5.

International Conference on Earthquake Engineering and Seismology ( 5ICEES )

8-11 OCTOBER 2019, METU ANKARA TURKEY

PROBABILISTIC SEISMIC ASSESSMENT OF RAILWAY BRIDGES IN

TURKEY
1 2 1
Ozakgul K. , Yilmaz M.F. , Caglayan B.O.
1
Assoc. Prof. Dr., Civil Engineering Department, Istanbul Technical University, Istanbul, Turkey
2
Res. Asst. Dr., Civil Engineering Department, Ondokuzmayıs University, Samsun, Turkey
Email: mehmetfatih.yilmaz@omu.edu.tr

ABSTRACT:

In this study, the results of the probabilistic seismic assessment of the bridges still in use in Turkish railway lines
are presented. The bridges were classified and selected considering their construction materials, structural systems,
span lengths, and ages. 3D finite element models of the selected bridges were generated based on the original
design drawings. To obtain the probabilistic seismic demand models (PSDMs) of each bridge, nonlinear time
history analyses using 3D finite element models were carried out under 60 different earthquake data with three
components. The earthquake data were selected considering different moment magnitude, peak ground
acceleration (PGA) and central distance. Probabilistic seismic demand models (PSDMs) were used to determine
relations between engineering demand parameters (EDP) and intensity measure (IM). IM parameter needs to have
a good correlation with EDP and well characterize earthquake hazard. PGA which is a characteristic parameter
obtained directly from earthquake records as the IM parameter was used. The greatest lateral displacements of the
bridge spans for different service velocities defined in the Eurocode were considered as a serviceability damage
state i.e. as EDP. Finally, fragility curves of the bridges were derived considering maximum damage probability
for the IM level and using maximum likely hood method. Thus, for the first time, railway bridges in Turkey were
investigated in terms of probabilistic seismic assessment.

KEYWORDS: Railway bridges, Engineering demand parameter, Intensity measure, Probabilistic seismic
assessment, Fragility curve,

1. INTRODUCTION

After earthquakes, transportation systems need to be sustained servicing to allow transport the basic needs.
Railway lines constituted an important part of the transportation systems. As there are many historical bridges in
Turkish railway lines, seismic assessment of these bridges needs to be done to reduce seismic losses. Fragility
curve is one of the effective tools used to determine the seismic performance of a bridge. There are mainly two
approaches to derive the fragility curve: analytical methods and empirical methods. To derive empirical fragility
curve past earthquake reports or experimental studies and surveys are required and it usually is not possible for
many bridges. Therefore, analytical fragility curves become more important. To derive analytical fragility curve,
linear or nonlinear dynamic analyses are used (Shinozuka et al. 2000a and 2000b).
The probabilistic seismic assessment of bridges has been studied by many researchers. However, there are few
studies related to railway bridges among them. Choi et al. (2004) developed fragility curves considering various
damage states of bearings and columns for four typical highway bridge types in the United States. Yi et al. (2007)
developed seismic fragility curves based on the return period using a probability density function (PDF)
interpolation technique. Pan et al. (2007) presented the results of seismic fragility analysis of a typical steel
highway bridge in New York State considering the uncertainties related to earthquake characteristics and the
bridge structural properties such as the estimation of material strength, bridge mass, the friction coefficient of
expansion bearings and expansion-joint gap size. Huang et al. (2010) proposed a methodology to obtain
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

