DYNAMIC ANALYSIS FOR FATIGUE SAFETY EXAMINATION OF

EXISTING SHORT SPAN CONCRETE RAILWAY BRIDGES

Mário Pimentel1*, Joaquim Figueiras1 and Eugen Brühwiler2
1
Laboratory for the Concrete Technology and Structural Behaviour (Labest)
University of Porto, Faculty of Engineering
Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
E-mail: mjsp@fe.up.pt
2
Division of Maintenance and Safety of Structures (MCS)
École Polytechnique Fédérale de Lausanne (EPFL)
ENAC IS MCS GC2B2402 Bâtiment GC Station 18, 1015 Lausanne, Switzerland
E-mail: eugen.bruehwiler@epfl.ch

Keywords: Fatigue, railway bridges, dynamic amplification, reinforced concrete.

ABSTRACT

In a previous work by the authors, it was shown that, in a scenario of increasing maximum
axle loads, the fatigue life of short span railway reinforced concrete bridges is drastically
reduced. In view of the great sensitiveness of the fatigue life to some simplified assumptions
usually adopted in the fatigue safety verification procedure, more detailed analysis methods
will certainly be very effective to prove the required fatigue safety of existing structures. This
communication focuses on the dynamic effects of heavy railway traffic on short span bridges.
A dynamic analysis model considering the interaction between the train and the bridge is
presented. The model uses a fully coupled formulation of the bridge/vehicle equilibrium
equations, avoiding iterative schemes at each time step. For maximum efficiency, the modal
decomposition technique in conjunction with a quasi-static correction procedure is used to
calculate the bridge dynamic response. It is shown that the dynamic amplification factor
proposed in the current design codes can be conservative in the case of short span ballasted
concrete bridges. The effect of the increasing train mass on the dynamic amplification factor is
discussed and it is shown that the dynamic amplification tends to decrease with increasing
axle loads.

1. INTRODUCTION
Modern railway systems sustainability demands for larger traffic volumes, higher traffic loads
and higher train speeds. This issue is already expressed in the recent studies of the European
Rail Research Institute regarding the formulation of new load model, the LM2000, to account
for the expected increase in the maximum axle loads in the design of new railway bridges [1].

E215
The existing bridge stock also needs to be checked against these demanding loading
conditions and decisions must be taken concerning their upgrade to withstand the new loading
scenario. The transport capacity increase of existing bridges by allowing axle loads up to 30
tonnes for freight traffic with moderate speeds is also one of the objectives of an ongoing
European research project (www.sustainablebridges.net).
The economical impact of strengthening measures for existing structures is much larger
than the additional cost due to a conservative design of a new structure. As the public funds
for the improvement or replacement of the increasingly growing existing bridge stock are
generally highly constrained, the safety examination of existing structures fully justifies more
detailed and sophisticated structural analysis methods those usually applied in the design of
new structures. It can be then understood that, in the case of the existing bridges safety
evaluation, besides the new traffic models, also the development and application of
methodologies enabling the exploitation of hidden safety reserves, usually not considered in
the design stage, is mandatory.
In a previous work by the authors [2] it was shown that, in a scenario of increasing
maximum axle loads, the fatigue life of short span reinforced concrete (RC) bridges can be
drastically reduced and possibly cannot be proven using usual analysis approaches. It was also
shown that in view of the great sensitiveness of the fatigue life to some of the underlying
assumptions usually adopted in the fatigue safety verification procedure, improved analysis
methods leading to even small reductions in the calculated stress spectrum may be crucial to
prove fatigue safety. The results from a parametric study [2] indicate that a 5% decrease in the
calculated reinforcement stress range spectrum may lead to a 50% increase of the remaining
fatigue life. From the multiplicity of factors that contribute to the sometimes significant
differences between the calculated and the real stress ranges, this work focuses on the
dynamic effects of heavy railway traffic on ballasted short span RC bridges.
In the current European codes regarding the railway traffic actions on bridges [3], the
proposed dynamic amplification factor (DAF) for the actions effects in railway bridges is
mainly based on experimental evidence and on numerical simulations of non-ballasted steel
bridges. Although some studies have been done with concrete ballasted bridges, it was
recognized that the presence of ballast will provide greater safety, as the dynamic effects are
reduced [4, 5]. Regarding the studies performed in the United States, in the paper [6] is
performed a statistical analysis of the results obtained from a large experimental campaign on
several type of railway bridges under controlled traffic actions (locomotives). Lower values
for the DAF were attained by ballasted bridges when compared to the non-ballasted ones.
In view of the sensitiveness of the fatigue life to the calculated stress range spectrum, the
applicability of the Eurocode 1 DAF for fatigue verifications of existing ballasted RC bridges
can be questionable and overly conservative predictions may be expected, fully justifying
research in this field.

