Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Vibration analysis of concrete bridges during a train pass-by using

various models
To cite this article: Qi Li et al 2016 J. Phys.: Conf. Ser. 744 012140

View the article online for updates and enhancements.

This content was downloaded from IP address 122.167.130.255 on 31/10/2018 at 10:01

MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

Vibration analysis of concrete bridges during a train pass-by

using various models

Qi Li, Ke Wang, Shili Cheng, Wuqian Li, Xiaodong Song

(Department of Bridge Engineering, Tongji University, Shanghai 200092, China)
e-mail: [email protected]

Abstract: The vibration of a bridge must be determined in order to predict the bridge noise
during a train pass-by. It can be generally solved with different models either in the time
domain or the frequency domain. The computation cost and accuracy of these models vary a lot
in a wide frequency band. This study aims to compare the results obtained from various models
for recommending the most suitable model in further noise prediction. First, train-track-bridge
models in the time domain are developed by using the finite element method and mode
superposition method. The rails are modeled by Timoshenko beam elements and the bridge is
respectively modeled by shell elements and volume elements. Second, power flow models for
the coupled system are established in the frequency domain. The rails are modelled by infinite
Timoshenko beams and the bridge is respectively represented by three finite element models,
an infinite Kirchhoff plate, and an infinite Mindlin plate model. The vibration at given
locations of the bridge and the power input to the bridges through the rail fasteners are
calculated using these models. The results show that the shear deformation of the bridge deck
has significant influences on the bridge vibration at medium-to-high frequencies. The Mindlin
plate model can be used to represent the U-shaped girder to obtain the power input to the
bridge with high accuracy and efficiency.

Key Words: railway bridges; vibration and noise; train-track-bridge interaction; power flow

1. Introduction
More and more elevated concrete bridges have been constructed for the growing metro lines in the
suburbs of large cities due to its lower construction costs compared with tunnels. However, the noise
from the elevated systems under the moving trains has adverse effects on the residents’ well-being and
health. Rolling noise and bridge noise are the two dominant noise of the elevated metro system when
the train speeds are less than 120 km/h. The prediction and control of rolling noise has been
extensively investigated in the past, and a comprehensive review on this area can be found in the book
by Thompson [1]. Nevertheless, the structure-borne noise from railway concrete bridges [2-8] only
gained lots of concerns recently.
The vibration analysis of a bridge under moving trains is the basis for the bridge noise prediction.
Janssens and Thompson [9] proposed a statistical energy analysis (SEA)-based method to predict
vibration and noise from railway steel bridges in a large frequency range. Zhang et al. [4] studied the
low-frequency concrete bridge noise by employing the three-dimensional finite element method for
the vibration analysis and boundary element method for the acoustical analysis. Li et al. [2, 5] adopted
the modal superposition method to simulate the train-track-bridge vibration in the time domain, and

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

the three-dimensional and 2.5-dimensional modal acoustic transfer vector techniques in the frequency
domain for the prediction of bridge noise below 200 Hz. To overcome the inefficiency of finite
element method based vibration analysis and boundary element method based acoustical analysis, Li
et al. [7] utilized the finite element and statistical energy analysis method to form a hybrid procedure
for bridge noise prediction. Another attempt to tackle the problem is the force-method based power
flow scheme proposed by Li and Wu [10] for the investigation of the train-track-bridge dynamic
interaction in the frequency domain. Li et al. [8] then combined this power flow method with the 2D
infinite/finite element method to predict the rail noise and bridge noise below 1000 Hz with good
balance between accuracy and efficiency.
In our previous studies, different methods have been proposed for the simulation of bridge
vibration during a train pass-by. It is then necessary to find which method can simulate the bridge
vibration with both efficiency and accuracy. In this study, three finite element models of the bridge are
first compared in the train-track-bridge interaction analysis by using both of the time domain and the
frequency domain methods. Then two infinite plate models are adopted in the frequency domain
vibration analysis to investigate the accuracy of the simple models for the representation of the
complex real bridge.

