A Numerical Investigation of Longitudinal Track–Bridge

Interaction in Simply Supported Precast Concrete

Girder Railway Bridges
Alper Ozturk 1; Eray Baran 2; and Aysegul Askan 3
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Abstract: In railway tracks, additional stresses occur in rails due to interaction between the rail and the bridge superstructure, which is a
phenomenon known as track–bridge interaction (TBI). The additional rail stresses (ARSs) develop primarily as a result of temperature effect,
vertical bending of bridge superstructure, and braking/acceleration of trains traveling over the bridge. One of the major parameters affecting
the TBI response is the behavior of interface elements simulating the coupling between the bridge superstructure and the rail. In TBI analysis
this coupling is defined by the so-called longitudinal resistance-displacement curves (RDCs). The present study deals with the longitudinal
aspect of TBI though a numerical investigation. The numerical modeling approach was verified with available analytical solutions and data
obtained from bridge monitoring. The primary interest was to illustrate how sensitive the TBI response is to changes in RDCs with both
separate and sequential analysis approaches, which has not been investigated thoroughly earlier. The parametric study was extended to
investigate the effects of expansion length and substructure stiffness, in addition to RDC type. For simply supported precast girder bridges,
RDCs that are experimentally proven specific to ballastless railway bridges can be used as an economic alternative to those specified by
provisions. DOI: 10.1061/PPSCFX.SCENG-1476. © 2024 American Society of Civil Engineers.
Author keywords: Track–bridge interaction (TBI); Additional rail stress (ARS); High-speed railway; Continuous welded rail (CWR);
Railway bridge.

Introduction difference between the rail and bridge deck and the vertical and
braking/accelerating forces of the train passing over the bridge.
Advancements in welding technology have led to replacement of To provide the required safety margin, design codes provide maxi-
jointed high speed rail tracks with continuous welded rails (CWRs) mum tensile and compressive stress limits for rails (Mirković et al.
in an attempt to have a safer track under high-speeds, reduced 2018). These limits, however, usually do not include the rail
maintenance costs, increased service life properties, increased stresses due to temperature changes in the rail.
passenger comfort, and reduced noise emission (Tzepushelov In numerical modeling of railway bridges, the interaction
and Troyitzky 1974). Although CWRs have several advantages between the bridge superstructure and the rails is simulated with
(Enshaeian and Rizzo 2021) when compared with conventional interface elements and the behavior of these elements is defined by
jointed tracks, continuous nature of these tracks prevents free the so-called longitudinal resistance-displacement curves (RDCs).
expansion/contraction under thermal and mechanical loading. Such Depending on the presence or absence of a vertical train loading,
a restraint results in additional compressive and tensile stresses in the RDC is termed as either loaded or unloaded. As the RDCs re-
the rails, which must be taken into consideration during track
present the degree of interaction between the bridge deck and the
design. International railway design codes usually provide limits
railway track, they have a significant effect on the TBI response of
for additional rail stresses (ARSs) and displacements for bridge
the bridge system.
superstructure under thermal and train loads [UIC 774-3R (UIC
Two aspects of TBI need special consideration in the design of
2001); EN 1991-2 (CEN 2003); DIN Fb 101 (DIN 2003); KR
high-speed railway bridges. The vertical aspect of the problem has
C-08080 (Korea Rail Network Authority 2014)]. ARSs represent
previously been investigated in detail (Kim et al. 2010; Chellini
the longitudinal stresses developing in the rail due to interaction
et al. 2011; Rocha et al. 2012; Kwark et al. 2004; Liu et al. 2009;
between the rail and the superstructure, which is a phenomenon
Song et al. 2003; Galvin et al. 2018; Karoumi et al. 2005; Cheng
known as track–bridge interaction (TBI) or rail-structure interac-
et al. 2001; Guo et al. 2012; Sung and Chang 2019; Alijani et al.
tion (RSI). The sources that lead to ARSs are temperature
2022). This aspect primarily focuses on vibration analysis of the
1
Ph.D. Candidate, Dept. of Civil Engineering, Middle East Technical system and utilizes dynamic properties of the bridge, track, and
Univ., Ankara 06800, Turkiye (corresponding author). ORCID: https:// train. The second aspect, which has not been investigated as thor-
orcid.org/0000-0001-8535-0955. Email: [email protected] oughly as the first one in the literature, is the longitudinal response
2
Professor, Dept. of Civil Engineering, Middle East Technical Univ., of the track-bridge system. The longitudinal response of TBI is
Ankara 06800, Turkiye. Email: [email protected] commonly studied by considering temperature and moving loads
3
Professor, Dept. of Civil Engineering, Middle East Technical Univ., according to several international standards [EN 1991-2 (CEN
Ankara 06800, Turkiye. Email: [email protected]
2003); DIN Fb 101 (DIN 2003); TB10015 (Chinese National
Note. This manuscript was submitted on September 29, 2023; approved
on January 9, 2024; published online on April 11, 2024. Discussion period
Standards 2012); KR C-08080 (Korea Rail Network Authority
open until September 11, 2024; separate discussions must be submitted for 2014); UIC 774-3R (UIC 2001)]. In this case, the adopted approach
individual papers. This paper is part of the Practice Periodical on Struc- generally involves treating the problem statically by ignoring the
tural Design and Construction, © ASCE, ISSN 1084-0680. dynamic effects.

© ASCE 04024023-1 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 Discussions on the phenomena of TBI date back to the effort simply supported steel truss bridge utilizing both ballasted and bal-
driven by Stokes (1867) to solve differential equations related with lastless tracks. A parametric study was conducted by Ramos et al.
the braking of trains over railway bridges. Frýba (1996) solved the (2019) to determine a limiting bridge length by adjusting rail and
problem analytically based on the linear elastic relation between deck expansion joints on PSC girder bridges. Chen et al. (2013)
rail and bridge deck under thermal loads. For the longitudinal as- developed a numerical analysis tool using an existing material li-
pect of TBI, some of the recent studies focus on the difference be- brary to establish a common computer model for TBI on arch
tween the so called separate and complete analysis approaches. The bridges with CWRs. Recently, Kang et al. (2021) indicated that
separate (i.e., simplified) analysis approach, as described in design TBI rail stress limits specified in design codes were developed
codes, assumes superposition of forces resulting from the individ- based on theoretical models and track systems that are outdated
ual effects of thermal, bending, and braking/accelerating. The nowadays due to the rapid advancement in the field of high-speed
rationale behind this assumption is that linear superposition of rail transportation. They provided recommendations to increase the
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

