long. The bridge deck model was installed qALV qALV qALV

332
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

q  0.5U 2 (3) this method was used to compute the flow field
where ρ is the air density; U represents the wind numerically around the vehicle-deck-tower
velocity magnitude referred to the vehicle; q is the system. The basic idea behind RANS is to
dynamic pressure; A is the frontal project area of average the instantaneous flow governing
the vehicle without wheels, which refers to the equations in the time domain. After being
project area in the X-Y plane, as shown in Fig. 7. averaged, the incompressible flow governing
equations become
ui
0 (4)
xi
FL

Y
MY
ui u p  ui u j
' '

  u j i     2ui  (5)
t x j xi x j
·

Wind
·
G rav ity C en ter where t is the time; xi is the coordinate in the ith
α
axis in the Cartesian coordinate system; ρ and μ
MP
X
are the density and dynamic viscosity coefficient
MR
FS of air, respectively; ui is the velocity component
Z
along the xi-axis; u′i is the fluctuation part of ui; p
FD is the pressure; the over bar represents the mean
value; and -ρu′i u′j is the so-called Reynolds stress
Fig. 7 Definitions of aerodynamic forces on vehicle. represented by the SST k-ω turbulence model in
this study. The use of Equations 4 and 5 to solve
Table 1 Measured aerodynamic coefficients of vehicle. the averaged flow is called the unsteady RANS
method. If the first term in Equation 5 is
CS CL CD CP CY CR neglected, it is then called the steady RANS
ZV=0 1.145 0.112 -0.136 -0.062 -0.670 -0.129 method. In this study, the unsteady RANS method
ZV= LV 4.120 -0.939 0.499 -0.225 -0.371 -0.251
was used in all of the numerical simulations,
except for the determination of the meshing
ZV=2LV 4.494 -0.837 0.490 -0.248 -0.473 -0.236
scheme, in which the steady RANS method was
used to reduce the extremely demanding
The measured averaged aerodynamic coefficients computational time. In this study, the governing
of the vehicle are listed in Table 1. It can be seen equations were discretized using the QUICK
that, from ZV=0 to ZV=2LV, the vehicle departs scheme, based on the finite volume method. The
gradually from the bridge tower. The side SIMPLEC algorithm was employed for the
coefficient increases but the pitching moment coupling of velocity and pressure. The time
coefficient decreases. The lift coefficient and the integration was performed using the first-order
rotating moment coefficient decrease first and implicit method. The CFD code Fluent was
then increase slightly, whereas the drag employed to solve these governing equations. The
coefficient and the yawing moment coefficient numerical simulation provided the pressure and
increase first and then decrease slightly. All shear stress distribution over the surfaces of the
coefficients show a steep gradient from ZV=0 to vehicle. The aerodynamic forces acting on the
ZV=LV. It is worthwhile to mention that the vehicle could then be acquired by integrating the
variations of both the side coefficient and the pressures and the shear stresses over the surfaces.
yawing moment coefficient have the same trend
as the test results presented by Charuvisit et al. 3.2 Computational domain and boundary
(2004). condition

3. STATIONARY VEHICLE SIMULATION To facilitate the validation of the computational

3.1 Simulation scheme
In consideration of the complexity of the problem
under consideration, this study considered only
the averaged aerodynamic forces on the vehicle.
Owing to the lower computation effort required
when using RANS to solve the averaged flow,
results, the same vehicle and bridge models used
in the wind tunnel tests were used in the
numerical simulations. The numerical model of
the bridge deck had the same size and
configuration as the wind tunnel model of the
bridge deck. The maintenance traces of the bridge
deck were also modelled in the light of the
principle of geometric similarity. The hand and

333
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

b_up
The flows at these boundaries were assumed to be
Inflow
Vehicle
Tower
uniform and the gradients of flow variables

b_right
Deck
(including wind velocity and pressure) normal to
b_left

VW

Dy
the boundaries were zero. The boundaries b_up
and b_down were set as no-slip wall boundaries.
3.5Bd Bd
b_down
10.5Bd The flow cannot penetrate the ground and the
Dx lorry surfaces, hence the no-slip wall was also
(a) Section in wind direction assigned to the deck, the tower and the vehicle
surfaces.
b_up

Tower
zV Top View
Vehicle

b_head
Deck
b_tail

Dy
LV

B_down
LD
Dz

Front View
Back View
(b) Section perpendicular to wind direction
Side View
Fig. 8 Computational domain sketch: vehicle-deck-
tower system.

