See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266567940

Experimental validation of the dynamic interaction analysis between high-

speed trains and the Sesia viaduct

Article

CITATIONS READS

4 390

3 authors:

Kai Liu Edwin Reynders

Academy of Disaster Reduction and Emergency Management KU Leuven
41 PUBLICATIONS   509 CITATIONS    98 PUBLICATIONS   2,870 CITATIONS   

SEE PROFILE SEE PROFILE

Guido De Roeck
KU Leuven
350 PUBLICATIONS   10,511 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Robust vibration serviceability design of structures subjected to human-induced loading View project

Effect of human-human interaction on the human-induced vibrations of footbridges View project

All content following this page was uploaded by Kai Liu on 18 August 2015.

The user has requested enhancement of the downloaded file.

Experimental validation of the dynamic interaction analysis
between high-speed trains and the Sesia viaduct
K.Liu, E. Reynders and G. De Roeck

Katholieke Universiteit Leuven, Dept. of Civil Engineering, Kasteelpark Arenberg 40, B-3001 Leuven, Belgium

Abstract
The EC research project ”DETAILS” aims to remove uncertainties on dynamic interaction effects of composite
railway bridges. As part of the experimental part of the project, the Sesia viaduct, located on the new Italian
high-speed line between Torino and Milano, has been tested under the excitation of Italian ETR500Y high-speed
trains. This paper presents the finite element modeling of the Sesia viaduct, as well as the train-bridge interaction
model and its implementation. To facilitate the calculation, a user friendly train-bridge interaction toolbox DATIS has
been developed in Matlab that communicates with a finite element program. The toolbox is validated by comparing
calculated accelerations and strains with measured results.

1 Introduction

Steel-concrete composite bridges have been more and more exploited in new lines of the European high-speed rail-
way network due to considerable advantages regarding the design, construction time, durability, and costs. Never-
theless, these new design solutions, which have been introduced during a period of quick expansion of the network,
amplified open problems related to dynamic effects and interaction phenomena, fatigue loading, structural modeling,
fatigue life and damage assessment.
The EC research project ”DETAILS” aims to remove uncertainties on dynamic train-bridge interaction effects of
these types of bridges. As part of the experimental part of the project, the Italian Sesia viaduct has been tested
under ambient excitation and the excitation of Italian ETR500Y high-speed trains.
This paper presents the dynamic experiments conducted at the Sesia viaduct and the experimental validation of
the train-bridge interaction model.

2 Experiment on SESIA viaduct

2.1 Introduction to the bridge

The Sesia viaduct, located on the new Italian high-speed line between Torino and Milano, is a composite railway
bridge formed by 7 spans, 46 m each, for a total length of 322 m (Fig. 1). Each simply supported girder span has
the same double box cross section (Fig. 2). The bottom steel box is composed by lower flanges and three webs.
The concrete slab, of 13.6 m width and 0.4 m thickness, is connected to the steel girder by studs.

Figure 1: Global view of Sesia viaduct. Figure 2: Cross section of Sesia viaduct.
2.2 Measurement setup
Dynamic measurements on the second span from Torino to Milano side have been performed both under ambient
excitation and the excitation of Italian ETR500Y high-speed trains. Accelerations and strains at different points on
the bridge were measured. Also some sensors were placed on the first and the third span to evaluate the vibration
transmission between two neighboring spans. The accelerometers layouts are shown in Figure 3. All measurement
channels are denoted by a label xYzzF, where the number x denotes the number of the span (1, 2 or 3), the character
Y indicates the main girder (A, B or C) or stiffener (D or E), the number zz donates the number of the section (1 to
15), the character F donates the direction (X, Y or Z).

Figure 3: Sensors layouts (a) global vertical layout (b) global horizontal layout.

The local strain field in the girder longitudinal direction was measured with fiber optic sensors (SOFO from
SMARTEC). These sensors measure the relative displacement between two points, 1 m or 1.2 m apart. A total of 8
sensors were placed at four different levels on two vertical lines in order to measure the local strains in the bottom
and upper part of the cross section. These measurements were carried out on two cross sections: one next to the
pier and another at a quarter of the span (Fig. 4 and Fig. 5).