the ARSs should always result in higher stresses. Such a linear current stress limits to allow for longer railway bridges. In addition,
superposition of various effects may produce inaccurate forces ARSs also develop as a result of earthquakes or other disasters.
when any of the components in the system exhibits a nonlinear Recently, change in ARSs due to earthquake loads on the tracks
behavior. In this case, the procedure specified in design codes for high-speed railway bridges was studied by several authors
may overpredict the actual ARSs and underpredict the rail/deck (Hu et al. 2022a, b). Zhang et al. (2020) proposed an equivalent
relative displacements. In the case of complete (i.e., sequential) modeling approach for analysis of multispan simply supported rail-
analysis approach, a separate load effect is considered at each way bridges under seismic excitations. The proposed model was
analysis step with the final condition in the previous load step taken reported to reduce the overall complexity of the model and hence
as the initial condition for the next step. The separate analysis ap- improve the computational efficiency. Wang et al. (2022) investi-
proach is more practical and common than sequential analysis gated the seismic response of a specific type of simply supported
among bridge designers in the light of international provisions, railway bridge system under near-fault ground motions with chang-
because it requires less effort and can be completed by many of ing fault distance, spatial location characteristics, and ground ac-
the available finite element (FE) software programs without a celeration. It was concluded that the regional ground motion
need to implement the loaded/unloaded states to the same model. analysis approach adopted in their study can be employed for se-
However, there might be cases where a separate analysis indicates lection of high-speed railway routes to improve line safety. Analy-
a safe design, whereas a sequential analysis indicates otherwise ses on railway bridges considering the interaction between train and
depending on the level of loading, geometric properties of the bridge have recently been conducted by several researchers (Peixer
bridge system, and the properties of the rail interface elements used et al. 2021; Kohl et al. 2023; Jin et al. 2022; Eroglu et al. 2023).
to model TBI. Peixer et al. (2021) investigated the effect of track irregularities on
Sanguino and Requejo (2009) showed the effect of the numeri- railway line safety by considering a continuous nine-span steel-
cal analysis method (i.e., separate versus sequential) on the results concrete composite bridge. Jin et al. (2022) studied the probability
of test cases given in the Appendix of UIC 774-3R (UIC 2001). of train derailment under the combined effect of seismic events and
Ruge and Birk (2007) were pioneers who investigated arbitrary track irregularities by utilizing a multispan simply supported bridge
sequence of loads using coupling interface based on the implemen- model. Eroglu et al. (2023) conducted dynamic analysis of a train-
tation of an exact formulation. Widarda (2009) studied the track– track-bridge system to investigate the effects of various parameters,
bridge interaction problem through dynamic analyses by utilizing such as train speed, bridge span length, and track stiffness. Kohl
appropriate dynamic properties of trains. Zhang et al. (2015) con- et al. (2023) considered many different vehicle models to study
ducted experiments on several types of fasteners used in urban rail the accuracy of the simplified approach (i.e., additional damping
transit and obtained RDCs to improve the accuracy of numerical approach) for vehicle-bridge interaction. Kang et al. (2018) pro-
TBI analyses. Huang et al. (2021) studied the effect of the mechani- vided a review of structural systems employed in high-speed rail-
cal behavior of the fasteners to seismic performance of urban rail way bridges in Germany and indicated that simply supported box
viaduct. Yan et al. (2012) analyzed a railway bridge consisting girder bridges are the most commonly used bridge type. The re-
of both simply supported and continuous spans and concluded that ported span lengths for these types of bridges are 25 m for small
the effect of the longitudinal stiffness of piers on TBI analysis re- overpasses, whereas longer spans of 44 and 58 m are typically used
sults are different for the simply supported and continuous spans. for valley crossings and large viaducts.
Yang and Jang (2016) proposed a practical solution using interface
elements that are adaptive to various loading cases based on their
experimental findings. Yun et al. (2019) monitored the last span Motivation for Present Study and Discussion of
of a 38-span simply supported prestressed concrete (PSC) girder Methodology
railway bridge with ballasted track to collect the temperature values
at different locations along the span, and to extract the unloaded The present study addresses the need for a detailed investigation of
RDCs. Furthermore, numerical analysis of the bridge system was the sensitivity of the TBI response to changes in RDCs with both
conducted for unloaded condition, and the numerical results agreed separate and sequential analysis approaches. Although RDCs are
with the measured response. Dai et al. (2020) provided an analyti- the essential variable that determines TBI in terms of separate and
cal algorithm for calculating longitudinal TBI forces especially for sequential analysis results, this aspect has not been studied in
long span bridges with rail expansion devices. Mirza et al. (2016) detail in the literature. Moreover, there is a need for a thorough
conducted an FE numerical investigation on the behavior of rails investigation of the ability of the RDCs specified in design
under thermal action after replacement of timber transoms with codes to represent the actual TBI response. The initial stiffness
fiber reinforced urethane cross beams in an existing railway. and resistance capacity of the track–bridge interface elements
Kašpárek et al. (2020) monitored a highly skewed arch bridge depend on many factors, such as the presence and type of ballast,
equipped with a special steering bar system to measure the resis- type of fastening system, type of sleepers, and the friction between
tance displacement behavior at track–bridge interface. Ryjáček and these components. With these multiple factors affecting the behav-
Vokáč (2014) conducted a long-term monitoring on a four-span ior of the track–bridge interface, there is a growing necessity for a

© ASCE 04024023-2 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Flowchart of analysis steps.

comprehensive investigation. Therefore, the primary interest in the located at the centroid of the composite section. These elements
present study is to fill these gaps in the literature. were connected to the nodes located at the top surface of the deck
The methodology adopted in this study involves a numerical and to the nodes located at girder supports with rigid beam ele-
investigation in an attempt to provide a better understanding of ments. Flexibility provided in the longitudinal direction due to
TBI for bridges incorporating CWRs. TBI analyses on several rail- girder support assembly and bridge pier was represented with a lin-
way bridges were conducted using a commercial FE software ear elastic spring. Rail elements were defined at the centroid of rails
MIDAS Civil (2012). Analytical verification of the numerical re- along the track. The geometric properties for the rail elements were
sults was performed by considering linear elastic RDCs for thermal determined based on the UIC60 EN 13674-1 (CEN 2002) type rail.
loads driven by Frýba (1996). Empirical data reported in previous Because the bridge cross section (i.e., deck and girders) and the
research by Yun et al. (2019) were used for further verification of rails were modeled with linear elements, only elastic material prop-
the FE modeling approach. Following the verification of the FE erties (i.e., modulus of elasticity and Poisson’s ratio) were included
model against analytical benchmark and real-world data, the model in the related material definitions.
was utilized to conduct a parametric study. The effort was focused The rail elements were connected to the bridge superstructure
on the combined effects of RDC type, expansion length, and sub- with bilinear springs. Multilinear elastic link elements available
structure stiffness on the TBI response in terms of ARSs and lon- in MIDAS Civil (2012) element library have been used for this
gitudinal displacements. A flowchart of the analysis steps followed purpose. As mentioned earlier, the interaction between rail and
in the study is given in Fig. 1. bridge superstructure in typical TBI analysis is represented with
resistance-displacement curves (RDCs). Several different RDCs
TBI Modeling that were reported based on laboratory and field measurements
and those specified by design standards were used in the present
study, as discussed in the following sections. The bilinear spring
Two-Dimensional Modeling Approach elements utilized in the numerical model were assigned initial stiff-
TBI modeling details adopted in the present study are depicted in ness values corresponding to the track resistance in the longitudinal
Figs. 2(a and b), respectively, for the cases of a single-span single- direction based on the RDC considered. In the transverse and ver-
track bridge and multi span double-track bridge. It is noted that tical directions, the rail elements were connected to the bridge
these figures are presented for the illustration of the general mod- superstructure through linear elastic springs with a relatively high
eling details, and the number of spans utilized in analyses varied, as stiffness. Spring stiffness in the vertical and transverse directions
mentioned in the subsequent sections. Deck and rails were modeled was taken to be 106 times the stiffness in the longitudinal direction,
using 2D beam elements considering both axial and flexural prop- based on the approach previously used by Petrangeli et al. (2008)
erties. Elements simulating the bridge deck and girders were and Liu et al. (2021). Behavior of the track–bridge interface springs

© ASCE 04024023-3 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Details of modeling for: (a) single-span single-track bridge; and (b) multispan double-track bridge.

in the longitudinal direction is a function of the vertical load A comparison of the separate and sequential analyses in terms of
available on the track. Thus, depending on whether there is a train rail resistance-displacement response is illustrated in Fig. 3. For the
loading (i.e., loaded case) or no train loading (i.e., unloaded case) case where the loaded track resistance is reached under a sequence
present on the track, different resistance-displacement behaviors of thermal, vertical and horizontal train loads, separate analysis in-
were assigned to the interface spring elements. dicates larger ARSs but smaller rail/deck relative displacements, as
A 300 m embankment length on both sides of the bridge was reported by Sanguino and Requejo (2009). Therefore, for bridges
considered for all studied cases. The track–bridge interface springs designed with rail stresses as the governing effect as opposed to
were typically provided at 1.0 m intervals. Utilizing a 1.0 m displacements, using separate analysis will be a conservative ap-
element length provides sufficiently accurate results both in terms
proach. However, it is necessary to check the displacement results
of ARSs and displacements, as reported previously by Yang and
carefully if separate analysis is to be used. For simply supported
Jang (2016). It is noted that the embankment length and element
test cases defined in appendix D of UIC 774-3R (UIC 2001),
length values utilized in the present study satisfy the corresponding
limits specified in point 1.7.3 of the UIC 774-3R Code (UIC 2001) Sanguino and Requejo (2009) reported higher rail/deck relative dis-
(i.e., a minimum embankment length of 100 m and a maximum placements, and absolute deck displacements in sequential analysis
element length of 2 m). as compared with separate analysis, indicating that linear superpo-
sition is not conservative in terms of displacements. The maximum
difference between separate and sequential analyses was reported
Difference between Separate and Sequential Analysis
to be 28% in these test cases. Therefore, in a sequential analysis
Approaches
special attention should be paid to the displacements when they
As specified in point 1.7.1 of UIC 774-3R (UIC 2001), the TBI are close to the code specified limits. The effect that the analysis
response can be obtained through either a separate analysis or a type (i.e., separate versus sequential) has on TBI response of rail-
complete analysis. The separate loading approach is simply based way bridges has been investigated earlier by several researchers
on the application of temperature, vertical, and braking/accelerating (Sanguino and Requejo 2009; Widarda 2009; Ruge and Birk 2007).
loads separately followed by the superposition of the corresponding The primary interest in the present study is to determine how sen-
results. Temperature loading is applied on a model utilizing un-
sitive the TBI response is to changes in RDCs with both separate
loaded resistance-displacement behavior for the interface elements,
and sequential analysis approaches. The initial stiffness and resis-
whereas the vertical and horizontal loads are applied on a separate
tance capacity of the track–bridge interface elements depend on
model utilizing loaded RDCs. The results from these two separate
models are then superimposed to obtain the total response. Because many factors, such as the presence and type of ballast, type of
this approach is based on the superposition of responses from fastening system, type of sleepers, and the friction between these
two separate analyses with RDCs exhibiting nonlinear behavior, components. Li et al. (2021) provided an insight on how to consider
the results are only approximate. The sequential analysis approach, the friction spring systems. With these multiple factors affecting the
on the other hand, utilizes a single model, where temperature behavior of the track–bridge interface, the ability of the RDCs
loads are applied first, followed by the application of vertical and specified in design codes to represent the actual response requires
braking/acceleration effects simultaneously with the interface ele- further investigation. The present study reports the findings of such
ments already deformed. an investigation.