protection rails were simplified as continuous

horizontal bars, based on the principle that the Bottem View
heights of the bars and the ventilation ratio were
identical to the physical model used in the wind Fig. 9 Grid distributions over vehicle surfaces.
tunnel tests. The numerical model of the road
vehicle was also constructed with the same size 3.3 Meshing
and configuration as the physical vehicle model
used in the wind tunnel tests. Fig. 8 shows the As seen in Figs. 2 and 3, the characteristic sizes of
vehicle-deck-tower model enclosed in a the components of the vehicle-deck-tower system
computational domain with the shape of cuboid. vary to a great extent, from the small size of the
The sizes of the domain are identified as Dx along rails to the large size of the tower. The complex
the wind direction, Dy vertical to the ground and geometric shape and the large variation in the
Dz along the length direction of the bridge deck. characteristic size lead to a complicated flow field
In order to obtain the same blockage ratio (the around the system in both time and space
ratio of the blockage area of the models to the domains. In order to obtain a relatively adequate
cross section area of upcoming winds) as in the spatial discretization grid system corresponding to
wind tunnel, the cross section perpendicular to the the capability of the computer, grid optimization
upcoming wind direction of the computational was performed. Since the simulation with the
domain shown in Fig. 8 is equal to the wind unsteady RANS method is time-consuming for
tunnel, that is, Dy=2m and Dz=15m. Bd is the the flow around the three-dimensional complex
width of the deck. LV and LD are the lengths of the geometric body, the steady RANS equations
vehicle and the deck, respectively. All coupled with the SST two-equation turbulence
boundaries, including the six outer boundaries model were solved during the process of grid
and the surfaces of the vehicle, the deck and the optimization. Four meshing schemes with
tower were assigned with boundary conditions different grid sizes were generated for the vehicle
approximating the experimental situation. b_left is to have a relative distance zV=LV from the tower.
the source of upcoming wind; a uniform wind In the four meshing schemes, the grid
speed of 10m/s, turbulence kinetic energy k of distributions over the vehicle surfaces and the
0.05 and special dissipation ratio of 2 were height of the first layer grid near the walls
assigned to this boundary. After the flow passes remained consistent, and the grid density in the
the vehicle, wind blows out of the domain left flow regions became the focus. The height of
through the outer boundary b_right. Thus, b_right the first layer grid near the vehicle’s surfaces was
was specified as flow outlet with zero pressure. 1×10-5 m, which ensured that the y+ of the first
The outer boundaries b_head and b_tail are layer grid near the walls was below 1 in the
parallel to the direction of the upcoming wind. simulation. The grid distributions over the

334
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

surfaces of the vehicle are shown in Fig. 9.
Totally, about 56 thousand grids were used for the
vehicle surfaces. Meshing scheme 1 was the
coarsest case, with 15.5 million grids, as shown in
Fig. 10. Meshing scheme 2 was generated through
the refinement of the grid sizes along the direction
of the deck length, while meshing scheme 3 was
meshed through the refinement of the grid sizes in
the plane perpendicular to the direction of the
deck length. Meshing scheme 4 was the result of
refining the grid sizes along and perpendicular to
the direction of the deck length. As a result,
meshing schemes 2, 3 and 4 had grid numbers of
20.6, 22.7 and 31.8 million, respectively. The
maximum y+ on the walls of the tower, the deck
and the up and down boundary faces were about
2, 5 and 5, respectively.
differences of the side coefficient and the rotating
moment coefficient are within 5% for different
meshing schemes. The maximum relative
difference is the lift coefficient among all of these
aerodynamic coefficients. For the lift coefficient,
the space underneath the bottom of the vehicle is
very small and the geometric shape of the bottom
surface is very complex. As a result, the flows
underneath the vehicle bottom and the pressure on
the bottom surface are hard to predict with a high
degree of accuracy in the numerical simulation.
The lift force, or the lift coefficient, is also hard to
predict because it is caused mainly by the
pressure difference between the top and bottom
surfaces of the vehicle. Fortunately, lift force is
less significant to the behavior of the vehicle
under crosswinds compared with other force
components such as side force and yawing and
rotating moments. Therefore, it can be concluded
that the computed aerodynamic coefficients are
not too sensitive to the mesh size and time step
within the range selected, hence meshing scheme
1 was adopted in this study with the least
computational effort.

