The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/1757-9864.htm

Fatigue
Fatigue assessment based on assessment
hot-spot stresses obtained from based on hot-spot
stresses
the global dynamic analysis and
local static sub-model
Cristiane Oliveira Viana Received 21 March 2019
Revised 14 August 2019
Federal University of Minas Gerais, Belo Horizonte, Brazil Accepted 20 August 2019

Hermes Carvalho
Department of Structural Engineering,
Federal University of Minas Gerais, Belo Horizonte, Brazil
José Correia
Department of Civil Engineering,
University of Porto Faculty of Engineering, Porto, Portugal
Pedro Aires Montenegro
University of Porto Faculty of Engineering, Porto, Portugal
Raphael Pedrosa Heleno
Federal University of Minas Gerais, Belo Horizonte, Brazil
Guilherme Santana Alencar
University of Porto Faculty of Engineering, Porto, Portugal
Abilio M.P. de Jesus
Department of Mechanical Engineering,
University of Porto Faculty of Engineering, Porto, Portugal, and
Rui Calçada
University of Porto Faculty of Engineering, Porto, Portugal

Abstract
Purpose – The purpose of this paper is to evaluate the fatigue process through the dynamic analysis of the
global structural model and local static sub-modelling in a critical detail using the hot-spot stress approach.
The detail was studied in three different positions at the “Alcácer do Sal” access viaduct, and the
methodologies from the IIW and Eurocode EN 1993-1-9 were compared.
Design/methodology/approach – In this study, the fatigue life process based on the hot-spot stress approach
was evaluated using a global dynamic analysis and a local sub-modelling based on a static analysis of welded
connections in the “Alcácer do Sal” railway structure, Portugal, taking into consideration the recommendations
from IIW and Eurocode EN 1993-1-9. The hot-spot stresses were calculated through the static analysis of the
sub-model of the welded connection for each vibration mode with the aim to obtain the temporal stresses using
the modal coordinates and modal stresses of the extrapolation points. The Ansys® and Matlab® softwares were
used for the numerical analysis and the hot-spot stress calculations, respectively.
Findings – The proposed methodology/approach to obtain fatigue assessment is based on the modal analysis
of the global structural model and local static sub-modelling. The modal analysis was used to extract the
boundary conditions to be used in the local model to determine the temporal stresses of the extrapolation points.
Based on the modal superposition method, the stresses as function of time were obtained for fatigue life
evaluation of a critical detail by the hot-spot stress approach. The detail was studied in three different positions.
Originality/value – In the present study, a global-local fatigue methodology based on dynamic analysis of the
global structural model and local static sub-modelling of the critical detail using the hot-spot stress approach is International Journal of Structural
proposed. Herein, the modal analysis of the global structural model supported by the modal superposition Integrity
© Emerald Publishing Limited
method was used to obtain the matrix of modal coordinates. The static analysis of the local sub-model for each 1757-9864
mode from the modal analysis of global structural model was done to estimate the hot-spot stresses. The fatigue DOI 10.1108/IJSI-03-2019-0021
IJSI damage calculation was based on S-N curve of the critical detail and rainflow method. The IIW recommendation
proved to be more conservative compared to the proposed rules in the Eurocode EN 1993-1-9. The global-local
modelling based on dynamic analysis is an important and effective tool for fatigue evaluation in welded joints.
Keywords Fatigue, Dynamic analysis, FEM analysis, Hot-spot stress, Local static sub-modelling,
Railway bridges, Welded connections
Paper type Research paper

1. Introduction
Railway bridges are constantly subjected to variable loads resulting from the passage of
trains, in addition to permanent loading. Thus, in this type of structure, the occurrence of
fatigue failures is common. In railway composite bridges with welded details, the fatigue
damage evaluation is made based on the hot-spot stress approach according to the IIW
(2008) recommendations and/or design rules proposed in Eurocode: European Committee for
Standardization – EN 1993-1-9:2005 (2005). Some aspects of fatigue are especially pertinent
to bridges of metallic and composite structures of steel and concrete (Nussbaumer et al.,
2011). The dynamic loads due to high-speed train passage can cause dynamic amplification
and thus increase fatigue damage (Alencar and Ferreira, 2018).
In this context, the present work analyses the fatigue failure process caused by the
dynamic loads in a critical detail of the access viaduct of the Sado railway structure in
“Alcácer do Sal”, Portugal. The access viaduct consists of a mixed structure of steel and
concrete, connecting the North viaduct to the bridge over the Sado River, as shown in
Figure 1. The steel used in the viaduct construction was S355NL; the concrete used
presented resistance class C40/50 and the steel reinforcement was CA-50 with yield limit
stress equal to 500 MPa (Ferreira, 2017).
The most usual fatigue approach uses the nominal stress and the classification of
geometric details to predict fatigue life, not covering connections that are more complex and
limited to the geometries previously tested and presented in the codes. Therefore, the
hot-spot stress approach, also known as structural stress or geometric stress approach,
presents a wider range of welded connections not included in the codes as it is based on
numerical simulations build on the finite element method (Heshmati, 2012). In this paper, in
addition to the hot-spot approach, a deterministic method was used to assess fatigue based
on the stress ranges. Other deterministic methods like the strain-life approach (Liu et al.,
2019) and the Fracture-Mechanics-based approach (Correia et al., 2016; De Jesus et al., 2011)
were also suggested in the literature. Jakubczak et al. (2006) proposed an algorithm for the