© ASCE 04024023-4 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Track resistance-displacement responses for separate and sequential analyses. (Data from MIDAS Information Technology 2012.)

Longitudinal Track Resistance-Displacement type fastening system. Zhang et al. (2015) also conducted labora-
Curves (RDCs) tory testing on ballastless tracks with a low resistance type fastener
The numerical analyses conducted as part of the present investiga- that is commonly used for urban rail transit in China. As evident in
tion included longitudinal track resistance-displacement curves de- the plots, the RDC reported by Zhang et al. (2015) has slightly
picted in Figs. 4 and 5 for the unloaded and loaded track cases, lower initial stiffness and approximately 50% higher resistance
respectively. The resistance values for the RDCs are specified in limit than those reported by Yang and Jang (2016). It is noteworthy
terms of resistance per unit length along the track. The curves that the RDCs reported in both studies consistently fall far below
are either based on experimental results reported in the literature the RDC specified in KR-C08080 (Korea Rail Network Authority
or taken from the related design codes. For ballastless track, the 2014) and point 1.2.1.2 of UIC 774-3R (UIC 2001) Codes for un-
Korean Code KR-C08080 (Korea Rail Network Authority 2014) loaded ballastless tracks, in terms of initial stiffness and resistance
and point 1.2.1.2 of UIC 774-3R (UIC 2001) specify the same ini- limit. In an attempt to examine the effect of changing track resis-
tial stiffness and resistance limit values for both unloaded and tance limit with the same initial stiffness, a new RDC was proposed
loaded cases. In Fig. 4(a), the RDC reported by Yun et al. (2019) in this study by extending the RDC reported by Yang and Jang
based on field measurements from a recently constructed ballasted (2016) up to a resistance limit of 20 kN=m. This new RDC is
track lies between the UIC good and moderate maintenance RDCs. labelled as “fictive extension in Fig. 4(b).
The measurements were collected on a ballasted track with Pandrol The RDCs investigated as part of the present study for loaded
e-clip fasteners and UIC60 rails supported on prestressed concrete track condition are presented in Fig. 5. For ballastless tracks, the
sleepers. It is noted that the RDC reported by Yun et al. (2019) has RDCs reported by Zhang et al. (2015) and Yang and Jang (2016)
a yield displacement of 2.7 mm, as opposed to a 2.0 mm yield based on laboratory measurements are provided in Fig. 5(b)
displacement in RDCs specified in point 1.2.1.2 of UIC 774-3R together with the RDC specified in KR-C08080 (Korea Rail
(UIC 2001). Network Authority 2014) and point 1.2.1.2 of UIC 774-3R (UIC
Unloaded ballastless RDCs utilized in the numerical study are 2001) Codes. Even though the RDCs that are based on experi-
given in Fig. 4(b). The RDC reported by Yang and Jang (2016) is mental measurements have similar initial stiffness values, there
based on the laboratory tests on tracks with resilient base plate pad is approximately 3.3 times difference between the resistance limits.

Fig. 4. Longitudinal resistance-displacement curves considered in the current study for unloaded track condition: (a) ballasted track; and
(b) ballastless track.

© ASCE 04024023-5 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Longitudinal resistance-displacement curves considered in the current study for loaded track condition: (a) ballasted track; and (b) ballastless
track.

This might be attributed to the fact that the vertical loads applied on was assumed to have a single span with one fixed and one move-
the tracks to simulate the loaded track condition were different in able support.
two studies. Yang and Jang (2016) tested their track with 30 kN of It is usually assumed that rail temperature does not have any
vertical load on the fastening system, whereas 40 kN of vertical influence on the bridge because the cross-sectional area of the rail
load was used by Zhang et al. (2015). In addition to such a differ- compared with the bridge is significantly small (Strauss et al.
ence in the amount of vertical loading applied on tracks, different 2018), whereas a change in bridge temperature results in a change
fastening systems were utilized in these studies. This can be another in rail stresses. The following equation is valid for each part of the
reason for the difference in the measured resistance limits reported investigated bar according to Hooke’s Law
by Zhang et al. (2015) and Yang and Jang (2016). The RDC speci-
fied in KR-C08080 (Korea Rail Network Authority 2014) and point N ¼ EAðu 0 − αΔtÞ ð1Þ
1.2.1.2 of UIC 774-3R (UIC 2001) Codes has approximately three
times larger initial stiffness than the experimentally determined re- where N = axial force in the bar element; EA = axial rigidity of
lations with the resistance limit in between the measured values but the bar; u ¼ uðxÞ = longitudinal displacement of the bar element;
somehow close to the one reported by Zhang et al. (2015). u 0 = relative elongation (strain) of the bar; α = coefficient of
thermal extension of the bar; and Δt ¼ t-t0 = temperature change
with respect to the initial temperature t0 .
Verification of Numerical Modeling Approach For a bar element with length dx, the force equilibrium equation
can be written as follows:
Verification with Analytical Solution of a Single Span −N þ N þ N 0 dx − kudx ¼ 0 ð2Þ
Bridge under Temperature Loading
Frýba (1996) provided a closed form solution for the differential where k = spring constant; and N ¼ NðxÞ; dx and u = as defined
equation for TBI under temperature loading assuming a linear elas- earlier.
tic response for the track–bridge interface elements. A system of i Eq. (2) can be rewritten using the Hooke’s law given in Eq. (1)
bars was connected to each other to represent the bridge and
CWRs. The problem was divided into four bar segments, where −EAu 00 þ ku ¼ 0 ð3Þ
segments 1–3 represent the rails, whereas segment 4 represents the
bridge itself (Fig. 6). The interface elements were assumed to be where EA, k, and u = as defined earlier.
distributed along the bridge with linear spring constants (ki ) per Solving the differential equation results in the displacements of
unit length of the ith bar. For the investigated case, the bridge individual segments as follows:

-x 0 +x 0 +x

Fig. 6. Modeling of railway bridge used for verification with analytical solution under temperature loading.