3.4 Time step and length

A characteristic time t* was defined as the ratio of

a characteristic length to a characteristic velocity
of the flow system. In the case of a vehicle
staying on the ground, wind blows around the
Fig. 10 Meshing scheme 1. vehicle, and the characteristic length and
characteristic velocity were selected as the width
Table 2 Simulated aerodynamic coefficients of vehicle of the vehicle and the upcoming wind velocity.
with different meshing schemes. As a result, the characteristic time was expressed
as
Meshing CS CL CD CP CY CR B
scheme t*  (6)
1 4.986 -0.070 0.378 -0.138 -0.611 -0.255 U
2 4.996 -0.089 0.347 -0.130 -0.623 -0.254 where B and U∞ are the width of vehicle and the
3 5.071 -0.137 0.339 -0.113 -0.567 -0.249 upcoming wind velocity, respectively.
4 4.958 -0.185 0.349 -0.116 -0.577 -0.246
The time step and length for the computation
were set based on t*. For the case zV=LV, two time
steps 0.1t* and 0.05t* were simulated to check the
Table 3 Simulated aerodynamic coefficients of vehicle
influence of time step on the simulation results.
with different time steps.
For all the cases of the stationary vehicle, the first
Time step CS CL CD CP CY CR 60t* was treated as a converging process and the
0.1t* 5.006 -0.116 0.350 -0.130 -0.626 -0.256
corresponding results were not taken into account.
The next 60t* was accepted as the time length for
0.05t* 4.998 -0.111 0.355 -0.130 -0.623 -0.256
the normal computational results. The following
results, including the aerodynamic coefficients,
The steady RANS method was utilized to obtain are all the averaged values on the last 60t*. The
the numerical results for the four grid systems. streamlines, velocity contour and pressure
The mean aerodynamic coefficients obtained from coefficients were taken from the time at 120t*.
1000 iterations after the first computation of 2000 The simulated aerodynamic coefficients from
iterations are presented in Table 2. The relative different time steps (0.1t* and 0.05t*) using the

335
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

unsteady RANS method can be seen in Table 3. It simulated side coefficients, lift coefficient and
can be seen that a time step of 0.1t* is accurate absolute yawing moments are larger than the
enough compared with a shorter time step. 0.1t* measured ones while the rotating moment
was thus selected as the time step to be used in coefficients are close to the measured ones. An
the computation of aerodynamic forces on the initial explanation about the differences is that the
stationary vehicle. velocities perpendicular to the lane in the wake of
the tower are overestimated, leading to a high side
Table 4 Simulated aerodynamic coefficients of vehicle force and yawing moment, but the velocities
and comparison with test results. along the lane are underestimated, leading to a
low drag force and pitching moment. The
CS CL CD CP CY CR
differences in aerodynamic coefficients between
SIM 2.005 0.164 -0.017 -0.030 -0.707 -0.108
ZV=0 the simulations and wind tunnel tests were
RD(%) 75 46 88 52 -6 16
expected in the current study, considering the
SIM 5.006 -0.116 0.350 -0.130 -0.626 -0.256 complexity of the problem and uncertainties
ZV= LV
RD(%) 22 88 -30 42 -69 -2 involved in wind tunnel tests and numerical
ZV=2LV
SIM 6.241 0.030 0.058 -0.181 -1.005 -0.303 simulations, but a further study is required to
RD(%) 39 104 -88 27 -113 -28 reduce further these differences when both
experimental and numerical simulation techniques
SIM: Simulation value; RD: Relative Difference: (Value are developed to a mature level for the problem
of SIM-Value of test)/| Value of test| *100% concerned. Nevertheless, it will be conservative
from the practical viewpoint of vehicle safety
under crosswinds if the simulated aerodynamic
forces are used. Furthermore, the CFD
simulations can provide further information on
flow fields around and pressure distributions over
the surfaces of the vehicle as demonstrated below.

3.6 Flow field and surface pressure

The flow field around the vehicle is illustrated in

(a) Section 1-1
(b) Section 2-2
Fig. 11 Projected instantaneous streamlines
velocity contours (ZV=0) (unit: m/s).
3.5 Simulated aerodynamic coefficients and
comparison with test results
The simulated aerodynamic coefficients of the
vehicle and their comparisons with wind tunnel
results are listed in Table 4. The maximum
relative difference between them is the yawing
moment coefficient, while the maximum absolute
difference occurs on the side coefficient. The
two main planes: section 1-1 and section 2-2 as
shown in Fig. 4. Cross section 1-1 is located at the
middle length of the vehicle’s trailer, while cross
section 2-2 is in the middle height of the vehicle.
The velocity vectors in the plane were first
extracted from the original 3D flow field. The
contours of the 2-norms of these velocity vectors
and the streamlines of the projected vehicle
vectors on the plane were taken to show the flow
field. At ZV=0, the vehicle was just behind the
bridge tower, and the projected streamlines and
velocity contours of the vehicle-deck-tower
system in the two selected main planes are
displayed in Fig. 11. Although the ratio of the
and vehicle length to the tower width in this study was
smaller, an obvious suction region can be
observed between the vehicle and the tower. This
region was also observed from the smoke
visualization done by Argentini et al. (2011).
From Fig. 11a, it is also noted that a large vortex
is generated near the upper boundary. This is the
by-products of the combination of strong
separation flows from the vehicle and the flow
drag from the up wall that should not exist in real
situation. Nevertheless, it can also be seen from
Fig. 11a that the flow velocities involved in the

