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

net/publication/255701699

The Dynamics of High-Speed Railway Bridges: A Review of Design Issues and

New Research for Lateral Dynamics

Article in International Journal of Railway Technology · April 2012

DOI: 10.4203/ijrt.1.1

CITATIONS READS

13 4,557

2 authors:

José M Goicolea Pablo Antolín

Technical University of Madrid / UPM Universidad Politécnica de Madrid École Polytechnique Fédérale de Lausanne
126 PUBLICATIONS 1,218 CITATIONS 68 PUBLICATIONS 871 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by José M Goicolea on 18 April 2014.

The user has requested enhancement of the downloaded file.

 Dynamics of High-Speed Railway Bridges: Review of Design Issues
and New research for Lateral Dynamics
International Journal of Railway Technology,
1(1), 27-55, 2012. doi:10.4203/ijrt.1.1.2
J.M. Goicolea and P. Antolin
School of Civil Engineering, Technical University of Madrid, Spain.

Keywords: dynamics, bridges, impact factor, resonance, vehicle-bridge interaction, lat-

eral dynamics, wheel-rail contact, nonlinear.

Summary
The dynamic effects of high-speed trains on viaducts are important issues for the design
of the structures, as well as for the consideration of safe running conditions of the trains.
In this work we start by reviewing the relevance of some basic design aspects. The signif-
icance of impact factor envelopes for moving loads is considered first. Resonance which
may be achieved for high-speed trains requires dynamic analysis, for which some key
aspects are discussed. The relevance of performing a longitudinal distribution of axle
loads, the number of modes taken in analysis, and the consideration of vehicle-structure
interaction are discussed with representative examples.
The lateral dynamic effects of running trains on bridges is of importance for laterally
compliant viaducts, such as some very tall structures erected in new high-speed lines.
The relevance of this study is mainly for the safety of the traffic, considering both internal
actions such as the hunting motion as well as external actions such as wind or earthquakes
[1]. These studies require three-dimensional dynamic coupled vehicle-bridge models, and
consideration of wheel to rail contact, a phenomenon which is complex and costly to
model in detail. We describe here a fully nonlinear coupled model, described in absolute
coordinates and incorporated into a commercial finite element framework [2]. The wheel-
rail contact has been considered using a FastSim algorithm which provides a compromise
between accuracy and computational cost, and captures the main nonlinear response of
the contact interface. Two applications are presented, firstly to a vehicle subject to a
strong wind gust traversing a bridge, showing the relevance of the nonlinear wheel-rail
contact model as well as the dynamic interaction between bridge and vehicle. The second
application is to a real HS viaduct with a long continuous deck and tall piers and high
lateral compliance [3]. The results show the safety of the traffic as well as the importance
of considering features such as track alignment irregularities.

References
[1] W. Guo, H. Xia, Y. Xu, “Running safety analysis of a train on the Tsing Ma Bridge
under turbulent winds”, Earthquake Engineering and Engineering Vibration, 9(3):
307–318, 2010.
[2] P. Antolin, J.M. Goicolea, J. Oliva, M.A. Astiz, “Nonlinear train-bridge lateral in-
teraction using a simplified wheel-rail contact method within a finite element frame-
work”, Journal of Computational and Nonlinear Dynamics, 2012, In press.
[3] F. Millanes, J. Pascual, M. Ortega, “’Arroyo de las Piedras’ viaduct: The first com-
posite steelconcrete high speed railway bridge in Spain”, Structural Engineering
International, 17(4): 292–297, 2007.
Abstract

The dynamic effects of high-speed trains on viaducts are important issues for the de-
sign of the structures, and also for the consideration of safe running conditions of the
trains. In this work we start by reviewing the relevance of some basic design issues.
The significance of impact factor envelopes for moving loads is considered first. Res-
onance which may be achieved for high-speed trains requires dynamic analysis, for
which some key aspects are discussed. The relevance of performing a longitudinal
distribution of axle loads, the number of modes taken in analysis, and the considera-
tion of vehicle-structure interaction are discussed with representative examples.
The study of lateral dynamics of running trains on bridges is of importance mainly
for the safety of the traffic, and may be relevant for laterally compliant bridges. These
studies require 3D coupled vehicle-bridge models, and consideration of wheel to rail
contact, a phenomenon which is complex and costly to model in detail. We describe
here a fully nonlinear coupled model, described in absolute coordinates and incorpo-
rated into a commercial finite element framework. Two applications are presented,
firstly to a vehicle subject to a strong wind gust traversing a bridge, showing the rel-
evance of the nonlinear wheel-rail contact model as well as the interaction between
bridge and vehicle. The second application is to a real HS viaduct with a long con-
tinuous deck and tall piers with high lateral compliance. The results show the safety
of the traffic as well as the importance of considering features such as track alignment
irregularities.

Keywords: dynamics, bridges, impact factor, resonance, vehicle-bridge interaction,

