Engineering Structures 106 (2016) 243260

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Dynamic analysis of a large span specially shaped hybrid girder bridge

with concrete-filled steel tube arches
Yan Li a,, C.S. Cai b, Yang Liu a, Yanjiang Chen c, Jiafeng Liu a
a

School of Transportation Science and Engineering, Harbin Institute of Technology, Harbin 150090, China
Department of Civil and Environmental Engineering, Louisiana State University, Baton Rouge, LA 70803, United States
c
Beijing Laboratory of Earthquake Engineering and Structural Retrofit, Beijing University of Technology, Beijing 100024, China
b

a r t i c l e

i n f o

Article history:
Received 8 October 2014
Revised 15 October 2015
Accepted 16 October 2015
Available online 11 November 2015
Keywords:
Concrete-filled steel tube arch bridges
Vehicle-bridge coupled vibration (VBCV)
Impact effect
Comfort evaluation
Dynamic test

a b s t r a c t
In the present study, the dynamic property of a specially shaped hybrid girder bridge with concrete-filled
steel tube (CSFT) arches is investigated based on experimental and numerical methods, especially under
moving vehicles. Before the inauguration of this bridge, a dynamic field test was conducted. A refined
three-dimensional finite element model is built to represent the complex structural mechanic property
of the bridge. The vehicle-bridge coupled vibration (VBCV) model with a 16 DOF vehicle model is established to simulate the dynamic behavior of the bridge with moving vehicles. The FE model is updated, and
the VBCV model for the bridge is verified, taking advantage of the aforementioned measured data. The
result indicates that the proposed VBCV numerical model can closely reproduce the measured response
and can be used to simulate the dynamic behavior of the bridge under various conditions. The impact
effect, ride and pedestrian comfort, and related parameters analysis for the bridge with moving vehicles
are studied by numerical simulations and experimental tests. The results indicate that the impact factor
formula from design standards significantly underestimates the dynamic impact effect, which may result
in an unfavorable influence on the bridge safety. Several conclusions are drawn for this bridge, and further research that is needed for this new bridge type is discussed.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
In recent years, a new type of concrete-filled steel tube (CFST)
arch bridges with specially shaped multi-arch ribs have been successfully constructed in certain cities in China, such as the Longjiang bridge in Zhangzhou, the Donggang bridge in Changzhou,
the Jiubao bridge in Hangzhou, the Yingzhou bridge in Luoyang,
and the Changfeng bridge in Ningbo [10]. This bridge type, one
with a strong spanning ability and a unique configuration style,
is a good candidate for constructing medium and large span urban
bridges. The structural features of this bridge type differing from
typical CFST arch bridges can be described as following: its arch
ribs are composed of main and auxiliary ribs, and all ribs are concentrated fixed in the same skewback; the auxiliary arch ribs are
outward-inclined in transversal space, and the transverse braces
and inclined braces are used to link different arch ribs; the girder
and arch ribs are connected with vertical or inclined suspensors
Corresponding author.
E-mail addresses: liyan2011@hit.edu.cn (Y. Li), cscai@lsu.edu (C.S. Cai),
ly7628@hit.edu.cn (Y. Liu), cyjrlx@sina.com (Y. Chen), liujiafeng0909@126.com
(J. Liu).
http://dx.doi.org/10.1016/j.engstruct.2015.10.026
0141-0296/ 2015 Elsevier Ltd. All rights reserved.

(a typical bridge example is shown in Fig. 1). The special and complicated configuration of this new bridge type leads to its particular
mechanical features, such as a higher center of gravity, more
degrees of indeterminacy than regular arches, and a different integrated anti-torsion system composed of an arch rib, girder and suspensors from traditional CFST arch bridges. All of the
aforementioned complex and distinct structural characteristics of
the bridge type may also induce a large difference in dynamic performance from regular CFST arch bridges, especially with moving
vehicles. The results of several initial simplified numerical analyses
also verify the above observations [47,28].
For regular CFST arch bridges, related studies on vehicle and
bridge coupled vibration (VBCV) have achieved a number of meaningful results in the past decade. A simplified impact factor formula
is suggested based on numerical analysis for half-through CFST
arch bridges with a span less than 200 m [19]. Taking a CFST arch
bridge with a pedestrian deck suspended under the girder as an
example, the natural vibration properties and dynamic response
analysis under moving vehicles is carried out, and the effect of
moving vehicles on the pedestrian and bridge is researched [45].
Another impact factor formula is proposed by statistically analyzing the data from field tests for regular CFST arch bridges [33].

244

Y. Li et al. / Engineering Structures 106 (2016) 243260

(a) Elevation view of the bridge (units: cm)

(b) Cross section of the bridge at mid-span and vehicle loading position (units: cm)

(c) Overview of the bridge

Fig. 1. Layout of Yitong River Bridge.

However, the suitability of general specifications and conclusions

from regular CFST bridges research for the dynamic response and
characteristic of this new bridge type is not clear. All of these questions and similar ones warrant further investigation. Until now,
studies on the above problem were rare, and related experimental
research cannot even be found in the literature. Therefore, further
research by combining theoretical analysis and experimental

testing on the dynamic performance for this new bridge type is

necessary.
In the present study, a large span irregular CFST arch bridge
with a hybrid girder in China, named Yitong River Bridge, is taken
as a case study, and the dynamic performance of the bridge,
especially under moving vehicle loading, is researched based on
experimental and numerical methods. The paper is organized by

Y. Li et al. / Engineering Structures 106 (2016) 243260

245

Fig. 2. The truck used in field tests.

the following main sections. Firstly, a brief introduction is given, and

a refined three dimensional finite element model is built for the
bridge. The VBCV numerical analysis method is detailed. Secondly,
a dynamic field test on the bridge is performed, and several main
experimental processes and important dynamic results are shown
in this part. Thirdly, based on experimental data, the FE model of
the bridge is updated by the zero-order approximation method,
and the VBCV model is verified by comparing numerical results
and measured responses. Finally, the dynamic performance of the
bridge, especially under moving vehicles, is studied. The dynamic
impact effect, such as riding and pedestrian comfort induced by
moving vehicles, is also emphasized and discussed. Conclusions are
drawn from the case study results at the end of the paper.

2. Description of the bridge

The Yitong River Bridge is located in Changchun, the capital city
of Jilin province in China. The main bridge is composed of three
spans of 51 m, 158 m and 51 m, with a total length of 260 m,
as shown in Fig. 1. The bridge was completed in the fall of 2009,

making it the largest span irregular CFST arch bridge in China. Its
bridge deck with a total width of 40 m includes eight-lane roadways and two 2.5 m-width walkways. The designed driving speed
is 60 km/h. The main components and structural details of the
bridge are introduced as following: the hybrid girder is composed
of a prestressed concrete (PC) beam in the side span and a steel box
girder in the main span with the same height; the PC girder is fixed
with V shape piers and the main arch foot; the steel box girder
with a length of 110.8 m in the main span is simply supported
on the cantilever of the PC girder (shown as Fig. 1(a)); the main
arch rib with its rise-span ratio of 1/4.23 has a triangular crosssection consisting of three steel tubes filled with C50 concrete with
an external diameter of 1.8 m and 1.2 m, respectively (shown as
Fig. 1(b)); the two assistant arch ribs, which have a 1.2 m external
diameter and are 21.8 degrees inclined from the vertical axis, are
connected to each other by transverse braces and linked with the
main arch rib by diagonal braces; the two groups of inclined suspensors, totaling 32, are used to connect the main arch rib with
the girder (shown as Fig. 1(b)); and the 6 horizontal tie rods are
arranged in the center insulation strip on the bridge deck to
balance the horizontal reaction.

Fig. 3. Vehicle model and vehicle-bridge interaction schematic.

