University of Nevada, Reno

Numerical Modeling Techniques of High-Speed Rail Bridge Structures

A dissertation submitted in partial fulfillment of the

requirements for the degree of Master of Science
in

Civil and Environmental Engineering

by

Andrew Stephenson

Dr. Mohamed A. Moustafa/ Thesis Advisor

August 2020
 THE GRADUATE SCHOOL

We recommend that the thesis

prepared under our supervision by

Andrew Stephenson

Hntitled
Numerical Modeling Techniques of High-Speed Rail Bridge
Structures

be accepted in partial fulfillment of the

requirements for the degree of

MASTER OF SCIENCE

Mohamed A. Moustafa, Ph.D., P.E.

$GYLVRU
Elnaz Ezmailzadeh Seylabi, Ph.D.
CoPPLWWHH0HPEHU

Hanif Livani, Ph.D.

*UDGXDWH6FKRRO5HSUHVHQWDWLYH

'DYLG:=HK3K''HDQ
*UDGXDWH6FKRRO

August, 2020
 i

ABSTRACT

High-speed rail (HSR) is a complex system incorporating various technical aspects such as
infrastructure, rolling stock (specially-designed train sets), telecommunications, operating
conditions, and equipment. The highly sophisticated technology combining these elements,
as well as the elements themselves continue to evolve as the new transportation mode
continues to expand and its intrinsic characteristics pose design issues unique to HSR
systems. With the requirements for deflections, rotations, and natural frequencies of HSR
bridge structures, comprehensive understanding of the HSR dynamic interactions is a topic
of growing interest. Accordingly, many studies over the past few decades have been
conducted, mostly internationally, with a focus on dynamic interaction between the
different components of HSR train/bridge systems through sophisticated structural models.
The focus of this research is to identify these modeling features and inherent characteristics
of HSR bridges, and to provide guidance and demonstration examples on how to develop
such models in OpenSees. Such models will aid researchers and designers in conducting
parametric studies to test the static, modal, and dynamic performance of future HSR bridge
designs to formulate a national standard for HSR infrastructure in the United States.

The main objective of this study was to create a comprehensive modeling guideline for
HSR bridge systems. To do so, a thorough literature review was conducted to synthesize
various methods of numerical modeling techniques used to model HSR systems. Literature
published from national and international sources were reviewed and compiled to
demonstrate how the individual components within a train system, track system, and bridge
system have been modeled in previous studies. The synthesis also identified the similarities
and differences regarding the different finite element modeling techniques for different
components. Based on the studies analyzed in the literature search, a prototype train system
and track-bridge system were selected to construct a fully detailed example HSR bridge
model. The prototypes were selected based on available information regarding the design
of the prototype components to minimize assumptions necessary to model the prototype
system. A step-by-step guide of the processes of formulating the model and analysis
 ii

parameters from start to finish were documented, accompanied by snapshots from a sample
OpenSees model input file for guidance and future use.

To exemplify potential use of the developed model for informing future designs using
OpenSees data output, sample static and dynamic analyses were performed with load cases
without train loading and with train loading on the prototype HSR bridge. Additionally, a
brief analytical study was performed to demonstrate the HSR bridge seismic performance
using three different ground motions. The ground motions were retrieved from the PEER
Ground Motion Database and were amplified to various degrees to perform nonlinear time
history analysis. The nonlinear analysis considered four load cases for unloaded bridge and
the bridge with a train on top in three sample load cases to observe the sensitivity of seismic
analysis based on the addition and location of train loading. From the preliminary analysis
results of the prototype HSR bridge modeled as a demonstration, the location of the train
loading did not show significant influence on the local and global response of the bridge.
At larger scale of ground motions, the bridge showed instances of higher nonlinearity with
load cases with train loading which suggest that the train-bridge interaction better be
considered when informing and optimizing future HSR bride designs in high-seismic areas.
 iii

ACKNOWLEDGMENTS

This study was funded by the Accelerated Bridge Construction University Transportation
Center (ABC-UTC) 2016 Grant led by Florida International University. This funding and
research opportunity associated with it is acknowledged.
A special thank you to Dr. Mohamed A. Moustafa for offering his knowledge, support,
mentorship, and friendship which I will cherish for the rest of my career. It was a great
pleasure to be able to work alongside him.
I am extremely thankful for my close friends and family who have given me endless
encouragement to successfully complete this dissertation. A sincere thank you to a friend
who I am truly fortunate and grateful to have been able to spend the entirety of my time as
a student at this university with.
 iv

TABLE OF CONTENTS

Abstract ................................................................................................................................ i
Acknowledgments.............................................................................................................. iii
Table of Contents ............................................................................................................... iv
List of Tables .................................................................................................................... vii
List of Figures .................................................................................................................. viii
1 Introduction ................................................................................................................. 1
Background and Motivation ................................................................................. 1

Problem Description ............................................................................................. 2

Research Objectives and Scope of Work ............................................................. 3

Organization of the Thesis ................................................................................... 4

2 Synthesis of National and International Studies on the Topic of Numerical Modeling

of High-Speed Rail Systems ............................................................................................... 6
Overview .............................................................................................................. 6

Modeling of Train Systems .................................................................................. 6

Traditional Vehicle System........................................................................... 6

Articulated Vehicle System .......................................................................... 8

Modeling of Railway Track Systems ................................................................... 9

Rail ................................................................................................................ 9

Ballasted Track System................................................................................. 9

Ballastless Track System ............................................................................ 10

Modeling of Bridge Systems .............................................................................. 12

Deck and Girder .......................................................................................... 12

Pier Column ................................................................................................ 13

Pier Column Foundation ............................................................................. 14

Isolation Bearing ......................................................................................... 15

 v

General Modeling Procedures ............................................................................ 15

Rigid Connection Arm ................................................................................ 15

Viscous Damping ........................................................................................ 16

3 HSR Bridge System Numerical Model: Selection of Prototype System and Modeling
Procedure .......................................................................................................................... 27
Introduction ........................................................................................................ 27

Selection of Prototype HSR System .................................................................. 27

Train System Prototype............................................................................... 27

Track and Bridge System Prototype ........................................................... 28

Numerical Model in OpenSees .......................................................................... 28

Basic Model Definitions ............................................................................. 29

Train System Model .................................................................................... 31

Track System Model ................................................................................... 36

Bridge System Model ................................................................................. 39

4 Demonstration of Gravity, Modal, and Seismic Analysis of HSR Bridge System ... 54
Overview ............................................................................................................ 54

Gravity Load Analysis ....................................................................................... 54

Gravity Load Analysis Setup ...................................................................... 54

Gravity Load Analysis Results ................................................................... 56

Modal Analysis .................................................................................................. 58

Seismic Load Analysis ....................................................................................... 59

Seismic Load Analysis Set Up .................................................................... 59

Seismic Load Analysis Results ................................................................... 62

5 Seismic Response of Prototype HSR Bridge System: More In-Depth Demonstration

88
 vi

Overview ............................................................................................................ 88

Maximum Response Tables ............................................................................... 89

Seismic Behavioral Graphs ................................................................................ 91

6 Summary and Conclusions ...................................................................................... 128

Summary .......................................................................................................... 128

Conclusions ...................................................................................................... 130

Research Impact ............................................................................................... 131

Validities and Limitations ................................................................................ 132

7 References ............................................................................................................... 135

APPENDIX A: OpenSees Commands ........................................................................... 140
APPENDIX B: Selected Scripts from OpenSees Input File ........................................... 147
 vii

LIST OF TABLES

Table 3-1. Prototype HSR Model Element and Material. ................................................. 45

Table 3-2. Dynamic Characteristics of Train Model. ....................................................... 46
Table 3-3. Masses for Track-Bridge System. ................................................................... 47
Table 3-4. Section parameters of elastic beam elements in track-bridge system as adopted
from Li et al., [22]. ............................................................................................................ 47
Table 4-1. Example HSR bridge system load cases based on the train position above the
bridge (the cases represent instances of the train crossing the bridge). ............................ 66
Table 4-2. Column Base Reactions (kN) in Direction 3 from Static Analysis. ................ 66
Table 4-3. Periods for first 10 modes................................................................................ 67
Table 5-1. Maximum Local Responses – Northridge 100% Scale. .................................. 97
Table 5-2. Maximum Local Responses – Northridge 200% Scale. .................................. 98
Table 5-3. Maximum Local Responses – Kobe 100% Scale. ........................................... 99
Table 5-4. Maximum Local Responses – Kobe 200% Scale. ......................................... 100
Table 5-5. Maximum Local Responses – Loma Prieta 100% Scale. .............................. 101
Table 5-6. Maximum Local Responses – Loma Prieta 200% Scale. .............................. 102
Table 5-7. Maximum Global Responses - Northridge 100% Scale. ............................... 103
Table 5-8. Maximum Global Responses - Northridge 200% Scale. ............................... 104
Table 5-9. Maximum Global Responses - Kobe 100% Scale. ........................................ 105
Table 5-10. Maximum Global Responses - Kobe 200% Scale. ...................................... 106
Table 5-11. Maximum Global Responses – Loma Prieta 100% Scale. .......................... 107
Table 5-12. Maximum Global Responses – Loma Prieta 200% Scale. .......................... 108
 viii

LIST OF FIGURES

Figure 2-1. China-star high-speed train [41]..................................................................... 18

Figure 2-2. Views of the KHST (a) panoramic view, (b) articulated bogie located between
the car bodies, (c) articulated bogie and (d) composition of the train (front power car) [19].
........................................................................................................................................... 18
Figure 2-3. Front view of the sprung-mass dynamic car model [29]................................ 18
Figure 2-4. Tradition train system modeled by He et al., [13].......................................... 19
Figure 2-5. Traditional train system modeled by Liu et al., [24]. ..................................... 19
Figure 2-6. Articulated train system modeled by Kwark et al., [19]. ............................... 19
Figure 2-7. Bogie–bridge interaction system in an articulated train system modeled by Song
et al., [36]. ......................................................................................................................... 20
Figure 2-8. Photo of ballasted track system [33]. ............................................................. 20
Figure 2-9. Ballasted track system modeled by Song et al., [36]. .................................... 21
Figure 2-10. Ballasted track system modeled by Montenegro et al., [29]. ....................... 21
Figure 2-11. Ballasted track system modeled by Guo et al., [11]. .................................... 21
Figure 2-12. Photo of ballastless track system [39]. ......................................................... 22
Figure 2-13. Japanese type RCRS slab track on grade [38]. ............................................ 22
Figure 2-14. Track system scheme with fasteners (a) and longitudinal boundary spring
hysteresis loop (b) by Li and Conte, [23]. ........................................................................ 22
Figure 2-15. Modeling schematic of ballastless track system modeled by Li et al., [22]. 23
Figure 2-16. Modeling schematic of track-bridge system by Montenegro et al., [29]. .... 23
Figure 2-17. Modeling schematic of track-bridge system by Li and Conte, [23]. ............ 24
Figure 2-18. Modeling schematic of bridge system by Li et al., [22]............................... 24
Figure 2-19. Modeling schematic of bridge system by He et al., [13]. ............................ 25
Figure 2-20. Concrete box girder modeled using shell elements by Song et al., [36]. ..... 25
Figure 2-21. Modeling schematic of bridge pier columns using fiber-based elements by
Kaviani et al., [17]. ........................................................................................................... 25
Figure 2-22. Pile foundation model using dynamic p-y approach: (a) schematic view of the
FE model, (b) pile cap mode [23]. .................................................................................... 26
Figure 2-23. Modeling schematic of rigid connections by Kaviani et al., [17]. ............... 26
 ix

Figure 3-1. Photo of KTX-Sancheon [6]. ......................................................................... 48

Figure 3-2. Schematic of the prototype bridge: a) Elevation layout of high-speed railway
bridge/cm, b) Schematic sketch of track and girder structure [22]. .................................. 48
Figure 3-3. Schematic of the prototype bridge typical cross-section of track and girder
structure [22]. .................................................................................................................... 49
Figure 3-4. Schematic drawing for the numerical modeling of train system (Top: Cross-
section, Bot: Elevation). .................................................................................................... 49
Figure 3-5. Schematic of track system. ............................................................................. 50
Figure 3-6. Schematic of track-bridge system. ................................................................. 50
Figure 3-7. Parameters of zero-length connection elements in the track-bridge system as
adopted from Li et al., [22]. .............................................................................................. 50
Figure 3-8. Force-deformation behavior of track system connection layers: (a) Fastener,
(b) CA mortar, (c) Shear reinforcement, (d) Sliding layer, and (e) Lateral blocking. ...... 51
Figure 3-9. Schematic of track-bridge system. ................................................................. 52
Figure 3-10. Force-deformation behavior of bridge bearings: (a) Fixed bearing, (b) Sliding
bearing............................................................................................................................... 53
Figure 3-11. Finite element model of bridge. ................................................................... 53
Figure 4-1. Train load cases used for Chapter 4. .............................................................. 68
Figure 4-2. Vertical bridge girder displacements under static analysis for both load cases.
........................................................................................................................................... 69
Figure 4-3. Vertical bridge girder displacements under Load Case 1. ............................. 70
Figure 4-4. Vertical bridge girder displacements under for Load Case 8. ........................ 70
Figure 4-5. Transverse bridge girder displacements under Load Case 1. ......................... 71
Figure 4-6. Transverse bridge girder displacements under Load Case 8. ......................... 71
Figure 4-7. Displacement time-history of column #2 under Load Case 1 in:................... 72
Figure 4-8. Displacement time-history of column #6 under Load Case 1 in:................... 73
Figure 4-9. Displacement time-history of column #8 under Load Case 1 in:................... 74
Figure 4-10. Displacement time-history of column #11 under Load Case 1 in:............... 75
Figure 4-11. Displacement time-history of column #2 under Load Case 8 in:................. 76
Figure 4-12. Displacement time-history of column #6 under Load Case 8 in:................. 77
 x

Figure 4-13. Displacement time-history of column #8 under Load Case 8 in:................. 78

Figure 4-14. Displacement time-history of column #11 under Load Case 8 in:............... 79
Figure 4-15. Force-displacement relationship of column #2, #6, #8, and #11 in the
longitudinal direction for: (a) Load Case 1, (b) Load Case 8. .......................................... 80
Figure 4-16. Force-displacement relationship of column #2, #6, #8, and #11 in the
transverse direction for: (a) Load Case 1, (b) Load Case 8. ............................................. 81
Figure 4-17. Bridge girder shear in the longitudinal direction (Vx) for Load Case 1. ..... 82
Figure 4-18. Bridge girder shear in the transverse direction (Vy) for Load Case 1. ........ 82
Figure 4-19. Bridge girder shear in the vertical direction (Vz) for Load Case 1. ............. 83
Figure 4-20. Bridge girder moment in the longitudinal direction (Mx) for Load Case 1. 83
Figure 4-21. Bridge girder moment in the transverse direction (My) for Load Case 1. ... 84
Figure 4-22. Bridge girder moment in the vertical direction (Mz) for Load Case 1. ....... 84
Figure 4-23. Bridge girder shear in the longitudinal direction (Vx) for Load Case 8. ..... 85
Figure 4-24. Bridge girder shear in the transverse direction (Vy) for Load Case 8. ........ 85
Figure 4-25. Bridge girder shear in the vertical direction (Vz) for Load Case 8. ............. 86
Figure 4-26. Bridge girder moment in the longitudinal direction (Mx) for Load Case 8. 86
Figure 4-27. Bridge girder moment in the transverse direction (My) for Load Case 8. ... 87
Figure 4-28. Bridge girder moment in the vertical direction (Mz) for Load Case 8. ....... 87
Figure 5-1. Train load cases used in the seismic analysis in Chapter 5. ......................... 109
Figure 5-2. PEER database ground motions used for the seismic performance assessment:
(a) Northridge, (b) Kobe, and (c) Loma Prieta. .............................................................. 110
Figure 5-3. Longitudinal displacement time-history for columns #3, #6, and #11 at 100%
– (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load
Case 6, and (3) Load Case 9). ......................................................................................... 111
Figure 5-4. Longitudinal displacement time-history for columns #3, #6, and #11 at 200%
– (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load
Case 6, and (3) Load Case 9). ......................................................................................... 112
Figure 5-5. Transverse displacement time-history for columns #3, #6, and #11 at 100% –
(Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load
Case 6, and (3) Load Case 9). ......................................................................................... 113
 xi

Figure 5-6. Transverse displacement time-history for columns #3, #6, and #11 at 200% –
(Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load
Case 6, and (3) Load Case 9). ......................................................................................... 114
Figure 5-7. Longitudinal force-displacement relationship for columns #6, #8, and #11 at
100% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 115
Figure 5-8. Longitudinal force-displacement relationship for columns #6, #8, and #11 at
200% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 116
Figure 5-9. Transverse force-displacement relationship for columns #6, #8, and #11 at
100% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 117
Figure 5-10. Transverse force-displacement relationship for columns #6, #8, and #11 at
200% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 118
Figure 5-11. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at
100% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 119
Figure 5-12. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at
200% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 120
Figure 5-13. Transverse moment-curvature relationship for columns #1, #6, and #10 at
100% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 121
Figure 5-14. Transverse moment-curvature relationship for columns #1, #6, and #10 at
200% – (Row: (1) Northridge, (2) Kobe, and (3) Loma Prieta, Column: (1) Load Case 1,
(2) Load Case 6, and (3) Load Case 9). .......................................................................... 122
Figure 5-15. Longitudinal displacement time-history for columns #3, #6, and #11 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 123
 xii

Figure 5-16. Transverse displacement time-history for columns #3, #6, and #11 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 123
Figure 5-17. Longitudinal force-displacement relationship for columns #6, #8, and #11 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 124
Figure 5-18. Transverse force-displacement relationship for columns #6, #8, and #11 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 124
Figure 5-19. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 125
Figure 5-20. Transverse moment-curvature relationship for columns #1, #6, and #10 at
Northridge 300% – (Left: Load Case 1, Middle: Load Case 6, Right: Load Case 9)..... 125
Figure 5-21. Force-deformation relationship of fasteners supporting rail 1 under Northridge
300%: (a) Above column #4, (b) Above column #6. ...................................................... 126
Figure 5-22. Force-deformation relationship of fasteners supporting rail 2 under Northridge
300%: (a) Above column #4, (b) Above column #6. ...................................................... 126
Figure 5-23. Force-deformation relationship of CA mortar layers supporting track 1 under
Northridge 300%: (a) Above column #4, (b) Above column #6. ................................... 127
Figure 5-24. Force-deformation relationship of sliding layers supporting track 1 under
Northridge 300%: (a) Above column #4, (b) Above column #6. ................................... 127
Figure A-1. model command parameters [32]. ............................................................... 140
Figure A-2. node command parameters [32]. ................................................................. 140
Figure A-3. fix constraint command parameters [32]. .................................................... 140
Figure A-4. equalDOF constraint command parameters [32]. ....................................... 140
Figure A-5. geomTransf Linear transformation command parameters [32]................... 141
Figure A-6. Steel01 material command parameters [32]. ............................................... 141
Figure A-7. Steel02 material command parameters [32]. ............................................... 141
Figure A-8. Concrete02 material command parameters [32]. ........................................ 141
Figure A-9. ViscousDamper material command parameters [32]. ................................. 142
Figure A-10. Elastic material command parameters [32]. .............................................. 142
Figure A-11. elasticBeamColumn element command parameters [32]. ......................... 142
Figure A-12. dispBeamColumn element command parameters [32]. ............................. 142
 xiii

Figure A-13. zeroLength element command parameters [32]. ....................................... 143

Figure A-14. twoNodeLink element command parameters [32]. .................................... 143
Figure A-15. section fiber command parameters [32]. ................................................... 143
Figure A-16. patch rect command parameters [32]. ....................................................... 144
Figure A-17. layer straight command parameters [32]. ................................................. 144
Figure A-18. section aggregator command parameters [32].......................................... 144
Figure A-19. mass command parameters [32]. ............................................................... 145
Figure A-20. eigen analysis command parameters [32]. ................................................ 145
Figure A-21. Rayleigh damping command parameters [32]. .......................................... 145
Figure A-22. timeSeries path command parameters [32]. .............................................. 145
Figure A-23. UniformExcitation pattern command parameters [32]. ............................. 146
Figure B-1. Predefined geometric locations for train nodes. .......................................... 147
Figure B-2. Node set up for rear power car. ................................................................... 148
Figure B-3. Node set up for rear intermediate passenger car. ........................................ 148
Figure B-4. Node set up for first intermediate passenger car. ........................................ 149
Figure B-5. Rigid elastic beam-column element for bogie arms in the x-direction. ...... 149
Figure B-6. Rigid elastic beam-column element for bogie arms in the y-direction. ...... 149
Figure B-7. Rigid elastic beam-column element for primary suspension arms in the .... 149
Figure B-8. Rigid elastic beam-column element for primary suspension arms in the .... 149
Figure B-9. Rigid elastic beam-column element for car-bodies. .................................... 149
Figure B-10. Primary suspension system model for the power cars............................... 150
Figure B-11. Power car primary suspension node MP-constraints with equalDOF. ..... 150
Figure B-12. Secondary suspension system model for the power cars........................... 151
Figure B-13. Power car secondary suspension node MP-constraints with equalDOF. .. 151
Figure B-14. Mass assignment for train car-bodies. ....................................................... 152
Figure B-15. Mass assignment for train bogies. ............................................................. 152
Figure B-16. Mass assignment for power and exterior passenger car axle wheels. ....... 153
Figure B-17. Mass assignment for intermediate passenger car axle wheels .................. 153
Figure B-18. Node set up for rail 1 of track 1. ................................................................ 154
Figure B-19. Node set up for base plate of track 1. ........................................................ 154
 xiv

