The Influence of Torsional Resistance of The Deck On The Dynamic Response

DEGREE PROJECT IN CIVIL ENGINEERING AND URBAN
MANAGEMENT,
SECOND CYCLE, 30 CREDITS
STOCKHOLM, SWEDEN 2017
The influence of torsional

resistance of the deck on the
dynamic response of a high-speed
railway bridge
Case study: Ulla River Viaduct
CLAUDIA SANROMAN CERVERO
KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ARCHITECTURE AND THE BUILT ENVIRONMENT
The influence of torsional
resistance of the deck on the
dynamic response of a high-
speed railway bridge
Claudia Sanroman Cervero
Master of Science Thesis

Stockholm, Sweden 2017
TRITA-BKN. Master Thesis 512, 2017 KTH School of ABE
ISSN 1103-4297 SE-100 44 Stockholm
ISRN KTH/BKN/EX 512 SE SWEDEN
© Claudia Sanroman Cervero, 2017

Royal Institute of Technology (KTH)
Department of Civil and Architectural Engineering
Division of Structural Engineering and Bridges
Abstract
Understanding how different parameters affect the dynamic response of high-speed

railway bridges is crucial to selecting an efficient structural form. Despite existing
numerous publications within this field, only few address the importance of torsional
deformations.
The main objective of this thesis is to investigate the influence of the torsional resistance
of the deck on the dynamic response of an existing bridge. Ulla River Viaduct is
presented as a case study, allowing to analyse some aspects of its design and what their
alteration entails. To this end, 6 different 3D FE models are compared, 5 of which show
a modification from the original configuration. In addition, several positions of the train
are considered to contrast the effects when the torsional modes are excited. The
performed dynamic calculations are based on the implicit direct integration procedure.
The analysis of the case study demonstrates the benefit of closing the torsional circuit of
the deck. The results also evidence the need of including torsional effects in its dynamic
assessment when low values of torsional rigidity are considered. All this is not easy when
simplified 2D or 3D beam models are used. As a final remark, the original design of the
Ulla River Viaduct is found highly efficient from a dynamical point of view.
Keywords: Dynamics, High-speed railway bridges, Torsional rigidity of the deck, FE

modelling, 3D Model, Ulla River Viaduct
i
Resumen
Entender cómo diversos parámetros afectan el comportamiento dinámico de puentes de

ferrocarril de alta velocidad resulta crucial en la elección de una configuración estructural
eficiente. A pesar de la existencia de numerosas publicaciones en este ámbito, pocas
abordan la importancia de las deformaciones torsionales.
La finalidad de este trabajo es investigar la influencia de la resistencia torsional del

tablero de un puente existente en su comportamiento dinámico. El Viaducto sobre el Río
Ulla se presenta como caso de estudio donde algunos aspectos relativos a su diseño son
estudiados, así como lo que su alteración conlleva. Para ello, se comparan 6 modelos de
elementos finitos en 3D, 5 de los cuales presentan una modificación del diseño original.
Asimismo, para contrastar los efectos provocados al excitar los modos torsionales se han
considerado varias posiciones de tren. Los cálculos dinámicos se han realizado por
integración directa.
El análisis del caso de estudio demuestra que un tablero con circuito torsional cerrado
resulta beneficioso. Los resultados, además, evidencian la necesidad de incluir los modos
de torsión en la evaluación del puente cuando se consideran bajas rigideces torsionales.
Todo esto no resulta sencillo cuando se utilizan modelos tipo viga simplificados, en 2D o
3D. Finalmente, cabe destacar que el diseño original del Viaducto sobre el Río Ulla es
altamente eficiente desde el punto de vista dinámico.
Palabras clave: Cálculo dinámico, Puentes de ferrocarril de alvta velocidad, Rigidez torsional
de tablero, Modelo FE, Modelo 3D, Viaducto Río Ulla
iii
Preface
This master thesis was conducted at the Royal Institute of Technology (KTH),
department of Civil and Architectural engineering, in collaboration with TYPSA. The
study was developed under the supervision of José Javier Veganzones, José Luis Sanchez
and Sonia Alonso to whom I wish to express my most sincere gratitude for their guidance,
advice, support and continuous feedback, but especially for letting me do research in this
field. My profound gratitude to the examiner, Prof. Raid Karoumi who always found
time to help me. Special thanks to John Leander and Andreas Andersson for their
valuable advice with my ABAQUS model, and to Heydar Beygi and Jing Yang for their
help with the moving loads simulation. I would additionally wish to thank Joan Ramon
Casas for his guidance every time I needed it. Finally, huge gratitude to Jose M. Goicolea
and Miguel Otega for sharing information that I found precious and essential for my
study.
Stockholm, June. 2017.
Claudia Sanroman Cervero
v
Notations
Classification coefficient
Stiffness-proportional Rayleigh damping coefficient [sec]
Displacements [m]
Wavelength [m]
Mass-proportional Rayleigh damping coefficient [sec-1]
Damping ratio [%]
Density [kg/m3]
Φ Dynamic factor or impact coefficient
, Coefficients for dynamic enhancement and track irregularities respectively
Circular frequency [rad/s]
a Vertical acceleration [m/s2]
C Number of intermediate coaches
d Bogie axle separation [m]
D Coach length [m]
e Eccentricity [m]
E Young Modulus [MPa]
Frequency [Hz]
External force vector
I Second moment of inertia [m4]
K Stiffness matrix
Span length [m]
Determinant length [m]
m Mass [kg]
M Mass matrix
p Load [kN]
Number of samples
Time [s]
T Total time [s]
u Nodal displacement vector
Speed of the train [km/h]
vii
viii
Abbreviations
2D Two-dimensional
3D Three-dimensional
DOF Degree of freedom
ERRI European Rail Research Institute
FE Finite element
FFT Fast Fourier Transform
HSLM High-Speed Load Model
HSR High-speed railway
LM Load Model
SRSS Square root of the sum of squares procedure
ix
Contents
Abstract ................................................................................................................... i
Resumen................................................................................................................. iii
Preface .................................................................................................................... v
Notations .............................................................................................................. vii
Abbreviations ......................................................................................................... ix
1 Introduction ................................................................................................... 1
1.1 Motivation .................................................................................................. 1
1.2 Aim and scope ............................................................................................. 3
1.3 Methodology ............................................................................................... 3
1.4 Assumptions and limitations ....................................................................... 4
1.5 Outline of the thesis .................................................................................... 4
2 High Speed Railway Bridges............................................................................ 7

2.1 Review of the special characteristics ............................................................ 7
2.1.1 Resonance and cancellation ............................................................. 7
2.2 Parameters that influence the dynamic response ......................................... 8
2.2.1 Effect of torsional rigidity .............................................................. 11
2.3 Longitudinal schemes ................................................................................ 12
2.4 Envelopes and verifications ....................................................................... 12
2.4.1 Train load models .......................................................................... 13
2.4.2 Envelope dynamic factor ............................................................... 14
2.4.3 Verifications .................................................................................. 17
2.4.4 Amplification factor....................................................................... 18
3 Dynamic analysis of bridges under high speed trains ....................................... 19

3.1 Introduction .............................................................................................. 19
3.2 FE models for torsional considerations ...................................................... 20
3.2.1 Two-dimensional models................................................................ 20
xi
3.2.2 Three-dimensional models ............................................................. 21
3.3 Dynamic analysis models for structures and vehicles ................................. 21
3.3.1 Moving mass problem .................................................................... 22
3.3.2 Moving load problem ..................................................................... 23
3.3.3 Moving oscillator (sprung mass) .................................................... 24
3.4 Dynamic analysis approaches .................................................................... 25
3.4.1 Time domain ................................................................................. 25
3.4.2 Frequency domain ......................................................................... 26
4 Case study: Ulla River Viaduct ...................................................................... 29

4.1 Introduction .............................................................................................. 29
4.2 Description of the structure ....................................................................... 30
4.3 FE models ................................................................................................. 32
4.3.1 Assumptions and limitations ......................................................... 32
4.3.2 Geometry and materials ................................................................ 33
4.3.3 Boundary conditions...................................................................... 34
4.3.4 Loading ......................................................................................... 35
4.3.5 Mesh .............................................................................................. 35
4.3.6 Different models ............................................................................ 36
4.3.7 Railway track locations ................................................................. 38
4.4 Model validation ....................................................................................... 39
4.4.1 Mass check .................................................................................... 39
4.4.2 Static loading test .......................................................................... 39
4.4.3 Modal analysis ............................................................................... 40
4.4.4 Boundary conditions...................................................................... 41
4.5 Total simulation and time step selection ................................................... 42
4.6 Moving mass-moving load comparison....................................................... 43
5 Results and discussion ................................................................................... 45

5.1 Introduction .............................................................................................. 45
5.1.1 Dynamic aspects examined ............................................................ 45
5.2 Modal analysis........................................................................................... 46
5.3 Measurement points .................................................................................. 47
5.4 Static displacements .................................................................................. 48
5.4.1 For high-speed trains (HSLM-A1) ................................................. 48
5.4.2 For nominal train LM-71 ............................................................... 50
5.5 Accelerations and dynamic displacements ................................................. 50
xii
5.6 Amplification factor .................................................................................. 54
5.7 Twist of the deck ....................................................................................... 55
5.8 Envelope dynamic factor ....................................................................... 55
5.9 Different locations of passing train ............................................................ 56
6 Conclusions ................................................................................................... 59
6.1 Conclusions ............................................................................................... 59
6.1.1 Regarding the different models ...................................................... 59
6.1.2 Regarding different positions of the trains ..................................... 60
6.1.3 Regarding the efficiency of bottom slab configuration .................... 60
6.2 Further research ........................................................................................ 60
Bibliography ........................................................................................................... 61
A FE Models ..................................................................................................... 65
A.1 Materials ................................................................................................... 65
A.1.1 Concrete .......................................................................................... 65
A.1.2 Steel ................................................................................................. 65
A.2 Cross sections ............................................................................................ 66
A.3 FE Models ................................................................................................. 68
B Mode shapes .................................................................................................. 71

B.1Oscillation modes MODEL 1 ......................................................................... 71
B2 Comparison eigenfrequencies ......................................................................... 75
C Static loading test .......................................................................................... 77

C.1 Vehicle definition ...................................................................................... 77
C.2 Loading positions ...................................................................................... 77
xiii
1.1. MOTIVATION
Chapter 1
Introduction
1.1 Motivation
During the last few decades, high-speed railway (HSR) traffic has become an interesting
alternative for connecting medium and long distances. Not only is it a sustainable and
comfortable way of travelling, but also a powerful tool of economic and social
development for the regions involved.
After Japan built their first passenger dedicated HSR line, it rapidly started its
expansion in Europe and East Asia. Nowadays, China boasts the longest HSR network,
reaching almost 13000 km. Spain covers about 3100 km coming to the second place in
the world and resulting in the first one in Europe. The entire world high-speed network
extends over 30000 km but it is forecasted to grow around 30% within the future ten
years (Figure 1.1). Detailed historic development has been addressed in the literature
[1].
Figure 1.1: Development of the world high-speed network [2]
Bridges help to overcome all the obstacles arising from the difficult orographic conditions
and to avoid altering existing infrastructures, becoming an indispensable asset in the
construction and operation of new HSR lines (Figure 1.2). For instance, in Spain it was
necessary to build around 1200 bridges to this end [3].
1
CHAPTER 1. INTRODUCTION
Figure 1.2: Bridge over the Yangcheng Lake in central China. It belongs to the Bejing-
Guangzhou HSR line [2]
These types of HSR bridges present some specific requirements clearly different from
those dedicated to conventional railway traffic. Apart from ensuring safety conditions,
a good comfort for the users has to be provided. The most relevant aspects in the design
of these structures are the magnitude of the live loads and the dynamics effects of passing
trains that may cause amplifications of forces and deformations. Hence, such effects are
to be thoroughly studied in order to avoid resonant phenomena that can result in non-
admissible forces and deformations, and consequently, altering the traffic conditions and
causing excessive vibrations inside the vehicles. However, their estimation usually
requires a great demand of computational resources and a deep knowledge of the
structure itself [4].
The selection of an efficient structural form is the basic goal of bridge design. Thus,
understanding how different parameters influence on their dynamic response is a crucial
concern. The effect of some parameters such as span length, number of spans, damping,
train speed, etc. on the said dynamic response has been the subject of numerous
researchers over the last decades [5, 3, 6, 7].
On the other hand, in many practical applications, design engineers usually consider
simplified 2D models for dynamic analysis of bridges, leading to a considerably large gain
in terms of time and computational sources. This is the main reason why a significant
amount of publications within this field focuses on linear and simple structures.
However, one may note that the vast majority of current bridges dedicated to HSR traffic
are composed of double railway-track, which torsional stiffness should be high enough to
cope with the torsional flows caused by the eccentric loads of only one train passing.
Moreover, steel-concrete composite bridges are becoming more and more popular to this
end [8]. Some of their main characteristics, in terms of dynamics, are their low values of
torsional natural frequencies and the similar values of deflections of the deck due to the
torsional and bending components, being necessary to consider both effects as coupled
[9]. With all this in mind, it results indeed very striking that little of the research effort
addresses the topic of the importance of torsional deformations.
2
1.2. AIM AND SCOPE
1.2 Aim and scope