2. Train-track-bridge interaction analysis in time domain

2.1. Calculation method

Train-track-bridge dynamic interaction analysis has been widely used to obtain the time-varying
responses of the coupled system. Li et al. [11] have proposed a mode superposition method based
scheme to deal with this problem with the aid of the developed computer program named as VBC2.0.
Vehicle
subsystem

Wheel-rail contact

Track subsystem
Bridge subsystem
Ground Fastener

Ground
Figure 1. The train-track-bridge interaction model in time domain.
In VBC2.0 program, a coupled train-track-bridge system is divided into the vehicle subsystem, the
track subsystem, and the bridge subsystem (see Figure 1). Each subsystem can be expressed by an
equation of motion. For train-track-bridge system, the equation of motion could be expressed as [2,
11]:
q  v = Φ Tv f v  2ξ v ω v q v  ω 2v q v
 (1)
q  t = Φ Tt f t  2ξ t ω t q t  ω 2t q t

 b = Φ Tb f b  2ξ b ω b q b  ω b2 q b
 q
where q , Φ , ω and ξ are the modal coordinate vector, modal shape matrix, modal frequency matrix,
and modal damping matrix of each subsystem respectively; f denotes the combination of the pseudo-
forces within each subsystem and the external forces exerted by other subsystems connected to it; the
subscript v, t and b represent vehicle, track and bridge subsystems respectively; and the superscript T
is the transpose operation of a matrix. Equation (1) can be solved using the fourth-order Runge-Kutta
method. Roughness spectrum provided in ISO3095:2005 standard [12] is used for the excitation
source of the train-track-bridge system by transforming it into samples of rail roughness in the space
domain.

2
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

2.2. Finite element models

In this study, a 30-metre-long one-span U-shaped girder (see Figure 2) is used to investigate the
effects of various models on the bridge vibration during a train pass-by. Three kind of finite element
models of the bridge are respectively developed by using SEHLL181, SHELL63 and SOLID95
elements in the software ANSYS. According to the principle of at least six elements per wavelength,
the element size should not exceed 0.20 m at the interest frequency range below 1500 Hz. The rails are
simulated by BEAM188 elements with the lengths 40m longer than the one-span girder. The rails
outside of the concerned span are connected to the rigid ground by rail fasteners. The parameters of
the bridge, vehicle and rail are listed in Table 1.
720 1825 1825 720
300
1800

1516

658 1887 1887 658

Figure 2. Cross-section of the bridge (unit: mm).
Table 1. Main parameters of the models.
Parameters Unit Value
Young’s modulus MPa 4 10
Bridge Density kg/ 2600
Poission’s ratio -- 0.2
Young’s modulus MPa 2.1 10
Rail Density kg/ 7800
Poission’s ratio -- 0.3
Vertical stiffness MN/m 60
Horizontal stiffness MN/m 20
Rail fastener Vertical damping kN s/m 80
Horizontal damping kN s/m 60
Trail Wheelset mass Kg 1150
Wheel Contact stiffness MN/m 1.08 10
Motor car Wheelset mass Kg 1900
Contact stiffness MN/m 1.12 10

2.3. Effect of different models on the bridge vibration

Figure 3 shows the one-third octave-band acceleration level spectra for the bottom slab at the
mid-span of the bridge on the pass-by time. It can be observed that various finite element models
lead to similar results below 630 Hz. However, the relative deviation becomes prominent
above 630 Hz. The bridge model using SHELL181 produces the computed vibration 7 dB larger
than that using the SHELL63 model, because the former include the shear deformation and
rotation inertia effects while the latter one neglects these effects. The Solid95 model gives larger
bridge acceleration levels than the SHELL181 model mainly due to the local deformation under the
concentrated forces of the fasteners.
Figure 4 illustrates the power transferred from the rails to the bridge using different models for
the bridge. It can be seen from the figure that various models have slight influences on the bridge
power for the frequencies above 630 Hz. The bridge models using SHELL181 elements and
SOLID95 elements give consistent results about 4 dB larger than that using the SHELL63 model.
The former two models both include shear deformation and rotation inertia effects while the latter
one ignores these effects. It is noted that the solid element model has 16554 elements while the shell

3
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

model has only 1048 elements. It is more suitable using shell element to model the U-shaped
girder, as it’s efficient and of high accuracy in terms of the computed power input to the bridge.
Acceleration Level (dB Re 1m/s2)

120 135

Power Level (dB Re 10-12 Watt)

130
115
125
110
120
105 115

100 110
105
95 Shell181 Shell181
Shell63 100 Shell63
90 Solid95 Solid95
95

85 90
20 40 80 160 315 630 1250 20 40 80 160 315 630 1250
Frequency (Hz) Frequency (Hz)
Figure 3. Vertical acceleration level spectra Figure 4. Power input to the U-shaped bridge
of bottom slab of bridge. by time domain method.