lateral dynamics, wheel-rail contact, nonlinear.
1 Introduction

Traffic loads on railway bridges are a major action to be considered for evaluating the
safety and functionality of structures. Not only the traffic loads are larger than for
road bridges, they also may produce significant dynamic effects due to their regular
spacing. Furthermore, the nature of the guided traffic on rails imposes more stringent
safety requirements than for road vehicles, resulting in limitations to deck accelera-
tions and deformations. With the advent of faster trains in new high-speed railways
and increased demands on the infrastructure the relevance of these dynamic effects
has become one of the key design factors for railway bridges.
In this work we start by reviewing some of the basic features related to the dynamic
response of bridges under traffic loads (section 2). Some of these features are key
aspects for the design and functionality of structures in the new high speed railway
lines, and have been incorporated into the recent engineering codes [3, 4, 7, 35]. The
discussion here has a conceptual approach, hence the models employed will be simple
in order to help in understanding the basic design issues and modelling options.
The basic solution of a moving load on a simply supported bridge [33] involves ne-
glecting the non-suspended mass (wheelsets) as well as the vibrations of the suspended
masses, a sufficiently approximate solution for many cases. It is easily obtained that
the maximum dynamic increment over static effects for a moving load is '0 = 77%,
using standard notation for railway structures engineering [34]. However significant,
this increase may be easily bounded by a factor in the design codes, the so-called im-
pact factor. This has been the basis of the approach followed in engineering codes up
to very recently [34], until high-speed trains have arrived.
The new high-speed trains introduce a potentially much greater dynamic effect,
the resonant response of the bridge from regularly spaced axle loads at speeds whose
effective frequency may coincide with the fundamental frequencies of the structure.
Resonance is not adequately covered by an impact coefficient and requires a dynamic
analysis of the bridge.
As a result of research carried out in Europe to investigate high speed traffic ac-
tions [11] the new codes for design of railway bridges take into account resonant
phenomena from traffic, in the Eurocodes [3], UIC code [35] or Spanish code [7]). In
addition to conservative static load models (LM71) and impact factor envelopes ( )
they prescribe dynamic analyses to check resonance under certain circumstances. Fur-
thermore they define a High Speed Load Model (HSLM) which provides a dynamic
envelope for European high-speed trains and enables interoperability within the trans-
European network [13].
In general, it is unlikely that the structural Ultimate Limit States (ULS) would be
reached from dynamic traffic actions in bridges designed according to modern stan-
dards. More often, the critical issues are Service Limit States (SLS) [21] such as the
maximum vertical accelerations of the bridge deck. This limit for ballasted track is
amax = 3.5 m/s2 [4], in order to avoid risk of destabilization of the ballast and unac-
ceptable safety risks to the railway traffic. High levels of acceleration are generally
observed in short span bridges, due to two reasons: the low overall mass of the bridge,
and the fact that the resonant action of the axles (bogies) may be more pronounced for
span lengths shorter than the vehicles.
In section 3 this paper deals with lateral dynamics of vehicles on viaducts. This
proves to be a phenomenon which may be significant in some cases and for which a
coupled full vehicle–structure interaction model must be employed. The interest for
this study originates from the observation of significant lateral vibrations in some Eu-
ropean railway bridges, with metallic open deck sections, high lateral compliance and
consequently low lateral eigenfrequencies. These vibrations affect the train as well as
the structure, and were studied under the auspices of UIC and ERRI [10]. The con-
clusions of this work have resulted in design limits for railway bridges incorporated
in the structural Eurocodes and UIC codes [4, 35].
A different type of railway structures with high lateral compliance in relative terms
are the long continuous viaducts with high piers erected in some high-speed railway
lines. Several such viaducts form part of the new Spanish HS lines [20]. In principle
these viaducts are not affected by the conclusions of the ERRI studies cited above;
however, they also have low frequencies for lateral vibration, but associated in this
case with modes of very long wavelengths. In spite of the above, some uncertainties
remain about the lateral response of these viaducts, being one of the objects of the
present work.
Lateral oscillations may arise from self-excited causes: track alignment irregular-
ities, nosing motion of the wheelsets, deck torsion specially on double track decks,
lateral bending of piers, or centrifugal forces. Another obvious origin is external ac-
tions, mainly cross winds and earthquakes. It becomes important to evaluate the safe
running conditions under these circumstances, such as the characteristic wind curves
(CWC) [14].
In this work we propose coupled train–bridge dynamic interaction models for eval-
uating the running safety of trains over railway viaducts. The models for these prob-
lems include the following ingredients [39]: 1) dynamic model for structure subsys-
tem; 2) dynamic model for the vehicle subsystem; 3) description of track irregularity
profiles; 4) wheel-rail contact models; and 5) adequate numerical solution algorithms
for the equations.
Most of the existing approaches to this problem assume linear models both for
the structure subsystem and also for the vehicle multibody system [30, 36–40]. In
these models quadratic velocity terms, such as gyroscopic effects, are also neglected.
Some approaches which consider nonlinear effects for vehicle and structures have
been proposed by [31, 32], and in [22], where a linear model is used for the vehicle
with additional gyroscopic terms included for the wheelsets.
In this work a fully nonlinear coupled model is proposed for vehicle-structure ver-
tical and lateral dynamic models. Vehicles are considered as fully three-dimensional
multibody systems, and the bridge structure is modeled by means of finite elements.
The model is developed in a general and modular and it may be easily implemented
within an existing finite element analysis software with multibody capabilities. For
this work Abaqus [28] finite element framework has been used. Both subsystems
(bridge and vehicles) are described with coordinates in absolute reference frames, as
opposed to alternative approaches which describe the multibody system with coor-
dinates relative to the base bridge motion. This facilitates the full consideration of
nonlinear inertia terms, without introducing additional difficulties for the structural
mechanical behavior [27].
Contrary to the majority of existing models for train–bridge dynamic interaction,
the formulation described here is capable of full consideration of geometrical and ma-
terial nonlinearities both in the structural subsystem (bridge) and in the multibody
subsystem (vehicle). The approach for wheel–rail geometrical interaction and me-
chanical contact model is fully nonlinear as well, not being limited neither to constant
conicity assumptions nor to linearized elastic contact forces. The particular model
reported in this work involves some simplifying assumptions that will be described
below.
In the rest of this paper we discuss first the dynamic response of bridges in section 2,
reviewing the “impact” action of moving loads, resonance, and models available for
dynamic analysis. Following, in section 3 we describe the models for lateral dynamics
of vehicles on viaducts. Two representative applications will be presented in section 4.
Finally, some concluding remarks are summarised in section 5.

2 Dynamic response of bridges

2.1 Dynamic response to moving load
The solution to a moving load on a simply supported bridge is well established and
available in closed form under certain assumptions. However, it is interesting to review
for several reasons. Firstly it provides a basis for defining a dynamic factor (or impact
factor) for design. Additionally, the closed form solution helps to identify clearly the
characteristics of the dynamic behavior.
From the dynamic equation of vibration of a beam, the solution may be performed
with a modal analysis [33], in which we shall take only the fundamental mode of
vibration of frequency f0 = !0 /2⇡ (further down we shall consider the implications
of taking more modes). For a load at constant speed v the wavelength is defined as
= v/f0 . Additionally, for a bridge of span L a non-dimensional parameter ↵ for the
load velocity may be defined as:
v
↵= = . (1)
2L 2f0 L
The displacement time history response at the center of the span, in terms of the
maximum static response ys = P L3 /48EI ⇡ 2P L3 /⇡ 4 EI and considering some
simplifications valid for small damping (⇣ ⌧ 1) is:
ys ⇥ ⇤
y (t) = sin (↵! 0 t) ↵e ⇣!0 t
sin (!0 t) , (2)
1 ↵2
where the first term within the brackets is due to the excitation from the external load
and the second to the free vibration of the bridge.
Figure 1 shows an application for a L = 15 m simply supported beam-type bridge.
The computed dynamic increment at 220 km/h, with 2% structural damping, is '0 =
59%. The critical velocity, for obtaining a maximum dynamic increment, would be
in this case 333 km/h, corresponding to an increase of '0 = 77% with no damping.
In real bridges, additional dynamic effects must be considered from the irregularities
of tracks and wheels (represented by parameter '00 in UIC standard railway structures
engineering notation [34]). However, these are generally of less importance for the
bridges: in this case they would only amount up to an additional 0.5'00 = 2% dynamic
increment for a well maintained track [34].

v= 220 km/h

Vertical displacement at centre of span (mm)

3
Load exits bridge
2

−1

P v −2
δsta

ϕ’δsta (dynamic increment)

−3

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Time (s)

Figure 1: Dynamic increment for moving load on simply supported L = 15 m bridge,

from catalog of ERRI D214 [11], with v = 220 km/h and damping ⇣ = 0.02.

The solution (2) is valid during the time the load is on the bridge, under this as-
sumption the maximum ydyn in terms of t may be computed (for very fast moving
loads when the maximum is reached after the load exits the bridge the response is
lower). For the most unfavourable case without damping (⇣ = 0) the maximum is
attained for ẏ= 0 ) !0 t = 12n↵ ⇡, with the result
 ✓ ◆ ✓ ◆
ydyn 1 ↵ 1
= sin 2⇡ ↵ sin 2⇡ . (3)
ys 1 ↵2 1 ↵ 1 ↵
This expression yields an envelope of the dynamic factor with respect to the non-
dimensional parameter ↵, plotted in figure 2. This envelope curve shows a maximum
response for a critical value of ↵c (and associated critical speed, vc = 2↵c f0 L):
✓ ◆
ydyn
↵c = 0.617 ) = (1+'0 dyn )max = 1.768. (4)
ys max
In figure 2 the code envelope '0 UIC is also plotted for comparison, which proves to be
sufficiently conservative.
As an example, for the bridges in the D214 ERRI study [11] the maxima of '0 dyn
correspond to vc = 333 km/h for the L = 15 m bridge and vc = 356 km/h for
L = 20 m, velocities which may be attained by modern high-speed trains.
 moving load on simply supported beam
2.4
1+ϕ’dyn
1+ϕ’UIC
2.2
max (0.760, 2.325)
impact factor (1+ϕ’)
2

1.8

1.6

1.4 max (0.617, 1.768)

1.2

1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
α=v/(2f0 L)

Figure 2: Envelopes of impact coefficient; '0 dyn from analytic solution (3) and '0 UIC
from [35]. The various lobes in '0 dyn correspond to slow loads for which the beam
performs more than one full oscillation during passage.

2.2 Dynamic analysis with moving loads

Resonance in bridges.— In order to motivate the need for more complete dynamic
analysis, figure 3 shows some measured results for a bridge in the Madrid-Sevilla HS
line, from an AVE S100 (ALSTHOM) train at 220 km/h. The bridge consists of a

Figure 3: Measured vertical displacement at center of simply-supported span in

viaduct over Tajo river, Madrid-Sevilla high-speed line, together with results of simu-
lation with moving load model [9]. AVE S-100 single unit train at 220 km/h

sequence of simply supported spans with L = 38 m. The measured and computed

results show an impact coefficient of approximately (1 + '0 ) = 2.0, measured with
respect to the highest static effects due to the locomotive loads. The effect would
have been even greater for a double unit train (approx. 400 m length). This dynamic
 v

P1 P2 P3 P4 P5 P6 P7

Figure 4: Load sequence from HS train for moving load dynamic analysis, showing
AVE S-102 (Talgo) HS train with regularly spaced single axles.

effect is not so much a problem for the Ultimate Limit State (ULS) of the bridge,
which in this case is covered by the safety margins embedded in the normative static
vertical load envelope LM71 and the impact coefficient employed for the design [3].
However, in this case the functionality of the bridge was impaired: vibrations induced
in the catenary posts proved to be excessive and these had to be relocated into new
positions. In other cases the resonant dynamic effects may be even of much greater
magnitude, and must be therefore avoided in the design of bridges.