The present thesis aims to study the influence of torsional resistance of the deck on the
dynamic behaviour of HSR bridges, more precisely, on the dynamic behaviour of a
complex and hyperstatic worldwide-recognised structure selected as a case study: Ulla
River Viaduct. This bridge uses an alternative solution for closing the torsional circuit
of the deck with respect to the traditionally one employed in similar bridges typologies.
More specifically, the objectives are to:
Demonstrate if the aforementioned design is efficient and investigate how its

behaviour is affected when it is altered, from a dynamical point of view.
Determine the extent of the contribution to the dynamic response of the mass and
stiffness of the piers.
Compare the effects produced when several locations of the train are considered.
As it can be gathered from previous lines, this thesis limits its scope to study a particular
bridge. Furthermore, due to the limited amount of time, it is also beyond the scope of
this work the consideration of more than one train model. All the calculations were
performed for high-speed load model A1, commonly denoted as HSLM–A1.
1.3 Methodology
First and foremost, an exhaustive literature review was carried out. Throughout this
report, some of the most relevant state-of-the-art reviews are presented, with particular
emphasis in those documents which subject of research focuses on structural dynamics
of high-speed railway bridges, all the relevant design codes and tutorials.
After having identified the most appealing aspects, a particular bridge was selected as a
case study: Ulla River Viaduct. Due to its complex typology, a full analysis of its dynamic
behaviour was considered of great interest. Furthermore, the aforementioned alternative
solution for closing its cross-section made it ideally suited for studying the influence of
its torsional resistance of the deck.
To this end, a complete 3D finite element (FE) model was performed reproducing in
detail the design drawings. Project information developed by IDEAM, courtesy of
TYPSA as technical assistance engineering at construction.
The finite element (FE) method is arguably the most suitable way of analysing
structures nowadays, both two- and three- dimensional. The commercial package
available for the FE modelling was ABAQUS [10]. However, applying moving loads is
not straightforward in this software, thus, the engineering tool MATLAB [11] was
needed as a support.
For the dynamic analysis, the implicit integration scheme has been used. However, few
checks in the frequency domain were performed as an endorsement for a better
understanding of the obtained results.
3
CHAPTER 1. INTRODUCTION
The quality of the results is highly dependent of the underlying assumptions, especially
when it comes to mass, stiffness and boundary conditions. Therefore, the refined quality
assurance included static as well as dynamic checks/verifications to ensure that the
model functioned adequately.
Finally, the core part of this work consisted of studying the influence of different aspects
on the bridge dynamic response. Thereby, different FE models have been developed in
which the following parts of the original bridge were altered:
Non-connected prefabricated plates of the bottom slab.

Torsional rigidity of the deck by considering open cross-sections.
Mass and stiffness of the piers.
Location of the railway track.
Subsequently, six different FE models were examined in which a high-speed train load
model travelled over three different positions. The train speeds vary from 200 km/h to
420 km/h with a step of 20 km/h, yielding more than 120 dynamic calculations.
1.4 Assumptions and limitations

As it will be explained through the following chapters, the 3D FE model reproduced in
detail the design drawings, seeking to simulate the real flexural and torsional rigidity of
the deck and piers. Nevertheless, the following simplifications have been made:
Only the central spans were modelled.

No interaction vehicle-structure was contemplated.
Each axle of the train was assumed to cause a concentrated load that moved along
one straight axis (singly railway-track lane).
No load distribution due to the existence of ballast or two rails was considered.
Only HSLM–A1 was contemplated.
On the other hand, the dynamic calculations were based on direct integration within the
time domain. This method requires of a very little time step if a realistic solution is
sought, especially for structures with low damping such as the case of bridges, leading to
a considerably large amount of resources in terms of computational memory and time.
The several aforementioned simplifications helped to reduce these problems, but they
could also diminish the accuracy of the results.
1.5 Outline of the thesis

Some knowledge of structural dynamics of HSR bridges is necessary beforehand to
understand the overall study; therefore, the basic theory and pertinent design codes
requirements are provided throughout chapters 2 and 3. Likewise, these chapters also
present some of the most relevant state-of-the-art review.
4
1.5. OUTLINE OF THE THESIS
Chapter 4 starts by describing in detail the Ulla River Viaduct, bridge selected as case
study. This chapter then proceeds by explaining the FE models analysed and its
validation by means of quality assurance checks/verifications.
All the results are shown and discussed in chapter 5. Finally, the 6th chapter includes the
subsequent conclusions and suggest some aspects for future research in this field.
5
2.1. REVIEW OF THE SPECIAL CHARACTERISTICS
Chapter 2
High Speed Railway Bridges
2.1 Review of the special characteristics

Essentially five aspects may constitute the basic differences between bridges dedicated
to conventional traffic and HSR lines, in terms of morphology and design [4]:
The magnitude of vertical loads: railway traffic has approximately double value than
conventional highway traffic.
Location of loads: the position of the live loads may only occur over the track.
Fatigue: repetitive loads, usually reaching its maximum values, may lead to fatigue
problems.
Braking and start-up forces.
Dynamic effects that can be manifested in two ways. The first one is that the static
effects are increased due to the effect of impact. This is particularly noticeable in
railways-tracks in poor maintenance. On the other hand, this kind of bridges can be
affected by the resonant phenomena (section 2.1.1).
2.1.1 Resonance and cancellation
Resonance can occur when the frequency of the external excitation coincides with the
main natural frequencies of a structure. For railway bridges, the resonant phenomenon
has been observed when the train is travelling at speeds exceeding 200 km/h, such as the
case of HSR lines.
It is possible to quantify these effects using the wavelength of excitation , shown in eq.
(2.1). It relates the train velocity to the fundamental frequency of the bridge 0 .
= (2.1)
Resonance will appear when such wavelength of excitation , or any of its multiples,
coincides with the separation of the axles of the given train , as shown in eq. (2.2).
7
CHAPTER 2. HIGH SPEED RAILWAY BRIDGES
= , = 1,2 (2.2)
Experimental results of real bridges as well as mathematical models show that the most
noticeable dynamic effects can be reached while the train is travelling and not necessarily
when it has left the bridge [12].
The completely opposite phenomenon, known as cancellation, may also occur under
particular circumstances in bridges with longer span than the train length and multi-
span bridges with continuous deck. It results in a suppression of the dynamics effects.
The reason is that several axles of the train will be passing over the bridge at different
phases. An illustrative example can be found in Figure 2.1, that explains how certain
train configurations can affect the dynamic response of a bridge, in terms of amplification
or cancellation effects [13]. The plots show the maximum accelerations for 10 HSLMs
and their normalised amplitude of free vibrations.
Figure 2.1: a) Maximum vertical accelerations for the train load-model, b) normalized
amplitude of free vibrations [13]
As a result, it is possible to find the most optimal bridge span length-to-vehicle ratio.
However, in real practice this is quite complicated due to the needs for interoperability,
where more than one train shall be assessed. For curious readers, cancellation effects in
the conception stage of bridge design have been addressed in [14], where the author
proposes taking them into account in order to enhance the dynamic response. Moreover,
in [15] an optimization process of cross-section for simply supported and single-track
bridges was performed based on vertical accelerations.
2.2 Parameters that influence the dynamic response

The recent development of high-speed railway lines had led to the necessity of upgrade
the existing infrastructures as well as the design of new ones. For this reason,
8
2.2. PARAMETERS THAT INFLUENCE THE DYNAMIC RESPONSE
ascertaining the most influential parameters on the dynamic response has been the topic
of numerous researchers during the last decades.
The European Rail Research Institute (ERRI) started the research carried out by Frýba
[5], where the effects of some of the most relevant parameters affecting the dynamic
response of a bridge were studied. The conclusions pointed out the train speed as the
most significant parameter for the design, affecting deflections, bending moment and
accelerations in a similar manner. It must be highlighted that as the span length l (Figure
2.2.a), damping theta (Figure 2.2.b) or bridge weight G increase, the vertical
accelerations diminish. Finally, Frýba also stated that the dynamic response of concrete
bridges is lower than for steel bridges due to their higher values of damping and mass.
a) Influence of span length l
b) Influence of damping theta. Case a out of resonance and case

b at resonance
Figure 2.2: Qualitative comparison of the effect of damping theta and span length l in vertical
accelerations under trains at 350 km/h [5]
It is not always possible to know beforehand some important parameters such stiffness,
damping or the fundamental frequency of a bridge, supposing their possible over- or
underestimation. The safety of the structures underlies on the accuracy of the models,
9
and therefore, Spanish design codes require a static and dynamic testing. The results of
such test, conducted in more than 119 bridges with different superstructures (from
precast to posttensioned girders), allowed to establish a relationship between span length
and fundamental frequency. Figure 2.3 clearly shows that the natural frequency tends
to decrease as the span length increases. This tendency is apparently unrelated to the
superstructure type [3].
Figure 2.3 Measured first natural frequency for 119 bridges [3]
The influence of number of spans is another popular subject of research. In that sense,
vertical accelerations were found lower for multi-span bridges in comparison to single-
span ones, as long as its natural frequencies were within the Eurocode limits [16]. On
the contrary, short continuous bridges with lower natural frequencies were observed to
yield higher accelerations [6].
Yau [7] demonstrated that continuous bridges lead to lower dynamic response than
simply supported ones, and consequently, an enhanced dynamic behaviour. However, his
observations also showed that more resonance peaks were produced (Figure 2.4).
Continuous multi-span bridges are almost always preferable therewith. With this
configuration, greater stiffness is achieved and more “noise” in the dynamic response,
and thus, less resonance problems [4].
Figure 2.4: Dimensionless speed versus impact factor I= (Rdy-Rsta)/Rsta. Where Rdy and
Rsta are the dynamic and the static responses respectively [7]
10
2.2. PARAMETERS THAT INFLUENCE THE DYNAMIC RESPONSE
2.2.1 Effect of torsional rigidity
Goicolea [17, 18] considered of especial importance the need of taking into account not
only the bending but also the torsional component of bridges’ deck. This last component
is particularly significant in those bridges with low torsional rigidity.
As a practical example, the case study of Las Piedras Viaduct, the first steel-concrete
composite continuous deck bridge in the Spanish HSR line is presented in [18]. The
authors studied its dynamic response and compared the results to those obtained when
altering its torsional rigidity, by means of considering open cross-section. Finally, they
concluded that when considering open cross-section the subsequent accelerations were
inadmissible (Figure 2.5).
a) Open cross-section b) Partially closed section, accelerations

are fulfilled
Figure 2.5: Maximum vertical acceleration when considering two different configurations of the
Las Piedras Viaduct cross-section [18]
The importance of considering coupled bending-torsional effects in dynamic analysis of

bridges using FE models was also investigated. The Río Cabra Viaduct, a continuous
deck bridge composed of hollow slab deck cross-section was proposed as a case study [9].
The author studied the applicability of some simplified methods for considering the
torsion by means of comparing the subsequent results to those obtained with a 3D
complete FE model of coupled bending-torsion. It was also compared such results to
those obtained when considering a box slab cross-section instead. This comparison
revealed that in bridges with high torsional stiffness the advantage gained by considering
coupled bending-torsion was insignificant, which is very interesting from a practical
point of view.
Lastly, the work developed in [19] is also perceived as essential for these thesis purposes.
The authors discussed the difference between several bridge typologies belonging to the
Spanish HSR line. The explanation of why it is considered indispensable for the present
study is that one of the bridges analysed is the Ulla River Viaduct, subject of our case
study. Two different calculations were performed in parallel, whether considering or not
the torsional modes. A significant increment in the dynamic response of such bridge was
11
observed when both mode types were considered. That is, reaching values five times as
those obtained when computing only vertical modes in terms of displacements, and twice
as much in terms of accelerations.
2.3 Longitudinal schemes

This section aims to summarize the most relevant aspects in terms of longitudinal
schemes and deck morphologies/ characteristics. To this end, the structures built up to
this moment in China and Spain were accounted. Both countries have been already
referred as those ones counting with the largest extension of high-speed network.
As commented previously in this work, the vast majority of bridges are composed of
double railway-ballasted track. The deck’s width generally adopted is 14 m. The
selection of their typology depends essentially on the span length, as it can be understood
from Table 2.1 [20].
Table 2.1: Common cross-sections used in Spain [20]
Span length (m)

Typology
5 10 15 20 25 30 35 40 45 50 55 60 65 ≥ 70
Slab section
solid slab
hollow slab
Box section
constant depth
variable depth
Prefabricated
simply supported
hyperstatics
Composite
It is noteworthy the work done in [1], where a cost efficiency criterion is discussed for
choosing the most appropriate structural forms in China. In general, for medium spans
(100-200 m) tied steel arch, rigid frames or hybrid steel arch bridge with concrete girder
are recommended. On the other hand, when it comes to long spans (>200 m) steel truss
arch and cable-stayed bridge with truss girder are the preferred ones.
2.4 Envelopes and verifications

Bridge response under dynamic loads can vary mainly because of the three factors below:
Different vertical loads in each point of the structure (vehicle speed and vibration of
structure).
Successive loads entering to the bridge at uniform and evenly spaced intervals.
Track and vehicles irregularities.
12
2.4. ENVELOPES AND VERIFICATIONS
Standing design codes [21, 16, 22, 23] usually consider the dynamic response of a railway
bridge through the magnification of the dynamic response for a single moving load with
respect to the static one, by the so-called dynamic factor or impact coefficient.
Nevertheless, due to the increment of speed and vehicle lengths, the resonant phenomena
became a crucial issue to be taken into account, and thus, the only use of this factor were
unacceptable [24]. For this reason, dynamic loads for high-speed traffic need to be
assessed by a specific and complete dynamic analysis.
ERRI carried out a research in order to update the design codes. They defined a set of
actions due to traffic loads that needs to be considered in the design of HSR bridges.
Those include from horizontal and vertical loads to static and dynamic ones. Their
combination together with all the other loads has to fulfil the ultimate state
requirements.
2.4.1 Train load models
An aspect worth emphasising is the convenience of interoperability of different families

of trains. Thus, the infrastructure needs to behave as desired when either of the existing
trains is passing, in terms of dynamics effects. Since the worst effects are not necessarily
reached for maximum train velocities, the referred effects shall be evaluated for all
possible trains speeds as well. The current vehicles have a broad variation between axles,
coach lengths, etc. and can be split in three categories (Figure 2.6):
Articulated trains: THALYS, AVE and EUROSTAR

