TRACK MODEL FOR PREDICTING DYNAMIC RESPONSES OF RAILROAD

BRIDGE APPROACHES

BY

WENTING HOU

DISSERTATION

Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy in Civil Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2020

Urbana, Illinois

Doctoral Committee:

Professor Erol Tutumluer, Chair

Professor Christopher Barkan
Professor Billie F. Spencer, Jr.
Associate Professor Hai Huang, Penn State University, Altoona
Associate Professor Debakanta Mishra, Oklahoma State University
 ABSTRACT

This research study focused on the development of a novel train-track-bridge model

that can analytically tackle the difficult problem of a railway track transition zone and

effectively simulate the dynamic response of a bridge approach subjected to moving train

loads. The scope of the research encompassed statistical field data analysis, analytical

model derivation, algorithmic developments, as well as experimental validation for the

track model to provide improved design and maintenance practices. The objective was to

better understand governing mechanisms of bridge approach problems that occur near

bridge abutments using a field validated train-track-bridge model, study problems

associated with differential settlement, hanging ties, impact loads and various lack of

support conditions, select mitigation methods for existing deficiencies, and as a result,

formulate improved bridge approach designs.

The first task in this PhD study involved identification of problematic track bridge

approach locations that experienced recurring differential movements for field

instrumentation, performance monitoring, and conducting field data collection and

interpretation along the US Amtrak Northeast Corridor lines near Chester, Pennsylvania.

Multidepth Deflectometers (MDDs) were installed at the selected bridge approaches to

monitor individual layer deformations of track substructure layers. Besides MDDs, strain

gauges were also mounted on the rail at the instrumented sites to measure vertical wheel

loads applied during the passage of a train. Statistical analyses were conducted on the

collected field data to quantify track transient response and performance trends at the

studied bridge approaches.

ii
 Field collected data indicated one significant problem occurring at the instrumented

bridge approaches, also known as the “hanging crosstie” problem, which is caused by

several sequential crossties near the bridge abutment experiencing lack-of-support. These

crossties with gaps formed underneath undergo oscillatory motion as the dynamic loading

from moving wheels push and pull these crossties through rail deformations. As a result,

the nonuniform support conditions of the track substructure worsen drastically and can

result in extremely high deformations including heave condition due to train passage.

This dissertation developed a ballasted train-track-bridge model and an analytical

solution scheme for both open track (regular track) and near bridge (with severe “hanging

crosstie” problem) locations. The simulation results from the developed ballasted train-

track-bridge model matched well with the deformation data collected from the field

instrumentation to properly validate the analytical solution scheme. Then, using the field

validated track model, various mitigation methods, including changing rail pad stiffness,

ballast stiffness, subgrade stiffness, and crosstie spacing, applicable to track bridge

approaches were studied and discussed. Furthermore, details on two successfully applied

mitigation methods, namely, (i) converting an open deck bridge to ballasted deck bridge

and (ii) installation of a new track panel consisting of approximately 30 concrete ties with

Under Tie Pads (UTPs) at a bridge entrance on Amtrak’s Northeast Corridor high speed

passenger lines, were presented.

iii
 ACKNOWLEDGMENTS

First and foremost, I would like to sincerely thank my advisor, Professor Erol

Tutumluer, for his continuous support and encouragement. His enthusiasm for research and

innovation has always been a source of inspiration for me. I am also grateful for the

numerous professional and academic opportunities he has provided me over the past six

years. Without his guidance, I cannot imagine myself to be who I am today.

This dissertation would not have been possible without the many wonderful people

with whom I had the pleasure to collaborate. I am grateful to Drs. Hasan Ozer and Issam

Qamhia for collaborating on the study of aggregate by-products. I am thankful to Huseyin

Boler for the great time working together on railroad track transition project. Besides, I am

indebted to Bin Feng and Wei Li for their hard work when we collaborated on the railroad

crosstie support condition project. I gratefully thank Drs. Christopher Barkan, Billie F.

Spencer Jr., Hai Huang, Debakanta Mishra for taking part in my Ph.D. thesis committee

and providing me with invaluable suggestions.

I would like to thank my friends, Yubing Liang and He Wang, who have provided

me with constant support and brought tremendous joy to my graduate life. My special

thanks are to my colleagues Priyanka Sarker and Wenjing Li, for their continuous kindness,

support, and encouragement over the years.