Figure B-20. Node set up for track plate of track 1. ....................................................... 154
Figure B-21. Elastic beam-column element for rail 3 of track 2. ................................... 154
Figure B-22. Elastic beam-column element for track plates of track 1. ......................... 154
Figure B-23. Elastic beam-column element for base plates of track 1. .......................... 154
Figure B-24. Zero-length element for fastener ............................................................... 155
Figure B-25. Zero-length element for lateral blocking. .................................................. 155
Figure B-26. Zero-length element for CA layer. ............................................................ 155
Figure B-27. Zero-length element for sliding layer. ....................................................... 155
Figure B-28. Zero-length element for shear reinforcement. ........................................... 156
Figure B-29. CA layer node MP-constraints with equalDOF. ....................................... 156
Figure B-30. Mass assignment for first two rail 1 nodes. ............................................... 156
Figure B-31. Node set up for the first bridge girder span. .............................................. 156
Figure B-32. Example elastic beam-column elements for bridge girder. ....................... 157
Figure B-33. Node set up for bearings supporting the first span of the bridge............... 157
Figure B-34. Zero-length elements for fixed bearings supporting the first span of the bridge.
......................................................................................................................................... 157
Figure B-35. Zero-length elements for sliding bearings supporting the first span of the
bridge. ............................................................................................................................. 158
Figure B-36. Bearing node MP-constraints with equalDOF. ......................................... 158
Figure B-37. Material properties for pier columns. ........................................................ 158
Figure B-38. Section designer for pier cross-section. ..................................................... 159
Figure B-39. Predefined geometric values for pier columns. ......................................... 159
Figure B-40. Node set up for first two columns. ............................................................ 160
Figure B-41. Displacement-based fiber-section beam-column elements for first pier
column............................................................................................................................. 160
Figure B-42. Node set up for column footings and ground. ........................................... 160
Figure B-43. Column footings and ground node SP-constraints using fix. .................... 161
Figure B-44. Zero-length element for bridge-soil interaction. ....................................... 161
Figure B-45. Rigid elastic beam-column element for footings of columns #1 and #2. .. 162
 xv

Figure B-46. Rigid elastic beam-column element for column-bearing connections at

column #1........................................................................................................................ 162
Figure B-47. Rigid elastic beam-column element for girder-bearing connections above
......................................................................................................................................... 162
Figure B-48. Rigid elastic beam-column element for first two girder-track system
connections. .................................................................................................................... 162
Figure B-49. Mass assignment for first two nodes of bridge girder. .............................. 162
Figure B-50. Mass assignment for first two nodes of column #1. .................................. 162
Figure B-51. Mass assignment for footings of column #1 and #2. ................................. 163
Figure B-52. Dead loads for train car-bodies and bogies. .............................................. 163
Figure B-53. Dead loads for power and extreme passenger car axle-wheels. ................ 164
Figure B-54. Dead loads for intermediate passenger car axle-wheels. ........................... 165
Figure B-55. Dead loads for rail 1 (first four nodes). ..................................................... 165
Figure B-56. Dead loads for track plate for track 1 (first four nodes). ........................... 165
Figure B-57. Dead loads for base plate for track 1 (first four nodes). ............................ 165
Figure B-58. Dead loads for first bridge girder span. ..................................................... 166
Figure B-59. Dead loads for first pier column. ............................................................... 166
Figure B-60. Dead loads for foundations. ....................................................................... 166
Figure B-61. Definition of gravity load analysis parameters. ......................................... 167
Figure B-62. Performance of gravity load analysis. ....................................................... 167
Figure B-63. Set up for modal analysis. ......................................................................... 167
Figure B-64. Set up for Rayleigh damping. .................................................................... 168
Figure B-65. Definition of seismic load analysis parameters. ........................................ 168
Figure B-66. Definition of ground motion parameters. .................................................. 168
Figure B-67. Application of ground motion in both directions. ..................................... 168
Figure B-68. Performance of seismic load analysis........................................................ 169
 1

1 INTRODUCTION

Background and Motivation

A transportation solution that has always been considered for the past few decades is the
high-speed rail (HSR). The successful commercial operation of the Japanese Shinkansen,
(bullet train) in 1964 marked the beginning of a new era for HSR and the development of
HSR spread throughout the world. Plans for HSR in the United States date back to the
High-Speed Ground Transportation Act of 1965 (Public Law 89-220, 79 Stat. 893) which
was the first attempt by the U.S. Congress to foster the growth of HSR. Although the United
States was one of the world’s first countries to have a high-speed rail service in place with
the Metroliner operating between Washington, D.C., and New York City in 1969, the trend
did not spread through the rest of the country. Various state and federal HSR propositions
followed but full implementation of an inter-state HSR has never been accomplished. The
closest the United States currently has to an HSR system is the Acela, formerly known as
Acela Express. The Acela is a high-speed service along the Northeast Corridor in the
Northeastern United States operated by Amtrak and replaced the aging Metroliner [4]. The
Acela provides a route from Washington, D.C. to Boston with 16 intermediate stops which
makes the service inter-state, but the top speeds of 240 km/h limits the service to be
categorized as a higher-speed rail (HrSR). Higher-speed rail is the jargon used to describe
inter-city passenger rail services that have top speeds higher than conventional rail but are
not high enough to be considered high-speed rail services [5]. Typically, an inter-city rail
service must have a minimum speed of 250 km/h to be considered as a high-speed rail
service.

In 2008, the California HSR network was authorized by voters with Proposition 1A which
would mark the largest project for American HSR, connecting the bay area to southern
California. At the time of the proposal, the project was sold to voters with a projected cost
of $33.6 billion; however, by 2018 the California High-Speed Rail Authority revised its
estimate to $77.3 billion and up to $98.1 billion anticipating a 2033 completion year [16].
Unfortunately, the fluctuating project cost estimates and delays has led to cancelation of
 2

major federal grants which funded the project. Construction for the maiden California HSR
infrastructure finally started in 2017 but all segments besides the Central Valley segment
from Bakersfield to Merced are indefinitely postponed due to cost overruns and delays as
of 2020.

On the contrary, an interstate project between California and Nevada and a project in Texas
is progressing towards success as of 2020. XpressWest, a passenger rail project connecting
Las Vegas and greater Los Angeles, has received the rights to build on the median of
Interstate 15 which runs through Southern California and Intermountain West. This
privately funded project was acquired by Florida-based passenger rail operator Virgin
Trains USA and anticipates its first service in 2023 [8]. An HSR line is also being proposed
between Dallas and Houston by a private railroad company called Texas Central. Current
plans include utilizing technology based on that of the Central Japan Railway Company
with rolling stock based on an international version of the N700 Series Shinkansen [7].

Independent of the California HSR progress, privately funded HSR projects are bringing
an upward trend to a successful implementation of monumental HSR in the United States.
Thus, providing guidance on the modeling, analysis, and design of HSR infrastructure and
structural systems could be greatly beneficial to inform future national and local HSR
research and projects within the United States.

Problem Description

Bridges are a key component of the HSR infrastructure because it can avoid the interruption
of existing roadways and the occupation of land. China, the world’s largest user of HSR,
incorporates bridges as a major part of their HSR infrastructure, covering more than 50%
of their total HSR mileage [43]. As of February 2020, China has over 35,000 km of HSR
track in operation and continues their advancement as the world’s unrivaled largest user of
HSR in operation with the next largest being Spain with 3,000 km [28]. Several other
European countries have built extensive HSR networks that now include several cross-
border international HSR links and the European Union continues to invest in the
development of HSR infrastructure. Countries within these regions have developed a
 3

standard design for their HSR infrastructure and stands as a great design reference for
future projects within the United States.

The inherent characteristics of HSR raise new problems beyond those found in typical
highway construction, so comprehensive numerical approaches on the bridge structure
modeling are needed. Good understanding of the sensitivity of a bridge span vertical
deflections and rotational deformations, as well as train-track-bridge dynamic interactions
and coupling vibrations are of great importance when designing HSR bridges. Compared
with a conventional railway bridge, the design of HSR bridges require a higher service
limit to minimize deformations and avoid excessive vibrations or resonance due to the
crossing of trains to improve the riding comfort for passengers. The focus of this research
is to identify these modeling features and inherent characteristics of HSR bridges and
provide guidance and demonstration examples on how to develop such models. These
models will aid researchers and designers in conducting parametric studies to test the static,
modal, and dynamic performance of HSR designs to inform and optimize future designs,
and eventually formulate a national design standard for HSR infrastructure in the United
States.

Research Objectives and Scope of Work

The main objectives of this study were to: (1) synthesize available national and
international literature on modeling and numerical simulation of HSR systems, (2) identify
critical modeling features needed to develop a detailed finite element model, based on
synthesized literature, that captures HSR train-track-structure interaction when simulating
service loads and extreme events such as earthquakes, and (3) develop a step-by-step guide
on the modeling and analysis of HSR bridge systems in OpenSees, an open source
framework developed by the Pacific Earthquake Engineering Center.

To achieve the first objective, modeling techniques from literature published by researchers
around the world were analyzed and compiled to understand the dynamic train-track-bridge
interactions. Studies modeling different types of high-speed train systems, track systems,
and bridge systems were explicitly researched to offer a comprehensive literature search
 4

that will allow the reader to gain insight on the modeling techniques of various HSR
systems.

From previous studies, a prototype train, track, and bridge system were selected based on
available information that can be incorporated into a prototype model. The selections were
then used to create a detailed HSR model in OpenSees using the modeling techniques
synthesized in the extensive literature search to achieve the second objective. The model
was created to demonstrate the functionality of the modeling techniques highlighted in the
first objective. The model was further tested under service loads and ground motion
excitations to demonstrate the various capabilities and analyses that can be performed.

To achieve the third objective, a walk-through of the steps taken to model the selected
prototype HSR system from start to finish was documented along with recommendations
and assumptions made during the process. Further demonstration of the nonlinear seismic
response of the prototype HSR bridge was presented through a brief analytical study. The
latter highlighted the performance under various train loading scenarios and ground
motions amplified to various degrees. This objective aims to encourage higher
understanding of HSR bridge behavior in high seismic areas. Overall, this study contributes
to the advancement of research involving HSR systems by creating a readily
comprehensible guideline for students, researchers, and bridge designers to embark on
creating their own HSR models for future studies.

Organization of the Thesis

This thesis is organized into six chapters and two appendices. Following the first
introduction chapter, Chapter 2 provides an in-depth literature review on the numerical
modeling of train, track, and bridge systems that make up HSR systems. Chapter 3 presents
a guide on modeling a sample high-speed rail system by selecting prototype train, track,
and bridge systems and demonstrating the numerical modeling techniques researched in
the literature. Chapter 4 provides a demonstration for gravity load analysis, modal analysis,
and seismic analysis of the structural model created in Chapter 3 along with interpretations
for the structural response from the respective analyses. Chapter 5 presents a more in-depth
seismic performance analysis of the structural model by conducting nonlinear time history
 5

analysis under three different ground motions with various intensities and for different train
load cases. Chapter 6 outlines the summary and conclusions from this research, along with
providing the research impact and recommendations for future works. The first appendix
provides screenshots for the key adopted OpenSees commands and syntax for reader
convenience. The second appendix provide the step-by-step script examples from a sample
OpenSees input file.
 6

2 SYNTHESIS OF NATIONAL AND INTERNATIONAL STUDIES ON THE TOPIC

OF NUMERICAL MODELING OF HIGH-SPEED RAIL SYSTEMS

Overview

Following the rapid growth of high-speed railway transportation and the advancement of
railway technology driven by an increasing demand for more efficient, cost-effective, and
safer railway transportation, precise analysis of dynamic interaction for vehicles and
bridges has become an issue of great significance. To encourage comprehensive
understanding of proper idealization of such systems, modeling techniques for train, track,
and bridge systems from national and international studies, and available design guidelines
have been studied and synthesized in their respective sections. The scope of the literature
search conducted herein focuses mainly on the modeling of superstructure components,
and only briefly touches upon the modeling methods of substructure components.

Modeling of Train Systems

High-speed train systems are mainly constituted by two vehicle systems: traditional vehicle
systems and articulated vehicle systems. A traditional vehicle system is characterized by
two bogies or trucks in the fore and rear parts of the car-body, and each passenger car
behaves independently (Figure 2-1). Each vehicle has one car-body, two bogies, and four
wheelsets. On the contrary, an articulated vehicle system as shown in Figure 2-2 connects
successive passenger cars by a single bogie frame (Figure 2-2b), but the power car and
motorized car at each end of the high-speed train are still supported by their own bogies
like a traditional vehicle system (Figure 2-2d). The articulated vehicle system restrains the
composition of the train but is proven to effectively improve the riding conditions
compared to traditional vehicle systems by reducing the vibration generated in each car
body [36].

Traditional Vehicle System

In early studies, vehicles were often approximated as a moving mass model to consider the
inertial effects of moving vehicles and to allow the problem to be solved analytically.
However, the effect of the suspension system must be considered for accurate vehicle
 7

response. The simplest model in this regard is a lumped mass supported by a spring-dashpot
unit, often referred to as the sprung-mass model [2, 9, 14, 15, 26, 27, 37, 40, 41, 44, 45,
50]. The sprung-mass dynamic system can reflect the motions of the vehicle in both the
vertical and horizontal directions. The car-body, bogies and wheelsets in each vehicle are
assumed as rigid bodies, neglecting elastic deformation, and are connected to each other
three-dimensionally by linear springs and dampers. The primary and secondary suspension
systems of the bogies are simplified as an elastic system with linear springs and viscous
dampers. Placement of the spring-dashpot units within each suspension system differ
slightly among studies depending on the type of HSR train system and the specific bogie
design, as can be seen by comparing the various train model schematics in Figure 2-3
through Figure 2-7.

Another method is to model the car-bodies, bogies, and wheelsets as beam finite elements
and the suspension system as a variation of bilinear and multilinear springs in the three
directions. Montenegro et al., [29] have modeled all springs characterized by a bilinear
behavior, except the one used to model the secondary transversal suspension which follows
a multilinear law to simulate the presence of rubber stoppers whose stiffness increases
gradually (Figure 2-5). Nonlinear springs can be used to model the suspension system, but
most of the studies have simplified the analysis by assuming a linear behavior.

The car-bodies and bogies are typically assumed to move along a well-maintained straight
track at a constant speed, and the wheels and the track to always keep in contact, neglecting
sliding, climbing or derailment phenomena [13, 24, 26, 36, 47, 50]. The assumption of
perfect contact between wheel and track is commonly represented as the vehicle-track
interaction by coupling the displacement degree-of-freedom (DOF) relationships between
the rail and wheel-set subsystems. A Hertzian contact spring can be placed in-between each
wheel and rail to accurately model the wheel-rail contact stiffness by consider the changing
contact area caused by the indentation of the rail due to the geometry of the wheel [3, 30,
35].

The main difference of vehicle modeling among studies is the selection of the DOFs to be
concerned in the car-body, bogies, and wheelsets. Each node has a maximum of six DOFs
 8

in finite element modeling but not every DOF is taken into consideration depending on the
study. Typically, each car-body and each bogie have five DOFs in consideration: lateral
displacement, roll displacement, yaw displacement, vertical displacement, and pitch
displacement. The sliding displacement is often omitted because the high-speed train is
assumed to be in motion and not stationary [2, 24, 37, 42]. Although rolling and sliding
motions would be excited due to torsional vibrations and track irregularities, these motions
are commonly constrained for efficiency of formulation [36]. On the contrary, Xia and
Zhang, [41] and Liu et al., [24] have included the rolling motion in the concerned DOFs.
If the train system is being modeled in a scenario where seismic loading is present, the
rolling motion should be accounted for because the seismic loading would heavily excite
the rolling motion in the car-bodies and bogies, as the wheel-sets are assumed to stay in
direct contact with the rails. The concerned DOFs for the wheelsets can be limited to the
lateral displacement, vertical displacement, and the roll displacement [24, 26]. The other
DOFs can be neglected because the wheelset is constantly in rotation and the wheels always
stay in contact with the track system. Various schematics of traditional vehicle systems are
shown in Figure 2-3, Figure 2-4, and Figure 2-5 as previously mentioned.

Articulated Vehicle System

For articulated vehicle systems, each passenger car no longer behaves independently, and
the behavior of each bogie will be affected by the dynamic behavior of the fore and rear
car-bodies. Aside from the coupling of intermediate passenger cars, the modeling
procedure of articulated vehicle systems are similar to the traditional vehicle system. The
model by Kwark et al., [19] individually modeled the car-bodies, the bogie in between, and
the wheels with DOFs as shown in Figure 2-6. Additional damping due to a central elastic
hinge in-between adjacent car-bodies was modeled by transverse springs and dampers, also
seen in the model by Xia et al., [42]. Another method is to model the fore and rear car-
body behavior as a single joint directly above the articulated bogie. In Song et al., [36]
study, the bouncing, swaying, pitching and yawing motions are considered for the non-
articulated power cars and these motions were condensed into two DOFs by the bouncing
motion and swaying motion at the joint for the articulated vehicles, as shown in Figure 2-
7. The bogie considered the bouncing, sliding, swaying, pitching, rolling, and yawing
 9

motion, so each car had a total of 16 DOFs. The car-body masses are lumped at the joints
and the bogies are connected through rigid bodies with masses. This method was also
followed by Rocha et al., [35].

Modeling of Railway Track Systems

Rail

Rails in HSR systems mainly rest on two types of foundations: ballasted foundations and
ballastless foundations. For both systems, a single track consists of two rails that are
designed to behave elastically as a capacity protected element. Therefore, they are modeled
as a series of linear elastic beam-column elements, and this method is consistent throughout
numerous research studies investigated for this report [22, 23, 24, 50]. If bridge abutments
are being modeled, the rail elements should be extended past the abutments to the
embankments to correctly represent the transition zone [23, 29].

When the train system is being modeled as a moving load, rail irregularity is commonly
considered to simulate the complex time-varying random dynamic behavior that occurs
when a high-speed train crosses over a bridge. Safety, stability, comfort, service-life of
train and track components, as well as the environmental noise of the train is influenced by
irregularity in the rails [25]. Vertical irregularity considers roughness of the rail surface,
elastic deformation, inelastic deformation, inconsistency of gap components, and uneven
subsidence of track foundations. Rail irregularities are approximately represented as
stationary and ergodic processes in space due to its random nature and is most frequently
characterized by power spectral density (PSD) functions [30, 35, 36, 49]. The PSD
functions are adjusted based on the characteristics of the rails used in each country.

Ballasted Track System

For ballasted track systems, rails rest on an elastic foundation composed of track ballast
and railroad ties (Figure 2-8). Ballast is the crushed material placed on the top layer of a
bridge superstructure to allow the embedment and support of railroad ties, also known as
sleepers. The ballast is traditionally made of interlocking sharp-edged hard stone to
stabilize the track system. Rails are fixed to railroad sleepers by fasteners. Rail pads are
 10

placed between the rail and tie to act as a damper that reduces fatigue cracking of fasteners
due to impact. Rail ties are rectangular wood or reinforced concrete supports placed
transverse to the rail and maintains correct gauge spacing between the rails.

A ballasted track system modeled by Song et al., [36] is shown in Figure 2-9. The figure
demonstrates a simple model with rails and sleepers as beam elements and ballast as
Winkler springs to idealize a two-parameter elastic foundation that models the interaction
between the track and the bridge deck. Ties were modeled as beam elements and lay on the
ballast, modeled similar to the Winkler foundation consisting of infinite closely spaced
linear springs. It is noted that the traditional Winkler foundation, based on the Winkler
hypothesis, does not consider interaction of springs. On the contrary, the additional second
parameter suggested by Zhaohua and Cook, [51] considers the effects of the interaction
between the linear spring-dampers which accurately represents characteristics of practical
foundations.

The ballasted track system modeled by Montenegro et al., [29] similarly modeled rails and
sleepers as beam elements (Figure 2-10). The stiffness and damping of the rail
pads/fasteners are combined and modeled as linear spring-dampers to simulate the dynamic
behavior of this layer. The ballast and non-structural elements such as safeguard and edge
beams of the deck were modeled as point mass elements. Spring-dampers are also used to
idealize the stiffness and damping of the ballast layer in the longitudinal, transversal, and
vertical directions.

Guo et al., [11] modeled both the sleepers and ballast as point mass elements at an interval.
The sleepers were connected to the rail through distributed spring-dampers simulating the
dynamic behavior of rail pads. The vertical and horizontal stiffness and damping of the
ballast were idealized with spring-dampers which also connect the ballast layer to the
sleepers. Shear stiffness of the ballast layer was also explicitly modeled as spring-dampers,
and rigid arms connected the ballast to the bridge deck (Figure 2-11).

