Engineering Failure Analysis 119 (2021) 105007

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Influence of different pathologies on the dynamic behavior and

against fatigue of railway steel bridges
Fran Ribes-Llario a, *, Clara Zamorano a, J.A. García Cerezo b, Julia Real c
a
Universidad Politécnica de Madrid, Calle del Prof. Aranguren, 3, 28040 Madrid, Spain
b
FGV, Spain
c
IMM, Universitat Politècnica de Valencia, 46022 Valencia, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: Steel railway bridges built throughout the twentieth century represent a very important asset in
Railway bridge the railway lines of the entire European network. These bridges, despite their advanced age,
Dynamics continue to be part of passenger and freight transport networks, that provide transport to a large
Fatigue
part of society. Given their advanced age, these structures are prone to present higher levels of
deterioration in spite of the maintenance tasks that are carried out.
The appearance of pathologies in the wheel-rail contact suppose an alteration in the dynamic
behavior of the via-vehicle system. This fact gains importance when they appear in bridges,
increasing the number of cycles undergone by the structure, breaking down structural elements.
In this paper, the structural dynamic behavior of a metallic bridge with the presence of
different pathologies is studied. The study is focused on a metal bridge built during the first half of
the twentieth century currently in service. The aim of the article is to evaluate the influence of
these elements on the remaining life to fatigue that they present.

1. Introduction

To fully understand track and vehicle dynamics is crucial when it comes to maintenance, safeness and operational matters. This
importance is even bigger when analyzing singular sections of the infrastructure, as it is the case of railway bridges.
According to [11], the analysis of the response of a bridge under the effect of a moving train is not trivial, because the excitation
forces exerted over the infrastructure involve not only the characteristics from a moving load but also repeated load pulses from
consecutive axles, bogies, and carriages.
On the other hand, the phenomenon of resonance takes place when frequencies similar to the eigenfrequencies of the bridge are
induced by the passage of the train. These frequencies may be associated to the distance between adjacent wheelsets, the distance
between adjacent sleepers, or irregularities in the wheel-rail contact among others.
Under these circumstances, the vibratory response of a track increases noticeably, thus enhancing the risk of damage to the
structure and, sometimes, leading to its collapse. These reasons have encouraged the study of dynamics effects in railway bridges over
the last few decades. Some authors as [7,13] studied the effect of high-speed trains on bridges and the procedures to calculate the
dynamical response of the bridge under resonance conditions.

* Corresponding author.
E-mail addresses: [email protected] (F. Ribes-Llario), [email protected] (C. Zamorano), [email protected] (J.A. García Cerezo),
[email protected] (J. Real).

https://doi.org/10.1016/j.engfailanal.2020.105007
Received 20 February 2019; Received in revised form 11 October 2020; Accepted 11 October 2020
Available online 15 October 2020
1350-6307/© 2020 Elsevier Ltd. All rights reserved.
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 1. Isolated span scheme.

The dynamic response of a railway bridge is influenced by numerous factors, such as structure natural frequencies and damping,
train speed, or the length of both the structure and the train. Within this context, [11] studied the influence of different parameters,
including the bridge-to-carriage length ratio, while [18] demonstrated that the primary frequencies in the bridge response might be
caused by the driving frequencies, related to the time the train spent crossing the bridge, and by the dominant frequencies, which are
caused by the repeated loads.
On the other hand, the actions generated by the trains on the railway bridges have this cyclic character necessary to generate the
breakage by fatigue, with very important load oscillations, being the tensional variation the most influential parameter. Furthermore,
when the speed of the trains increases, the phenomenon worsens due to the following reasons:

• The number of tension cycles increases when a greater number of free vibrations are generated.
• The possible resonance input of the structural elements leads to a significant growth in both the magnitude of the tensional os­
cillations and the number of cycles.