© ASCE 04024023-6 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 Table 1. Material and geometrical properties used in verification of the ΔT ¼ ΔT 4 ¼ t4 − t40 ð6Þ
numerical model with analytical solution
Element Description Properties The corresponding bar forces are determined as follows:
 
Rail Cross-sectional area, A (m2 ) 0.1534 α ΔT
N 1 ðxÞ ¼ −EAαΔt 0
λ
Elasticity modulus, E (kN=m2 ) 2.1 × 108 ½1 − ð1 þ λl2 Þe−2 eλx
2λ
Thermal expansion coefficient, αi (/°C) 1.0 × 10−5  
Rail length on span, l2 (m) 60 α0 ΔT −λx −λ2−x
N 2 ðxÞ ¼ −EAαΔt ½2λx þ e − ð1 þ λl2 Þe 
Bridge Span length, l4 (m) 60 2λ
 
α ΔT
N 3 ðxÞ ¼ −EAαΔt 0
λ
Embankment Embankment length, l1 , l3 (m) 100 ðλl2 − 1 þ e−2 Þe−λx
2λ
Spring Spring constant per unit length, ki (kN=m2 ) 1 × 105
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

N 4 ðxÞ ¼ 0 ð7Þ

where N, EA, ΔT, α, α0 , λ, l2 , and x are as defined earlier. The

α ΔT stress σi for each bar is determined by dividing the axial force N i by
u1 ðxÞ ¼ 0
λ
½1 − ð1 þ λl2 Þe−2 eλx
2λ the rail area A.
α ΔT A numerical model of the bridge and track system for which the
u2 ðxÞ ¼ 0
λ
½2λx þ e−λx − ð1 þ λl2 Þe−2−x 
2λ analytical solution is presented above has been created and ana-
α ΔT lyzed under temperature loading following the modeling assump-
u3 ðxÞ ¼ 0
λ
ðλl2 − 1 þ e−2 Þe−λx tions as given in Fig. 2(a). Linear elastic springs were used for the
2λ
bridge-track interface elements in the numerical model, and the
u4 ðxÞ ¼ α0 ΔTx ð4Þ
bridge was assigned a temperature increase of 35°C. The geometric
where αi = coefficient of thermal extension of rails; and α0 ¼ α4 = and material properties used in the numerical model and the ana-
lytical solution are given in Table 1. Comparisons between the ana-
coefficient of thermal extension of the bridge; λ2 ¼ λ2i ¼ ðki =
lytical and numerical solutions are provided in Fig. 7 in terms of
Ei Ai Þ ¼ ðk=EAÞ assuming the same modulus of elasticity and areas
ARSs and longitudinal displacements. Displacement of the bridge
for segment 1–3; l2 = length of the rail segment over the bridge;
increases linearly from the fixed support to the movable support
x indicates the distance along the rail; and Δt is the difference in
[Fig. 7(c)], whereas the rail displacement has a nonlinear variation
temperature for rail bars from the initial temperature ti0
along the bridge [Fig. 7(b)]. As a result of such bridge and rail dis-
Δt ¼ Δti ¼ ti − ti0 ; for i ¼ 1 − 3 ð5Þ placement distributions, tensile and compressive stresses develop
within the rail with the distribution shown in Fig. 7(a). As evident,
where ΔT = difference in temperature for bridge bar from initial the maximum compressive rail stress occurs at the movable support
temperature t40 location, whereas the maximum tensile stress occurs at a location

Fig. 7. Comparison of numerical result with analytical solution: (a) rail stresses; (b) rail displacements; and (c) bridge displacements.

© ASCE 04024023-7 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 between the fixed support and midspan. The agreement between values along the final span of the bridge and the abutment. As for
the two sets of results in terms of displacements and rail stresses the relative displacement between the rail and the deck, the numeri-
indicates that the assumptions and the element size adopted in the cal results underestimate the measured values at the roller support
numerical model are acceptable for the investigated problem. of the last span.
The effect of the temperature drop within the bridge deck on the
Verification with Measured Response from a Multispan ARSs is shown in Fig. 9 at four different locations along the in-
Bridge with Unloaded Ballasted Track vestigated span of the bridge. For all four locations considered,
ARSs increase almost linearly with increasing temperature drop
Rail displacements, ARSs, and relative displacements between rail applied to the bridge deck. The rate at which ARSs change with
and bridge have previously been monitored by Yun et al. (2019) on temperature drop increases when moving from the midspan loca-
a ballasted track. The studied bridge has 38 simply supported spans tion toward the right support. As depicted in Figs. 8 and 9, the nu-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

with each span consisting of 25 m long precast/prestressed concrete merically predicted rail/deck relative displacements and ARSs
girders. The investigation by Yun et al. (2019) focused on the final agree well with the trend that is based on the measured values.
span of the bridge, which is the most critical span in terms of TBI The final span of the bridge reported by Yun et al. (2019) was
under thermal loading. As part of the present study, the last six further analyzed under a temperature decrease of 30°C by con-
spans of the same bridge, that had previously been monitored by sidering both the moderate maintenance and good maintenance bal-
Yun et al. (2019), were modeled based on the geometric and
lasted RDCs specified in point 1.2.1.2 of UIC 773-4 (UIC 2001), in
material properties given in the study. This is done to increase
addition to the RDC that is based on measurements. As mentioned
modeling efficiency because after six spans the analysis results
earlier, Yun et al. (2019) reported a resistance limit of 18 kN=m and
converged to the same rail/deck relative displacements and ARSs.
a corresponding longitudinal displacement of 2.7 mm based on
According to the measurements reported by Yun et al. (2019),
their measurements on the bridge. For ballasted tracks under un-
the rail-bridge attachment devices have a longitudinal resistance
loaded condition with good maintenance UIC 774-3R (UIC 2001)
limit of 18 kN=m and a corresponding longitudinal displacement
specifies 20 kN=m resistance limit and 2 mm longitudinal displace-
of 2.7 mm. Numerical analyses discussed in this section utilized the
RDC values that are based on the authors’ measurements. ment. The resistance limit drops to 12 kN=m for the case of mod-
Rail stresses and relative displacements between the bridge deck erate maintenance. The relation between these three RDCs is shown
and the rail are compared, respectively, in Figs. 8(a and b). The in Fig. 4(a). Results from the analyses performed by utilizing these
figures show the numerical results from the present study and three RDCs are presented in Fig. 10. The type of RDC does not
the measured values reported by Yun et al. (2019) for temperature have an appreciable effect on relative displacements as presented
drops of 10°C, 20°C, and 30°C in the bridge deck which were re- in Fig. 10(b). However, the corresponding rail stresses are affected
trieved from daily temperature change data collected on site. Due to to a certain degree. As evident, the level of ARSs developing as a
temperature decrease in the bridge deck, the bridge contracts in result of similar levels of rail/deck relative displacements depends
each span at the roller support location. As a result of this motion, on resistance and stiffness levels of the RDC used. The maximum
the relative displacement between the bridge deck and the rail difference in ARSs is 14% when the RDC proposed by UIC for
changes sign within each span. The amount of rail stress that de- good maintenance track is used instead of the RDC reported by
velops as a result of bridge temperature change depends directly on Yun et al. (2019). In this case, the UIC RDC produces more
the level of relative longitudinal displacement between the bridge conservative values. The RDC proposed by UIC for moderate
deck and the rail. A comparison of Figs. 8(a and b) indicates that maintenance track, on the other hand, resulted in 15% lower maxi-
relatively small stresses develop within the rail at locations where mum rail stress than the experimental RDC reported by Yun et al.
the rail/deck relative displacements are small (i.e., near midspan (2019). The results indicate that for this specific bridge with 25 m
location). Similarly, largest relative displacements occur at the right long simply supported concrete girders, utilizing the RDC that is
end (i.e., roller support) of the investigated span, and this results in specified in point 1.2.1.2 by UIC 774-3R (UIC 2001) for good
relatively large rail stresses at this location. The numerically deter- maintenance tracks is conservatively acceptable in the absence of
mined rail stresses provide an accurate estimation of the measured measured data from monitoring of the tracks.

Fig. 8. Variation of: (a) additional rail stress; and (b) bridge-rail relative displacement along span for the bridge studied by Yun et al. (2019).

© ASCE 04024023-8 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Variation of ARS for the bridge studied by Yun et al. (2019) with temperature change at: (a) 13 m; (b) 18 m; (c) 24 m; and (d) 25 m from left
end of the last span.

Fig. 10. Effect of RDC for the bridge studied by Yun et al. (2019): (a) ARS; and (b) rail-deck relative displacement.