Ballastless Track System

As the name suggests, ballastless track systems utilize slabs instead of ballast (Figure 2-
12). The typical design includes continuous welded rails, track plates, base plates, and
 11

connecting members [22, 23]. Connecting members can vary depending on regional design
standards. In the study by Li et al., [22], the China Railway Track System (CRTS) II
ballastless track was adopted and includes sliding layers, shear cogging, concrete asphalt
(CA) mortar layers, shear reinforcement, fasteners, and lateral blocks as connection
members. Similarly, the Japanese reinforced concrete roadbed system (RCRS) slab track
utilizes fasteners, track slabs and CA mortar (Figure 2-13). The study by Li and Conte, [23]
for the California High Speed Rail (CHSR) Authority adopted connecting members of
direct fixation fasteners for rail-track slab attachment and cylinder bollards as shear
reinforcement to anchor the track slab to the concrete base plate. Figure 2-14(a)
demonstrates the modeling schematic of a CHSR ballastless track system by Li and Conte,
[23]. The rails were connected to the rigid deck through direct fixation fasteners modeled
as a series of three elastic and inelastic springs to represent the behavior between the rails
and track base.

To represent the rail-structure interaction, linear springs were used to model the vertical
and transverse stiffness, and an elastic–perfectly–plastic (EPP) spring was used to model
the resistance of the track base against the relative longitudinal displacement of the rail
track. Additionally, longitudinal boundary springs were modeled at each rail end because
of the finite length modeling of the rail extensions to accurately capture seismic response
performance. A nonlinear spring model, defined as a single element, denoted as series-
parallel (S-P) spring model, was developed to represent the longitudinal boundary spring.
A mechanical model was developed to calibrate and validate the rail boundary spring
model, and the cyclic hysteresis behavior of the mechanical and S-P model is shown in
Figure 2-14(b). The closeness of the behavior validates the S-P model.

In the China Railway Track System (CRTS) study by Li et al., [22], the track plate and
base plate were modeled using linear elastic beam-column elements with their respective
cross-section parameters because they are designed to behave elastically as capacity
protected elements (Figure 2-15). The connection components consisting of the sliding
layer, CA mortar layer, fastener, shear reinforcement, and lateral block are simulated using
nonlinear zero-length elements.
 12

Modeling of Bridge Systems

Deck and Girder

Concrete box girder bridges were found to be the common bridge type used in HSR
systems. Such type is commonly modeled using three-dimensional linear elastic beam-
column elements, even when representing bridges in highly seismic areas, since they are
structurally designed to be capacity protected elements that need to remain essentially
elastic [19, 22, 23, 29]. Figure 2-16 and Figure 2-17 schematically show example box-
girder bridge idealization and modeling as relates to the track modeling for HSR systems
from two previous studies. As shown in the figures, bridge spans are discretized into
several nodal increments to allow for the representation of different section properties at
the ends of each spans and to accommodate the rail track-to-deck connections and deck-
to-bearing connections. Each increment was connected using linear elastic beam-column
elements defined by the cross-sectional characteristics of the actual bridge being modeled,
and rigid arms were used to connect the bridge girder to the rail and bearing systems. The
increment lengths should be adjusted relative to the actual bridge span dimensions and
based on the desired accuracy of bridge response values. Bridges have also been modelled
as an assemblage of three-dimensional beam elements in the elastic domain with six DOFs
at each node as illustrated in Figure 2-18 and Figure 2-19 [13, 22].

Three-dimensional shell elements have also been used to idealize bridges. Song et al., [36]
utilized nonconforming flat shell elements (NFS-series) formulated by a linear
combination of the nonconforming membrane element with drilling DOF (NMD-series)
and the nonconforming plate bending element (NPB-series). NFS elements with six DOFs
per node are used to model the box-girder structure as shown in Figure 2-20. In-plane and
out of-plane deformations are coupled and the consistent mass matrix of the NFS element
is lumped at the element joints using the HRZ lumping scheme [36]. When the
superstructure and track system are modeled using NFS elements, consisting of four nodes
with six DOFs per node, it is common engineering practice to use a relatively fine finite
element grid in areas of high stress gradients due to abrupt geometrical changes or
concentrated loading and a course finite element grid in areas of uniform stress gradients.
 13

Transition zones between the fine and coarse grids are modeled using variable-node NFS
elements [36].

In another study, a combination of flat plate elements and beam elements were used to
model a steel plate girder bridge. In Kim et al., [18] study, a steel girder bridge was
idealized by modeling the concrete decks as flat plate elements with four nodes and the
steel girders, cross beams, and guard rails of the bridge as linear elastic beam elements with
six DOF nodes. As a similar steel bridge, a steel box girder bridge has been idealized by
modeling the concrete deck as a solid element and the steel box as shell elements [24].
Headed shear studs that connect the concrete deck to the steel boxes are modeled as linear
spring elements in the longitudinal direction and coupled in other directions [34].

Pier Column

Pier columns can be modeled using a number of fiber-based elements such as

displacement-based fiber-section beam-column elements [23], fiber-based force-based
beam finite elements [17], and three-dimensional elastoplastic fiber elements [22]. Fiber
based elements account for material nonlinearity, geometric nonlinearity, and bond slip
effect of anchoring steel in joints, making it an accurate plastic hinge representation.
Integration points are placed along the length of the element in each column to allow for
inelastic behavior at every point. Column cross sections are discretized into fibers in polar
coordinates as shown in the Section A-A examples in Figures 2-17, 2-18, and 2-21, with a
specific nonlinear uniaxial material model assigned to each fiber, i.e. unconfined concrete,
confined concrete, and steel rebar [17, 22, 23]. To obtain the behavior of the nonlinear
column section, the fiber behavior over the column cross-section is integrated. Potential
plastic hinge regions (bottom of column for seismically isolated bridges, and both top and
bottom of column for non-isolated bridges) are modeled using a single element with length
equal to the plastic hinge length, approximated as half the column diameter, to ensure mesh
objectivity of the finite element response prediction. The portion of the column-bent
embedded in the superstructure was modeled as a rigid element attached to the top of the
nonlinear beam-column element, and the length of this rigid element is set equal to the
 14

distance between the top of the column and the centroid of the soffit-flange of the box-
girder.

If a bridge is being modeled to observe the response under moderate earthquakes, the
columns may be modeled with a linear elastic behavior, because unlike highway bridges,
the HSR bridge columns generally do not experience significant damage in this case. An
alternate methodology by Montenegro et al., [29] estimated the effective stiffness of the
columns performed in the elastic domain, considering reduction in stiffness due to
cracking. The material behavior of the columns should be decided based on the magnitude
of the excitation applied to the structural model and the overall purpose of the model. A
number of studies have completely omitted the modeling of bridge piers and limited their
model to the train, track, and deck/girder system [11].

Pier Column Foundation

Column supports can be modeled with a variety of complexities depending on the intended
study or analysis emphasis on soil-structure interaction. If the focus of the model is to
analyze the train-track-structure interactions, the soil-structure interaction can be
simplified to a few springs modeled between the fixed base and the bottom of the column
footing elements. He et al., [13] modeled the elastic effects of column footings, pile
structures and the surrounding soil by placing longitudinal and transversal ground springs
at the bottom of each column.

Li and Conte, [23] have extensively modeled HSR bridge deep pile foundations using a
variety of elements. The schematic from their study is shown in Figure 2-22, along with
the geometric and material properties that represent the bridge site considered in their
study. The well-established p-y approach was used in modeling the pile foundations and
each pile was modeled through displacement-based nonlinear fiber-section beam-column
elements. These piles were supported by a series of springs distributed along the length of
the pile representing the resistance of the surrounding soil, p-y springs for horizontal
resistance and t-z springs for vertical resistance. These springs represented the horizontal
and vertical resistance of the surrounding soil, and Q-z springs were placed at the pile tips
to represent the vertical soil end-bearing. Pile caps were considered essentially rigid and
 15

rigidly connected to the top of each pile, thus modeled as quasi-rigid beam elements to
capture the various geometric offsets. Hyperbolic p-y springs were attached to the pile caps
to represent the lateral soil resistance. Similarly, Li et al., [22] have modeled pile
foundations as three-dimensional elastoplastic fiber elements. The fiber elements were
divided into 1 m intervals and connected to the soil through three translational and three
rotational springs with constant spring values to simulate the pile-soil interaction (Figure
2-18).

Isolation Bearing

A bridge bearing is a component of the bridge placed between the bridge superstructure
girders and substructure pier/bent. Bearings transfer deck loads to piers or bents and allow
specific movements and rotations of the superstructure. Studies that include bearings are
limited but explicitly modeling bearings allows the user to capture the interaction between
bridge decks and columns. Li and Conte, [23] idealized a generic seismic isolation device
with a material of bilinear inelastic force-deformation behavior. Each bearing is modeled
as a zero-length element combined with two uncoupled bilinear inelastic materials for the
horizontal behavior: one in the longitudinal direction and the other in the transverse
direction of the bridge. Li et al., [22] similarly idealized bearings as zero-length nonlinear
connection elements. Each girder span was supported by four steel bearings, with
alternation between fixed and spherical bearings to minimize torsional effects. An elastic-
perfectly-plastic force-deformation material behavior was used to model the nonlinear
characteristics of the bearings. Linear spring-dampers were used to idealize bearing
supports in a study by Montenegro et al., [29] for moderate earthquakes.

General Modeling Procedures

Rigid Connection Arm

Connections between bridge and track elements are commonly modeled using a type of
rigid arm or element. The use of rigid arms allows the user to simplify structural
components connecting these elements to each other and allow load transfer throughout
the structure. For this study, rigid arms are used to connect the centroid of bridge girders
to the track system and bridge girder supports in a similar way to what have been adopted
 16

in previous studies and illustrated in Figure 2-16, Figure 2-17, Figure 2-18, Figure 2-22,
and Figure 2-23.

Linear elastic beam-column elements assigned with exceedingly stiff properties, referred
to as quasi-rigid objects, can be used to represent the rigid offset between respective
element nodes such as the rail and deck. Quasi-rigid objects allow the user to extract the
internal forces between the two nodes in connection. The finite element model scheme
utilizing quasi-rigid beam elements by Li and Conte, [23] is displayed in Figure 2-17. The
figure illustrates the use of quasi-rigid beam elements to connect the centroidal axis of the
box girder deck to the track system along a single span. The rigid element also connects
the isolation system to the column substructure and box girder deck at the ends of each
bridge span.

Another method for modeling rigid arms is to use rigid links. A rigid link is an explicit
command in different analysis platforms such as OpenSees that allows the user to constrain
DOFs between a master node and slave node. The command offers two types: bar/rod and
beam. The bar/rod type rigid link constrains only the translational DOFs of the slave node
to be the exactly the same as those at the master node. The beam type rigid link constrains
both the translational and rotational DOFs of the slave node to the master node. The
advantage of using rigid links is the simplification of the element stiffness matrix. Rigid
links reduce computational effort but does not allow the user to extract the internal forces
between the two nodes connected by the rigid link. A modeling schematic by Montenegro
et al., [29], utilizing rigid links, is shown in Figure 2-16. The placement and use of rigid
links are almost identical to quasi-rigid objects discussed previously.

Viscous Damping

Energy dissipation can be idealized in finite element models through inelastic materials
applied to elements, as mentioned in previous sections, and a method of viscous damping.
Although the hysteretic damping included within the elements with nonlinear behavior can
dissipate the majority of energy introduced by a seismic load, energy dissipation due to
inherent non-hysteretic damping must be accounted for through the application of viscous
damping to obtain a realistic result. A Rayleigh damping scheme with a specified damping
 17

ratio at two selected modes is commonly used to idealize such damping due to vibration,
and applies to all structural components of the bridge model that are not highly nonlinear
elements [10, 20, 21, 40, 50]. The Rayleigh damping scheme forms the damping matrix
through a linear combination of the stiffness and mass matrices of the numerical model,
and a damping ratio of 2% has been commonly used for HSR bridges [23, 29, 36]. Higher
values of 3% and 5% have also been reported and used in other studies [13, 48]. The
damping coefficients are usually estimated based on the dominant transverse and
longitudinal vibration modes, which are estimated from an eigenvalue analysis that uses
the tangent stiffness matrix of the bridge system after application of the gravity loads
through static analysis.
 18

Figure 2-1. China-star high-speed train [41].

Figure 2-2. Views of the KHST (a) panoramic view, (b) articulated bogie located
between the car bodies, (c) articulated bogie and (d) composition of the train (front power
car) [19].

Figure 2-3. Front view of the sprung-mass dynamic car model [29].
 19

Figure 2-4. Tradition train system modeled by He et al., [13].

Figure 2-5. Traditional train system modeled by Liu et al., [24].

Figure 2-6. Articulated train system modeled by Kwark et al., [19].

 20

Figure 2-7. Bogie–bridge interaction system in an articulated train system modeled by

Song et al., [36].

Fastener

Rail
Sleeper/Tie

Ballast

Figure 2-8. Photo of ballasted track system [33].

 21

Figure 2-9. Ballasted track system modeled by Song et al., [36].

Figure 2-10. Ballasted track system modeled by Montenegro et al., [29].

Figure 2-11. Ballasted track system modeled by Guo et al., [11].

 22

Figure 2-12. Photo of ballastless track system [39].

Figure 2-13. Japanese type RCRS slab track on grade [38].

Figure 2-14. Track system scheme with fasteners (a) and longitudinal boundary spring
hysteresis loop (b) by Li and Conte, [23].
 23

Figure 2-15. Modeling schematic of ballastless track system modeled by Li et al., [22].

Figure 2-16. Modeling schematic of track-bridge system by Montenegro et al., [29].

 24

Figure 2-17. Modeling schematic of track-bridge system by Li and Conte, [23].

Figure 2-18. Modeling schematic of bridge system by Li et al., [22].

 25

Figure 2-19. Modeling schematic of bridge system by He et al., [13].

Figure 2-20. Concrete box girder modeled using shell elements by Song et al., [36].

Figure 2-21. Modeling schematic of bridge pier columns using fiber-based elements by
Kaviani et al., [17].
 26

Figure 2-22. Pile foundation model using dynamic p-y approach: (a) schematic view of
the FE model, (b) pile cap mode [23].

Figure 2-23. Modeling schematic of rigid connections by Kaviani et al., [17].

 27

3 HSR BRIDGE SYSTEM NUMERICAL MODEL: SELECTION OF

PROTOTYPE SYSTEM AND MODELING PROCEDURE

Introduction

This chapter presents the process of formulating a sophisticated train-track-structure

interaction model of a prototype HSR system. A prototype bridge, track, and train system
were selected from the studies researched in the literature search. The prototype track-
bridge system was selected based on the completeness of the design guideline provided in
the reference study, such as bridge dimensions and cross-sectional properties. Assumptions
were made where information was omitted in the reference study. This was not a major
issue because the purpose of this study was to demonstrate how to model an HSR system
as opposed to discuss or assessing the viability of a certain design. Similarly, the prototype
train system was selected from a reference study that explicitly stated the masses of the
various train components, as well as the stiffness and damping properties of the primary
and secondary suspension systems, which are critical to accurately simulating the dynamic
behavior of an HSR system.

Selection of Prototype HSR System

Train System Prototype

The prototype train system selected for this study is the KTX-Sancheon high-speed train
which is shown in Figure 3-1. Formerly known as the KTX-II, the KTX-Sancheon is the
second commercial high-speed train operated in South Korea as part of the Korea Train
eXpress (KTX), making its debut in 2010 [6]. The KTX-Sancheon consists of two power
cars at both ends and an articulated set of eight intermediate passenger cars in-between. As
mentioned previously, an articulated bogie system couples a passenger car with the fore
and rear passenger car, improving riding conditions of the train. As can be seen in Figure
3-1, the power cars have two standard bogies, and the extreme intermediate passenger cars
have a standard bogie and an articulated bogie coupling them with the intermediate
passenger cars.
 28

Track and Bridge System Prototype

The prototype track-bridge system selected for this study is a ballastless track prestressed
concrete double-track simply supported girder bridge used in a publication by Li et al.,
[22]. The track-bridge system is from the Beijing to Xuzhou section of the Beijing-
Shanghai high-speed railway. The bridge has 10 equal spans of 31.95 m with a total length
of 319.5 m. The bridge superstructure is made of C50 concrete and is 13.40 m wide at the
top, 5.74 m wide at the bottom, and 3.09 m deep from the top to bottom surface. Each
girder end is supported by two spherical steel bearings that rest on the 11 single column
bents of 13.5 m height, made of C50 concrete and HRB335 steel bars. The bridge properties
and overview as obtained from the reference study is shown in Figure 3-2.

The CRTS II slab ballastless track was adopted for the track system and comprises of base
plates, track plates, rails and connecting members. The connecting members include sliding
layers, shear cogging, CA layers, shear reinforcement, fasteners, and lateral blocks. The
CHN60 rails are fixed to the base plate through WJ-8C fasteners. The track plate is made
of C55 concrete and has a width and thickness of 2.55 m and 0.20 m, respectively. The
track plate is connected to the C30 concrete base plate of 2.95 m width and 0.19 m thickness
through the CA layer. Shear reinforcement bars are placed at the girder ends in the CA
layer to withstand the deformation caused by rotation, and the sliding layer is arranged
between the bridge deck and the base plate. The sliding layer, CA layer and fasteners allow
for longitudinal slippage relative to the bridge and the lateral blocking provides support in
the transverse direction relative to the bridge. The layout of the connection layers is shown
in Figure 3-2(b) and Figure 3-3.

Numerical Model in OpenSees

OpenSees is an object-oriented, open source software framework that allows users to create
both serial and parallel finite element computer applications for simulating the response of
structural and geotechnical systems subjected to earthquakes and other hazards [32].
OpenSees allows the user to build a structural model by using the numerous commands
available in the program. The commands used in the model for this study are discussed in
this section. For the convenience of the reader, the syntax and input parameter of the key
 29

OpenSees commands or functions used throughout this study are presented via series of
screenshots provided in Appendix A. Moreover, sample scripts that represent or form the
main sections of a typical HSR bridge model in OpenSees are provided in Appendix B. In
the discussion presented in this section as well as the next chapter, specific figures from
both Appendix A and Appendix B are explicitly referenced in the text for completeness
and convenience. Figures from Appendix A and Appendix B use a numbering sequence
that starts with A or B, respectively, such as Figure A-5 or Figure B-11 for instance.

Basic Model Definitions

To start a model, the user must define the spatial dimensions (1, 2, or 3) and the number of
DOFs (1, 3, or 6) at each node, using the model command shown in Figure A-1. Since a
three-dimensional model was created for this study, the spatial dimension was specified as
3 and the DOF at each node was specified as 6 to account for all translational and rotational
movement. The user can then construct numerous nodes which will be used to construct
the framework of the structure. The node command requires a unique tag number and the
x, y, and z-coordinates to define the location (Figure A-2). OpenSees uses the numbers 1,
2, 3, 4, 5, and 6 to define the three translational and three rotational DOFs, respectively.
For this specific model, the x-coordinates were modeled in direction 1, the y-coordinates
in direction 2, and the z-coordinates in direction 3.

Single-point (SP) homogeneous boundary constraints can be implemented using the fix
command, and multi-point (MP) constraint between nodes can be defined using the
equalDOF command (Figure A-3 and Figure A-4). The fix command is typically used at
the base of the structure and was used at the foundation in this model. The equalDOF
command was used to maintain structural stability between zero-length elements where
stiffness was not defined for every DOF. The way in which the local coordinates of the
elements correlate to the global coordinates of the model is defined using the geomTransf
command (Figure A-5). This command defines how OpenSees transforms the stiffness and
resisting forces of the beam element from the local system to the global-coordinate system.
Specifically, the basic linear geometric transformation method was selected for this study.
Careful attention should be given towards assigning the vector orientations for elements
 30

since this could result in element cross-section properties such as inertia in the local y and
z axis to be flipped if defined incorrectly. A very helpful visual demonstration is provided
in the OpenSeesWiki, [32] which should be referenced.

The next step is to define material properties used in the model. OpenSees has a wide
variety of uniaxial materials, including steel and concrete materials. The uniaxialMaterial
command is used to construct a material object which represents uniaxial stress-strain
relationships [32]. Steel01, Steel02, Concrete02, ViscousDamper and Elastic material
commands were used in this study to model the nonlinear behavior of the train, track, and
bridge system components (Figure A-6 through Figure A-10). The Steel01 material was
used to simulate the behavior of bearings and the connection layers in the track system.
Steel02, Concrete02 and Elastic materials were used to simulate the pier columns, and
ViscousDamper materials were used to model the train suspension system. These materials
were then specified as a parameter for the construction of elements.