Models for analysis.— The Impact factor is derived from moving load envelopes
for analysis of real trains at conventional speeds (i.e.  200 km/h). It does not cover
resonant effects, which may be much greater than moving load effects; in order to
consider resonance an impact factor approach would be either unsafe or excessively
conservative. For circumstances in which there is a possibility of resonance it is neces-
sary to perform a dynamic analysis of the whole train taking into account the complete
load sequence (figure 4). In order to discuss these concepts, we shall consider in what
follows models for the simplest case, a straight beam subject only to vertical bending,
for which the differential equation governing the dynamics is
N
X
00 00
mü + (EI u ) = p (x, t) = Pk h (x + dk vt)i , (5)
k=1

for a train with N concentrated axle loads Pk with offsets dk , where x is the longitu-
dinal coordinate, u(x) the beam vertical displacements, m the mass per unit length,
and (·) is the Dirac delta function. The brackets h·i have the meaning h (⇠)i = (⇠)
if 0 < ⇠ < L (load within bridge) or 0 otherwise. Superposed dots (•) ˙ represent
time derivatives and primes (•0 ) derivatives with respect to x. This equation may
be generalized for arbitrary structures either with 3D beams, torsional effects, shear
deformation, or more general continuum-type descriptions. In particular, torsion is
added straightforwardly to the above equation.
The solution to equation (5) is obtained in two steps, first in space (x) and then in
time (t). For simple cases, such as the simply supported beam, the spatial solution
can be performed analytically through modal analysis, obtaining uncoupled modal
equations for the amplitude of vibration of each mode [6]. Considering mode shapes
i (x) and associated circular frequencies !i :

N
X
Mi ÿi +2⇣i !i Mi ẏi + !i2 Mi yi = Pk h i (x + dk vt)i, (6)
k=1

where yi is the amplitude for mode i , Mi is the corresponding modal mass and ⇣i
the damping ratio. These equations may be integrated in time by direct numerical
algorithms (either coded directly or available within finite element software).
A more general procedure is to employ finite element (FE) software for the dis-
cretization of the dynamic equations in space, enabling solution of fully 3D problems.
The only feature which is special for these dynamic problem, as compared to other
structural dynamics problems, is the adequate definition of the actions from the mov-
ing loads, which needs an ad-hoc preprocessing. As a result of FE discretization the
following matrix system of ordinary differential equations is obtained
Mü + Cu̇ + Ku = f (t) , (7)
where M, C and K are respectively the mass, damping and stiffness matrices and f (t)
the load vector obtained from the moving loads. At this point within FE software it
may be chosen to do a direct time integration of the coupled equations (7) with a nu-
merical scheme, or to perform a modal analysis of the discretised system and obtain
numerical mode shapes and frequencies. These will then be available as uncoupled
equations identical to (6) and may be integrated in time individually. The modal anal-
ysis option has several advantages. The number of modes to consider can be chosen
thus avoiding high-frequency components from higher modes, which are not signif-
icant for the bridge response. Moreover, the solution is generally much faster. The
alternative approach of performing a direct time integration of the complete system
provides a more general method, which may be necessary in some cases, for instance
to consider nonlinear effects such as contacts.

Example with enhanced impact.— Following we discuss an application, solved

with a dynamic moving load analysis. It serves as an example of a case with an
enhanced impact effect with respect to the single moving load solution in section 2.1,
due to a phenomenon which may be termed “repeated impact”. Figure 5 shows the
case of an articulated multiple unit train (MU) [29] at the critical moving load velocity
(↵c = 0.617, vc = 353 km/h), together with the quasi-static response neglecting
inertial vibration. The impact factor obtained is (1 + '0 ) = 2.45, clearly larger than
that for a moving load (figure 2). We remark the response in this case is not a case of
resonance, but does represent a case for which the simple moving load coefficient is
not sufficient.

Longitudinal distribution of loads.— The axle loads from railway traffic are trans-
mitted by the track structure, which will perform a longitudinal distribution of these
  

           

   

5
quasi-static
moving loads
0
[mm]

0 10 20 30 40 50 60 70 80 90 100 110
x [m]

Figure 5: Response for v = 353 km/h load model for articulated MU AB 2D [29],
5 coach unit with individual axle loads P = 21.5 t; vertical displacements at mid-
span for L = 20 m bridge from ERRI D214 [11], considering moving loads; The
horizontal axis represents the distance x = vt allowing synchronized comparison at
different speeds.

loads due to the stiffness of rail, railpads and ballast [12] as they attain the deck of the
bridge. This distribution may have some significance for short bridges, as compared to
the consideration of concentrated point loads [3]. A reasonable assumption for usual
track stiffness is to consider a 14 – 12 – 14 distribution on 3 consecutive sleepers sep-
arated 0.60 m. Figure 6 shows a representative example, the maximum acceleration
envelope in the bridge deck as a function of train speed for a 10 m span bridge and a
reference MU load model, with a reduction of approximately 10% in dynamic effects
at the critical speed.

Number of modes to consider.— An issue which may be of importance is the num-

ber of modes to consider in the modal analysis. Some indications to this purpose are
given in the Eurocode [4] for checking accelerations on the bridge deck. For a dis-
placement analysis of a simply supported beam, it may be generally carried out with
only the first (fundamental) mode. Acceleration analysis or the extraction of stresses
or sectional resultants will often require more modes to be considered.
As an example, we present comparative results for the case of the ERRI D214
L = 30 m bridge under the ICE3 HS train in figure 7. The fundamental (first sym-
metric) mode frequency is in this case f1 = 3 Hz, and the second symmetric mode is
f3 = 27 Hz (the 2nd mode is skew-symmetric and has no influence on mid-span deflec-
tions). The analysis is performed here for a resonant velocity. It is clear from figure 7
(left) that for displacements only the fundamental mode is significant. However, a
noticeable influence is seen in figure 7 (right) for accelerations from the second sym-
metric mode. Further modes or even the direct integration with the complete model
  

           

   
Max. acceleration [m/s2 ] 4
Distributed loads ( 14 - 12 - 14 )
3 Concentrated point loads

0
80 100 120 140 160 180 200 220 240 260 280 300
v [km/h]

Figure 6: Envelopes for maximum accelerations showing the effect of track load dis-
tribution; reference load model for articulated MU AB 1D [29], 5 ⇥ 5 coach units
with individual axle loads P = 21.5 t. Simply supported bridge L = 10 m, m = 12
t/m, f0 = 12.464 Hz, ⇣ = 2.2%.
mid-span acceleration [m/s2 ]

20
mid-span deflection [mm]

1 symm. mode 1 symm. mode

2 symm modes 5 2 symm modes
10

0 0

10
5
20
0 1 2 3 4 5 0 1 2 3 4 5
time [s] time [s]

Figure 7: Influence of number of modes for simple bridge, on displacements and

accelerations. ICE3 HS train at v = 268 km/h on L = 30 m bridge from ERRI
D214 [11]

yield only minor increases to these accelerations.