2. DEVELOPED MATHEMATICAL MODEL

Classically, the problem of bridge-vehicle interaction can be approached by two fundamental
formulation types: uncoupled or coupled formulations. In the formulations of the uncoupled
type, the vehicle and the bridge equations of motion are solved separately and an iterative
procedure is necessary so that, at each time step, the vehicle and the bridge displacements
differ less than a pre-defined tolerance. Examples of this type of formulations can be found in
[7-10]. In [11] a different implementation of an uncoupled formulation is described, in which
the bridge equations of motion are solved in the frequency domain and the iterative procedure
is performed after each complete passage of the vehicle. In the coupled formulations the

E215
equations of the bridge and the vehicle are solved simultaneously and, once linear elastic
behaviour is assumed for the components, no iterative procedure is required. In this case, the
stiffness and damping matrices of the coupled system are time dependent, as they depend on
the vehicle position. Formulations of the coupled type can be found in [12-15], among others.
In the present work a coupled formulation is developed and implemented in Matlab. For
maximum efficiency, the modal decomposition technique in conjunction with a quasi-static
correction procedure is used to calculate the bridge dynamic response.

2.1 Bridge
Let us consider the general case of a bridge deck subjected to a set of Nl moving forces, Pl(t),
not necessarily constant in time and each of them uniformly distributed over a surface Sl =
dx,l* dy,l (see Figure 1).

Figure 1. Bridge deck under moving forces.

Assuming linear elastic behaviour for the bridge, which is a reasonable assumption under
operational loading conditions, and considering small displacements, the partial differential
equation of motion describing the problem is given by

∂ 2 w(x, t ) ∂w(x, t )
L[w(x, t )] + m (x ) + C (x ) = p (t ) (1)
∂t 2
∂t

with the external load

Nl
Pl (t )  x − v x tl   y − v y tl 
p(t ) = ∑ ∫R  d x ,l
rect rect  dx (2)
l =1 d x ,l d y ,l
  dy 
   ,l 

In the expressions above the following notation is valid: x = ( x, y ) is the position vector;
v = (v x , v y ) is the velocity vector; w(x, t ) is the bridge deck displacement at time t and at the
position x ; L[w(x, t )] is a linear operator with respect to spatial variables; m is the mass per
unit surface area; C is a viscous damping operator with respect to spatial variables;
t l = t − d1,l v where d 1,l is the distance between the first and the lth axle force; and rect (a )
is the rectangular function which is defined as 1 if a ≤ 1 2 and 0 otherwise.
The solution to this problem can be advantageously found by transforming the geometric
displacement coordinates w(x, t ) into modal-amplitude or normal coordinates, using the
mode-superposition method. For the general case of a 3D problem, this transformation reads

∞
w(x, t ) = ∑ Y (t )φ (x )
n n (3)
n =1

where Yn (t ) is the time-dependent amplitude of the nth mode shape and φ n (x ) is the nth mass-
normalized mode shape satisfying the standard orthogonality conditions.

E215
 Replacing (3) in (1), pre-multiplying both sides by φ k (x ) , integrating along the domain
and applying the orthogonality conditions, the uncoupled equations of motion are obtained,

Y&&k + 2 ξ k ω k Y&k + ω k2 Yk = Pb, k ( t ) , k = 1,..., ∞ (4)

where the kth modal excitation function is given by

Nl
Pb ,k ( t ) = ∑ φ (v t ) P ( t )
med
k l l (5)
l =1

with φkmed (v tl ) being the mean value of the kth mode shape function over the loaded area at
instant t and under the lth axle:

 x − v x tl   y − v y tl 
φ kmed (v t l ) =  φ (x )dx
1 rect 
d x ,l d y , l ∫R  d x ,l
rect
  dy  k
(6)
   ,l 