Despite the importance of the phenomena, few studies have demonstrated the decreased fatigue life with corrugated rail profiles or
other wheel-rail pathologies. One of these studies was performed by [6], whose analyzed a two-span steel bridge at the Facility for
Accelerated Service Testing (FAST). Previous test results shown the importance of reducing dynamic loads by removing rail joints. The
tests results indicate that corrugated rail can also produce high dynamic loads resulting in a significant increase in fatigue damage. For
operation of HAL traffic over bridges, it is recommended that the rail surface be maintained as smooth as possible and free of dis­
continuities in order to minimize dynamic loading on the structure. It was observed a strong correlation between rail irregularities and
increased rate of fatigue crack initiation and growth. The irregularities of the rail have a strong influence in the dynamic behavior of
the structure, and it can be separated between two parts: (a) static track irregularity, due to the inherent defects of the rails and the
train wheels and (b) dynamic track irregularity, due to rail elastic deformation under train loads.
Recently, [1] carried out an in-depth literature review of the history of fatigue cracking of welded railway bridges, discussing the
structural behavior of welded bridges under dynamic traffic loads and the different causes of fatigue damage of welded connections.
The studied cases cover a period from the early 1930s until now, revisiting well-known cases of cracks documented in the literature due
to their historic relevance and presenting also recent cases of the last decades due to new traffic conditions related to Heavy Traffic
Loads (HAL) and High-Speed Trains.
The aim of the current research is to study how different pathologies in either the infrastructure, or the structure may affect the
dynamic behavior of a vehicle-track-structure system. This affection is analyzed at different vehicle speeds, and it is presented in terms
of displacements and structure accelerations. It is also intended to determine how these pathologies can reduce the fatigue life of the
different structural elements that make up the bridge.

2. Methodology

In order to identify the affection of different pathologies to the dynamic response of the Bridge caused by the traffic of vehicles, and
to the structural affection that it implies, a FEM model is developed. A steel railway bridge built in the beginning of the XX century and
located in the province of Alicante will be the focus of the research. Measurements obtained from a data gathering campaign on the
structure will be used for calibrating the FEM model.
Once the model has been calibrated, a series of scenarios with different pathologies will be simulated in both wheel-rail contact and
structure. The aim is to know the dynamic behavior, on the one hand, and its affection by the fatigue resistance of the materials on the
other.

2.1. Finite elements model

The current section presents the finite elements model developed in order to analyze the dynamic response of a track built over a 7
isostatic steel spans bridge built on the early XX century (see Fig. 1). To that aim, the dynamic model has been set so as to reproduce the
characteristics, conditions, and constraints available in a real stretch of track located close to Alicante, and the outputs of the model
have been compared to the data gathered in the aforementioned track. The description of the model is followed by the presentation of
the calibration process performed to validate the accuracy of the model. Finally, a modal analysis of the structure-track system is
described.

2
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 2. Bridge FEM model.

Table 1
Main track parameters.
Element E (MPa) υ ρ (kg/m3)
Rail 210,000 0.3 7,850
Railpad 600 0.45 7,850
Sleeper 1,310 0.35 700
Steel 210,000 0.225 7,850
Piers 2,000 0.3 1,890

The structure-track-vehicle system is simulated by means of a finite elements model, implemented with ANSYS LS-DYNA V17. It is
divided into two sub-models: the structure-track model, and the vehicle model.
Then, the dynamic response of each sub-model is calculated by relating the internal forces of the system to the external ones. This
relationship can be expressed through the equation of motion:
[M]{ü} + [C]{u̇} + [K]{u} = {F a (t)} (1)

Being [M] the global mass matrix, [C] the damping matrix, [K] the stiffness matrix and {u}, {u̇} and {u}
¨ the vectors of displacement,
speed and acceleration respectively. { Fα (t)} represents the time-dependent external forces vector.
Following [15–16], the damping matrix [C] can be expressed as:
[C] = α[M] + β[K] (2)

where α and β are the so-called Rayleigh coefficients.