45 m

Figure 1.
Railway composite
bridge in Alcácer do
Sal on the Sado River
(Portugal) Sources: Alencar, Ribeiro, Calçada and Silva (2016) and Reis and
Pedro (2013)
simulation based fatigue life assessment approaches considering the uncertainties from the Fatigue
fatigue material properties, loading and structural components. assessment
Fatigue analysis of structural connections from numerical models, including the analysis based on hot-spot
based on the hot-spot stress approach, generally requires the construction of refined
numerical models according to IIW (2008) recommendations. However, the numerical stresses
models based on the finite elements analysis using shell and solid elements have a high
computational cost. As the use of these elements over the entire length of the structure is
very costly, the use of a sub-modelling methodology is justified. The sub-modelling
technique is based on the principle of St Venant, where the stress field of a region far from
the stress concentrations is replaced by a set of equivalent displacements. However, the
stress and strain distribution only suffers significant alterations in regions with notches and
geometry changes. Such effects of stress concentration are localized. In case of the
sub-model in which the boundaries are far enough from the stress concentration, the
sub-model can provide accurate results (Ansys Inc, 2017).
There are several advantages of using the sub-modelling technique, as mesh transition
zones are not necessary in the computational model of the structural detail. The
sub-modelling analysis is used in the detailed assessment of specific regions of a structural
model. This methodology includes a global model that represents the entire system under
study and a sub-model of the detail evaluated. According to Leite et al. (2018), the global
structural model is analyzed using beam or shell finite elements, whereas specific regions
may be studied through models with shell or solid finite elements including mesh
refinement. This technique is important to ensure a good transfer of the boundary
conditions between the different models used.
According to Reis et al. (2010), the structure design under evaluation, especially in the
central regions of the spans, was managed by the fatigue limit state. In this way, the steel
cost of the structure was directly related to the fatigue verifications and a more conservative
verification method may significantly impact the cost of the work. On the contrary, it is
necessary to ensure the safety of the designed structure, in such a way as to ensure that no
cracks occur in the structure before it has reached the end of its useful life.
In the present study, fatigue life was assessed based on hot-spot stresses obtained from a
global dynamic analysis and local static sub-model in a critical detail of the access viaduct of
a high-speed railway bridge in “Alcácer do Sal” (Portugal). The hot-spot stresses are
estimated through the static analysis of the sub-model of the welded connection for each
vibration mode with the aim of obtaining the temporal stresses using the modal coordinates
and modal stresses of the extrapolation points. Additionally, the extrapolation points are
defined according to IIW (2008) recommendations and design rules presented in Eurocode:
European Committee for Standardization – EN 1993-1-9:2005 (2005) standard. The
high-speed train used for the dynamic analysis is the “Alfa Pendular” passenger train.
Nowadays in Europe, railways are commonly used in public transportation, and high-speed
trains have many advantages, such as low emission of carbon dioxide and fast travel. This
study may be useful in fatigue designs and in residual fatigue life assessments of railway
infrastructures, especially high-speed trains.
When it comes to fatigue in bridges situated on high-speed lines, there are other
important factors to be considered, such as the effect of dynamic amplification on the
structure. The dynamic amplification effects can significantly decrease the fatigue life of the
structural components when the bridge is excited at certain speeds (Alencar, Ribeiro,
Calçada and Silva, 2016). To evaluate the need of dynamic analysis for a bridge project, the
flowchart presented in Eurocode: European Committee for Standardization – EN 1991-2
(2003) was used as reference. The parameters used in the flowchart are railroad maximum
speed, V (km/h); span, L (m); natural frequency of the first vertical bending mode, n0 (Hz);
and, natural frequency of the first torsion mode, nt (Hz). In this case study, the maximum
IJSI railroad speed is greater than 200 km/h and the structure is simply supported with 45 m of
span length. The natural frequency of the first vertical bending mode (2.48 Hz) corresponds
exactly to the lower limit frequency value for a span of 45 m (2.48 Hz). In this case, the global
dynamic analysis is considered in the case study as the fatigue limit of the natural frequency
is within the regulatory limits, even though very near its boundaries, as recommended in
Eurocode: European Committee for Standardization – EN 1991-2 (2003). The frequency of
the first torsion mode (3.05 Hz) is greater than 1.2 times the frequency of the first bending
mode (1.2×2.48 ¼ 2.98 Hz). Therefore, this case study requires a dynamic evaluation and
should considerer both bending and torsion modes in the proposed fatigue methodology.
For this reason, the present work evaluates the fatigue life according to the demands on
the bridge for the traffic of high-speed trains, encompassing global dynamic effects as well
as considering a global-local numerical modelling. Recently, several authors have used the
global-local methodology for fatigue evaluation in critical details. A fatigue study through
the sub-modelling methodology of the central span of the “Alcácer do Sal” bridge over the
Sado River in Portugal was performed by Albuquerque (2015). The global model was
developed in shell elements, whereas the sub-model was developed in solid elements.
Fatigue assessments in a critical detail in the “Alcácer do Sal” railway structure were
performed by Alencar and Ferreira (2018) and Ferreira (2017). Fatigue evaluation of a
critical connection of the historical bridge Hercílio Luz, located in “Florianópolis”, Brazil,
was performed by Liu et al. (2019) through the global-local methodology. According to
Leite et al. (2018), a sub-modelling analysis was performed of a riveted detail present on a
metal bridge, the overall bridge model being developed with beam elements and the local
model with solid elements. The compatibility of displacements was made through rigid
elements, connected to reference nodes.