The basis of the modal superposition method is that, once the structural response is
uncoupled into orthogonal modal coordinates, each individual mode responds to the applied
loading independently. The dynamic amplification effects of the applied loading can then be
evaluated separately for each mode in the same way as any single-degree-of-freedom
oscillator. It can be shown that the response component associated to high order modes is
negligible, meaning that the dynamic amplification can be accurately determined using only a
small number of terms of the series given in expression (3). However, non negligible errors
may arise from this truncation when computing the elastic forces (bending moments and shear
forces) since a large number of mode shapes may be necessary just to reproduce the
correspondent static influence line (or influence field). To overcome this fact a quasi-static
correction procedure was implemented, which basically uses the static influence lines to
obtain the contribution of all the modes not explicitly considered in the equations of motion
and that are assumed to respond in a static way:

N mod e
 Pb,k ( t )  Nl
ζ (x, t ) = ∑ Aζ ,k (x ) Yk ( t ) − + ∑ης (x, v t ) P ( t ) (7)
ω 2 l l
k =1  k  l =1

In the expression above ς (x, t ) is the dynamic effect under consideration (shear force,
bending moment, strain component, etc), ης (x, v tl ) is the static influence field (or influence
line) of the effect, Aς ,n (x ) is the contribution to ς (x, t ) due to a unit displacement in the kth
mode and Nmode is the number of modes included in the response. All these functions can be
analytically derived in simple cases or most generally be obtained from a detailed finite
element model.
The bridge equations of motion can be then expressed in the matrix form:

&& + C Y
Mb Y & + K Y = P (t ) (8)
b b b

with M b = I , C b = diag (2 ξ k ω k ) , K b = diag ω k2 ( ) and Pb (t ) a column vector whose

elements are given by (5).

E215
 In the assessment of existing structures a proper calibration of the numerical models with
the actual structural behaviour is mandatory. In the case of short to medium span railway
bridges, besides the additional stiffness provided by non-load bearing elements and
elastomeric supports, the composite action between the track and the deck has been shown to
have a large influence on the dynamic properties of the structural system (mode shapes and
respective natural frequencies) [16, 17]. For an accurate dynamic behaviour examination,
prior system identification and model updating is a crucial task, as these effects can hardly be
predicted theoretically with sufficient accuracy. In order to fulfil these requirements, in the
present approach the mode shapes and the natural frequencies of the bridge can be extracted
from detailed 3D finite element models and input into the developed program.

2.2 Vehicle and track

Different types of running gear can be found on European railway lines. The standard
suspension system for two–axle freight wagons is called the UIC standard suspension, see
Figure 2a). Each wagon has four of these suspensions, consisting of a leaf spring and linkages
(the UIC links). The deflection of the leaf spring provides the vertical stiffness while
dissipative dry friction forces are generated by the relative displacements in between the
leaves. Due to this form of friction damping the force-displacement relation is non-linear,
exhibiting hysteresis cycles [18]. The carbody is connected with the leaf spring through the
UIC links. The system of links works as a pendular suspension system with friction in the
joints, providing the horizontal suspension [19] and exhibiting a stepwise progressive
behaviour due to existing clearances between the links and the suspension bracket. The axle
motion is restricted by the axle guidance and the existing clearances both in the horizontal and
transversal direction are an additional source of nonlinearity [20].

a) b) c)
Figure 2. Types of running gear: a) UIC standard suspension; b) UIC link bogie; c) Y25 bogie.

In the case of four-axle freight wagons two types of bogies are commonly used in Europe.
In Figure 2b) is depicted the UIC link bogie (typeG70). The vertical and horizontal suspension
characteristics are also governed by leaf springs and UIC links. The bogie is connected to the
car body by a central spherical pivot. In Figure 2c) is depicted the Y25 bogie. The primary
suspension of the Y25 consists of pairs of coil springs. The load-dependent friction damping
is provided by a normal force, which is generated by the so-called Lenoir link, acting on
friction surfaces on both sides of the axle box. A central spherical pivot connects the bogie to
the car body. For further details refer to [21]. More recently, a hydraulic damped type of
running gear, the TF25 bogie series (TF25SA for two axle wagons), is being introduced in the
European lines, allowing 25t axle loads (30t axle loads already running on selected Swedish
lines) as well as speeds up to 160km/h.
The traditional freight vehicles exhibit a highly complex nonlinear dynamic behaviour due
to dry friction damping and motion delimiters causing impacts. The detailed modelling of this
behaviour is essential in order to investigate the vehicle running safety, wear of the
mechanical systems and track degradation. Concerning the dynamic effects on ballasted
concrete bridges the focus is on the analysis of the vertical degrees of freedom, which largely