Parametric Investigation Two tracks were considered on the bridge for vertical and hori-
zontal loadings, and the loading arrangements were adjusted such
To investigate the effect of major parameters on TBI, additional that maximum compressive ARS occurs at roller support location
numerical analyses were conducted on the precast concrete girder of the last span. Accordingly, positive temperature loading was
bridge reported by Yun et al. (2019) considering both ballasted and considered together with braking/acceleration force directions
ballastless track cases. Similar to the model used for the experimen- and train positions indicated in Fig. 11. As indicated, the train
tal verification study, the last six spans of this bridge were modeled on track-2 is 30 m into the bridge and creates an acceleration force
based on the geometric and material properties given by Yun et al. of 33 kN=m, whereas the front end of the train on track-1 is about
(2019). As part of the investigation, the RDCs given in Figs. 4 and 5 to leave the bridge with a braking force of 20 kN=m. A uniformly
were utilized with 35°C thermal loading in bridge deck, and vertical distributed load of 80 kN=m was used for vertical moving loading
and horizontal (i.e., braking and acceleration) loadings. For bal- based on the LM71 train model. For longitudinal loading, acceler-
lasted track condition, the UIC moderate maintenance RDC and ation force was taken as 33 kN=m for 30 m, and braking force was
the RDC reported by Yun et al. (2019) were not included in the taken as 20 kN=m for 300 m. The horizontal braking/acceleration
analyses, because they cannot be used in the loaded track case. forces and the vertical loading values employed in the analyses are

© ASCE 04024023-9 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Train loading positions considered in numerical models.

those specified in point 1.4.3 by UIC 774-3R (UIC 2001) and mostly remain within the initial linear portion of the correspond-
Annex G of EN 1991-2 (CEN 2003). ing RDC.
ARS behaviors with changing RDCs under braking and accel-
eration forces are presented in Figs. 12(c–e). Longitudinal forces
Effect of RDC on ARSs and Relative Displacements created on rails due to acceleration and braking of trains are trans-
ARS distributions under various loading effects are presented in ferred from rail to bridge deck based on the RDC stiffness. Plots
Fig. 12, both for the cases of ballasted and ballastless track. An in Figs. 12(c–e) indicate that the RDC with higher initial stiffness
increase in maximum compressive ARS is observed under temper- [i.e., the one specified by KR-C 08080 (Korea Rail Network
ature loading with increasing initial stiffness of RDC. Among all Authority 2014) and UIC 774-3R (UIC 2001)] results in smaller
unloaded RDCs considered, the one specified in point 1.2.2 by rail stress under braking and acceleration forces than the other
UIC 774-3R (UIC 2001) and KR-C08080 (Korea Rail Network RDCs. The reason is that this RDC develops the resistance capacity
Authority 2014) for ballastless track has the largest stiffness of under a relatively small relative displacement of 0.5 mm and results
80 kN=mm=m. On the other hand, the RDC that has the smallest in large slip between rail and bridge deck, which in turn causes
stiffness (i.e., 10 kN=mm=m) is the one specified in point 1.2.1.2 relatively small ARSs. Furthermore, the RDC with higher yield dis-
by UIC 774-3R (UIC 2001) for ballasted track with good mainte- placement capacity [i.e., the one suggested by Zhang et al. (2015)]
nance. These two RDCs, representing the largest and smallest creates higher rail stress than the other RDCs due to delayed yield-
stiffness, result in, respectively, 39.9 and 23.7 MPa additional com- ing. This behavior is observed for the loading cases of braking on
pressive stress in rails under a temperature increase of 35°C in the track-1 [Fig. 12(c)] and acceleration on track-2 [Fig. 12(d)], and the
bridge deck. The ratio of the maximum and minimum thermal ARS simultaneous application of these two loadings [Fig. 12(e)].
values is 1.68, which is much smaller than the stiffness ratio of the The total ARSs are determined as the summation of stresses due
two RDCs. The reason for the amount of change in ARSs being to temperature, vertical, and braking/acceleration loadings. Total
much smaller than the difference in the RDC stiffness is due to ARS distributions determined this way are shown in Fig. 12(f).
the track–bridge interface spring elements reaching the resistance Among all loadings, the largest portion of ARSs is caused by ther-
limit (i.e., yielding) under the effect of considered loading and mal loading. Another observation is that ARS in none of the cases
undergoing significant rail/deck relative displacements. Variation exceeds the compressive stress limit of 72 MPa defined in point
of rail/deck relative displacements along the span length under ther- 1.5.2 by UIC 774-3R (UIC 2001).
mal loading is illustrated in Figs. 13(a and b) for the two RDCs
specified in point 1.2.1.2 by UIC 774-3R (UIC 2001) for ballasted
Effect of Span Length on ARSs and Relative
track with good maintenance and ballastless track. The span length
Displacements
was divided into five equal segments, and the range of rail/deck
relative displacements within each segment is indicated on the As mentioned in the previous section, temperature loading contrib-
corresponding RDC curve. As evident, all track–bridge interface uted the most to the total ARS in the investigated bridge. The ther-
spring elements in segment-4 and segment-5 reached the resistance mal component of the ARS changes directly in relation with the
limit for both RDCs. In the case of ballastless track, segment-1 also expansion length. To investigate the effect of expansion length on
reached the resistance limit. This is due to the lower displacement ARS due to different loading components, the bridge reported by
resistance limit of 0.5 mm for ballastless track. For segment-2, Yun et al. (2019) was further investigated by changing the span
ballasted track spring elements remain in the linear elastic zone, length from 25 to 40 m with 5 m increments. The maximum span
whereas a portion of the elements reached the resistance limit in length was considered to be 40 m due to the practical span limit
the case of ballastless track. of precast girder bridges. The parameters regarding material and
ARS distribution under vertical moving loading is given in section properties were based on the following assumptions. The
Fig. 12(b). Results are provided only for track-2, because vertical cross-sectional dimensions of the bridge were increased with in-
loading on this track creates the most unfavorable condition due to creasing span length such that the ratio of vertical displacement
the neighboring span being only partially loaded (Fig. 11). As to span length under uniformly distributed load remains constant.
evident in Fig. 12(b), stiffness of track–bridge interface elements The cross-sectional area and total height of the bridge cross section
has a profound impact on the compressive ARSs for vertical mov- were increased linearly with the span length. The relative location
ing loads. On the other hand, the resistance value in loaded RDCs of the center of gravity of the bridge cross section was kept constant
does not have an appreciable influence on the resulting ARSs. This in relation to the section height. The parameters determined
is because under vertical moving loading the interface elements based on these assumptions are tabulated in Table 2. A modulus of

© ASCE 04024023-10 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. ARS distributions with ballastless track: (a) thermal loading; (b) vertical loading; (c) braking on track-1; (d) acceleration on track-2;
(e) braking on track-1 and acceleration on track-2; and (f) total.

elasticity value of 35.22 GPa was used for the bridge superstructure 56% of the total stress in 25 m span length, whereas it increases to
to represent the C40 grade concrete. 63% for 40 m span length, indicating that after a certain point rail
Maximum compressive ARS results are plotted in Fig. 14. As stresses due to braking/acceleration and vertical loadings are not
evident, the total rail compressive stress is dominated by thermal increasing as much as the thermal loading. Increasing the span
stresses. The rail stress due to temperature loading corresponds to length for simply supported bridges leads to an increase in the

© ASCE 04024023-11 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Range of track/deck relative displacements along bridge span with: (a) ballasted track RDC with good maintenance (data from UIC 774-3
2001); and (b) ballastless track RDC (data from UIC 774-3 2001; KR-C08080 2014).

expansion length, which affects thermal stresses the most. The boundary conditions. In addition to these idealized boundary con-
higher increase in the thermal stresses may also be attributed to an ditions, the bridge model was also analyzed with linear elastic
obvious increase of the cross-sectional area of the bridges as the springs provided at girder supports to represent the substructure
span length increases. stiffness. Two spring stiffness value of 2,500 mN=m (25 m span ×
Similar to the ARS, a significant portion of the total displace- 100 mN=m=m) and 250 mN=m (25 m span × 10 mN=m=m) were
ments is produced by thermal loading. Temperature loading con- utilized in these analyses. The bridge model was further analyzed
tributes 84%–86% of the total bridge top deck displacement, with bridge piers and elastomeric girder supports included in the
and 82%–86% of the total rail/deck relative displacement depend- model instead of the elastic springs simulating the substructure
ing on the span length. Braking/acceleration loading produces very stiffness. This type of modeling approach is believed to represent
small bridge top deck displacement and rail/deck relative displace- the actual substructure response more accurately. In these analyses,
ment with almost no change with the span length. Therefore, the two pier heights of 2 and 10 m were used together with a 2 m
impact of the span length on the relative displacements due to diameter circular pier cross section. The 2 and 10 m heights can
braking/acceleration and vertical loadings is insignificant. be considered to represent a short and medium length pier,
respectively.
Effect of Substructure Stiffness on ARSs and Relative
Displacements
Another design parameter that has an impact on TBI response is the
longitudinal stiffness present at bridge supports. Petrangeli et al.
(2008) stated that longitudinal stiffness of a short abutment/pier
roughly corresponds to 100 mN=m=m. Based on this substructure
stiffness value, additional analyses were conducted on the test
bridge reported by Yun et al. (2019) with different substructure
stiffness levels using the RDC specified in point 1.2.1.2 by UIC
774-3R (UIC 2001) for ballasted track. It is noted that the analyses
mentioned in previous sections utilized idealized roller and pin

Table 2. Parameters used for various span lengths

Moment of Depth of
Span length, inertia, Area, girder, Center of gravity
L (m) I (m4 ) A (m2 ) D (m) from bottom (m)
25 7.19 11.09 2.60 1.7
30 12.42 19.16 3.12 2.04
35 19.73 30.42 3.64 2.38 Fig. 14. Maximum compressive ARS distribution with different span
40 29.45 45.41 4.16 2.72 lengths.