336
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

large vortex are very small and therefore the large

vortex is expected to have slight effects on the
computed aerodynamic forces on the vehicle. At
ZV= LV and ZV=2LV, the vehicle has moved out of
the tower. The projected streamlines and velocity
contours of the vehicle-deck-tower system are
shown in Figs. 12 and 13 for ZV=LV and Z V=2LV,
respectively. Clearly, the flow fields shown in (a) Section 1
Figs. 12 and 13 are quite different from those
shown in Fig. 11. The location of the vehicle
relative to the tower affects the flow field around
the vehicle significantly. In view of section 1-1,
the wake region around the vehicle-deck is much
higher for the ZV=0 case compared with the ZV=LV
case. The effects of the tower on the flow around
the vehicle are limited for the ZV=2LV case in
view of section 2-2.
At the position ZV=0 and ZV= LV of the vehicle,
the pressure coefficient distributions are displayed
in Fig. 14. The pressure coefficient was defined as
the ratio of the pressure to the dynamic pressure
(b) Section 2
q. Since the length of the vehicle was slightly
longer than the width of the tower, the tail part of Fig. 13 Projected instantaneous streamlines and
the vehicle was not shielded by the tower at the velocity contours (Zv= 2Lv) (unit: m/s).
position ZV=0. High positive pressures occur on
the tail part of the windward side surface of the
vehicle as in Fig. 14. From the case ZV=0 to the
case ZV= LV, the positive pressures extend to most
parts of the windward side surface of the vehicle.

(a) ZV=0

(a) Section 1

(b) Section 2 (b) ZV= LV

Fig. 12 Projected instantaneous streamlines and Fig. 14 Instantaneous pressure distributions over
velocity contours (Zv= Lv) (unit: m/s). surfaces of vehicle.

337
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)
Inflow
Vehicle
Deck
VW
b_lef
t
Dynamic mesh
region
3.5Bd Bd
(a)
Vehicle
VV
LL=10m
(b)
Fig. 15 Computational domain for moving vehicle-
deck-tower system.
4. MOVING VEHICLE SIMULATION
4.1 Simulation scheme
The dynamic meshing method implemented in the
software Fluent was activated to simulate the
movement of the vehicle. The total computational
domain was divided into two regions, as seen in
Fig. 15a. The one surrounding the vehicle is a
dynamic mesh region, where the meshes move
with the same velocity as the vehicle moves. The
other region is a stationary mesh region without
the motion of meshes. The interface defined in
Fluent was assigned to the contact faces between
the two regions. For the flow in the stationary
region, the common governing equations (Eqs. 5
and 6) were applied. In the region of the dynamic
mesh, the conservation equation for a general
variable  , which represents the flow velocity
component ui , turbulence kinetic energy k or
special turbulence dissipation rate ω for the
turbulence model used in this study, on an
arbitrary grid volume V can be written as (Ansys
13.0 help):
d
dt V
 dV    (u  ug )  dA     dA   S dV
V V
where u is the flow velocity vector; u g is the
velocity vector of the moving mesh;  is the
diffusion coefficient; S is the source term of  ;
V is the boundary face of the control volume;
and A is the area vector of V . The discritization
methods are the same as used in Section 3.1.
It can be seen from Fig. 15a that the dynamic
mesh region is enclosed by the contact faces with
b_up
the stationary mesh region and a part of the deck
surface. After each time step, offsets of the
Tower
meshes occured at the contact faces between the
b_right
Dy
stationary and dynamic mesh regions. The
locations of the meshes in the dynamic mesh
region were updated using the dynamic layering
b_down model. The layer split factor and layer collapse
10.5Bd
Dx
factor were both set as 0.5. The detailed
information about the dynamic layering model
can be found in the help document of Fluent.
Moreover, part of the deck surface moved
Tower
together with the dynamic mesh region in the
ZV
Deck numerical simulation, but at the same time the
LR=5m
moving velocity of the vehicle was oppositely set
on that part of the deck surface to keep the deck
VV stationary.
4.2 Computational cases
To ensure the proper simulation of flows around
the moving vehicle when it passes by the tower,
the length of the bridge deck had to be increased
from 5m to 15m, as shown in Fig. 15b. The
vehicle then moved from the left to the right. A
distance of 10m was arranged at the left side of
the tower so that it was long enough for
developing a rational flow around the moving
vehicle before it passed by the tower. The
increase of the deck length resulted in a higher
blockage ratio. To keep the blockage ratio
unchanged, the height of the computational
domain was also increased, from 2m to 4.43m. As
a result, the number of grids increased to 20.7
million, with the same grid distribution around the
vehicle-deck-tower surfaces as the stationary
case. The outer boundary settings of the
computational domain were the same as those
used in the stationary case except for the surfaces
of the vehicle and the wind inflow boundary
b_left. The surfaces of the vehicle were also non-
slip wall boundaries, but with a moving velocity
VV along the deck. The vehicle was positioned at
ZV = -8.52m initially and then accelerated to the
designated velocity at a distance of 2.7m (equal to
5LV). A relative yaw angle  was defined to
(7) reflect the relative magnitude of wind velocity to
V
vehicle velocity:
VW
  arctan( ) (8)
VV
where VW and VV are the designated wind velocity
and the designated vehicle velocity, respectively.
Three moving cases of  =0°(VW=0, VV=10m/s),
30°(VW=5m/s, VV=8.66m/s) and 60° (VW=8.66m/s,
VV=5m/s) were computed. Therefore, different
wind velocities were assigned to the inflow
338
 Engineering
Engineering Applications
Applications of Computational
of Computational Fluid Mechanics
Fluid Mechanics Vol. 8, Vol.
No. 3,8, pp.
No.3 (2014) (2014)
330–344