Last but not least, my deepest love and gratitude are for my family, my father

Kaifeng Hou, my mother Jun Wang, and my husband Tuo Zhao, for their unconditional

love, constant support, encouragement, companionship, humor, and strength. No words

can describe how grateful I am for everything they have done for me. It is with great

pleasure that I dedicate this dissertation to them.

iv
 TABLE OF CONTENTS

CHAPTER 1: INTRODUCTION ....................................................................................... 1

CHAPTER 2: LITERATURE REVIEW .......................................................................... 10

CHAPTER 3: QUANTIFICATION OF RAILROAD BRIDGE APPROACH

PERFORMANCE ............................................................................................................. 46

CHAPTER 4: DISCRETELY SUPPORTED TRAIN-TRACK-BRIDGE MODEL ....... 83

CHAPTER 5: FIELD VALIDATION OF TRACK MODEL ........................................ 106

CHAPTER 6: EVALUATION OF MITIGATION METHODS .................................... 135

CHAPTER 7: CONCLUSIONS AND RECOMMENDATIONS .................................. 168

REFERENCES ............................................................................................................... 173

APPENDIX A : FIELD GPR DATA AND LAYER PROFILES ................................. 190

APPENDIX B : FIELD MONITORED TRANSIENT DATA ..................................... 194

APPENDIX C : FOURIER TRANSFORM BASED MODEL ..................................... 208

APPENDIX D : CODE FOR TRAIN-TRACK-BRIDGE MODEL.............................. 211

v
CHAPTER 1: INTRODUCTION

1.1 Introduction

While most railroad track structures will not experience sudden changes over a

short distance, there still are cases where the adjacent track structures exhibit totally

different characteristics. Track transition zone is referred to as a certain section of track

which results in different characteristics noted as discontinuities along the track structure.

Due to the sudden change in track stiffness, the “stiff” side of a track transition, e.g. bridge

deck, undergoes lower deformations due to train loading, compared to the “less stiff” side,

e.g. open track side. Such differential movement caused geometry problems at railway

transitions have been well recognized, and they present a significant challenge to

railroaders as far as maintenance of track profile is concerned (Li and Davis, 2005;

Banimahd et al. 2012).

Design standards including European Rail Research Institute (ERRI), UIC code

719, Design Standards for Railway Structures and Commentary, and AREMA all involve

the designs of a transition zone (Muramoto, 2013). Track transition zones require frequent

maintenance; when neglected, they deteriorate at an accelerated rate compared to a regular

track (Kerr and Maroney, 1993). The Association of American Railroads (AAR) reported

an annual expenditure of approximately USD 200 million by US railroads to maintain track

transitions (Sasaoka and Davis, 2005). Similarly, more than USD 110 million was spent

annually on transition zones in Europe by 1999 (Hyslip et al. 2009). According to the AAR

data center, America's freight railroads spend $25 billion a year on average from 2006 to

1
2018 on capital expenditures and maintenance expenses related to locomotives, freight cars,

tracks, bridges, tunnels and other infrastructure and equipment.

The most common transition problems are often encountered at bridge approaches,

with the approach track on either side of the bridge abutment being much less stiff

compared to the bridge deck often supported by deep foundations. Figure 1.1 illustrates a

typical structural view of the transition zone at track bridge approach. The “embankment

side” on the right-hand side represents regular ballasted track structure. And the “bridge

side” on the left-hand side represents an open deck bridge where there is no ballast layer.

As shown in the figure, the abutment at bridge side is often non-deformable and modulus

properties of the layers on both sides could be significantly different. In accordance, the

tracks on a bridge deck undergo significantly lower deformations under loading compared

to the approach tracks. This sudden change in track deformation behavior at the transition

point may result in extreme dynamic loading conditions and may ultimately lead to rapid

deterioration of the track and bridge structural components (Mishra et al. 2012).

Figure 1.1 Typical Profile of Railroad Track Structure at Bridge Approaches.

2
 One of the most influential results of the differential movement occurring at bridge

approaches is the “hanging crosstie” problem, which is caused by several sequential

crossties near the bridge abutment experiencing lack-of-support (Kerr and Maroney, 1993;

Li and Davis, 2005; Muramoto et al. 2012; Banimahd et al. 2012; Sanudo et al. 2016).

These crossties with gaps formed underneath undergo oscillatory motion as the dynamic

loading from moving wheels push and pull these crossties through rail deformations. As a

result, the non-uniform support conditions of the track substructure worsen drastically and

can result in extremely high deformations including heave condition due to train passage.

Figure 1.2 illustrates an example of unequal tie-ballast gap modeled by Stark et al. (2015)

Figure 1.2 Illustration of Unequal Tie-Ballast Gap (Stark et al. 2015)

To develop effective technologies for improving designs of railway transitions at

bridge approaches, improve safety, and ensure passenger comfort, the mechanism of

dynamic responses at railroad bridge approach must be properly identified. This requires

the combined application of field instrumentation for measuring dynamic track responses

as well as a field validated approach of effective numerical modeling.