2.3 Models with vehicle-structure interaction

The above models for dynamic analysis consider the traffic actions as moving loads of
fixed values, whereas in reality these loads provide from vehicles which have their own
dynamics and will not be constant. More realistic approaches include coupled vehicle-
bridge dynamic models with the interaction between both subsystems. In general, the
consideration of interaction with vehicles will allow part of the energy of vibration
 dtd dBd

M, J
G

Ks ,c s Ks ,c s
M b ,Jb M b ,Jb
Kp,c p Kp,c p

Mw Mw

deB
L

Figure 8: Schematic representation of models for vehicle-bridge interaction

of the bridge to be transferred to the vehicles, and consequently will predict lower
vibrations on the bridge.
Vehicle subsystems may be generally considered as rigid bodies, with masses for
wheelsets Mw , bogies Mb and Vehicle box M , and concentrated springs and dampers,
as shown in figure 8. Models for vertical dynamics need include only vertical trans-
lation degrees of freedom and pitch rotations for vehicle body or bogies, resulting in
relatively simple multibody models. Models for lateral dynamics include transver-
sal displacement and generally also roll and yaw rotations. These need complete 3D
descriptions and should be handled with general multibody models, either linear or
nonlinear, as described in section 3.

2.3.1 Simplified Interaction Model

Often full vehicle-bridge interaction models such as shown in figure 8 are not required,
as some parts of the vehicle will not interact with the bridge, depending on the fre-
quencies of vibration. Bodies with frequencies which are very low with respect to the
bridge will not be excited by the bridge motion and behave as constant moving loads.
Bodies with frequencies which are much higher will behave as added masses.
For railway passenger vehicles under vertical motion one can identify three ba-
sic characteristic frequencies for vertical vibration of the vehicle: 1) vibration of
wheelsets considering Hertz contact with the rails, in the order of 100 Hz; 2) vi-
bration of bogies on the primary suspension, in the order of 4 Hz; 3) vibration of the
vehicle box, in the order of 1 Hz. In practice, the types of bridges which show greater
dynamic vibrations have fundamental frequencies in the order of 3 to 7 Hz, hence the
interaction will provide mainly from bogie masses and primary suspensions. Wheelset
masses behave (from the point of view of bridge dynamics) as rigidly attached to rails,
and vehicle box masses as moving loads of fixed value.
For such cases a simplified interaction model is convenient (figure 9): the bridge
will be modelled by modal analysis with i = 1 . . . n modes, the train with j = 1 . . . k
interaction elements (one per wheelset, suspended mass + moving load). Under these

Figure 9: Simplified models for vehicle-bridge interaction including suspended bogie

masses and primary suspension

assumptions, the equations for each mode of vibration ( i = 1 . . .n, with amplitudes
yi ) are (details in [9]):
k
X ⌦ ↵
Mi ÿi + Ci ẏi + Ki yi = i djrel pj + mjb z̈ j , (8)
j=1

and for each interaction element (zj = 1 . . . k):

" n
#
j j
X ⌦ j
↵
mb z̈ + k j z j yi i (drel )
i=1
" n n
#
X ⌦ j
↵ X ⌦ ↵
0 j
+ cj ż j ẏi i (drel ) yi v i (drel ) = 0 . (9)
i=1 i=1

These equations are simple to implement and represent only a minor extension to the
moving load modal analysis equations (6), with very small computational cost.
Reduction in predicted dynamic effects from consideration of interaction more rel-
evant in short bridges and with frequencies of the same order as that of the primary
suspension, and for critical resonant velocities. As a representative example we show
in figure 10 the results as compared to moving load models. The case corresponds
to a CAF articulated MU train, whose primary suspension frequency is f1 = 5.97
Hz on a L = 20 m bridge with fundamental frequency f0 = 4 Hz. The maximum
deck accelerations obtained for the critical speed v = 280 km/h are 10.9 m/s2 with
moving loads and 7.8 m/s2 with interaction, a reduction of ⇡ 40%. The Envelope of
maxima obtained with respect to train speed shows that the reduction obtained with
vehicle interaction is relevant for critical resonant velocities, but is not significant for
non resonant scenarios.

3 Lateral dynamics of vehicles on bridges

3.1 General features of model
The consideration of lateral dynamics for the response of railway vehicles on bridges
requires three dimensional models, including degrees of freedom for lateral displace-
ment, rolling and yawing in the vehicles. It is also necessary to employ coupled
  

           

    

mid-span acceleration [m/s2 ]

Max. acceleration [m/s2 ]

10 moving loads 10 moving loads
interaction interaction

0
5

10
0
0 0.5 1 1.5 2 2.5 3 100 150 200 250 300
time [s] v [km/h]

Figure 10: Accelerations obtained with moving loads and with interaction with pri-
mary suspension: L = 20 m bridge ERRI D214 [11], f0 = 4 Hz, ⇣ = 1%; CAF
TEMD train, primary suspension frequency f1 = 5.97 Hz. Time history for resonant
critical velocity v = 280 km/h (top) and envelope as a function of velocity (bottom).

vehicle-structure models, which must take into account such features as nosing mo-
tion of wheelsets on the rails as well as track alignment irregularities. These features
give rise to substantially more complex models than those used for vertical dynamics
described in the previous section.
In this work a fully nonlinear coupled model is proposed for the lateral dynamic
analysis of vehicles on viaducts. Vehicles are considered as three-dimensional multi-
body systems, and the bridge structure is modeled by means of finite elements. The
model is developed in a general and modular way so that it may be easily implemented
within an existing finite element analysis software with multibody capabilities. For
this work Abaqus [28] finite element framework has been used. Both subsystems
(bridge and vehicles) are described with coordinates in absolute reference frames, as
opposed to alternative approaches which describe the multibody system with coordi-
nates relative to the bridge motion at the base of the vehicle. This facilitates the full
consideration of nonlinear inertia terms, without introducing additional difficulties for
the structural mechanical behavior [27].
Contrary to the majority of existing models for train–bridge dynamic interaction,
the formulation described here is capable of full consideration of geometrical and ma-
terial nonlinearities both in the structural subsystem (bridge) and in the multibody
subsystem (vehicle). The approach for wheel–rail geometrical interaction and me-
chanical contact model is fully nonlinear as well, not being limited neither to constant
conicity assumptions nor to linearized elastic contact forces. However, the particular
model and applications reported here involves some simplifying assumptions that will
be described below.
 A key aspect to address is the contact interface between bridge and vehicle, for
which several approaches could be followed:

1. A perfectly guided predefined path for the vehicle wheelsets, for the more direct
and simplest models; in these contact points between wheels and rails share
position and velocity. An assumed hunting movement can be introduced as a
prescribed motion of wheelsets [8, 30, 36, 39].

2. Linear models for taking into account relative motion between wheelset and
track; these rely on the classical assumption of constant conicity in wheel and
rail profiles and simplified contact [37, 38, 40].

3. Nonlinear models with more detailed treatment of wheel to rail contact, such as
proposed in [22, 32].

In this work we follow the last option, including a fully nonlinear wheel–rail inter-
action, based on the elastic contact forces approach [25]. The approach for contact
point determination at each wheel may be completely general, however the model
here reported employs a contact point determination based on a pre-computed geo-
metric lookup table. A single hertzian contact point [15] is considered at each wheel.
The tangential contact is solved by the FastSim algorithm [17]. Following the basic
features of the model are summarized. A more detailed description may be found
in [2].

3.2 Kinematics of wheelset and track

The wheelsets are considered as rigid bodies, within the vehicle multibody subsys-
tem. The bridge deck cross sections are also assumed to be rigid within the structural
finite element model, i.e. no distortion of the plane cross section is considered. This
assumption is automatically enforced by standard 3D beam-type finite elements. How-
ever, both vehicle and bridge can undergo large displacements and rotations. In order
to establish the contact interface with sufficient precision a detailed description of the
positions and velocities of the wheel and rail points is required, whose key concepts
are summarised below. A more general description of multibody kinematics may be
found in [23] or more specifically for railway applications in [24, 26].