2. Structure description
The “Alcácer do Sal” composite bridge was built in 2010 and is part of the railway link
between Algarve and Lisbon (Portugal), to transport passengers using high-speed trains.
This bridge was designed for speeds of up to 250 km/h and is also part of the logistics of
good flows through the Port of Sines, on the Atlantic coast (Albuquerque, 2015; Ferreira,
2017). Plate 1 presents a view of the “Alcácer do Sal” bridge.
The structure is composed by the central span of the “Sado” River and the north and
south access viaducts, totalling a length of 2700 meters, with the north viaduct and the
central span straight and the south viaduct in a curve. Most of spans of the structure are

Plate 1.
View of the railway
composite bridge in
“Alcácer do Sal”
Sources: Mourão (2010) and Alencar, Calçada and Silva (2016)
continuous, with few interposed simple supported spans (Ferreira, 2017). This study focuses Fatigue
on a simple supported span of the North viaduct, as previously shown in Figure 1, and the assessment
main dimensions of the cross section being shown in Figure 2. based on hot-spot
stresses
3. Computational model description
The fatigue life evaluation was performed using numerical models developed in the ANSYS
Mechanical APDL software, simulating a passage of the “Alfa Pendular” train over the
bridge span through the sub-modelling dynamic analysis, thus enabling refining regions of
critical details for fatigue. The modal superposition was used in the dynamic solution of the
problem, and the calculation of fatigue life in the sub-model was performed by the hot-spot
stress method, considering quadratic extrapolation and according to the recommendations
of Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005) and IIW
(2008). In addition, fatigue damage was evaluated.

3.1 Global model

The complete model of the access viaduct was developed in the ANSYS Mechanical APDL
program, based on the Finite Element Method, using “shell181” shell elements, which have
four nodes and six degrees of freedom in each node – translation in x, y, z and rotation
around these three axes. This element is suitable for analysis in thin to moderately thick
structural elements. In the numerical model, shell elements were considered in the middle
plane of the steel plates. The following material properties were used in the analysis: steel
young modulus 210 GPa, steel density 7850 kg/m3, steel Poisson coefficient 0.3, concrete
young modulus 35 GPa, concrete density 2500 kg/m3 and concrete Poisson coefficient 0.15.
The compatibility of displacements between top flange and slab was developed through
rigid links composed of “link180” elements – link elements (uniaxial traction and
compression) with two nodes and three degrees of translation per node – with high
transverse area (106 m2) and low density (10−9 kg m−3) in order to neither influence the total
mass of the model nor the natural frequencies. The masses of the non-structural elements
presented on the slab such as rails, poles for illumination and ballast, were represented by
elements of mass “mass21” distributed in the nodes of the slab. The mass elements were
considered as three-dimensional and without rotational inertia, with the equivalent mass
applied in the y direction (gravity direction). Figure 3 is the overall model of the bridge.
The boundary conditions of the span were implemented as described in
Malveiro et al. (2018), with the two supports on the South side (direction of positive z)

1,585
>257
100 167 225 225
200 225 100 225 125 125
25 50 25 100
83
20 60

1,570 260
Figure 2.
Cross-section of the
350 350
simple supported span
of the North viaduct
Source: Alencar, Calçada and Silva (2016)
IJSI (a) (b)

Figure 3.
View of the structural
model of the access
viaduct
Notes: (a) Global model; (b) local model

restricted in x, y and z, a support on the North side restricted in x and another restricted in y
only, as shown in Figure 4.
The train selected to perform the dynamic analysis was the CPA 4000 series train, also
known as “Alfa Pendular”. This train went into circulation in 1999, connecting Lisbon with
Porto, Portuguese cities. “Alfa Pendular” is an electric train with an active pendulum system
with six wagons (Figure 5), totalling a length of 158.9 m and maximum speed equal to
220 km/h (Ribeiro, 2012).
The type of load consideration over the selected span was the mobile load, due to its
simplicity when compared to more complex models, which take into consideration the train-
structure interaction. It is important to note that the mass of the train over the bridge is not