Conventional trains: ICE2, ETR-Y and VIRGIN
Regular trains: TALGO AV
Figure 2.6: Different typologies of high speed trains according to [19]
The consideration of all the trains might be quite time consuming and laborious. For
this reason, together with the possibility of appearance of new train configurations,
ERRI D214 committee established a High-Speed Load Model (HSLM) [25]. Being
HSLM-B for bridges with span lower than 7 m and HSLM-A for all the rest. HSLM-A
13
consist of 10 different train load models that represent fictitious universal trains which
signature (Figure 2.7) is the envelope of all the real high-speed trains.
Figure 2.7: Envelope of dynamic signatures for all European high-speed trains and HSLM-A
model [17]
The characteristics of these universal fictitious trains are summarised in Table 2.2. The
complete dynamic analysis should be performed for all 10 HSLM and considering series
of speed up to 1,2xMaximum Permitted Vehicle Speed, every 10 km/h.
Table 2.2: Characteristics of HSLM-A
Number of
Universal Coach length D Bogie axle Point force p
intermediate
train [m] spacing d [m] [kN]
coaches C
A1 18 18 2 170
A2 17 19 3.5 200
A3 16 20 2 180
A4 15 21 3 190
A5 14 22 2 170
A6 13 23 2 180
A7 13 24 2 190
A8 12 25 2.5 190
A9 11 26 2 210
A10 11 27 2 210
2.4.2 Envelope dynamic factor
2.4.2.1 Vertical train load model
Figure 2.8 shows the train load model LM -71 [23]. In order to calculate the dynamic
factor, such model has to be considered by means of a static calculation. It aims to
14
represent the maximum effects that the railway traffic may cause on an infrastructure.
For one railway-track loaded it is defined as follows:
a) Four axles of 250 kN each, separated a distance of 1.6 m. Located in the worst
position along the axis of the railway lane.
b) A uniformly distributed load of 80 kN/m. This load shall be applied in the worst
situation (location and extension) for reaching the maximum absolute values of
displacements.
Figure 2.8: Load Model LM-71 [23]
2.4.2.2 Definition
From the static point of view, the Ultimate Limit State is verified through the dynamic
factor Φ [16], also known as impact coefficient in the Spanish code [21]. Its definition
is based on statistical analysis of existing bridges and it covers all the dynamic effects
arising from railway traffic, including track imperfections. However, such effects cannot
be split because they are derived from measurements. It may be calculated according eq.
(2.3) for railway tracks well maintained.
1,44
Φ = + 0,82 1,00 ≤ Φ ≤ 1,67 (2.3)
− 0,2
Where relates the determinant length defined in Table 6.2 in [16]. Accordingly, its
application is limited to conventional structures, in other words, to structures included
in this table.
Note that this dynamic factor Φ is only applicable when the design velocity is lower
than 220 km/h and for fundamental frequencies within the limits specified in the
Eurocode. This is because it omits the possibility of resonant phenomenon occurring and
assumes that the maximum accelerations do not overpass the limiting values set in the
referred code under any circumstances.
In the event that the aforementioned conditions are not satisfied, it results necessary to
carry out a specific dynamic analysis. Nevertheless, in these cases, the same concept may
be applicable by means of the envelope dynamic factor defined in eq. (2.4).
Φ = max (2.4)
15
Where,
is the static maximum deflection caused by the load model LM-71, for two
railway-track loaded.
is the maximum dynamic deflection caused by real trains. To consider two trains
travelling simultaneously, the values calculated for one railway-track loaded shall be
combined using the square root of the sum of squares procedure (SRSS). In other
words, the results are multiplied by √2.
The determination of not only includes the dynamic effects produced by the train
loads themselves but also by the track irregularities. Its decomposition can be expressed
as shown in eq. (2.5).
= (1 + + φ'') (2.5)
Where,
is the quasi static deflection caused by real high-speed trains. It may be

calculated by considering a train speed slow enough.
is a coefficient that represents the enhancement of the dynamic effect with respect
to the static one. Its analytical expression is shown as follows:
= , being =
1− + 2
φ'' is a coefficient that takes into account track irregularities. Values of " are ≈ 0
for long span bridges and low natural frequencies. It is defined in expression B.12 in
[21]:
= 0.56 + 0.5 −1
80
Where again, represents the determinant length and 0 the fundamental frequency.
2.4.2.3 Treatment in different design codes
The requirements of the Eurocode [16] with respect to this factor Φ are in accordance
to the Spanish code [21]. In this context, the main differences between them are that,
for double railway-track bridges, the first one considers only one track loaded whereas
the second one considers both tracks loaded. Another distinction rests in the values of
the classification coefficient . In the Spanish code it represents an increment of 21% in
the response caused by the train LM-71. This percentage was set aiming to keep the
validity of the infrastructures if the vehicles increase weight in 30 tn. Finally, the
Eurocode requires the application of extra loading models for continuous bridges or
freight transportation in opposition to the only model LM-71 required in the Spanish
code. Further definition and utilization of this coefficient according others European
codes is addressed in the literature [12, 24].
16
2.4.3 Verifications
Traffic safety is ensured when the limitations related to the parameters listed below are
satisfied [16]. The most relevant ones for this thesis are addressed in the following
subsections.
Vertical acceleration of the deck.

Torsion of the deck.
Vertical deformation of the deck.
Horizontal deformation of the deck.
2.4.3.1 Vertical accelerations
The vertical accelerations are limited to 3.5 m/s2 for ballasted tracks and 5 m/s2 for un-
ballasted ones, in order to ensure ballast stability. For its calculation, it is necessary to
consider frequencies up to the greatest value of either 30 Hz or 1.5 times the fundamental
frequency but including at least the three first oscillation modes. These frequencies
should be filtered according to the acceleration criterion, and can be done through any
of the following methods [26]:
Suppressing all modes with higher frequencies, but keeping enough mass.
Filtering through the time step integration election.
Filtering the whole complete calculation.
2.4.3.2 Vertical displacements
The serviceability limit state criteria establish an upper bound of the maximum vertical
displacements shown in Figure 2.9. However, this plotting is used for simply supported
bridges with three spans or more, and in case of continuous beams they should be
multiplied by a coefficient of 0.9.
As a final remark, according to the relevant Spanish codes, if the span length of the
examined bridge is greater than 10 m, a dynamic and load testing should be performed
to verify the accuracy of the calculation models.
Figure 2.9: Maximum permissible vertical deflections for bridges for railway traffic
corresponding to a permissible vertical acceleration of 1 m/s2, figure A.2.3 of the code [16]
17
2.4.4 Amplification factor
The amplification factor, also known as the deformation response factor, represents the
amplification in the displacements caused by dynamic effects. It is defined as the ratio
of the dynamic displacements to the static one . Figure 2.10 shows this factor
for an undamped system subjected to a harmonic force plotted against the frequency
ratio / . Where corresponds to the forcing frequency of excitation in rad/s and
the natural frequency of the system. Note that for resonant frequency this factor becomes
maximum under resonant circumstances [27].
Figure 2.10: Amplification factor (Deformation response factor) for an undamped system
subjected to a harmonic force [27]
18
3.1. INTRODUCTION
Chapter 3
Dynamic analysis of bridges under

high speed trains
3.1 Introduction
There are available several dynamic analysis methods for assessing railway bridges:
1. Static calculation through the dynamic factor or impact coefficient.

2. Time integration of the dynamic equations for the structure. In general, by
considering one of the following models: dynamic calculation by moving loads or by
vehicle-bridge interaction.
3. Simplified models using the dynamic train signature.
Such calculations are usually easy to apply to simple and isostatic structures. In such
structures, the first eigenmodes predominate, permitting to characterise their dynamic
response. On the contrary, when it comes to hyperstatic structures, more complex
analysis methods are needed. That is because many vibration modes contribute to its
dynamic response [9].
The first method is based on calculating a dynamic factor Φ (section 2.4.2) that shall
be applied to the envelope of the internal reactions obtained from a static analysis. The
bridge then can be designed with these new values of internal reactions. Due to its limited
applicability to train speeds ≤ 220 km/h, a specific dynamic analysis is needed for
HSR.
The dynamic train signature can be understood as a function that describes the dynamic
effect caused by a train in a bridge. Such function is characterized as a combination of
damped harmonics. This fact limits its applicability to simply supported bridges that
can be defined, in dynamic terms, by their fundamental oscillation mode [17].
For all these reasons, this chapter focuses in the calculation method: time integration of
the dynamic equations for the structure. To facilitate the explanation, the following
paragraphs treat the vehicle-bridge interaction problem as part of the moving load
problem. In other words, it expounds the three different moving load problems that are
generally discussed in the literature: moving mass problem, moving load problem itself
19
CHAPTER 3. DYNAMIC ANALYSIS OF BRIDGES UNDER HIGH SPEED TRAINS
and moving oscillator (vehicle-bridge interaction). The general procedure to solve the
moving load problem is by means of FE method.
3.2 FE models for torsional considerations

The FE models can be 2D or 3D. Eccentric loads, as it is the case of only one train
travelling over a double railway-track bridge, produce deformations not only in terms of
flexion but also of torsion. For this reason, an adequate calculation FE model that allows
capturing the deformations for all the relevant modes involved is crucial. In other words,
the coupling between bending and torsion must be properly captured. This section
contains the type of models and elements more recommendable, taking into account
these torsional considerations.
3.2.1 Two-dimensional models
It omits the transversal direction; a plane (typically X-Z) embeds the structure that only
varies along its longitude X. It is the simplest way of modelling and it can contain a
combination of shell and wire elements.
There exists a broad range of elements to be used in such models: two-dimensional

continuum solid elements, truss elements or beam elements, depending on the structure
geometry and characteristics. The very simplest way of modelling bridges in 2D is using
a beam models.
Beam elements are also separated in sub-families depending on whether the cinematic
hypotheses adopted is Euler-Bernoulli or Timoshenko. The selection of one of these sub-
families rests in the slenderness span length l to thickness t ratio. The recommendations
according to [28] are presented as follows:
/ ≥ 10 Bernoulli beam elements

2 ≤ / < 10 Timoshenko beam elements
Railway bridges usually present slenderness higher than 10/1 and the shear deformation
for span lengths superior than 15 m is not significant. This supposes that the most
recommended elements are Bernoulli beam elements, being the base for many dynamic
simplified models. It is possible to assess the torsional effects in a simplified manner by
using a bridge beam model which associated DOFs shall include the torsional twist of
the transverse section (Figure 3.1). Due to the definition of its nature, such models are
still treated as 2D. The torsional response of the structure shall be included specifically
in the dynamic calculations, generally introducing the effects produced by the
eccentricity e of a load p [12].
20
3.3. DYNAMIC ANALYSIS MODELS FOR STRUCTURES AND VEHICLES
Figure 3.1: Geometric variables influencing the torsional response of a beam model [12]
3.2.2 Three-dimensional models
The model is embedded in the coordinate system X-, Y-, Z and can be a combination of
3D solid, shell and wire elements. These models are necessary for assessing the torsional
effects in non-conventional bridges, in other words, bridges that cannot be reduced as a
2D beam model. The torsional effects are included directly by defining the loads in the
space.
For dynamic calculations, the torsional modes of vibration shall be captured properly. If
the transversal and torsional stiffness are overestimated, the model will yield higher
torsional modes and therefore its contribution to the dynamic response could be
undervalued. Figure 3.2 shows the deformed mesh of a detailed 3D FE model of the Río
Milanillos viaduct after having applied a torsional moment to the deck. This bridge
belongs to the Spanish Segovia-Garcillán HSR line and it is composed of a two box beams
separated a distance of 7 m and joined together through a flexible deck. The resulting
deformation revealed that the transversal section cannot be defined by the hypothesis of
rigid body motion, which is the same as saying that the transversal stiffness is not infinity
[9].
Figure 3.2: Detailed 3D FE model of the Río Milanillos viaduct deformed mesh caused by a
torsion moment [9]
3.3 Dynamic analysis models for structures and

vehicles
A train crossing a bridge is treated as a moving load problem, where the load location
varies along time. The so-called vehicle-bridge interaction problem is so far one of the
most studied type of moving load problem. It is handled as a moving elastic subsystem
over a primary elastic system. While this kind of problems is well defined in terms of
21
numerical and mathematical models, their application in commercial calculation

software is usually limited [29].
Various typical moving load problems are presented as an easy following tutorial in [30].
The aforementioned work also describes a brief but complete state-of-the-art review,
starting from Fryba’s monograph [31] and including a great deal of literature that covers
several types of bridges, making its reading very enjoyable and highly recommended.
Even though it is not a popular subject of investigation, some researchers have also
studied the effects of separation and reattachment of the moving subsystem from the
supporting structure. That may occur when the travelling speed of the first one is high
enough, causing an impact during the reattachment. An overview of this problem is
exposed in the literature [32, 29] and it is considered of interest for the case of high-
speed trains.
The three approaches of moving load problem (moving mass problem, moving load
problem itself and moving oscillator) are explained in detail in the following paragraphs.
3.3.1 Moving mass problem
In this case the loads are simulated by point masses moving along the bridge (Figure
3.3). The moving mass problem is the most intuitive and straightforward way of
modelling moving loads using ABAQUS [10]. The following steps must be taken [29]:
1. Model the primary structure (i.e. deck: shell elements).