E215
simplifies the problem. In this work the vehicles are modelled as a set of lumped masses,
dampers with viscous behaviour and linearized springs.
The vehicle equations were derived using Langrange’s formulation, as follows:

d  ∂T 
 − ∂T + ∂V +
∂WD
=0 (9)
dt  ∂u& v ,i  ∂u
 v , i ∂u v , i ∂u& v , i

where T is the total kinetic energy of the system, calculated using the mass, mass moments of
inertia and rotational as well as translational velocities of the system components, V is total
potential energy, which is computed from the spring stiffness and relative displacements, and
WD is dissipated energy, obtained from the relative velocities and damping coefficients.
A library of vehicles has been built and parameterized covering a wide range of
locomotives and freight cars. In Figure 5 are depicted two types of freight wagons. The
presented vehicles are 2D, although 3D vehicles can be formulated within the same
framework.

a) b)
Figure 5. Freight wagon models: a) Two axle wagons; b) Four axle wagons.
The track stiffness is taken into account by considering a spring and a damper bellow the
vehicle wheels. This approximation is acceptable in the low frequency range, which is the
range of interest when studying the dynamic effects on bridge elastic forces.
The vehicle equations of motion can then be expressed in the matrix form as:

&& v + C v u& v + K v u v = Pv (t )
Mv u (10)

where Mvv, Cvv and Kvv are the (Nvdof x Nvdof) mass, damping and stiffness vehicle matrices,
&& v , u& v
respectively (Nvdof is the total number of degrees of freedom of the specified train); u
and u v are the (Nvdof x 1) vectors with the accelerations, velocities and displacements
associated to each degree of freedom. The right hand side vector components are given by

Pv ,i (t ) = k t ,i [r (v t i ) + w(v t i )] + ct ,i [r&(v t i ) + w& (v t i )] + Fstat ,i , i = 1,.., N vdof (11)

if the ith degree of freedom refers to a wheel mass vertical displacement and equals 0
otherwise. In the expression above, kt,i and ct,i are the track spring stiffness and damping
coefficient, respectively, r(xi) is the irregularity coordinate under the correspondent axle and
v.ti is the axle position in correspondence with the ith degree of freedom. The dot refers to time
differentiation. The force exerted by the lth axle, is given by

P l (t ) = k t ,l [u v ,l (t ) − r (v t l ) − w(v t l )]+ ct ,i [u& v ,l (t ) − r&(v t l ) − w& (v t l )] ≥ 0, l = 1,..., N l (12)

if the lth axle is on the bridge and equals 0 otherwise. The displacements uv,l are measured
from the position where the springs kv,l are undeformed.

E215
2.3 Coupled equations of motion
The bridge and vehicle equations can be coupled after expressing the vehicle excitation force
vector Pv, with the Nvdof components given by eq. (11), and the vector Pb with the Nmode bridge
modal excitations functions Pb,k, given in eq. (5) and (12), in terms of the system unknowns,
Yn and uv,i (and respective time derivates). To accomplish that, it must be noted that:

w(v t l ) = Φ(v t l ) Y(t ) ; w& (v t l ) = v Φ ′(v t l ) Y(t ) + Φ(v t l ) Y (t ) ; r&(v t l ) = v r ′(v t l ) (13)
T T T &

where the last two expressions were obtained by the chain rule, (-)´ refers to space
differentiation along the trajectory followed by the vehicle wheels and Φ(v t l ) is a column
vector with the Nmode mode shapes evaluated at vtl. Replacing (13) in (10) and (8), it is
possible to obtain the (Nmode + Nvdof) coupled system of second order differential equations:

M b 0  Y &&  C b + C bb C bv   Y&  K b + K bb K bv   Y  Fb 

 0  +  +  =  (14)
 M v  u
&& v   C vb C v  u& v   K vb K v  u v  Fv 

where the time-dependent coupling matrices Cbb, Cbv, Cvb, Kbb, Kbv and Kvb, and the right
hand side vectors, Fb and Fv, are given by

Nl
C bb = ∑c Φ med (v t l )Φ(v t l ) ;
T
t ,l
l

∑ [k ]
Nl
K bb = Φ med (v t l ) Φ(v t l ) + v ct ,l Φ med (v t l ) Φ ′(v t l ) ;
T T
t ,l
l