UX, UY, UZ UX, UY, UZ

South Side

Z
Train direction: Z– to Z+
Y X

Track location: X+

Figure 4.
Boundary conditions
of the bridge and
direction of train North Side
movement considered
in the analysis
UX, UY UY
considered in this method, which may lead to an overestimation of the natural frequencies, Fatigue
nor the damping effect of the train suspension (Albuquerque, 2008). assessment
In order to implement the analysis with mobile loads, a matrix was elaborated containing based on hot-spot
the values of nodal forces, according to the load and speed of the “Alfa Pendular” train. The
distribution of the load along the train is shown in Figure 5. The matrix of the nodal forces stresses
was created from a code made in Matlab software. Each column of this matrix contains the
equivalent train loads for each time instant (row) and for each node over the bridge (column).
The considerations made are presented in Table I. In the computational model in ANSYS,
two lines represent the rails, the lines being divided into the positions where the forces are
applied and have equal space of 0.1 m between them, as the application of the forces occurs
directly on the nodes.
The time steps of Table I, for all speeds, were lower than those stipulated by the
recommendations of ERRI D214/RP9 (2001) and Ribeiro (2004). A wide range of speeds from
140 to 440 km/h was selected to evaluate the dynamic behaviour of the bridge, and this
range included the maximum train operating speed – 220 km/h (Ferreira, 2017) and the
bridge design maximum speed – 250 km/h (Alencar, Ribeiro, Calçada and Silva, 2016).
The flowchart describing the complete process is presented in Figure 6, and it presents
the steps of the sub-modelling methodology used to evaluate the fatigue process. The global
model is created, and the modal analysis is simulated. The boundary conditions are obtained
from each mode and they are applied in the sub-model. Modal stresses were obtained from
the results of the sub-model static analysis and they were converted in temporal stresses by
multiplying for the modal matrix coordinates. Hot-spot stresses were calculated for fatigue
assessment, using the Rainflow method and the accumulated linear damage theory.

3.2 Sub-model for fatigue evaluation

The sub-model is presented in Figure 7. It was chosen according to the critical detail, taken
from the stress range for fatigue and is not among the possible details of evaluation by the
nominal stress method. Therefore, evaluation of fatigue life of this detail becomes unfeasible
with only the codes used in steel structure projects. The dimensions considered for the local

3.85 2.7 16.3 2.7 4.2 2.7 16.3 2.7 4.2 2.7 16.3 2.7

132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8

Figure 5.
2.7 16.3 2.7 4.2 2.7 16.3 2.7 4.2 2.7 16.3 2.7 3.85 Distribution of the
“Alfa Pendular” axle
132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 132.8 load – forces in kN
and distance in meters
Sources: Alencar and Ferreira (2018)

Train characteristics Alfa Pendular

Longitudinal spacing between e ¼ 0.1m Table I.

rail nodes, e (m) Parameters used for
Train speeds, v (km/h) v from 140 to 440 km/h (10–10 km/h) analysis with
Time steps considered, Δt (s) Δt ¼ 3.6⋅e/v, if v ¼ 140–190 km/h; Δt ¼ 2⋅3.6⋅e/v, if v ¼ 200–440 km/h mobile loads
IJSI
Modal analysis Static analysis for each mode Rainflow

Modal Superposition Modal stress xM, zM, xzM FATcor = FAT/m

Matrix of modal coordinates Temporal stress x, z, xz Selection of S-N Curve

Figure 6.
Flowchart of dynamic
Hot Spot x HS, z HS, xz HS  n/N = D
sub-modelling and
fatigue evaluation
Global Model Sub-model Fatigue analysis

Figure 7.
Sub-model modelled
for the
fatigue evalution

model were based on the project (GRID et al., 2006), with the implementation of some
adaptations necessary for the boundaries of the local model to be centralized to the elements
of the global model. The contour was positioned far from the detail, ensuring that there were
no significant changes in the distribution of stresses and deformations between global and
local models in the contour region, validating the sub-model analysis. The boundary
conditions in both models varied according to the discretization method employed, as
previously described.
Regarding the detail in this work, the IIW (2008) and Eurocode: European Committee for
Standardization – EN 1993-1-9:2005 (2005) recommend a category of detail (FATRef ) equal
to 100 MPa, but the present study contemplated a lower detail category, equal to 90 MPa, Fatigue
considering that the detail is subjected to multi-axial loading and the weld is classified with assessment
moderate/low tensile stresses. based on hot-spot
stresses
3.3 Fatigue assessment methodology
The hot-spot stresses in x – σxHS, y – σyHS and xy – τxyHS directions are calculated based on
Equation (1). The stresses σ0.4t, σ0.9t e σ1.4t are the ones at the quadratic extrapolation points
according to IIW (2008). Subsequently, the principal stress in function of time – σ1 is
calculated according Equation (2):

sH S ¼ 2:52 s0:4t 2:24s0:9t þ0:72s1:4t ; (1)

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s 2
sx HS þsy H S x H S sy H S
s1 ¼ þ þt2xy H S : (2)
2 2
The values of stress ranges (Δσ), to be applied in damage calculation, are obtained using the
Rainflow method. The detail category (FAT) for the evaluated component is corrected by
dividing the FAT by the partial safety factor according to Equation (3). The corrected detail
category (FATcor) indicates the S-N curve to be used. From the S-N curve, the service life (N )
corresponding to each stress range obtained in the Rainflow method (Δσ) is calculated. The
occurrence number of each stress range (n) divided by the corresponding service life (N )
represents the damage fraction associated with this stress range and the summation of all
damage fractions related to all stress ranges is equal to the total damage (D), as shown in
Equation (4):

FAT cor ¼ FAT= gm ; (3)