3. Train-track-bridge interaction analysis in frequency domain

3.1. Calculation method

Power flow analysis in the frequency domain could be used to investigate the vibration of a coupled
system in the medium-to-high frequency range. Li et al [8, 10] proposed a force-method-based power
flow approach to solve the train-track-bridge coupling system shown in Figure 5. Only the rigid wheel
and the linearized contact spring are considered for the train modelling in the medium-to-high
frequency range. The rail is modeled by infinite Timoshenko beam. The bridge can be represented by
either a finite element model or an infinite plate model.
Rigid masses(wheel)

Infinite Timoshenko Wheel-rail

beam(rail) contact spring

Ground FE model or Infinite plate(bridge)

Fastener

Ground
Figure 5. The force-method-based power flow model of coupled train-track-bridge system.
The compatibility equation of the coupled system can be obtained by considering the internal
forces of the spring-dashpot pairs within the system as unknowns [8, 10]
δ   F    ΔP     Λ   F   （2）

where is the angular frequency; is the unknown force vector of the spring-dashpot pairs;
is the relative compression displacement vector of the spring-dashpot pairs caused by the unit
harmonic excitation exerted on the released rail through the active wheel; is the dynamic
flexibility matrix of the released structure; and is the dynamic flexibility matrix corresponding to
the stiffness and damping of the spring-dashpot pairs.
The unknown spring-force can be obtained from equation (2). Then the vibration velocities
and power of the rail and bridge could be calculated based on the released structures under known
forces. The wheel-rail contact force could be obtained through the compatibility condition at
the wheel-rail interface [1].
The moving roughness method [1] is used in the frequency domain analysis but the wheel position
changes when the trains are moving on the bridge. In this study, two typical load cases were chosen to

4
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

include the effect of the wheel positions. In case A, the front and rear bogies of two adjacent vehicles
locate at the middle of the bridge symmetrically, and in case B the two bogies of one vehicle load on
the ends of the bridge symmetrically. In both load cases, each wheel is respectively chosen as the
active wheel in turn. And the vibration of the bridge is calculated by averaging the results obtained
from various load cases and active wheel scenarios.

3.2. Finite element models and infinite plate models of the bridge
Finite elements of the bridge can be used to calculate the dynamic flexibility matrix of the released
bridge in Equation (2) by using mode superposition method. Thus the three bridge models used in the
time domain analysis are also adopted in the frequency domain analysis. In addition, the flexibility of
the bottom slab of the U-shaped girder can be represented by an infinite plate [13]. As there are rail
support blocks on the slab, the thickness of infinite plate could be set to 1.4 times of the bottom slab
thickness according to parameter analysis. The infinite Kirchhoff plate (thin plate theory) and Mindlin
plate (thick plate theory) [13] are used to model the bridge separately in the power flow analysis.

3.3. Effect of different bridge models on the bridge vibration

Figure 6 shows the acceleration levels of the bottom slab beneath the rail by using different bridge
models. It can be seen from Figure 6.(a) that three finite element models of the bridge give similar
results in the frequency range of 40 Hz to 400 Hz. For the frequency above 630 Hz, the difference
becomes distinct, which matches the trend obtained through the time domain analysis. Figure 6.(b)
also indicates the effects of shear deformation and rotation inertia do have large influence on the
bridge vibration above 630 Hz.
Acceleration Level (dB Re 1m/s2)
Acceleration Level (dB Re 1m/s2)

130 130

120 120

110 110

100 100

90 90 Kirchhoff
Mindlin
Shell63
80 Shell181 80
Solid95
70 70
20 40 80 160 315 630 1250 20 40 80 160 315 630 1250
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 6. Vertical acceleration level spectra of bottom slab of bridge: (a) finite element model;
and (b) infinite plate model.
140
Power Level (dB Re 10-12 Watt)

140
Power Level (dB Re 10-12 Watt)

130 130

120 120

110 110

100 Shell63 100 Kirchhoff

Shell181 Mindlin
Solid95
90 90
20 40 80 160 315 630 1250 20 40 80 160 315 630 1250
Frequency (Hz) Frequency (Hz)
(a) (b)
Figure 7. Power flow to the U-shaped bridge: (a) finite element model;
and (b) infinite plate model.

5
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

Figure 7 shows the bridge power obtained with different finite element models and infinite plate
models. It can be observed that the bridge power will be underestimated by up to 10 dB in the
frequency range from 315 Hz to 1250 Hz if the shear effect is not included.