2. Model the secondary moving structure (i.e. point mass) and apply a non-structural
inertia and its corresponding weight as a load.
3. Select an appropriate contact between both of them. In ABAQUS it is possible to
specify Node-to-surface contact with frictionless tangential behaviour and hard
contact in the normal direction. Such contact allows capturing the separation and
reattachment of the moving structure.
4. Apply the speed of the moving structure as a constraint, by either prescribing velocity
or time history displacements.
Figure 3.3: Moving mass problem [33]
On the other hand, the drawbacks of this modelling approach, among others, are:
The coupling stiffness between the primary and secondary systems are assumed
infinity.
22
3.3. DYNAMIC ANALYSIS MODELS FOR STRUCTURES AND VEHICLES
In order to apply more than one moving mass, each one has to be modelled separately
with its corresponding contact properties, resulting in an awkward and time-
consuming way of modelling moving loads.
3.3.2 Moving load problem
This approach assumes that the axles of the train cause a constant concentrated load
that moves along with it. Thus, the vibrations produced in the vehicle are neglected as
well as its inertia effect. Nevertheless, in the literature it is considered accurate enough
for practical applications [9].
In this respect, the vehicle is treated as a set of concentrated loads separated a

determined distance depending on the geometry of the given train (Figure 3.4) which
travels at a constant speed v as can be schematized in Figure 3.5.
Figure 3.4: Moving loads pattern of a real train. Each axle force is separated a distance
[33]
There are basically two available ways of applying moving loads in ABAQUS and its
implementation is based on defining load histories that are applied to each node. The
first one is using the already implemented subroutines (i.e. DLOAD) using a Fortran
compiler.
Figure 3.5: Moving load problem. Concentrated loads travelling at a speed v [33]
The second option is applying the time history loads as tabular amplitudes. First of all,
it is necessary to define such tabular amplitudes following the steps below:
1. Numerate the nodes that belong to the load trajectory.

2. Define a reference time that represents the initial time where the first axle of the
train enters the bridge.
23
3. Determine the instant that corresponds to the arrival time of the first axle of the
train to each node j. The axle force causes triangular amplitude to this node as
shown in Figure 3.6.
4. Determine the arrival time steps of the posterior axles of the train to each node j
and its corresponding triangular amplitudes.
Figure 3.6: Load history definition of an axle force p travelling at a speed v and causing a
triangular amplitude to each node j [9]
3.3.3 Moving oscillator (sprung mass)
The moving oscillator is the most realistic approach; the coupling stiffness is finite and
considers inertial effects of the moving structure. The train is represented by point
masses, bodies and springs as shown in Figure 3.7.
Figure 3.7: Moving oscillator. The axle loads combine masses and springs [33]
These kinds of models can be complete or simplified. In both cases they are represented
by sprung and unsprung masses as well as by the primary suspension of each axle. The
completed ones also consider the length, mass and inertia of the bogies, vehicle body
geometry and secondary suspension of bogies.
The interaction between vehicle and bridge is accounted, therefore the complexity of the
model increases considerably and so the computational time does. That is why they
result highly interesting for research aims but not useful for many practical applications
[9].
24
3.4. DYNAMIC ANALYSIS APPROACHES
Furthermore, according to Gabaldón [24], for long span bridges or continuous deck
bridges, the accuracy gained by using interaction models are usually very small. For
these reason, their application are only when the passenger comfort inside the vehicles is
evaluated, but is it not considered necessary for design purposes.
3.4 Dynamic analysis approaches

FE method can generally be applied to any arbitrary structure, even if non-linear effects
are considered. It is based in the discretization of the structure into elements in space
and nodal coordinates in time, yielding to a N-degrees of freedom (DOFs) system of
equations shown in eq. (3.1).
+ + = (t) (3.1)
Where,
is the mass matrix

is the damping matrix
represents the stiffness matrix
( ) is the external loads vector
is the nodal displacements vector, unknown
These equations are usually coupled and therefore need to be solved simultaneously. The
solution is given by multiplying the nodal coordinates by shape functions.
The dynamic analysis of the structure may be performed either through the integration
along time of the -DOFs system or via modal superposition (MS) method, separating
space from time in the coupled equations [27].
3.4.1 Time domain
This procedure solves the whole set of N differential equations for each time step by
means of direct integration. Since these equations are coupled, they need to be solved
simultaneously. The stability of the solution of this problem depends only on the
transient response because no steady state solution exists.
3.4.1.1 Explicit schemes
Use the equation of motion at a time = plus assumptions to find the solution at a
time = + ∆ . In other words, at the end of each time step the matrices are updated
and the system of equations is solved. Those methods are conditionally stable with
respect to the time step election, if the increments are small enough the solution will be
accurate, otherwise it will diverge because the equilibrium is not enforced.
25
3.4.1.2 Implicit schemes
In this case, uses the equation of motion at a time = − ∆ plus assumptions to find
the solution at = . After each increment, the equilibrium is enforced by iterating
with the Newton-Raphson algorithm. Consequently, it allows using larger time steps and
the accuracy is generally higher than the explicit case.
Nevertheless, the mesh selected has to be coarse enough to capture the deformation of
the structure and the time step small enough to capture the higher frequency sought.
3.4.1.3 Time step selection
Recommendations for selecting the time step can be found in the literature [12]. Some
of them are summarized below in this subsection, and are differentiated according to the
aspect considered for the dynamic analysis. In this context, the time step ∆ can be taken
according to:
Higher natural frequency max of the bridge to be accounted:
1
∆ ≤ (3.2)
8
Minimum number of time intervals that an axle needs to travel across the shortest
span min of the bridge at a speed :
∆ ≤ (3.3)
200
Number of natural frequencies n considered, length of the shorter span of the bridge
and train speed :
∆ ≤ (3.4)
4
3.4.2 Frequency domain
For systems with a large number of DOFs the computational effort for solving
simultaneously may be very extensive. Performing a modal analysis using the mode-
superposition method allows solving a reduced system of uncoupled equations,
considering only ( ≪ ) oscillation modes. The procedure consists of transforming
these equations to its modal coordinates expressing the subsequent solution in terms of
modal contributions [27].
The accuracy of the solution, hence, underlies on the number of modes considered.
Calculating moment and shear forces requires a larger number of modes than
displacements and accelerations, due to the major contribution of higher oscillation
modes.
26
3.4. DYNAMIC ANALYSIS APPROACHES
Note that since this approach expresses the response as a superposition, it is only
applicable for linear elastic structures, as it is generally the case of structural dynamics
of bridges.
3.4.2.1 Fast Fourier transform (FFT)
Any arbitrary periodic signal can be expressed in either time or frequency domain, being
the last one the so-called spectrum of a signal. Such spectrum may be constructed as a
linear combination of different periodic functions associated to a different frequency
(Figure 3.8). This transform is already implemented in MATLAB [11].
Figure 3.8: Fast Fourier Transform1
Nyquist frequency, expressed in eq. (3.5), establishes the limit frequency necessary to
reconstruct a signal, the sampled frequency should be greater.
1
= (3.5)
2
Where relates the number of samples recorded in a total time .
1
Figure retrieved from http://mri-q.com/fourier-transform-ft.html
27
4.1. INTRODUCTION
Chapter 4
4.1 Introduction
In the present chapter, a case study is presented, where the methodology explained up
to this point will be applied. The structure selected for this study is the Ulla River
Viaduct (Figure 4.1), which constitutes the most remarkable intervention for the High-
Speed Atlantic Railway Line [34]. It is worldwide recognised and resulted finalist for
the 2016 Outstanding Structure Award given by the International Association for Bridge
and Structural Engineering (IABSE). It is located in Galicia, Spain, as shown in Figure
4.2. IDEAM was responsible of the design project and TYPSA oversaw the technical
assistance at construction, but many other companies were involved in the entire project.
Figure 4.1: Ulla River Viaduct
29
CHAPTER 4. STUDY CASE: ULLA RIVER VIADUCT
The reasons for choosing this structure were:
Its complex typology made this structure attractive for a fully study of its dynamic
behaviour.
The torsional circuit of the deck is closed by an alternative solution different to the
typically employed in bridges with similar typologies.
Detailed information of the bridge, including design drawings and testing results of
the real built bridge, has been provided.
Figure 4.2: Location of Ulla River Viaduct
4.2 Description of the structure

The structure has a total length of 1620 m, distributed in 12 spans of 225+240+225 m
the main ones, several approaching spans of 120 m and side spans of 50 m and 80 m
(Figure 4.3).
Given the project constraints, seeking to minimize the number of piers and avoid the
environmental impact to certain extent, the resultant solution is composed of a steel
lattice composite bridge, with double steel-concrete composite action near the supports.
Figure 4.3: Elevation view of Ulla River Viaduct
The cross-section is variable in depth, from 17.5 m near the supports and 8.75 m at
midspans. The total width is 14 m, and it is equipped with two ballasted railway-tracks.
The concrete top slab has its maximum depth of 0.77 m over the upper chords axis, the
minimum one of 0.39 m in the edges and a central part of 0.46 m. The two steel lattices
are rotated approximately 45º with respect to the horizontal, the upper chords are
separated a distance of 6 m to each other. The characteristic cross-section is shown in
Figure 4.4. More information about materials and cross-sections can be found in
Appendix A.
30
4.2. DESCRIPTION OF THE STRUCTURE
Figure 4.4: Characteristic cross-section of Ulla River Viaduct
In the main central spans, the piers are rigidly connected to the deck resulting in frames.
This configuration gives the sufficient stiffness required for the passing trains. The two
central piers (P-6 and P-7) are hollow boxes cross-section, whereas the two resting ones
(P-5 and P-8) are non-connected shafts, allowing the movements due to imposed
displacements.
The double composite action near the supports is achieved thanks to the cast-in-place
bottom slab (Figure 4.5). Where this action is not required, non-connected prefabricated
plates of 2 m long are resting on the bottom chords (Figure 4.6). This configuration
results in an increment of the torsional rigidity of the deck due to the closed section.
Figure 4.5: Views of the bottom slab taken from the real bridge
31
Figure 4.6: Plan view and detail of the prefabricated plates (bottom slab)
4.3 FE models
Due to the complexity of the structure together with the aim of the study, the 3D FE
base model (hereon named as MODEL 1) reproduced in detail the design drawings of
the bridge. Several parts of it were altered for study purposes. Accordingly, six different
FE models and three different train load positions have been examined.
ABAQUS [10] is composed of separated modules dedicated to the different aspects

necessary in the modelling process. In that sense, for developing the FE model was
required to define the geometry, material, constraints, loads and generate the mesh using
the modules Part, Property, Assembly, Step, Interaction, Load and Mesh in that order.
While moving along the modules an input file was created and modified to generate the
moving loads. It was thereupon submitted to calculation.
Dynamic analyses using this commercial package can be quite time and memory
consuming. In this context, MODEL 1 resulted in 104237 elements, 143137 nodes and a
total number of 674292 variables in the model, including all degrees of freedom plus
maximum number of any Lagrange multiplier variables. Nevertheless, one of its benefits
that it is possible to select only a few nodes to store the results and use multiple
processors in parallel leading to a noticeably computing time and computational memory
reduction.
This section organization starts with a detailed description of the MODEL 1, including
the geometry, mass and cross-section and finally some aspects related to the mesh. It
proceeds later by summarizing the total six models and train load locations.
4.3.1 Assumptions and limitations
All models only considered the main central spans and two approaching spans. In this
respect, the models comprised from pier P-4 to pier P-9 (Figure 4.3). The greater
dynamic effects were expected in these central main spans, but on the other hand, the
excitation caused by the train when entering and leaving the bridge is always higher in
the extreme spans, thus, some approaching spans had to be also modelled.
32
4.3. FE MODELS
For simplification, the bracings of the upper truss have been neglected, among other
reasons, because their placement is mainly related to construction purposes and not so
much to its dynamic response. Soil-structure interaction contribution has been also
neglected and the stiffness of the piers P-4 and P-9 (see Figure 4.3) was assumed to be
infinite in the vertical direction.
The railway track was assumed to be one beam, which is the same as saying that both
wheels of the train were concentrated in one axle.
4.3.2 Geometry and materials
This subsection describes the geometry and materials employed in MODEL 1. The
materials remained the same for all the other models, whereas the geometry could vary.
Both geometry and materials of the model were based on the design drawings, see
Appendix A for further detail.
The following members of the bridge were modelled in separated parts: truss, top slab,
bottom slab, head, base and body of the piers. Figure 4.7 shows the geometry of the
assembled MODEL 1.
Figure 4.7: Elevation view of the assembled parts of MODEL 1 in ABAQUS
The modal damping ratio applied to all the modes during the simulation is 0.005,
according to Table 4.1. Moreover, the mass- and stiffness- proportional Rayleigh
damping were introduced as and β respectively, calculated as shown in eq. (4.1).
2 ( − ) 2( − )
= , = (4.1)
− −
Where,
and are the cyclic natural frequencies of the i and j modes, in rad/s
and are the critical damping ratio for the i and j modes, in this case it has
been considered 0.5% for all modes (Table 4.1).
33
Table 4.1: Lower bound values of the structural damping to be used in design of railway
bridges [16]
Lower limit of percentage of critical damping [%]