The external forces, represented by {Fα (t)}, are the outputs from the vehicle sub-model that will be detailed below. Thus, dis­
placements, velocities and accelerations can be calculated for each node of the model solving Eq. (1).
Full three-dimensional finite elements models present a significant drawback in terms of computational time. Although they
provide reliable results, in some occasions they are not the best choice when it comes to calculate numerous hypothetical scenarios.
That is the case of the current piece of research, and therefore, despite it is a three-dimensional model, some simplifications have been
made to achieve reasonable computational time without harming results’ accuracy.
The bridge model is composed by BEAM189 elements (see Fig. 2), a quadratic beam element (3-node) in 3-D with six degrees of
freedom at each node. The degrees of freedom at each node includes translations in x, y, and z directions, and rotations about the three
axis. Warping of cross sections is assumed to be unrestrained. This element is based on Timoshenko beam theory. Shear deformation
effects are included.
The spans are composed by structurally independent (except the track) steel lattice girders built as a pratt truss.
The union of the different elements of the deck and partial sections of the elements (flange and web, compound profiles, etc.) is
carried out by means of angular joining profiles and semi-spherical head bolts.
The bridge has a single track composed by a UIC 54 rail with a metric gauge, Wooden sleepers with a dimensions of 220 mm × 180
mm, according to NRV 3-1-0.0., and both are fixed by a KD54P railpad. Despite the focus of this study are the bridge behavior, the
continuity of the track it is an important conditioning of the behavior of the whole system. The boundary conditions of the track have
been simulated avoiding the displacement on the longitudinal directions, simulating the constraint of the continuous beam.
The calibrated parameters of the elements in the track-bridge system are detailed in the Table 1. Some of them are well known, due
to its achievement through test, and are reliable for the authors. However, there are some parameters which uncertainty is relatively
high. Has been for these last parameters for which the calibration process has been carried out. More specifically, the parameters which

3
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 3. Vehicle model scheme.

has been calibrated have been the railpad, which stiffness could vary over such a long time of use, and the piers, which parameters
could be approximated but not exactly.
Regarding to the vehicle sub-model, it has been designed considering a three-mass system, accounting the wheel, bogie and the
carbody masses. The masses are linked by springs and dampers in parallel, which simulate the primary and secondary suspensions.
The vehicle is modeled as a multi-mass system of 1\8 of the rail vehicle. In it, 2 vertical degrees of freedom are considered con­
nected by means of spring-damping elements that represent the primary and secondary suspensions of the vehicle. There are two mass
elements, one corresponding to 1/8 of the mass of the vehicle, and another that represents 1/4 of the mass of the bogie. Both are
connected to the non-suspended masses which are in contact with the rail. The contact is modeled as a Herzian contact, which will be
detailed later. These idealizations are commonly used when the studied is the bridge response. This model is not valid if we would
assess the passenger confront, as it is not the case of this paper [5].
The scheme that represent the 1/8 movement of the vehicle are (see Fig. 3):
Where k2v, c2v, k1v, c1v represent the stiffness and damping coefficients of the primary and secondary suspensions.
The problem is governed by the following equation systems:
⎡ ⎤ ⎡ ⎤
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ ⎡ ⎤
Mc 0 0 0 ⎢ ẍc ⎥ cs − cs 0 0 ⎢ ẋc ⎥ ks − ks 0 0 xc Mc 0
⎢ 0 Mb 0 0 ⎥⎢ ẍb ⎥ ⎢ − cs cp + cs − cp 0 ⎥⎢ ẋb ⎥ ⎢ − ks kp + ks − kp 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥⎢ ⎥ + ⎢ ⎥⎢ ⎥ + ⎢ ⎥⎢ xb ⎥ == ⎢ Mb ⎥g + ⎢ 0 ⎥ (3)
⎣ 0 0 Mu 0 ⎦⎢ ẍu ⎥ ⎣ 0 − cp cp 0 ⎦⎢ ẋu ⎥ ⎣ 0 − kp kp + kH − kH ⎦⎣ xu ⎦ ⎣ Mu ⎦ ⎣ 0 ⎦
⎣ ⎦ ⎣ ⎦
0 0 0 0 üC 0 0 0 0 u̇c 0 0 − kH kH uc 0 FC