Figure 4: Vertical view of local strain layout. Figure 5: Local strain layout at a quarter span.

Figure 6: Installation of SOFO sensors.

 2.3 Modal analysis
A modal model of the bridge was identified using stochastic subspace identification(SSI) [1]. 8 frequencies were
extracted from the ambient vibration data (Tab. 1). According to the mode shapes of adjacent spans, the modes of
the bridge can be divided into symmetrical and anti-symmetrical ones (Fig. 7).

(a) (b)

Figure 7: Identified modes (a) symmetrical (b) anti-symmetrical

Modes Symmetrical Anti-symmetrical

Frequency Damping ratio Frequency Damping ratio
[Hz] ξ [Hz] ξ
First vertical bending mode 4.14 2.17% 3.62 1.70%
First torsional mode 9.00 1.84% 8.35 1.79%
Second vertical bending mode 10.44 2.64% 10.00 1.84%
Second torsional mode 14.28 1.69% 11.26 2.20%

Table 1: Identified eigenfrequencies and damping ratios from ambient tests.

50 50
40 40
30 30
20 20
4 3
2
3
2 10 1 10
1 0
Z

−1
Z

0 0
42 −10 42 −10
0
−2 X 0 X
−4 −20 −2
−4 −20
Y Y
(a) (b)

50 50
40 40
30 30
4
2 20 3 20
0 2
1
Z

−2 10 0 10
Z

−1
0 0
42 −10 5 −10
0
−2 X 0 X
−4 −20 −5 −20
(c) Y (d) Y

Figure 8: Identified mode shapes of symmetrical model (a) first vertical bending mode at 4.14 Hz (b)
first torsional mode at 9.00 Hz (c) second vertical bending mode at 10.44 Hz (d) second torsional mode
at 14.28 Hz.

From the modal analysis of the free vibration measurement, immediately after the train leaves the considered
second span, 3 modes were identified which correspond to the symmetrical modes identified from the ambient
vibration tests (only the second torsional mode was missed). By comparing the eigenfrequencies and mode shapes
identified from the ambient vibration tests and from the free vibration tests (Tab. 2), It can be concluded that the
symmetrical modes are predominately excited when the train passes the bridge.
From the system identification of a train passage, using the data when the train is still on the considered second
span of bridge, the first and the second bending modes which correspond to the symmetrical patterns (by comparing
mode shapes) were clearly identified (Tab. 3), which confirms that the symmetrical modes are more excited when
 Mode Free vibration Ambient vibration MAC
Frequency Damping ratio Frequency Damping ratio value
[Hz] ξ [Hz] ξ
First vertical bending mode 3.90 2.30% 4.14 2.17% 0.9881
First torsional mode 9.13 1.25% 9.00 1.84% 0.9620
Second vertical bending mode 10.41 2.38% 10.44 2.64% 0.9727

Table 2: Comparison of eigenfrequencies between ambient vibration tests and free vibration tests.

the train passes the bridge. The natural frequencies of the vehicle range from 0.26 Hz to 6.04 Hz, including a
frequency of 3.95 Hz which is close to the first bending frequency. This explains the decrease of the first bending
frequency while the second bending frequency reamains almost unchanged.

Mode Train forced vibration Ambient vibration MAC

Frequency Damping ratio Frequency Damping ratio value
[Hz] ξ [Hz] ξ
First vertical bending mode 3.67 2.81% 4.14 2.17% 0.9901
Second vertical bending mode 10.54 4.27% 10.44 2.64% 0.8316

Table 3: Comparison of eigenfrequencies between ambient vibration tests and train forced vibration
tests.

3 Numerical modelling of the Sesia viaduct

The numerical model (Fig. 9) of Sesia viaduct has been created with the finite element program Ansys. As described
in section 2.3, the symmetrical modes are predominately excited when the train passes, so the numerical model is
tuned according to the symmetrical modal properties derived from the ambient vibration measurements. The ap-
propriate boundary condition is applied in longitudinal direction of the rail and the ballast to simulate their continuity.
A track-ballast model (Fig. 10) is introduced to represent the track substructure. The sleepers are modeled as rigid
bodies with point mass. Linear springs and viscous dampers are used to simulate the vertical connections between
rail and sleepers [2]. The element types and material properties are illustrated in Table 4.