3
1.2 Problem Statement

At bridge approaches, differential settlement of the foundation and unsupported ties

have been found in the vicinity of bridge abutments (Sanudo et al., 2016). The working

conditions of bridge approaches under dynamic loading have great impact on the rider

comfort and operational safety. Track geometry also suffers from accelerated deterioration

at bridge approaches.

In order to simulate dynamic responses of a track system at bridge approach, it is

essential to consider the coupled effect of moving train, open track, hangint crossties and

bridge deck together. To mend the gap in literature, this dissertation focuses on developing

a model to ideally consider discretely supported track for both the open track side and the

bridge side. The open track side includes multiple layers underneath the rail including

crosstie, ballast, and subgrade. The bridge side adequately models the effect of the stiff

open deck bridge. In addition, the hanging crosstie problem that often occurs at the bridge

approach is demonstrated in the model to simulate the nonlinear nature of track

substructure when there is a gap between crosstie and ballast particles. The developed

model can be utilized to simulate responses at bridge approaches under real train moving

and guide designs for smoother transition zones.

Various combinations of improvement plans, and mitigation methods could be

potentially beneficial to track behavior under loading. However, it is often difficult and

expensive to try different combinations of mitigation methods in the field. Therefore, the

developed model can be further utilized to evaluate effectiveness of potential mitigation

methods.

4
1.3 Objectives

In this PhD study, field monitoring data collected from railroad track bridge

approaches are analyzed to determine track dynamic responses and deformation behavior.

A dynamic track model aimed at simulating track behavior at bridge approaches under

moving load is developed. Both open track and bridge deck are included in the simulation

model to study bridge approach dynamic loading conditions. A nonlinear relationship

between crosstie and ballast contact forces and relative displacements are considered in the

dynamic track model. The detailed objectives of this research study are listed as follows:

(1) Develop a train-track-bridge model at the bridge approach that can realistically

predict responses under dynamic train loads, for example, due to the observed

difficult cases of hanging crossties and lack of support conditions, and validate the

proposed model with field collected data;

(2) Use the validated model as an analytical solution scheme to study the effects of

various mitigation methods applicable to track bridge approaches including

changing crosstie spacing and substructure layer properties, converting open deck

bridge to ballasted deck bridge, and installation of under tie pads.

This research study therefore contributes to the knowledge and state-of-the-art in

science and engineering with the development of a dynamic track model at transition zones,

specifically ballasted railway bridge approaches, that will eventually help engineers to

better simulate dynamic response behavior of track structures under moving loads. The

validated track model is expected to serve as an analysis tool to provide improved design

and maintenance practices.

5
1.4 Methodology and Scope

As part of recent track transition zone research efforts at UIUC, major field

instrumentation tasks were undertaken at several railroad bridge approaches from July to

August of 2012. Multi-Depth Deflectometers (MDDs) and rail strain gauges were installed

on three Northeast Corridor bridge approaches of Amtrak passenger lines near Chester,

Pennsylvania, USA. Transient deformation and wheel loading data from both passenger

and freight trains were recorded at the instrumented sites at different time intervals.

As part of the study scope, the collected field data are first reduced and analyzed

using statistical approaches including 95% confidence interval. Next, individual layer

transient displacements from the MDD data are used to the backcalculate the instrumented

track substructure layer moduli using the field measured wheel loads in order to evaluate

changes in properties and deformation behavior of track substructure layers at different

locations (open track vs. near bridge). This step employed GEOTRACK, a well-known

elastic layered software program, to generate approximate track substructure layer

properties that could provide a basic approximation of the field condition.

An incremental progression of the train-track-bridge model development is

followed to simulate the behavior of railroad bridge approaches under dynamic loading.

First, a numerical model of discretely supported train-track-bridge model is developed to

consider the realistic coupled conditions of the open track and the bridge deck sides of the

bridge approach. The derived model is then solved with Newmark integration method and

implemented in MATLAB for numerical results. This step is intended to simulate open

track behavior at bridge approaches under dynamic single moving wheel load and compare

6
the predicted results with field measured responses on the open track side of the bridge

approach to validate effectiveness of the solution scheme.