Wheelset kinematics.— The position of a wheel point A, such as the wheel-rail

contact, at a certain time instant, is defined by (Figure 11)

rA = rw + ⇢A = rw + ⇤w ⇢¯A , (10)

where rw is the position vector of the wheelset center of mass, ⇢A the relative position
of the contact point A, ⇢¯A the convective vector referred to the body reference sys-
tem {w, ewi }, and ⇤w the rotation tensor that defines the absolute orientation of the
Figure 11: Reference frames and vectors for wheelset kinematics: the inertial refer-
ence frame {I, ei }; the body reference frame {w, ew
i }; and an intermediate frame
which does not consider wheelset spin{w, evi }

wheelset. We shall consider that the spin of the wheelset is a given rate '. ˙ It is ad-
vantageous to employ an intermediate reference frame {w, evi } which does not inherit
this spin, as in this frame the contact point of one wheel will be a constant vector ⇢˜A .
Considering a given parametrization ✓v for rotations (Euler angles, Euler parame-
ters. . . ), the angular velocity !w can be expressed in general as

!w = !' + !v = !' + G ✓˙ v , (11)

where G is a non-constant matrix operator whose structure depends on the choice of

parametrization and whose coefficients depend on the rotation coordinates ✓v .

Track kinematics.— We consider a bridge deck section at a position on the deck

defined by the longitudinal coordinate sw (Figure 12). The position vector of a point
C on the rail, in a similar way as for the wheelset, may be expressed as

rC = rb + ⇤b ⇢˜t + ⇤t ⇢¯C (12)

where rb defines the position of the reference point on the deck section. Two reference
frames are used for the track: the bridge section reference system b, ebi which is
attached to the deck section at point b, and the track reference frame {t, eti } attached
to the track point t located at the mid point between the top of both rails and at the
same deck section. ⇤b and ⇤t are, respectively, the rotation tensors that relate both
reference systems and the inertial frame.
The position of deck section reference point rb is obtained by interpolation of the
finite element nodal coordinates along the bridge deck,

rb = rb0 + Td (sw ) qB , (13)

where qB is the nodal coordinates vector of the bridge, Td (sw ) the displacements
interpolation operator, consistent with the FE approximation employed, and rb0 the
 (a) (b)

Figure 12: Reference frames and vectors for track kinematics: intermediate frame
attached to the deck section b, ebi and the track coordinate system {t, eti }

initial position vector of the deck section. In the same way, rotation parameters will
be interpolated as
✓b = ✓b0 + T✓ (sw ) qB , (14)
where T✓ (sw ) is the rotations interpolation operator and ✓b0 the initial values of ro-
tation parameters. The above expression contains the implicit assumption of small
rotations for the bridge deck, for which case the rotation vectors may be interpolated
linearly, as opposed to finite rotations which would require an interpolation of the
exponential mapping.

3.3 Wheelset-track interaction

In order to compute the contact forces between wheels and rails the problem may be
split in three consecutive stages:

1. Contact geometry: the main geometric variables involved in the problem are
computed;

2. Normal contact: the dimensions and shape of the contact area and the normal
stress distribution are determined;

3. Tangential contact: the resultant forces and moment of the tangential stresses,
which appear as a consequence of the rolling contact, are obtained.

Assuming that the material properties of wheel and rail are the same (material sym-
metry), the normal and tangential problems may be considered uncoupled [19] and
solved consecutively after the determination of the geometric variables.

Contact geometry.— The geometric problem relies on the key assumption that both
the track section and the wheelset are supposed to move as planar rigid bodies within
the plane cross section of the bridge (i.e. perpendicular to vector ebx , Figure 12). Under
this assumption, yaw rotation is neglected when assessing the main geometric vari-
ables involved in the wheel-rail contact. For a high-speed train running on a straight
track, which is the case considered in this work, yaw rotation remains small and this
approximation may be considered reasonable. Under this assumption the relative dis-
placements between the wheelset and the track may be defined by 3 coordinates, the
lateral and vertical relative displacements ( yw and zw respectively) and the rela-
tive rotation within the cross section w . Furthermore, two additional hypotheses
are assumed:

1. The contact between the wheel and the rail occurs at only one point (area) at
each wheel;

2. No separation occurs between wheels and rails.

Under these hypotheses, the relative vertical displacement žw and rotation ˇw ,
due to purely geometric considerations, can be computed as a function of the relative
lateral displacement yw for given wheel and rail profiles (in Figure 13 these are
plotted for the S1002 wheel and UIC60 rail profiles):

žw = žw ( yw ) , (15a)

ˇw = ˇw ( yw ) . (15b)

The relative coordinates žw and ˇw , the contact angles (Figure 14), rolling
radii r and transversal curvatures at both wheels are stored in a lookup table, from
which these geometric variables will be obtained by interpolation for a given value of
yw .

10
ˇ w [mrad]

5
Dy

0
Dzˇ w [mm],

5 Dzˇ w
ˇw
Dy
10
8 6 4 2 0 2 4 6 8
Dyw [mm]

Figure 13: Compatible relative vertical Figure 14: Contact reference frame
motion žw and rotation ˇw for S1002 {C, ei } and definition of angle at con-
c

wheel profile and UIC60 rail tact point of one wheel

Normal and tangential contact forces.— At every wheel-rail pair a single hertzian
contact [15] is assumed, for which the contact area is considered to be an ellipse with
 3.4. Problema tangencial

ma matemática como:
⌧ (xc + xc , yc ) = ⌧ˆ(xc + x c , yc )
major and minor semiaxes a and b respectively. The normal stress distribution is as-
si |ˆ sumed
⌧ (xc + xc to
, ybe
c )|an
<ellipsoid
µ zc (xcwhose + xresultant
c , yc ) is the normal contact force N . The wheel and
rail curvatures at the contact point may be computed as a function of yw as shown
above. From these one may compute their mechanical , properties,
(3.30) the resultant normal
force N , the ellipse semiaxes a and
⌧ˆ(xc + xc , yc ) b and the normal stress distribution, according to
⌧ (xc + xc , yc ) =Hertz
µ ztheory.
c (xc + x c , yc )
|ˆ
⌧ (xc + xc , yc )|
The solution for the tangential contact forces at each wheel contact is a complex
nonlinear problem for which a reference solution may taken as Kalker’s exact three-
si |ˆ
⌧ (xc + xc , yc )| µ zc (xc + xc , yc )
dimensional rolling contact theory [18]. The coupled simulation of trains on bridges
nto, si no existe will require numerous
deslizamiento, evaluations,
la tensión at each time-step
hipotética calculadaandseeachda wheel of the train. In
order toelbring
ida, en caso contrario, valordown computational
máximo de la tensióncosts totangencial
reasonable bounds
será el while maintaining suf-
o por la Ley de Coulomb, y la dirección será la de la tensión hipotética.algorithm [17] is used
ficient precision for the simulations, the simplified FastSim
here, involving a compromise between accuracy and computational cost. The FastSim
method 3.29,
tiendo de la Ecuación has been implemented
y haciendo usoby dethe la
authors as a user
Ecuación subroutine
3.30, es within the Abaqus
software employed [28] (further details
obtener la distribución de tensiones tangenciales (xc , yc ) en toda la in [1]).
de contacto. Para cubrirFor ade given
unafriction
formacoefficient
discreta la the input variables
µ, totalidad for evaluation of tangential
de la elipse,
sario discretizar, además de en la dirección Txc , en la dirección yc . Parasemiaxes and the creep-
rolling forces are the the normal stress distribution, the ellipse
ages, defined
a cabo la discretización as ⇠ = un
se toma y , ⇠r ] which are defined as:
[⇠x , ⇠número de divisiones según la
c
(vA v
ón xc (pxc ), y otro según la yc (pyc ). Valores C ) · e{x,y} son px(!
razonables w 20!yt ) · ez
= c
⇠{x,y} = 2 ⇠r = c , (16)
0, que es lo adoptado para este trabajo . Una v vez elegido el número v
njas pyc , el valor with the wheelset
del vancho de cadalongitudinal
una envelocity. The FastSim
esa dirección es model
yc =involves a discretisation
in strips within
Cada franja de ancho the contact
yc tiene ellipse and
una longitud integration
variable, poralong eachlos
lo que strip, as shown in Figure
15. In thispara
entos xc serán distintos figure a representative
cada case for distribution
franja y dependerán del valorofytangential
c de stresses is shown
also.
ntos medios de cada rectángulo, siendo xc (yc ) = 2 a 1 (yc /b)2 /pxc
iormente se explica el porqué del signo menos). Ver Figura 3.21.