The passage of the vehicle through a certain section of track generates a series of normal and tangential stresses in the wheel-rail
contact. The value of these forces will depend, among other variables, on the relative displacements, velocities and accelerations of
both bodies. Depending on the objective of the study to be carried out, the modeling of the wheel-rail contact can be considered as only
acting vertical forces, or considering the normal and tangential forces. If the second option is required, it is necessary to go to specific
software for modeling vehicle dynamics. In this work, we will only be focused on vertical dynamics, considering the vehicle-rail
interaction. The contact surface is considered an ellipse contained in a plane, the distribution of normal pressures being a semi-
ellipsoid.
According to Hertz’s theory, the elastic deformation of two bodies can be determined as:
( )1/3
9Fn
δ= *Fe (e) (4)
16E2 Re
Being Fn the normal contact force, Re the equivalent radius of curvature and E a relationship between the elastic properties of both
bodies. In the case of rail-wheel contact, the wheel and rail are considered to have the same elastic properties, the radius of curvature in
the direction of the rail is infinite, and that there is only contact in the wheel’s tread.
From the previous expression it is that the out of touch is:

δv 2/3
Fv = (5)
CH

2E
where, Ch = (rr rw )1/4
3(1 − ν2 )

Δ is the relative vertical deformation, rw is the radius of the beam at the point of contact, rr the radius at the railhead.
A non-linear Herzian spring is used to model the wheel-rail contact consisting of two nodes, one in the centre of the wheel, and
another in contact with the rail. This type of model is suitable even if the normal contact force significantly exceeds the static force.
To establish the transmission of loads and the compatibility of displacements between the vehicle and the track, contact between

4
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Table 2
Main vehicle parameters.
Car body mass (Mc ) 18,920 kg
Bogie mass (Mb ) 6,000 kg
Unsprung mass (Mu ) 1,500 kg
Hertzian contact stiffness (kH ) 1600E6 N/m
Primary suspension stiffness (kp ) 1,200,000
Primary suspension damping (cp ) 30,000 N/m
Secondary suspension stiffness (ks ) 550,000
Secondary suspension damping (cs ) 98,000 N/m

Fig. 4. Installed sensors. Red: Accelerometers; Blue: Strain sensors; Green: Temperature sensors. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

Table 3
Numerical and experimental modes.
Experimental results Numerical results Variation (%)

First flexural mode 9.66 9.406 2.62%

Second flexural mode 23.24 22.09 4.91%
Third flexural mode 27.78 27.28 1.76%

the lower node of the hertzian element and the contact surfaces of the railhead will be required. The contact process is composed of two
stages. First, the identification of the contact areas occurs and then the kinematic relationships, necessary to impose the corresponding
contact restrictions, are established. For this, a contact without relative slippage is imposed and the contact restrictions are established
from the Lagrange multipliers. When the node comes into contact with the surface, first of all, the point of contact X must be
determined, which is responsible for materializing the contact between the lower node of the spring and the rail surface. Once
determined X the non-penetration condition is defined as:
( )
gN = X1 − X *n ≥ 0 (6)

The mechanical characteristics of the vehicle system, which has been provided by the railway administrator, are the following (see
Table 2).

2.2. Data gathering campaign

In order to count on real data for the development of the present study, a load test was carried out with the collaboration of railway
administrator. Strain gauges and accelerometers were used for the test with a QuantumX MX1615B datalogger.
A first static load test and a second dynamic test were performed with speeds of 50 km/h. In both cases a Diesel Train Series 2500
was used, being the one that usually circulates through the railway network that occupies the studied bridge.
The points where the sensors have been installed are indicated in the following Fig. 4.

2.3. Calibration of the model

For the calibration of the model, the results of the records taken by the accelerometers has been used, first, a modal analysis were
performed, and the results measured by the sensors were compared with those achieved by the model (see Table 3). Then, the
deflection value of the center of the span has been used for validating the model.

2.3.1. Modal analysis

A modal analysis has been performed assuming that materials present a linear-elastic behavior, neglecting the damping and dis­
regarding those actions external to the system. Following these premises, the following equation is to be solved:

5
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 5. Flexural modes of vibration.

Fig. 6. Dynamic deflection comparison.

(− ω2i [M] + [K]){ϕi } = {0} (7)

where ωi are the natural circular frequencies of the system, {ϕ} the eigenvectors, [M] the matrix of mass and [K] the matrix of stiffness.
Natural frequencies f i are then calculated as:
ωi
fi = (8)
2π
Due to geometric and loading symmetry in the structure, torsional modes are not excited, and therefore the main natural fre­
quencies that may be involved in this scenario are those related to the flexural shape modes. The first flexural modes of vibration are
shown in Fig. 5, and they are related to frequencies of 9.66 Hz, 23.14 Hz, and 27.78 Hz respectively accordingly to the obtained from
acceleration results.