Next, the linear train-track-bridge model is further developed and refined to

consider nonlinear stiffness, which represents unsupported crossties near the bridge

abutment. The nonlinear train-track-bridge model is developed to employ advanced

numerical simulation method Newmark-𝛽 Integration Method for numerical results. Given

that the focus of this study is the track substructure of bridge approaches, the bridge is

simplified as non-deformable beam. This advanced nonlinear train-track-bridge model is

then validated with field measured responses collected from the bridge approaches (near

bridge location) at Chester, PA, Northeast Corridor. Similar geometries, loading

parameters, and structural properties of the field conditions are used to accomplish model

validation. The validated model can be used to conduct sensitivity analysis on train speed

and train loading or weight effects.

Finally, several mitigation methods are studied using the validated train-track-

bridge model. The proposed model is employed to study vertical transient deformations of

track structure under moving loads with the help of possible mitigation methods such as

changing crosstie spacing and substructure properties, converting open deck bridge to

ballast deck bridge, and installation of crossties with under tie pads (UTPs) at transition

zones. Results from the model simulations are used to propose practical solutions for bridge

approach problems, i.e., differential movement.

The ultimate goal is to develop a numercial model to simulate dynamic behavior at

railroad track bridge approaches and utilize the validated train-track-bridge model to aid in

selecting mitigation methods for remedying existing problematic bridge approaches and

7
developing improved transition zones. For example, by identifying the effects of track

properties, vehicle speed, and hanging crossties on the performance trends of both open

track and near bridge locations, the designs of transition zones could be improved, which

would result in adequate track modulus and support with little differential movement.

1.5 Organization of the Thesis

This thesis consists of seven chapters. The remaining chapters are organized as

follows:

Chapter 2 provides a comprehensive literature review focused on definitions, field

performance, previous published modeling approaches and possible mitigation methods of

railroad track transition zones as the bridge approaches.

Chapter 3 summarizes field instrumentation efforts at the bridge approaches on

Amtrak’s Northeast Corridor lines. Next exploratory data analyses on the collected field

instrumentation data are presented. Chapter 3 also presents inferential statistical data

analysis of field instrumentation data. Finally, backcalculation results of substructure layer

moduli using field MDD deformations and wheel load data are presented. These calculated

moduli are used as a base line for estimating layer stiffness characteristics in the developed

numerical track model.

Chapter 4 describes the detailed model development, mathematical derivations as

well as numerical and analytical solution schemes of the developed ballasted train-track-

bridge model for both the linear model at regular embankment track structure and the

nonlinear model at near bridge location. The proposed nonlinear track model adequately

considers “hanging tie” problem at bridge approaches.

8
 Chapter 5 presents the numerical simulation results for two case studies utilizing

the established ballasted train-track-bridge model. The developed track model simulates an

Amtrak passenger train moving on top of track bridge approaches. Results for both open

track location (linear model) and near bridge location (nonlinear model) are compared with

field instrumentation data for validation purposes. The validated model is then used to

generate more key indexes of track dynamic response. In addition, the validated model is

also utilized to study the train speed and train loading or weight effects on track dynamic

responses.

Chapter 6 utilizes the validated train-track-bridge model to investigate the effect of

different mitigation methods for railroad track bridge approaches including changing

crosstie spacing and substructure properties, converting open deck bridge to ballasted deck

bridge, and installation of under tie pads on several crossties at the bridge approach

transition zones.

Finally, Chapter 7 summarizes the research effort, findings and conclusions. Future

recommendations are also provided in this chapter.

9
CHAPTER 2: LITERATURE REVIEW

2.1 Railway Track Structure

This chapter provides an introduction to railway track structures and terms, types

and performance records of track transitions as bridge approaches. Recent work on bridge

approach instrumentation is described first. Analytical and numerical modeling of bridge

approaches is then reviewed. Mitigation efforts are also summarized in this chapter.

Compared to other transportation modes, railroad is one of the most economical

and energy efficient ways to transport both people and freight. Figure 2.1, for example,

shows that intercity rail is the most efficient mode for the fuel economy among the 9

transportation modes listed.

Figure 2.1 Different Modes of Transportation (Davis & Boundy, 2018).

There are two major types of railroads from the perspective of structural layered

systems. One is called ballasted track, which is a traditional type of track structure,

consisting of rail, crossties, and ballast layer. The other type of track system is non-
10
ballasted track, which substitutes ballast with concrete slabs. The two types of track

systems both have their own advantages and are used widely worldwide. Ballasted track is

a traditional railroad track system that is usually used in the normal speed passenger and

freight lines, while ballastless track has been more and more adopted in high-speed lines.

Ballasted track requires a lower initial construction cost but higher maintenance cost while

ballastless track requires a higher construction cost but lower maintenance cost.