C bv , ij = − φimed (v t j )ct , j ; K bv , ij = − φimed (v t j )k t , j ;

C vb, ji = − φ i (v t j )ct , j ; K vb, ji = − v φi′(v t j )ct , j − φi (v t j )k t , j ;

∑ φ (v t )[− k r (v t ) − v c r ′(v t )];

Nl
Fb , i = i
med
l t ,l l t,l l
l =1

Fv , j = k r (v t ) + v c r ′(v t ) + F
t, j j , t, j j stat , j i = 1,..,Nmode ; j = 1,…,Nvdof

The expressions for the elements the of Cbv, Cvb, Kbv, Kvb coupling matrices and Fv vector
are valid only if the jth vehicle degree of freedom corresponds to the vertical displacement of a
wheel mass, being equal to 0 otherwise.

2.4 Wheel/Track irregularities

In railway tracks several types of track defects can occur. These can be of the localized type,
like an indentation of the rail head, a raise on a welding joint, a dipped rail joint or a
depression due to an unsupported sleeper, or of the periodic type like the rail corrugations.
Wheel defects, like wheel flats, are also responsible for a periodic type of defect.
Irregular track patterns (due to wear, subsidence, ballast compaction, insufficient track
maintenance, etc) are usually considered as random irregularities which are described by its
statistical characteristics. These irregularities are normally considered as stationary and
ergodic process in space and are fully characterized by a power spectral density (PSD)
function, from which particular realizations can be generated. In [22] a complete set of
different PSD functions for elevation, alignment and gauge irregularities are presented.

E215
 Concerning the study of the dynamic amplification of bridge elastic forces, only the
irregularities with a wavelength and amplitude above certain thresholds are considered to be
relevant. Low wavelength small amplitude irregularities (like rail corrugation) generate a high
frequency excitation content which is damped by the track system and hardly affects the
bridge (which is also inert to very high frequency contents). It is also common practice to
execute rail joints outside the bridges, precisely to avoid unwanted dynamic effects, and
therefore the effects of welding and dipped joints will not be considered in this study.
The function

1  2π 
r (x ) = Arr 1 − cos x  (15)
2  Lrr 

was adopted to model the shape of rounded wheel flats, worn rail head indentations and
depressions in the track due to unsupported sleepers, whereas the random irregularity patterns
where generated according to

Grr ( f rr ,i )∆f rr cos(2π f rr ,i x − θ i )

N harm
r (x ) = ∑2 (16)
i =1

where frr (cycles/m) is the inverse of the irregularity wavelength, Grr is the appropriate PSD
function and θi is a random phase angle uniformly distributed between 0 and 2π.

2.5 Computational implementation

The presented mathematical model was implemented in Matlab. The coupled system of
differential equations is solved with the fourth order Runge-Kutta method and an
automatic/adaptative time stepping scheme. An event driven integration procedure was
adopted to deal with the situations where loss of contact between the wheels and the bridge is
detected. An event corresponds to a zero of the function (12), which physically may signify
both a loss of contact or contact recover situation. When the contact is lost the integration is
stopped, the correspondent kt,l and ct,l parameters are made equal to zero in (14), the numeric
integration restarted, and vice-versa.
The model was validated with solutions obtained by other authors with different theoretical
models [11-13, 23] and total agreement was found.

3. PARAMETRIC STUDY
The principal factors influencing the dynamic amplification of the elastic response of concrete
ballasted railway bridges are the traffic speed across the bridge, the span length L (or the
influence line length of the element under analysis), the bridge natural frequencies, the mass
of the bridge; the bridge damping, the number of axles and the axle spacing, the mass of the
trains, the suspension characteristics of the vehicles, the track irregularities and vehicle
defects. In the following the influence of some of these factors is analysed.
The results presented herein refer to two simply supported short span bridges of 5 and 10m
spans (see Table 1). The analyses were 2D and the axle loads distributed within a 1m length.
The bridge damping ratios were conservatively defined following the lower bound values
proposed in [3] for RC ballasted bridges. Four modes were included in the response and only
the midspan bending moment results are presented. The train was composed of 3 wagons of
two axles each (see Figure 5a)), with 7m axle spacing and 4m between axles of consecutive

E215
wagons. The vehicle properties are summarized in Table 1. The two wagon masses, mc,
indicated correspond to 200kN and 300kN axle loads.