Figure 9: Numerical model of Sesia viaduct.

 Rail
Krs Crs
Sleepers
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa Ballast
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa Bridge
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

Figure 10: Track-ballast model.

Element type E ρ ν Mass Stiffness Damping

[N/m2 ] [kg/m3 ] [kg] [N/m] [N.s/m]
Steel Box shell 63 2.056E+011 7850 0.3 − − −
Ballast solid 45 3.2E+08 1750 0.28 − − −
Concrete Slab solid 45 3.3E+10 2500 0.17 − − −
Lateral panel solid 45 1.55E+08 2300 0.17 − − −
Rail beam 4 2.056E+011 7850 0.3 − − −
Sleepers mass21 − − − 290 − −
Krs combin14 − − − − 500E+06 −
Crs combin14 − − − − − 200E+03

Table 4: Element types and material characteristics of numerical model.

The calculated eigenfrequencies are compared to the frequencies of symmetrical modes identified from the
ambient tests (Tab. 5). There is a good correspondence between calculated and experimented results.

Mode Frequeny
[Hz]
experiment calculation
First vertical bending mode 4.14 4.15
First torsional mode 9.00 9.05
Second vertical bending mode 10.44 10.38
Second torsional mode 14.28 14.76

Table 5: Comparison of eigenfrequencies between measurements and numerical model.

4 The dynamic train-bridge interaction system

The dynamic model for the train-bridge interaction system is composed of the train subsystem and the bridge
subsystem. The two subsystems are linked by coupling the displacement relationships between rail and wheel.

4.1 Train system

The excitation train in the experiments is the Italian ETR500Y type. It is composed of a locomotive followed by 8
passenger cars and another locomotive. The average static axle loads for the locomotives and passenger cars are
176.4 kN and 112.9 kN respectively. The measured speed reached 288 km/h. The vehicle system is assumed
as following: each vehicle is considered as an independent entity with 1 car body, 2 bogies and 4 wheel sets [3].
Both the 1st and 2nd suspending system can be simplified as an elastic system. The bogies and the wheel sets
are linked by horizontal and vertical springs and dampers. There are horizontal and vertical springs (KH ,KV ) and
dampers (CH ,CV ) at each side of each wheel set, also horizontal and vertical springs (KHH ,KV V ) and dampers
(CHH ,CV V ) at each side of each bogie. So in a vehicle there are 8 horizontal and vertical springs and dampers in
the 1st suspending system, 4 horizontal and vertical springs and dampers in the 2nd suspending system. Each car
body and each bogie has 5 degrees of freedom (DOFS): the displacement in vertical direction (Z) and longitudinal
direction(Y ), and rotations around the X-axis(RX ), Y -axis(RY ) and Z-axis(RZ ). Each wheel set has 3 DOFs: the
displacement in veretical direction(Z) and longitudinal direction(Y ), and rotation around the X-axis(RX ). In this
way, a vehicle model of 27 DOFs is established, in which the DOFs of the wheel sets are linked to the movement of
the bridge and only the remaining 15 ones are taken as independent DOFs in vehicle equations. The model of the
vehicle system is shown in Figure 11.

Figure 11: Vehicle model.

The displacement of a vehicle is defined by the following vector:

{Y1 ,RX1 ,RZ1 ,Z1 ,RY 1 ,Y2 ,RX2 ,RZ2 ,Z2 ,RY 2 ,Y3 ,RY 3 ,Z3 ,RY 3 }
Where subscript 1 stands for the front bogie, subscript 2 stands for the car body and subscript 3 stands for the
rear bogie.
The equation of dynamic motion of the vehicle can be expressed as:
[MV ]{V̈V } + [CV ]{V̇V } + [KV ]{VV } = {PV } (1)
Where [MV ] , [CV ]and [KV ] are mass, damping and stiffness matrices, {VV }, {V̇V } and {V̈V } are the dis-
placement, velocity and acceleration vectors of the vehicle system; {PV } is the interaction force vector transferred
to the bogies by the first suspension system. The mass matrix of the vehicle system is a diagonal matrix, expressed
as:
[MV ] = diag[M1 , IX1 , IZ1 , M1 , IY 1 , M2 , IX2 , IZ2 , M2 , IY 2 , M1 , IX1 , IZ1 , M1 , IY 1 ] (2)
Where M1 is the mass of the bogie, IX1 , IY 1 , IZ1 are the mass moments of inertia of the bogie around the
X-axis, Y -axis and Z-axis; M2 is the mass of the car body, IX2 , IY 2 , IZ2 are the mass moments of inertia of car
body around the X-axis, Y -axis and Z-axis.
The stiffness matrix of the vehicle system is expressed as:
 H
sym