According to Association of American Railroads (AAR), the length of the US

freight railroad network is nearly 140,000 miles. Its general components include ballast,

crossties, tie plates and rail, as shown in Figure 2.2. Note that crossties are referred as

“sleepers” in Figure 2.2. Ballast is used as the granular layer to provide ease in

constructability, stability, load distribution, and drainage ability to the main track

component of a ballasted railway system for several decades. The ballast layer is mostly

manufactured from local materials, preferably crushed stone. Ballast is installed under

crossties for transferring from loads from passing trains. It is required to maintain the track

in both good horizontal and vertical alignment. Note that track geometry deterioration can

be tolerated effectively by ballast. Furthermore, the ballast layer is essential to provide

proper drainage to the track structure. Kaewunruen and Tang (2019) noted that to fulfill

the requirements, ballast must have the following characteristics:

1. Durable to absorb the loads from crossties and transfer the loads to the subgrade

without undue breakdown;

2. Hard-wearing with high abrasion resistance;

3. Angular particles with sufficient bulk density to resist track movement;

11
 4. Suitable particle size to allow packing and load transmitting, and sufficient void

space to allow free draining.

Another significant component is railway crossties. Crossties are used to withstand

loads and transmit them to the ballast to maintain uniform rail gauge for safety. They can

also improve stiffness and stability of track structure by providing lateral track resistance.

Crossties are usually made of timber, steel, concrete and plastic or composite materials,

while concrete is the most adopted material in high speed routes. The materials can largely

influence the deterioration process of crossties. (Silva et al. 2017)

Tie plates are used to support rails and fix the rail fastening system to provide a

smooth and uniform bearing surface for rail and crossties, prevent the rail from cutting into

crossties, maintain gauge, and help distribute wheel load to crossties.

Rail is mostly made of steel, which can help guide the train, withstand the wheel

loads and transmit the loads to crossties. Rail must provide a smooth and continuous rolling

surface for train wheels.

Figure 2.2 General Components of Ballasted Railway.

12
2.2 Track Transition: Definition, Types, and Performance Records

While most railroad track substructures do not experience sudden changes in the

operating areas, there are still many cases where intentionally or passively, the track

substructures across a very short distance would exhibit tremendous differences. Examples

of railway track transitions include tunnels, special track work, highway/rail at-grade

crossings, and the most common transition encountered at bridge approaches.

In the U.S railroad network system, there are over 100,000 bridges, generating great

discontinuities along the track. A track transition zone is defined as a certain section of

track which results in different characteristics noted due to such discontinuities along the

track structure. Due to the sudden change in track stiffness, the “stiff” side of a track

transition, e.g. bridge deck, undergoes lower deformations under loading, compared to the

“less stiff” side, e.g. open track side.

Many problematic issues arise at railroad track transition zones. One of the most

significant problems is differential settlements, which occur between two sides of the

transition point, and has been a global problem for both track maintenance personnel and

passengers. The Association of American Railroads (AAR) reported an annual expenditure

of approximately USD 200 million to maintain track transitions (Sasaoka et al. 2005,

Hyslip et al. 2009), whereas more than USD 110 million was spent annually on transition

zones in Europe by 1999 (Hyslip et al. 2009, ERRI 1999).

Railway track transition zones present a significant challenge as far as maintenance

of track profile is concerned (Woodward et al. 2007, Banimahd et al. 2012). Due to the

sudden change in track stiffness, the “stiff” side of a track transition undergoes lower

deformations under loading, compared to the “less stiff” side. This differential movement

13
often results in the formation of a “bump” in the track profile. Differences in track system

stiffness and/or damping characteristics, settlement of the ballast layer due to degradation

and/or fouling, and settlement of the subgrade and/or fill layers are some of the factors

commonly reported as mechanisms contributing to the differential movement at track

transitions. Proper understanding of different mechanisms contributing to this phenomenon

requires the combined application of field instrumentation with analytical and numerical

track modeling.

2.3 Bridge Approaches as the Most Significant Track Transition Type

Bridge approaches, as one of the most significant track transitions, have been

investigated by many researchers. The “bump” and “dip” formations at the end of a railroad

bridge were illustrated by Nicks (2009) as shown in Figure 2.3. Due to drastic differences

in substructure and loading conditions, the tracks on a bridge deck undergo significantly

lower deformations due to train loading compared to the approach tracks. This sudden

change in track deformation behavior at the transition point results in extreme dynamic

loading conditions and ultimately leads to rapid deterioration of the track and bridge

structural components. These structural damages often manifest themselves as track

geometry defects. Differential movement at bridge approaches often results in the

development of a “bump” usually within 15 m from the bridge end (Plotkin and Davis

2008).