246

Y. Li et al. / Engineering Structures 106 (2016) 243260

3. Vehicle-bridge coupled vibration system

Currently, numerical analytical method has been widely applied
to relevant VBCV studies because it has fewer restrictions and its
inexpensive cost compared with field experimental research
[31,37,48,12]. The literature indicated that the results from numerical analysis can agree well with those measured from field tests
[18,34,24,1,2]. Generally VBCV problems can be solved by iterative
or coupled methods. The iterative method usually has a high
computation cost to achieve numerical convergence [6,4]; for the
coupled method, vehicle and bridge coupled motion equations are
assembled and solved by direct integration methods such as the
Newmark and RungeKutta method [15,23,11,26]. Furthermore,
modal superposition is a general technique used in solving VBCV
problems because it significantly reduces the calculation effort by
reducing the size of the matrices in the equation [38,40,42].
In the present study, the coupled method is used to solve the
VBCV problems. Firstly, the two sets of equations of motion for
the bridge and the vehicle are characterized respectively. Then
the two subsystems are coupled through the contact condition
and assembled into a system motion equation. Furthermore, the
modal superposition technique is used to describe the bridges
dynamic behavior based on the obtained bridge mode shapes
and natural frequencies. Finally the Newmark-b method is used
to solve the VBCV system.

bridge is performed by a standard three-axle truck (as shown in

Fig. 2) with a total weight of 34,300 kg, where P1 is 6300 kg, P2
and P3 are 14,000 kg, respectively, where D1 is 4.15 m, D2 is
1.90 m, and D3 is 1.40 m. This truck type is usually taken as a standard vehicle used in loading tests on highway bridges in China.
Based on the actual condition of trucks used in field testing, a
3D three-axle vehicle model is built, which is combined by several
rigid bodies connected by several axle mass blocks, springs, and
damping devices and consists of 16 independent degrees of freedom (DOF), namely six vertical and six lateral DOFs for the six
wheels, and vertical, pitch, lateral and rolling DOFs for the vehicle
body (as shown in Fig. 3).
The detailed mechanical and geometric parameters for the vehicle model are listed in Table 1. The dimensions, axle loads, and
total weight of the vehicle were actually measured and can then
be treated as reliable information. The values of suspension stiffness and damping might not be the same as the values for the
actual trucks used in the test, which were obtained from references
and may lead to some errors. However, the inaccuracy can be generally accepted based on previous studies since reliable information for these parameters are lacking [3,30,16,39,43].
The equation of motion for the vehicle is built based on the
equilibrium state under self-weight and expressed as follows:

n o
n o
v C v  X_ v K v fX v g fF v g
M v  X

n
where fF v g col 0; 0; 0; 0; F 1wv L ; F 1wv R ; F 2wv L ; F 2wv R ; F 3wv L ; F 3wv R ; F 1wyL ; F 1wyR ;

3.1. Vehicle model

The vehicle models adopted in previous studies can be classified
into three types in terms of complexity: one-dimensional (1D)
models [5,41], in which the vehicle is modeled as a sprung mass
of one or two degrees of freedom (DOFs); two-dimensional (2D)
models [6,9], in which a planar model with multiple axles is considered; and 3D complete vehicle models, which was validated
by experiments to well predict the truck dynamic response
[17,32]. In the present study, the dynamic loading test for the

F 2wyL ; F 2wyR ; F 3wyL ; F 3wyR g is the vector of the wheel-road contact forces
acting on the vehicle, and the detailed expression is listed in
h
Eq. (13). Mv  diag M v r ; Iv r ; Jv r ; Mv r ; M 1sL ; M 1sR ; M2sL ; M2sR ; M3sL ; M3sR ; M1sL ;
n
M1sR ; M 2sL ; M 2sR ; M3sL ; M3sR  and fX v g Z v r ; hv r ; uv r ; Y v r ; Z 1sL ; Z 1sR ; Z 2sL ; Z 2sR ;
Z 3sL ; Z 3sR ; Y 1sL ; Y 1sR ; Y 2sL ; Y 2sR ; Y 3sL ; Y 3sR gT are the mass and displacement vectors of the vehicle, respectively; The stiffness and damping matrices
K v  and C v  can be expressed as following:

Table 1
Major parameters of the vehicle model.
Parameters

Value

Unit

Mass of rigid body (Mv r

31,280
710

kg
kg

Axle mass (M 1sR ; M 1sL

Axle mass (M 2sR ; M 2sL ; M3sR ; M 3sL
Pitching moment of inertia (Iv r
Rolling moment of inertia (J v r

Upper vertical spring stiffness (kv uL ; kv uR

2

Upper vertical spring stiffness (kv uL ; kv uR ; kv uL ; kv uR

Lower vertical spring stiffness (kv lL ; kv lR

Lower vertical spring stiffness (kv lL ; kv lR ; kv lL ; kv lR

1

Upper lateral spring stiffness (kyuL ; kyuR

2

400

kg

172,160
61,496
242.604

kg m2
kg m2
kN/m

951.586

kN/m

266.67

kN/m

533.34

kN/m

102.302

kN/m

501.5155

kN/m

Lower lateral spring stiffness (kylL ; kylR ; kylL ; kylR ; kylL ; kylR

120

kN/m

Upper vertical damping coefficient (c1v uL ; c1v uR

2.19

kN s/m

3.941

kN s/m

1.9

kN s/m

Upper lateral spring stiffness (kyuL ; kyuR ; kyuL ; kyuR

1

Upper vertical damping coefficient (c2v uL ; c2v uR ; c3v uL ; c3v uR

Lower vertical damping coefficient (c1v lL ; c1v lR

Lower vertical damping coefficient (c2v uL ; c2v uR ; c3v uL ; c3v uR

Upper lateral damping coefficient (c1yuL ; c1yuR

3.8

kN s/m

1.69

kN s/m

Upper lateral damping coefficient (c2yuL ; c2yuR ; c3yuL ; c3yuR

2.935

kN s/m

Lower lateral damping coefficient (c1ylL ; c1ylR ; c2ylL ; c2ylR ; c3ylL ; c3ylR

kN s/m

Distance between the front axial and the center of the truck (L1
Distance between the second axial and the center of the truck (L2
Distance between the third axial and the center of the truck (L3
Distance between the center and driving seat of the truck (L4
Elevation difference between the center and driving seat of the truck (h1
Distance between the right and left wheels (b1

1.12
3.03
4.43
1.42
1
0.8

m
m
m
m
m
m

247

Y. Li et al. / Engineering Structures 106 (2016) 243260

2
6
6
K v  6
6
4

k11

k12   k1;16

k21
..
.

k22   k2;16
.. . .
..
.
.
.

3
7
7
7;
7
5

k16;1 k16;2   k16;16

ci11 ci12
6 i
6 c21 ci22
6
C v i  6 .
6 .
...
4 .
ci16;1 ci16;2

  ci1;16

Based on the above finite element model for the bridge, the
motion of the bridge structure around its statically deformed position can be described as:

7
  ci2;16 7
7
7
..
... 7
.
5
  ci16;16

n o
n o
b C b  X_ b K b X b fF b g
M b  X
2

The matrix elements of K v  and C v  are obtained based on

Lagrange equation method [20] and listed as follows:

k11

3 
3





X
X
i
i
i
i
1
1
kv uL kv uR ; k12
Li kv uL kv uR  L1 kv uL kv uR ;
i1

i2

3 

D2 X
i
i
k13
k  kv uL ;
2 i1 v uR
1

k15 kv uL ; k16 kv uR ; k17 kv uL ;

k18 kv uR ; k19 kv uL ;
3


X
3
i
i
k1;10 kv uR ;k22
L2i kv uL kv uR ;
i1
2

 D 

D2 X
i
i
2
3
3
1
Li kv uL  kv uR  L3 kv uL  kv uR ; k25 L1 kv uL ;
k23
2 i1
2
1

k26 L1 kv uR ; k27 L2 kv uL ; k28 L2 kv uR ;

3 

D2 X
3
i
i
k2;10 L3 kv uR ; k33
kv uL kv uR ;
4 i1
1

k38 bkv uR ; k39 bkv uL ; k3;10 bkv uR ;

3 

X
i
i
1
1
k44
kyuL kyuR ; k4;11 kyuL ; k4;12 kyuR ;
i1
2

k4;15 kyuL ; k4;16 kyuR ; k55 kv uL kv lL ;

1

k66 kv uR kv lR ; k77 kv uL kv lL ; k88 kv uR kv lR ;

3

k99 kv uL kv lL ; k10;10 kv uR kv lR ;
1

1
1
k11;11 kyuL kylL ;

k12;12 kyuR kylR ; k13;13 kyuL kylL ;

2
2
k14;14 kyuR kylR ;

3
3
k15;15 kyuR kylL ;

   qN b

T

UB fqB g

where Nb is the total number of modes for the bridge under consideration; {Ui} and qi are the ith mode shape and its generalized coordinates. Each mode shape is normalized by the mass, i.e. {Ui}T[Mb]
{Ui} = 1. Substituting Eqs. (4) to (3), the mode motion equation for
the bridge is obtained and written as:

B g C B fq_ B g K B fqB g fF B g
MB fq

k4;13 kyuL ; k4;14 kyuR ;

1

q2

where Mb ; C b  and K b  are the mass, damping, and stiffness matrices of the bridge, respectively; {X b } is the node displacement vector
for all DOFs of the bridge, and fF b g is the force vector acting on the
bridge through the vehicles wheels.
The direct integration of the motion equations including so
many DOFs for the bridge in the time domain is very cumbersome.
To simplify the modeling procedure in the VBCV system, the mode
superposition technique is used to model the bridge system.
Firstly, the bridge mode shape and the corresponding natural circular frequencies are obtained from bridge modal analysis by
ANSYS. Upon the introduction of the modes orthogonality, hundreds of FEM equations coupled with each other can then be
decoupled, which makes the bridge calculation model become
the superposition of independent modal equations. Thus the computational effort can be significantly reduced. The accuracy and
availability of the mode superposition method has been verified
and used comprehensively in previous VBCV studies [38,46,7].
Based on the mode superposition method, the bridge dynamic
response {X b } can be expressed as:



 
q1
UN b
fX b g fU1 g fU2 g   

k35 bkv uL ; k36 bkv uR ; k37 bkv uL ;

2

3
3
k16;16 kyuR kylR :

The matrix elements not given here have a value of zero. The
[C v ] is the same as [K v ] in form and can be obtained by replacing
the stiffness, k, with the corresponding damping coefficient, c.