Xn
ni
¼ D: (4)
i¼1
N i

4. Results and discussion

4.1 Modal analysis and model validation
The natural frequencies obtained from the modal analysis of the access viaduct are
presented in Table II, and the values compared with the frequencies determined by Ferreira
(2017) are presented in Figure 8. The ratio of effective mass to total mass using 100 modes
is 0.79998 in x – transversal direction, 1.31983 in y – vertical direction, and 0.79217 in
z – longitudinal direction.
The maximum difference between frequencies was 8.6 per cent and for the first five most
significant modes the difference reduced to 1.5 per cent, indicating a significant similarity
between the models. The frequencies obtained in the present work are, in general, slightly
higher than those presented by Ferreira (2017), indicating that the model presents greater
rigidity or lower mass assigned. The first three modes obtained in this work and in Alencar,
Ribeiro, Calçada and Silva (2016) are presented below (Figures 9–11).
The first three modes obtained in this work and in Alencar, Ribeiro, Calçada and Silva
(2016) coincide in their description, but they present small behavioural differences, possibly
due to differences in the details of the geometries of the models, which influence the rigidity
of both, or by small differences in mass considerations. However, such differences were not
considered significant, and the modelling of the bridge was representative.
IJSI Mode no. Frequency (Hz) Mode no. Frequency (Hz) Mode no. Frequency (Hz) Mode no. Frequency (Hz)

1 2.4837 26 11.822 51 19.345 76 23.575

2 3.0522 27 12.011 52 19.409 77 24.443
3 3.5933 28 12.401 53 19.664 78 24.721
4 5.3137 29 12.857 54 19.712 79 25.195
5 5.5007 30 13.772 55 20.226 80 25.558
6 6.0281 31 13.797 56 20.538 81 26.224
7 6.5516 32 14.024 57 20.597 82 26.429
8 7.2245 33 15.171 58 20.619 83 26.591
9 7.6185 34 15.228 59 20.671 84 26.795
10 8.1107 35 15.249 60 20.756 85 27.035
11 8.3667 36 15.267 61 20.96 86 27.237
12 8.7675 37 15.288 62 21.009 87 27.328
13 8.7781 38 15.344 63 21.037 88 27.417
14 8.779 39 15.758 64 21.163 89 27.591
15 8.7804 40 15.88 65 21.22 90 27.879
16 8.7916 41 16.09 66 21.314 91 27.96
17 8.7925 42 16.417 67 21.645 92 28.008
18 8.8052 43 17.713 68 21.883 93 28.147
19 9.7519 44 17.729 69 21.904 94 28.185
20 10.099 45 17.904 70 22.005 95 28.353
Table II. 21 11.474 46 17.949 71 22.115 96 28.773
Frequencies obtained 22 11.552 47 18.378 72 22.176 97 28.975
in the analysis of the 23 11.585 48 18.435 73 23.01 98 29.341
global model of 24 11.621 49 18.509 74 23.462 99 29.678
the bridge 25 11.635 50 18.844 75 23.519 100 29.695

30

25
Frequency (Hz)

20

15 Frequencies
10
Figure 8. Frequencies Ferreira
Comparison of 5 (2017)
frequencies with those
presented by 0
1 10 19 28 37 46 55 64 73 82 91 100
Ferreira (2017).
Mode Number

Figure 9. Y

First mode of
vibration in Alencar,
Ribeiro, Calçada and
Silva (2016) and
in the present
work, respectively
Note: Global bending
4.2 Dynamic analysis of the global model Fatigue
Dynamic analyses were performed based on the developed global model. The graphs of assessment
Figures 12 and 13 show the maximum vertical displacements and accelerations at the centre based on hot-spot
of the span as a function of train speed. Figure 12 illustrates maximum vertical displacement
and Figure 13 shows maximum vertical acceleration of the span. stresses
It was observed that there was a significant amplification in the values of vertical
displacements and accelerations at a speed close to 220 km/h due to the resonance effect.
However, vertical accelerations are always less than 3.5 m/s2, according to Eurocode:
European Committee for Standardization – EN 1990 (2001). Figures 14 and 15 illustrate the
dynamic response at the centre of the span as a function of time for two speeds, one speed
without the resonance effect, 140 km/h and the other with the effect, 220 km/h.
The graphs shows that the amplification phenomenon cannot be ignored, and the
differences between the peaks of displacements and accelerations indicate significant
variations in the damage values according to train speed. The damage results are presented
in the next section.

Figure 10.
Y

Z X Second mode of
vibration in Alencar,
Ribeiro, Calçada and
Silva, (2016) and in
the present
work, respectively
Note: Global torsion

Y
Figure 11.
Z X Third mode of
vibration in Alencar,
Ribeiro, Calçada and
Silva (2016) and in the
present work,
respectively
Note: Global torsion

0.0510

0.0505
Vertical displacement (m)

0.0500

0.0495

0.0490

0.0485
Figure 12.
0.0480 Maximum absolute
vertical displacement
0.0475 at the mid span in
140 190 240 290 340 390 440
function of train speed
Train speed (km/h)
IJSI 1.2

1.0

Vertical Acceleration (m/s2)

0.8

0.6

0.4

Figure 13. 0.2

Maximum vertical
acceleration at the
mid span in function 0.0
140 190 240 290 340 390 440
of train speed
Train Speed (km/h)

–4.10E–02
Train Speed 140 km/h
Train Speed 220 km/h
–4.30E–02
Vertical Displacement (m)

–4.50E–02

–4.70E–02

Figure 14. –4.90E–02

Vertical displacement
at the mid span in
function of –5.10E–02
train position 0 50 100 150 200
Position (m)