A survey of railroads in North America, Australia, and Europe conducted in 2006

indicated that approximately 50% of bridge approaches developed a low approach, usually

6 to 102 mm in depth and 1.2 to 15.2 m in length, that adversely affected ride quality

(Briaud et al. 2006). Development of sudden dips adjacent to the bridge deck increases the

14
dynamic impact loads significantly. Koch and Hudacsek (2017) reported vertical dynamic

loads about twice the static wheel load level for coal gondolas at track transitions. Read

and Li (2006) concluded that the bump problem at track transitions was more significant

as a train moves from a high-stiffness track to a low-stiffness track. According to Read and

Li (2006), the problem of differential movement was more critical at the exit end of a bridge,

whereas the sudden track stiffness increase at the bridge entrance led to rail surface fatigue,

tie deterioration and rail seat pad deterioration.

Figure 2.3 The Bump (left) and Dip (right) at End of Railway Bridge (Nicks, 2009).

Moreover, the differential movement at track transitions is particularly problematic

for high-speed rail infrastructure as the “bump” is accentuated at high speeds. The issue is

even more critical for shared corridors carrying both freight and high-speed lines.

Transitions along shared corridors need to be maintained to satisfy the high ride quality

requirements associated with high-speed trains. Additionally, these transitions also need to

withstand the heavy loads imposed by slow-moving freight trains without undergoing

excessive deformations. With the current impetus for development of high-speed lines in

the U.S and the challenges associated with shared corridors for operation of passenger

trains at increased speeds, preventing and mitigating the problem of differential movement

at bridge approaches and other track transitions has become more significant.

15
 One of the most influential results of the differential movement occurring at bridge

approaches is the “hanging tie” problem, which is caused by several sequential crossties

near the bridge abutment experiencing lack-of-support (Kerr and Maroney, 1993; Li and

Davis, 2005; Muramoto et al. 2012; Banimahd et al. 2012; Sanudo et al. 2016). These

crossties with gaps formed underneath undergo oscillatory motion as the dynamic loading

from moving wheels push and pull these ties through rail deformations. As a result, the

nonuniform support conditions of the track substructure worsen drastically and can result

in extremely high deformations including heave condition due to train passage.

The root causes and major influential factors affecting bridge approach problems

are not fully agreed upon by researchers. From the investigation of four bridge approaches

with concrete ballast-deck bridges and concrete ties, Li and Davis (2005) reported

inadequate ballast and subballast layer performance to be the primary cause of track

geometry degradation. Using settlement rods installed in the test sections, they observed

no significant subgrade movements, but instead reported significant track geometry

deterioration for a site with cement-stabilized backfill. On the other hand, Selig and Li

(1994) identified subgrade stiffness to be the most influential parameter affecting the

moduli of ballasted tracks. As track transition problems are often related to the stiffness of

the approach trackbed, this would indicate that the subgrade layer plays the most significant

role in governing the differential movement at track transitions. Yet, recent field

instrumentation data to be discussed in detail in this dissertation clearly identify the ballast

layer as the primary contributor to the differential movement problem (Mishra et al. 2012,

Tutumluer et al. 2012, and Mishra et al. 2014). As the ballast “densifies” under repeated

16
train loading, related ballast movement and degradation often lead to recurrent ballast

settlement at the transition zones.

In spite of different factors being identified as most critical in affecting the

differential movement at track transitions, a general consensus exists regarding the list of

plausible mechanisms. Sasaoka and Davis (2005) attributed track transition problems to

three primary factors: (1) differential settlement, (2) differences in stiffness characteristics,

and (3) discrepancies in track damping properties between adjacent sections. Similarly, Li

and Davis (2005) listed (1) track stiffness change, (2) ballast settlement, and (3)

geotechnical issues as the major causes of bridge approach problems. Nicks (2009) listed

the following ten factors identified by researchers as contributing to the “bump”

development at railway bridge approaches: (1) differential track modulus, (2) quality of

approach fill, (3) impact loads, (4) ballast material, (5) drainage, (6) damping, (7) abutment

type, (8) bridge joint, (9) traffic considerations, and (10) quality of construction. Note that

although most researchers list “track stiffness difference” as an important factor

influencing the differential movement and other track deterioration problems at transitions,

Plotkin and Davis (2008) used five different analysis methods to conclude that stiffness

differences did not play an important role as far as track behavior and ride quality at track

transitions were concerned.