probabilistic seismic demand models for reinforced concrete highway bridges with one single-column bent
considering structural and statistical uncertainties and model errors. Park and Choi (2011) developed seismic
fragility curves of track-on steel-plate-girder bridges without ballast in conventional Korean railway lines using
probabilistic seismic demand model. Tavares et al. (2013) evaluated the seismic vulnerability of a three spans
continuous highway bridge in Quebec using fragility analyses based on probabilistic seismic demand model
defined for the abutments, bearings, and columns by nonlinear time history analyses of three-dimensional FE
model. Huo and Zhang (2013) used a fragility function method to investigate the effects of pounding and skewness
on the seismic behaviors of typical multi-span reinforced concrete highway bridges.
Zakeri et al. (2014) investigated the effect of skew angle on bridge seismic fragility for skewed single-frame
concrete box-girder bridges with single or two-column bents, integral or seat-type abutments. Parool and Rai
(2015) obtained fragility curves of multi-span simply supported bridges with drop spans and steel bearings based
on the incremental dynamic analysis performed using 3D nonlinear finite element models of bridge and steel
bearings. Jeon et al. (2015) investigated the effect of vertical ground motions on the seismic fragility of typical
older reinforced concrete highway bridges under horizontal and vertical ground motions. Yang et al. (2015)
developed analytical fragility curves for six highway bridge types namely multi-span simply supported (MSSS)
concrete bridges, multi-span continuous (MSC) concrete bridges, MSSS steel bridges, MSC steel bridges, single-
span (SS) concrete bridges, and SS steel bridges using 3D nonlinear analytical models considering nonlinear
behavior of the columns, bearings, and abutments and investigated the effect of skew angle on the seismic response
of bridges using fragility parameters in the fragility curves considering the effect of skew. Jeon et al. (2016)
investigated the effect of geometric parameters such as horizontal curvature, abutment skew and column height on
the seismic fragility of curved multi-frame concrete box girder highway bridges with in-span hinges. Stefanidou
et al. (2017) investigated effects of soil-structure interaction on seismic fragility of reinforced concrete highway
bridges, considering the vulnerability of critical components, namely piers, bearings and abutments of bridge and
soil-foundation-pier and abutment-embankment interactions. Martinez et al. (2017) determined fragility curves of
typical non-skewed highway bridges in Chile using incremental dynamic analysis.
Wei et al. (2018) investigated the effect of vertical ground motions on the seismic vulnerabilities of a continuous
concrete box girder bridge widely used in Chinese high-speed railway and concluded that the seismic damage
probabilities of bridge and track components increase with the increase of vertical ground motion. Mangalathu et
al. (2019) calculated seismic vulnerability of skewed concrete box-girder highway bridges in California subjected
to near-field and far-field ground motions using fragility curves of bridge components and system based on for the
material, geometric, and structural uncertainties. Abbasi and Moustafa (2019) developed component and overall
system fragility curves of older and newly-designed straight and skewed multi-frame reinforced concrete box-
girder bridges in California considering four different damage states: slight, moderate, extensive, and complete,
and five skew angles (0–60°). Kabir et al. (2019) derived fragility curves of a three-span continuous reinforced
concrete highway bridge using nonlinear incremental dynamic analysis under near-fault, far-field and long-
duration ground motions. Barbieri DM (2019) assessed the seismic vulnerability of a railway masonry arch bridge
with a double-track railway line using non-linear static analysis and a probabilistic approach.

2. PROBABILISTIC SEISMIC DEMAND MODEL

Probabilistic seismic demand model (PSDM) is derived to determine relations between engineering demand
parameters (EDPs) and intensity measures (IMs) or used to describe seismic demand of a structure in terms of
intensity measure, as follows,

ln(d ) − ln( EDP)

P[ EDP  d IM ] = 1 −  ( ) (1)
 EDP IM
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

The median EDP can be estimated by a power model described with Equation 2 or a linear logarithm model is
given in Equation 3.

EDP = aIM b (2)

ln( EDP) = ln(a) + b ln( IM ) (3)

where IM is the seismic intensity measure, a and b are regression coefficients,  is the standard normal cumulative
distribution function, EDP is the median value of engineering demand, d is the limit state used to assess the
damage level and  EDP IM (dispersion) is the conditional standard deviation of the regression as given in Equation
4.

 EDP IM 
 (ln(d ) − ln(aIM
i
b
)) 2
(4)
N −2

Equation 3 is used to determine the PSDM of the bridge based on the results of the nonlinear time history analysis.
To derive PSDMs, the linear or nonlinear analysis need to be done. The nonlinear time history analysis gives more
realistic results. There are three methods to derive PSDM depend on nonlinear time history analysis, cloud,
incremental dynamic analysis and stripe method (Mackie and Stojadinovic, 2005). In this study, the cloud method
was used. The Cloud method includes results of nonlinear time history analyses achieved using a group of
earthquake records without scaling. As the results of the nonlinear analysis are depending on selected
earthquake records, the earthquake record domain has an important effect on the PSDMs.