. Elipse de contacto discretizada. En el centro de cada rectángulo de la discretización se

n los valores de las tensiones tangenciales según las dos direcciones locales, mediante FastSim,
rmitirá conocer una distribución de tensiones en toda la huella.
Figure 15: Discretisation of contact ellipse with FastSim and evaluation of tangential
forces in rail-wheel contact
poniendo conocida la tensión (xc , yc ), mediante la Ecuación 3.29 es
determinar el valorThe de resultants
la tensiónof the
(xcshear
+ xstress
c , yc ) distribution at a point
(Figura 3.22). A are the forces Tx and Ty
Inicial-
and the moment Mz whose directions are, respectively, ex , ecy and ecz :
c
elección pxc y pyc debe ser un compromiso entre la precisión y la eficiencia del
fC,A = Tx ecx + Ty ecy + N ecz , (17a)
mC,A = Mz ecx , (17b)
 Finally, in order to establish the geometric compatibility between wheels and rails
set in equation (15), two constraint equations are defined at every wheelset:
⇢ ⇢
zw žw ( yw ) 0
{ w} = ˇ = . (18)
w w ( y w ) 0
These constraints are enforced through a penalty method, with coefficients computed
by a linearization of the Hertz contact.

3.4 Vehicle-bridge system dynamics

Considering wheelset contact forces and moments from (17) (fC , mC ), applied ex-
ternal loads (fE , mE ) and loads transmitted by the suspension systems (fS , mS ), the
Newton-Euler equations for a single wheelset can be written as

m r̈w + fS = fE + fC , (19a)
J !˙ w + !w ⇥ (J !w ) + mS = mE + mC , (19b)

being m the mass and J the inertia tensor for the wheelset.
For a general case of parametrization of rotations G as defined in (11), equation
(19b) can be expressed as

J G ✓¨v + J Ġ✓˙ v + !w ⇥ (J !w ) +mS = mE + mC , (20)

| {z }
mQ
where the quadratic velocity terms have been included within the term mQ . The
wheelset dynamic set of equations (19) may be assembled as

Mw q̈w + Fw w w
I = F E + FC , (21)

being 
m1 0 ⇥ ⇤T
w
[M ] = , {qw } = rw ✓v , (22)
0 JG
and similarly FwI , FE , FC the assembled force vectors.
w w

Employing standard multibody dynamics models for the remaining of the vehicle,
these equations may be assembled into a full system of equations for the complete
vehicle. On the other hand, the discretised equations for the bridge structural dynamics
may be obtained following standard finite element procedures, and expressed in a
similar fashion. Finally, the two nonlinear coupled sets of differential equations may
be written as

MB q̈B + FB B B
I = FE + F C , (23a)
V
M q̈ + V
FVI = FVE + FVC , (23b)

where the superscript B is referred to the bridge and V to the vehicles, (FB I , FI ) are
V

the internal force vectors, (FB

E , FE ) the external loads and (FC , FC ) the contact force
V B V
vectors. For the bridge the internal force vector corresponds to the forces and moments
which appear as a consequence of the structural deformation. In the case of vehicles,
forces and moments produced by suspension systems and quadratic velocity terms are
included in this vector.
The interaction forces and moments of every wheelset are arranged in FVC . For
the bridge, interaction forces applied on the track are extrapolated to the correspond-
ing nodal coordinates of finite element model and assembled in FB C . These vectors
establish the coupling between the equations of both subsystems.
The nonlinear set of differential equations (23) may be solved in time using an
implicit integration method. In this work the HHT-↵ algorithm [16] has been used,
included in Abaqus software [28]. This method has a good stability and robustness for
the coupled problems described, and includes an inherent tunable numerical damping
for the high frequency noise. The constraints inherent to the multibody subsystem are
solved efficiently with augmented lagrangian procedures.

4 Applications in lateral dynamics

4.1 Vehicle running on continuous deck bridge
Two representative applications of the proposed models will be shown. First the re-
sponse of a representative high-speed vehicle when it crosses over a multi-span bridge
with continuous deck is analyzed. The mechanical data for the vehicle are detailed in
Appendix A.
The structure is a single track continuous bridge consisting of six equal spans of
L = 50 m each. The bridge is supported on two end abutments and five intermediate
piers. Torsional rotation is allowed at the piers but constrained at the abutments. The
deck cross section properties are uniform along the bridge length, see Appendix A.
The first lateral and vertical bending eigenfrequencies are equal, of value 2.18 Hz; the
first torsion eigenfrequency is 1.10 Hz. Euler-Bernoulli beam elements of 1 m length
with linear elastic behaviour have been used.
The vehicle runs at a constant velocity of 100 km/h, with a lateral transient wind
gust load defined by a chinese hat function [5] applied, with maximum value Fmax =
270 kN and gust duration ⌧ = 0.1 s (Figure 16). This load is applied on the vehicle
car-body along y direction (transversal) when the last wheelset of the vehicle enters
the bridge.
In Figure 17 the lateral response of the last wheelset of the vehicle when it crosses
the bridge is shown, compared with the case for a perfectly rigid track (i.e. no struc-
ture). It can be seen that wheel flanges contact the lateral part of rail heads, the limits
are indicated in the graph with dotted lines. We remark that for assessment of these
running safety scenarios a nonlinear model such as proposed here is essential, as linear
models cannot reproduce wheel–flange impacts.
Figure 18 shows the lateral response of the vehicle car-body and Figure 19, the
 Fmax
5 With structure
Without structure

Dyw [mm]
F(t)

5 Wheel flange contact

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400

t [s] x [m]

Figure 16: Lateral force history on vehicle Figure 17: Lateral response of last
car-body: wind gust load corresponding wheelset; dotted lines indicate limit for
to a “chinese hat” function wheel flange contact

100 With structure 10

Without structure
50
yc [mm]

Ty [kN]

0
0
With structure
50
10 Without structure
0 100 200 300 400 0 100 200 300 400
x [m] x [m]

Figure 18: Lateral response of the car- Figure 19: Tangential contact forces at the
body under an applied wind gust load left wheel of the last wheelset, showing
peaks for wheel–flange contact

tangential force Ty time history. This force is expressed at each instant in the local
tangent plane to the whee contact (i.e. frame {C, eci }) of the left wheel of the last
vehicle wheelset; it cannot be interpreted as a lateral load.
Figure 17 shows a significantly different response for wheelset displacements when
the structure flexibility is taken into account: not only in terms of amplitude of oscil-
lations, but also of frequency. A similar remark can be made for the car-body dis-
placements, Figure 18. Moreover, contact forces in Figure 19 exhibit peaks which
correspond to the flange impacts, these impacts differ significantly in both models.

4.2 Application for viaduct “Arroyo las Piedras”

A second application is performed for a real bridge within a new HS line, the “Arroyo
las Piedras” viaduct, in operation in the Córdoba–Málaga high speed line. This is
a long viaduct (total length 1209 m), with a double-track composite steel-concrete
continuous deck beam, supported on concrete piers with pot-type bearings. The spans
of the deck are of 65 m, and the section has upper and lower concrete slabs performing
a so-called double composite action, with special provisions for ensuring adequate
torsional stiffness. The tallest piers are of 94 m height. The first natural frequency
of this viaduct is 0.313 Hz and corresponds to a lateral deformation mode (figure 20).
More complete details of the structure may be found in [20].

Figure 20: Viaducto “Arroyo las Piedras” in Córdoba–Málaga HS line: first mode of
vibration 0.313 Hz

As described above, the model for the viaduct is based on 3D beams with appro-
priate kinematic constraints. The Rayleigh method has been used for the damping
matrix of the structure subsystem, with 0.5% damping centered in the two first natural
frequencies. The train that has been used in the calculations is an approximation of the
Siemens ICE 3 composed of 8 cars, each of 24.775 m length. The ICE3 is a multiple
unit (distributed power) train, and all cars are supposed to have the same geometrical
and mechanical properties.
Calculations have been carried out for train speeds of 250, 300 and 350 km/h, with
several models:

(1) Bridge only model with moving loads, consisting only of the bridge dynamic
subsystem, with the vehicle wheelsets simplified as moving loads of fixed mag-
nitude. This amounts to neglecting the dynamic effects of vehicle vibration, and
serves as a basic result with which to compare the influence of the vehicle vi-
bration on the results. It also serves the purpose of obtaining a so-called virtual
path for the wheelsets of the vehicles, which can be later applied to these in a
sequential approach to the vehicle-bridge dynamics: the bridge history of dis-
placements is obtained in a first step, and these histories are then applied in a
second step to the vehicle wheelsets to obtain the response of the train.