2.3.2. Quasi-static deflection

One the modal analysis has been performed; the deflection results will be used to calibrate the model. During the data gathering
campaign, a train passing at a speeds within 40 and 50 km/h where used for obtaining, among other parameters, the maximum
deflection of the span. These registers have been used for validating the amount of deflection of the span with a quasi-static analysis.
On Fig. 6, can be seen a superposition of 10 real registers with the results of the validated model (Blue thick line).

6
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

0.3

0.2
Amplitude (mm)

0.1

0
0.00 1.00 2.00 3.00 4.00 5.00 6.00
-0.1

-0.2

-0.3
Rail Profile (m)

0.5 m Corrugation 2 m Corrugation

Fig. 7. Rail corrugation profiles.

1.2
1
Amplitude (mm)

0.8
0.6
0.4
0.2
0
0.00 1.00 2.00 3.00 4.00 5.00 6.00
Rail Profile (m)

Fig. 8. Squat rail profile.

3. Influence of pathologies on the structure dynamics

Despite the maintenance operations conducted over railway tracks and vehicles along their lifecycle, the existence of pathologies in
any of the elements of the vehicle-track-infrastructure system is not uncommon, especially in old steel bridges. On this research, once
calibrated the model, it is analyzed the impact of different pathologies (wheel-rail contact and fatigue cracks) on structural dis­
placements and vibrations. The main objective is to assess to which extent the behavior shown for undamaged tracks is altered by the
considered pathologies at the normal speed of the section (50 km/h).
In the current section, the circulation of the whole vehicle has been simulated by overlapping the results for a single bogie, where
the diphase between two consecutive sets is set in accordance to the distance between consecutive axles. As stated by [15–16], this
approach allows to simulate the whole vehicle accurately while saving computational resources.

3.1. Wheel-rail contact pathologies

For this analysis, two wheel-rail contact pathologies will be considered, rail corrugation and squat.
Rail corrugation is a common defect, that can be simulated as a sinusoidal wave, and therefore it is defined by its amplitude and
wavelength. In the current study, the amplitude of the defect has been set to 1 mm and two different wavelengths are introduced on the
model (0.5 m and 2 m). The following Fig. 7 shows the rail profile for both wavelengths.
In normal conditions, the higher the amplitude of corrugation and the lower its wavelength, the higher the displacements, ac­
celerations and stresses generated by the passage of the vehicle. However, there are other variables that can modify severely that logic.
For instance, vehicle speed has proved to be a variable that should be taken into consideration when it comes to bridge-track-vehicle
dynamics.
The same hypothesis could be applied to the squat. Squat is a punctual rail pathology, which can produce large dynamic overloads.
Its height ranges from micrometers to almost a centimeter. The following Fig. 8 shows in an illustrative manner of how it is the
equivalent rail profile for the studied case.
Random track irregularities are often idealized as stationary random processes, described by Power Spectral Density (PSD)
functions. Different PSD functions have been defined by different railway authorities, generally derived from field measurements and
experimental analysis. The PSD functions describe the severity of the irregularities as a function of the spatial frequency. Some of these
functions have been developed by SNCF, the German railway authorities, the FRA, and some others derived from studies as are the
Chinese PSD ([3] and the ISO 3095 [19] in [4].Theoretical PSD functions have also the shortcoming that they can not reproduce the
isolated defects that are present in real track profiles. Therefore, they tend to produce profiles with less variation in maximum

7
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 9. Simulated Fracture location.

Fig. 10. Results calculated points.

deviation. The standard deviations of the theoretical profiles are scaled to the means of the measured running standard deviations.
As has been previously mentioned, this study pretend to offer a practical approximation of what the effect of irregularities could
generates in railway bridges, is for that reason that, it has been opted for a deterministic analysis, including a single wavelength that
must be analysed at the time. It allows the reader to understand the effect of each individualized wavelength, which will be more
difficult if different wavelengths are integrated in the same track section.