3. SELECTION OF THE EARTHQUAKE RECORDS

Selection of an earthquake record is one of the important steps to derive analytical fragility curve. Characteristic
properties of selected earthquake record constitute an important uncertainty in seismic demand (Kwon and
Elnashai, 2006). It is aimed to represent different earthquake hazards in the selection of the earthquake records.
One of the most important parameters affecting the characteristics of the earthquake is the soil type. In this study,
totally 60 different earthquake records were selected considering different soil types, moment magnitude, PGA
and central distance. The moment magnitudes are varying between 4.9 and 7.4 and PGAs are changing from 0.01g
to 0.82g and the central distance of earthquake records are ranging from 2.5 km to 217.4 km. Distribution of
moment magnitude to PGA and central distance is shown in Figure 1. As one of the most important parameters
affecting the characteristics of the earthquake is soil type, the maximum accelerations were classified between 0.1s
and 0.3s for soil type A, 0.15s and 0.5s for soil type B and 0.1s and 0.9s for soil type C. The selected earthquake
records were used for time history analysis without scaling.

4. ANALYTICAL FRAGILITY CURVES OF THE BRIDGES

In past earthquakes, ground-motion-induced various damages of bridges were reported in literature as shear failure
of piers, unseating of bridge spans due to excessive relative movement of spans, loss of supports due to liquefaction
and excessive lateral movements, embankment failures, track damages such as broken rails and joints, buckled
tracks, parapet and spandrel wall failures due to outward movement of piers and abutments parapets and spandrel
walls, bearing and anchor bolt damages, derailments and overturning collapse of locomotives and cars due to
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

settlement of tracks and ground shaking (Byers 2004, Kawashima 2012). In this study, as lateral displacement of
bridges is a destructive case for the most important vulnerability, lateral displacements of the bridge spans were
considered as a serviceability damage state. The lateral displacement limits are given in EN1990-Annex A2 (2001)
for different service velocities in railway bridges were used (see Table 1). The horizontal deflections of the bridge
deck are limited by EN 1990-Annex A2 (2001) to sustain traffic safety of the railway line. To endure traffic safety,
fragility curve of the bridges can be derived depending on horizontal displacement limits. UIC2000 train loads
applied for masonry and steel girder bridge. these bridges have very limited deformation under self-weight. Only
self-weight is considered for steel truss bridge and important lateral deformations are visualized. Fragility curves
of the bridge were derived considering maximum damage probability for all IM level and using probabilistic
seismic demand model. Serviceability limit states were considered as slight damage (Pitilakis et al. 2014).

Earthquake Data Distribution

8
7,5
7
6,5
Mw

6
5,5 Earthquake
5 Data
4,5
4
0 50 100 150 200 250
Central Distance (kM)
Figure 1. Moment magnitude and central distance distribution of the earthquake records

Table 1. Maximum angular variation and minimum radius of curvature (EN1990-ANNEX A2, 2001).
Speed range (km/h) Rotation Curvature
(rad) (1/m)
V≤120 0.0035 1700
120<V≤200 0.0020 6000
V>200 0.0016 14000

The existing railway bridges were classified according to service time, structural system, span length, structural
material, traffic load, train speed and location within the scope of the TUBITAK 1001 project. Then, from each
classification, 37 bridges were selected for investigation as seen in Figure 2. In this paper, seven bridges among
them were presented.

Figure 2. Bridges investigated within the scope of the TUBITAK 1001 project
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

4.1 Km 235+451 bridge in Haydarpaşa-Eskişehir railway line

The bridge is located in Haydarpaşa-Eskişehir railway line at 235+451 km and built at 1930 by Flender
Aktiengesellschaft für Eisenbrücken und Shiffbau, Benrath–RH Düseldorf. There are 7 steel truss spans on the
bridge. The multi-span truss girder bridge is supported by abutments at the edges and 6 piers at the middle spans.
One edge of steel truss span is simply supported, and the other edge is designed as sliding support on rollers and
length of a span is 25.7m while the total length of the bridge is 187 m. The bridge has a 300 m curve radius. Steel
girders on spans are composed of angle section, IPN, UPN hot-rolled sections, steel plates, and built-up sections
(see Figures 3 and 4).

Figure 3. General view of the km 235+451 railway bridge

Figure 4. 3D finite element model of the km 235+451 railway bridge

Probabilities of exceeding of serviceability limit states are shown in Figure 5.

Figure 5. Fragility curves of the km 235+451 railway bridge

4.2 Km 284+400 bridge in Malatya-Yolçatı railway line

The bridge located between the cities of Malatya and Elazig and crossing Karakaya Dam Lake consists of 29
simply-supported steel trusses, each having a 64m span length, giving a total bridge length of 2029.60 m (see
Figures 6 and 7).
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

Figure 6. General view of the km 284+400 railway bridge

Figure 7. 3D finite element model of the km 284+400 railway bridge

Probabilities of exceeding of serviceability limit states are shown in Figure 8.