1.25
Train Speed 140 km/h
Train Speed 220 km/h
0.75
Vertical Acceleration (m/s2)

0.25

–0.25

Figure 15. –0.75

Vertical acceleration
at the mid span in
function of –1.25
0 50 100 150 200
train position
Position (m)

4.3 Determination of the critical speeds of train passage

Preliminary damage assessment was performed for all speeds studied, between 140 and 440
km/h, and for all positions under study considering a reduced free vibration time. This
evaluation allows the selection of the most critical speeds for fatigue, i.e., the speeds that
result in a greater amplitude of stress range. Once the critical velocities are identified, Fatigue
complementary simulations taking into consideration these speeds and a longer free assessment
vibration time, equal to 5 s, will indicate the maximum values obtained from a dynamic based on hot-spot
analysis. The damage values regarding the passage of an “Alfa Pendular” train on the
bridge as a function of train speed for the three different positions of the detail are presented stresses
in accordance with the recommendations of the IIW (2008) and Eurocode: European
Committee for Standardization – EN 1993-1-9:2005 (2005), respectively. The three positions
of the detail correspond to: the detail closest to the support of the section; the detail more to
the centre of the span; and the detail intermediate to the previous two (Figures 16 and 17).
It is noteworthy that the region where maximum dynamic amplification occurs is similar
in the three positions evaluated, with the observed peak of damage at a speed of 220 km/h. It
is interesting to note that the same is true for the two standards under review, Eurocode:
European Committee for Standardization – EN 1993-1-9:2005 (2005) and IIW (2008). The five
most critical speeds identified are 190, 210, 220, 230 and 240 km/h and the most critical
position is at the centre. New tests were performed considering the five most critical speeds
and a longer time of free vibration, 5 s. The comparison between the damage calculated
through Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005) and
IIW (2008) is presented in Figure 18.
It can be observed from the values presented that recommendation IIW (2008) in this
situation is more severe than Eurocode: European Committee for Standardization – EN
1993-1-9:2005 (2005). The partial safety coefficient of IIW (2008) is equal to 1.30 and lower

9.0E–07
8.0E–07
7.0E–07
6.0E–07
Damage

5.0E–07
4.0E–07 Detail close to support
3.0E–07 Interjacent detail Figure 16.
2.0E–07 Central detail Damage due to train
1.0E–07 speed “Alfa Pendular”
according to the IIW
0.0E+00 (2008), considering 100
140 160 180 200 220 240 260 280 300 320 340 360 380 400 420 440
years of service life
Train speed (km/h)

6.0E–07

5.0E–07

4.0E–07 Figure 17.

Damage

Detail close to support Damage due to train

3.0E–07
Interjacent detail speed “Alfa Pendular”
according to
2.0E–07 Central detail Eurocode: European
Committee for
1.0E–07 Standardization – EN
1993-1-9:2005 (2005),
0.0E+00 considering 100 years
140 160 180 200 220 240 260 280 300 320 340 360 380 400 420 440
of service life
Train speed (km/h)
IJSI than the Eurocode coefficient (Eurocode: European Committee for Standardization – EN
1993-1-9:2005, 2005), which is equal to 1.35, but the S-N curve presented by IIW (2008) has a
more conservative profile and considers lower stress ranges. Another important detail is
that in order to compare the values between both methodologies and to normalize all data,
the damage obtained by IIW (2008) is multiplied by 2, always considering the maximum
damage equal to 1. In IIW (2008), the evaluated case would fail when total damage reached
0.5, which corresponds to multiplying the damage by 2 and considering the limit value for
damage equal to 1, equal to that of Eurocode: European Committee for Standardization – EN
1993-1-9:2005 (2005).
Fatigue life estimation of the most critical detail was performed considering the following
loads: light traffic, with two “Alfa Pendular” trains passing over the bridge per day; medium
traffic, with six “Alfa Pendular” trains going per day; and heavy traffic, with the passage of
ten trains per day; as proposed by Alencar, Ribeiro, Calçada and Silva (2016). The speeds
tested were the three most critical to fatigue life: 210, 220 and 230 km/h. The results obtained
are presented below (Figures 19 and 20).
It is observed that the shortest useful life corresponds to heavy traffic at the speed of
220 km/h, as expected. Using the IIW (2008) formulas, estimated life expectancy is
estimated to be 210 years and using the Eurocode: European Committee for
Standardization – EN 1993-1-9:2005 (2005) formulas, estimated life expectancy is
estimated to be 326 years. There is a significant drop in working life with increased train
flow. It is important to point out that this is only a comparative study and does not

1.4E–06
1.2E–06

Figure 18. 1.0E–06

Damage values for
Damage

8.0E–07 IIW
passing an “Alfa
Eurocode
Pendular” train over 6.0E–07
the bridge at the most
critical speeds for the 4.0E–07
detail located closest 2.0E–07
to the span centre and
considering 100 years 0.0E+00
190 210 220 230 240
of service life
Speed (km/h)

4,400
4,000
3,600
3,200
210 km/h
2,800
Life (years)

2,400 220 km/h

2,000 230 km/h
Figure 19. 1,600
Estimated life by the 1,200
IIW calculation
methodology (2008) 800
for the detail near the 400
centre of the bridge, 0
considering different Light traffic Medium traffic Heavy traffic
traffic and speeds
Source: IIW calculation methodology (2008)
 4,400 Fatigue
4,000 assessment
3,600
based on hot-spot
3,200
210 km/h
stresses
2,800
Life (years)

2,400 220 km/h

2,000 230 km/h
1,600
1,200 Figure 20.
800 Estimated life by the
Eurocode EN 1993-1-
400 9:2005 (2005) for the
0 detail near the centre
Light traffic Medium traffic Heavy traffic of the bridge,
considering different
Source: Eurocode: European Committee for Standardization – EN 1993- traffic and speeds
1-9:2005 (2005)

correspond to the reality of the bridge, as other types of trains, such as conventional and
freight transport, pass over it. This study also does not take into account the progressive
increase in train flow over the years.