K11
V
 H K11
 
H

 K21 K 22

[KV ] =  V V
 (3)

 K21 K22 

 KH32 KH 33

KV32 KV33
where
 
4KH + 2KHH −4KH h3 + 2KHH h2 0
[KH
11 ] =  −4KH h3 + 2KHH h2 4KH h23 + 4KV b21 + 2KHH h22 + 2KV V b22 0 
0 0 4KH q02

 
4KV + 2KV V 0
[KV11 ] =
0 4KV q02
  
−2KHH −2KHH h2 0
[KH
21 ] =  2KHH h1 2KHH h1 h2 − 2KV V b22 0 
−2KHH s −2KHH sh2 0

 
4KHH −4KHH h1 0
[KH
22 ] =  −4KHH h1 4KHH h1 + 4KV V b22 0 
0 0 4KHH s2

 
−2KV V 0
[KV21 ] =
2KV V s 0

 
4KV V 0
[KV22 ] =
0 4KV V s2

 
−2KHH 2KHH h1 2KHH s
[KH
32 ] =  −2KHH h2 2KHH h1 h2 − 2KV V b22 2KHH sh2 
0 0 0

 
4KH + 2KHH −4KH h3 + 2KHH h2 0
[KH
33 ] =  −4KH h3 + 2KH h2 4KH h23 + 4KV b21 + 2KHH h22 + 2KV V b22 0 
0 0 4KH q02

 
−2KV V −2KV V s
[KV32 ] =
0 0

 
4KV + 2KV V 0
[KV33 ] =
0 4KV q02
Where KV and KH are the vertical and horizontal spring stiffness coefficients at each side of the wheel set
in the first suspension system, KV V and KHH are the vertical and horizontal spring stiffness at each side of the
bogie in the second suspension system, q0 is half of the distance between the wheel sets, s is half of the distance
between the bogies, b1 and b2 are the distance between the springs of the first and the second suspension systems,
h1 is the distance between the car body and the second suspension system, h2 is the distance between the second
suspension system to the bogie and h3 is the distance between the bogie and the wheel set.
The damping matrix can be derived from the stiffness matrix just by replacing stiffness coefficients(KV , KH ,
KV V , KHH ) by damping coefficients (CV , CH , CV V , CHH ), in which CV and CH are vertical and horizontal damping
coefficients at each side of bogie in the first suspension system, CV V and CHH are vertical and horizontal damping
coefficients at each side of car body in the second suspension system.
For the DOFs of the front bogie and the rear bogie, the interaction force {PV }={PV 1 ,PV 2 ,. . . ,PV 15 } transferred
to the bogies by the first suspension system can be expressed in terms of the displacements and the velocities of
the wheel sets:



 PV 1 = 2CH (ẎW 1 + ẎW 2 ) + 2KH (YW 1 + YW 2 )
PV 2 = 2CH [b21 (ṘXW 1 + ṘXW 2 ) − h3 (ẎW 1 + ẎW 2 )] + 2KH [b21 (RXW 1 + RXW 2 ) − h3 (YW 1 + YW 2 )]




PV 3 = 2CH q0 (ẎW 1 − ẎW 2 ) + 2KH q0 (YW 1 − YW 2 )








 PV 4 = 2CV (ŻW 1 + ŻW 2 ) + 2KV (ZW 1 + ZW 2 )
PV 5 = 2CV q0 (−ŻW 1 + ŻW 2 ) + 2KV (ZW 1 + ZW 2 )

(4)