Three types of elements were used in the model: elastic beam-column elements,
displacement-based beam-column elements, zero-length elements, and two-node links
(Figure A-11 through Figure A-14). The elastic beam-column elements were used to model
the elastic capacity protected elements like the bridge girder. This element command
requires the section properties and not the material behavior because they remain elastic.
Displacement-based beam-column elements were used to model the pier column. To
accurately model the behavior of the columns, the cross-section must be modeled using the
section fiber command (Figure A-15). The patch and layer commands allow the
construction of several fibers within a predefined cross-section to model the behavior of
cover concrete, core concrete, and steel reinforcement with the material properties that
were defined (Figure A-16 and Figure A-17). The specific details will be explained later in
Section 3.3.4.3. The fiber section can then be aggregated into an existing elastic material
using the section aggregator command (Figure A-18). The new aggregated material can
then be used as the material parameter for the displacement-based beam-column elements.
zeroLength element were used together with the Steel01 material to simulate the bridge
bearings and track connection layers. twoNodeLink elements were used together with the
 31

ViscousDamper material to simulate the damping in the train suspension system, and the
stiffness in the train suspension system was simulated using an elastic material. A complete
list of elements and materials used in the prototype model is presented in Table 3-1.

The mass of each component in the model can be defined using the mass command in
OpenSees (Figure A-19). The mass command allows the user to set the nodal mass values
corresponding to each DOF. Defining masses allows the user to perform modal and
dynamic analyses but is not required for static analysis. For this study, analysis of the modal
and dynamic behavior of the structure was of interest, so the mass command was used to
set translational and rotational mass values at every appropriate node. Mass values were
applied at the nodes representing the centroid of the train system components and bridge
footings, and the masses of the rest of the track-bridge system components were distributed
at every node along the entire length of the rails, track and base plates, bridge girder, and
pier columns.

Train System Model

To model the KTX-Sancheon, a study by Kwark et al., [19] was used as a reference due to
the similarity of the train prototype selected. The train selected by Kwark et al., [19] is a
Korean High-Speed Train (KHST) with an articulated bogie system. Based on the train
configuration described in the study and the year the paper was published, the prototype
train system selected by Kwark et al., [19] was assumed to be the KTX-I, which is the first
set of trains used by the Korea Train eXpress (KTX). The 20-car formation (380.15 m long)
of the high-speed train entered service in 2004 and is optimized for high capacity. In
comparison, the KTX-Sancheon is the second commercial high-speed train operated in
South Korea and was created as a shorter companion to the KTX-I. Initially, the same train
prototype was considered for this study; however, the train was exceptionally long (20 cars
with a total length of 380.15 m) and was conceived as unfit for the prototype bridge
selected. The transition was made to the KTX-Sancheon which has similar car-body and
bogie systems with roughly half the total length (193.15 m). The configuration and
numerical model discretization of the prototype train model used in this study is shown in
Figure 3-4.
 32

3.3.2.1 Train System Model Geometry

Before defining the train nodes, lateral and vertical distances for the general location and
geometric design of the train system were predefined to simplify the modeling process and
allow for easy modification when necessary. As mentioned before, the track system of the
prototype HSR bridge selected is a double-track, which means there is a right (R) and left
(L) track relative to the center of the bridge. From here onwards the right and left tracks
will be referred to as track 1 and 2, respectively. Train dimensions retrieved from the
reference study by Kwark et al., [19] were used to define the train nodes. The train axle
wheels are 3 m apart in the x-direction (w) and 2 m apart in the y-direction (wr), so the rails
for track system 1 were defined as R1 and R2 and are 1 m to the right and left of the track
center line, respectively. Similarly, the rails for track system 2 were defined as R3 and R4.
As previously mentioned, Appendix B provides scripts from the developed OpenSees
model input file for completeness and step-by-step guidance. Figure B-1 in Appendix B is
the first screenshot in the series of model definiton figures which shows the predefined
gemoetric locations for train nodes. The lateral lengths of the power car (Lp), extreme
passenger car (Lm), and intermediate passenger car (Lc) were defined respectively as 14.0
m, 18.7 m, 18.7 m, as well as the total length of the bridge system (LT) as 193.15 m. The
distance between the axle wheels of the power car and extreme passenger car is 3.275 m
(wp) [19].

Various height parameters for the train system were also predefined. The rail height (hr)
was defined as 16.59 m, which is the sum of the column height (13.5 m) and girder depth
(3.09 m). The height of centroid for the bogies (hb) were defined as 0.56 m and the height
of centroid for the power and passenger car-bodies (h) were defined as 1.72 m and 1.627
m, respectively. These values were retrieved from a study by Song et al., [36] who similarly
modeled a Korean high-speed train assumed to be the KTX-I based on the dynamic
properties of the mass constituent elements. The vertical distance between the bottom of
the car-body and center-of-mass of the power car (hp), extreme passenger car (hm), and
intermediate passenger car (hc) were defined respectively as 0.605 m, 0.420 m, and 0.508
m. These values were taken from the reference study by Kwark et al., [19]. To expedite the
process of shifting the train system along the length of the bridge, all train nodes were
 33

defined with an initial variable (x), which is the x-coordinate of the last wheel assuming
the train is moving in the positive x-direction. This practice was beneficial to analyze
various train load cases as part of the seismic analysis conducted in Chapter 5 and is
recommended for future studies. The value (x) is adjusted depending on the load case being
analyzed. Figure B-1 shows how the aforementioned parameters were defined and the “x”
value shown in the snippet is for the load case where the train is loading the second to
seventh spans of the bridge. A summary of all the parameters used for the train system is
shown in Table 3-2.

3.3.2.2 Train System Nodes

Train nodes are created by defining the parameters specified for the node command (Figure
A-2). For large scale structural models for an OpenSees model to be filled with thousands
of nodes, which can be very confusing if the node tags (NodeTags) are not organized. Since
this study is modeling the train system running on track 1, the train node tags were
organized where any tag starting with a 7 specified an alignment on the right side of the
train over R1 (rail 1), a 8 specified an alignment on the left side of the train over R2 (rail
2), and a 6 specified an alignment on the centerline of track 1 (R). This can be seen in the
y-coordinate for the nodes defined in Figure B-2, Figure B-3, and Figure B-4. These figures
in Appendix B are snippets of the rear power car, rear extreme passenger car, and first
intermediate passenger car to demonstrate how they are defined in OpenSees. The second
value of the node tag specifies the vertical grid of the train system as can be seen in the
train model schematic (Figure 3-4). The value 0 is for the wheel nodes, 1 is for the bogie
nodes, 2 is for the primary suspension nodes, and 3 is for the car-body nodes. The second
to last number in the node tag specifies the bogie that the wheel, bogie, or suspension node
is associated with, and the last number further specifies the location of the node within axle
(1 or 2), bogie (1 to 3), or suspension system (1 to 3). For example, NodeTag 70042
designates the node for wheel 2 on the right side of bogie 4, and NodeTag 71052 designates
the node for bogie 5’s center node. This trend is not followed for the car bodies. Instead,
the last digit of the car-body node tags ranges from 1 to 23. Each car-body is constituted
by three nodes and car-bodies for the articulated system share a node as can be seen in
Figure 3-4.
 34

All coordinates are defined using the predefined parameters as discussed in Section 3.2.2.1
above. This allows for simple adjustment of the train dimensions in the case of a parametric
study or adjustment to a potential design. For the intermediate passenger cars, a value “n”
was set to represent the respective number of the 6 intermediate passenger cars. A value of
1 was set for the first intermediate passenger car which was used to define the x-coordinates
of the nodes, and each successive intermediate passenger car nodes were defined by
increasing the n value by 1. The variable “x” previously defined and shown in Figure B-1
is included in the x-coordinates of every train node to shift the location of the entire train
system along the length of the bridge. The z-coordinates were defined with the predefined
train system heights as shown in Figure 3-4. Wheel nodes were modeled at the same height
as rail nodes under the assumption of perfect contact and the height of the bogie nodes
were modeled as the sum of the rail height and bogie height relative to the rail. The z-
coordinate of car-bodies were defined as the sum of the height of their center-of-mass (h)
assumed in Section 3.3.2.1 and the height of the rail (hr). and the top node of the secondary
suspension system as the sum of car-body height (h) and the height of the rail (hr), minus
the respective cars vertical distance between the car-body center of mass to the bottom of
the car-body. The node set up for the rear power car, rear intermediate passenger car, and
first intermediate passenger car are illustrated in Appendix B in Figure B-2, Figure B-3,
and Figure B-4.

3.3.2.3 Train System Rigid Connections

The car-body and bogie are modeled as elastic beam-column elements with exceedingly
stiff properties. The cross-sectional area, Young’s modulus, shear modulus, torsional
moment of inertia of the cross-section, and second moment of area about the local z and y-
axis were assigned exceptionally large values to create a rigid element. Exceptionally stiff
elements can potentially cause convergence issues depending on the type of convergence
test type for analysis, so the values should be defined accordingly. The cross-section values
used for this study as defined in Figure B-5, which were determined to provide appropriate
stiffness relative to the rest of the elements in the model. Examples of the rigid elastic
beam-column elements defined for the bogies are shown in Figure B-5 and Figure B-6.
Similarly, Figure B-7 and Figure B-8 demonstrate the rigid elements for the primary
 35

suspension system. Since the KTX-Sancheon has an articulated bogie system, the
passenger cars act as a coupled unit. The car-bodies for the extreme and intermediate
passenger cars are modeled as rigid beam-column elements in series; however, the power
cars are disconnected from the rest of the system. This is demonstrated in Figure B-9 where
Node 63003 of the power car is not connected to Node 63004 of the extreme passenger car.

3.3.2.4 Train System Suspensions

Flexibility is provided in the train system through the primary suspensions system between
the axles and bogies, and the secondary suspension system between the bogies and car-
bodies. The primary and secondary suspension system of the train were modeled using the
twoNodeLink link element command in OpenSees. This command allows the user to
construct a zero or non-zero length element defined by two nodes and apply material
behavior to any transverse or rotational DOFs for a three-dimensional model. Uniaxial
elastic materials were used to model the stiffness in the translational DOFs, and uniaxial
viscous damper materials were used to model the vertical damping within the suspension
system. Stiffness and damping coefficients for the suspension system of the power car,
extreme passenger car, and intermediate passenger car were defined as given in the
reference study [19]. The parallel material command was used to combine the stiffness
and damping material in the z-direction to a single material. These materials were then
used as the material parameters for the two-node link elements. The i-nodes shown are the
bogie nodes and the j-nodes are the axle wheel nodes. The materials defined were applied
in their respective directions and the orient command was used to manually instruct
OpenSees of the element vector components. Since the primary suspension system only
applies stiffness in the three translational DOFs, the equalDOF command was used to
constrain the remaining DOFs between the bogie and axle nodes. Figure B-10 and Figure
B-11 demonstrates how the primary suspension system of the power cars were modeled.

Similar process was performed for the secondary suspension systems; however, damping
for the z-rotational DOF was also applied in addition to any translational damping (Figure
B-12). As shown in the train model schematic in cross-section of the train model in Figure
3-4, the secondary suspension system has three layers: left, middle, and right. The left and
 36

right layers supply stiffness and damping in the translational DOFs and the middle layer
supplies damping in the z-rotational DOF. Due to this DOF not having any stiffness, the
DOF must be constrained for the stability of the model. However, if the displacement
between the two-nodes constituting the middle layer of the secondary suspension system
were constrained using the equalDOF command, the z-rotational damping would not
activate due to the lack of displacement (x). Therefore, a relatively small stiffness value (1
kN/m) was applied in the z-rotational DOF to allow for the activation of the damping, and
the rest of the DOFs were constrained using the equalDOF command (Figure B-13).

3.3.2.5 Train System Masses

The train masses were modeled using the values given in the reference study [19], included
in Table 3-3. Since the extreme passenger car for the KTX-Sancheon is not motorized,
unlike the KTX-I in the reference study, the translational mass and inertial mass values for
the intermediate passenger car were used for the extreme passenger car as well. The masses
were defined at the center-of-mass nodes for each car-body and bogie. The masses for the
wheels are defined at every wheel node. Figure B-14 through Figure B-17 demonstrate
how the car-body, bogie, and axle masses were defined in OpenSees. The inertial masses
were used to define the rotational nodal masses.

Track System Model

The track system comprises of rails, track plates, base plates, and the connection layers in
between these components. The rails, track plates, and base plates were modeled as
elasticBeamColumn elements and the connection layers were modeled as zeroLength
elements. The rails, track plates, and base plates were discretized into equal intervals of
3.195 m and the connection layers were modeled at the end nodes of each interval. The
train-track interaction was modeled by including and connecting the train wheel nodes as
a member of the series of nodes creating the rail elements. This directly transfers the train
loads to the track system, which then transfers the loads down to the bridge system through
rigid arms connecting the track system to the bridge girder. The bridge girder was also
discretized into equal increments of 3.195 m, which allowed for the track-bridge interaction
to occur at an equal distribution along the entirety of the bridge length. A general schematic
 37

of the track system is shown in Figure 3-5 and Figure 3-6. The steps taken to model the
track system nodes, elements, and masses are further discussed in detail in this chapter.

3.3.3.1 Track System Elastic Elements

The rails, track plate, and base plate were modeled as linear elastic beam-column elements
because they are all designed to remain elastic as capacity protected elements. The location
of the track plate and base plate nodes are the same, and rail nodes are located to the right
and left of the track plate/base plate nodes by half the transverse train wheel spacing,
defined earlier as R1 and R2 for track 1 and L1 and L2 for track 2, respectively. Figures B-
18, B-19, and B-20 in Appendix B show sample node setup for rail, base plate, and track
plate of one of the tracks, respectively. The elements were assigned cross section
parameters as given in the study by Li et al., [22]. The rail, track plate, and base plate
elements span the entirety of the bridge length. The process of modeling rail, track plate,
and base plate elements are shown in Figure B-21, B-22, and Figure B-23, respectively.

To connect the train system to the track system, wheel nodes of the train were connected
to neighboring rail nodes using the same linear elastic beam-column elements used for the
rails. Since the train was placed on track 1 consisting of rails 1 and 2, the wheel nodes were
modeled at the same y and z-coordinates as the rail nodes. The sequential order of the wheel
nodes and rail nodes were organized offline and defined in OpenSees accordingly. This
was done under the assumption that the train wheels are always in contact with the rails, as
researched in Chapter 2 to be a common assumption.

3.3.3.2 Track System Connection Layers

Zero-length elements were used to simulate the nonlinear behavior of the sliding layer, CA
layer, shear reinforcement, lateral blocking, and fasteners. The nonlinear material behavior
was assigned to the zero-length elements using the Steel01 material in OpenSees. The yield
strengths were assigned as given by Li et al., [22] and the initial elastic tangent was found
by a quotient of the yield strength and relative displacement. The strain hardening ratio was
assigned a value of zero to mirror the perfectly elastic-plastic behavioral graph from the
reference study. Figure 3-7 first shows the generalized elastic-plastic behavior along with
the parameters of the different zero-length connection elements in the track-bridge system
 38

as adopted from Li et al., [22]. Next, dedicated plots were generated to demonstrate the
behavior of five of those connection component in track systems and shown in Figure 3-8.
Fasteners and lateral blocking were modeled between the duplicate rail nodes as
demonstrated in Figure B-24 and Figure B-25 in Appendix B, respectively. The CA mortar
layer was modeled between the track plate and base plate (Figure B-26), and the sliding
layer was modeled between the base plate and rigid arm connecting the track system to the
bridge girder (Figure B-27). Sample shear reinfrocement definition is also shown in Figure
B-28. The fasteners, CA mortar layer, and sliding layer allow for longitudinal slippery
relative to the bridge length. Multi-point constraints were used to constrain the remaining
DOFs of the connection layer nodes that stiffness was not applied to through zero-length
elements. For example, stiffness was applied in the longitudinal direction for the sliding
layer to allow for movement based on the behavior of the material, so the equalDOF
command was used to constrain the remaining 5 DOFs (Figure B-29).

3.3.3.3 Track System Rigid Connections

Rigid elements were used in the track system to connect the track plate nodes to the rails.
Specifically, the rigid arms branch out from each track plate node to duplicate rail nodes
that were not used to model the rail elements. The rigid section properties to model rigid
arms out of elastic beam-column elements were kept the same as what was used for the
train system rigid bodies. Rigid arms were modeled at 3.195 m intervals for both tracks 1
and 2, which is the same intervals as the track system nodes. The location of the rigid arms
can be seen in Figure 3-9.

3.3.3.4 Track System Masses

The masses for the rails, track plates, and base plates were assumed using approximate
densities of steel and concrete. The steel rails were assumed to have a density of 7,700
kg/m3, and the concrete track plate and base plate were assumed to have a density of 2,400
kg/m3. These are very generic values and accurate densities should be utilized to accurately
model the dynamic performance of HSR systems because the mass matrix is one of the key
components of solving the equation-of-motion of the model. Mass per node was found by
dividing the product of the given cross-sectional area and the length of the bridge by the
 39

number of nodes constituting the entire length (110 nodes). General mass moment of inertia
equations for rectangular sections were used to solve for the moment of inertia in the three
rotational DOFs. The masses used for the track system in this study is shown in Table 3-3.
The mass per node was used for the nodal mass value in the translational DOFs and the
inertial masses were used for the rotational DOFs (Figure B-30).

Bridge System Model

The bridge system comprises of girders, bearings, pier columns, and footings. Girders were
modeled as elastic beam-column elements, and bearings were modeled as zero-length
elements. Pier columns were modeled as displacement based elastoplastic fiber elements
and columns footings were modeled as rigid elements. Rigid arms were used to connect
each bridge component to one another as illustrated in the track-bridge system schematic
shown in Figure 3-9.

3.3.4.1 Bridge System Girder

The prestressed concrete box-girder bridge is designed to be elastic, i.e. capacity protected
component for seismic considerations, so linear elastic beam-column elements with
equivalent section characteristics were used to model the superstructure. Each span was
discretized into 10 equivalent lengths of 3.195 m by creating 11 nodes per girder span.
Figure B-31 demonstrates how the nodes for the first two bride girder spans were defined.
A 0.05 m gap was created between each bridge girder span to simulate the isolated
movement allowed to each girder span by four steel bearings, two fixed and two sliding.
The cross-sectional area, Young’s modulus, shear modulus, torsional moment of inertia of
the cross-section, and second moment of area about the local z and y-axis were assigned
the values given by Li et al., [22] and shown in Table 3-4. To simulate the process of bridge
design, the Young’s Modulus was decreased from 3.45e7 kN/m2 to 2.45e7 kN/m2 and the
moment of inertia values were reduced by 30% to account for the reduction in concrete
stiffness due to cracking. The process of modeling the first span of the bridge girder is
shown in Figure B-32. For the first girder span, Node 90001 to Node 90011 were modeled
in series with the elastic beam-column element, using predefined cross-sectional
 40

parameters. The distance between Node 90011 and Node 90012 demonstrate the gap
between girders, so these nodes are not connected using the elastic element.

3.3.4.2 Bridge System Bearings

The spherical steel bearings were modeled using zero-length elements. To use zero-length
elements, the OpenSees user must create two nodes with the same coordinates, hence the
zero-length. Since the bearings are located at the ends of each bridge span, two-sets of
nodes were created accordingly. The fixed and sliding bearings were assumed to be 4 m
apart, based on the box-girder dimensions, in the direction transverse to the bridge at the
top of the 13.5 m tall pier columns. The nodes for the bearings supporting the first bridge
span are shown in Figure B-33. One set of the bearing nodes were used to connect the
bearing system to the bridge girder, and the other set of nodes were used to connect the
bearings to the top of the pier columns, both through rigid arms.

The OpenSees material command Steel01 was used to define the bilinear behavior of the
steel bearings within the zero-length elements. The required parameters for the zero-length
elements for the steel bearings are shown in Figure 3-7. The yield strength was defined as
given by the reference study in Figure 3-7 with a value of 5000 kN for the fixed bearing
and 470 kN for the sliding bearing, and the elastic tangent or slope of the elastic region was
found by a quotient of the yield strength and relative displacement also given in Figure 3-
7. As previously mentioned, the strain-hardening ratio was set as 0 and the uniaxial material
was applied into directions 1 and 2 to apply stiffness in the lateral translational DOFs. The
behavior of the fixed and sliding bearing is shown in Figure 3-10. The fixed and sliding
bearings were alternated as shown in Figure 3-11 to mirror the design of the actual bridge.

As previously mentioned, stiffness was only applied in the longitudinal and transverse
DOFs, so the vertical DOF and the three rotational DOFs were constrained for structural
stability. The high stiffness value for the fixed bearing idealizes the resistance it provides
to constrain movement and the low value for the sliding bearing idealizes the slight
resistance it provides despite allowing movement. The fixed and sliding bearings modeled
to support the first span of the bridge are shown as examples in Appendix B in Figure B-
34 and Figure B-35, respectively. For this study, the equalDOF command was used to
 41

constrain the rest of the DOFs and make sure duplicate bearing nodes will have the same
movement (Figure B-36).