© ASCE 04024023-12 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Effect of substructure stiffness ARS distribution: (a) thermal loading; (b) vertical loading; (c) braking on track-1; (d) acceleration on track-2;
(e) braking on track-1, acceleration on track-2; and (f) total.

ARS results from the bridge models mentioned above are plot- the smallest ARSs occurring with 10 m tall piers. Any increase in
ted in Figs. 15(a–f). Maximum ARSs under thermal loading vary bridge substructure stiffness provides a larger restraint at girder
from 18.9 to 23.4 MPa [Fig. 15(a)], with the largest ARSs occur- supports against longitudinal movement, which in turn, results in
ring with idealized roller and pin girder boundary conditions, and larger rail stresses under temperature loading. A similar result is

© ASCE 04024023-13 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 also valid for vertical loading, as shown in Fig. 15(b). This is due to Table 3. Analysis parameters used for comparison of separate and
the increased substructure stiffness resulting in higher top deck dis- sequential analysis results
placement, which leads to higher ARS. Span Support
The effect of substructure stiffness on rail stresses due to length (m) condition RDC
braking/acceleration loading is different than that of temperature 25 Pin-roller UIC 774-3 (2001) ballasted
and vertical loading. As braking/acceleration loading is applied di- 25 250 mN=m UIC 774-3 (2001) ballasted
rectly on rails, any increase in substructure stiffness results in 40 Pin-roller UIC 774-3 (2001) ballasted
smaller rail stresses. This is the reason for the model with 10 m 40 400 mN=m UIC 774-3 (2001) ballasted
tall piers (i.e., the model with smallest substructure stiffness) hav- 25 Pin-roller UIC 774-3 (2001) and
ing the largest maximum compressive ARS among all cases studied KRC 08080 (2014) ballastless
[Figs. 15(c–e)]. Under the combined effect of braking on track-1 25 250 mN=m UIC 774-3 (2001) and
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

and acceleration on track-2, maximum compressive ARSs vary KRC 08080 (2014) ballastless
from 12.1–35.4 MPa [Fig. 15(e)]. Such a variation indicates that 40 Pin-roller UIC 774-3 (2001) and
KRC 08080 (2014) ballastless
substructure stiffness has a significant influence on ARSs under
40 400 mN=m UIC 774-3 (2001) and
braking/acceleration loading. KRC 08080 (2014) ballastless
Variation of ARSs along the bridge span under the combined
effect of all loadings (i.e., thermal, vertical, and braking/accelera-
tion) with all studied substructure stiffness levels is presented in
Fig. 15(f). Maximum compressive rail stress occurs in models with compressive ARS is 68.5 and 62.8 MPa, respectively, for separate
10 m tall piers and 250 mN=m support stiffness. It is noteworthy and sequential analyses [Fig. 16(b)]. The reason for having 9%
that none of the ARSs exceeds the compressive stress limit of difference between these two ARS values can be explained by the
72 MPa defined in point 1.5.2 UIC 774-3R (UIC 2001). A com- yielding of the longitudinal RDCs. The RDCs exceed the 0.5 mm
parison of the stress plots in Fig. 15(f) indicates that the resistance limit with a support stiffness of 250 MN=m. As a result,
100 mN=m=m longitudinal stiffness as suggested by Petrangeli the displacements under vertical and longitudinal moving loads re-
et al. (2008) is a proper representation of stiffness level for short sult in higher rail/deck relative displacements, which in turn, leads
piers (i.e., 2 m pier height with 2 m diameter circular pier cross to additional force in the loaded track in the case of separate analy-
section). For the bridge model considered, the idealized pin and sis. A similar behavior is also valid for the span length of 40 m.
roller girder support conditions closely represent this substructure When this span length is considered together with a ballastless
longitudinal stiffness level. As evident in the plots, in the presence track and 400 MN=m support stiffness (i.e., equivalent to a stiff-
of medium length or slender piers, braking/acceleration loading be- ness of 10 MN=m per m of span length) the maximum ARS of
comes the dominant effect in terms of rail stresses. It is noted that 84.6 and 76.9 MPa is obtained, respectively from sequential and
using idealized pin and roller girder support conditions without separate analyses [Fig. 16(d)]. These values indicate an increase
modeling the support stiffness leads to unconservative rail stress of 10% in the maximum compressive ARS when changing the
estimates. This is primarily due to the effects caused by braking analysis method from separate to sequential. Even though, this
and acceleration forces getting larger when bridge piers are of much increase may seem insignificant, conducting a sequential
medium or high slenderness, as discussed previously by Wenner analysis can eliminate using rail expansion joints for threshold
bridges.
et al. (2019).
The results presented in Fig. 16 indicate that performing a sep-
arate analysis instead of a sequential analysis does not result in an
Comparison of Separate and Sequential Analysis appreciable difference in ARS values for ballasted tracks and for
Results ballastless tracks over bridges with relatively high substructure
stiffness. It is noteworthy that the maximum compressive ARSs
As illustrated in Fig. 3, separate and sequential analyses might lead presented in Fig. 16(b) remain below the compressive stress limit
to different track–bridge interaction responses. This difference may of 72 MPa defined in point 1.5.2 of UIC 774-3R (UIC 2001),
be especially critical in the cases where the ARSs are close to the whereas those in Fig. 16(d) exceed this limit, irrespective of
limits specified in the standards. To investigate the extent of differ- whether a sequential or a separate analysis is used.
ence between the responses obtained with separate and sequential
analysis approaches, the verified numerical model of the bridge
monitored by Yun et al. (2019) was further studied. Table 3 shows Conclusions
the analysis parameters used for this purpose. In these analyses,
thermal loads were defined as 35°C increase in the bridge deck, In this study, a numerical investigation was conducted to provide a
and the train loadings were applied on the numerical model as illus- better understanding of TBI for bridges with CWRs. The numerical
trated in Fig. 11. modeling approach was verified with available analytical solutions
The level of difference between separate and sequential analysis and data obtained from bridge monitoring. The primary interest
results was analyzed with different support conditions, span was to illustrate how sensitive the TBI response is to changes in
lengths, and RDC behaviors. Results of this investigation are RDCs with both separate and sequential analysis approaches. The
presented in Figs. 16(a–d) in terms of total ARSs. There is no ap- parametric study was extended to investigate the effects of expan-
preciable difference in ARS values from separate and sequential sion length and substructure stiffness, in addition to RDC type, on
analyses for the case of ballasted track, irrespective of the span the TBI response. The following conclusions are drawn based on
length and support condition. Among the cases investigated, the the numerical analysis results:
largest difference between ARS values from separate and sequen- 1. As the expansion length, (i.e., span length of a simply supported
tial analyses occurs when ballastless track is considered, and the bridge) increases, the contribution of thermal loading in total
substructure stiffness is included in the analysis model. With ARSs increases due to increased cross-sectional area of the
250 MN=m support stiffness (i.e., equivalent to a stiffness of bridge. The same conclusion cannot be derived for braking/
10 MN=m per m of span length) and 25 m span, the maximum acceleration and vertical loading, because these forces are

© ASCE 04024023-14 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Effect of separate versus sequential analysis on ARS distribution: (a) ballasted 25 m span; (b) ballastless 25 m span; (c) ballasted 40 m span;
and (d) ballastless 40 m span.