 PV 11 = 2CH (ẎW 3 + ẎW 4 ) + 2KH (YW 3 + YW 4 )
PV 12 = 2CH [b21 (ṘXW 3 + ṘXW 4 ) − h3 (ẎW 3 + ẎW 4 )] + 2KH [b21 (RXW 3 + RXW 4 ) − h3 (YW 3 + YW 4 )]








 PV 13 = 2CH q0 (ẎW 3 − ẎW 4 ) + 2KH q0 (YW 3 − YW 4 )
PV 14 = 2CV (ŻW 3 + ŻW 4 ) + 2KV (ZW 3 + ZW 4 )




PV 15 = 2CV q0 (−ŻW 3 + ŻW 4 ) + 2KV (ZW 3 + ZW 4 )


For the DOFs of the car body:

PV 6 = PV 7 = PV 8 = PV 9 = PV 10 = 0 (5)
4.2 Bridge system
The equation of dynamic motion of the bridge can be expressed as:

[Mb ]{V̈b } + [Cb ]{V̇b } + [Kb ]{Vb } = {Pb } (6)

In the present numerical model developed for the Sesia viaduct, more than 8000 nodes and 11000 elements are
included. Therefore, the Modal Superposition Method is adopted to solve the motion equations of the bridge. It is
assumed that only the first N0 modes of bridge are contributing to the interaction computation and the modal data
are normalized to the mass matrix:
ΦT [Mb ]Φ = 1 (7)
Thus the mass matrix of the bridge subsystem in modal coordinates is a N0 -order unit matrix, the damping and
stiffness matrices of the bridge subsystem are diagonal matrices:

1
 

 1 

∗
[Mb ] = 
 1 
 (8)
 .. 
 . 
1

2ε1 ω1
 

 2ε2 ω2 

[C∗b ] =
 2ε3 ω3 
 (9)
 .. 
 . 
2εN0 ωN0
ω12
 
 ω22 
ω32
 
[K∗b ] = (10)
 

 .. 
 . 
2
ωN 0

Where ωi and εi are the frequency and the damping ratio of the ith mode.
Thus the motion equations of the bridge can be expressed as:

q̈ + C∗b q̇ + K∗b q = P∗b (11)

Where q represents the displacement in modal coordinates. P∗b is the normalized force. It can be expressed as:

P∗b = ΦT Pb (12)

Where Pb represents the force vector caused by the wheel-rail interaction. It is determined by position, move-
ment status and mass of the wheel sets. The horizontal, torsional and vertical forces produced by wheel set i of
bogie j corresponding to the deck displacement can be deduced in terms of the equilibrium displacement of the
wheel:

 FY W 1 = −MW 1 ŸW 1 + 2CH [Ẏ1 + q ṘZ1 − h3 ṘX1 − ẎW 1 ] + 2KH [Y1 + qRZ1 − h3 RX1 − YW 1 ]
FZW 1 = −MW 1 Z̈W 1 + 2CV [Ż1 + qRY 1 − ZW 1 ] + 2CV [Z1 + qRY 1 − ZW 1 ] + FG1 (13)
FRXW 1 = −IW 1 R̈XW 1 + 2CV Db1 (RX1 − RXW 1 ) + 2KV Db1 (−RX1 + RXW 1 )


 FY W 2 = −MW 2 ŸW 2 + 2CH [Ẏ1 + q ṘZ1 − h3 ṘX1 − ẎW 2 ] + 2KH [Y1 + qRZ1 − h3 RX1 − YW 2 ]
FZW 2 = −MW 2 Z̈W 2 + 2CV [Ż1 + qRY 1 − ZW 2 ] + 2CV [Z1 + qRY 1 − ZW 2 ] + FG2 (14)
FRXW 2 = −IW 2 R̈XW 2 + 2CV Db1 (RX1 − RXW 2 ) + 2KV Db1 (−RX1 + RXW 2 )


 FY W 3 = −MW 3 ŸW 3 + 2CH [Ẏ3 + q ṘZ3 − h3 ṘX3 − ẎW 3 ] + 2KH [Y3 + qRZ3 − h3 RX3 − YW 3 ]
FZW 3 = −MW 3 Z̈W 3 + 2CV [Ż3 + qRY 3 − ZW 3 ] + 2CV [Z3 + qRY 1 − ZW 3 ] + FG3 (15)
FRXW 3 = −IW 3 R̈XW 3 + 2CV Db1 (RX3 − RXW 3 ) + 2KV Db1 (−RX3 + RXW 3 )