3.3.4.3 Bridge System Pier Columns

Materials for the pier column cross-section were defined using uniaxial materials available
within OpenSees and material strengths were input as parameters. The core concrete, cover
concrete, and reinforcing steel strength assumptions were adopted from a sample code
provided by the OpenSeesWiki, [32] since the design guideline for the selected prototype
HSR bridge used herein did not provide sufficient information on specific material
specifications for the bridge columns. The assumptions used for the concrete and
reinforcing steel properties and input parameters are shown in Figure B-37. The cover and
core concrete were modeled using the Concrete02 material and the longitudinal
reinforcement was modeled using the Steel02 material in OpenSees; a typical modeling
practice for bridge elements that has been adopted in many of the reviewed studies such as
the one Li and Conte, [23]. For the Steel02 command, the R0, cR1, and cR2 parameters
were defined as 15, 0.925, and 0.15, respectively, as recommended for general reinforcing
bar by the OpenSeesWiki.

The pier cross-section was created using the fiber section command (Figure B-38). The
cover and core concrete were defined within the section using the patch rect command to
generate fibers over a rectangular cross-sectional area. The reinforcing steel was defined
using layer straight commands to generate fibers along a straight line for the four sides of
the rectangular cross-section. The material tag (matTag) for these commands reflect what
was defined for the cover, core, and reinforcing steel materials.

The geometry of cross-section design, as well as the coordinates required in the command
parameters to create the cross-section were predefined as shown in Figure B-39. A
reinforcement ratio of 1.30% was assumed for the cross-section and this led to a
preliminary design of 176- #11 bars, split into 60 bars on the long face and 28 bars on the
short face of the cross-section. Transverse reinforcement was assumed as #4 bars and a
clear cover of 0.04 m was also assumed. The design used for the cross-section does not
reflect the actual design of the pier columns, but since the details are unknown, a general
 42

design was done based on engineering judgement. The design specified in the section Fiber
command was then aggregated into a uniaxial elastic material section using the section
Aggregator command to create a single section force-deformation model. The torsion
force-deformation (T) was selected as the force-deformation quantity parameter to be
modeled by the section object.

The rectangular bridge pier columns were modeled as a series of four three-dimensional
displacement based elastoplastic fiber elements using the dispBeamColumn command with
the nonlinear fiber cross-section that was defined. Each pier was constituted by five nodes
with equal 3.375 m intervals with five integration points each (Figure B-40). Integration of
fiber characteristics over the pier cross-section allowed for the obtainment of nonlinear
section characteristics. The process of modeling the first pier column is shown in Figure
B-41.

3.3.4.4 Bridge System Column Footings and Soil

Column footing dimensions of the prototype bridge selected were not explicitly noted in
the reference study, so generic dimensions of 4 m for the depth and 11 m for the width
were assumed. The nodes were defined at -2 m to create nodes at the centroid of the
footings. The column footings were modeled as rigid elements via the same method for all
other rigid elements to connect the column base nodes to the footing nodes. Figure B-42 in
Appendix B shows a sample for footing nodes and ground.

Due to the focus of the study being the dynamic interactions between the train-track-bridge
systems, a simplistic method was used to model the interaction between the bridge and soil.
Since California is projected to be the home of the largest HSR system in the United States,
soil spring constants from a study by Abbasi, [1] were used to simulate the general soil
properties of California. Since multi-column box-girder bridges in California typically
have the pinned connection details in the foundation, there are no rotational stiffness
defined at the column footings. Abbasi, [1] considered a wide range of soil profiles and
foundation systems over the state of California and determined the stiffness of translational
springs to be 115 MN/m. However, adjustments were made to accommodate the single
column bent design of the bridge piers. Single column bents typically utilize fixed-base
 43

connections to provide stability to the cantilevered system. Accordingly, the footing nodes
were fixed in the non-translational DOFs and the foundation nodes were fixed in all 6 DOFs
to create a base for the entire model (Figure B-43).

The structure-soil interaction was simplified in-part due to the lack of information
regarding the soil spring constants required to model the pile-soil interaction and the focus
of the study being the train-track-structure interaction. If this information is available, a
sophisticated soil-structure interaction model is recommended by explicitly modeling the
piles as displacement based elastoplastic fiber elements, as done by Li et al., [22] and Li
and Conte, [23]. The process of modeling the column footings and the interaction with the
soil for the model in place is shown in Figure B-43 and Figure B-44.

3.3.4.5 Bridge System Rigid Connections

Rigid elements are used in the bridge system to connect the bridge girder, bearing, pier
column, and footing to one another. For the model in-place, the track system is connected
to the bridge girder through two diagonal arms at an interval of 3.195 m, along the entire
bridge length. Additionally, two diagonal rigid arms connected the bridge girder to the steel
bearings isolating the bridge girder from the pier columns, meaning the two nodes defining
the ends of each bridge girder span had a total of four rigid arms. The bearings are
connected to the pier columns through two horizontal arms in the y-direction at the top of
the pier columns, and the column footings are idealized as a rigid arm. The location of rigid
arms is shown in the track-bridge system schematic in Figure 3-9. The same rigid section
properties were used as the rigid arms in the train and track system. Examples of all the
rigid elastic beam-column elements used in the bridge system are shown in Figure B-45
through Figure B-48.

3.3.4.6 Bridge System Masses

For the dynamic equation of motion, masses for the concrete deck, pier column, and footing
were assumed using a standard density of 2,400 kg/m3. General mass moment of inertia
equations for rectangular sections were used to solve for the very approximate mass
moment of inertia in the three rotational DOFs. The masses of the bridge girder were
distributed along the 10 spans, consisting of 11 nodes each. The masses of each pier column
 44

were distributed along the five nodes constituting the entire column. The masses were
applied at the center-of-mass node for each footing. The masses for the bridge system in
this study is shown in Table 3-3 as previously mentioned. Moroever, the process of
applying the masses for sample different bridge components, i.e. box-girder, columns, and
footings, are shown in Figure B-49, Figure B-50, and Figure B-51, respectively.
 45

Table 3-1. Prototype HSR Model Element and Material.

 46

Table 3-2. Dynamic Characteristics of Train Model.

Extreme Intermediate
Property Power Car
Passenger Car Passenger Car
Mass of car-body (kg) [M] 54960 26000 26000
Primary sprung mass per bogie
(kg) [mt] 2420 2514 3050

Unsprung mass per axle (kg) [ma] 2050 2050 2000

Primary stiffness per axle box 40000, 9000, 40000, 9000, 55000, 11000,
(kN/m) [kx, ky, kz] 1250 1250 800
Secondary stiffness per bogie
side (kN/m) [kax, kay, kaz] 303, 303, 1270 100, 150, 370 100, 170, 303

Primary damper per axle box

(kN-s/m) 0, 0, 10, 4230 0, 0, 10, 4230 0, 0, 6, 240
[cx, cy, cz, cϕ]
Secondary damper per bogie
side (kN/m) [cax, cay, caz] 0, 100, 20 0, 30, 20 0, 0, 0

Moment of inertia of car-body 59.4, 1132.8, 33.94, 971.81, 33.94, 971.81,

(Mg-m2) [Ix, Iy, Iz] 1112.9 971.81 971.81
Moment of inertia of bogie 1.645, 2.593, 2.07, 3.26, 2.03, 3.20,
(Mg-m2) [Itx, Ity, Itz] 3.068 3.86 3.79
Moment of inertia of wheel 1.03, 0.0008, 1.03, 0.0008, 1.03, 0.0008,
(Mg-m2) [Iax, Iay, Iaz] 1.03 1.03 1.03
Length of car-body (m)
[Lp, Lm, Lc] 14.0 18.7 18.7

Height of centroid (m)

[h, hb] 1.72, 0.56 1.627, 0.56 1.627, 0.56

Height from secondary

suspension arm to centroid (m) 0.605 0.420 0.508
[hp, hm, hc]
 47

Table 3-3. Masses for Track-Bridge System.

Moment of Moment of Moment of

Mass
Inertia 1 Inertia 2 Inertia 3
(Mg/node)
(Mg-m2) (Mg-m2) (Mg-m2)
Girder 63.7359 159.1817 61.1692 189.1868
Column 7.9940 27.2587 11.7515 23.8342
Footing 629.7408 7859.6900 7859.6900 14122.9870
Rail 0.1693 0.0025 0.1459 0.1446
Track
3.5878 1.9561 3.0640 4.9961
Plate
Base
3.9466 2.8739 3.3691 6.2193
Plate

Table 3-4. Section parameters of elastic beam elements in track-bridge system as adopted
from Li et al., [22].
 48

Figure 3-1. Photo of KTX-Sancheon [6].

Figure 3-2. Schematic of the prototype bridge: a) Elevation layout of high-speed railway
bridge/cm, b) Schematic sketch of track and girder structure [22].
 49

Figure 3-3. Schematic of the prototype bridge typical cross-section of track and girder
structure [22].

Figure 3-4. Schematic drawing for the numerical modeling of train system (Top: Cross-
section, Bot: Elevation).
 50

Figure 3-5. Schematic of track system.

Figure 3-6. Schematic of track-bridge system.

Figure 3-7. Parameters of zero-length connection elements in the track-bridge system as

adopted from Li et al., [22].
 51

Figure 3-8. Force-deformation behavior of track system connection layers: (a) Fastener,
(b) CA mortar, (c) Shear reinforcement, (d) Sliding layer, and (e) Lateral blocking.
 52

Figure 3-9. Schematic of track-bridge system.

 53

Figure 3-10. Force-deformation behavior of bridge bearings: (a) Fixed bearing, (b)
Sliding bearing.

Figure 3-11. Finite element model of bridge.

 54

4 DEMONSTRATION OF GRAVITY, MODAL, AND SEISMIC ANALYSIS OF

HSR BRIDGE SYSTEM

Overview

In OpenSees, an analysis is performed through the aggregation of component objects. The

component objects define the type of analysis that is performed on the model and consists
of the following: constraints handler, DOF numberer, integrator, solution algorithm,
system-of-equation constructor and solver, and convergence test. This chapter will discuss
the component objects defined for the gravity load static analysis and the seismic load
dynamic analysis, as well as how the modal analysis was performed. Static and dynamic
analysis were performed for a load case without the train and an example load case with
the train. The recorded data was analyzed to verify and observe the responses within the
HSR bridge. This chapter serves to demonstrate the selection of analyses component
objects for the prototype HSR model and present example studies that can be performed to
understand the behavior of the model under various loading.

Gravity Load Analysis

Gravity Load Analysis Setup

To perform a linear or nonlinear static gravity load analysis, loads must be applied to
represent the self-weight of each structural component. Masses do not have to be defined
for static analysis because inertial and damping effects are neglected. The masses defined
in Section 3 were instead converted into forces (kN) and applied as vertical loads at the
same nodes as the masses. This was done through the pattern plain command which allows
the user to apply loads to specific nodes and elements. Train system car-body, bogie, and
axle wheel and bridge foundation dead loads were applied at their center-of-mass nodes,
and track-bridge system rail, track plate, base plate, bridge girder, and pier column dead
loads were distributed to each node formulating their respective elements. The train system,
track system, and bridge system had a total weight of 3,989 kN, 16,992 kN, and 184,230
kN, respectively, with a total static weight of 205,211 kN. The static weights of the train-
track-bridge system were used to verify the load transfer within the HSR model through
 55

comparison with column base reactions. As previously mentioned and shown in Chapter
3, the step-by-step type of model and analysis definition demonstration is provided in
Appendix B. As part of Appendix of B, Figure B-52 through Figure B-60 demonstrate the
process of applying dead loads to each component of the HSR bridge system.

The constraints command handles how the constraint equations are enforced in the
analysis. Constraint equations enforce a specified value for a DOF, or a relationship
between DOFs [31]. The type of constraint selected should depend on the type of
constraints implemented in the user’s model, homogeneous single-point constraints or non-
homogenous single-point constraints. For this study, multi-point constraints were used
(equalDOF), so the Transformation command was used to enforce the constraints using
the transformation method.

The numberer command determines the mapping between equation numbers and DOF,
and how DOF are numbered. The use of the plain numberer is recommended mostly for
very small problems and for the sparse matrix solvers which provide their own numbering
scheme. For this study, the RCM option was used for the numberer in the case of this large-
scale system model. The RCM (Reverse Cuthill-McKee) algorithm optimizes node
numbering to reduce bandwidth using a numbering graph, and outputs a warning when the
structure is disconnected. The system command constructs the linear system-of-equations
and solver objects to store and solve the linear system-of-equations (K.u = R), and each
solver is tailored to a specific matrix topology. The UmfPack command was used to
construct a large sparse system-of-equations object which will be factored and solved
during the analysis using the UmfPack solver.

To perform nonlinear analysis, the user must define how OpenSees will deem whether the
model has converged to the correct solution. The test command is used to select
convergence test to determine if convergence has been achieved at the end of an iteration
step. The command parameters allow the user to define the convergence tolerance, the
maximum number of iterations that will be performed before OpenSees returns “failure to
converge”, and a flag to instruct OpenSees on how to print information on convergence.
The NormDispIncr test type selected in this study uses the norm of the left-hand side
 56

solution vector of the matrix equation to determine if convergence has been reached. The
test returns positive for convergence if the displacement increment in the linear system-of-
equation is less than the specified tolerance. For this model, a tolerance of 1.0e-6 and a
maximum number of iterations of 100 was deemed reasonable. A flag value of 1 was
selected to instruct OpenSees to print convergence information on each step to monitor
whether the model was operating correctly, but this does not affect the actual analysis.

The next step is to define a solution algorithm to instruct OpenSees on the sequence of
steps to take to solve the nonlinear equation. The Newton command was used to solve the
nonlinear residual equation using the Newton-Raphson algorithm, which is the most widely
used robust method for solving nonlinear algebraic equations [31]. The integrator
command is used to determine the predictive time step for the analysis, specify the tangent
matrix and residual vector at any iteration, and determine the corrective time step based on
the displacement increment. The LoadControl integrator type was selected and an initial
load-increment factor (pseudo-time step) was defined as 0.1 to apply a tenth of the dead
loads defined earlier at each step. The gravity load was applied through 10 loading steps to
avoid convergence issues that might have happened if the large gravity loads is applied in
one step.

Finally, the analysis command was used to specify a static analysis and the analyze
command was used with the number of load steps parameter, to slowly apply the
gravitational loads in 10 steps. The loadConst command was used to instruct OpenSees to
maintain constant gravity loads and reset the time to zero before the transient analysis. This
entire process of setting up the gravity analysis parameters then performing the analysis is
demonstrated in Figure B-61 and Figure B-62, respectively.

Gravity Load Analysis Results

Sample studies were performed to demonstrate behavioral analysis that can be performed
using the gravity analysis results obtained from the model. In high seismic areas, the main
design considerations for HSR bridges are usually dictated by resonance and seismic
forces. Nonetheless, the static analysis was performed as a precursor to the dynamic
analysis and for verification of load transfer within the structure. Several loading scenarios
 57

could be considered for analyzing the HSR bridge system with respect to train position on
the bridge as the train approaches and crosses the bridge. A list of 16 different scenarios
that could be considered for the system in hand is provided in Table 4-1 as an example.
Only few selected cases are included in this research, but the list is still provided to
highlight and indicate how train position over the bridge can be represented. For gravity
load analysis, two load cases from Table 4-1 were considered for the demonstration
purposes as sample studies: (1) Load Case 1 where the train is not on the bridge, and Load
Case 8 where the train is loading spans 2 through 7. The load cases are illustrated in Figure
4-1. For Load Case 1, the train model and train model gravity loads were completely
omitted, leaving just the track and bridge model, along with their respective gravity loads.
For Load Case 8, the very first train wheel was determined to be located 30.815 m along
the bridge, the train system was connected to the track system accordingly.

The first exercise performed with the static analysis results was the verification of load
transfer within the HSR system. Since loads were applied within the track and bridge
subsystems, an error within either subsystem could cause the loads to incorrectly transfer
through the structure. To perform this exercise, node recorders were used to extract the
reactions at the column bases under Load Case 1 without the train and Load Case 8 with
the train. The column base reactions in the vertical direction were tabulated in Table 4-2
for both load cases, and the distribution and sum of the reactions were observed to check
for any red flags regarding the incorrect transfer loads. The sum of column base reactions
in both load cases were found to be equal to the total loads applied for each load case,
described in Section 4.2.1, which indicates all the loads were able to transfer to the column
bases. The distribution of the interior column base reactions for Load Case 1, to the left
and right of the center pier column #6, was symmetrical. The exterior columns had a
difference of 33 kN which is not exceptionally large considering the scale of the reactions.
For Load Case 8, an increase in the reactions for columns #2 through #8 were observed.
This behavior verifies that the train loading over bridge girder spans #2 through #7 was
properly supported by the pier columns supporting those respective spans. The rest of the
pier columns maintained the same reactions as Load Case 1 since they were not affected
by the static loading of the train.
 58

As a verification of static behavior of the model, vertical displacements of the bridge box-
girder were analyzed for both load cases. Node recorders were used to output vertical nodal
displacements along the entire bridge length. The recorded values were post-processed
using Matlab to organize the data and plot a graph demonstrating the deformed shape of
the bridge girder under gravity loads. An exaggerated view of the deflection in each bridge
span under the loading scenarios of Load Case 1 and Load Case 8 is shown in Figure 4-2.
The bridge span displacements were nearly identical among all the spans for Load Case 1.
A maximum vertical displacement for the bridge was recorded at -0.408 mm at the center
node of each span. For Load Case 8, an increase in vertical displacements for the spans
loaded by the train was visibly apparent in the graph. Larger displacements were recorded
at span 2 and span 7, which is due to these spans supporting the fore and rear power cars
of the KTX-Sancheon model. The maximum vertical displacement for the bridge under
Load Case 8 was recorded at -0.452 m at spans #2 and #7. As seen in Figure 4-2, the mass
of the power cars is greater than two times that of the passenger cars, so the displacement
trend observed from Load Case 8 were deemed reasonable.

Modal Analysis

Analyzing modal characteristics is imperative to designing HSR bridges for seismic

stability and riding comfort by minimizing resonance within the structure. Modal analysis
of the bridge system was performed by using the eigen command which uses the overall
mass and stiffness of the structure to determine the various vibration frequencies (or
periods) along with mode shapes. The eigen command performs a generalized eigenvalue
problem to determine a user specified number of eigenvalues and eigenvectors. For this
study, the parameter for number of eigenvalues (λ) was defined as 10, for the first 10 modes
which were then used to solve for the periods (T) of the structural model (Figure A-20).
An open vector for the periods and the value for pi (π) were defined. The tcl commands
foreach and lappend were used to instruct OpenSees to take each eigenvalue from the
lambda index and solve for periods using equations (1) and (2) below. An output file was
then specified and a tcl command, open, was used to open the output file and the foreach
and puts commands were used to record the periods that were solved. The output file was
 59

then closed using the tcl command, close, to allow OpenSees to continue with the rest of
the analyses.

𝜔 = √𝜆 (1)
2𝜋
𝑇= (2)
𝜔
The modal analysis process covered in this section is demonstrated for a step-by-step
procedure in Appendix B in Figure B-63. The first 10 periods obtained for the bridge
system under the two sample load cases, i.e. without the train and with the train covering
spans 2 through 7 of the bridge, are tabulated in Table 4-3. The values shown in the table
show that the first two modes are likely the dominant bridge modes in the transverse and
longitudinal direction that are not sensitive to the train loading. Higher modes varied
slightly which is attributed to the added train mass and specific train-track-bridge system
vibration modes.

Seismic Load Analysis

Seismic Load Analysis Set Up

To start off the set up for the seismic analysis, structural damping must be applied first to
model the inherent damping and energy dissipation mechanisms within the structure. The
Rayleigh command was used to apply classical Rayleigh damping, i.e. viscous damping
proportional to a linear combination of mass and stiffness, to all previously-defined
elements and nodes in the structural model as demonstrated in Figure B-64. Due to the
nature of the bridge system and model, the natural frequencies of the first and sixth modes
were selected to solve the alpha and beta parameters for the Rayleigh command as defined
from the OpenSees syntax shown in Appendix A in Figure A-21. A typical damping ratio
of 2% was used for this study.

The set up for the seismic load analysis is overall similar to the gravity load analysis, with
some differences to accommodate the transition from static analysis to transient analysis
as depicted in Figure B-65. For the constraint handler, the transformation method was used
again due to the use of multi-point constraints in the model. The RCM algorithm was also
used as the DOF numberer to optimize node numbering and reduce bandwidth, and the
 60

Newton-Raphson method was used to advance the analysis to the next time step. The
convergence test type was changed to the energy increment test which uses the dot product
of the solution vector and norm of the right-hand side of the matrix equation to determine
if convergence has been reached. The test returns positive for convergence if one half of
the inner-product of the unbalanced load and displacement increments at the current
iteration is less than the specified tolerance. The tolerance was decreased to 1.0e-8 to
increase accuracy of the analysis and the maximum number of iterations was increased to
1000 to raise the chances of the model correctly converging. The OpenSees Manual does
not recommend a type of convergence test for static or dynamic analysis, but this is one
area where informed user input is needed to properly obtain correct convergence. The same
linear equation solver, UmfPack, was used to store and solve the system-of-equations in
the analysis.