Bridge Type Span < 20 m Span ≥ 20 m
Steel and composite = 0.5 + 0,125 (20 − ) = 0.5
Prestressed concrete = 1.0 + 0,07 (20 − ) = 1.0
Filler beam and reinforced
= 1.5 + 0,07 (20 − ) = 1.5
concrete
The altered density of the concrete top slab accounted the dead loads of non-structural
elements such as ballast, edge beams, railway-track and safety equipment. More
precisely, such dead loads have been accounted according to [4], considering that the
weight of two ballasted tracks and all bridge finishes has typical values of 120 kN/m.
This weight was assumed to be concentrated in the central part of the deck, to be more
realistic, where the ballast was resting (Figure 4.8). Accordingly, the total density of the
concrete top slab was 31750 kg/m distributed in an area of 3 m with density =
2500 kg/m and 4.36 m with density = 5552 kg/m .
Figure 4.8: Equivalent density distribution over concrete top slab
4.3.3 Boundary conditions
Despite the only modelling of the central spans of the bridge, the continuity of the real
one should be simulated by applying boundary conditions.
According to the design drawings, the supports in piers P-4 and P-9 allow displacements
in direction u1 (longitudinally) and u1 and u2 (transversally) each, as shown in Figure 4.9.
A realistic model would have taken into account the stiffness of the POT supports and
piers, in their respective direction. However, for simplification, the boundary conditions
in these extreme nodes were considered as free in all DOFs but u3 (vertical direction).
That simplification led to significant lower eigenfrequencies than if a spring support,
with the correspondent stiffness, would have been applied instead.
Finally, the restraint of the piers from the foundations was assumed as fixed (all twists
and displacements kept).
34
4.3. FE MODELS
a) Plan view Ulla River Viaduct. Central spans
b) Allowed displacements in supports P-4 and P-9
Figure 4.9: Supports piers P-4 and P-9
4.3.4 Loading
For the static calculations, the concentrated loads corresponding to the train load models
were applied as pressure loads to small surfaces (Figure 4.10). This led to a more realistic
approach load-structure interaction.
Figure 4.10: Concentrated loads modelled as pressure in surfaces
On the other hand, the moving loads were directly applied as concentrated loads to the
nodes of beam elements that simulated the railway track.
4.3.5 Mesh
Different members of the bridge model were associated to a particular mesh element. In
order to capture with accuracy the structural response against dynamic loads, the models
combined beam, shell and solid elements.
35
The lattice members were represented by beam elements rigidly connected to each other
simulating the welding unions. The element type chosen was B31, which uses linear
interpolation, based on Euler-Bernoulli beam theory but allows specifying transverse
shear strain (becoming suitable for both thick and thin beams).
Conversely, despite allowing large magnitude axial strains, it is assumed to be small

when computing torsional shear strain. For thick beams this could be an important issue
in dynamic analysis, but for slender beams is usually insignificant. In any case, the
rotatory inertia does not affect to the same extent that shear deformation effects.
Regarding the concrete members, the shaft and head of the piers were shell elements,
connected by solid elements capable to transfer the forces in all directions. Likewise, the
top slab was composed of shell elements accurately partitioned longitudinally to
reproduce the variable transversal depth. Finally, the cast-in-place and precast bottom
slabs, when considered, were also shell elements. The shell elements available were: linear
triangular (SR3) and quadrilateral (SR4), quadratic triangular (STRI65) and
quadrilateral (S8R). SR4 is a general-purpose shell element that is perfectly suitable for
thick and thins shell models in terms of robustness and accuracy, and for that reason
was employed in all shell elements.
Choosing the most appropriate mesh size required a convergence analysis, but some
aspects needed to be studied carefully when meshing. ABAQUS offers a wide range of
constraints modelling. In order to prevent the relative movement between two separate
edges or faces, the tie constraint is the preferred one for its generality. It is based in a
strict algorithm master-slave. Two approaches can be specified for the discretization
method: surface-to-surface and node-to-surface; where only the master is treated as a
surface. The slave nodes pass through the normal direction of the master surface and can
be penetrated by them, for that reason the surface mesh has to be coarser.
The mesh size of the diagonals was taken large enough to be simulated by only one
element each. These members were not subjected to significant bending moments or
shear deformations. Due to the interaction between the chords and the concrete slabs,
the mesh of the first members was thinner: element size of 0.5 m. However, due to the
configuration of the bridge, these members were not subjected to significant bending
moments or shear deformations either.
The mesh size for the solid elements that simulates the base of the piers was taken as 0.8
m, the size of the elements composing the body of the piers was thinner: 0.5 m. The
concrete top slab mesh size was 0.8 m. Finally, the shell elements of the bottom slab had
a size of 1 m.
4.3.6 Different models
The object of the present study is to study the dynamic response of the bridge when the
following aspects were altered:
Bottom slab configuration, considering open and closed cross-sections. The original
one is shown in Figure 4.11.
Mass and stiffness of the piers.
36
4.3. FE MODELS
Figure 4.11: Definition of the bottom slab, indicating the length of cast-in-place concrete and
prefabricated plates
As a consequence, six different models were necessary, hereon denominated as MODEL

1, MODEL 2, MODEL 3, MODEL 4, MODEL 5 and MODEL 6 and which
characteristics are described throughout this section. See Appendix A/section A.3 for
further detailed description of the models.
MODEL 1:
The most realistic and detailed model reproduced with fidelity de design drawings of
the bridge. The original configuration of the bottom slab was respected (Figure 4.11).
The cast-in-place bottom slab was variable in depth and rigidly connected to the
bottom chord. All shell elements of the prefabricated plates were non-connected to
each other and simply supported over the bottom chords, what means no rotational
degrees of freedom allowed in the interaction between these parts. Figure 4.6 shows a
3D view of the central span of MODEL 1 and a zoom of its middle part.
Figure 4.12: Central span view of MODEL 1 and a zoom of the prefabricated plates separation
MODEL 2:
This model did neglect the contribution of the non-connected prefabricated plates of
the bottom slab. Therefore, only the cast-in-place bottom slab was modelled (see
Figure 4.11), keeping the double steel-concrete composite action near the supports. It
was performed to study two aspects: 1. if the only length of cast-in-place bottom slab
was sufficient to transfer the torsional flow to the piers and 2. to determine up to what
extent the placing of the prefabricated plates affected its dynamic response. Figure
4.13.a shows a zoom of the central span of this FE model.
37
MODEL 3:
This model aimed to determine if the discontinuity in the prefabricated plates of the
bottom slab offered more advantages than if those plates were connected to each
other, from a dynamic point of view. Consequently, the bottom slab was continuous
along all the span length as it can be discernible from the zoom presented in Figure
4.13.b.
MODEL 4:
In this model, an open cross-section of the deck was considered, decreasing noticeably
its torsional rigidity. For this purpose, neither cast-in-place nor prefabricated plates
of the bottom slab were modelled (see Figure 4.13.c). It is worth highlighting that in
this configuration the double steel-concrete composite action near the supports was
supressed, affecting the flexural longitudinal behaviour as well.
a) b) c)
Figure 4.13: Zoom of the central span of: a) MODEL 2, b) MODEL 3 and c) MODEL 4
MODEL 5:
To determine how much the mass of the piers contributed in the dynamic response of
the bridge, this model assumed null density for the concrete of the piers.
MODEL 6:
This model did not consider the body of the piers. On the contrary, the head of the
piers that allows the rigid connection to the truss remained. The boundary conditions
applied (as fixed) resulted in a configuration like a continuous beam with fixed
intermediate supports.
4.3.7 Railway track locations
As it can be drawn from subsection 4.3.6, the torsional response of the bridge in front of
dynamic loads plays a crucial role, and therefore the position of the railway-track must
excite it. For this purpose, three different configurations were contemplated (hereon
denominated as Position A, Position B and Position C), accounting one or two trains
travelling in the same direction depending on the case. According to Figure 4.14:
Position A: One train travelling directly over one upper chord axis – Train load (1).
Position B: Two trains travelling in the same direction, each of them directly over the
axis of its respective upper chord. This configuration did not excite torsion – Train
loads (1) and (3).
Position C: One train travelling across the centre of the deck; this position does not
excite torsion either - Train load (2).
38
4.4. MODEL VALIDATION
Figure 4.14: Definition of the train load (1), (2) and (3) locations over the top slab
4.4 Model validation

Incorrect inputs lead to poor results, especially when a dynamic analysis is involved.
Some relevant aspects such as mass, stiffness or boundary conditions may affect
considerably the subsequent results. For that reason, the quality control of MODEL 1
included checks/verifications from a static as well as dynamic point of view. The
available information to validate the model was the static loading testing of the real
bridge plus its expected natural frequencies and mode shapes.
4.4.1 Mass check
The total weight of the model was 76360 t corresponding to a relative error of 6 % from
the expected weight calculated according to the drawings (75905 t). This error was
considered acceptable on the part of the author. When neglecting the weight of the piers,
the total mass of the model was reduced to 59905 t, which equivales to 65 t/m.
4.4.2 Static loading test
MODEL 1 was also verified against experimental measurements regarding static

deflections. In [35] the results obtained in the static loading testing of the real viaduct
are provided. Several locations of the train were considered in order to achieve the
maximum deflection for each span. The measurements were obtained using displacement
transducers with an expected minimum accuracy of 5 %, located in the following points,
referred to the corresponding span and pier numbering shown in Figure 4.15:
Span 6: Bottom chord, midspan. Approximately at 112.5 m from P-5 and P-6 pier
axis.
Span 7: Upper chord, midspan. Approximately 120 m distance far from P-6 and P-7
pier axis.
Span 8: Bottom chord, midspan. Approximately at 112.5 m from P-7 and P-8 pier
axis.
39
Figure 4.15: Measurement points
The railway vehicle used in the loading testing is composed of four sets of one locomotive
333.3 and a maximum of six hoppers RENFE-80T (see Appendix C, Figure C.1). The
loading positions are defined in Appendix C, they accounted two tracks loaded in all the
cases. Due to the only modelling of the central spans four loading positions were checked,
named from loading position 11 to 14.
The measured displacements together with the calculated ones are shown in Table 4.2.
Table 4.2: Measured displacements vs. calculated for loading testing
Loading Measured Calculated

Span [%]
position [mm] [mm]
6 -98.48 -96.50 2.05
11
8 -89.86 -96.99 7.35
12 7 -72.68 -88.18 17.58
6 -95.58 -83.26 14.80
13 7 -66.82 -73.13 8.63
7 -66.50 -71.80 7.38
14
8 -86.98 -83.48 4.19
4.4.3 Modal analysis
A 2D plane truss model was developed by IDEAM [35] to determine the first natural
frequencies of the bridge, aiming to compare the results calculated with the experimental
ones obtained during the dynamic load testing. In this respect, the dynamic load testing
should aim to excite the first bending mode of the main span.
Figure 4.16: First bending mode for the main span correspondent to a natural frequency of
0.813 Hz
Despite the absence of experimental results, the calculated theoretical first bending mode
of the main span was 0.765 Hz. The respective mode obtained in ABAQUS for MODEL
1 was associated to a frequency of 0.813 Hz, resulting in a relative error of 6.27%, which
was considered within acceptable limits.
40
4.4. MODEL VALIDATION
On the other hand, TYPSA provided information about the first 120 natural frequencies
and mode shapes that were employed to study the influence of the torsional modes on
the dynamic response of the bridge [19]. The FE model (Figure 4.17) was performed
using the software SAP 2000 [36]. The comparison of natural frequencies between both
models is shown in Appendix B/section B.2.
Figure 4.17: SAP 2000 model of the bridge developed by TYPSA
4.4.4 Boundary conditions
As explained in subsection 4.3.3, the boundary conditions in the side spans allowed the
movements in all directions except the vertical one. In order to analyse the effect of this
assumption in the overall dynamic response of the bridge, the movements in the
transversal direction were restrained in a modified MODEL 1. The obtained results in
terms of accelerations and displacements for this modified MODEL 1 and the original
MODEL 1 were compared (Figure 4.18, Figure 4.19).
0.1
Span 6, u2 free
0.08 Span 6, u2 restrained
Accelerations [g]
Span 7, u2 free
0.06
Span 7, u2 restrained
0.04 Span 8, u2 free
0.02
0
200 250 300 350 400
Speed [km/h]
Figure 4.18: Accelerations for MODEL 1 and MODEL 1 modified boundary conditions
40
Span 6, u2 free
35 Span 6, u2 restrained
30 Span 7, u2 free
[mm]
25 Span 7, u2 restrained
Span 8, u2 free
20
15
10
200 250 300 350 400
Speed [km/h]
Figure 4.19: Displacements for MODEL 1 and MODEL 1 modified boundary conditions
41
One might note that despite the undertaken simplification highly affected the oscillation
modes, the dynamic response when a train HSLM-A1 was travelling across the bridge
was almost identical. The peak observed for velocities around 280 km/h in MODEL 1
was not produced in the modified MODEL 1. However, the author assumed that this
very small difference would not have affect the final conclusions and thus could be
neglected for this study.
4.5 Total simulation and time step selection