Figure 8. Fragility curves of the km 284+400 railway bridge

4.3 Km 199+325 bridge in Afyon-Manisa railway line

The bridge located in the Afyon-Manisa railway line was built by Ateliers De Construction De Jambes Namur
Company at 1923. The bridge is composed of 6 spans simply supported steel truss bridge with 30m length each
and has 5 truss piers which are 12.5 m, 19m, 26m, 33m and 40 m in height and 300m horizontal curve radius (see
Figures 9 and 10).

Figure 9. General view of the km 199+325 railway bridge

5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

Figure 10. 3D finite element model of the km 199+325 railway bridge

Probabilities of exceeding of serviceability limit states are shown in Figure 11.

Figure 11. Fragility curves of the km 199+325 railway bridge

4.4 Km 146+070 bridge in Narlı-Malatya railway line

This existing bridge is a reinforced concrete arch railway bridge with a total length of 285 m having seven spans
of 35 m, a height of 34 m. The bridge located about 105 km south of Malatya city in Narlı-Malatya railway line
was constructed by Nydqvist & Holm A.B.; J. Saabye & O. Lerche; Kampmann, Kierulff & Saxild in 1928 (see
Figures 12 and 13).

Figure 12. General view of the km 146+070 railway bridge

5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

Figure 13. 3D finite element model of the km 146+070 railway bridge

Probabilities of exceeding of serviceability limit states are shown in Figure 14.

Figure 14. Fragility curves of the km 146+070 railway bridge

4.5 Km 307+222 bridge in Konya-Fevzipaşa-Hudut railway line

The bridge is a plain concrete arch railway bridge with a total length of 215 m having three major spans
of 30 m, a height of 69 m and 19 ‰ slope. The bridge located about 60 km northwest of Adana city on
Konya Adana railway line and having a single-track railway line was designed and built by German
design contractor Philipp Holzmann AG in 1916 (see Figures 15 and 16).

Figure 15. General view of the km 307+222 railway bridge

5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

Figure 16. 3D finite element model of the km 307+222 railway bridge

Probabilities of exceeding of serviceability limit states are shown in Figure 17.

Figure 17. Fragility curves of the km 307+222 railway bridge

5. CONCLUSIONS

This paper presents the results of the probabilistic seismic assessment of the bridges selected considering their
construction materials, structural systems, span lengths, and ages. 3D finite element models of the bridges were
generated based on the original design drawings using SAP2000 software. To obtain the probabilistic seismic
demand models (PSDMs) of each bridge, nonlinear time history analyses using 3D finite element models were
carried out under 60 different earthquake data with three components. The earthquake data were selected
considering different moment magnitude, peak ground acceleration (PGA) and central distance. Finally, fragility
curves of the bridges were derived considering maximum damage probability for the IM level. Thus, for the first
time, all railway bridges in Turkey were investigated in terms of probabilistic seismic assessment. %50 probability
of acceding of limit states for three velocity limit states are determined for each bridge. The results show that the
decrease in speed range increases the safety of the bridge. Derived fragility curve allows engineers and
governments to fast capacity analysis of the investigated bridge.

6. ACKNOWLEDGMENTS

This research presented in this paper was supported by TCDD and TUBITAK 114M332 project. Any opinions
expressed in this paper are those of authors and do not reflect the opinions of the supporting agencies.
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

REFERENCES

Abbasi M, Moustafa MA (2019). Probabilistic seismic assessment of as-built and retrofitted old and newly designed skewed
multi-frame bridges. Soil Dynamics and Earthquake Engineering 119: 170-86.

Barbieri DM (2019).Two methodological approaches to assess the seismic vulnerability of masonry bridges. Journal of
Traffic and Transportation Engineering 6(1): 49-64.

Byers WG (2004). Railroad lifeline damage in earthquakes. Proceedings of the 13th World Conference on Earthquake
Engineering, 1-6 August, Vancouver, B.C., Canada.

Choi E, DesRoches R, Nielson B (2004). Seismic fragility of typical bridges in moderate seismic zones. Engineering
Structures 26(2): 187-99.

EN1990-ANNEX A2 (2001). Application for bridges, Eurocode: Basis of Structural Design, European Committee for
Standardization, Brussels, Belgium.

Huang Q, Gardoni P, Hurlebaus S (2010). Probabilistic seismic demand models and fragility estimates for reinforced concrete
highway bridges with one single-column bent. ASCE Journal of Engineering Mechanics 136(11): 1340-53.