For the transient analysis, a numerical integrator is needed to solve the dynamic equation
of motion that is needed to account for inertial and damping effects. For this study, the
classical Newmark method was used to perform the numerical integration. The Newmark
method is a two-parameter time-stepping method developed by Nathan M. Newmark. The
gamma (g) and beta (b) parameter values depend on whether the average acceleration

method or linear acceleration method is selected. For this study, the average acceleration
method was selected because it is unconditionally stable, i.e. independent of the analysis
time step, and the gamma = 0.5 and beta = 0.25 values were defined accordingly. Dynamic
analyses could use any of several explicit or implicit integrator types as per the list provided
in the OpenSeesWiki or OpenSees Manual, and users could select from the available
methods based on the application or so. The analysis command was then used to instruct
OpenSees to conduct a transient analysis with the parameters defined above.

Once the specifics of the transient analysis were defined, the ground motions to be used as
the transient loads were defined. The ground motion selected for the sample transient
analysis is from the 1994 Northridge earthquake recorded at the LA-Sepulveda VA
Hospital. The acceleration time-history was retrieved from the Pacific Earthquake
Engineering Research Center (PEER) ground motion database provided by the University
 61

of California, Berkeley. The downloaded acceleration time-history file was placed in the
same OpenSees bin folder as the tcl file of the structural model to allow the code to call
out the ground motion. The ReadSMDFile, available on the OpenSeesManual, [31] and
OpenSeesWiki, [32] online, was sourced to convert the PEER ground motion to a format
readable by OpenSees. The sourced file removes the header text in the PEER ground
motion file and converts the file extension from AT2 to g3. This process can be seen in
Figure B-66. The analysis time-step (DT) and total number of steps (Nstep) were defined as
0.005 seconds and 9557, respectively, with maximum duration of the ground motion being
47.785 seconds.

Using the converted acceleration time-history file and the ground motion parameters
defined, the timeSeries path command was used to define the time-series information for
both ground motions (see Figure A-22 in Appendix A for OpenSees command details). A
gravitational acceleration value of 9.81 𝑚⁄𝑠 2 was applied as the factor to retrieve the
acceleration time-history values from the multiples of [g] format. The factor can be further
increased if amplification of the ground motion is of interest. Unique load tags were created
for each excitation, and the ground motions were then applied to the model using the
UniformExcitation pattern command. The parameters required in the UniformExcitation
pattern command are shown in Figure A-23. The respective unique pattern tag
(patternTag), ground motion direction, and time-series information for each excitation
defined earlier were used in the command. The process of applying the ground motion in
both directions is shown in Figure B-67.

After completing the definition of dynamic analysis parameters and the transient loads, the
analyze command was used to instruct OpenSees to perform the dynamic analysis with the
time-stepping parameters previously defined for the ground motion. Figure B-68
demonstrates a loop function created to run the dynamic analysis and engage additional
algorithms and convergence test types if the initial dynamic analysis parameters are
incapable of converging the model. The analyze command set to return “ok = 0” if the
analysis at a time-step successfully converged to a solution. The loop command is set to
start if “ok != 0”, which means that the “ok” value is not 0. While the current time-step is
 62

less than the maximum duration of the ground motion, the loop attempts to converge the
model using a norm displacement increment convergence test and the Newton-Raphson
algorithm with initial stiffness iterations, the Broyden algorithm, and the Newton-Raphson
algorithm with line search, in order.

Seismic Load Analysis Results

After the gravity load analysis was completed and damping was applied, dynamic analysis
of the model was performed. The same two load cases were considered for the dynamic
analysis: (1) Load Case 1 where the train is not on the bridge, and (2) Load Case 8 where
the train is loading spans 2 through 7. Several sample exercises were conducted using the
results from the two load cases to analyze the maximum forces and moments experienced
by the prototype HSR bridge and observe the sensitivity of the results with and without
train loading. This section aims to demonstrate the variety of studies that can be performed
using the data output by OpenSees and the sample results presented shall not be taken as a
reference for design.

As an extension to the exercise done for the static analysis, the vertical displacements of
the bridge girders under seismic loading were plotted for both load cases. The maximum
vertical displacement was recorded as -0.657 mm at girder spans #1 and #10 for Load Case
1. The bridge girder displacements at the end of the static analysis (start of dynamic
analysis) and at a time-step of 4.185 seconds during the dynamic analysis, when the
maximum displacement was recorded for Load Case 1, were plotted in Figure 4-3 as
sample. For Load Case 8, the maximum vertical displacement of -0.636 mm was recorded
at girder spans #2 and #7. The bridge girder displacements at the end of the static analysis
and at a time-step of 3.915 seconds during the dynamic analysis, where the maximum
displacement for Load Case 8 was observed, were also plotted as samples and shown in
Figure 4-4. The vertical displacement trends for both load cases under seismic loading were
found to be very similar to that of the static analysis. This behavior is understandable
because only the two horizontal components of the ground motion were considered (which
excites the lateral directions of the bridge) and the vertical excitation component was
neglected. The box-girder is also designed to be a capacity protected element, meaning
 63

inelastic deformation is not expected to be caused by the ground motions. The minor
increase in displacements are most likely caused by rotations at each girder-span end above
the pier. It is noted that the box-girder is not continuously supported over the pier and the
gap between each two successive girder spans allow for some minor rotation.

The second exercise conducted was the observation of transverse bridge displacement
trends, which are crucial for seismic performance assessment. To observe the
displacements experienced by the bridge during the ground motion, the transverse
displacements were analyzed at the time-step at which the bridge experienced the largest
transverse displacement between both load cases and the final time-step of the ground
motion to see whether any residual displacements were observed. The maximum
displacement during the ground motion between both load cases occurred at a time-step of
4.735 seconds for Load Case 1, with an absolute value of 291.7 mm. The maximum
transverse displacement recorded for Load Case 2 was 282.2 mm at a time-step of 4.750
seconds. The transverse displacements at the end of the ground motion were also analyzed
to observe the residual displacements caused by the nonlinear effects of dynamic loading,
and plastic damage, if any. The prototype HSR bridge under Load Case 1 had a residual
displacement of 111.2 mm and Load Case 8 had a residual displacement of 116.6 mm. The
displaced shapes of the bridge for the selected time steps mentioned above is shown in
Figure 4-5 for Load Case 1 and Figure 4-6 for Load Case 8.

Similar to the previous displacement exercises, time-histories of pier column and girder
end displacements were plotted to better understand the bridge behavior with and without
train loading. The time-history graphs compare the relative drift between girder ends and
the supporting columns and indicate whether residual displacements were observed due to
nonlinear/plastic deformations induced by the cyclic loading of the ground motions. Four
pier columns and their respective girder ends were considered in the shown sample time-
history analysis: #2, #6, #8, and #11, to observe the magnitudes of drift along the bridge.
Pier column displacements were recorded by outputting the transverse and longitudinal
displacements of the top nodes and their histories were plotted through the total duration
of the ground motion. Similarly, the translational displacements of the nodes defining the
 64

ends of each girder span were recorded and plotted. The displacement time-histories from
the four piers are shown in Figure 4-7 through Figure 4-10 for Load Case 1 and Figure 4-
11 through Figure 4-14 for Load Case 8. The figures include two sub-plots, which are
designated as “a” and “b” to represent the displacement trends in the longitudinal and
transverse direction, respectively. Based on the longitudinal displacement trends, the
shapes are nearly identical between both load cases with Load Case 8 showing slightly
larger drift between the column and girder for columns #6 and #8. From the displacement
time-histories for the transverse direction shown in Figure 4-8(b), Figure 4-9(b), Figure 4-
12(b), and Figure 4-13(b), all four columns showed similar trends within each load case.
Comparing the displacement trend between the load cases, Load Case 1 had larger
displacements in the 8 to 15 second range, and Load Case 8 had larger displacements in
the 15 to 20 second range and showed larger oscillations throughout the rest of the ground
motion which can be a result of additional mass due to train loading.

To further demonstrate other seismic performance metrics, hysteresis loops for the pier
columns as obtained from force-displacement relationships were plotted. The same four
columns (#2, #6, #8, and #11) were selected from the displacement time-history analysis
and were analyzed under both load cases. Column forces were extracted from OpenSees
by assigning element recorders with the force parameter for the fiber-based column
element that was modeling the bottom of the pier columns. The shear force-displacement
relationships from the two load cases were plotted in the two lateral directions 1 and 2, i.e.
longitudinal and transverse directions, in Figure 4-15 and Figure 4-16, respectively. The
main objective of graphing the force-displacement behavior of the pier columns was to
identify extent of nonlinearity and damage in the columns. The nonlinearity is observed by
observing whether the loading and unloading behavior follows a similar slope which
signifies the column remains within the elastic region. From the hysteresis loops provided,
the force-displacement behavior can be observed to be relatively linear for the four columns
under both load cases with the transverse direction showing slight instances of nonlinearity.
The residual displacements previously shown are also indicators of nonlinear behavior.
Given the observed residual displacements, this might be attributed to other components
yielding or damage (e.g. bearings). However, it is beyond the scope of this study to interpret
 65

or assess the seismic behavior especially that no proper design was conducted for the bridge
components and only demonstration is desired here.

As the last exercise in this part of the study, the internal forces and moments within the
bridge girders were observed by plotting shear force and bending moment diagrams. Girder
straining actions are usually more important for gravity load checks and design. However,
for better demonstrations selected cases of girder straining actions are shown under the
seismic loading as it accounts for gravity loads already in addition to any extra demands
from the seismic loading. Forces in the bridge girder elements were recorded by assigning
element recorders to all 100-elastic beam-column elements used to model the bridge with
the force parameter. The recorders export the axial force, and shear forces and moments in
the local y and z-axis of the element cross-section. The forces and moments were plotted
along the length of the bridge for each load case at an arbitrary time-step of 4.600 seconds
during the peak of the Northridge ground motion. The shear force diagrams and bending
moment diagrams for Load Case 1 and Load Case 8 are shown in Figure 4-17 through
Figure 4-22, and Figure 4-23 through Figure 4-28, respectively. Again, analyzing the
obtained shear and bending moment values is not the goal here.
 66

Table 4-1. Example HSR bridge system load cases based on the train position above the
bridge (the cases represent instances of the train crossing the bridge).

Table 4-2. Column Base Reactions (kN) in Direction 3 from Static Analysis.
Column Base Reactions (kN)
Column Load Case 1 Load Case 8
1 14528.6 14520.8
2 19132.7 19790.2
3 19071.1 19758.7
4 19072.6 19506.9
5 19072.6 19530.6
6 19072.6 19534.3
7 19072.6 19766.4
8 19072.6 19681.1
9 19071.6 19066.3
10 19115.3 19115.4
11 14939.0 14939.1
Total 201221.3 205209.8
 67

Table 4-3. Periods for first 10 modes.

Period (seconds)
Mode Load Case 1 Load Case 8
1 0.691 0.704
2 0.691 0.699
3 0.560 0.662
4 0.407 0.594
5 0.349 0.561
6 0.264 0.546
7 0.209 0.537
8 0.204 0.513
9 0.170 0.504
10 0.147 0.463
 68

Load Case 1

Load Case 8

Figure 4-1. Train load cases used for Chapter 4.

 69

Figure 4-2. Vertical bridge girder displacements under static analysis for both load cases.
 70

Figure 4-3. Vertical bridge girder displacements under Load Case 1.

Figure 4-4. Vertical bridge girder displacements under for Load Case 8.
 71

Figure 4-5. Transverse bridge girder displacements under Load Case 1.

Figure 4-6. Transverse bridge girder displacements under Load Case 8.

 72

(a)

(b)
Figure 4-7. Displacement time-history of column #2 under Load Case 1 in:
(a) Longitudinal, (b) Transverse directions.
 73

(a)

(b)
Figure 4-8. Displacement time-history of column #6 under Load Case 1 in:
(a) Longitudinal, (b) Transverse directions.
 74

(a)

(b)
Figure 4-9. Displacement time-history of column #8 under Load Case 1 in:
(a) Longitudinal, (b) Transverse directions.
 75

(a)

(b)
Figure 4-10. Displacement time-history of column #11 under Load Case 1 in:
(a) Longitudinal, (b) Transverse directions.
 76

(a)

(b)
Figure 4-11. Displacement time-history of column #2 under Load Case 8 in:
(a) Longitudinal, (b) Transverse directions.
 77

(a)

(b)
Figure 4-12. Displacement time-history of column #6 under Load Case 8 in:
(a) Longitudinal, (b) Transverse directions.
 78

(a)

(b)
Figure 4-13. Displacement time-history of column #8 under Load Case 8 in:
(a) Longitudinal, (b) Transverse directions.
 79

(a)

(b)
Figure 4-14. Displacement time-history of column #11 under Load Case 8 in:
(a) Longitudinal, (b) Transverse directions.
 80

(a)

(b)
Figure 4-15. Force-displacement relationship of column #2, #6, #8, and #11 in the
longitudinal direction for: (a) Load Case 1, (b) Load Case 8.
 81

(a)

(b)
Figure 4-16. Force-displacement relationship of column #2, #6, #8, and #11 in the
transverse direction for: (a) Load Case 1, (b) Load Case 8.
 82

6000

4000

2000
Shear Force (kN)

-2000

-4000

-6000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-17. Bridge girder shear in the longitudinal direction (Vx) for Load Case 1.

6000

4000

2000
Shear Force (kN)

-2000

-4000

-6000

-8000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-18. Bridge girder shear in the transverse direction (Vy) for Load Case 1.
 83

4000

3000

2000

1000
Shear Force (kN)

-1000

-2000

-3000

-4000

-5000
0 32 64 96 128 160 192 224 256 288 320

Length (m)

Figure 4-19. Bridge girder shear in the vertical direction (Vz) for Load Case 1.

4000

3000

2000
Moment (kN-m)

1000

-1000

-2000

-3000

-4000
0 32 64 96 128 160 192 224 256 288 320

Length (m)

Figure 4-20. Bridge girder moment in the longitudinal direction (Mx) for Load Case 1.
 84

35000
30000
25000
20000
Moment (kN-m)

15000
10000
5000
0
-5000
-10000
-15000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-21. Bridge girder moment in the transverse direction (My) for Load Case 1.

10000
8000
6000
4000
Moment (kN-m)

2000
0
-2000
-4000
-6000
-8000
-10000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-22. Bridge girder moment in the vertical direction (Mz) for Load Case 1.
 85

6000

4000

2000
Shear Force (kN)

-2000

-4000

-6000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-23. Bridge girder shear in the longitudinal direction (Vx) for Load Case 8.

6000

4000

2000
Shear Force (kN)

-2000

-4000

-6000

-8000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-24. Bridge girder shear in the transverse direction (Vy) for Load Case 8.
 86

4000

3000

2000

1000
Shear Force (kN)

-1000

-2000

-3000

-4000

-5000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-25. Bridge girder shear in the vertical direction (Vz) for Load Case 8.

4000

3000

2000
Moment (kN-m)

1000

-1000

-2000

-3000

-4000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-26. Bridge girder moment in the longitudinal direction (Mx) for Load Case 8.
 87

35000
30000
25000
20000
Moment (kN-m)

15000
10000
5000
0
-5000
-10000
-15000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-27. Bridge girder moment in the transverse direction (My) for Load Case 8.

10000
8000
6000
4000
Moment (kN-m)

2000
0
-2000
-4000
-6000
-8000
-10000
0 32 64 96 128 160 192 224 256 288 320

Length (m)
Figure 4-28. Bridge girder moment in the vertical direction (Mz) for Load Case 8.
 88

5 SEISMIC RESPONSE OF PROTOTYPE HSR BRIDGE SYSTEM: MORE IN-

DEPTH DEMONSTRATION

Overview

Seismic loads pose a great threat to the stability of HSR bridges that can be built in high
seismic regions, such as California in the United States. A proper design guideline and
code are required to assess the seismic performance of an HSR bridge, which is not fully
mature and developed for the United States yet. Nonetheless, this chapter further extends
the brief seismic analysis presented in Section 4.4 by providing a more in-depth
demonstration of the attributes of a comprehensive analysis of the structural behavior of
HSR system with focus on bridge components. The more in-depth demonstration of
nonlinear time history analysis of HSR bridge systems performed in this chapter considered
three load cases and three ground motions applied with various intensities.

The three load cases were again selected from the 16 sample cases previously outlined in
Table 4-1 for the selected train and bridge prototypes used in this study. These are Load
Case 1, Load Case 6, and Load Case 9. Load Case 1 was selected similar to the sample
analysis conducted in Chapter 4 to demonstrate the HSR bridge response without any
loading from the train. Load cases 6 and 9 were selected to demonstrate the prototype HSR
bridge behavior under partial and full train loading. The load cases are illustrated in Figure
5-1. The prototype HSR bridge model under each of these load cases was subjected to three
ground motions sourced from the PEER Ground Motion Database by the University of
California, Berkeley. The acceleration time histories of the three selected ground motions
are shown in Figure 5-2. The first record is the same 1994 Northridge earthquake record
from the Sepulveda VA Hospital station as used before in Chapter 4. The two additional
ground motions include one from the 1995 Kobe earthquake recorded at the Takatori
station, and another one from the 1989 Loma Prieta earthquake recorded at the LGPC
station. Each of the three ground motions were applied with two intensity levels at 100%
and 200% scale of the original record. An additional analysis was performed for the
Northridge record scaled at 300% to explore the seismic response of the HSR bridge at
higher seismic demands.
 89

In addition to what was presented in Chapter 4 as sample seismic analysis, this chapter
provides a deeper look at both global and local behavior of selected bridge components
from the 100% and 200% scale ground motion runs. A comprehensive summary of the
maximum selected local and global responses of the HSR bridge are tabulated and provided
here. Additional displacement time-histories, force-displacement relationships, and
moment-curvature relationships are plotted to compare the effect of ground motion
intensity and train loading scenarios on the HSR bridge. Moreover, results from the 300%
scale Northridge record to assess the extent of nonlinear and inelastic behavior of the HSR
bridge columns as well as the force-deformation behavior of selected track-bridge
interaction zero-length elements to observe the load transfer within the system during
seismic events.

Maximum Response Tables

The behavior of the prototype HSR bridge was analyzed by tabulating the maximum
responses under the various loading scenarios. A total of 12 tables were created to analyze
the maximum responses of the prototype HSR bridge. The local maximum responses of
the pier columns and bridge girder spans under each load case (1, 6, and 9) were tabulated
for the three ground motions at an amplification of 100% and 200%, resulting in 6 tables.
The shear, moment, and curvature in the transverse and the longitudinal directions were
recorded for the pier columns. However, only the longitudinal shear and moment for the
bridge box-girder spans were recorded at each end of the spans since the in-plane responses
were not of interest. The other 6 tables demonstrate the global maximum displacement and
acceleration of the bridge girder nodes directly above the pier columns for the same load
variations. The values in the tables represent the absolute maximum responses (positive or
negative) and the maximum response within each category was highlighted to help
visualize the trends under each load case.

Observing the tabulated maximum local responses of the pier columns and girder spans
presented in Table 5-1 through Table 5-6, there is an obvious increase in magnitude for all
presented values when comparing the maximum response under the original 100% scaled
ground motion to the 200% scaled ground motion. The columns experienced a significant
 90

increase due to the larger seismic forces applied at the base of the model connected to the
column footings through translational springs. Column shear, moment, and curvature
showed an average increase of 70%, 28%, and 32% about the longitudinal axis, and an
average increase of 56%, 19%, and 30% about the transverse axis. The box-girder sections
were assumed to be less affected by the earthquake intensity because they are capacity
protected elements and should not see higher demands beyond what is dictated by the
columns’ capacity.

The magnitude of the maximum local responses for Load Case 1, 6, and 9 were compared
among all of the considered loading scenarios to identify the impact of train loading. The
Load Case 6 train loading is heavily shifted to one side of the bridge and imposes less total
weight of the train on the bridge, relative to full train load in Load Case 9, due to a portion
of the train not being on the bridge. Yet, the bridge seismic response due to both load cases
with partial and full train load on top of the bridge were similar. Comparing the average
responses between Load Case 1 with no train loading to Load Cases 6 and 9 with train
loading, the most notable change was in the maximum longitudinal moment response
where an average increase of 10% and 12% was observed for Load Case 6 and 9,
respectively. The maximum column shear response showed small increases of less than
2% and the maximum transverse column moment increased by 4% for both load cases.
Load Case 9 showed 6% increase for the maximum column curvature in both directions
and Load Case 6 increased by 4% for both directions. The in-plane girder shear and
moment also increased by 5% for Load Case 6 and 6% for Load Case 9. When comparing
the two load cases with train loading, Load Case 9 had slightly larger responses on average
when compared against Load Case 6.