Vehicle Bridges
mc= 38t, 68t kv=1/2 mc (2 π fv1)2 L=10m L=5m
Ic=7mc cv= 2 π fv1 mc ξv f1=12Hz f1=22Hz
mw=1.0t kt=80MN/m, 160MN/m mb=8.3t/m mb=7.0t/m
d1=d2=3.5m ct=0 ξ=2.20% ξ=2.55%
Table 1: Values used in the parametric study.

In Figure 7a) are depicted the linearized parabolic leaf spring stiffness for various
maximum axle loads and corresponding tolerances (values taken from http://www.langen-
sondermann.de/). In this study, the body bounce natural frequency, fv1, which is related to the
suspension stiffness by the approximate relation indicated in Table 1, was chosen as basic
parameter for the suspension properties definition. Three values for the body bounce
frequency (fv1 =1.8, 2.0 and 2.2Hz) were selected, as indicated in Figure 7.

3500 2.8
Susp. stiffness 1/2 kv [kN/m]

3000 2.6
2.2Hz
Frequency fv1 [Hz]

2.4
2500 2.2Hz
2.0Hz 2.2
2.0Hz
2000 1.8H 2
z 1.8Hz
1500 1.8
1.6
1000
1.4
500 1.2
0 1
100 150 200 250 300 100 150 200 250 300
Max. Axle Load [kN] Max. Axle Load [kN]

a) b)
Figure 7. Parabolic spring stiffness, 1/2 kv, (a) as a function of the maximum axle load and
correspondent body bounce natural frequencies, fv1 (b).

The equivalent viscous suspension damping was defined by the equivalent damping ratio,
ξv, (see Table 1) which was assumed to vary between 0.05 and 0.1.
Four types of irregularities were considered. For the first three types the function given by
expression (15) was used: irregularity of type 1 intends to reproduce a large depression in the
track (Arr=6mm, Lrr=3m), irregularity of type 2 a medium depression (Arr=2mm, Lrr=1m) and
irregularity of type 3 the effect of a large wheel flat or railhead indentation (Arr=1.3mm,
Lrr=0.100m). These irregularities were conservatively placed at the midspan of the bridges, so
that the dynamic effect occurs simultaneously with the maximum static effect of the bending
moment. The irregularity of type 4 consists of 10 random profiles generated using (16), with a
wavelength content between 1m and 100m, and the PSD function proposed by the Federal
Railway Administration (FRA) for class 4 tracks (class 1 corresponds to a poorly maintained
track and class 6 to the highest track standard). In this case, the presented values for the DAF
correspond to the mean value obtained with the 10 random profiles.
In this work, the dynamic amplification, 1 + ϕ of a given effect is defined as the ratio
between the maximum dynamic and static responses.
In Figures 8 and 9 are presented the variation of the DAF with the train speed and train
mass for various types of irregularities. The results obtained with a moving load algorithm
(neglecting the interaction) are also presented for comparative purposes. Although the

E215
conclusions are limited due to the fact that only one train type and two bridges were
considered, some trends can be extracted from these preliminary results: the DAF decreases
with increasing axle loads, especially when the DAF attains significant values; the 10m span
bridge is more affected by the long track depression (Type 1) and the 5m span bridge by the
medium track depression (Type 2); the short wavelength irregularity (Type3) induces higher
DAF values at lower speeds; in the case of the 10m span bridge, the mean bending moment
DAF value obtained with the 10 random irregularity patterns (Type 4) compares reasonably
well with the EC1 fatigue DAF [3], which is already a mean value reflecting the average
dynamic effect over the bridge lifespan; in the case of the 5m span bridge, the EC1 fatigue
DAF is clearly above the computed results.
In Figure 10 are presented the results of the sensitivity study concerning the suspension
parameters. It can be seen that the DAF tends to increase with increasing suspension stiffness
and decreasing suspension damping. The DAF obtained with the irregularity of type 3 is
insensitive to the suspension characteristics, at least within the considered range of variation.