3.2. Bridge model

A three dimensional refined finite element model is set up with
ANSYS to represent the complex dynamic property of the bridge.
Firstly, the arch ribs, inclined braces and lateral braces are simulated with space beam elements (Beam44). Secondly, the concrete
components including the V shape rigid frame, prestressed concrete girder, skewback and column are modeled by solid elements
(Solid65). Thirdly, the suspenders and tied bars are simulated with
cable elements (Link10). Lastly, the flat steel box girder, bridge
deck, and pavement are simulated with shell elements (Shell63).
In addition, for the boundary conditions of the model, fixed
constraints are given to the bottom of the skewback, and other
displacement and rotation constraints are applied based on the different bearing types. The FE model is used in the following VBCB
simulation analysis (shown in Fig. 4). A total of 421,358 elements
are included in the model.

where the modal mass, damping, stiffness matrix and acting force
vectors for the bridge can be expressed as:

MB  fUB gT M b fUB g I

C B  fUB gT C b fUB g 2xi ni I



K B  fUB gT K b fUB g x2i I

fF B g UB T fF b g

where xi denotes the natural circular frequency of the bridge and fi

is the percentage of the critical damping for the bridges ith mode.
Submitting Eqs. (6)(9) to (5), Eq. (5) can be rewritten as



B g 2xi ni Ifq_ B g x2i I fqB g fF B g
Ifq

10

3.3. Road surface condition

The road surface condition (RSC) is a very important factor that
affects the dynamic responses of both the bridge and vehicles. The
RSC is usually assumed to be a zero-mean stationary Gaussian
random process and can be generated by an inverse Fourier transformation based on a power spectral density (PSD) function as:

rx

N p
X
2unk Dn cos2pnk x hk

11

k1

where hk is the random phase angle uniformly distributed from 0 to

2p, u() is the PSD function (m3/cycle) for the road surface elevation,
and nk is the wave number (cycle/m). The PSD functions for road

248

Y. Li et al. / Engineering Structures 106 (2016) 243260

B: Steel girdersection

A: PC girder section

D: Main arch rib section

C: V frame

Fig. 4. Finite element model of the bridge.

surface roughness were developed by Dodds and Robson [13], and

the PSD function was simplified by Huang and Wang [17] as:

2
n
un un0
n0

n1 < n < n2

12

where n is the spatial frequency (cycle/m); n0 is the discontinuity

frequency of 1/2p(cycle/m); un0 is the roughness coefficient
(m3/cycle) whose value is chosen depending on the road condition;
and n1 and n2 are the lower and upper cut-off frequencies, respectively.
The International Organization for Standardization [21] has proposed a road roughness classification index from A (very good) to H
(very poor) according to different values of u(n0 . In this paper the
classification of road roughness from the ISO is used. Considering
expected management and maintenance conditions for this
important urban bridge, the RSC will be kept at a level better than
average during its whole service time. Therefore, the average
level is considered as the possible poorest road roughness condition in the subsequent study.
3.4. Assembling the vehicle and bridge coupled system
The motion equations for the vehicle and bridge can be coupled
through the displacement and interaction force relationship at the
contact points. The left wheel of the ith axle for the vehicle is taken
as an example to explain the interaction between the vehicle and
bridge (shown as Fig. 3). The action force on the left wheel due
to the relative displacement between the bridge deck and wheel
can be expressed as:

F iwyL
F iwzL

"

#(

kylL
i

kv lL

DiylL
Div lL

"

#(

ciylL
civ lL

D_ iylL
D_ i

v lL

13

where F iwzL and F iwyL are the vertical and lateral forces on the left
wheel of the ith axle from the bridge deck, respectively. Div lL and

DiylL are the vertical and lateral deformations of the lower springs
of the left wheel. For the contact point, the equations of compatible
displacement can be written in a matrix form as follows:
( i ) (
) ( ) 

DylL
0
Y isL
Y ib



14
r i x
Z isL
Z ib
Div lL

D_ iylL
D_ i

v lL

Y_ isL
Z_ i
sL

(


Y_ ib
Z_ i

0
r_ i x


15

where r_ i x drdxx dx
drdxx Vt r0 V; Y isL and Z isL are the vehicle axle
dt
suspension displacements in the lateral and vertical direction
(shown as Fig. 3); rx is the road surface profile at the wheel and
deck contact point, and Vt is the vehicle velocity. Y ib and Z ib are
the lateral and vertical displacement of bridge at wheel-deck contact points (shown as Fig. 3(b)), which can be written in generalized
mode coordinates as:

Y ib
Z ib

(
)
Nb
X
qn /nh xi
n1

qn /nv xi

16

where Nb is the mode number of the bridge involved in analysis.

/nh xi and /nv xi are the horizontal and vertical components of
the nth mode shape, respectively. xi is the wheel position of the
vehicle on the bridge and qn is the generalized mode coordinates.
The acting force on the bridge deck from the left wheel of the ith
axle of the vehicle F ibL is written as:

249

Y. Li et al. / Engineering Structures 106 (2016) 243260

Fig. 5. Measurement points arrangement for dynamic test.

n o
F ibL
where

F Li
G

F ibyL
F ibzL

0
F iGL

F iwyL

F iwzL

17



 Miv rL MisL g is the gravity force on the bridge at the

location of the left wheel of the ith axle for the vehicle due to the
weight of the vehicle. F ibyL and F ibzL are the horizontal and vertical
forces acting on the bridge, respectively.
The nth mode generalized force F Bn acting on the bridge is
defined as follows:

F Bn

3 

X
/nh xiL F ibyL /nv xiL F ibzL /nh xiR F ibyR /nv xiR F ibzR
i1

18
where xiL and xiR are the locations of the left and right wheels of the
ith axle on the bridge, respectively. For the case of multiple vehicles
on the bridge, the total contact forces on the bridge can be accumulated by the above equation for all vehicles.
Substituting Eqs. (13)(18) into Eq. (5), the nth bridge mode
motion equation can be written as:

n C Bn q_ n K Bn qn F Bn
M Bn q

19

where the MBn ; C Bn ; K Bn , and F Bn are the nth modal mass, damping,
stiffness and generalized force for the bridge, respectively.
Assembling Eqs. (19) and (1), the motion equation of the VBCV
system can be obtained and written as:

( )
( _ )
Cv
CvB
Xv
Xv

B
C Bv C B C vB
0 MB
q
q_ B

  ( r )
Fv
Kv
K vB
Xv

K Bv K B K vB
qB
F rG
B

Mv

20

where subscripts B and v represent the bridge and vehicle,

respectively; superscript v in the stiffness (K and damping (C
terms refers to the contributions due to vehicles; subscripts Bv
and vB refer to the vehicle and bridge coupled terms; superscripts
r and G represent the acting force due to road roughness and
gravity force of the vehicles, respectively. In comparison with Eqs.
(1) and (5), there are additional terms C Bv ; C v B ; C vB ; K Bv ; K v B and K vB
in Eq. (20). The detailed formulations of the terms are listed as
follows:

2


v

KB KB

6
6
6
6
6
4

x21 K Bv 11
K Bv 21
..
.

v

CB CB

6
6
6
6
6
4

v Nb 1

2n1 x1 C Bv 11
C Bv 21
..
.

v Nb 1

CB



..
.

..

v Nb 2

v 1Nb

v 2Nb

7
7
7
7
7
5

KB

x22 K Bv 22   

KB
2

K Bv 12

KB
..
.

v Nb Nb

   x2Nb K B

KB



CB

2n2 x2 C 22
B
..
.


..
.