The implicit integration scheme (see section 3.4) has been employed for the dynamic
calculations in ABAQUS. The stability of the solution for a moving subsystem over a
primary one is determined by the transient response [29].
ABAQUS/Standard allows to specify transient fidelity categorise in the direct

integration step. In applications with a physical time scale, parameters to set the wished
level of accuracy can be introduced. This implies that the software would be in charge of
choosing the time step increments. Likewise, it is possible to establish a lower and an
upper limit of time step for the resolution of the system of equations. In this case, when
a large number of iterations are needed, the time step is automatically taken as minimum
whereas if only a few iterations are sufficient, this time step will increase up to the
maximum defined [37].
All the simulations lasted until the last axle of the train had left the bridge plus 5 s, the
total time depended thus on the train speed. Regarding the time step and despite the
literature recommendation (see subsection 3.4.1.3) of choosing at least one eighth of the
minimum period considered, the accuracy gained was not worth the time consumption.
In that sense, an example for a particular train speed of 300 km/h the total simulation
time is of 9.21 h, 3.98 h and 2.81 h corresponding to the time step of 0.002 s, 0.005 s and
0.008 s respectively. For that reason, the upper limit of time step for the dynamic
calculations was chosen to be 0.008 s. Figure 4.20 and Figure 4.21 show the dynamic
responses when different time steps were contemplated, revealing that the acceleration
peaks are slightly shifted towards the right, but the maximum values are almost the
same.
0.1
Span 6, t=0.008s
0.08 Span 6, t=0.005s
Accelerations [g]
Span 6, t=0.002s
0.06 Span 7, t=0.008s
Span 7, t=0.005s
0.04 Span 7, t=0.002s
Span 8, t=0.008s
0.02 Span 8, t=0.005s
Span 8, t=0.002s
0
200 250 300 350 400
Speed [km/h]
Figure 4.20: Accelerations for different time steps
42
4.6. MOVING MASS-MOVING LOAD COMPARISON
40
Span 6, t=0.008s
35 Span 6, t=0.005s
Span 6, t=0.002s
30
Span 7, t=0.008s
[mm]
25 Span 7, t=0.005s
Span 7, t=0.002s
20
Span 8, t=0.008s
15 Span 8, t=0.005s
Span 8, t=0.002s
10
200 250 300 350 400
Speed [km/h]
Figure 4.21: Displacements different time steps
4.6 Moving mass-moving load comparison

As a final verification, the relevance of neglecting the inertial effects when simulating
the train as moving loads has been studied. To this effect, a moving point mass and a
moving load, both corresponding to 170 kN, were applied to MODEL 1. The results in
terms of displacements (Figure 4.22) exposed an insignificant difference, supporting that
the selection of modelling the train as moving loads was correct.
Moving mass Moving load
Midspan 7, v=200km/h
0.5
0
Displacement [mm]
-0.5
-1
-1.5
0 5 10 15
time [s]
Figure 4.22: Comparison Moving load-Moving mass
43
5.1. INTRODUCTION
Chapter 5
Results and discussion
5.1 Introduction
This chapter presents the results obtained along this work as well as a brief discussion
of them. It first starts by the outcome obtained from the modal analysis. It then proceeds
to show the static and dynamic calculations.
5.1.1 Dynamic aspects examined
Ulla River Viaduct is an existing bridge and therefore all the design code requirements
were already verified in its design stage. For that reason, this work does not contemplate
checks/verifications but it does examine some of the most relevant aspects in terms of
dynamics. Those are:
Vertical accelerations. The limitation of the vertical accelerations is a requirement

present in all the relevant design codes (see subsection 2.4.3) because it involves from
safety (ballast stability) to comfort conditions for the users.
Vertical deformations. In long span bridges the vertical deformations tend to be the
demanding parameter. They are limited for comfort purposes (see subsection 2.4.3).
Amplification factor. It represents the increment of the dynamic deflection with
respect to the static one. The last one may be obtained for train speeds slow enough
(see subsection 2.4.4). In this case it was considered 50 km/h, despite not being
extremely slow the subsequent results were in the safe side.
Envelope dynamic factor. It indicates if the calculated load effects due to high-speed
are larger than those caused by conventional railway traffic (characterised by a
nominal train load model). It was calculated according subsection 2.4.2.
The dynamic calculations were carried out for train speeds from 200 km/h to 420 km/h
with a step of 20 km/h. Results for the entire range of train speeds were obtained by
interpolation using cubic splines. There is available an already implemented function in
MATLAB for splines. All the results are represented as envelopes of the maximum values
obtained over time.
45
CHAPTER 5. RESULTS AND DISCUSSION
Finally, it is also presented the twist of the deck caused by torsion, for a given train
speed. It corresponds to the angle, in radians, that the deck rotates around the
longitudinal axis.
5.2 Modal analysis

Appendix B shows the natural frequencies and mode shapes of MODEL 1. MODELS 1
to 3 had similar values of natural frequencies, whereas the ones of MODEL 4 were lower
due to the significantly low-value of torsional inertia of the deck. The first torsional mode
for this last referred model corresponded to mode 10 and it was associated to a frequency
of 0.77 Hz (Figure 5.1). MODEL 6 had totally different mode shapes that were similar
to those expected for a continuous beam with fixed supports.
Figure 5.1: First torsional mode for MODEL 4 corresponding to mode 10 and 0.77 Hz
According to the design codes, the cut off frequencies needed to calculate using the mode
superposition method is 30 Hz. MODEL 1, resulted in more than 370 oscillation modes
of the deck, after all non-relevant local modes had been supressed. The influence of the
oscillation modes contribution on the dynamic response can be observed in Figure 5.2,
where the Fast Fourier Transform (FFT) of MODEL 1 and MODEL 4 are compared for
vertical accelerations. These results were extracted in span 7 for a train speed of 200
km/h. The time step and total time considered were 0.008 s and 25 s respectively. Its
associated Nyquist frequency was 62.5 Hz, however, the plot shows values up to 30 Hz
because of scale reasons. These transforms were normalised with respect to its maximum
values.
The maximum peak for both models was produced around 3 Hz. The explanation rests
in the fact that HSLM-A1 train’s axles separation is 18m, for this particular velocity,
resulted in an impact every 0.324 s (3.08 Hz).
Note that all the peaks found for MODEL 4 were slightly shifted towards the right
because of its lower values of natural frequencies. In this sense, the points indicated as 1
and 2 corresponded to the first bending mode of this span, 0.76 Hz and 0,81 Hz
respectively. Points named 3 and 4 corresponded to torsional modes only existing in
MODEL 4.
46
5.3. MEASUREMENT POINTS
Span 7
1
FFT normalised az 0.8 MODEL 1

MODEL 4
0.6
o 2 o 3
0.4
o 1
o 4
0.2
0
0 5 10 15 20 25 30
frequency [Hz]
Figure 5.2: FFT of the accelerations obtained in span 7 for MODEL 1 and MODEL 4 when a
train is travelling at 200 km/h, normalised with respect to its maximum values
5.3 Measurement points

Due to the modelling of the train as a singly-lane, located over the same vertical plane
as the upper chord, the overall behaviour of the bridge will not be symmetrical because
of the torsional excitation. As a consequence, the results obtained were expected to vary
significantly depending on the measured node, not only in the longitudinal direction but
also in the transversal one.
In Figure 5.3 is indicated the name used hereon when presenting the results. The
numbering of the spans respected those assigned in the design drawings, and were kept
the same in the entire present document. Extrapolating to the FE model, the
measurement point named as deck relates to the central node of the shell elements that
represents the concrete top slab. On the other hand, upper chord refers to the central
node of the upper chord belonging to the vertical plane where the train load was applied.
Figure 5.3: Measurement points
47
5.4 Static displacements
5.4.1 For high-speed trains (HSLM-A1)
The static deflection caused by high-speed trains, in this case HSLM-A1, was
needed for calculating the amplification factor as well as the envelope dynamic factor. It
was obtained as the maximum values reached in all three mid spans, when the train was
travelling at a speed low enough. The speed considered in this work was 50 km/h and
despite not being extremely slow, the maximum results obtained were in the safe side.
That is because the dynamics effects produced were almost negligible and the maximum
displacements obtained were greater than if lower velocities would have been taken.
Figure 5.5 and Figure 5.5: Static displacement caused by train HSLM-A1 travelling at
50km/h. Measurement point: deckFigure 5.5 present the results for MODEL 1 and
MODEL 6 respectively. The effect of the pier´s stiffness was particularly important for
the maximum displacements obtained. In this context, in MODEL 1 the stiffness of the
piers P-5 and P-8 (Figure 4.3) conferred certain flexibility to the displacements. Besides,
when the train passed over one span, it influenced the surrounding ones. All of this led
to reaching the maximum displacements in the extreme spans. On the contrary, MODEL
6 yielded to lower deflections in general, since the piers’ DOFs were restrained in all
directions. The maximum values were reached in the central span due to its larger length.
It can also be drawn from the aforementioned figure that the history output of the static
deflections can be understood as an envelope of the critical position of the train.
10
0
[mm]
-10
-20
0 20 40 60 80 100
time [s]
Figure 5.4: Static displacement caused by train HSLM-A1 travelling at 50km/h. Measurement
point: deck
48
5.4. STATIC DISPLACEMENTS
[mm] 0
-5
-10
-15
0 20 40 60 80 100
time [s]
Figure 5.5: Static displacement caused by train HSLM-A1 travelling at 50km/h. Measurement
point: deck
The values obtained for all models and spans, when measuring in both upper chord and
deck, are gathered in Table 5.1 and Table 5.2 respectively.
Table 5.1: Static deflection caused by a train travelling at 50 km/h. Measurement point: deck
[mm]
Span 6 Span 7 Span 8
Model 1 99.15 90.30 99.15
Model 2 86.68 91.82 86.68
Model 3 99.43 91.82 99.43
Model 4 123.07 113.15 123.07
Model 5 99.15 90.30 99.15
Model 6 58.48 53.24 58.48
Table 5.2: Static deflection caused by a train travelling at 50 km/h. Measurement point: upper
chord
[mm]
Model 1 18.97 15.64 19.04
Model 2 19.50 16.11 19.59
Model 3 19.63 16.17 19.45
Model 4 19.50 16.11 19.59
Model 5 18.97 15.64 19.04
Model 6 19.04 18.97 15.64
49
5.4.2 For nominal train LM-71
The static deflection caused by the nominal train load model was needed for
calculating the envelope dynamic factor. The loading position depended on the span
wished to analyse, so that the maximum deflection was reached. That is, loading the
span considered and its alternates ones. The results obtained in the centre of each span,
measuring in the deck and in the upper chord are resumed in Table 5.3 and Table 5.4
respectively. All these results are already multiplied by the classification coefficient of
= 1.21.
Table 5.3: Static deflection for train model LM-71 [23] (classification coefficient = 1.21).
Measurement point centre of deck
[mm]
Model 1 99.15 90.30 99.15
Model 2 86.68 91.82 86.68
Model 3 99.43 91.82 99.43
Model 4 123.07 113.15 123.07
Model 5 99.15 90.30 99.15
Model 6 58.48 53.24 58.48
Table 5.4: Static deflection for train model LM-71 [23] (classification coefficient = 1,21).
Measurement point centre of upper chord
[mm]
Model 1 98.22 90.71 98.22
Model 2 85.22 90.79 85.22
Model 3 98.34 90.76 98.34
Model 4 121.77 113.57 121.77
Model 5 98.22 90.71 98.22
Model 6 52.42 57.624 52.42
5.5 Accelerations and dynamic displacements

Figure 5.6 and Figure 5.7 show the accelerations calculated for the measurement points
upper chord and deck, caused by the load model HSLM-A1. As it can be observed, in all
the cases the design requirements for ballasted bridges ( < = 0.35g) were
fulfilled.
These results show the influence of the effects produced when the train is entering to and
leaving the bridge, reaching the maximum values of accelerations in span 6 and span 8.
When the train was entering to the structure, it caused forced excitations whereas when
it was exiting the structure was allowed to vibrate freely.
50
5.5. ACCELERATIONS AND DYNAMIC DISPLACEMENTS
Figure 5.6 clearly reveals that when the measurement point was the centre of the deck,
all the models followed the same pattern. A peak response in accelerations was observed
around 360-380 km/h. This peak was shifted towards 380-400 km/h for MODEL 6
because de frequencies and mode shapes for this model were quite differentiated from
the others.
MODEL 4 gave rise to larger accelerations within the speed range 320-420 km/h than
all the other models, excluding MODEL 6, for midspans 6 and 8. This fact was
particularly noticeably in midspan 7.
MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6
Midspan 6. Deck
0.1
Accelerations [g]
0.08
0.06
0.04
0.02
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
Midspan 7. Deck
0.1
Accelerations [g]
0.08
0.06
0.04
0.02
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
Midspan 8. Deck
0.1
Accelerations [g]
0.08
0.06
0.04
0.02
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
Figure 5.6: Accelerations in the deck caused by HSLM-A1
Higher accelerations were reached in the upper chord where the axle of the train was
passing. In this member, both torsion and bending were important. In this case, the
effects of the train entering the bridge were different for each model. Conversely, same
pattern has been observed in midspans 7 and 8 in all the models. Likewise, the
51
acceleration peaks in midspan 6 were produced for different speeds than in midspan 8,
with the exception of MODEL 6. It is important to highlight that the accelerations
increased as the train speed did (Figure 5.7).