7
0° 30° 60° 2.5
0° 30° 60°
6
2.0

1.5
4
CS

1.0

CL
3

2 0.5

1
0.0

0 -BT LV
-0.5
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

ZV (m) ZV (m)

(a) side coefficient (b) lift coefficient

0° 30° 60°
0.3
0.4
0° 30° 60°

0.2
0.0

0.1
-0.4
CD

CP

-0.8 0.0

-1.2
-BT LV -0.1
-BT LV

-1.6 -0.2
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

ZV (m) ZV (m)

(c) drag coefficient (d) pitching moment coefficient

0.6 0.1
0° 30° 60° 0° 30° 60°
0.4

0.0
0.2

0.0
-0.1

-0.2
CR
CY

-0.4 -0.2

-0.6

-0.3
-0.8

-BT LV -BT LV
-1.0
-0.4
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

ZV (m) ZV (m)

(e) yawing moment coefficient (f) rotating moment coefficient

Fig. 16 Aerodynamic coefficients of moving vehicle.

boundary b_left. For each case, the time step was the vehicle is entering the tower region. At ZV=0,
set as 0.01m/ VV. the head surface of the vehicle is aligned with the
right boundary of the tower, which means the
4.3 Aerodynamic coefficients and comparison vehicle is exiting the tower region. At ZV=LV, the
with stationary vehicle tail surface of the vehicle is aligned with the right
edge of the tower, and the entire vehicle is just
The simulated aerodynamic coefficients of the out of the tower region.
vehicle during the period of passing by the tower At 0°, there was no action of crosswind. The
are shown in Fig. 16, from ZV= -3.78m to ZV= influences of the tower on the aerodynamic forces
3.78m for the three velocity cases. At ZV=-BT in of the vehicle are thus limited. The side
Fig. 16, the head surface of the vehicle is aligned coefficient, the yawing moment coefficient and
with the left boundary of the tower, which means the rotating moment coefficient are almost zero.

339
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

The negative lift coefficient of the vehicle at 0°
indicates that the vehicle’s ability to attach to the
ground is enhanced.
When the vehicle is at the location without the
influence of the tower (eg. ZV=-3.78), its side
coefficient, yawing moment coefficient and
rotating moment coefficient increase with the
increase of crosswind velocity. Different from
these coefficients, the lift and pitching moment
coefficients are related not only to the crosswind
velocity but also to the vehicle running velocity.
Therefore, they do not show a monotonic
relationship with . Intuitively, the drag force
decreases with the decrease of vehicle velocity.
However, a large drag force occurs at the front
surface of the trailer at a low yaw angle, which
makes the drag force at 30° larger than those at 0°
and 60°.
The aerodynamic coefficients of the moving
vehicle at positions ZV =0, LV and 2LV are
compared with those of the stationary vehicle at
the same positions in Table 5. It can be seen that
the side coefficient of the moving vehicle and the
absolute value of the yawing moment coefficient
of the moving vehicle increase with the relative
angle and are lower than those of the stationary
vehicle at the same positions. There are no
obvious magnitude relations for other
aerodynamic coefficients between the moving and
stationary vehicles.
The shielding effects of the tower on the
aerodynamic forces of the vehicle are significant
in the cases where =30° and =60°. As the
vehicle approaches the tower, the absolute values
of the side coefficient, the yawing moment
coefficient and the rotating moment coefficient
increase continuously to their respective peak
values due to the accelerated flow at the side of
the tower. The side coefficient and the rotating
moment coefficient then decrease to the lowest
values as the vehicle is totally immerged by the
tower (about ZV=0) because the flows are shielded
by the tower. These two coefficients reach peak
values again as the vehicle moves out of the tower,
because the flow is accelerated at the right side of
it. The value of the right peak is higher than that
of the left peak, which is the same as the test
results presented by Charuvisit et al. (2004). The
situation becomes different for the yawing
moment coefficient. The side forces act mainly on
the trail part of the vehicle as it enters the tower
region and on the head part of the vehicle as it
moves out of the tower region. Therefore, the
yawing moment reaches a maximum value when
the side force decreases at the location near -BT,
and it then reduces and increases again. The
accelerated flows at the sides of the tower also
yield peak values for the lift and pitching
moment. At the position of the vehicle directly
behind the tower (about ZV=0), the lift and the
pitching moment are small. At the two sides of
the tower, the accelerated flows lead to a higher
relative yaw angle of wind to the vehicle. As a
result, the drag force decreases at these positions.