CB

CB

v 1Nb

C Bv 12

v Nb 2

21

v 2Nb

..
.

v Nb N b

7
7
7
7
7
5

   2nNb xNb C B

22

250

Y. Li et al. / Engineering Structures 106 (2016) 243260

where

v nm

KB

8 i
9
i
i
i
n
m
n
m
n
m
n
m
3 < k / xLi / xLi k / xLi / xLi k / xRi / xRi k
X
h
h
ylL h
ylR h
v lL v
v
v lR /v xRi /v xRi =
h
i

;
n
n
: ci /n xLi @/h xLi ci /n xRi @/h xRi ci /n xLi @/mv xLi ci /n xRi @/mv xRi V ;
i1
ylL h
ylR h
v lL v
v lR v
@x
@x
@x
@x

C Bv nm

3 h
X

n
m
n
m
n
m
i
i
i
civ lL /nv xLi /m
v xLi cylL /h xLi /h xLi cv lR /v xRi /v xRi cylR /h xRi /h xRi

1 6 n; m 6 Nb :

i1

h
K v B  K Bv T K 1v B

K 2v B

   K v Bb

h
C v B  C Bv T C 1v B

C 2v B

   C v Bb

where

23
24

3
0
7
6
0
7
6
7
6
0
7
6
7
6
7
6
0
7
6
6 k1 /m xL1  c1 @/mv xL1 V 7
7
6
v lL v
v lL @x
7
6 1 m
m
6 kv lR /v xR1  c1 @/v xR1 V 7
v lR @x
7
6
m
7
6
2 @/v xL2
7
6 k2v lL /m
x


c
V
v L2
v lL @x
7
6
m
7
6 2 m
1 @/v xR2
6 kv lR /v xR2  cv lR @x V 7
7
6
6 k3 /m x  c3 @/mv xL3 V 7
7;
6
L3
v
lL v
v
lL
Km

@x
vB
7
6 3
6 k /m x  c1 @/mv xR3 V 7
7
6 v lR v R3
v lR @x
7
6
6 k1 /m x  c1 @/mh xL1 V 7
L1
7
6
ylL h
ylL
@x
7
6
6 k1 /m x  c1 @/mh xR1 V 7
7
6
R1
ylR h
ylR
@x
7
6
@/m xL2
2
7
6
m
6 kylL /h xL2  c1ylL h@x V 7
7
6
m
7
6
2
2 @/h xR2
6 kylR /m
V7
h xR2  c ylR
@x
7
6
m
7
6
3 @/h xL3
7
6 k3ylL /m
x


c
V
L3
h
ylL
@x
5
4
m
3
m
3 @/h xR3
kylR /h xR3  cylR @x V
3
2
0
7
6
0
7
6
7
6
0
7
6
7
6
7
6
0
7
6
7
6
c1v lL /m
v xL1
7
6
7
6
m
1
cv lR /v xR1
7
6
7
6
m
2
7
6
cv lL /v xL2
7
6
m
2
7
6
c
/
x

v lR v R2
7; m 1; N b
6
Cm
vB 6
7
m
3
/
x

c
7
6
L3
v
lL v
7
6 3 m
m
6 cv lR /v xR3  c1ylL /h xL1 7
7
6
7
6
c1ylR /m
7
6
h xR1
7
6
m
2
7
6
c
/
x

L2
h
ylL
7
6
7
6
m
2
7
6
cylR /h xR2
7
6
m
7
6
3
c
/
x

L3
5
4
ylL h
m
3
cylR /h xR3
2

(a) Testing vehicles

(b) Dynamic response test under one moving truck load

8
9
1
1
1
0
>
>
>
>
> 0; 0; 0; 0; kv lL rxL1 cv lL r xL1 V; kv lR rxR1
>
>
>
>
>
2
<
=
1
0
2
0
r
x
V;
k
rx

c
r
x
V;
c
R1
L2
L2
v lL
r
v lR
v lL
F v col
2
3
2
0
>
>
>
>
kv lR rxR2 cv lR r xR2 V; kv lL rxL3
>
>
>
>
>
>
: 3 0
;
3
3
0
cv lL r xL3 V; kv lR rxR3 cv lR r xR3 ; 0; 0; 0; 0; 0; 0
25

(c) Dynamic response data recording

Fig. 6. The pictures for the dynamic loading test process.

Y. Li et al. / Engineering Structures 106 (2016) 243260

251

Fig. 7. Typical dynamic response obtained from field testing.

The nth modal force acting on the bridge from the vehicle can
be written as:

F rG
bn

3 nh
X

i
i
kv lL rxLi civ lL r0 xLi V F LiG /nv xLi

i1

o
h
i
i
n
kv lR rxRi civ lR r 0 xRi V F Ri
G /v xRi

26

The VBCV system in Eq. (20) contains only the modal properties
of the bridge and the physical parameters of the vehicles. As a
result, the complexity of solving the vehicle-bridge coupled equations is greatly reduced. Based on the methodology discussed
above, the authors developed a Matlab program to automatically
assemble the equation of motion of the VBCV system and solve it
in the time domain by the Newmark-b method.

252

Y. Li et al. / Engineering Structures 106 (2016) 243260

Table 2
Comparison of dynamic properties from testing and fe analysis.
No.

1
2
3
4
5
6
7
8
9
10
11

Tested
frequency
f t (Hz)

Calculated
frequency f c

100(f c  f t =f t (%)

MAC

Before
updating

After
updating

Before
updating

Before
updating

After
updating

0.606
1.617
1.879

2.787
2.904
3.432

3.513
4.946
5.315

0.525
1.619
1.717
2.583
2.894
3.125
3.351
3.493
3.832
4.711
4.842

0.582
1.635
1.796
2.572
2.648
2.791
3.37
3.521
3.62
4.88
5.096

0.914
0.907
0.922

0.901
0.931
0.959

0.935
0.977
0.924

0.911
0.912
0.931

0.912
0.925
0.937

0.942
0.948
0.917

After
updating

13.37
0.12
8.62

3.84
7.61
2.36

9.08
4.75
8.90

3.96
1.11
4.42

4.99
3.89
1.81

3.05
1.33
4.12

Vertical acceleration
limit (m/s2)

Lateral acceleration
limit (m/s2)

CL1
CL2
CL3
CL4

<0.50
0.501.00
1.002.50
>2.50

<0.10
0.100.30
0.300.80
>0.80

Best
Average
Poor
Not acceptable

Description of the vibration mode shape

1.48
0.85
0.59

1.12
1.77
4.44

2.86
2.12
2.96

Arch rib: symmetric lateral bending

Arch rib: asymmetric lateral bending
Bridge deck: asymmetric vertical bending
Whole bridge: asymmetric vertical bending
Bridge deck: symmetric vertical bending
Whole bridge: vertical bending, two sine waves
Bridge deck: torsion,one and half sine waves
Bridge deck: torsion, one and half sine waves
Bridge deck: vertical bending and torsion, two sine waves
Whole bridge: vertical bending, one sine wave
Bridge deck: torsion, one sine wave

DH3817 data recording system (produced by DH Company in

China). The displacement response of the bridge under loading
by moving trucks is measured by IBIS-S, which is a new type of
deformation monitoring and measurement instrumentation with
0.01 mm measurement accuracy and maximum sample rate
200 Hz, produced by IDS Company in Italy.

Table 3
Personnel vibration comfortability criterion in EN03.
Comfortable level

Damping
ratio (%)

4.2. Dynamic testing under controlled moving vehicle and ambient

excitation
Dynamic load testing is performed with two three-axle trucks
as shown in Fig. 6. The moving vehicle test includes two loading
cases (shown as Fig. 1(b)), one truck pass on the right lane 3 and
two truck passes on the right lane 3 and lane 4 at different speeds
(30 km/h, 40 km/h and 60 km/h).
Most of the passes are repeated to check the validity of the measurements. Additionally, a 60-mm-high wedge-shaped wooden
plank is placed across the roadway at the middle and quarter span
for the purpose of exciting dominant flexural modes. Consequently,
a total of 8 different test cases are performed. In addition to the
moving vehicle testing, the acceleration response of the bridge
under ambient excitations is also recorded during two different
time periods lasting 15 min each. Several typical collected data
points are plotted in Fig. 7. The identification and discussion for
the dynamic parameters of the bridge are presented in the following section.

4. Experimental identification of dynamic properties for the

bridge
Before the inauguration of the Yitong River Bridge in November
of 2009, a dynamic test composing of moving vehicle loading and
ambient excitation testing was performed to check and verify the
dynamic performance of the bridge. Because the vehicle speed,
weight and position are known, dynamic testing is especially suited for the experimental identification of the dynamic properties
of bridges. The process and result of the dynamic test are introduced in detail in this part. The related results demonstrate the
dynamic characteristic of this bridge and also provide a basis to
verify the numerical analysis model.