3.2. Structural cracks

The reported bridge is a metal structure built in the early twentieth century executed by the technique of riveting. In this type of
structure, the elements that tend to suffer fatigue are the longitudinal members and joints. The reason is that the circulation of a vehicle
produces a higher number of cycles than presented by the main elements of the structure, being affected, in addition by higher impact
coefficients. This is the reason why, in this type of structures is not difficult to find cracks in similar locations on different bridges, as
have been assessed by railway administration. In this study we will explicitly simulate a crack in the transverse and the longitudinal
elements on a span-central section in order to assess the performance of the bridge under these conditions.
The following Fig. 9 shows schematically the place where the cracks has been simulated.

3.3. Results comparison

Once the different scenarios have been planned and simulated, a comparison of these results will be conducted. The purpose is to
quantify in which condition each of the pathologies presents with respect to the original pattern in terms of structural dynamics.

3.3.1. Displacements
In terms of displacements, the following figures show the deflection of the track in the center of the span, and the deflection of the
studied points shown in Fig. 10.

8
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

0.002

0
0.00 1.25 2.50 3.75 5.00 6.25

-0.002
Deflection (m)

-0.004

-0.006

-0.008

-0.01
Time (s)
0,5m Corrugation 2m Corrugation Longitudinal element Crack Transversal element Crack
Original Case SQUAT Wheel plane

Fig. 11. Point A deflection.

0.002

0
0.00 1.25 2.50 3.75 5.00 6.25

-0.002
Deflections (m)

-0.004

-0.006

-0.008

-0.01
Time (s)
0,5m Corrugation 2m Corrugation Longitudinal element crack Transversal element crack
Original SQUAT Wheel plane

Fig. 12. Point B deflections.

Analyzing the results, can be seen how for the rail, the wheel-rail contact pathologies (corrugations and squat) generate a greater
amplitude of movement with respect to the original case, but it does not mean an increase of the deflection. This same pattern is
repeated for the maximum deflections at the center of the structure.
For the case with the crack on the point E, it is observed a considerable increase of the deflection in the rail (A), reaching to increase
up to 1 mm, this is 12% higher. This pattern, however, is not repeated for the second studied point (B). The is because the mechanical
fracture, the measured point and the load are located at the same vertical. When the measured point is located out of that plane, the
deformations are distributed over the rest of the elements that do resist. The deflections in the elements immediately inferior to the
track will be of a magnitude similar to that presented by the railway (see Fig. 11 and Fig. 12).

3.3.2. Accelerations
When accelerations are analyzed for the studied scenarios (Fig. 13, Fig. 14), a clear pattern is found. On the one hand, analyzing the
accelerations when an element is fractured, an increase of the maximum peaks from 3.3 to 5.5 m/s2 is observed, this is an increase of
80% for the rail and from 1.2 to 1.8 m/s2, (an increase of 50%) for the structure. Something similar happens for the corrugation of 2 m
wavelength, there is an increase of 100% for the rail and a 30% for the structure. Based on this we can observe that the increase in
vibrations due to the fracture is more noticeable in the structure than in the railway, and the opposite for the corrugation of 2 m
wavelength.
However, for the corrugation of 0.5 m, we find a much higher increase in accelerations for both points. It is a 530% higher for the
railway and 1000% higher for the structure. This represents what is described by other authors, where a shorter wavelength is

9
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

20

15

10
Accelerations (m/s2)

0
0. 00 1 .2 5 2.50 3 .7 5
-5

-10

-15

-20
Time (s)
0,5m Corrugation 2m Corrugation Longitudinal element Crack Transversal element Crack Original Case

Fig. 13. Point A accelerations.

15

10

5
Accelerations (m/s2)

0
0.00 1.25 2.50 3.75

-5

-10

-15

-20
Time (s)
0,5m Corrugation 2m Corrugation Longitudinal element crack Transversal element crack Original case

Fig. 14. Point B accelerations.

associated with a greater increase in the dynamic response.

In the case of SQUAT we find that the increase in accelerations is exponential. In the case of the rail values of 1000 m/s2 are reached
(not shown for authors clarity). This increase is also observed for the case of point B.