Table 5 Comparison of aerodynamic coefficients between moving and stationary status.

CS CL CD CP CY CR
0° -0.043 -0.224 -0.735 -0.066 0.005 0.000
30° 0.889 0.516 -1.036 0.069 -0.243 -0.059
ZV=0
60° 1.671 0.302 -0.268 -0.022 -0.492 -0.264
Stationary 2.005 0.164 -0.017 -0.030 -0.707 -0.108
0° 0.021 -0.212 -0.746 -0.066 -0.011 0.000
30° 2.742 1.405 -1.241 0.065 -0.210 -0.179
ZV= LV
60° 5.361 1.668 -0.407 0.098 -0.620 -0.283
Stationary 5.006 -0.116 0.350 -0.130 -0.626 -0.256
0° 0.003 -0.217 -0.748 -0.067 -0.001 0.000
30° 2.785 1.342 -1.274 0.091 -0.284 -0.159
ZV=2LV
60° 5.160 0.766 -0.441 -0.080 -0.556 -0.096
Stationary 6.241 0.030 0.058 -0.181 -1.005 -0.303

340
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)

Table 6 Characteristic values of aerodynamic coefficients.

Peak Transition Transition Non-dimensional

Standard Valley ratio
ratio distance (m) time (s) time
30° 2.376 1.253 0.374 0.665 0.077 1.667
CS
60° 4.575 1.239 0.288 0.545 0.109 1.363
30° 1.369 1.150 0.236 0.605 0.070 1.516
CL
60° 0.953 1.844 0.286 0.524 0.105 1.313
30° -1.373 0.680 1.066 1.642 0.190 4.114
CD
60° -0.562 0.338 1.281 0.438 0.088 1.100
30° 0.092 1.669 -1.418 0.526 0.061 1.321
CP
60° 0.108 0.922 -0.973 0.283 0.057 0.713
30° -0.220 -0.568 1.980 0.378 0.044 0.953
CY
60° -0.475 -0.743 1.616 0.404 0.081 1.013
30° 0.141 1.261 0.317 0.443 0.051 1.104
CR
60° -0.235 0.320 1.276 0.567 0.113 1.413

(a) section 1-1 (b) section 2-2

Fig. 17 Flow velocity contour and projected streamlines (α=0°) (unit: m/s).

(a) section 1-1 (b) section 2-2

Fig. 18 Flow velocity contour and projected streamlines (α=30°) (unit: m/s).

(a) section 1-1 (b) section 2-2

Fig. 19 Flow velocity contour and projected streamlines (α=60°) (unit: m/s).

341
 Engineering Applications
Engineering of Computational
Applications Fluid Fluid
of Computational Mechanics Vol. 8,Vol.
Mechanics No.8,3,No.
pp. 3330–344
(2014) (2014)

The drag force approaches to the same value as

the case of no crosswind if the vehicle is just
behind the tower. This is why the drag coefficient
decreases at 30° and increases at 60° when the
vehicle is behind the tower.
In general, the vehicle is shielded from the
crosswind by the bridge tower with a low side
force, yawing moment and rotating moment.
When it moves out of the shelter it enters a sharp-
edged crosswind gust. All the aerodynamic
coefficients experience fluctuation. Table 6 lists
the standard value, the maximum peak ratio, the
(a) α=0°
minimum valley ratio, the transition distance and
time, the non-dimensional transition time for each
of six aerodynamic coefficients. The standard
value was defined as the average of the two
values at both ends of Fig. 16. The maximum
peak ratio and the minimum valley ratio are the
ratios of the maximum peak value and the
minimum valley ratio to the standard value. The
transition distance and the transition time are the
traveling distance and the traveling time between
the maximum peak and the minimum valley. The
side coefficient, the yawing coefficient and the
rotating moment coefficient are the most
important quantities for the safety analysis of the (b) α=30°
vehicle under the crosswinds. Of them, the
yawing moment coefficient experiences the
sharpest change between its maximum peak ratio
and its valley ratio in the shortest times of 0.044
and 0.081s at 30° and 60°, respectively. A non-
dimensional time was defined as the ratio of
transition time to a reference time (the vehicle
velocity divided by the width of the tower) and
this is also listed in Table 6. For the prototype
bridge with a width of 10m, a vehicle velocity of
17.3m/s and a wind velocity of 10m/s, the
transition time of yawing moment derived from
the non-dimensional time is 0.55s, which is lower (c) α=60°
than the median driver reaction time of 0.66s
specified in AASHTO (2004). Therefore, the Fig. 20 Mean pressure distributions over surfaces of
wind environment for the vehicle passing by the vehicle (ZV= LV).
bridge tower deserves serious attention.
crosswind. For section 1-1, air circulations are
4.4 Flow field and surface pressure generated at the upper, bottom and leeward side
surfaces of the vehicle at 30°, and the flow
The projected streamlines and velocity contours separates at the windward corners of the vehicle
of the vehicle-deck-tower system in the two at 60°. This indicates that the flow around the
selected planes (see Fig. 4) of the vehicle when it vehicle changes from the lengthwise feature to the
enters behind the tower are displayed in Figs. 17 crosswise feature as the relative yaw angle
to 19. At 0°, it is obvious that the head of the increases. For section 2-2, the flow around the
vehicle pushes the air away and the air then vehicle interacts with the wake of the windward
converges at the tail of the vehicle. At other tower and the symmetric wake feature behind the
angles, the air near the surfaces of the vehicle is windward tower is disturbed by the motion of the
driven by both the motion of the vehicle and the vehicle. The developing direction of the wake