 FY W 4 = −MW 3 ŸW 4 + 2CH [Ẏ3 + q ṘZ3 − h3 ṘX3 − ẎW 4 ] + 2KH [Y3 + qRZ3 − h3 RX3 − YW 4 ]
FZW 4 = −MW 3 Z̈W 4 + 2CV [Ż3 + qRY 3 − ZW 4 ] + 2CV [Z3 + qRY 1 − ZW 4 ] + FG4 (16)
FRXW 4 = −IW 3 R̈XW 4 + 2CV Db1 (RX3 − RXW 4 ) + 2KV Db1 (−RX3 + RXW 4 )


Where FY W i , FZW i , FRXW i (i=1,2,3,4) represent the forces and moment under the ith wheel set. MW is the
mass of the wheel set, IW is the inertia around the X-axis of the wheel set, FG is the vehicle weight on the wheel, D
 N1 Fi N2
d1 d2

Figure 12: Forces transfer to adjacent nodes.

is the gauge of the rail. The forces caused by wheel-rail force between nodes are transferred to the two neighboring
rail nodes (Fig. 12).
 
α1 0 0
  0 α1 0 
   

 0 FY i
FiN1 0 α1  

= α2 0 ∗ FZi  (17)
FiN2 0 

 0 α2 0 
 FRXi
0 0 α2
Where N1 and N2 are the neighboring rail nodes of the ith wheel position, FiN1 , FiN2 are the vectors of forces
transferred to the nodes N 1 and N 2 respectively, FY i , FZi , FRXi are the forces under the ith wheel. α1 = d1d+d
2
2
,
d1 th
α2 = d1 +d2 , where d1 and d2 are the distances between the i wheel position and the neighboring rail nodes N1
and N2 .
The displacements of the bridge can be calculated by Equation 18.

υb = Φq (18)

In this paper, the wheel hunting and the track irregularities are not yet considered. The displacement of wheel
set i under bogie j is taken equal to that of the same position on the structural member on which the vehicle runs.

4.3 Computation of the coupled train-bridge system

The vehicle system and the bridge system are linked by coupling the displacement relationships between wheel and
rail. As described before, the dynamic equations of the vehicle system and the bridge system can be expressed as:

MV ẌV + CV ẊV + KV XV = PV
(19)
q̈ + C∗b q̇ + K∗b q = P∗b

The dynamic analysis of the bridge response due to the forces transmitted by the wheel is alternated with the
dynamic analysis of the vehicle motion due to the bridge displacements. The integration of the motion equations is
performed according to the Newmark − β method:

Ẋi = Ẋi−1 + [(1 − γ)Ẍ + γ Ẍi ]∆t
(20)
Xi = Xi−1 + Ẋi−1 ∆t + [( 12 − β)Ẍi−1 + β Ẍi ]∆t2
Where β=0.25 and γ=0.5 according to the trapezoidal rule, which is stable for any time increment.
A Matlab toolbox DATIS(Dynamic Analysis of Train-bridge Interaction System) is developed for the train-bridge
interaction computation. To facilitate the input, a GUI is provided as shown in Figure 13.

5 Numerical results and comparison with experiments

According to [4], for the determination of the maximum deck acceleration, the frequencies in the dynamic analysis
should be considered up to a maximum of : i)30Hz; ii)1.5 times the frequency of the first mode shape of the
structural element being considered, including at least the first three modes. For the Sesia viaduct, the first mode is
the first lateral mode with a frequency of 2.7885 Hz, therefore these two frequencies are 30 Hz and 4.2 Hz. Thus
the threshold of low-pass filtering is 30 Hz. The first 300 modes, with natural frequencies ranging from 2.7885 Hz
to 41.185 Hz, were used for train-bridge interaction computation, with a uniform damping ratio equal to 2.5% for all
the modes. The speed of the train is set to 288 km/h. Figure 14 shows the computed vertical displacement history
of mid-span point 2b08 (see Fig. 3), the maximum displacement reaches 1.9 mm.
Figure 15 compares the experimental and computed time history and frequency content of the vertical acceler-
ation at 2c12 (see Fig. 3) after a filtering to 30 Hz. A good agreement between the experimental and numerical
results is observed. The acceleration has approximately the same maximum value in time domain. Furthermore,
it can be observed both the experimental and numerical results show a dominant frequency of about 4 Hz, which
corresponds to the first bending frequency of the bridge.
 Figure 13: Matlab GUI for train-bridge interaction computation.