The maximum global response in terms of the displacement and acceleration measured at
the girder nodes directly above the respective pier columns were obtained under the three
different ground motions and are tabulated in Table 5-7 through Table 5-12. Each table
compares results from the three selected load cases. Thus, the six tables represent the six
ground motion scenarios: 3 different records × 2 different seismic intensities. On average,
the higher intensity ground motions at 200% scale increased the longitudinal and transverse
 91

maximum global displacements by 111% and 87%, respectively, as well as the longitudinal
and transverse maximum global accelerations by 54% and 55%, respectively. When
comparing Load Case 1 to Load Case 6 and 9, the most notable change was increase in the
average maximum longitudinal displacement by 4% for both load cases. The maximum
transverse displacement increased by 3% for Load Case 6 but did not change for Load Case
9. The increase in maximum acceleration for either load case was insignificant with less
than 1% increase and the transverse acceleration for Load Case 9 even decreased by 3%.
The minimal increase in the longitudinal acceleration and decrease in the transverse
acceleration for the load cases with train loading can be assumed to follow the fundamental
concept of Newton’s Second Law of Motion. The addition of train loading increases the
mass and in-turn decreases the acceleration of the bridge to maintain force equilibrium;
however, this is assuming a perfectly linear system which is not the case for this study since
inelastic material behavior have been modeled. Seismic response of the prototype HSR
bridge will vary as the stiffness of the structure changes throughout the cyclic loading of
the seismic forces and the overall mass changes based on the load case. Ultimately, the
lack of major change in local and global response due to additional train loading could be
a result of the inherent conservative design nature of HSR bridges. Compared to similar
highway bridges, HSR bridge columns are designed to be much stiffer to minimize lateral
deformations to improve the train and track stability as well as the riding comfort of
passengers. HSR bridges feature massive columns with larger force and moment capacities,
relative to equivalent highway bridges, which indirectly result in HSR bridges withstanding
larger earthquake forces before failure.

Seismic Behavioral Graphs

The behavioral graphs plotted for the additional seismic analysis conducted in this chapter
include displacement time-histories, force-displacement relationships, and moment-
curvature relationships of selected columns in the longitudinal and transverse directions.
The displacement time-history graphs demonstrated the displacement amplitudes and
trends along with residual displacements at the end of the ground motion duration. The
force-displacement and moment-curvature relationships graphs serve to demonstrate the
full range of response of the prototype HSR bridge system throughout the course of the
 92

ground motions whether it remains linear elastic or started getting nonlinear. The graphs
were plotted for the data retrieved from the prototype HSR bridge response under the three
earthquakes at 100% and 200% amplification.

Displacement time-histories for Load Cases 1, 6, and 9 under all three ground motions are
shown in Figure 5-3 through Figure 5-6 for the transverse and longitudinal directions and
at 100% and 200% seismic intensity. Each of the four figures provides nine subplots where
each subplot compares the displacement at the girder end node above columns #3, #6, and
#11 to visually assess the displacement trends of the interior and exterior columns. The
nine subplots represent the three different ground motion records × the three train loading
cases. Observing the figures for the 100% scale, the time-histories for the Kobe and Loma
Prieta earthquakes oscillated about the 0 mm displacement mark for both directions, i.e. no
residual displacements were observed to indicate the columns among other components
stayed linear throughout the ground motion duration. The time-histories for the Northridge
earthquake were shifted to oscillate about the 40 mm mark for the longitudinal direction
and about the 110 mm mark for the transverse direction. These are residual displacements,
i.e. plastic damage, which indicate that either the columns underwent nonlinear inelastic
behavior or other components simulating the train-track-superstructure-substructure
interaction might have yielded. This was previously noted in Chapter 4. However, given
that the 200% run of the Northridge record rendered higher force demands in the columns,
the columns were obviously well below their capacities as a result of the 100% run.
Therefore, the residual displacements observed in the 100% or 200% Northridge
earthquake cases are not likely associated with the columns, which motivated an additional
analysis case at 300% as discussed later in this chapter. It is also noted from the Figures
5.2 through 5.5 that the overall displacement trends for the three load cases were nearly
identical between Load Case 1, 6, and 9 for each direction barring any apparent variations
in the displacement amplitudes after the 8 second mark.

For the 200% scale, larger residual drift between the interior and exterior columns become
apparent for all three ground motions in the longitudinal direction. The relative drift stayed
similar between the three load cases for the Northridge and Loma Prieta earthquakes, and
 93

showed a slight increase for the load cases with train loading for the Kobe earthquake. The
transverse displacements heavily increased for the Northridge earthquake, oscillating about
the 240 mm line for the load cases with train loading and the 140 mm line for the load case
with no train loading. In comparison, the Kobe and Loma Prieta earthquakes had small
residual transverse displacements which were nearly consistent among the load cases.

Based on the displacement time-history graphs for both scales, the addition of train loading
had higher influence towards the displacement trends for ground motions scaled at 200%.
The displacement trends under the Loma Prieta earthquake lacked any variation among the
load cases for either scale, but the Northridge and Kobe earthquakes showed definite signs
of increased residual displacement for the load cases with train loading under the 200%
earthquakes. Displacement time-histories for Load Case 6 and 9 also oscillate at a larger
magnitude towards the middle to end of the ground motion for the transverse direction
which proves the addition of train loading does increase the magnitude of bridge vibration
despite the peak displacement values being relatively similar for all the load cases.

Similar to the displacement time-history graphs, the force-displacement and moment-

curvature behavioral graphs were compiled in four figures, with each figure presenting a
respective direction and ground motion scale. Observing the force-displacement
relationships shown in Figure 5-7 and Figure 5-9 for columns #6, #8, and #11 and the
moment-curvature relationships shown in Figure 5-11 and Figure 5-13 for columns #1, #6,
and #10, the columns showed glimpses of inelastic response but stayed relatively linear
elastic. However, the columns clearly demonstrate signs of nonlinearity under the 200%
scale Northridge earthquake in the force-displacement graphs for both directions, shown
in Figure 5-8 and Figure 5-10, where larger or fatter hysteresis loops were recorded. The
moment-curvature graphs for the 200% scale ground motions presented in Figure 5-12 and
Figure 5-14 also showed instances of large nonlinearity for all of the ground motions. In
comparison to the transverse moment-curvature graphs, the longitudinal moment-curvature
relationship behaved along a lower slope. This can be assumed to be a result of the
geometric orientation of the rectangular pier columns providing higher resistance to
rotation in the transverse direction compared to the longitudinal direction.
 94

Although the force-displacement behaviors were similar among the three load cases, the
moment-curvature behaviors showed that the columns experienced larger responses for
Load Cases 6 and 9 for the ground motions scaled at 200%, which was an observation also
seen in the displacement time-histories. In general, the influence of train loading becomes
more apparent when the columns start to experience some nonlinearity due to large seismic
loading. This can be tied to the inherent design of HSR bridges being very stiff and high
capacity, which results in a bridge that can behave consistently regardless of various
loading scenarios but only up to a certain seismic demand level. However, further research
is necessary to fully validate this observation and tie it to proper seismic design and
assessment framework.

Regardless of the onset of nonlinear column behavior shown under the 200% scale runs, it
is not conclusive whether any of the columns reached its ultimate capacity already. Thus,
it was of interest to pick the most damaging ground motion out of the three utilized ones,
i.e. the Northridge record, and apply it at 300% scale. This mainly aimed at understanding
whether the residual displacements observed at least at the 200% scale were related to the
column’s nonlinear behavior. It was also desired to confirm whether the column reached
its capacity during the 200% run or still had more capacity that can be rendered at an even
larger seismic intensity. The displacement time-history, force-displacement, and moment-
curvature relationships are shown in both the transverse and longitudinal directions under
the 300% Northridge record in Figure 5-15 through Figure 5-20. Observing the
displacement time-histories, it can be confirmed that the columns approached their capacity
and might have failed under excessive nonlinear demands that reached about 1400 mm as
suggested by the residual displacement values that surpassed 500 mm for both directions.
Unlike the response at 100% and 200% scales, no other bridge component is likely to lead
to 500 mm residual displacements except the main lateral support system, i.e. columns.

The force-displacement and moment-curvature graphs for both directions confirm the large
nonlinear response and inelasticity within the columns as demonstrated through the large
hysteresis loops that stray from the core elastic behavior. Analyzing the seismic
performance of the prototype HSR bridge under the 300% scale further supports the
 95

perspective that a by-product of the HSR bridge column’s large stiffness requirement is the
large force and moment capacity that can help the columns remain almost linear elastic
under moderate seismic intensities. In other words, the large column nonlinearities were
not observed until the 300% intensity where the force and moment values suggest that these
are at the capacity of the analyzed columns. A formal design guideline and code would be
necessary in the near future to do a proper seismic assessment of HSR bridge behavior
under simultaneous train and seismic loading, which is a future work that can stem from
the research presented in this study.

Finally, the force-deformation behavior of selected track-bridge interaction elements for

the prototype HSR bridge were obtained and plotted under the Northridge record scaled at
300% and under the same train loading cases. Force and deformation were output for the
zero-length elements idealizing the fasteners, CA layers, and sliding layers at locations
directly above columns #4 and #6, which were selected arbitrarily. The force-deformation
behavior for fasteners supporting rail 1 and rail 2 of track 1 is shown in Figure 5-21 and
Figure 5-22 respectively. Similarly, the force-deformation behavior for the CA layers
supporting track 1, and the sliding layers supporting track 1 at the locations indicated above
are plotted in Figure 5-23 and Figure 5-24, respectively. These graphs compare the demand
and performance of the interaction elements under three levels of seismic intensity.

From this brief analysis, it is apparent that the fasteners and CA layers operate within its
elastic capacities which were defined as part of the modeling of the material behaviors (see
Figure 3-7 and Figure 3-8 in Chapter 3). Contrarily, the sliding layer has clearly exceeded
its yield capacity and is deforming heavily due to the lack of capacity. The sliding layer in
a ballastless track system connects the track system to the bridge deck and is prone to be
firstly damaged under earthquakes. The sliding layer is also implemented in ballastless
track systems to effectively dissipate seismic energy through the damage of the layer [12].
However, the damage observed in the sliding layer for this study is excessive and does
raise some concern. An obvious issue could be the lack of resistance provided in the
interaction layers of the track system due to the large sub-spans or intervals used to model
elements and springs along the length of the bridge. The reference study that the prototype
 96

track-bridge system was based off modeled each girder span as 50 elements of identical
lengths opposed to the 10 elements used for this study, which was a limitation to expedite
the modeling process given the overall goal that the model in-place is for demonstration
purposes. This modeling limitation significantly decreased the amount of springs modeled
per interaction layer because the springs were modeled at intervals five times larger than
that of the reference study for instance. Nonetheless, it is again noted that the provided
analysis in this chapter or previous ones were intended to only demonstrate the capabilities
associated with the developed HSR bridge system model, and touch on the potential
response metrics that could be assessed against a formal future design framework.
 97

Table 5-1. Maximum Local Responses – Northridge 100% Scale.

 98

Table 5-2. Maximum Local Responses – Northridge 200% Scale.

 99

Table 5-3. Maximum Local Responses – Kobe 100% Scale.

 100

Table 5-4. Maximum Local Responses – Kobe 200% Scale.

 101

Table 5-5. Maximum Local Responses – Loma Prieta 100% Scale.

 102

Table 5-6. Maximum Local Responses – Loma Prieta 200% Scale.

 103

Table 5-7. Maximum Global Responses - Northridge 100% Scale.

 104

Table 5-8. Maximum Global Responses - Northridge 200% Scale.

 105

Table 5-9. Maximum Global Responses - Kobe 100% Scale.

 106

Table 5-10. Maximum Global Responses - Kobe 200% Scale.

 107

Table 5-11. Maximum Global Responses – Loma Prieta 100% Scale.

 108

Table 5-12. Maximum Global Responses – Loma Prieta 200% Scale.

 109

Load Case 1

Load Case 6

Load Case 9

Figure 5-1. Train load cases used in the seismic analysis in Chapter 5.
 110

(a)

(b)

(c)

Figure 5-2. PEER database ground motions used for the seismic performance assessment: (a) Northridge, (b) Kobe, and (c) Loma
Prieta.
 111

Figure 5-3. Longitudinal displacement time-history for columns #3, #6, and #11 at 100% – (Row: (1) Northridge, (2) Kobe, and
(3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 112

Figure 5-4. Longitudinal displacement time-history for columns #3, #6, and #11 at 200% – (Row: (1) Northridge, (2) Kobe, and
(3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 113

Figure 5-5. Transverse displacement time-history for columns #3, #6, and #11 at 100% – (Row: (1) Northridge, (2) Kobe, and (3)
Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 114

Figure 5-6. Transverse displacement time-history for columns #3, #6, and #11 at 200% – (Row: (1) Northridge, (2) Kobe, and (3)
Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 115

Figure 5-7. Longitudinal force-displacement relationship for columns #6, #8, and #11 at 100% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 116

Figure 5-8. Longitudinal force-displacement relationship for columns #6, #8, and #11 at 200% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 117

Figure 5-9. Transverse force-displacement relationship for columns #6, #8, and #11 at 100% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 118

Figure 5-10. Transverse force-displacement relationship for columns #6, #8, and #11 at 200% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 119

Figure 5-11. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at 100% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 120

Figure 5-12. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at 200% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 121

Figure 5-13. Transverse moment-curvature relationship for columns #1, #6, and #10 at 100% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 122

Figure 5-14. Transverse moment-curvature relationship for columns #1, #6, and #10 at 200% – (Row: (1) Northridge, (2) Kobe,
and (3) Loma Prieta, Column: (1) Load Case 1, (2) Load Case 6, and (3) Load Case 9).
 123

Figure 5-15. Longitudinal displacement time-history for columns #3, #6, and #11 at Northridge 300% – (Left: Load Case 1,
Middle: Load Case 6, Right: Load Case 9).

Figure 5-16. Transverse displacement time-history for columns #3, #6, and #11 at Northridge 300% – (Left: Load Case 1, Middle:
Load Case 6, Right: Load Case 9).
 124

Figure 5-17. Longitudinal force-displacement relationship for columns #6, #8, and #11 at Northridge 300% – (Left: Load Case 1,
Middle: Load Case 6, Right: Load Case 9).

Figure 5-18. Transverse force-displacement relationship for columns #6, #8, and #11 at Northridge 300% – (Left: Load Case 1,
Middle: Load Case 6, Right: Load Case 9).
 125

Figure 5-19. Longitudinal moment-curvature relationship for columns #1, #6, and #10 at Northridge 300% – (Left: Load Case 1,
Middle: Load Case 6, Right: Load Case 9).

Figure 5-20. Transverse moment-curvature relationship for columns #1, #6, and #10 at Northridge 300% – (Left: Load Case 1,
Middle: Load Case 6, Right: Load Case 9).
 126

(a) (b)

Figure 5-21. Force-deformation relationship of fasteners supporting rail 1 under

Northridge 300%: (a) Above column #4, (b) Above column #6.

(a) (b)

Figure 5-22. Force-deformation relationship of fasteners supporting rail 2 under

Northridge 300%: (a) Above column #4, (b) Above column #6.
 127

(a) (b)

Figure 5-23. Force-deformation relationship of CA mortar layers supporting track 1 under

Northridge 300%: (a) Above column #4, (b) Above column #6.

(a) (b)

Figure 5-24. Force-deformation relationship of sliding layers supporting track 1 under

Northridge 300%: (a) Above column #4, (b) Above column #6.
 128

6 SUMMARY AND CONCLUSIONS

Summary

High-speed rail (HSR) is a complex system that involves critical infrastructure components
such as bridges, that in turn, poses several design challenges unique to the nature of the
HSR systems. With the requirements for deflections, rotations, and natural frequencies of
bridge spans, comprehensive understanding of the HSR dynamic interactions among train-
track-bridge structures is a topic of great importance. Accordingly, national and
international research studies have focused on such dynamic interaction through
sophisticated structural models. The main objective of this study was to synthesize existing
knowledge to create a comprehensive modeling guideline for HSR bridge systems. To do
so, an extensive literature search was performed to compile the modeling techniques for
various HSR systems and identify common modeling practices. A prototype HSR system
model was constructed using the modeling methods researched and a follow through of the
steps taken to create a detailed model in OpenSees was documented and discussed
sequentially. Due to the lack of a complete design guideline for a full HSR model, a train
system, train-track system, and soil properties from separate studies were combined under
the assumption that they are compatible. Sample static and dynamic analyses were
performed for a variety of train loading scenarios, and the data was used to analyze the
behavior of the HSR superstructure. The analysis aimed to showcase some of the
capabilities associated with the developed OpenSees model. Amore elaborate summary of
the different components of this study are provided in the next few paragraphs.

A thorough literature review was conducted to synthesize the various methods of numerical
modeling techniques used to model HSR systems. Literature published from national and
international sources were reviewed and compiled to demonstrate and how the individual
components within a train system, track system, or bridge system have been modeled in
previous studies and the similarities and differences regarding the finite element modeling
techniques. Doing so, the reader can gain insight on how to model different types of train,
track, and bridge systems and apply this knowledge to the formulation of their own HSR
system model. This task also aided the selection of the prototype train-track-bridge system
 129

modeled to demonstrate the application of the modeling techniques highlighted in the

literature search.

Based on the studies analyzed in the literature search, a prototype train system and track-
bridge system were selected to construct an example HSR model. The prototypes were
selected based on available information regarding design. Although the model is for
demonstration purposes, a realistic design would produce results that can be comprehended
and allows for easier identification of any errors in the formulation of the model. The
modeling procedures for each component of the HSR model in-place followed the methods
presented in their respective studies. Any information that was not stated in the reference
study was assumed using knowledge gained from the literature search. A step-by-step
guide of the process of formulating the model and analysis parameters from start to finish
were documented, accompanied by snapshots from the OpenSees model in-place for visual
demonstration.

To exemplify potential data analysis with the variety of data that can be output by
OpenSees, sample static and dynamic analyses were performed with a load case without
train loading on the HSR bridge and with train loading on the HSR bridge. Additionally, a
more in-depth set of nonlinear seismic analyses were performed to set the stage for
potential future seismic performance assessment. The analyses used three ground motions
retrieved from the PEER Ground Motion Database and scaled at 100% and 200%. Three
different load cases with no, partial, and full train loading were considered to observe the
sensitivity of seismic response of the bridge with respect to the train loading scenarios.
Although the train was modeled to be stationary during the seismic loading, this simulates
a scenario where a train would be called to a stop after notice of an earthquake early
warning. Local and global response of the prototype HSR bridge was presented through
maximum response tables and behavioral graphs including displacement time-history,
force-displacement, and moment-curvature, and a comparative seismic assessment for the
response was conducted.
 130

Conclusions

The focal point of this dissertation was the presentation of numerical modeling methods of
HSR bridge systems including train-track-structure interaction. The modeling details
provided in Chapter 3 along with the complementary step-by-step procedure and scripts
provided from an example OpenSees input file in Appendix B are the main outcome of this
research study. Thus, the impact is more of a product as opposed to set of conclusions
based on analytical studies. Nonetheless, the study provided a demonstration of the seismic
response of HSR bridges through a prototype HSR model created based off previous
studies. The analysis results presented in that part of the study are based on a prototype
HSR bridge system assumption that were generously assumed to be applicable to one
another. However, general conclusions can be still drawn from the performance of the
prototype HSR bridge from a broad perspective, which at least could serve as a foundation
for future research, as provided next.

Based on the seismic performance of the model in-place, the location of train loading for
Load Case 6 and 9 did increase the local and global response within the bridge girders and
columns. The maximum longitudinal moment response in the bridge columns under train
loading experienced an average 10% and 13% increase throughout the three ground
motions scaled to a 100% and 200% for Load Case 6 and Load Case 9, respectively.
Column curvature also increased in the longitudinal and transverse directions by 4% and
6% on average for Load Case 6 and Load Case 9, respectively, and the maximum transverse
moments in the columns showed an average increase of 5% for both of the load cases with
train loading. The columns did not experience a significant increase in maximum shear
forces due to additional train loading with less than 2% increase on average due to train
loading. As for the global responses, bridge girders under Load Case 6 and Load Case 9
had an average increase of 4% for the maximum longitudinal displacement. Acceleration
at the girder level for either direction experienced insignificant effects, even decreasing by
3% for the acceleration under Load Case 9 in the transverse direction.

Although the maximum response of the HSR bridge experienced variation due to the
addition of train loading, the behavioral trends documented in the force-displacement and
 131

moment-curvature graphs were nearly identical with and without train loading for the
original scale of the ground motions and showed slight instances of increased nonlinear
loading-unloading loops for the 200% scale. Increase in displacements throughout the
course of the ground motion were observed at the bridge girder level in the transverse
displacement time-histories. Exceptionally large nonlinearities were not observed until
analyzing the HSR bridge under the Northridge earthquake at 300% scale where apparent
inelastic behavior was observed in all of the behavioral graphs plotted for Load Case 9.

The similarities in the seismic performance of the HSR columns between the load cases
may be attributed to the intrinsic design, where force and moment capacities are much
higher compared to typical railway or highway bridges; a by-product of the desired
excessively large stiffness for HSR systems. In other words, the HSR bridge started to
show response variation due to static train loading when the linear elastic limit had been
breached. However, the inherent design complications for HSR bridges may be influenced
largely by the dynamic loading of the train system which was not included in this study.
To fully understand and design for the operation of HSR systems under the paramount
safety, future studies are recommended to analyze the seismic performance of HSR bridges
under the dual loading of dynamic train loading and dynamic seismic loading.