4.1. Measurement system for the dynamic test

4.3. Identification of dynamic parameters of the bridge
During the dynamic testing, the displacement, strain and acceleration time histories are measured at selected points and sections
(as shown in Fig. 5) where structural response is typical and significant based on theoretical analysis results.
A data acquisition system with 32 channels (IMC CL1032,
German) is used to record the time history response for all acceleration sensors. The dynamic strain response is collected by the

Based on the measured deflection and acceleration of the

bridge, several dynamic characteristics such as structural modal
parameters, dynamic impact factor, and maximum values of acceleration can be determined.
The natural vibration frequencies, mode shapes, and damping
ratios are estimated by analyzing the acceleration time history

Table 4
Structural material properties before and after model updating.
Components

Main arc rib

Assistant arc rib
Lateral brace
Inclined brace
Suspender
Steel girder
PC girder
PC V shape frame

Modulus of elasticity(E/MPa)
Before updating

After updating

2.06  105
2.06  105
2.06  105
2.06  105
1.95  105
2.06  105
3.45  104
3.45  104

2.25  105
2.25  105
2.07  105
2.15  105
2.12  105
2.07  105
3.65  104
3.49  105

Change (%)

9.22
9.22
0.49
4.37
8.72
0.49
5.80
1.16

Density P(kg/m3)

Change (%)

Before updating

After updating

12,400
7850
7850
7850
8010
7850
2550
2550

12,700
8230
7140
8150
8550
7210
2740
2630

2.42
4.84
9.04
3.82
6.74
8.15
7.45
3.14

Y. Li et al. / Engineering Structures 106 (2016) 243260

data of the bridge under ambient excitation and moving trucks.

The Frequency Domain Decomposition (FDD) method [29] is used
to identify the modal parameters for the bridge. The identified and
numerical results are listed in Table 2, where f t ; f c and 1 represent
the measured frequency, calculated frequency, and damping ratio,
respectively. The order of the total identified eleven frequencies is
decided based on the match degree with the calculated mode
shapes, i.e., the MAC. The first two mode shapes are lateral bending
modes of arch ribs, which indicates that the out-of-plane stiffness
of the arch ribs is smaller than that of the in-plane one. The first
vertical bending mode shape for the bridge deck appears in the
third mode with the 1.879 Hz measured frequency. The torsional
mode shape for the bridge deck in the main span appears and
repeats several times from the 7th mode. That result indicates that
the torsional mode shapes play an important role in the total
dynamic characteristics of the bridge, and the dynamic data from
the experiment and simulation will further verify this statement
later. The MAC value is employed to evaluate the mode shape correlation between the tested and calculated ones. The result with all
values over 0.90 indicates the computed mode shape closely
matches the tested ones. Several possible modes are not identified
for the reason of modes overlapping or mixing based on the tested
data, such as the 4th and 8th modes. The measured damping ratios
range from 0.59% to 4.44% (as shown in Table 2) and are basically
in the range of empirical values for concrete and steelconcrete
composite girder bridges.
Another cause for concern for the bridge is the impact effect due
to moving vehicles. The impact factor is commonly used to express
the dynamic amplification effect for bridges induced by moving
vehicles. In this study, the impact factor is defined as:

Rd x  Rs x
Rs x

27

where Rs x and Rd x denote the maximum static and dynamic

response at location x for the bridges under moving trucks,

253

respectively. Generally, the structural deflection is chosen as a representative response to determine the impact factor of the bridges.
Rs x is also usually assumed as a quasi-static response. In the
design specifications for highway bridges in China [8], the impact
factor is defined as a function of the basic frequency of the bridges
and expressed with Eq. (28), which is used for all bridge types.

8
f < 1:5 Hz
>
< 0:05
l 0:1767 ln f  0:0157 1:5 Hz 6 f 6 14 Hz
>
:
0:45
f > 14 Hz

28

where f is the first in-plane modal frequency of the bridge.

For the Yitong River Bridge, the maximum vertical deflection
appears at the measured point A (shown as Fig. 5) in the midspan of the bridge deck under two asymmetric loading cases
(shown as Fig. 1(b)) based on the computed results from the
designer, and therefore the impact factor of the measured point
A in Deck3 (shown as Fig. 5) is taken as a representative of the
whole bridge. The tested maximum impact factors for the bridge
are 0.073 and 0.069 caused by one truck on lane 3 and two trucks
on lanes 3 and 4 passing the bridge, respectively, corresponding to
the vehicle speed of 60 km/h. That is basically consistent with the
designed impact factor 0.082 calculated by Eq. (28).
Additionally, the bridge does not only supply 8 roadways but
also provides two walkways for pedestrians. Therefore, the compliance with maximum acceleration limits is especially important to
ensure the serviceability of the bridge. The maximum values of
acceleration are gained in the measurement point V1 in the
mid-span. For the one and two truck passing cases, the peak vertical
accelerations are 1.25 m/s2 and 0.59 m/s2, and the lateral ones are
0.17 m/s2 and 0.11 m/s2, respectively, corresponding to the vehicle
speed of 60 km/h. In the present study, the criterion from the
Design of Footbridges Guide in Germany [14] is used to evaluate
the vibration comfort for pedestrians (shown in Table 3). The
comfort level reaches poor and average for the case of two trucks
passing side by side under the vertical and lateral acceleration limit,

Fig. 8. Comparison of measured and numerical results in typical measurement points.

254

Y. Li et al. / Engineering Structures 106 (2016) 243260

and for the one truck passing case the result is average. Actually,
the possibility of two heavy trucks passing the bridge side by side is
very low, so the vibration comfort for pedestrians with an average
level can be considered acceptable.
5. Model updating and numerical analysis program verification
Due to the limitation of experimental conditions and cost, it is
very difficult and even unpractical to perform a complete test for
all responses of key structural components considering various
possible loading cases for a real bridge. Fortunately, a verified
numerical approach can be employed to complete the task and
predict the structural dynamic behavior. In the following part,
model updating and verification of the vehicle-bridge coupled
vibration numerical program is carried out for the Yitong River
Bridge. Firstly, a model updating technology is used to make the
initial FE model match better with the real structure based on
the measured dynamic parameters. Then, the aforementioned
numerical method and program for the vehicle-bridge coupled
vibration is verified by comparing the tested and simulated
dynamic data.
5.1. Model updating
As shown in Table 2, the mean and maximum deviation of the
calculated modal frequencies compared with the measured
response are 5.94% and 13.37%, respectively, and the MAC value
ranges from 0.901 to 0.977. Because the error of the initial FE
model is obvious, a model updating process is necessary to make
the dynamic features of the model agree better with that of the

actual bridge. Generally speaking, the error between the computed

and measured modal frequency is mainly affected by the following
factors: (1) discretization of the continuous structure; (2) the
uncertainty of structural geometry and boundary conditions; (3)
variability of material properties; (4) the error from dynamic
response recording and processing. Among these factors, (2) and
(3) are usually emphasized and considered in FE model updating.
For the bridge in the present study, the initial FE model is built
based on the construction drawing and some measured parameters
(axial forces of suspenders and horizontal tie rods). The structural
boundary conditions of the bridge are also very simple and clear.
Therefore, the geometric parameters and boundary conditions
can be assumed to be correct and do not need further modifications. As a result, the structural material properties are chosen as
the main modification variables in the model updating process. A
zero-order approximation method [27] is used to update the initial
FE model of the bridge. The relative error sum of squares for the
calculated and measured frequencies is taken as the optimization
object function, of which an error less than 5% is taken as the
constraint condition, and the variation range of the parameters is
specified as 15% of the variation of the initial material properties
value. More details for the model updating and optimization
method can be found in Li et al. [27], and only some major model
updating results are discussed here.
As shown in Table 4, the change of the design variables during
the model updating process is from 0.49% to 9.22%. Particularly, the
modulus of elasticity of the arch rib, assistant arch rib and suspender, and material density of girder change over 8%, which indicates that these variables have high sensitivity to the structural
dynamic characteristics. The natural vibration frequencies of the

Fig. 9. Predicted dynamic response of the bridge under one moving truck on right lane 3.