1.4 1.4
20t - Irreg. Type 1 20t - Irreg. Type 3
1.3 30t 1.3 30t

Amplification (1 + ϕ)
Amplification (1 + ϕ)

20t - Irreg. Type 2 20t - Irreg. Type 4

30t 30t EC1 fatigue DAF
1.2 1.2

1.1 1.1

1 1
Moving Loads Moving Loads
0.9 0.9
40 60 80 100 120 140 40 60 80 100 120 140
Speed [km/h] Speed [km/h]

Figure 8. Computed midspan bending moment DAF for varying train speed, axle load and
track irregularity type: 10m span bridge, fv1=2Hz, ξv=0.1, kt=80MN/m.
1.4 1.4
20t - Irreg. Type 1 20t - Irreg. Type 3
30t 30t EC1 fatigue DAF
1.3 1.3
Amplification (1 + ϕ)
Amplification (1 + ϕ)

20t - Irreg. Type 2 20t - Irreg. Type 4

30t 30t
1.2 1.2

1.1 1.1

1 Moving Loads 1
Moving Loads
0.9 0.9
40 60 80 100 120 140 40 60 80 100 120 140
Speed [km/h] Speed [km/h]

Figure 9. Computed midspan bending moment DAF for varying train speed, axle load and
track irregularity type: 5m span bridge, fv1=2Hz, ξv=0.1, kt=80MN/m.

In Figure 11 the effect of the ballast stiffness is analyzed. It can be seen that just in the case
of the short wavelength irregularity the effect of the track stiffness can be distinguished. This
type of irregularity excites mostly the wheel mass (the suspension keeps the wagon mass
isolated from the high frequency content) which turns the axle forces highly sensitive to the
track stiffness. Nonetheless it must be noted that the considered track/wheel defect exceeds by
far any defect that can be found in a European heavy traffic railway line with standard
maintenance procedures. In Figure 12 are depicted the maximum, minimum and mean values
of the DAF computed with the 10 random irregularity patterns (type 4). The maximum values
are in the range of the values obtained by [24] for prestressed concrete ballasted bridges with
similar spans.

E215
 1.4 1.4
fv1=2.0Hz - Irreg. Type 3 psiv=0.1 - Irreg. Type 3
fv1=2.2Hz EC1 fatigue DAF psiv=0.05 EC1 fatigue DAF
1.3 1.3
Amplification (1 + ϕ)

Amplification (1 + ϕ)
fv1=1.8Hz
fv1=2.0Hz - Irreg. Type 4 psiv=0.1 - Irreg. Type 4
fv1=2.2Hz psiv=0.05
1.2 fv1=1.8Hz 1.2

1.1 1.1

1 1
L=5m, 30t axles, ξv=0.1, kt=80MN/m L=5m, 30t axles, fv1=2Hz, kt=80MN/m
0.9 0.9
40 60 80 100 120 140 40 60 80 100 120 140
Speed [km/h] Speed [km/h]

Figure 10. Computed DAF for varying suspension parameters: a) Varying fv1;; b) Varying ξv.
1.4 1.4
kt=80MN/m - Irreg. Type 3 kt=160MN/m - Irreg.Type 4
kt=160MN/m EC1 fatigue DAF EC1 fatigue DAF
1.3

Amplification (1 + ϕ)
1.3
Amplification (1 + ϕ)

kt=80MN/m - Irreg. Type 4 Max

kt=160MN/m 1.2
1.2

1.1 1.1 Mean

1 1
L=5m, 30t axles, fv1=2Hz, ξv=0.1 Min
0.9 0.9
40 60 80 100 120 140 40 60 80 100 120 140
Speed [km/h] Speed [km/h]

Figure 11. Computed DAF for varying kt. Figure 12. Dispersion of the DAF values.

6. CONCLUSIONS
Due to the high sensitiveness of the fatigue life of short span RC bridges to the expected
increase in the axle loads, a detailed study on the traffic action effects is being undertaken
with the objective of performing more accurate fatigue life predictions for existing RC
bridges. In this work a numerical model for the analysis of vehicle/bridge interaction problems
was presented. The algorithm proved highly efficient and some preliminary results were
discussed. Although still of limited extension, the parametric study revealed a clear trend for
the DAF to decrease with increasing axle loads and lower DAF values than the proposed in
the EC1 for fatigue verifications. The parametric study will now be extended and field tests
are underway to validate some modeling assumptions.

ACKNOWLEDGMENTS

The financial support by the Portuguese Foundation for Science and Technology (FCT)
through the PhD grant SFRH/BD/24540/2005 attributed to the first author is greatly
acknowledged.