The overall performance of the prototype HSR bridge was well as it showed its ability to
behave within its linear capacity. The performance was particularly good under the original
scale of the ground motions. The HSR bridge columns were able to behave within its elastic
capacity and showed slight nonlinearities when analyzed under the 200% scaled ground
motions. Thus, at moderate ground motion intensities, it is safe to say the HSR bridge
columns behaved essentially linearly or at least did not get into a large range of
nonlinearities and were not at their force and moment capacities as well.

Research Impact

The work presented in this thesis is critical and timely as the implementation of HSR as a
major mode of transportation in the United States is coming into fruition. Due to the recent
advances in HSR research, national studies regarding this topic are still very limited and
heavily rely on the publications from researchers abroad in Europe and East Asia where
 132

HSR systems are widely used as a major method of transportation. This study resulted in
the following new and important contributions:

• The main contribution of this study is the walk-through of the processes of modeling a
prototype HSR system, including the train-track-bridge system in high detail. This
guide will allow future students and researchers with minimal experience in numerical
modeling or modeling in OpenSees to formulate their own HSR model. This guide can
also be of benefit to researchers or designers who may need some guidance, as existing
publications regarding this topic focus mainly on the analysis and results rather than
the specific methods that were used to model each sub-system.
• Sub-systems of HSR have evolved over the years as technological advancements
continue to improve the safety and efficiency of HSR. The extensive literature search
presented in this study synthesizes the modeling methods that have been used by
national and international researchers to idealize variety of train, track, and bridge
systems. Future researchers can access this study to understand how specific HSR sub-
systems are modeled and can pursue the publications referenced within this study for
further details since.
• The design and analysis of HSR bridges presents many challenges in comparison to the
design of highway bridges and conventional railway bridges. Consequently, this study
demonstrates a variety of potential methods for analyzing the seismic performance of
an HSR bridge through post-processing OpenSees output which would allow the
verification of design. Although the seismic performance assessment demonstrated in
this study is not meant to prove the soundness of the prototype HSR bridge modeled,
future work can be built off of the research presented to formulate a national code and
design guideline for HSR bridges.

Validities and Limitations

For completeness, a statement on the validities and limitations of this study are presented
here and discussed to provide points of future recommendations and improvements. Due
to the recent emphasis on implementing HSR systems as a mode of transportation in the
United States, the literature available is heavily limited to a few national studies and foreign
 133

studies that have been translated to English and published to journals. This results in
limitation of reference studies that can be researched for the purpose of understanding the
methods of numerical modeling of HSR systems.

Another issue is the validity of the prototype model analysis results due to the lack of
available design information regarding the prototype train, track, or bridge system that have
been selected from the reference studies. This is mainly due to the limitation of content that
can be included in such journal papers which could lead to the omission of detail that is not
the emphasis of the respective study. To combat this, many assumptions were made when
formulating the prototype model as discussed in Chapter 3. A design assumption example
being the cross-sectional design and strength of concrete and reinforcing steel of the pier
columns for the prototype bridge from the Beijing to Xuzhou section of the Beijing-
Shanghai high-speed railway. Although the cross-sectional area and height of the pier
columns were specified, the reinforcement layout and strength design were omitted so
generic assumptions were made regarding reinforcement ratio and strength of core
concrete.

For this study, the train-track-structure interaction was the focus of the model. Accordingly,
soil-structure interaction was simplified to a few springs between the column bases and the
fixed boundaries of the model as discussed in Chapter 3. Future studies should elaborate
on the modeling of soil-structure interaction by creating a sophisticated footing model with
pile-soil interaction and abutments at bridge ends. In addition, elements were not
discretized as precisely as recommended for a study focusing on analysis results, since the
focus is to demonstrate the process of modeling and analyzing a prototype model. The
prototype HSR bridge model in place is a primitive design combining a train system from
Korea, a track-bridge system from China, and general soil properties from California under
the assumption that they are all compatible for the sake of demonstrating a model.

A proper seismic analysis of any structural system requires a design guideline and code
that acts a standard for the performance of the structural design. Since there is no such
standards in-place for HSR bridges in the United States as of yet, the performance of the
prototype HSR bridge was based on engineering judgement and preexisting knowledge
 134

based on highway bridges. The analysis presented should not be taken as a

recommendation for design, but as a demonstration of potential seismic analysis that can
be conducted with a formal design guideline and code.

The seismic analysis presented was performed under earthquakes applied biaxially in the
longitudinal and transverse directions and applied as identical support excitations.
Although this is a common assumption when conducting seismic analysis of structures,
there are limitations to the validity of the analysis. Vertical excitations can impact the
response of girders with large spans, and multi-support excitations might be considered to
accurately analyze the response of multi-support structures under incoherent ground
motions. Future research is recommended to consider such limitations to expand the
comprehensive understanding of HSR bridge performance.
 135

7 REFERENCES

[1] Abbasi, Mohammad. Seismic Vulnerability Assessment of As-Built and

Retrofitted Multi-Frame Box-Girder Bridges. Diss. 2018.
[2] Du, X. T., Y. L. Xu, and H. Xia. "Dynamic interaction of bridge–train system
under non‐uniform seismic ground motion." Earthquake Engineering &
Structural Dynamics 41.1 (2012): 139-157.
[3] Connolly, D., A. Giannopoulos, and M. C. Forde. "Numerical modelling of
ground borne vibrations from high-speed rail lines on embankments." Soil
Dynamics and Earthquake Engineering 46 (2013): 13-19.
[4] En.wikipedia.org. 2020. Acela. [online] Available at:
<https://en.wikipedia.org/wiki/Acela> [Accessed 25 July 2020].
[5] En.wikipedia.org. 2020. High-Speed Rail In The United States. [online]
Available at: <https://en.wikipedia.org/wiki/High-
speed_rail_in_the_United_States> [Accessed 23 July 2020].
[6] En.wikipedia.org. 2020. KTX-Sancheon. [online] Available at:
<https://en.wikipedia.org/wiki/KTX-Sancheon> [Accessed 27 May 2020].
[7] En.wikipedia.org. 2020. Texas Central Railway. [online] Available at:
<https://en.wikipedia.org/wiki/Texas_Central_Railway> [Accessed 25 July
2020].
[8] En.wikipedia.org. 2020. Xpresswest. [online] Available at:
<https://en.wikipedia.org/wiki/XpressWest> [Accessed 25 July 2020].
[9] Garg, Vijay. Dynamics of railway vehicle systems. Elsevier, 2012.
[10] González, Arturo, Eugene J. OBrien, and P. J. McGetrick. "Identification of
damping in a bridge using a moving instrumented vehicle." Journal of Sound
and Vibration 331.18 (2012): 4115-4131.
[11] Guo, W. W., et al. "Integral model for train-track-bridge interaction on the Sesia
viaduct: Dynamic simulation and critical assessment." Computers & Structures
112 (2012): 205-216.
 136

[12] Guo, Wei, et al. "Simplified seismic model of CRTS II ballastless track
structure on high-speed railway bridges in China." Engineering Structures 211
(2020): 110453.
[13] He, Xingwen, et al. "Numerical analysis on seismic response of Shinkansen
bridge-train interaction system under moderate earthquakes." Earthquake
engineering and engineering vibration 10.1 (2011): 85-97.
[14] Hurty, Walter C., and Moshe F. Rubinstein. "Dynamics of structures (Text on
response of structures to large variety of dynamic excitations)." ENGLEWOOD
CLIFFS, N. J., PRENTICE-HALL, INC., 1964. 455 P (1964).
[15] Hutton, S. G., and Y. K. Cheung. "Dynamic response of single span highway
bridges." Earthquake Engineering & Structural Dynamics 7.6 (1979): 543-553.
[16] Hsr.ca.gov. 2018. 2018 Business Plan. [online] Available at:
<https://hsr.ca.gov/docs/about/business_plans/2018_BusinessPlan.pdf>
[Accessed 23 July 2020].
[17] Kaviani, Peyman, Farzin Zareian, and Ertugrul Taciroglu. "Seismic behavior of
reinforced concrete bridges with skew-angled seat-type abutments."
Engineering Structures 45 (2012): 137-150.
[18] Kim, Chul Woo, Mitsuo Kawatani, and Ki Bong Kim. "Three-dimensional
dynamic analysis for bridge–vehicle interaction with roadway roughness."
Computers & structures 83.19-20 (2005): 1627-1645.
[19] Kwark, J. W., et al. "Dynamic behavior of two-span continuous concrete
bridges under moving high-speed train." Computers & structures 82.4-5 (2004):
463-474.
[20] Lee, Chang Hun, et al. "Dynamic response analysis of monorail bridges under
moving trains and riding comfort of trains." Engineering Structures 27.14
(2005): 1999-2013.
[21] Lee, Chang Hun, et al. "Dynamic response of a monorail steel bridge under a
moving train." Journal of Sound and Vibration 294.3 (2006): 562-579.
[22] Li, Haiyan, et al. "Nonlinear random seismic analysis of 3D high-speed railway
track-bridge system based on OpenSees." Structures. Vol. 24. Elsevier, 2020.
 137

[23] Li, Yong, and Joel P. Conte. "Effects of seismic isolation on the seismic
response of a California high‐speed rail prototype bridge with soil‐structure and
track‐structure interactions." Earthquake Engineering & Structural Dynamics
45.15 (2016): 2415-2434.
[24] Liu, Kai, et al. "Experimental and numerical analysis of a composite bridge for
high-speed trains." Journal of Sound and Vibration 320.1-2 (2009): 201-220.
[25] Lu, Dan, Futian Wang, and Suliang Chang. "The research of irregularity power
spectral density of Beijing subway." Urban Rail Transit 1.3 (2015): 159-163.
[26] Mao, Jianfeng, et al. "Random dynamic analysis of a train-bridge coupled
system involving random system parameters based on probability density
evolution method." Probabilistic Engineering Mechanics 46 (2016): 48-61.
[27] Matsuura, Akio. "A study of dynamic behavior of bridge girder for high speed
railway." Proceedings of the Japan Society of Civil Engineers. Vol. 1976. No.
256. Japan Society of Civil Engineers, 1976.
[28] McCarthy, N., 2020. Infographic: The World's Longest High-Speed Rail
Networks. [online] Statista Infographics. Available at:
<https://www.statista.com/chart/17093/miles-of-high-speed-rail-track-in-
operation-by-country/> [Accessed 23 July 2020].
[29] Montenegro, P. A., et l. "Running safety assessment of trains moving over
bridges subjected to moderate earthquakes." Earthquake Engineering &
Structural Dynamics 45.3 (2016): 483-504.
[30] Nguyen, Dinh-Van, Ki-Du Kim, and Pennung Warnitchai. "Simulation
procedure for vehicle–substructure dynamic interactions and wheel movements
using linearized wheel–rail interfaces." Finite Elements in Analysis and Design
45.5 (2009): 341-356.
[31] Opensees.berkeley.edu. 2020. Opensees User Manual. [online] Available at:
<https://opensees.berkeley.edu/OpenSees/manuals/comparisonManual/1417.h
tm> [Accessed 6 August 2020].
 138

[32] Opensees.berkeley.edu. 2020. Openseeswiki. [online] Available at:

<https://opensees.berkeley.edu/wiki/index.php/Main_Page> [Accessed 6
August 2020].
[33] Plasseramerican.com. 2020. Plasser American - Machines & Systems - Ballast
Bed Cleaning. [online] Available at:
<http://www.plasseramerican.com/en/machines-systems/ballast-bed-
cleaning.html> [Accessed 23 July 2020].
[34] Queiroz, F. D., P. C. G. S. Vellasco, and D. A. Nethercot. "Finite element
modelling of composite beams with full and partial shear connection." Journal
of Constructional Steel Research 63.4 (2007): 505-521.
[35] Rocha, J. M., A. A. Henriques, and R. Calçada. "Probabilistic safety assessment
of a short span high-speed railway bridge." Engineering Structures 71 (2014):
99-111.
[36] Song, Myung-Kwan, Hyuk-Chun Noh, and Chang-Koon Choi. "A new three-
dimensional finite element analysis model of high-speed train–bridge
interactions." Engineering Structures 25.13 (2003): 1611-1626.
[37] Tanabe, M., Yamada, Y., & Hajime, W. (1987). Modal method for interaction
of train and bridge. Computers & Structures, 27(1), 119-127.
[38] Tayabji, Shiraz & Bilow, David. (2001). Concrete Slab Track State of the
Practice. Transportation Research Record. 1742. 87-96. 10.3141/1742-11.
[39] Wang, Jinfeng, et al. "Performance of Cement Asphalt Mortar in Ballastless
Slab Track over High-Speed Railway under Extreme Climate Conditions."
International Journal of Geomechanics 19.5 (2019): 04019037.
[40] Wu, Yean-Seng, and Yeong-Bin Yang. "Steady-state response and riding
comfort of trains moving over a series of simply supported bridges."
Engineering Structures 25.2 (2003): 251-265.
[41] Xia, He, and Nan Zhang. "Dynamic analysis of railway bridge under high-speed
trains." Computers & Structures 83.23-24 (2005): 1891-1901.
 139

[42] Xia, He, Nan Zhang, and Guido De Roeck. "Dynamic analysis of high speed
railway bridge under articulated trains." Computers & Structures 81.26-27
(2003): 2467-2478.
[43] Yan, Bin, Gong-Lian Dai, and Nan Hu. "Recent development of design and
construction of short span high-speed railway bridges in China." Engineering
Structures 100 (2015): 707-717.
[44] Yang, Yeong-Bin, and Bing-Houng Lin. "Vehicle-bridge interaction analysis
by dynamic condensation method." Journal of Structural Engineering 121.11
(1995): 1636-1643.
[45] Yang, Yeong-Bin, and Jong-Dar Yau. "Vehicle-bridge interaction element for
dynamic analysis." Journal of Structural Engineering 123.11 (1997): 1512-
1518.
[46] Yang, Yeong-Bin, and Yean-Seng Wu. "A versatile element for analyzing
vehicle–bridge interaction response." Engineering structures 23.5 (2001): 452-
469.
[47] Yu, Zhi-wu, et al. "Non-stationary random vibration analysis of a 3D train–
bridge system using the probability density evolution method." Journal of
Sound and Vibration 366 (2016): 173-189.
[48] Yu, Zhi-wu, and Jian-feng Mao. "Probability analysis of train-track-bridge
interactions using a random wheel/rail contact model." Engineering Structures
144 (2017): 120-138.
[49] Yu, Zhi-wu, and Jian-feng Mao. "A stochastic dynamic model of train-track-
bridge coupled system based on probability density evolution method." Applied
Mathematical Modelling 59 (2018): 205-232.
[50] Zeng, Zhi-Ping, et al. "Random vibration analysis of train–bridge under track
irregularities and traveling seismic waves using train–slab track–bridge
interaction model." Journal of Sound and Vibration 342 (2015): 22-43.
[51] Zhaohua, Feng, and Robert D. Cook. "Beam Elements on Two-Parameter
Elastic Foundations." Journal of Engineering Mechanics, vol. 109, no. 6, 1983,
pp. 1390-1402.
 140

APPENDIX A: OPENSEES COMMANDS

For the convenience of the reader, this Appendix provides the syntax and input parameter
definition (in form of screenshots as obtained from OpenSeesWiki, [32]) for the key
OpenSees commands used in creating the HSR bridge model.

Figure A-1. model command parameters [32].

Figure A-2. node command parameters [32].

Figure A-3. fix constraint command parameters [32].

Figure A-4. equalDOF constraint command parameters [32].

 141

Figure A-5. geomTransf Linear transformation command parameters [32].

Figure A-6. Steel01 material command parameters [32].

Figure A-7. Steel02 material command parameters [32].

Figure A-8. Concrete02 material command parameters [32].

 142

Figure A-9. ViscousDamper material command parameters [32].

Figure A-10. Elastic material command parameters [32].

Figure A-11. elasticBeamColumn element command parameters [32].

Figure A-12. dispBeamColumn element command parameters [32].

 143

Figure A-13. zeroLength element command parameters [32].

Figure A-14. twoNodeLink element command parameters [32].

Figure A-15. section fiber command parameters [32].

 144

Figure A-16. patch rect command parameters [32].

Figure A-17. layer straight command parameters [32].

Figure A-18. section aggregator command parameters [32].

 145

Figure A-19. mass command parameters [32].

Figure A-20. eigen analysis command parameters [32].

Figure A-21. Rayleigh damping command parameters [32].

Figure A-22. timeSeries path command parameters [32].

 146

Figure A-23. UniformExcitation pattern command parameters [32].

 147

APPENDIX B: SELECTED SCRIPTS FROM OPENSEES INPUT FILE

This Appendix provides selected, but detailed, scripts from a sample OpenSees TCL file
for modeling and analyzing a full HSR bridge system. The input files for a given bridge
configuration and various train positions over the bridge vary from 17,000 to 18,000 lines
and could be provided upon request from the author. Nonetheless, the provided scripts
herein should be sufficient to reproduce or generate full input files.

Figure B-1. Predefined geometric locations for train nodes.

 148

Figure B-2. Node set up for rear power car.

Figure B-3. Node set up for rear intermediate passenger car.

 149

Figure B-4. Node set up for first intermediate passenger car.

Figure B-5. Rigid elastic beam-column element for bogie arms in the x-direction.

Figure B-6. Rigid elastic beam-column element for bogie arms in the y-direction.

Figure B-7. Rigid elastic beam-column element for primary suspension arms in the
y-direction.

Figure B-8. Rigid elastic beam-column element for primary suspension arms in the
z-direction.

Figure B-9. Rigid elastic beam-column element for car-bodies.

 150

Figure B-10. Primary suspension system model for the power cars.

Figure B-11. Power car primary suspension node MP-constraints with equalDOF.
 151

Figure B-12. Secondary suspension system model for the power cars.

Figure B-13. Power car secondary suspension node MP-constraints with equalDOF.
 152

Figure B-14. Mass assignment for train car-bodies.

Figure B-15. Mass assignment for train bogies.

 153

Figure B-16. Mass assignment for power and exterior passenger car axle wheels.

Figure B-17. Mass assignment for intermediate passenger car axle wheels .
 154

Figure B-18. Node set up for rail 1 of track 1.

Figure B-19. Node set up for base plate of track 1.

Figure B-20. Node set up for track plate of track 1.

Figure B-21. Elastic beam-column element for rail 3 of track 2.

Figure B-22. Elastic beam-column element for track plates of track 1.

Figure B-23. Elastic beam-column element for base plates of track 1.

 155

Figure B-24. Zero-length element for fastener

Figure B-25. Zero-length element for lateral blocking.

Figure B-26. Zero-length element for CA layer.

Figure B-27. Zero-length element for sliding layer.

 156

Figure B-28. Zero-length element for shear reinforcement.

Figure B-29. CA layer node MP-constraints with equalDOF.

Figure B-30. Mass assignment for first two rail 1 nodes.

Figure B-31. Node set up for the first bridge girder span.
 157

Figure B-32. Example elastic beam-column elements for bridge girder.

Figure B-33. Node set up for bearings supporting the first span of the bridge.

Figure B-34. Zero-length elements for fixed bearings supporting the first span of the
bridge.
 158

Figure B-35. Zero-length elements for sliding bearings supporting the first span of the
bridge.

Figure B-36. Bearing node MP-constraints with equalDOF.

Figure B-37. Material properties for pier columns.

 159

Figure B-38. Section designer for pier cross-section.

Figure B-39. Predefined geometric values for pier columns.

 160

Figure B-40. Node set up for first two columns.

Figure B-41. Displacement-based fiber-section beam-column elements for first pier

column.

Figure B-42. Node set up for column footings and ground.

 161

Figure B-43. Column footings and ground node SP-constraints using fix.

Figure B-44. Zero-length element for bridge-soil interaction.

 162

Figure B-45. Rigid elastic beam-column element for footings of columns #1 and #2.

Figure B-46. Rigid elastic beam-column element for column-bearing connections at

column #1.

Figure B-47. Rigid elastic beam-column element for girder-bearing connections above
column #1.

Figure B-48. Rigid elastic beam-column element for first two girder-track system
connections.

Figure B-49. Mass assignment for first two nodes of bridge girder.

Figure B-50. Mass assignment for first two nodes of column #1.
 163

Figure B-51. Mass assignment for footings of column #1 and #2.

Figure B-52. Dead loads for train car-bodies and bogies.

 164

Figure B-53. Dead loads for power and extreme passenger car axle-wheels.
 165

Figure B-54. Dead loads for intermediate passenger car axle-wheels.

Figure B-55. Dead loads for rail 1 (first four nodes).

Figure B-56. Dead loads for track plate for track 1 (first four nodes).

Figure B-57. Dead loads for base plate for track 1 (first four nodes).
 166

Figure B-58. Dead loads for first bridge girder span.

Figure B-59. Dead loads for first pier column.

Figure B-60. Dead loads for foundations.

 167

Figure B-61. Definition of gravity load analysis parameters.

Figure B-62. Performance of gravity load analysis.

Figure B-63. Set up for modal analysis.

 168

Figure B-64. Set up for Rayleigh damping.

Figure B-65. Definition of seismic load analysis parameters.

Figure B-66. Definition of ground motion parameters.

Figure B-67. Application of ground motion in both directions.

 169

Figure B-68. Performance of seismic load analysis.