5. Conclusions
Evaluation of the modal analysis of the access viaduct showed that the frequencies obtained
in the present study, when compared with the frequencies obtained by Ferreira (2017),
presented a maximum difference of 8.6 per cent, and for the first five most significant modes,
the difference was reduced to 1.5 per cent. The natural frequencies of vibration in the
present work was generally slightly higher than those presented by Ferreira (2017),
indicating greater rigidity or lower mass attributed than the model by Ferreira (2017).
Evaluation of fatigue damage considering quadratic extrapolation and reduced free
vibration time was performed in the range of speeds between 140 and 440 km/h. It was
observed that the most significant and critical damages were obtained at speeds of 190, 210,
220, 230 and 240 km/h, speeds at which the highest dynamic amplification occurs, as at these
speeds, the train excites the access viaduct in frequencies close to the natural frequency of the
first mode of span vertical bending vibration. The dynamic modal superposition analysis was
again performed for each of the critical speeds, using a free vibration time of 5 s.
Three detailed positions along the bridge were evaluated: the position closest to the
support, the closest to the centre, and the intermediate position of the two. The most critical
damage was always identified in the detail positioned near the centre of the bridge, where
the greatest longitudinal stress range is expected.
Fatigue life was calculated using a hot-spot stress approach as found in IIW (2008) and in
Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005). It was
observed that the IIW (2008) recommendation generated more severe results than those of
Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005). Although
the partial safety coefficient of (IIW, 2008) is slightly lower than that of Eurocode: European
Committee for Standardization – EN 1993-1-9:2005 (2005) – 1.30 and 1.35, respectively, the
S-N curve of IIW (2008), for the same category of detail, is more conservative than the
S-N curve of Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005).
In addition, the maximum damage recommended by IIW (2008) is half that presented by
Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005) – 0.5 and 1.0,
which significantly influences fatigue life evaluation.
IJSI Three different flows of trains were considered for the evaluation of fatigue life: flow with
light traffic, which the passage of two “Alfa Pendular” trains per day; medium traffic, with the
passage of six trains per day; and heavy traffic, with the passage of 10 “Alfa Pendular” trains
per day. Among the evaluated evaluations, the shortest fatigue life was detected in the closest
detail to the centre and with train passage 220 km/h. Calculated life estimation due to fatigue
was 210 years by IIW (2008) and 326 years by Eurocode: European Committee for
Standardization – EN 1993-1-9:2005 (2005). Increased train flow significantly reduced fatigue life.

Acknowledgments
This work was financially supported by: national funds through FCT – Fundação para a
Ciência e a Tecnologia; UID/ECI/04708/2019 – CONSTRUCT – Instituto de I&D em Estruturas
e Construções funded by national funds through the FCT/MCTES (PIDDAC); post-doctoral
grant SFRH/BPD/107825/2015 provided by FCT to the first author; NORTE-01-0145-FEDER-
000022 – SciTech – Science and Technology for Competitive and Sustainable Industries,
co-financed by Programa Operacional Regional do Norte (NORTE2020), through the Fundo
Europeu de Desenvolvimento Regional (FEDER); IN2TRACK2 – Research into enhanced track
and switch and crossing system 2 funded by European funds through the H2020
(SHIFT2RAIL Innovation Programme); and Institute of Construction (IC, University of Porto).
Additionally, this research work was also supported by CAPES (Coordination for the
Improvement of Personnel of Graduation), CNPQ (National Council for Scientific and
Technological Development) and UFMG (Universidade Federal de Minas Gerais).