4. Influence of pathologies on fatigue

Carrying out an evaluation of the fatigue behavior of the structure would be a complex and imprecise task, since the history of loads
and axes that the structure has suffered throughout its history is totally unknown.
Given the difficulty to determine the fatigue life consumed by any certain element, the methodology used to estimate fatigue life is
to calculate the total number of load cycles supported by each proposed scenario with a frozen conditions and compare it with the
section without pathologies. The hypothesis is that all the structure-life cycle is influenced by the same pathology, with the aim of
obtaining the influence of it in the service life of the element in terms of service-life decreasing.
The procedure followed is, first, the total number load cycles that the bridge can support are calculated considering the track in
perfect condition at a certain speed. The cyclic loads considered are the produced by the vehicle used in the data gathering campaign, a
2500 series train, described above. Then, the same parameter is calculated for the different scenarios considered. From these data, it is
obtained the reduction of service life of the studied elements as a cause of the different pathologies considered.
The process for obtaining the remaining fatigue life has been obtained using the nominal stress method and fatigue analysis is based
on Miner’s cumulative damage rule. The stresses have been obtained directly from the finite element methods, and the fatigue life has

10
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

Fig. 15. Fatigue resistance curve for normal stress strokes, fig. 42.6 a of the EAE. Fuente:Fuente especificada no válida.

35,00,000.00 90.00%

80.00%
30,00,000.00

70.00%
25,00,000.00

Lifespan decrement (%)

Supported Load Cycles

60.00%

20,00,000.00 50.00%

15,00,000.00 40.00%

30.00%
10,00,000.00
20.00%

5,00,000.00
10.00%

0.00 0.00%
Original 0,5 m corrugation 2 m corrugation Wheel plane Squat

Fig. 16. Cycles supported by the stringers with the passage of trains at 50 km/h.

been estimated following the nest procedure:

First, the stress history of any point is obtained for any cyclical load (pass of the train), it will allow to define the number and height
of cycles of each passing train. The stressess are obtained directly from the finite element’s method for each studied point.
Then, the stress spectrum is obtained, it is a histogram which classify the different stress ranges involved in any train passing cycle.
The stress spectrum is used for obtaining the number of N supported cycles on the S-N curve for each train pass. To obtain the number
of cicles from the S-N curve (figure 42.6 from the EAE), the details are categorized with as a (71).
Fatigue analysis is based on Miner’s cumulative damage rule. The analysis is reduced to simplify the efforts to which the piece is
subjected and simplify it in a set of simple efforts and analyze the total damage as the sum of the various accumulated damages due to
the different efforts:

11
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

35,00,000.00 90.00%

80.00%
30,00,000.00

70.00%
25,00,000.00

Lifespan decrement (%)

Supported Load Cycles

60.00%

20,00,000.00 50.00%

15,00,000.00 40.00%

30.00%
10,00,000.00
20.00%

5,00,000.00
10.00%

0.00 0.00%
Original 0,5 m corrugation 2 m corrugation Wheel plane Squat

Fig. 17. Cycles supported by the stringers with the passage of trains at 50 km / h.

35,00,000.00 90.00%

80.00%
30,00,000.00

70.00%
25,00,000.00

Lifespan decrement (%)

60.00%
Supported Load Cycles

20,00,000.00 50.00%

15,00,000.00 40.00%

30.00%
10,00,000.00
20.00%

5,00,000.00
10.00%

0.00 0.00%
Original 0,5 m corrugation 2 m corrugation Wheel plane Squat

Fig. 18. Cycles supported by the stringers with the passage of trains at 150 km/h.

∑i
nj n1 n2 n3 ni
= + + +⋯+ ≥1 (9)
j=1
Nj N1 N2 N3 Ni

The center section of the span has been studied, since it will be the most requested due to the load cycles produced by the circulation
of trains.
This process must be carried out for the different studied elements, considering the different sections that correspond in each case
(see Fig. 15).

4.1. Stringers fatigue life estimation from the results of the numerical models.

On this section, the results for the analysis are shown. In the following graph, the calculation has been made at a speed of 50 km/h.