2.4 Summary of Past Track Transition Instrumentation Efforts

To better understand the mechanism of differential settlement at railroad track

transition zones, it is of vital importance to conduct in-field measurements to learn the

structural behavior under moving trains. Several attempts have been made in the past to

characterize “bump” development at track transitions. For instance, strain gauges (see

17
Figure 2.4) are widely used for measuring wheel loads on the rail. Namura and Suzuki

(2007) used axle box acceleration data to evaluate the wheel load on vehicles and placed

strain gauges on rail webs to measure the wheel loads on the rail. The field measured data

were compared with analytical model results to validate effectiveness of the modeling

approach. Mitigation methods were then proposed for minimizing the track geometry

degradation at track transitions. Sakurai et al. (2013) adopted a set of instruments for

monitoring performance of pre-stressed ballast track. The crossties of pre-stressed ballast

track were pulled down against ballast by tie rods fixed to the anchors laid under trackbed.

Strain gauges were placed on tie rod to measure reactions (Sakurai et al. 2013). Similarly,

Hayano et al. (2013) investigated the effect of ballast thickness and tamping repair

approaches on the settlement characteristics of ballasted tracks using a one-fifth scale

model developed in the laboratory with displacement gauges installed to measure

settlement of footing under cyclic loading. Computer vision techniques, such as 3D images,

were also used by researchers to measure soil mass density of ballast material (Hayano et

al. 2013). They introduced the method of using 3D digital camera to obtain stereo pair

images of excavated holes and estimate volume of the holes accordingly. The results

showed this method was comparable to water replacement technique.

18
 Figure 2.4 Dual-Element Shear Strain Gauges Installed at the Rail Neutral Axis for
Wheel Load and Tie Reaction Measurements.

As a quite effective method to study dynamic track response due to train loading,

Multidepth Deflectometers (MDDs), which are the subject of this research effort, have

been used to measure track substructure layer deformations. The MDD technology was

first developed in South Africa in the early 1980s to measure individual layer deformations

in highway pavements (Scullion et al., 1989). MDDs typically consist of up to six linear

variable differential transformers (LVDTs) installed vertically at preselected depths in a

small diameter (typically 45 mm diameter as also used in the current study) hole to measure

the displacements of individual substructure layers with respect to a fixed anchor buried

deep in the ground (De Beer et al., 1989). Although several studies in the U.S. have

implemented MDDs to measure layer deformations in highway and airfield pavements,

their use in railroad applications has been limited. Two previous studies in the U.S.

(Sussmann and Selig 1998, Bilow and Li 2005) used MDDs to monitor the deformations

19
in railway track substructure layers. It is noteworthy that the use of MDDs to monitor

railway track performance has been extensively pursued in South Africa (Grabe and Shaw

2010, Priest et al. 2010, Vorster and Grabe 2013). The vertical deformations are measured

from a reference head near the top surface to predetermined depths in a borehole with the

overall vertical deformation referenced to a fixed anchor in the subgrade (see Figure 2.5).

Coelho et al. (2011) studied typical transition zone in the Netherlands. In their research,

accelerations and velocities of the track, soil, and approach slabs in response to passenger

trains were measured for the calculation of displacements. In addition, track settlements

and pore water pressures were monitored over a 1-year period.

(a) (b)

Figure 2.5 Multi-Depth Deflectometer (MDD) Installed to Measure Vertical Deformation

Contributed by Each Layer of a Multi-Layer Strata (Weinmann et al., 2004).

20
2.5 Analytical and Numerical Modeling of Railroad Track Transitions

2.5.1 Analytical modeling approaches

There have been various mathematical models developed to interpret and predict

the dynamic response of railroad track. Early models were one- or two-dimensional

composed of a beam as the rail on a Winkler foundation under moving force as wheel load.

Mise and Kunii (1956) developed a theoretical framework for the vibrations of a flexible

beam supporting a moving locomotive. In their simplified model, only a beam under a

moving force was considered. Kalker (1996) introduced discretely supported rails under a

travelling vertical point load to better represent the irregular discrete support of the

crossties. Also, closed-form solutions were derived for the dynamic response of simple

beam under high-speed trains (Yang et al. 1997). Huang et al. (2009) developed the

analytical solution of a railroad track model under moving loads with asphalt trackbed used

underneath ballast layers (see Figure 2.6). Basu and Rao (2013) studied the steady state

responses of an infinite beam resting on a viscoelastic foundation where shear resistance

of soil was also included in the analysis.

Figure 2.6 Discretely Supported Track Model (Huang et al., 2009).