Huo Y, Zhang J (2013). Effects of pounding and skewness on seismic responses of typical multispan highway bridges using
the fragility function method. ASCE Journal of Bridge Engineering 18(6): 499-515.

Jeon JS, Shafieezadeh A, Lee DH, Choi E, DesRoches R (2015). Damage assessment of older highway bridges subjected to
three-dimensional ground motions: Characterization of shear–axial force interaction on seismic fragilities. Engineering
Structures 87: 47-57.

Jeon JS, DesRoches R, Kim T, Choi E (2016). Geometric parameters affecting seismic fragilities of curved multi-frame
concrete box-girder bridges with integral abutments. Engineering Structures 122: 121-43.

Kabir MR, Billah AHMM, Alam MS (2019). Seismic fragility assessment of a multi-span RC bridge in Bangladesh
considering near-fault, far-field and long duration ground motions. Structures 19: 333-48.

Kawashima K (2012). Damage of bridges due to the 2011 Great East Japan earthquake. Journal of Japan Association for
Earthquake Engineering 12(4): 319-38.

Kwon OS, Elnashai A (2006). The effect of material and ground motion uncertainty on the seismic vulnerability curves of
RC structure. Engineering Structures, 28: 289-303.

Mackie KR, Stojadinovic B (2005). Comparison of incremental dynamic, cloud and stripe methods for computing
probabilistic seismic demand models. Proceedings of the ASCE Structures Congress 2005, 20-24 April, New York, USA.
Mangalathu S, Jeon JS, Jiang J (2019). Skew adjustment factors for fragilities of California box-girder bridges subjected to
near-fault and far-field ground motions. ASCE Journal of Bridge Engineering 24(1): 1-13.

Martinez A, Hube MA, Rollins KM (2017). Analytical fragility curves for non-skewed highway bridges in Chile.
Engineering Structures 141: 530-42.

Pan Y, Agrawa AK, Ghosn M (2007). Seismic fragility of continuous steel highway bridges in New York State. ASCE
Journal of Bridge Engineering 12(6): 689-99.

Park J, Choi E (2011). Fragility analysis of track-on steel-plate-girder railway bridges in Korea. Engineering Structures 33:
696-705.
5. International Conference on Earthquake Engineering and Seismology ( 5ICEES )
8-11 OCTOBER 2019, METU ANKARA TURKEY

Parool N, Rai DC (2015). Seismic fragility of multispan simply supported bridge with drop spans and steel bearings. ASCE
Journal of Bridge Engineering 20(12): 1-11.

Pitilakis K, Crowley H, Kaynia AM (2014). SYNER-G: Typology Definition and Fragility Functions for Physical Elements
at Seismic Risk, Geotechnical, Geological and Earthquake Engineering, Springer, Dordrecht, Netherlands.

Shinozuka M, Feng MQ, Lee J, Naganuma T (2000a). Statistical analysis of fragility curves. ASCE Journal of Engineering
Mechanics 126(12): 1224–31.

Shinozuka M, Feng MQ, Kim HK, Kim SH (2000b). Nonlinear static procedure for fragility curve development. ASCE
Journal of Engineering Mechanics 126(12): 1287-95.

Stefanidou SP, Sextos AG, Kotsoglou AN, Lesgidis N, Kappos AJ (2017). Soil-structure interaction effects in analysis of
seismic fragility of bridges using an intensity-based ground motion selection procedure. Engineering Structures 151: 366-
380.

Tavares, DH, Suescun JR, Paultre P, Padgett JE (2013). Seismic fragility of a highway bridge in Quebec. ASCE Journal of
Bridge Engineering 18(11): 1131-39.

Wei B, Zuo C, He X, Jiang L, Wang T (2018).Effects of vertical ground motions on seismic vulnerabilities of a continuous
track-bridge system of high-speed railway. Soil Dynamics and Earthquake Engineering 115: 281-90.

Yang CSW, Werner SD, DesRoches R (2015). Seismic fragility analysis of skewed bridges in the central southeastern
United States. Engineering Structures 83: 116-28.

Yi JH, Kim SH, Kushiyama S (2007). PDF interpolation technique for seismic fragility analysis of bridges. Engineering
Structures 29: 1312-22.

Zakeri B, Padgett JE, Amiri GG (2014). Fragility analysis of skewed single-frame concrete box-girder bridges. ASCE
Journal of Performance of Constructed Facilities 28(3): 571-82.