applied directly to the rail and to specific limited lengths accord- especially for ballastless systems to avoid exceeding the total
ing to the position of the train. ARS limits if the bridge is close to threshold in terms of TBI.
2. Resistance limit in RDCs is reached close to roller support 4. Substructure stiffness has a crucial effect on the ARSs especially
locations under thermal and braking/acceleration loading, under acceleration/braking loads. Increased substructure stiff-
whereas RDCs remain in initial linear range under vertical load- ness leads to higher ARSs due to thermal and vertical effects,
ing. The main reason is the direction of the loading. Relative whereas the ARSs due to braking/acceleration loads decrease.
displacements between track and bridge from vertical loads Considering the combined contribution of all load effects, lower
are not as significant as in the longitudinal loadings. In the case substructure stiffness results in higher ARSs in multispan sim-
of braking/acceleration loading, RDCs with higher initial stiff- ply supported bridges. Therefore, as the substructure becomes
ness and lower relative yield displacements tend to yield earlier, more slender, the TBI is more vulnerable to braking/acceleration
resulting in lower ARSs where higher longitudinal forces are forces in terms of ARSs. The ARS difference due to braking/
transferred from track to bridge. acceleration forces was 192% between substructure with lowest
3. Highest ARSs are obtained with UIC 774-3R (UIC 2001) bal- stiffness (10 m height of reinforced concrete piers with 2 m of
lastless track RDC. Utilizing the RDCs that were reported based diameter) and highest stiffness (idealized pin condition). As sub-
on field or laboratory measurements instead of those specified structure stiffness decreases, ARSs due to braking/acceleration
by design standards may lead to smaller ARSs. Therefore, forces become significantly high and start to dominate the to-
conducting physical testing on fastening systems to identify tal ARSs.
more realistic RDCs can be an approach for obtaining longer 5. Performing a separate analysis instead of a sequential analysis
expansion lengths, and hence more economical designs. In other does not result in an appreciable difference in ARS values for
words, according to the type of rail, sleepers, and fastening sys- ballasted tracks and for ballastless tracks over bridges with rel-
tem on tracks, specific RDCs can be used by bridge designers atively high substructure stiffness. For the case of ballastless

© ASCE 04024023-15 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 track with slender substructure, sequential analysis produces ap- and critical assessment.” Comput. Struct. 112–113 (Dec): 205–216.
proximately 10% smaller ARS compared with separate analysis. https://doi.org/10.1016/j.compstruc.2012.09.001.
In this study, short to medium span bridges with simply sup- Hu, Z., B. Wei, L. Jiang, S. Li, and H. Min. 2022a. “Track structural dam-
ported precast concrete girders were considered. Consequently, age index for high-speed railway girder bridges considering residual
the above conclusions were validated only for this type of bridge deformations due to earthquake.” Bull. Earthquake Eng. 20 (12):
6587–6609. https://doi.org/10.1007/s10518-022-01448-2.
and with the static moving load model utilized in analyses. One
Hu, Z., B. Wei, L. Jiang, S. Li, Y. Yu, and C. Xiao. 2022b. “Assessment of
particular aspect of precast concrete girder bridges is that as the
optimal ground motion intensity measure for high-speed railway girder
girders are prestressed almost exclusively, the concrete remains bridge (HRGB) based on spectral acceleration.” Eng. Struct. 252 (11):
in the uncracked state under service loading. This condition vali- 113728. https://doi.org/10.1016/j.engstruct.2021.113728.
dates the use of uncracked cross-section properties for girders in Huang, Y., J. Wang, W. Le, L. Zhang, and J. Su. 2021. “Study on mechani-
analysis. Future analyses may focus on bridges of other types, cal behaviours of rail fasteners and effects on seismic performance of
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

and continuous span bridges with longer span lengths and dynamic Urban Rail Viaduct.” Structures 33 (Oct): 3822–3834. https://doi.org
train loads. TBI response of horizontally curved bridges may also /10.1016/j.istruc.2021.06.105.
be addressed in future studies. Jin, Z., W. Liu, and S. Pei. 2022. “Probabilistic evaluation of railway ve-
hicle’s safety on bridges under random earthquake and track irregularity
excitations.” Eng. Struct. 266 (Sep): 114527. https://doi.org/10.1016/j
.engstruct.2022.114527.
Data Availability Statement Kang, C., S. Schneider, M. Wenner, and S. Marx. 2018. “Development of
design and construction of high-speed railway bridges in Germany.”
All data, models, or code that support the findings of this study are
Eng. Struct. 163 (May): 184–196. https://doi.org/10.1016/j.engstruct
available from the corresponding author upon reasonable request. .2018.02.059.
Kang, C., M. Wenner, and S. Marx. 2021. “Background investigation
on the permissible additional stresses due to track/bridge interaction.”
References Eng. Struct. 228 (Feb): 111505. https://doi.org/10.1016/j.engstruct
.2020.111505.
Alijani, A., M. Abadi, and S. Ghyasvand. 2022. “Numerical and experi- Karoumi, R., J. Wiberg, and A. Liljencrantz. 2005. “Monitoring traffic
mental analysis of intact and cracked railway bridges.” Structures loads and dynamic effects using an instrumented railway bridge.”
38 (2): 1018–1030. https://doi.org/10.1016/j.istruc.2022.02.040. Eng. Struct. 27 (12): 1813–1819. https://doi.org/10.1016/j.engstruct
CEN (European Committee for Standardization). 2002. Railway applications— .2005.04.022.
Track—Rail—Part 1: Vignole railway rails 46 kg=m and above. EN Kašpárek, J., P. Ryjáček, T. Rotter, M. Polák, and R. Calçada. 2020. “Long-
13674-1. Brussels, Belgium: CEN. term monitoring of the track–bridge interaction on an extremely skew
CEN (European Committee for Standardization). 2003. Eurocode 1: steel arch bridge.” J. Civ. Struct. Health Monit. 10 (Jul): 377–387.
Actions on structures. Traffic loads on bridges (Part 2). EN 1991-2. https://doi.org/10.1007/s13349-020-00389-1.
Brussels, Belgium: CEN. Kim, H. B., J. Lee, and D. H. Lee. 2010. “Extracting modal parameters
Chellini, G., L. Nardini, and W. Salvatore. 2011. “Dynamical identification of high-speed railway bridge using the TDD technique.” Mech. Syst.
and modelling of steel-concrete composite high-speed railway bridges.” Signal Process. 24 (3): 707–720. https://doi.org/10.1016/j.ymssp.2009
Struct. Infrastruct. Eng. 7 (11): 823–841. https://doi.org/10.1080 .11.010.
/15732470903017240. Kohl, A. M., K. D. Clement, J. Schneider, A. Firus, and G. Lombaert.
Chen, R., P. Wang, and X. Wei. 2013. “Track-bridge longitudinal interac- 2023. “An investigation of dynamic vehicle-bridge interaction effects
tion of continuous welded rails on arch bridge.” Math. Probl. Eng.
based on a comprehensive set of trains and bridges.” Eng. Struct.
2013 (Jan): 1–8. https://doi.org/10.1155/2013/494137.
279 (Mar): 115555. https://doi.org/10.1016/j.engstruct.2022.115555.
Cheng, Y. S., F. T. K. Au, and Y. K. Cheung. 2001. “Vibration of railway
Korea Rail Network Authority. 2014. Track-bridge longitudinal interaction
bridges under a moving train by using bridge-track-vehicle element.”
analysis (in Korean). KR C08080. Daejeon, Korea: Korea Rail Network
Eng. Struct. 23 (12): 1597–1606. https://doi.org/10.1016/S0141-0296
Authority.
(01)00058-X.
Kwark, J. W., E. S. Choi, Y. J. Kim, B. S. Kim, and S. I. Kim. 2004. “Dy-
Chinese National Standards. 2012. Code for design of railway continuous
namic behavior of two-span continuous concrete bridges under moving
welded rail. TB 10015-2012. Beijing: Chinese National Standards.
high-speed train.” Comput. Struct. 82 (Jun): 463–474. https://doi.org/10
Dai, G., G. Chen, R. Zheng, and Y. F. Chen. 2020. “A new bilinear resis-
.1016/S0045-7949(03)00054-3.
tance algorithm to analyze the track-bridge interaction on long-span
steel bridge under thermal action.” J. Bridge Eng. 25 (2): 04019138. Li, S., B. Wei, H. Tan, C. Li, and X. Zhao. 2021. “Equivalence of friction
https://doi.org/10.1061/(ASCE)BE.1943-5592.0001505. and viscous damping in a spring-friction system with concave friction
DIN (Deutsches Institut für Normung e.V). 2003. “DIN-Fachbericht 101.” distribution.” J. Test. Eval. 49 (1): 20190885. https://doi.org/10.1520
In Einkwirkungen auf bracken. Berlin: DIN. /JTE20190885.
Enshaeian, A., and P. Rizzo. 2021. “Stability of continuous welded rails: A Liu, K., E. Reynders, G. De Roeck, and G. Lombaert. 2009. “Experimental
state-of-the-art review of structural modeling and nondestructive evalu- and numerical analysis of a composite bridge for high-speed trains.”
ation.” Proc. Inst. Mech. Eng., Part F: J. Rail Rapid Transit 235 (10): J. Sound Vib. 320 (1–2): 201–220. https://doi.org/10.1016/j.jsv.2008
1291–1311. https://doi.org/10.1177/0954409720986661. .07.010.
Eroglu, M., M. A. Koc, I. Esen, and R. Kozan. 2023. “Realistic modelling Liu, W., H. Lai, G. Dai, S. Rao, D. Wang, and B. Wu. 2021. “Numerical
for analysis of train-structure and ballasted-track interaction for high- study on track–bridge interaction of integral railway rigid-frame bridge.”
speed trains.” J. Vib. Eng. Technol. 2023 (Jun): 1–33. https://doi.org/10 Appl. Sci. 11 (3): 922. https://doi.org/10.3390/app11030922.
.1007/s42417-023-01029-w. MIDAS Information Technology. 2012. Midas civil: Analysis reference
Frýba, L. 1996. Dynamics of railway bridges. London: Thomas Telford. manual. Seongnam-si, Korea: MIDAS Information Technology.
Galvín, P., A. Romero, E. Moliner, and M. D. Martínez-Rodrigo. 2018. Mirković, N. B., J. P. Popović, A. P. Pustovgar, L. M. Lazarević, and A. V.
“Two FE models to analyze the dynamic response of short span simply- Zhuravlev. 2018. “Management of stresses in the rails on railway
supported oblique high-speed railway bridges: Comparison and exper- bridges.” FME Trans. 46 (Jun): 636–643.
imental validation.” Eng. Struct. 167 (Aug): 48–64. https://doi.org/10 Mirza, O., S. Kaewunruen, C. Dinh, and E. Pervanic. 2016. “Numerical in-
.1016/j.engstruct.2018.03.052. vestigation into thermal load responses of railway transom bridge.” Eng.
Guo, W. W., H. Xia, G. D. Roeck, and K. Liu. 2012. “Integral model for Fail. Anal. 60 (Jun): 280–295. https://doi.org/10.1016/j.engfailanal.2015
train-track-bridge interaction on the Sesia viaduct: Dynamic simulation .11.054.