References
Albuquerque, C.M.C. (2008), “Comportamento Dinâmico de Pontes com Tabuleiro Ortotrópico em Vias
de Alta Velocidade”, Dissertação de Mestrado , Engineering Faculty, University of Porto, Porto.
Albuquerque, C.M.C. (2015), “Advanced methodologies for the assessment of the fatigue behavior of
railway bridges”, doctor’s thesis, Engineering Faculty, University of Porto, Porto.
Alencar, G. and Ferreira, G. (2018), “Fatigue assessment of a high-speed railway composite steel-
concrete bridge by the hot-spot stress method”, International Journal of Structural Integrity,
Vol. 9 No. 3, pp. 1-18.
Alencar, G., Calçada, R. and Silva, J.G.S. (2016), “Fatigue assessment of approach viaducts of the new
sado river railway crossing”, Proceedings of the 5th International Conference on Integrity-
Reliability-Failure, Porto, pp. 22-33.
Alencar, G., Ribeiro, D., Calçada, R. and Silva, J.G.S. (2016), “Análise Dinâmica e verificação à Fadiga dos
Viadutos de Acesso da Nova Ponte Ferroviária sobre o Rio Sado”, Proceedings of the XXXVII Iberian
Latin-American Congress on Computational Methods in Engineering, ABMEC, Brasília, November.
Ansys Inc (2017), Release 18.2 – Documentation for ANSYS, Ansys Inc, Canonsburg.
Correia, J.A.F.O., Blasón, S., De Jesus, A.M.P., Fernández-Canteli, A., Moreira, P.M.G.P. and Tavares, P.J.S.
(2016), “Fatigue life prediction based on an equivalent initial flaw size approach and a new
normalized fatigue crack growth model”, Engineering Failure Analysis, Vol. 69, November, pp. 15-28.
De Jesus, A.M.P., Figueiredo, M.A.V., Ribeiro, A.S., De Castro, P.M.S.T. and Fernandes, A.A. (2011),
“Residual lifetime assessment of an ancient riveted steel highway Bridge”, Strain, Vol. 47 No. S1,
pp. e402-e415.
ERRI D214/RP9 (2001), “Railway bridges for speeds W200 km/h”, final report, European Rail Research
Institute (ERRI), Utrecht.
Eurocode: European Committee for Standardization – EN 1990 (2001), “Eurocode 0: basis of structural
design – Annex 2: application for bridges (normative)”, European Committee for
Standardization, Brussels.
Eurocode: European Committee for Standardization – EN 1991-2 (2003), “Eurocode 1: actions on
structures – part 2: traffic loads on bridges”, European Committee for Standardization, Brussels.
Eurocode: European Committee for Standardization – EN 1993-1-9:2005 (2005), “Eurocode 3: design of Fatigue
steel structures”, Parts 1–9: Fatigue, European Committee for Standardization, Brussels. assessment
Ferreira, G.C. (2017), “Análise À Fadiga de um Viaduto Ferroviário para Tráfego de Comboios de Alta based on hot-spot
Velocidade”, master’s thesis, Engineering Faculty, University of Porto.
GRID, GREISH, BEG, FERBRITAS (2006), “Variante de Alcácer: Projecto de Execução do
stresses
Atravessamento do Rio Sado”, In REFER.
Heshmati, M. (2012), “Fatigue life assessment of bridge details using finite element method”, master’s
thesis, Programme Structural Engineering and Building Performance Design, Chalmers
University of Technology, Gothenburg.
IIW (2008), Hobbacher. A. Recommendations for Fatigue Design of Welded Joints and Components,
doc. XIII-2151r4-07/XV-1254r4-07, International Institute of Welding, Paris.
Jakubczak, H., Sobczykiewicz, W. and Glinka, G. (2006), “Fatigue reliability of structural components”,
International J. Materials and Product Technology, Vol. 25 Nos 1/2/3, pp. 64-83.
Leite, R.G.C., Jesus, A.M.P., Correia, J., Raposo, P., Jorge, R.N., Parente, M.P. and Calçada, R. (2018),
“A methodology for a global-local fatigue analysis of ancient riveted metallic bridges”,
International Journal of Structural Integrity, Vol. 9 No. 3, pp. 355-380.
Liu, Z., Correia, J., Carvalho, H., Mourão, A., Jesus, A., Calçada, R. and Berto, F. (2019), “Global-local
fatigue assessment of an ancient riveted metallic bridge based on submodelling of the critical
detail”, Fatigue & Fracture of Engineering Materials & Structures, Vol. 42 No. 2, pp. 546-560.
Malveiro, J., Ribeiro, D., Sousa, C. and Calçada, R. (2018), “Model updating of a dynamic model of a
composite steel-concrete railway viaduct based on experimental tests”, Engineering Structures,
Vol. 164, pp. 40-52, 41.
Mourão, N. (2010), “CP 4000 series Pendolino train with an Alfa pendular service to Oporto”, seen in the
Alcácer Bypass, CC BY 2.0.
Nussbaumer, A., Borges, L. and Davaine, L. (2011), “Fatigue design of steel and composite structures:
eurocode 3: design of steel structures part 1-9 – fatigue eurocode 4: design of composite steel and
concrete structures”, ECCS Eurocode Design Manuals, 311p.
Reis, A., Cremer, J.M., Lothaire, A. and Lopes, N. (2010), “The steel design for the new railway bridge
over the River Sado in Portugal. Ernst & Sohn Verlag für Architektur und technische
Wissenschaften GmbH & Co. KG, Berlin”, Steel Construction, Vol. 3 No. 4, pp. 201-211.
Reis, A.J. and Pedro, J.J.O. (2013), “Pontes Ferroviárias com Tabuleiro Misto Aço-Betão”, II Congresso
Luso-Africano de Construção Metálica Sustentável, Maputo, Moçambique, July.
Ribeiro, D.R.F. (2004), “Comportamento Dinâmico de Pontes sob Acção de Tráfego Ferroviário a Alta
Velocidade”, master’s thesis, Faculty of Engineering, University of Porto, Porto.
Ribeiro, D.R.F. (2012), “Efeitos Dinâmicos Induzidos por Tráfego em Pontes Ferroviárias: Modelação
Numérica, Calibração e Validação Experimental”, doctor’s thesis – Faculty of Engineering,
University of Porto, Porto.

Corresponding author
Cristiane Oliveira Viana can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]