Y. Li et al. / Engineering Structures 106 (2016) 243260

updated model for the bridge (listed in Table 2) show a very good
match with the measured values, and the average error of the calculated and tested frequencies decreases from the previous 5.94%
to the current 2.99%. The MAC values change very slightly before
and after model updating. All the results indicate that the updated
FE model can better represent the dynamic property of the real
bridge and can be used in the following vehicle-bridge coupled
vibration simulation.
5.2. Verification of vehicle and bridge coupled vibration program
To verify the numerical model and program of the vehiclebridge coupled vibration system for the Yitong River Bridge, one
actual dynamic test case (one truck moving on the right lane 3 at
the speed of 60 km/h) is reproduced under the measured road surface roughness sample via the aforementioned simulation program. The vehicle model is plotted in Fig. 3 and the parameters
are listed in Table 2. The time-dependent response from the field
test and numerical analysis is compared on several typical measurement points as shown in Fig. 8. The experimental and numerical dynamic response in all the selected measurement points
agree reasonably with both the time history curve shape and the
peak values. The relative errors of the response extreme values
are less than 5% for the displacement, suspender force and stress,
and the errors for the acceleration response are less than 8%. The
good match of the measured and simulated results for the different
components and response types demonstrates that the proposed
vehicle-bridge coupled vibration model and program work very
well and possess high accuracy and applicability. The result also
further proves that the updated FE model can represent the
dynamic property of the actual bridge well. Based on the above
analysis, the proposed model and program can be used to predict
the structural dynamic behavior under some other moving truck
loading conditions that are not included in the previous onsite
testing.
6. Dynamic analysis of the bridge under moving vehicles
6.1. Impact effect analysis of the bridge due to moving vehicles
Due to the limitation of experimental conditions, some impact
factors and dynamic responses cannot be included and measured
in the field test. In the subsequent study, the verified numerical
simulation methodology is used for a complete and systematic
dynamic analysis of the Yitong River Bridge. Several main factors
including vehicle number and velocity, truck position on bridge

255

lanes, and road surface condition, are taken into account in the
simulation to comprehensively reveal the vibration performance
of the bridge under moving vehicle loading.
6.1.1. Dynamic behavior characteristics of the bridge under moving
vehicles
Heavy trucks are usually in the inside lanes, i.e., lanes 3 and 4
for the Yitong River Bridge. To study the general dynamic behavior
of the bridge, a typical case for one truck crossing the bridge on the
right lane 3 at the speed of 50 km/h under good RSC is computed.
The maximum vertical displacement response occurs at mid-span
for the bridge deck as shown in Fig. 9(a). This result is consistent
with the previous analysis from the designers, and also indicates
that the dynamic impact factor of the whole bridge can be represented by this section. The transverse distribution of the vertical
deflection at the mid-span section is linear as displayed in Fig. 9
(b), the maximum deflection of 6.8 mm appears in the same side
walkway with the truck, and a 3.1 mm opposite deflection occurs
in another side walkway. On the other hand, the torsional deflections of the bridge deck at the mid-span change very little along
the transverse direction, and two extreme values appear in both
walkways with values 2.23  104 rad and 2.65  104 rad,
respectively, as shown in Fig. 9(c). The above deflection result indicates that while the torsional stiffness of the steel girders is very
high, the torsional stiffness of the main span of the bridge is relatively weak. Therefore, the dynamic behavior of the bridge under
moving vehicles on one side lane is given more attention in the following study. The peak value distribution of axial forces for the
suspenders is shown in Fig. 9(d). The extreme value decreases from
suspenders 2 to 8. In the following analysis, suspenders 2, 5 and 8
are taken as the representation of the component.
6.1.2. Dynamic impact effect analysis of the bridge under moving
vehicles
In this part, the influence of several key factors (speed, lane
position, truck number, and road surface condition) on the
dynamic impact effect of the bridge under moving vehicles is analyzed and discussed. The corresponding quasi-static transversal
loading position of the two typical cases is shown in Fig. 1(b),
and the boundary condition of the bridge model is same as that
described in Section 4.2.
Firstly, the influence of vehicle speed and road surface condition
is discussed. Considering possible speeding and the actual maintenance level for the bridge, the speed with a range from 20 km/h to
80 km/h and the RSC with very good, good and average classifications are included in the analysis. The displacement impact

Fig. 10. Predicted impact factors of the whole bridge under one and two moving trucks.

256

Y. Li et al. / Engineering Structures 106 (2016) 243260

Fig. 11. Predicted axial force impact factors of the suspenders for the bridge under two moving trucks.

factors of the bridge for the measured point A position (shown as

Fig. 5) at the mid-span under the two loading cases, two trucks
on lane 3 and 4 and one truck on lane 3 (shown as Fig. 1(b)), with
various velocities and RSCs are shown in Fig. 10. The impact factor
increases with a degrading RSC, but the linear influence relationship does not reappear for the vehicle velocity. Several peak values
for each RSC basically occur in the speed range from 60 km/h to
70 km/h. More specifically, for the average RSC, the extreme
impact factor 0.205 under two trucks is larger than the 0.138 under
one truck, which is obviously different from the observation that
the impact factor decreases or has no change with the increase of
acting truck numbers for regular CFST arch bridges [19]. Under
the good RSC, the maximum impact factors are 0.079 and 0.075
corresponding to one and two trucks, respectively. The result indicates that the impact factor from numerical analysis under good
RSC agrees with the tested value and is slightly less than the design
value 0.082. However, the maximum impact factor under the average RSC is 2.5 times the design value. This difference deserves
more attention in future research and practice.
For the suspenders, only the results under two trucks are plotted in Fig. 11. Obviously the shorter suspender undertakes more
significant impact under moving vehicles. The peak impact factors
reach 0.255, 0.221 and 0.204 for suspenders 2, 5 and 8, respectively, under the two moving trucks in the speed range from
20 km/h to 80 km/h for the average RSC. When considering the
design vehicle speed of 60 km/h, the corresponding maximum
impact factors are 0.162, 0.221 and 0.154. For the good RSC, the
maximum impact factor 0.093 occurs in suspender 2 under the

speed of 60 km/h and is 13% more than the design value 0.082.
Therefore, taking the impact factor of short suspenders as the representation is more conservative for the bridge.
Secondly, the influence of the heavy trucks lateral position on
the bridge dynamic behavior is also an important issue for this
wide bridge with 8 lanes. The case with one truck moving in
different lanes at the speed of 60 km/h under the average RSC is
analyzed to investigate this effect. The maximum vertical displacement on the sidewalk at the mid-span and maximum axial force
for suspender 2 are shown in Fig. 12. The two typical dynamic
responses increase with the truck being closer to the sidewalk
due to the transversal distribution effect. However, the trucks lateral position does not have significant effect on the impact factor of
the bridge. The impact factor for the bridge does not decrease with
the increase of the response amplitude as the results obtained from
regular CFST arch bridges [19,36].

Table 5
Discomfort Rating criterion in ISO 2631-1.
Discomfort rating

aw (m/s2) Range

ISO 2631-1 Discomfort level

1
2
3
4
5
6

<0.315
0.3150.63
0.501.0
0.81.6
1.252.5
>2.0

Not uncomfortable
A little uncomfortable
Fairly uncomfortable
Uncomfortable
Very uncomfortable
Extremely uncomfortable

Fig. 12. Predicted dynamic response and impact factor (IF) of the bridge under one truck running on different lanes at the speed of 60 km/h and average RSC.

257

Y. Li et al. / Engineering Structures 106 (2016) 243260

6.2. Evaluation for vibration comfort

A bridges vibration due to moving traffic is important for two
reasons. In addition to the dynamic impact effect from moving
vehicles as discussed above, another reason is that excessive vibrations may be noticeable to persons walking or driving on the
bridge. Although the vibrations are not related to issues of safety,
it may have the psychological effect of impairing public confidence.
In this part, the vibration evaluation on the ride and pedestrian
comfort for the bridge is performed.
6.2.1. Ride comfort evaluation
An examination of the literature on ride comfort assessment
indicates that the method based on ISO 2631-1 [22] has the most
extensive applications [44]. It uses root-mean-square (RMS) magnitudes of vibration accelerations as the standard for ride comfort
as shown in Table 5.
For vibrations in more than one direction, the weighted RMS
acceleration aw determined from vibrations in orthogonal coordinates is calculated as

h
i1=2

2
2
2
aw kx awx ky awy kz awz

29

where awx , awy and awz are the RMS of longitudinal, lateral and vertical acceleration from the driving seat position, respectively, and
the corresponding weighting factors are kx 1:4, ky 1:4, and
kz 1:0.
The RMS of each directional weighted acceleration is calculated
as


awj jx;y;z

80

1=2
W 2 f Gaj f df

30

where Gaj f is the power spectrum density (PSD) in the jth direction obtained from the acceleration time history. The acceleration
ax , ay and az at the driving seat position can be obtained based on
the vehicle body vibration via: ax av h h1 ; ay av y  av u h1 ;
az av z  av h L4  av u b1 . Wf is a frequency weighted function
and is determined by:

x; y direction Wf

1:0 1 Hz < f 6 2 Hz

2 Hz < f 6 80 Hz
8
1:0 1 Hz < f 6 4 Hz
>
<
2
4 Hz < f 6 8 Hz
z direction Wf f
>
:8
8 Hz < f 6 80 Hz
f
2
f

31

The cases with one truck passing the bridge under different
speeds and RSC on lane 3 are computed to assess the ride comfort
as shown in Table 6. The results indicate that the RSC has a remarkable influence on the ride comfort, which decreases from not
uncomfortable to very uncomfortable with the RSC changing
from very good to average. Under the same RSC, the aw value
increases with the increase of the vehicle speed except for a few
points. Specifically for the bridge, when the RSC is kept at a good
level, the ride comfort will change from a little uncomfortable
to fairy uncomfortable as the velocity exceeds 40 km/h; when
the RSC degrades to the average level, the ride comfort reaches
the very uncomfortable level except for a vehicle speed being
less than 20 km/h. On the whole, the ride comfort of the Yitong
River Bridge is not so satisfactory and below the ordinary level.
The ride comfort for the bridge obviously is worse than that of
the regular CFST arch bridge [45,35,36]. Such a conclusion also suggests that paying more attention on the ride comfort problem and
related research is necessary in the engineering practice for this
bridge type.

Table 6
Comfort rating result for the bridge under different speed and RSC.
Speed (km/h)

aw (m/s2) Very good

Discomfort level

aw (m/s2) Good

Discomfort level

aw (m/s2) Average

Discomfort level

20
30
40
50
60
70
80

0.230
0.314
0.287
0.373
0.405
0.467
0.417

Not uncomfortable
Not uncomfortable
Not uncomfortable
A little uncomfortable
A little uncomfortable
A little uncomfortable
A little uncomfortable

0.523
0.566
0.604
0.746
0.870
0.930
0.941

A little uncomfortable
A little uncomfortable
A little uncomfortable
Fairly uncomfortable
Fairly uncomfortable
Fairly uncomfortable
Fairly uncomfortable

0.931
1.207
1.318
1.511
1.739
1.840
1.601

Fairly uncomfortable
Very uncomfortable
Very uncomfortable
Very uncomfortable
Very uncomfortable
Very uncomfortable
Very uncomfortable

Fig. 13. Extreme values of accelerations on walkways under one truck with a good RSC.

258

Y. Li et al. / Engineering Structures 106 (2016) 243260

6.2.2. Pedestrian comfort analysis for the bridge

Because the bridge also provides walkways for pedestrians, the
compliance with maximum acceleration limits is especially important to ensure the serviceability of the bridge. In the present study,
the criterion from EN03 (Table 3) is applied to evaluate the vibration comfort of pedestrians. During the dynamic test for the bridge,
only a few points on the sidewalks are measured for the acceleration response. More complete vibration information on the sidewalks along the whole bridge is numerically predicted to rate the
pedestrian comfort. Fig. 13 shows the maximum and minimum
envelopes of accelerations along the walkway of the bridge under
one truck and a good RSC at several typical vehicle speeds based on
the numerical analysis. The maximum vertical and lateral accelerations are found in the range from 75 m to 100 m along the bridge
length, not at the mid-span, similar to the maximum displacement
response, which suggests that the pedestrian comfort rating result
under dynamic testing based on several points, such as the midspan section, may have limitations and errors. Therefore, a series
of numerical analyses are performed to gain a more complete evaluation of the pedestrian comfort for the bridge based on the verified vehicle-bridge coupled vibration program. The number of
vehicles, speed and RSC are considered during the analysis process.
For each analysis case, just one absolute maximum acceleration
value obtained from the response time histories for all the nodes
along the walkways is recorded and used for evaluating the
pedestrian comfort as shown in Fig. 14 under various occasions.
Under the good RSC, the maximum vertical accelerations are up
to 0.88 m/s2 and 1.92 m/s2 due to one and two moving trucks,

respectively; the corresponding pedestrian comfort is at the

average and poor level based on the criteria shown in Table 3,
respectively, when the speed limit of 60 km/h is observed. If considering possible speeding to 80 km/h, the corresponding absolute
maximum acceleration value changes to 1.22 m/s2 and 2.01 m/s2,
respectively, which means the pedestrian comfort will downgrade
to the poor level. For the lateral vibration, the maximum accelerations are 0.12 m/s2 and 0.19 m/s2 under one and two trucks, respectively, which means that the pedestrian comfort is in the average
level. When the RSC changes into the average level, the maximum
vertical acceleration goes up to 1.74 m/s2 and 3.72 m/s2 due to one
and two trucks passing at a speed below 60 km/h, and the pedestrian
comfort falls in the poor and not acceptable levels, respectively.
When considering possible speeding to 80 km/h, the pedestrian comfort level all stays in the not acceptable level. On the other side, the
maximum lateral acceleration goes up to 0.24 m/s2 and 0.38 m/s2
under one and two trucks, respectively, which will lead to the
average and poor levels for the pedestrian comfort.
These observations indicate that the RSC has a remarkable
effect on the pedestrian comfort of the bridge; the comfort level
decreases with the number of moving vehicles changing from
one to two; the rating result is controlled by the vertical vibration
rather than the lateral one; for the bridge just any one of the two
situations (two trucks passing in parallel or speeding) appears,
the pedestrian comfort will be down to the poor and not acceptable level for the good and average RSC, respectively; and the
pedestrian comfort level of the bridge is lower than the usual
best or average level for the regular CFST arch bridges [25,45].

Fig. 14. Predicted maximum acceleration of the bridge on sidewalk under various conditions.

Y. Li et al. / Engineering Structures 106 (2016) 243260

7. Conclusions
In the present study, the dynamic performance of a large span
new type irregular CFST arch bridge with hybrid girders with moving vehicles is studied experimentally and numerically. The refined
finite element model of the bridge is devised to represent its complicated mechanical characteristics better. Based on experimental
data, the FE model is updated, and the proposed vehicle-bridge
coupled vibration analysis model is verified. A comprehensive
numerical analysis, supplementing the dynamic test, is conducted
to predict the dynamic property of the bridge under moving vehicles, including the dynamic impact effect, ride comfort, pedestrian
comfort and some related influence factors. For the present study,
the following conclusions are drawn:
1. Based on the results of modal analysis and identification for the
bridge, the arch ribs have weak out-of-plane stiffness, and the
bridge deck in the main span has obviously close torsional
and bending coupled modes. The above characteristics deserve
more attention in future investigations for this bridge type.
2. The comparison of numerical predictions and field measurements indicates that the vehicle-bridge coupled vibration
numerical model with an updated bridge model can be confidently applied to simulate and predict the dynamic behavior
of vehicle-bridge coupled systems.
3. The tested and numerical impact factor of the bridge matches
very well with the design value 0.082 from the specifications
of China under good RSC. However, when RSC changes into
the average level (which is very possible during the service period), the maximum impact factor is 2.5 times the design value.
The impact factor formula from design standards significantly
underestimates the dynamic impact effect, which may result
in an unfavorable influence on the bridge safety.
4. The parametric analysis shows that while the speed and truck
lane position do not have a significant direct effect on the
impact factor, the impact factor increases with the RSC degrading; considering two trucks passing on the two side lanes is
more suitable than one truck when determining the impact factor of the bridge.
5. The influence on the ride comfort from RSC and speed is remarkable for the bridge. The ride comfort changes sharply from not
uncomfortable to very uncomfortable under various vehicle
speeds and RSCs for the bridge. The ride comfort for the bridge
is obviously worse than that of the regular CFST arch bridge.
The ride comfort problem for this bridge type is worth giving
more concern in the future engineering practice and research.
6. The pedestrian comfort level of the bridge ranges from average to not acceptable. The structural features of the simply
supported steel girder at mid-span with weak torsional stiffness
and intensive bending-torsion coupled vibration modes, are the
leading reason for the poor pedestrian comfort of the bridge.
The pedestrian comfort level for the bridge is lower than the
usual best or average level for the regular CFST arch bridges.
It is noted that these conclusions are drawn based on the
dynamic study on the Yitong River Bridge and are difficult to
extend to the general situation. However, this study presents several noteworthy dynamic features for this new bridge type in
regards to mode characteristics, determination of design impact
factors, and evaluation of ride and pedestrian comfort.

Acknowledgments
The authors greatly appreciate the financial support provided
by the National Science Foundation of China (Grant No.