342
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)
behind the trail of the vehicle is along the
direction of the resultant velocity of the crosswind
and the vehicle velocity. The pressure coefficient
distributions over the vehicle surfaces at the
location ZV=LV are displayed in Fig. 20. At 0°,
positive pressures exist only on the head surface.
Positive pressures extend to the windward side
surface at 30° and further to the leeward side
surface, the roof and the bottom surface at 60°.
5. CONCLUSIONS
The computational fluids dynamics (CFD)
technique has been used to obtain the
aerodynamic forces acting on a stationary/moving
vehicle as it passes by a bridge tower. The
stationary vehicle cases were first simulated for
three different positions of the vehicle relative to
the bridge tower. The simulated results were
compared with the wind tunnel test results. It was
found that the simulated aerodynamic coefficients
are larger than those measured from the wind
tunnel tests. Through the simulation of the
moving vehicle passing by the bridge tower using
the dynamic mesh method, the dynamic
aerodynamic coefficients were computed.
Significant shielding effects of the tower on the
aerodynamic forces of the vehicle were found.
The transition time of aerodynamic forces due to
the sudden gust and the shielding effect is so short
that there may not be enough time for a driver to
react. With the further advancement of computer
capacity, the instantaneous aerodynamic forces of
a moving vehicle on a bridge deck passing by a
bridge tower should be investigated in the future.
Moreover, if wind tunnel test technique is
advanced to be able to conduct such an
experiment, the verification of the numerical
simulation scheme presented in this paper against
the experimental results should be performed.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the financial
supports from The Hong Kong Polytechnic
University through a PhD studentship to the first
author and the Research Grants Council of the
Hong Kong through a competitive grant (PolyU
5311/07E) to the second author. The authors also
thank the Wuxi Supercomputing Center for the
computation service they provided.
REFERENCES
1. American Association of State Highway
Officials (2004), A Policy on Geometric
343
Design of Highways and Streets. Washington
DC, USA.
2. Andersson AG, Andreasson PA, Lundstrom
TS (2013). CFD-modelling and validation of
free surface flow during spilling of reservoir
in down-scale model. Engineering
Applications of Computational Fluid
Mechanics 7(1): 159-167.
3. Ansys Inc. Fluent help in Ansys 13.0.
4. Argentini T, Ozkan E, Rocchi D, Rosa L,
Zasso A. (2011). Cross-wind effects on a
vehicle crossing the wake of a bridge pylon.
Journal of Wind Engineering and Industrial
Aerodynamics 99(6-7): 734-740.
5. Baker CJ (1986). A simplified analysis of
various types of wind-induced road vehicle
accidents. Journal of Wind Engineering and
Industrial Aerodynamics 22(1): 69-85.
6. Baker CJ (1987). Measures to control vehicle
movement at exposed sites during windy
periods. Journal of Wind Engineering and
Industrial Aerodynamics 25(2): 151-161.
7. Baker C, Cheli F, Orellano A, Paradot N,
Proppe C, Rocchi D (2009). Cross-wind
effects on road and rail vehicles. Vehicle
System Dynamics 47(8): 983-1022.
8. Baker CJ, Reynolds S (1992). Wind-induced
accidents of road vehicles. Accident Analysis
& Prevention 24(6): 559-575.
9. Bettle J, Holloway AGL, Venart JES (2003).
A computational study of the aerodynamic
forces acting on a tractor-trailer vehicle on a
bridge in cross-wind. Journal of Wind
Engineering and Industrial Aerodynamics
91(5): 573-592.
10. Bruno L, Khris S (2003). The validity of 2D
numerical simulations of vertical structures
around a bridge deck. Mathematics and
Computer Modeling 37: 795-828.
11. Cai CS, Chen SR (2004). Framework of
vehicle–bridge–wind dynamic analysis.
Journal of Wind Engineering and Industrial
Aerodynamics 92(7-8): 579-607.
12. Charuvisit S, Kimura K, Fujino Y (2004).
Experimental and semi-analytical studies on
the aerodynamic forces acting on a vehicle