0.5

0
Displacement [mm]

−0.5

−1

−1.5

−2
0 2 4 6 8 10
Time [s]

Figure 14: Computed displacement time history at 2b08.

Figure 16 compares the one-third octave band RMS values of the acceleration at 2c12. There is a good agree-
ment between the predicted and the measured results.
Figure 17 compares the prediction and measurement for strain at S16 (see Fig. 5). The comparison shows
a convincing agreement. Both the prediction and the measurement almost have the same magnitude and the
contribution of the axles can be clearly distinguished. These strain results are very meaningful for fatigue tests and
for fatigue analysis.
Figure 18 shows the predicted maximum displacement of mid-span point 2b08 under varying train speeds.
Computations were made in the the interval [250,400] km/h with a step of 5 km/h. The vertical displacement
response is represented in Figure 18. The resonance speed is reached at 380 km/h.
 0.3
0.5

Acceleration [m/s /Hz]

0.25
Acceleration [m/s ] 0.2
2

2
0

0.1
−0.25

−0.5 0
0 2 4 6 8 0 5 10 15 20 25 30
(a) Time [s] (b) Frequency [Hz]

Figure 15: Comparison of acceleration at 2c12 between prediction(red) and measurement(black) (a)
time history (b) frequency content.

0
10
Acceleration [m/s ]

−1
2

10

−2
10

−3
10 0 1
10 10
Time [s]

Figure 16: Comparison of one-third octave band RMS spectra between predicted(red) and mea-
sured(black) acceleration at 2c12 in frequency domain.

40

30
Strain [10 6]

20
−

10

−10
0 2 4 6 8 10
Time [s]

Figure 17: Comparison of predicted strain histories(red) with measured ones(black) at S16.

The passage of successive loads with uniform spacing can excite the structure in resonance when critical speed
 2.4

2.3

Displacement [mm]
2.2

2.1

1.9

1.8
250 275 300 325 350 375 400
Speed [km/h]

Figure 18: Maximum displacement at 2b08 under varying train speeds.

is reached, which can be computed as follows [5]:

l0
Vcri = × fn , i = 1, 2, 3, · · · , n (21)
i
For Italian high speed train ETR500Y, l0 =26.1 m, i=1, and f1 =4.15 Hz, The critical speed is 389 km/h, which
matches well with the train-bridge interaction analysis.

6 Conclusions

The main objective of the work presented in this paper is to present the dynamic experiments of the Sesia viaduct
and the developed tools for estimating the vibrations due to train passages. From these measurements and the
modeling application, it can be concluded that using proper equipment and suitable layouts of sensors, the dynamic
behavior of bridges can be assessed by vibration measurements, opening the way for continuous health monitoring.
This study also allows a better understanding of the train-bridge interaction phenomenon. The developed Matlab
toolbox DATIS for train-bridge interaction calculation has been validated by comparing predictions with measure-
ments.

References

[1] B. Peeters and G. De Roeck. Reference-based stochastic subspace identification for output-only modal analysis.
Mechanical Systems and Signal Processing, 13:855–878, 1999.
[2] C.Rebelo C.Rigueiro and L.S.Silva. Vibration of the railway track-viaduct system under moving vehicles taking
into account the interaction effect. In Proceedings of ISMA, pages 1233–1247, Leuven, 2006.
[3] N. Zhang H. Xia and G. De Roeck. Dynamic analysis of high speed railway bridge under articulated trains.
Computers Structures, 81:2467–2478, 2003.
[4] En1990-prannex A2, basis of structural design-annex A2: Application for bridges (normative), final draft. Euro-
pean committee for standardization. Cen, 2002.
[5] L. Fryba. Vibration of solids and structures under moving loads. Thomas Telford, 1999.