REFERENCES
[1] ERRI-D221, ERRI D221/RP1 - Background document for rail traffic actions, in Vertical
Rail Traffic Actions, Combination Factors. European Rail Research Institute (ERRI),
Utrecht, 1999
[2] Pimentel, M., E. Brühwiler, and J. Figueiras, Fatigue Life of Short Span RC Railway
Bridges. Structural Concrete, Journal of the fib, 2007. (accepted for publication).
[3] CEN, Eurocode 1: Actions on structures - Part 2: Traffic loads on bridges. EN 1991-2.
Brussels, 2003.
[4] James, G., Analysis of Traffic Load Effects on Railway Bridges, in Structural Engineering
Division. Doctoral Thesis, Royal Institute of Technology, Stockholm, 2003.

E215
[5] ERRI-D23:Specialists'Committe, RP17: Determination of dynamic forces in bridges -
General survey of the work of the D23 Comittee and preparation of general formulae for
calculating the dynamic effect. International Union of Railways UIC, Office for Research
and Experiment (ORE), Utrecht, 1970.
[6] Byers, W.G., Impact from Railway Loading on Steel Girder Spans. Journal of the
Structural Division, ASCE, 96: p. 1093-1103, 1970
[7] Karoumi, R., Response of Cable-Stayed and Suspension Bridges to Moving Vehicles,
Doctoral Thesis. Royal Institute of Technology (KTH), Stockholm, 1999.
[8] Hwang, E.-S. and A.S. Nowak, Simulation of Dynamic Load for Bridges. Journal of
Structural Engineering, 117(5): p. 1413-1434, 1990.
[9] Calçada, R., A. Cunha, and R. Delgado, Analysis of Traffic-Induced Vibrations in a
Cable-Stayed Bridge. Part II: Numerical Modelling and Stochastic Simulation. Journal of
Bridge Engineering, 10(4): p. 386-397, 2005.
[10] Broquet, C., Comportment dynamique des dalles de roulement des ponts en béton
sollicités par le trafic routier. Thèse de doctorat No 1964, École Polytechnique Fédérale
de Lausanne (EPFL), Lausanne, 1999.
[11] Green, M.F. and D. Cebon, Dynamic Response of Highway Bridges to Heavy Vehicle
Loads: Theory and Experimental Validation. Journal of Sound and Vibration, 170(1): p.
51-78, 1994.
[12] Frýba, L., Vibration of Solids and Structures under Moving Loads. 3rd ed., London:
Thomas Telford, 1999.
[13] Yang, Y.-B. and J.-D. Yau, Vehicle Bridge Interaction Element for Dynamic Analysis.
Journal of Structural Engineering, 123(11): p. 1512-1518, 1997.
[14] Zhang, Y., et al., Vehicle-Induced Dynamic Perfomance of FRP versus Concrete Slab
Bridge. Journal of Bridge Engineering, 11(4): p. 410-419, 2006.
[15] Zhang, Q.-L., A. Vrouwenvelder, and J. Wardenier, Numerical simulation of train-bridge
interactive dynamics. Computer & Structures, 79: p. 1059-1075, 2001.
[16] Chellini, G. and W. Salvatore. Updated models for steel-concrete composite HS railway
bridges. EVACES'07, Porto, Portugal, 2007.
[17] Zabel, V., M. Brehm, and C. Bucher. Seasonal changes of the dynamics of railway
bridges with steel girders embedded in concrete. EVACES'07, Porto, Portugal, 2007.
[18] Fancher, P.S., et al., Measurement and representation of the mechanical properties of
truck leaf springs. Technical Paper Series 800905, Society of Automotive Engineers,
1980.
[19] Piotrowski, J., Model of the UIC link suspension for freight wagons. Archive of Applied
Mechanics, 73: p. 517-532, 2003.
[20] Hoffmann, M., Dynamics of European two–axle freight wagons, Doctoral Thesis,
Technical University of Denmark, Lingby, 2006.
[21] Stichel, S., On freight wagon dynamics and track deterioration. Proceedings of the
Institution of Mechanical Engineers - Part F, 213: p. 243-254, 1999.
[22] Fryba, L., Dynamics of Railway Bridges, London: Thomas Thelford, 1996.
[23] Yang, Y.-B. and Y.-S. Wu, A versatile element for analysing vehicle bridge interaction.
Engineering Structures, 23: p. 452-469, 2001.
[24] Chu, K.H., V.K. Garg, and T.L. Wang, Impact in Railway Prestressed Concrete Bridges.
Journal of Structural Engineering, 112(5): p. 1036-1051, 1986.

E215