passing through the wake of a bridge tower in
cross wind. Journal of Wind Engineering and
Industrial Aerodynamics 92(9): 749-780.
13. Chen SR, Wu J (2008). Performance
enhancement of bridge infrastructure systems:
Long-span bridge, moving trucks and wind
with tuned mass dampers. Engineering
Structures 30(11): 3316-3324.
14. Cochran L, Derickson R (2011). A physical
modeler’s view of computational wind
 Engineering Applications of Computational Fluid Mechanics Vol. 8, No. 3 (2014)
engineering. Journal of Wind Engineering
and Industrial Aerodynamics 99(4): 139-153.
15. Chen S, Chen F (2010). Simulation-Based
assessment of vehicle safety behavior under
hazardous driving conditions. Journal of
Transportation Engineering 136(4): 304-315.
16. Coleman SA, Baker CJ (1990). High sided
road vehicles in cross winds. Journal of Wind
Engineering and Industrial Aerodynamics
36(2): 1383-1392.
17. Corson D, Jaiman R, Shakib F (2009).
Industrial application of RANS modeling:
capabilities and needs. International journal
of Computational Fluid dynamics 23(4): 337-
347.
18. Fransos D, Bruno L (2010). Edge degree-of-
sharpness and free-stream turbulence scale
effects on the aerodynamics of a bridge deck.
Journal of Wind Engineering and Industrial
Aerodynamics 98(10-11): 661-671.
19. Guilmineau E, Chometon F (2009). Effect of
side wind on a simplified car model:
experimental and numerical analysis. Journal
of Fluids Engineering-transactions of the
ASME 131(2): 021104-1-021104-12.
20. Gunes D (2010). On the similarity of wind
tunnel experiments and numerical simulation
of heavy-duty trailer flow. Progress in
Computational Fluid Dynamics, An
International Journal 10(3): 168-176.
21. Hargreaves D, Morvan H (2008). Initial
validation of cross wind effects on a static
high-sided vehicle. International Journal of
CFD Case Studies 7: 17-31.
22. Huang L, Liao H (2011). Identification of
flutter derivatives of bridge deck under multi-
frequency vibration. Engineering
Applications of Computational Fluid
Mechanics 5(1): 16-25.
23. Li L (2009). Aerodynamic Coefficients of
Typical Vehicles and Disturbance of Bridge
Deck and Tower. Unpublished MS thesis,
Tongji University, Shanghai, China (in
Chinese).
24. Menter FR (2009). Review of the shear-stress
transport turbulence model experience from
an industrial perspective. International
Journal of Computational Fluid Dynamics
23(4): 305-316.
25. Mannini C, Šoda A., Voß, R., Schewe G..
(2010). Unsteady RANS simulations of flow
around a bridge section. Journal of Wind
Engineering and Industrial Aerodynamics
98(72): 742-753.
26. Murakami S (1997). Current status and future
trends in computational wind engineering.
344
Journal of Wind Engineering and Industrial
Aerodynamics 67 & 68: 3-34.
27. Sterling M, Quinn AD, Hargreaves DM,
Cheli F, Sabbioni E, Tomasini G, Delaunay
D, Baker CJ, Morvan H (2010). A
comparison of different methods to evaluate
the wind induced forces on a high sided lorry.
Journal of Wind Engineering and Industrial
Aerodynamics 98(1): 10-20.
28. Wu CL, Chau KW (2006). Mathematical
model of water quality rehabilitation with
rainwater utilization - a case study at
Haigang. International Journal of
Environment and Pollution 28(3-4): 534-545.
29. Xu YL, Guo WH (2003a). Dynamic behavior
of high-sided road vehicles subject to a
sudden crosswind gust. Wind & Structures
6(5): 325-346.
30. Xu YL, Guo WH (2003b). Dynamic analysis
of coupled road vehicle and cable-stayed
bridge systems under turbulent wind.
Engineering Structures 25(4): 473-486.
31. Yoo I, Kwak E, Lee S, Kim BS, Park SH
(2009). Computational study on
aerodynamics of long-span bridges. Journal
of Mechanical Science and Technology 23(3):
802-813.
32. Zhu Z, Gu M, Chen Z (2007). Wind tunnel
and CFD study on identification of flutter
derivatives of a long span self anchored
suspension bridge. Computer-Aided Civil and
Infrastructure Engineering 22(8): 541-554.
33. Zingg DW, Godin P (2009). A perspective on
turbulence models for aerodynamic flows.
International Journal of Computational Fluid
Dynamics 23 (4): 327-335.