12
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

35,00,000.00

30,00,000.00

25,00,000.00
Supported Load Cycles

20,00,000.00

15,00,000.00

10,00,000.00

5,00,000.00

0.00
Original 0,5 m corrugation 2 m corrugation Wheel plane Squat

50 km/h 90 km/h 150 km/h

Fig. 19. Cycles supported by the stringers with the passage of trains at different speeds and different wheel-rail pathologies. Original.

It can be observed that with the appearance of a 2-meter wavelength wear, which hardly had any implications from a dynamic point of
view, we find a 10% decrease in the useful life. Considering the rest of the defects we find values of decrease between 60 and 90%,
which shows the great influence of these defects in the resistance to fatigue of these elements in metal bridges (see Fig. 16).
When increasing the speed to 90 and 150 km/h we find the same trends as in the case of 50 km/h. There are variations in the
percentages that decrease, but the trend continues in each of them. The data can be observed in the following Figs. 17 and 18.
In the following graph, the number of cycles for each defect is compared with the speed. It is observed that speed is not a factor that
affects both the stresses that the elements are subject to as well as the defects (see Fig. 19).

5. Conclusions

In the current research the dynamic performance of damaged and undamaged vehicle-track-systems built over a multi-span bridge
is studied. This analysis is carried out in terms of displacements and accelerations, which are calculated at different checkpoints of the
track and the structure.
From the first analyses, the following conclusions are drawn:

• displacements of the center of span sections are not significantly affected by the rail pathologies. However, these are significantly
affected in cases of fracture, especially in the longitudinal elements.
• As far as accelerations are concerned, there are very significant increases in all cases. The most notable cases being short wave­
length wear and specially, squat, reaching values several times higher.

Once studied the dynamic behavior of the structure, we proceed to study the affection of these pathologies to fatigue life of the
elements. The study of the fatigue life of the different elements is clearly influenced by any pathology that appears. In long-standing
bridges, the need for periodic inspections is obvious, since the different elements will be influenced by numerous pathologies due to
their high age.
The great influence of the pathologies in the life to fatigue of the elements becomes noticeable, much more significant of what is the
speed.
It is worth noting the great influence of short wave undulatory wear and squat. This fact shows that there is a great relationship
between the increase of vibrations and the reduction of supported cycles.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

13
F. Ribes-Llario et al. Engineering Failure Analysis 119 (2021) 105007

References

[1] G. Alencar, A. de Jesus, J.G. da Silva, R. Calçada, Fatigue cracking of welded railway bridges: A review, Eng. Fail. Anal. 104 (2019) 154–176.
[3] A.R. Berawi, Improving railway track maintenance using power spectral density (PSD), 2013.
[4] E.G. Berggren, A new approach to the analysis and presentation of vertical track geometry quality and rail roughness, Wear 265 (9–10) (2008) 1488–1496.
[5] D.A. Cantero, Train–track–bridge modelling and review of parameters, Struct. Infrastruct. Eng. 12 (9) (2016) 1051–1064.
[6] S.M. Dick, R.J. Connor, D. Otter, J. Lloyd, Cracks in welded steel girders on the steel bridge at FAST, AREMA Annu. Conf., 2013, 711–728.
[7] J. Dominguez, Railways bridges dynamics for high-speed railways: calculus methods and study of resonance, 2001.
[11] L. Mao, Y. Lu, P. Woodward, Frequency characteristics of railway bridge response to moving trains with consideration of train mass, Eng. Struct. 42 (2012)
9–22.
[13] P. Museros, Vehicle-structure interaction and resonance effects in isostatic railways bridges for high-speed lines, 2002.
[15] J. Real, C. Zamorano, R. Comendador, T. Real, Computational considerations on 3-D finite element method models of railway vibration prediction in ballasted
tracks, J. Vibroengineering 16 (4) (2014) 1709–1722.
[16] J. Real, C. Zamorano, C. Hernandez, R. Comendador, T. Real, Computational considerations of 3-D finite element method models of railway vibration prediction
in ballasted tracks, J. Vibroengineering 16 (4) (2014) 1709–1722.
[18] Y. Yang, C. Lin, Vehicle bridge interaction dynamics and potential applications, J. Sound Vib. (2005) 247–259.
[19] Standard, I. S. O. (2013). 3095: 2013. Acoustics–Railway applications–Measurement of noise emitted by railbound vehicles, Geneva.

14