© ASCE 04024023-16 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023

 Peixer, M. A., P. A. Montenegro, H. Carvalho, D. Ribeiro, and T. N. Sung, D., and S. Chang. 2019. “Nonlinear behavior of rail fastening system
Bittencourt. 2021. “Running safety evaluation of a train moving over on slab track at railway bridge ends: FEA and experimental study.” Eng.
a high-speed railway viaduct under different track conditions.” Eng. Struct. 195 (Sep): 84–95. https://doi.org/10.1016/j.engstruct.2019.05
Fail. Anal. 121 (Mar): 105133. https://doi.org/10.1016/j.engfailanal .098.
.2020.105133. Tzepushelov, A., and L. Troyitzky. 1974. “Design, laying, maintenance and
Petrangeli, M., C. Tamagno, and P. Tortolini. 2008. “Numerical analysis of repair of continuously welded rails on the USSR railways.” Rail Int.
track: Structure interaction and time domain resonance.” J. Rail Rapid 1974 (Sep): 589–605.
Transit 222 (4): 345–353. https://doi.org/10.1243/09544097JRRT168. UIC (Union Internationale des Chemins de fer). 2001. “774-3R Track/
Ramos, Ó., F. Schanack, G. Carreras, and J. de Vena Retuerto. 2019. bridge interaction.” In Recommendations for calculations. Paris: UIC.
“Bridge length limits due to track-structure interaction in continuous Wang, Z., L. Jiang, L. Jiang, W. Zhou, and Y. Du. 2022. “Seismic response
girder prestressed concrete bridges.” Eng. Struct. 196 (Oct): 109310. of high-speed railway simple-supported girder track-bridge system con-
https://doi.org/10.1016/j.engstruct.2019.109310. sidering spatial effect at near-fault region.” Soil Dyn. Earthquake Eng.
Rocha, J. M., A. A. Henriques, and R. Calçada. 2012. “Safety assessment
Downloaded from ascelibrary.org by "Indian Institute of Technology, Patna" on 08/13/24. Copyright ASCE. For personal use only; all rights reserved.

158 (Jul): 107283. https://doi.org/10.1016/j.soildyn.2022.107283.

of a short span railway bridge for high-speed traffic using simulation Wenner, M., T. Meier, F. Wedel, G. Schacht, and S. Marx. 2019. “Exper-
techniques.” Eng. Struct. 40 (Jul): 141–154. https://doi.org/10.1016/j imental determination of the longitudinal pier stiffness of a long railway
.engstruct.2012.02.024. viaduct.” Front. Built Environ. 5 (Apr): 45. https://doi.org/10.3389
Ruge, P., and C. Birk. 2007. “Longitudinal forces in continuously welded /fbuil.2019.00045.
rails on bridge decks due to nonlinear track–bridge interaction.” Com-
Widarda, D. R. 2009. “Longitudinal forces in continuously welded rails due
put. Struct. 85 (7–8): 458–475. https://doi.org/10.1016/j.compstruc
to nonlinear track-bridge interaction for loading sequences.” Ph.D.
.2006.09.008.
thesis, Faculty of Civil Engineering, Dresden Technical Univ.
Ryjáček, P., and M. Vokáč. 2014. “Long-term monitoring of steel railway
Yan, B., G. Dai, and H. Zhang. 2012. “Beam-track interaction of high-
bridge interaction with continuous welded rail.” J. Constr. Steel Res.
speed railway bridge with ballast track.” J. Central South Univ.
99 (Aug): 176–186. https://doi.org/10.1016/j.jcsr.2014.04.009.
Sanguino, M. C., and P. G. Requejo. 2009. “Numerical methods for the 19 (5): 1447–1453. https://doi.org/10.1007/s11771-012-1161-8.
analysis of longitudinal interaction between track and structure.” In Yang, S. C., and S. Y. Jang. 2016. “Track–Bridge interaction analysis using
Track bridge interaction on high-speed railways, edited by R. Calçada, interface elements adaptive to various loading cases.” J. Bridge Eng. 21 (9):
R. Delgado, A. C. Matos, and J. M. Goicocela, 95–108. London: 04016056. https://doi.org/10.1061/(ASCE)BE.1943-5592.0000916.
Taylor & Francis. Yun, K.-M., H.-U. Bae, J. Moon, J.-J. Kim, J.-C. Park, and N.-H. Lim.
Song, M. K., H. C. Noh, and C. K. Choi. 2003. “A new three-dimensional 2019. “Quantification of ballasted track-bridge interaction behavior
finite element analysis model of high-speed train-bridge interactions.” due to the temperature variation through field measurements.” NDT &
Eng. Struct. 25 (13): 1611–1626. https://doi.org/10.1016/S0141-0296 E Int. 103 (Apr): 84–97. https://doi.org/10.1016/j.ndteint.2019.01.009.
(03)00133-0. Zhang, J.-L., D. J. Wu, and Q. Li. 2015. “Loading-history-based track–
Stokes, G. G. 1867. “Discussion of a differential equation related to the bridge interaction analysis with experimental fastener resistance.”
breaking of railway bridges.” Trans. Cambridge Philos. Soc. 8 (5): Eng. Struct. 83 (Jan): 62–73. https://doi.org/10.1016/j.engstruct.2014
707–735. https://doi.org/10.1016/S0141-0296(03)00133-0. .11.002.
Strauss, A., S. Karimi, M. Šomodíková, D. Lehký, D. Novák, D. M. Zhang, Y., L. Jiang, W. Zhou, Y. Feng, Z. Tan, and X. Chai. 2020. “Study of
Frangopol, and K. Bergmeister. 2018. “Monitoring based nonlinear bridge-subgrade longitudinal constraint range for high-speed railway
system modeling of bridge–continuous welded rail interaction.” Eng. simply-supported beam bridge with CRTSII ballastless track under
Struct. 155 (Jan): 25–35. https://doi.org/10.1016/j.engstruct.2017.10 earthquake excitation.” Constr. Build. Mater. 241 (Apr): 118026.
.053. https://doi.org/10.1016/j.conbuildmat.2020.118026.

© ASCE 04024023-17 Pract. Period. Struct. Des. Constr.

Pract. Period. Struct. Des. Constr., 2024, 29(3): 04024023