(2) Vehicle model, consisting only of the vehicle subsystem, in two different sce-
narios:

(a) Vehicle on rigid track (i.e. no bridge) with prescribed profiles of irreg-
ularities. This model will enable to compare the influence of the bridge
deformation on the vibrations of the vehicle.
(b) Vehicle on virtual path, as described above, this path results from previous
analysis of the bridge with moving loads. The geometric track irregular-
ities are added to the virtual path in order to consider their effect on the
vehicle.
 (3) Bridge-Vehicle model, performing the calculation for the global coupled sys-
tem with wheel-rail contact interaction. In this case, two scenarios have been
considered:
(a) Model with interaction but without track irregularities
(b) Model with interaction and with track irregularities

Following we show only the results for train speed of v = 350 km/h, as they
correspond to the greatest effects. Firstly we present results for the deformation of
the bridge deck: displacements and rotations at the center of span 11, corresponding
to the tallest pier, are shown in figure 21. These correspond to the three different
scenarios defined above, the cases enumerated as 1, 3a and 3b. It may be clearly
seen that the influence of the vehicle vibration and of the track irregularities on the
bridge deformation is negligible. It is also seen that these deformations are small,
with maximum lateral displacement of 9 mm, which for a viaduct of 1209 m length is
very small.

1
no interaction (moving loads)
0 interaction with irregularities
interaction w/o irregularities
0.8
Deck torsion Lateral (mrad)
Lateral displacement (mm)

-2
0.6

-4
0.4

-6
0.2

-8 0
no interaction (moving loads)
interaction with irregularities
interaction w/o irregularities
-10 -0.2
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Longitudinal distance (m) Longitudinal distance (m)

Figure 21: Displacements and torsional rotation of bridge deck at center of span 11 of
Arroyo de las Piedras viaduct, for v = 350 km/h train speed.

Regarding the response of the vehicle, the accelerations of vehicle body on coach
4 are shown in figure 22. Four scenarios are shown, as defined in section 4.2, the
cases enumerated as 2a, 2b, 3a and 3b. Maximum accelerations experienced are in the
order of 0.2 – 0.3 m/s2 . We remark that very clearly irregularities are the main cause
responsible for vehicle vibrations, as in the case 2a accelerations are negligible com-
pared to other cases with irregularities. In other words, the vehicle body accelerations
attributable to deformation of the bridge are of the order of 0.05 m/s2 .

5 Conclusion
The dynamic effects of high-speed trains on viaducts are important issues for the de-
sign of the structures, as well as for the consideration of safe running conditions of the
trains. An adequate analysis of the resonant conditions and effects on railway bridges
 0.4 0.4
rigid track with irregularities virtual path with irregularities
0.3 0.3
LAteral body acceleration (m/s2)

LAteral body acceleration (m/s2)

0.2 0.2

0.1 0.1

0 0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Longitudinal distance (m) Longitudinal distance (m)

0.4 0.4
interaction w/o irregularities interaction with irregularities
0.3 0.3
LAteral body acceleration (m/s2)

LAteral body acceleration (m/s2)

0.2 0.2

0.1 0.1

0 0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Longitudinal distance (m) Longitudinal distance (m)

Figure 22: Acceleration of coach body over Arroyo de las Piedras Viaduct, for v =
350 km/h train with different models

is a key aspect for the design of bridges, which may be of special relevance in short
span bridges. Some aspects such as the longitudinal distribution of loads, the num-
ber of modes for calculation and the consideration of interaction with the vehicles are
discussed.
The study of lateral dynamics of running trains on bridges is of importance mainly
for the safety of the traffic on laterally compliant bridges. These studies require 3D
coupled vehicle-bridge models and consideration of wheel to rail contact. A fully
nonlinear coupled model is proposed here, described in absolute coordinates and in-
corporated into a commercial finite element framework. The applications presented
demonstrate the relevance of the coupling effect between vehicle and bridge, as well
as for considering nonlinear contact wheel-rail models.

Acknowledgments
The authors are grateful for support from Ministerio de Ciencia e Innovación of Span-
ish Government in the project Integración de la Monitorización de Viaductos Fer-
roviarios en el Sistema de Gestión y Mantenimiento de Infraestructuras “VIADIN-
TEGRA” (IPT-370000-20190-012), of the subprogram INNPACTO. The authors ac-
knowledge also the support from the Technical University of Madrid, Spain.
References
[1] P. Antolı́n, J.M. Goicolea, M.A. Astiz, “Strategies for Modeling Train-Bridge
Lateral Dynamic Interaction”, in H. Xia, G. De Roeck, J.M. Goicolea (Editors),
Bridge Vibration and Controls: New Research, Chapter 6, page In press. Nova
Science Publishers, Inc., Hauppauge, NY, 2012, ISBN 978-1-62100-868-2.

[2] P. Antolin, J.M. Goicolea, J. Oliva, M.A. Astiz, “Nonlinear train-bridge lateral
interaction using a simplified wheel-rail contact method within a finite element
framework”, Journal of Computational and Nonlinear Dynamics, 2012, In press.

[3] CEN, EN1991-2:2003 Eurocode 1 – Actions on structures, Part 2: Traffic loads

on bridges, European Committee for Standardization, september 2003.

[4] CEN, EN1990:2002/A1:2005 Eurocode 0 – Basis of Structural Design, Am-

mendment A1; Annex A2, Application for bridges, European Committee for
Standardization, december 2005.

[5] CEN, EN14067-6: Railway applications – Aerodynamics – Part 6: Require-

ments and test procedures for cross wind assessment, European Committee for
Standardization, 2010.

[6] R. Clough, J. Penzien, Dynamics of Structures, Mac Graw-Hill, 1993.

[7] M. de Fomento, IAPF-07: Instrucción sobre las acciones a considerar en el

proyecto de puentes de ferrocarril, Government of Spain, 2007.

[8] R. Dias, J. Goicolea, F. Gabaldon, M. Cuadrado, J. Nasarre, P. Gonzalez, “A

study of the lateral dynamic behaviour of high speed railway viaducts and its
effect on vehicle ride comfort and stability”, in K.. Frangopol (Editor), Bridge
Maintenance, Safety, Management, Health Monitoring and Informatics, pages
724–735. IABMAS08: Fourth International Conference on Bridge Maintenance,
Safety and Management, Taylor & Francis Group London, Seoul, 14-16 jul 2008,
ISBN 978-0-415-46844-2.

[9] J. Domı́nguez Barbero, Dinámica de puentes de ferrocarril para alta velocidad:

mtodos de cálculo y estudio de la resonancia, PhD thesis, Technical University
of Madrid, 2001, http://oa.upm.es/1311.

[10] ERRI, Committee D181 final report: Forces Latérales sur les Ponts Ferrovi-
aires, European Railway Research Institute, Utrecht, The Netherlands, 1996.

[11] ERRI, Committee D214 final report: Design of Railway Bridges for Speed up
to 350 km/h; Dynamic loading effects including resonance, European Railway
Research Institute, Utrecht, The Netherlands, 1998.

[12] C. Esveld, Modern Railway Track, MRT–Productions, The Netherlands, 2 edi-

tion, 2001.
[13] European Railway Agency, Directive 96/48/EC Interoperability of the Trans-
European High Speed Rail System; Technical Specification for Interoperability,
Infrastructure Sub-System, Official Journal of the European Union, march 2008.

[14] W. Guo, H. Xia, Y. Xu, “Running safety analysis of a train on the Tsing Ma
Bridge under turbulent winds”, Earthquake Engineering and Engineering Vi-
bration, 9(3): 307–318, 2010.