Midspan 6. Upper chord
0.4
Accelerations [g]
0.3 a =0,35g
lim
0.2
0.1
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
0.4
Accelerations [g]
0.3
a =0,35g
lim
0.2
0.1
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
0.4
Accelerations [g]
0.3
a =0,35g
lim
0.2
0.1
0
200 220 240 260 280 300 320 340 360 380 400 420
Speed [km/h]
Figure 5.7: Accelerations in upper chord caused by HSLM-A1
The displacements are represented in Figure 5.8, for both measurement points: deck and
upper chord. Their values were observed to increase together with the train speed,
especially for high velocities (380-420 km/h).
An important aspect to highlight is that, considering an open cross-section had more

influence on the deflections than the stiffness of the piers. In that sense, MODEL 6
resulted in the lower displacements due to the clamping of the piers whereas MODEL 4
yielded the maximum displacements due to the torsion. One may note as well, that in
span 7 the magnitude order of the displacements was similar for MODEL 1, MODEL 2,
52
5.5. ACCELERATIONS AND DYNAMIC DISPLACEMENTS
MODEL 3 and MODEL 5 in comparison to MODEL 6, implying that the two central
piers gave sufficient stiffness.
The results obtained in the deck and upper chord are presented side by side for the sake
of clarity. It is straightforward to see the effects due to the torsion of the deck; the
displacements reached in both measurement points were similar for all the models
excepting MODEL 4. The torsional rotation of the deck is noticeably for this last model.
That fact implies that the geometry was distorted, the two steel lattices were working
separately and on their own, increasing considerably the dynamic effects.

Midspan 6. Deck Midspan 6. Upper chord
40 40
35 35
30 30
[mm]
[mm]
25 25
20 20
15 15
10 10
200 250 300 350 400 200 250 300 350 400
Speed [km/h] Speed [km/h]
40 40
35 35
30 30
[mm]
[mm]
25 25
20 20
15 15
10 10
200 250 300 350 400 200 250 300 350 400
40 40
35 35
30 30
[mm]
[mm]
25 25
20 20
15 15
10 10
200 250 300 350 400 200 250 300 350 400
Figure 5.8: Dynamic displacements induced by HSLM-A1
In overall, these results showed a similar magnitude order in terms of accelerations and
displacements for all the models with exception of MODEL 4, where the maximum
values were generally reached. This fact is highly meaningful because it demonstrates
the importance of closing the torsional circuit of the deck.
53
It can be also drawn that the mass of the piers did not have a significant influence on
the dynamic response. As a final remark, it is important to note that generally, MODEL
1 yielded results comprised between MODEL 3 and MODEL 4 because it corresponded
to the intermediate case. In addition, these results were substantially alike to those
obtained from MODEL 2.
5.6 Amplification factor

Figure 5.9 depicts the amplification factor for the deck and the upper chord, calculated
as the ratio of the dynamic displacements to the static ones induced by HSLM-
A1.

2.5 2.5
2 2
dyn sta
dyn sta
/
1.5 1.5
1 1
200 250 300 350 400 200 250 300 350 400
2.5 2.5
2 2
dyn sta
dyn sta
/
/
1.5 1.5
1 1
200 250 300 350 400 200 250 300 350 400
2.5 2.5
2 2
dyn sta
dyn sta
/
1.5 1.5
1 1
200 250 300 350 400 200 250 300 350 400
Figure 5.9: Amplification factor. Deck
54
5.7. TWIST OF THE DECK
As it happened with the displacements, this factor increased together with the train
speed, fact particularly evident for high velocities (380-420 km/h). The amplification
factor was found near the unity in all the models except MODEL 4. In this model, the
results were dramatically different since the amplification factor could reach values near
to 2. The explanation rests on the fact that the torsional component was crucial in the
amplification of the dynamic deflections with respect to the static ones. For this same
reason, when measuring in the upper chord, these values were even higher; the twist of
the deck played an essential role.
5.7 Twist of the deck

Figure 5.10 presents the twist of the deck produced for a given train speed of 360 km/h
in MODELS 1-4. The results are only shown in midspan 7 because when observing the
dynamic displacements (Figure 5.8) the rotation of the deck was expected similar for all
the spans. Likewise, the train speed was selected in accordance to the peaks in the
dynamic response produced at this velocity (Figure 5.6). As expected, the results
reflected an outstanding twist of the deck for MODEL 4 in comparison to MODELS 1-
3.
Figure 5.10: Torsional twist of the deck in rad
5.8 Envelope dynamic factor

The displacements under HSLM-A1 presented in subsection 5.4.2, obtained by a
dynamic analysis served as the basis for calculating the envelope dynamic factor
according eq. (2.4). The subsequent results were compared to the dynamic factor Φ for
train speeds v ≤ 220 km/h, calculated according eq. (2.3). Considering the determinant
length = 279m resulted in a value of Φ = 0.907, taking consequently Φ = 1.0.
Figure 5.11 shows the values of the envelope dynamic factor Φ for HSLM-A1. For all the
models excepting MODEL 4 the magnitude order was the same. As occurred with the
amplification factor (section 5.6), when measuring in the upper chord the torsional
contribution led to higher values in this model.
Note that for all spans and models the value of impact coefficient for the universal train
was lower than the unit, which indicates that the deflections caused by LM-71 [19] were
larger than the dynamic deflections caused by HSLM-A1. In other words, the calculated
55
load effects for HSR traffic were lower than those to conventional railway traffic,
characterised by a nominal train.
This result is very useful from a practical point of view, because it implies the suitability
of adopting the impact coefficient Φ for evaluating the dynamic effects in the design
stage, in terms of deflections.

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
200 250 300 350 400 200 250 300 350 400
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
200 250 300 350 400 200 250 300 350 400
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
200 250 300 350 400 200 250 300 350 400
Figure 5.11: Impact coefficient for real trains. Deck
5.9 Different locations of passing train

Figure 5.12 presents the accelerations and dynamic displacements produced in MODEL
1 and MODEL 4 for different locations of the train. It allows several observations
regarding three different aspects: the train was entering to or leaving the bridge, the
56
5.9. DIFFERENT LOCATIONS OF PASSING TRAIN
induced effects were symmetrical/non-symmetrical and the consideration of an

open/closed cross-section of the deck.
MODEL 1. Position A MODEL 1. Position B MODEL 1. Position C

MODEL 4. Position A MODEL 4. Position B MODEL 4. Position C
Figure 5.12: Different location of passing trains
The maximum value of accelerations was reached for Position B. Two clearly
differentiated acceleration peaks were produced for train speeds ranges 320-360 km/h
and 360-400 km/h in both models for this position. Such peaks values were similar for
MODEL 1 and 4 in spans 6 and 7 but moderately larger for MODEL 4 in span 8. On the
other hand, Position A resulted in similar values of maximum accelerations but different
peaks location in the span 6. The explanation may rest in the fact that, even if the impact
was produced in the upper chord in both situations, the bridge response for Position B
was symmetrical. On the contrary, a slight peak was produced for Position C and train
57
speeds between 340-400 km/h, the values were minimum because the impact was not
produced in the upper chord.
It is noteworthy that the effect of the train entering to the bridge was more accused for
Position A than the train exiting it. Especially when considering open cross-sections.
When looking at the results obtained in span 7, the accelerations pattern was almost
identical for Position A and B. In this case, the stiffness that the piers conferred to this
span played a crucial role.
The values of the displacements were greater for Position B, as was expected. The static
displacements caused by two trains should be exactly twice as large as those caused by
one train. However, the dynamic displacements were not exactly double.
For all the cases, MODEL 4 resulted in the largest values of displacements and
accelerations. This implies that bridges with low torsional rigidity of the deck, a 2D
model would neglect significant accelerations and dynamic displacements.
58
6.1. CONCLUSIONS
Chapter 6
Conclusions
6.1 Conclusions
HSR traffic requires in general structures with greater stiffness than conventional traffic
ones. Currently, the vast majority of these bridges are equipped with a double railway-
track, fact that emphasises the importance of considering the torsional flows caused by
the eccentric loads of only one train passing.
This research focuses on studying the particular bridge of Ulla River Viaduct, where the
bottom steel truss traditionally used in composite steel-concrete lattice structures for
closing the cross-section torsional circuit is replaced by a concrete bottom slab. For the
study, a model that is capable to combine the dynamics of the truss members and the
deck was required.
The most important remarks that can be drawn from this study are presented in the
following paragraphs. This section starts commenting the findings concerning different
models and train locations and finally it proceeds to explain the conclusions regarding
the efficiency of the solution of concrete bottom slab configuration.
6.1.1 Regarding the different models
The study demonstrated the importance of closing the torsional circuit of the deck.
Models with open cross-sections yield the highest values, in terms of acceleration and
especially deflections, despite remaining far from the limiting values proposed in the
relevant codes. This fact became even more evident when the amplification factor of
deformations was examined, observing values twice as large as those corresponding
closed cross-sections. Finally, it is remarkable that while this bridge did not present
any problem associated to dynamic effects, its peculiar design could be of great
interest for other case studies where such dynamic effects pose a critical issue.
The mass of the piers mobilised did not have a significant influence on the dynamic
behaviour of this particular bridge. The calculated results were almost identical
whether it was contemplated or not. Conversely, totally different dynamic effects
were observed when the supports were assumed as fixed. This fact highlighted the
importance of modelling adequately their real stiffness to capture the real dynamic
59
CHAPTER 6. CONCLUSIONS
response of the bridge. It is noteworthy that the central piers conferred sufficient
stiffness to the main span; similar displacements were observed when the supports
were assumed as fixed.
6.1.2 Regarding different positions of the trains
Considerable influence of the train location on the accelerations and dynamic

displacements due to the excitation of different mode shapes. This evidenced the need
to perform 3D models that allow capturing the torsional modes properly. This is not
easy when simplified 2D or 3D beam models are used.
6.1.3 Regarding the efficiency of bottom slab configuration
No significant difference was found in the results when including in the model the
prefabricated plates of the bottom slab, regardless of being considered continuous or
not. That fact allowed concluding that the length of cast-in-place concrete was
sufficient for transferring torsional flows to the piers. Nevertheless, its original deck
configuration also results in more advantages non-related to its dynamic behaviour.
Among others, the extension of the bottom slab along the entire span facilitates
maintenance and inspection tasks as well as allows closing the formal section from
the bottom view, being more aesthetical. With all of that in mind, it may be remarked
that the solution of this bridge not only resulted very clever, but also more
economical than the traditional solution of placing a bottom steel truss.
6.2 Further research

The proposed possible future lines of work are summarized as follows:
Provide more validity of the results by modelling the entire bridge and contemplating
all 10 HSLM
Research if there would be a pronounced variation of the results when a complex
bridge-vehicle interaction model is performed.
Quantify the minimum torsional rigidity (with respect to the longitudinal one), for
which the torsional effects can be neglected and a 2D model could be considered.
Examine a greater number of HSR bridges with similar characteristics in order to
contrast and generalize the conclusions.
Broaden the study to viaducts with completely different structural forms. Determine
if the conclusions obtained in this thesis are applicable.
Investigate possible simplified models capable to capture the coupled torsional-
bending response of the studied bridge taking into account the stiffness that the piers
confer.
60
7. BIBLIOGRAPHY
Chapter 7
Bibliography
[1] N. HU y et al., «Recent development of design and construction of medium and