4. Comparison and justification of the two methods

In this study, bridge vibration under a train pass-by was computed both in the time domain and
frequency domain. The two methods provide similar results in terms of the effect of shear deformation
on the bridge vibration, though they have some difference in the vibration amplitudes, as can be seen
in Figure 8. Overall, both the time domain and frequency domain method can be used to compute the
bridge vibration under a train pass-by with fairly good accuracy. And the Mindlin plate model which
includes shear deformation and rotation inertia effects can be used in the power flow analysis in the
medium-to-high frequency range above 200 Hz accurately and efficiently.
140
Power Level (dB Re 10-12 Watt)

135
130
125
120
115
110
105 Time Domain--Solid95
Frequency Domain--Solid95
100 Frequency Domain--Mindlin
95
40 80 160 315 630 1250
Frequency (Hz)
Figure 8. Power flow to the bridge obtained from different methods and models.

5. Conclusions
In this study, three finite element models have been used to calculate the bridge vibration during a
train pass-by in both the time domain and frequency domain. Two theoretical plate models have also
been adopted in the frequency domain analysis. The conclusions obtained from the calculated results
can be summarized as follows:
(1) Both of the time domain and frequency domain analysis method could provide reasonable
vibration results of bridge in the coupled train-track-bridge system. It is better to use the time
domain method in the low frequency range (below 200 Hz) for better accuracy and the
frequency domain method in medium-to-high frequency range for higher efficiency.
(2) The Mindlin plate model can give good results compared with the solid finite element model
but the former is more efficient in the modelling and computing procedures.
The simulated bridge power by the train-track-bridge interaction analysis can be used to predict the
bridge noise by combing it with the noise radiation models. The equivalent thickness of the infinite
plate for a real bridge should be further investigated in the future.

Acknowledgments
The authors are grateful for the financial support from the Natural Science Foundation of Shanghai
(No.15ZR1442800) and China Scholarship Council.

References
[1] Thompson D J 2009 Railway noise and vibration: mechanisms, modelling and means of control
(UK: Elsevier Science Ltd) p 6
[2] Li Q, Xu YL and Wu DJ 2012 Concrete bridge-borne low-frequency noise simulation based on
train-track-bridge dynamic interaction J. Sound Vib. 331 pp 2457-70
[3] Wu TX and Liu JH 2012 Sound emission comparisons between the box-section and U-section
concrete viaducts for elevated railway Noise Control Eng. J. 60 pp 450–57

6
MOVIC2016 & RASD2016 IOP Publishing
Journal of Physics: Conference Series 744 (2016) 012140 doi:10.1088/1742-6596/744/1/012140

[4] Zhang X, Li XZ, Li YD, et al 2013 Train-induced vibration and noise radiation of a prestressed
concrete box–girder Noise Control Eng. J. 61 pp 425–35
[5] Li Q, Song XD, Wu DJ 2014 A 2.5-dimensional method for the prediction of structure-borne
low-frequency noise from concrete rail transit bridges J. Acoust. Soc. Am. 135 pp 2718-26
[6] Li XZ, Zhang X, Zhang ZJ et al 2015 Experimental research on noise emanating from concrete
box-girder bridges on intercity railway lines P. I. Mech. Eng. F-J. Rai. 229 pp 125–35
[7] Zhang X, Li XZ, Zhang JQ, Song LZ, and Li YD 2015 A hybrid model for the prediction of
low–frequency noise emanating from a concrete box-girder railway bridge P. I. Mech. Eng. F-J.
Rai. DOI: 10.1177/095440971 5605127
[8] Li Q , Li WQ, Wu DJ, and Song XD 2016 A combined power flow and infinite element
approach to the simulation of medium-frequency noise radiated from bridges and rails. J. Sound
Vib. 365 pp 134-56
[9] Janssens MHA and Thompson DJ 1996 A calculation model for the noise from steel railway
bridge J. Sound Vib. 193 pp 295-305
[10] Li Q and Wu DJ 2013 Analysis of the dominant vibration frequencies of rail bridges for
structure-borne noise using a power flow method J. Sound Vib. 332 pp 4153-63
[11] Li Q, Xu YL, Wu DJ, and Chen ZW 2010 Computer-aided nonlinear vehicle-bridge interaction
analysis J. Vib. Control 16 pp 1791-816
[12] ISO 3095: 2005 Railway applications–acoustics–measurement of noise emitted by railbound
vehicles
[13] Cremer L, Heckl M, Petersson B A T 2005 Structure-borne Sound: Structural Vibrations and
Sound Radiation at Audio Frequencies (Berlin: Springer). p 283