259

51108132) and the Fundamental Research Funds for the Central

Universities (Grant No. HIT.NSRIF.2012059). The contents of the
paper reflect only the views of the authors.
References
[1] Ashebo DB, Chan THT, Yu L. Evaluation of dynamic loads on a skew box girder
continuous bridge part I: field test and modal analysis. Eng Struct 2007;29
(6):105263.
[2] Ashebo DB, Chan THT, Yu L. Evaluation of dynamic loads on a skew box girder
continuous bridge part II: parametric study and dynamic load factor. Eng
Struct 2007;29(6):106473.
[3] Bergman W. Theoretical prediction of the effect of traction on cornering force.
SAE Trans 1961;69(4):4556.
[4] Broquet C, Bailey SF, Fafard M, Brhwiler E. Dynamic behavior of deck slabs of
concrete road bridges. J Bridge Eng 2004;9(2):13746.
[5] Chang D, Lee H. Impact factors for simple-span highway girder bridges. J Struct
Eng 1994;120(3):70415.
[6] Chatterjee PK, Datta TK, Surana CS. Vibration of continuous bridges under
moving vehicles. J Sound Vib 1994;169(5):61932.
[7] Chen ZW, Xu YL, Li Q, Wu DJ. Dynamic stress analysis of long suspension
bridges under wind, railway, and highway loadings. J Bridge Eng 2011;16
(3):38391.
[8] Chinas Ministry of Transport. General code for design of highway bridges and
culverts. JTGD60-2004. Beijing: Chinas Ministry of Transport; 2004.
[9] Chompooming K, Yener M. The influence of roadway surface irregularities and
vehicle deceleration on bridge dynamics using the method of lines. J Sound Vib
1995;183(4):56789.
[10] Dai ZZ, Zheng SF. Urban bridge space. Nanjing: Southeast University Press;
2003.
[11] Deng L, Cai CS. Development of dynamic impact factor for performance
evaluation of existing multi-girder concrete bridges. Eng Struct 2010;32
(1):2131.
[12] Deng L, Yu Y, Zou Q, Cai C. State-of-the-art review of dynamic impact factors of
highway bridges. J Bridge Eng 2015;20(5):04014080.
[13] Dodds CJ, Robson JD. The description of road surface roughness. J Sound Vib
1973;31(2):17583.
[14] EN03. Design of footbridges guideline. Germany; 2008.
[15] Fafard M, Laflamme M, Savard M, Bennur M. Dynamic analysis of existing
continuous bridge. J Bridge Eng 1998;3(1):2837.
[16] Gim G, Nikravesh PE. Analytical model of pneumatic types for vehicle dynamic
simulations part 2. Comprehensive slips. Int J Vehicle Des 1991;12(1):1939.
[17] Huang DZ, Wang TL. Impact analysis of cable-stayed bridges. Comput Struct
1992;43(5):897908.
[18] Huang D, Wang T-L, Shahawy M. Impact studies of multigirder concrete
bridges. J Struct Eng 1993;119(8):2387402.
[19] Huang DZ. Dynamic and impact behavior of half-through arch bridges. J Bridge
Eng 2005;10(2):13341.
[20] Humar JL, Kashif AH. Dynamic response analysis of slab type bridges. J Struct
Eng 1995;121(1):4862.
[21] ISO. Mechanical vibration-road surface profiles-reporting of measured data.
ISO 8068: (E), Geneva; 1995.
[22] ISO. Mechanical vibration and shock-Evaluation of human exposure to whole
body vibrationpart I: general requirements. ISO 2631-1:1997E, Geneva;
1997.
[23] Kim CW, Kawatani M, Kim KB. Three-dimensional dynamic analysis for bridgevehicle interaction with roadway roughness. Comput Struct 2005;83(1920):162745.
[24] Kwasniewski L, Li H, Wekezer J, Malachowski J. Finite element analysis of
vehicle-bridge interaction. Finite Elem Anal Des 2006;42(11):9509.
[25] Li BN, Dai H. Analysis of dynamic comfort of a concrete-filled steel tubular arch
bridge under moving vehicle load. Steel Constr 2011;26(151):2832.
[26] Li Huile, Xia He, Soliman Mohamed, Frangopol Dan M. Bridge stress calculation
based on the dynamic response of coupled train-bridge system. Eng Struct
2015;99:33445.
[27] Li X, Zhang DY, Yan WM, Chen YJ. Operational model identification and
calibration of CFST arch bridge based on ambient excitation. Eng Mech
2013;30(9):8190.
[28] Li Yan, Chen Yan-jiang, Huang Xin-yi. Impact effect study on long-span
irregular concrete filled steel tube arch bridge under moving vehicles. J Harbin
Inst Technol 2010;42(1). 109105.
[29] Liu Y, Li H. Modal parameters identification under multi-operational grades
and its application to a cable-stayed bridge. In: The 6thInternational
conference on bridge maintenance, safety and management. Stresa, Italy;
2012. p. 4317.
[30] Livingston DI, Brown JE. Physics of the slipping wheel-Force and torque
calculations for various pressure distributions. Rubber Chem Technol 1969;43
(2):4556.
[31] Moghimi Hassan, Ronagh Hamid R. Development of a numerical model for
bridgevehicle interaction and human response to traffic-induced vibration.
Eng Struct 2008;30:380819.
[32] OBrien EJ, Cantero D, Enright B, Gonzlez A. Characteristic dynamic increment
for extreme traffic loading events on short and medium span highway bridges.
Eng Struct 2010;32(12):382735.

260

Y. Li et al. / Engineering Structures 106 (2016) 243260

[33] Sun C, Wu QX, Chen BC. Analysis on dynamic behavior of concrete filled steel
tubular arch bridge under vehicle moving loads. J Highway Transp Res Develop
2007;24(12):549.
[34] Wang T-L, Huang D, Shahawy M. Dynamic behavior of slant-legged rigidframe highway bridge. J Struct Eng 1994;120(3):885902.
[35] Wu QX, Chen BC, Takahashi K, Xi LZ. Vehicle-bridge dynamic analysis and
riding comfort evaluation of New Saikai Bridge. J Highway Transp Res Develop
2008;25(5):617.
[36] Wu QX, Huang WK, Chen BC. Vehicle-induced vibration research and dynamic
analysis for concrete-filled steel tubular arch bridges. Eng Mech 2013;30
(1):14755.
[37] Wyss Jean-Charles, Su Di, Fujino Yozo. Prediction of vehicle-induced local
responses and application to a skewed girder bridge. Eng Struct
2011;33:108897.
[38] Xia H, Xu YL, Chan T. Dynamic interaction of long suspension bridges with
running trains. J Sound Vib 2000;237(2):26380.
[39] Xu YL, Guo WH. Dynamic analysis of coupled road vehicle and cable-stayed
bridge system under turbulent wind. Eng Struct 2003;25:47386.
[40] Xu YL, Li Q, Wu DJ, Chen ZW. Stress and acceleration analysis of coupled
vehicle and long-span bridge systems using the mode superposition method.
Eng Struct 2010;32:135668.

[41] Yang YB, Lin CW, Yau JD. Extracting bridge frequencies from the dynamic
response of a passing vehicle. J Sound Vib 2004;272(3-5):47193.
[42] Yin Xinfeng, Fang Zhi, Cai CS, Deng Lu. Non-stationary random vibration of
bridges under vehicles with variable speed. Eng Struct 2010;32:216674.
[43] Yin XF, Cai CS, Fang Z, Deng L. Bridge vibration under vehicular loads tire
patch contact versus point contact. Int J Struct Stabil Dynam 2010;10
(3):52954.
[44] Yin XF, Cai CS, Liu Y, Fang Z. Experimental and numerical studies of nonstationary random vibrations for a high-pier bridge under vehicular loads. J
Bridge Eng 2013;18(10):100520.
[45] Yoshimura M, Wu QX, Takahashi KSN, Furukawa K. Vibration analysis of the
Second Saikai Bridge-a concrete filled tubular (CFT) arch bridge. J Sound Vib
2006;290:388409.
[46] Yau JD, Yang YB. Vibration of a suspension bridge installed with a water
pipeline and subjected to moving trains. Eng Struct 2008;30(3):63242.
[47] Zhang H, Zhang ZC, Xie X, Gong RM. Theoretical study on the impact factor of
crescent-shaped multi-rib concrete filled steel tube truss arch bridges. Eng
Mech 2008;25(7):11824.
[48] Zhong Hai, Yang Mijia, Gao Zhili. Dynamic responses of prestressed bridge and
vehicle through bridge-vehicle interaction analysis. Eng Struct 2015;87:
11625.