long span high-speed railway bridges in China,» Engineering structures, vol. 74,
pp. 233-241, 2014.
[2] U. I. U. o. Railways, «HIGH SPEED RAIL Brochure,» 2015.
[3] M. Vicente, D. C. González y Gongka, «Static and Dynamic Testing of High-Speed
Rail Bridges in Spain,» M.ASCE3 J. Bridge Eng, vol. 20(2), p. 1, 2015.
[4] A. Aparicio, «Differences in designing high-speed railway bridges and highway
bridges,» de In Workshop Bridges for High-Speed railways, Faculty of
Enegineering, University os Oporto, Delgado, R.; Calçada, R.; Campos, A, 2009,
pp. 221-236.
[5] L. Frýba, «Dynamic behaviour of bridges due to high-speed trainspages,» de In
Workshop Bridges for High-Speed railways, Faculty of Enegineering, University os
Oporto, R. Delgado, R. Calçada, A. Campos, 2004, pp. 125-142 .
[6] A. Andersson, «Höghastighetsprojekt - Bro - Delrapport 1: Befintliga krav och
erfarenheter samt parameterstudier avseende dimensionering av järnvägsbroar för
farter över 200 km/h,» Stockholm, 2010.
[7] J. Yau, « Resonance of continuous bridges due to high speed trains,» Journal of
Marine Science and Technology, vol. 9, pp. 14-20, 2001.
[8] F. Millanes, L. Matute, M. Ortega, D. Martínez y E. Bordó, «Development of steel
and composite solutions for,» Hormigón y Acero, vol. 62, nº 259, pp. 7-27, 2011.
[9] F. Gabaldón, J. Navarro, F. Riquelme y J. Domínguez, «Dynamic analysis of
hyperstatic structures under high-speed Railway bridges,» de In Workshop
Bridges for High-Speed railways, Faculty of Enegineering, University os Oporto,
R. Delgado, R. Calçada, A. Campos, 2009, pp. 143-158.
[10] ABAQUS, SIMULIA Inc., 2014.
[11] MATLAB, MathsWorks, 2014.
[12] J. Domínguez, «Dinámica de puentes de ferrocarril para alta velocidad: métodos
de cálculo y estudio de la resonancia. PhD dissertation,» Madrid, 2001.
[13] C. Svedholm, «Efficient Modelling Techniques for Vibration Analyses of Railway
Bridges. PhD dissertation,» Stockholm, 2017.
61
CHAPTER 0. BIBLIOGRAPHY
[14] J.-R. Cho, «Determination of the optimal span length to minimize resonance
effectsin bridges on high-speed lines,» J Rail and Rapid Transit 2014, 2014.
[15] C. Mellier, «Optimal Design of Bridges for High-Speed Trains. Single and double-
span bridges . Master dissertation,» Stockholm, 2010.
[16] E. C. f. S. (CEN), «UNE-EN 1991-2. Eurocode 1: Actions on structures – Part 2:
Traffic loads on bridges,» 2003.
[17] J. Goicolea, La consideración de los fenónmenos dinámicos en el proyecto de
puentes ferroviarios, http://w3.mecanica.upm.es: E.T.S. Ingenieros de Caminos,
Canales y Puertos, Universidad Politécnica de Madrid.
[18] F. Gabaldón, J. Goicolea, S. Ndikuriyo y J. Navarro, «Cáculo dinámico de un
viaducto continuo de sección mixta. Technical report in Spanish for IDEAM and
MC-2,» Madrid, 2002.
[19] J. Sánchez, J. Vaquero y R. Campoamor, «Dynamic response comparison between
four different structural bridge typologies used in the Spanish High-Speed Rail
Network,» de IABSE Congress. , Stockholm. September , 2016.
[20] I. Bisús, « Tipología de viaductos en las líneas de alta velocidad en España. Master
dissertation,» Barcelona, 2010.
[21] IAPF, «Instrucción sobre las Acciones a Considerar en el Proyecto de Puentes de
Ferrocarril Dirección General de Ferrocarriles,» Ministerio de Fomento, Madrid,
2010.
[22] E. C. f. S. (CEN), «EN1990-A2. 2004. Eurocode – Basis of Structural design,
Annex A2: Application for Bridges,» 2004.
[23] U. I. U. o. Railways, «Loads to be considered in railway bridge design,» nº 776-
1R, 2006.
[24] J. Goicolea, F. Gabaldón, J. Dominguez y J. Navarro, «Dynamic loads in new
engineering codes for railway bridges in Europe and Spain,» de In Workshop
Bridges for High-Speed railways, R. Delgado, R. Calçada, A. Campos, 2009, pp.
31-46.
[25] C. E. D214.2, « Utilisation de convois universels pour le dimensionement
dynamique de ponts-rails. Synthèse des résultats du D214.2,» European Rail
Research Institute (ERRI), 2002.
[26] D. Martin, «Railway bridges for high speed lines and Eurocodes,» de In Workshop
Bridges for High-Speed railways, Faculty of Enegineering, University os Oporto,
R. Delgado, R. Calçada, A. Campos, 2009, pp. 21-30.
[27] A. K. Chopra, Dynamic of structures: Theory and Applications to Earthquake
Engineering, Prentice-Hall International Series in Civil Engineering and
Engineering Mechanics, 2014.
[28] E. D192, «4th. report: Study of the construction costs of railway bridges with
consideration of the live load diagram,» Technical report, European Rail Research
Institute, 1995.
[29] A. Saleeb y A. Kumar, «Automated Finite Element Analysis of Complex
Dynamics of Primary System Traversed by Oscillatory Subsystem,» Comput.
Methods Eng. Sci. Mech., vol. 12(4), pp. 184-202, 2011.
[30] H. Ouyang, «Moving-load dynamic problems: A tutorial (with a brief overview),»
Mechanical Systems and Signal Processing , p. 2039–2060, 2001.
62
7. BIBLIOGRAPHY
[31] L. Fryba, «Vibration of Solids and Structures under Moving Loads,» Noordhoff,
Groningen, 1972.
[32] L. Baeza y H. Ouyang, «Dynamics of a Truss Structure and Its Moving-Oscillator
Exciter with Separation andImpact-Reattachment,» Mathematical, Physical and
Engineering Sciences, vol. 464, pp. 2517-2533, 2008.
[33] P. Antolín, «Efectos dinámicos laterals en vehículos y puentes ferroviarios
sometidos a la acción de vientos transversales, PhD dissertation,» Madrid, 2013.
[34] L. M. &. M. O. F. Millanes, «Viaduct over Ulla River in the Atlantic high speed
railway line: A composite (steel–concrete) truss world record,» Hormigón y Acero,
vol. 66(277), pp. e1-25, 2015.
[35] IDEAM, «Pruebas de carga estática y dinámica. Viaducto del Río Ulla, eje
Atlántico de alta velocidad. Technical report,» Madrid, 2015.
[36] SAP 2000, Computers&Structures, 2015.
[37] Hibbit, Karlsson y Sorensen, ABAQUS/Standard User's Manual, Pawtucket,
2006.
[38] J. Sobrino. Bridges for the high speed railway lines in Spain. Design criteria and
case studies, in: R. Delgado, R. Calçada, A. Campos (eds.). 2004. In Workshop
Bridges for High-Speed railways, pages 71-93 Faculty of Enegineering, University
os Oporto.
[39] Raid Karoumi. Course material KTH (AF2011). Structural Dynamics for civil
engineers.
63
A.1. MATERIALS
Appendix A
FE Models
A.1 Materials
A.1.1 Concrete
Strength Poisson Young modulus

Element Quality [ / ] coefficient E [MPa]
Precast Precast slabs C40/50 f ≥ 40 N/mm 0.3 35000
Piers C70/85 f ≥ 70 N/mm 0.3 41000
Cast-in-
Top Slab C35/45 f ≥ 35 N/mm 0.3 34000
place
Bottom slab C50/60 f ≥ 50 N/mm 0.3 37000
A.1.2 Steel
Strength Poisson Young modulus

Element Quality [ / ] coefficient E [MPa]
Structural
Steel profiles S 460 ML f ≥ 460 N/mm 0.2 210000
steel
65
APPENDIX A. FE MODELS
A.2 Cross sections
Figure A.1: Box cross-section
UPPER CHORD
PROFILE NAME t [m] b [m] a [m]
Box c1x0,8x30 0.03 1 0.8
Box c1x0,8x35 0.035 1 0.8
Box c1x0,8x40 0.04 1 0.8
Box c1x0,8x45 0.045 1 0.8
Box c1x0,8x50 0.05 1 0.8
Box c1x0,8x60 0.06 1 0.8
Box c1x0,8x65 0.065 1 0.8
Box c1x0,8x70 0.07 1 0.8
Box c1x0,8x75 0.075 1 0.8
Box c1x0,8x90 0.09 1 0.8
Box c1x0,8x110 0.11 1 0.8
BOTTOM CHORD
Box c1,2x0,8x25 0.025 1.2 0.8
Box c1,2x0,8x30 0.03 1.2 0.8
Box c1,2x0,8x35 0.035 1.2 0.8
Box c1,2x0,8x40 0.04 1.2 0.8
Box c1,2x0,8x45 0.045 1.2 0.8
Box c1,2x0,8x50 0.05 1.2 0.8
Box c1,2x0,8x55 0.055 1.2 0.8
Box c1,2x0,8x60 0.06 1.2 0.8
Box c1,2x0,8x65 0.065 1.2 0.8
Box c1,2x0,8x70 0.07 1.2 0.8
Box c1,2x0,8x90 0.09 1.2 0.8
66
A.2. CROSS SECTIONS
DIAGONALS
Box d1x0,7987x15 0.015 1 0.7987
Box d1x0,7987x20 0.02 1 0.7987
Box d1x0,7987x25 0.025 1 0.7987
Box d1x0,7987x30 0.03 1 0.7987
Box d1x0,7987x35 0.035 1 0.7987
Box d1x0,7987x40 0.04 1 0.7987
Box d1x0,7987x50 0.05 1 0.7987
Box d1x0,7987x55 0.055 1 0.7987
Box d1x0,7987x60 0.06 1 0.7987
Box d1x0,7987x65 0.065 1 0.7987
Box d1x0,7987x70 0.07 1 0.7987
Box d1x0,7987x75 0.075 1 0.7987
CONCRETE MEMBERS
PART NAME t [m]
Deck Deck_e1 0.770
Deck Deck_e2 0.460
Cast-in-place bottom slab HA50/30mm 0.030
Piers body P5 1.500
Piers body P6 0.600
Exterior edge. Pier head BECP 1.400
Transversal cover pier TTP 0.600
67
APPENDIX A. FE MODELS
A.3 FE Models
MODEL 1
- Discontinuity in the
prefabricated plates of
the bottom slab (middle
of spans)
- Cast-in-place bottom
slab near the supports
MODEL 2
MODEL 3
- Continuity in the
prefabricated plates of
the bottom slab (middle
of spans)
68
A.3. FE MODELS
MODEL 4
- Open cross-section of
the deck. No bottom
slab.
MODEL 5
- Original bottom slab

configuration (MODEL
1).
- Null density in the
concrete material of the
piers.
MODEL 6 -
- Original bottom slab

configuration (MODEL
1).
- Fixed supports.
69
B.1OSCILLATION MODES MODEL 1
Appendix B
Mode shapes
B.1Oscillation modes MODEL 1
Figure B. 1: Mode 1. = 0.21 Hz Figure B. 2: Mode 2. = 0.22 Hz
71
APPENDIX B. MODE SHAPES
72
B.1OSCILLATION MODES MODEL 1
Figure B. 14: Mode 14. = 1.28 Hz


73
APPENDIX B. MODE SHAPES
74
B2 COMPARISON EIGENFREQUENCIES
B2 Comparison eigenfrequencies
Natural frequencies f [Hz]
MODE SAP 2000 ABAQUS MODEL 2 er (%)
1 0.2865 0.2727 5.07
2 0.3053 0.2890 5.63
3 0.5010 0.5234 4.28
4 0.5301 0.5617 5.62
5 0.5773 0.6010 3.94
6 0.6883 0.6659 3.36
7 0.8005 0.9067 11.71
8 0.9130 0.9401 2.87
9 1.0312 1.0331 0.18
10 1.0447 1.1454 8.79
11 1.0878 1.1616 6.35
12 1.1956 1.1767 1.61
13 1.4503 1.3860 4.64
14 1.6927 1.6325 3.69
15 1.7242 1.6618 3.75
16 1.7364 1.8719 7.24
17 1.8642 1.9441 4.11
18 1.8781 1.9769 5.00
19 2.0604 2.0983 1.81
20 2.0817 2.2745 8.48
21 2.0830 2.3273 10.50
22 2.0939 2.3678 11.57
23 2.2764 2.5368 10.27
24 2.2895 2.5718 10.98
25 2.4969 2.5843 3.38
26 2.8710 2.7093 5.97
27 2.9072 2.9355 0.96
28 2.9413 2.9431 0.06
29 2.9946 2.9927 0.06
30 3.0256 3.0065 0.63
75
C.1. Vehicle definition
Appendix C
Static loading test
C.1 Vehicle definition
Figure C.1: Locomotive 333.3 composed of 6 axles of 20 t each and Hopper RENFE-80T
composed of 4 axles of 20 t each
C.2 Loading positions
Figure C.2: Loading position 11
77
APPENDIX C. STATIC LOADING TESTING
78
TRITA -BKN. Master Thesis 512, 2017
ISSN 1103-4297
ISRN KTH/BKN/EX 512 SE
www.kth.se

The Influence of Torsional Resistance of The Deck On The Dynamic Response

Uploaded by

Copyright:

Available Formats

The Influence of Torsional Resistance of The Deck On The Dynamic Response

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Influence of Torsional Resistance of The Deck On The Dynamic Response

Uploaded by

Copyright:

Available Formats

DEGREE PROJECT IN CIVIL ENGINEERING AND URBAN

The influence of torsional

CLAUDIA SANROMAN CERVERO

KTH ROYAL INSTITUTE OF TECHNOLOGY

Master of Science Thesis

© Claudia Sanroman Cervero, 2017

Understanding how different parameters affect the dynamic response of high-speed

Keywords: Dynamics, High-speed railway bridges, Torsional rigidity of the deck, FE

Entender cómo diversos parámetros afectan el comportamiento dinámico de puentes de

La finalidad de este trabajo es investigar la influencia de la resistencia torsional del

Stockholm, June. 2017.

Claudia Sanroman Cervero

Notations .............................................................................................................. vii

2 High Speed Railway Bridges............................................................................ 7

3 Dynamic analysis of bridges under high speed trains ....................................... 19

4 Case study: Ulla River Viaduct ...................................................................... 29

5 Results and discussion ................................................................................... 45

B Mode shapes .................................................................................................. 71

C Static loading test .......................................................................................... 77

Figure 1.1: Development of the world high-speed network [2]

1.2 Aim and scope

Demonstrate if the aforementioned design is efficient and investigate how its

Non-connected prefabricated plates of the bottom slab.

1.4 Assumptions and limitations

Only the central spans were modelled.

1.5 Outline of the thesis

High Speed Railway Bridges

2.1 Review of the special characteristics

2.1.1 Resonance and cancellation

2.2 Parameters that influence the dynamic response

a) Influence of span length l

b) Influence of damping theta. Case a out of resonance and case

2.2.1 Effect of torsional rigidity

a) Open cross-section b) Partially closed section, accelerations

The importance of considering coupled bending-torsional effects in dynamic analysis of

2.3 Longitudinal schemes

Table 2.1: Common cross-sections used in Spain [20]

Span length (m)

2.4 Envelopes and verifications

2.4.1 Train load models

An aspect worth emphasising is the convenience of interoperability of different families

Articulated trains: THALYS, AVE and EUROSTAR

Figure 2.6: Different typologies of high speed trains according to [19]

Table 2.2: Characteristics of HSLM-A

2.4.2 Envelope dynamic factor

2.4.2.1 Vertical train load model

Figure 2.8: Load Model LM-71 [23]

is the quasi static deflection caused by real high-speed trains. It may be

2.4.2.3 Treatment in different design codes

Vertical acceleration of the deck.

2.4.3.1 Vertical accelerations

2.4.3.2 Vertical displacements

2.4.4 Amplification factor

Dynamic analysis of bridges under

1. Static calculation through the dynamic factor or impact coefficient.

3.2 FE models for torsional considerations

3.2.1 Two-dimensional models

There exists a broad range of elements to be used in such models: two-dimensional

/ ≥ 10 Bernoulli beam elements