[15] H. Hertz, “Über die berührung fester elasticher körper and über die härtean”, J.
für reine und agewandte Mathematik, 92: 156–171, 1882.

[16] H.M. Hilber, T.J.R. Hughes, R.L. Taylor, “Improved numerical dissipation for
time integration algorithms in structural dynamics”, Earthquake Engineering &
Structural Dynamics, 5: 283–292, 1977.

[17] J.J. Kalker, “A Fast Algorithm for the Simplified Theory of Rolling Contact”,
Vehicle System Dynamics, 11(1): 1–13, 1982, ISSN 0042-3114, URL http:/
/www.informaworld.com/10.1080/00423118208968684.

[18] J.J. Kalker, “The principle of virtual work and its dual for contact problems”,
Ingenieur-Archiv, 56(6): 453–467, 1986.

[19] J.J. Kalker, Three-Dimensional Elastic Bodies in Rolling Contact (Solid Me-
chanics and Its Applications), Springer, 1990, page 344.

[20] F. Millanes, J. Pascual, M. Ortega, “’Arroyo de las Piedras’ viaduct: The first
composite steelconcrete high speed railway bridge in Spain”, Structural Engi-
neering International, 17(4): 292–297, 2007.

[21] J. Nasarre, “Serviceability limit states in relation to the track in railway bridges”,
in R. Calçada, et al. (Editors), Bridges for High-Speed Railways, pages 211–220.
Taylor & Francis, London, 2009.

[22] D.V. Nguyen, K.D. Kim, P. Warnitchai, “Dynamic analysis of three-dimensional

bridge-high-speed train interactions using a wheelrail contact model”, Engineer-
ing Structures, 31(12): 3090–3106, 2009.

[23] P.E. Nikravesh, Computer-Aided Analysis of Mechanical Systems, Prentice Hall,

Inc., Englewood Cliffs, NJ, 1988, page 370.

[24] K. Popp, W. Schiehlen, Ground Vehicle Dynamics, Springer Berlin Heidelberg,

Chennai, 1st edition, 2010, ISBN 978-3-540-68553-1, page 348.

[25] A.A. Shabana, K.E. Zaazaa, J.L. Escalona, J.R. Sany, “Development of elastic
force model for wheel/rail contact problems”, Journal of Sound and Vibration,
269(1-2): 295–325, Jan. 2004, ISSN 0022460X, URL http://dx.doi.
org/10.1016/S0022-460X(03)00074-9.
[26] A.A. Shabana, K.E. Zaazaa, H. Sugiyama, Railroad Vehicle Dynamics: A Com-
putational Approach, CRC Press, 2008.

[27] J.C. Simo, L. Vu-Quoc, “The role of non-linear theories in transient dynamic
analysis of flexible structures”, Journal of Sound and Vibration, 119(3): 487–
508, Dec. 1987, ISSN 0022460X, URL http://dx.doi.org/10.1016/
0022-460X(87)90410-X.

[28] Simulia Ltd., Abaqus 6.10 User’s manual, Providence, RI, 2010.

[29] SNCF, “New categories for defining the compatibility interface between multi-
ple units and infrastructure; Definition of Reference Load Models for new Multi-
ple Units”, Technical report, Societé Nationale des Chemins de Fer, Département
des Ouvrages d’Art, 6 avenue Francois Mitterrand, 93574 La Plaine St Denis,
june 20, 2011, In the framework of CEN TC256 WG10 SG03 – Update of the
EN15528.

[30] M.K. Song, H.C. Noh, C.K. Choi, “New three-dimensional finite element anal-
ysis model of high-speed train-bridge”, Engineering Structures, 25: 1611–1626,
2003.

[31] M. Tanabe, H. Wakui, N. Matsumoto, H. Okuda, M. Sogabe, S. Komiya, “Com-

putational model of a Shinkansen train running on the railway structure and the
industrial applications”, Journal of Materials Processing Technology, 140(1-3):
705–710, 2003.

[32] M. Tanabe, H. Wakui, M. Sogabe, N. Matsumoto, T. Y, “An efficient numerical

model for dynamic interaction of high speed train and railway structure including
post-derailment during and earthquake”, in G. De Roeck, G. Degrande, G. Lom-
baert, G. Müller (Editors), 8th International Conference on Structural Dynamics,
EURODYN 2011. K. U. Leuven, Leuven, Belgium, 2011.

[33] S.P. Timoshenko, Vibration problems in engineering, Van Nostrand, 1928.

[34] UIC, Code UIC 776-1: Charges a prendre en consideration dans le calcul des
ponts-rails, Union Internationale des Chemins de Fer, 3 edition, jul 1979.

[35] UIC, Code UIC 776-1: Charges a prendre en consideration dans le calcul des
ponts-rails, Union Internationale des Chemins de Fer, 5 edition, august 2006.

[36] H. Xia, N. Zhang, G.D. Roeck, “Dynamic analysis of high speed railway bridge
under articulated trains”, Computers & Structures, 81: 2467–2478, 2003.

[37] Y.L. Xu, N. Zhang, H. Xia, “Vibration of coupled train and cable-stayed bridge
systems in cross winds”, Engineering Structures, 26: 1389–1406, 2004.
[38] Y.B. Yang, J.D. Yau, Y.S. Wu, Vehicle-Bridge Interaction Dynamics:
With Applications To High-Speed Railways, World Scientific Publishing
Company, Singapore, 1st edition, 2004, ISBN 9812388478, page 530,
URL http://www.amazon.com/Vehicle-Bridge-Interaction-
Dynamics-Applications-High-Speed/dp/9812388478.

[39] N. Zhang, H. Xia, W. Guo, “Vehicle-bridge interaction analysis under high-

speed trains”, Journal of Sound and Vibration, 309(3-5): 407–425, 2008.

[40] N. Zhang, H. Xia, W.W. Guo, G. De Roeck, “A Vehicle-Bridge Linear Interac-

tion Model and Its Validation”, International Journal of Structural Stability and
Dynamics, 10(02): 335, 2010.

Appendix A: Vehicle and bridge mechanical properties

Item Unit Value Item Unit Value

mc kg 53500.0 Jcx m4 70000.0
Jcy m4 2621500.0 Jcz m4 2621500.0
mb kg 3500.0 Jbx m4 560.0
Jby m4 315.0 Jbz m4 1715.0
mw kg 1800.0 Jwx m4 1000.0
Jwy m4 100.0 Jwz m4 1000.0
kx1 kN/m 120000.0 ky1 kN/m 12500.0
kz1 kN/m 1200.0 cx1 kN · s/m 27.9
cy1 kN · s/m 9.0 cz1 kN · s/m 10.0
kx2 kN/m 12000.0 ky2 kN/m 240.0
kz2 kN/m 350.0 cx2 kN · s/m 600.0
cy2 kN · s/m 30.0 cz2 kN · s/m 20.0
hc m 1.4 hb m 0.5
r0 m 0.46 dc m 17.375
db m 2.5 dw m 0.753

Table 1: Mechanical properties of the vehicle considered in Section 4.1. m refers to

masses, J to rotary inertias, k corresponds to suspension stiffnesses and c to damping;
subscripts x, y and z indicate the orientation of the suspension properties, and 1 and 2
are referred to primary and secondary suspensions; c, b and w correspond to car-body,
bogies and wheelsets, respectively; hc and hb are the center of gravity heights of car-
body and bogies; and dc is the longitudinal distance between two bogies of the same
vehicle and db between two wheelsets of the same bogie.
 Item Unit Value Item Unit Value
E GPa 30.0 G GPa 12.5
A m2 1.0 Jxx m4 0.189
Jyy m4 1.0 Jzz m4 1.0
⇢ kg/m3 2500.0 ez m 1.50

Table 2: Mechanical properties of bridge deck cross section considered in section 4.1,
being E and G the Young and shear modulus; A the cross section area; Jxx , Jyy and
Jzz the moments of inertia corresponding to the longitudinal, transversal and vertical
directions, respectively; ⇢ the material density and ez the vertical offset of the track
center as explained in Section 3.2.

View publication stats