21
 A discretely supported rail system with coupling of soil’s shear resistance can

adequately represent the track and the substructure in the field. However, these models do

not include the mass of train and the interacting force between wheel and track or

irregularity of track profile. The work of Lee (1998) introduced dynamic response of a

Timoshenko beam on Winkler foundation subjected to moving mass load and showed the

limitation of neglecting inertial effects of the mass when only equivalent moving force was

considered. Furthermore, Lei and Noda (2002) developed a computational model for the

vehicle and track coupling system where the system was divided into two parts. Upper

structure is a whole locomotive with two layers of spring and damping system in which

vehicle and bogie are involved. The lower structure is a railway track where rails are

considered as beams rested on a double layer continuous elastic foundation. Random

irregularity of the track vertical profile is also simulated in their work. Zhai and Sun (1994)

also investigated a model for vertical interaction between vehicle and track. Vehicle

subsystem is modeled as a multi-body system with 10 degrees of freedom (DOFs) running

on the track with a constant velocity, and the track substructure as an infinite Euler beam

supported on a discrete continuous elastic foundation consisting of the three layers of rail,

crosstie, and ballast. Lei and Rose (2008) analyzed track vibration by Fourier transform

technique with random irregularity of track vertical profile.

Note that the above analytical models do not include the three-dimensional (3D)

dynamic wave field generated in the ground due to the passage of a train. Therefore, they

are restricted to the cases where the velocity of the train is much smaller than a critical

velocity (Krylov, 1995). Krylov (1995) studied ground vibrations generated by superfast

trains theoretically. He found that superfast railway trains moving with speeds approaching

22
or exceeding the Rayleigh wave velocity in the ground could cause very large increases in

ground vibration levels when compared to those generated by conventional trains. The

most typical Rayleigh wave velocity values for soils are 250-500m/s. Holm et al. (2002)

confirmed that “critical speed” is the speed at which trains induce a large resonant response

in the track with soft ground conditions resulting in excessive vertical vibrations. Based on

these theories, three-dimensional models on the soil considering ground vibrations were

proposed. Bian et al. (2008) conducted simulation of high-speed train induced ground

vibrations using a 2.5D finite element method. The train induced track and ground

vibrations have been studied using analytical-numerical combined method. Later on, a fully

coupled 3D train track soil model was also introduced by Huang et al. (2014) where three-

dimensional soil parts are included. Li et al. (2016) applied a 3D finite element model with

a cycle domain constitutive model for track settlement simulation. This study adopted a

cyclic densification model (CDM), which is formulated as a viscoplastic model, for 3D

continuum modelling of ballast settlement. The CDM could calculate effective crosstie

loads, while the contact pressure between crosstie and ballast was also calculated for every

vehicle passage. The method considered that the time of loading applying on adjacent

crossties was not the same. A full scale test track was built in Spain to validate the model

and calibrate chosen parameters. The results showed this model could predict the track

settlement under heterogeneous loading conditions and track characteristics.

Most of the developed analytical models are linear in nature and therefore limited

when nonlinear aspects become significant, for example, when there is a hanging tie

problem. To avoid the limitation, Tanabe et al. (2003) built a 3D numerical analysis model

where nonlinear springs and dampers were employed. Nielsen and Oscarsson (2004)

23
expanded their previous model to account for state-dependent track properties, separated

into linear contributions corresponding to an unloaded track, and non-linear contributions

that are dependent on the time-variant state of the different track components due to the

dynamic loading from a moving train model. Ling et al. (2014) developed a 3D dynamic

model of a nonlinear HSR coupled with a flexible ballast track. In this model, the vehicle

was modeled as a 42 degrees of freedom multi-body system, the track was modeled as a

traditional three-layer discrete elastic support model, the rails were assumed to be

Timoshenko beans supported by discrete ties, which were modeled as Euler beam and the

ballast bed was treated as equivalent rigid ballast body. After validation using numerical

comparison with software SIMPACK, this model was proved that it considered the mutual

influence of the adjacent vehicles and tracks, and the fast dynamics calculation on a long

train running on an infinitely long flexible track was possible to be carried out.

There are very few analytical solutions for track transition problem, in which case

the track is resting on inhomogeneous foundation. Frohling et al. (1996) developed a

mathematical model to study the influence of spatially varying track stiffness on the

performance of the wheel and the track at low frequencies. Biondi et al. (2005) investigated

the 2D dynamic interaction among a running train, a track structure and a supporting bridge

where both rails and bridge were modeled as Bernoulli Euler beams. The work of Coelho

et al. (2009) showed the importance of hanging crossties in the transition zone due to large

settlement of embankment and the higher dynamic impact forces induced by passing trains

on these areas. Zhai et al. (2013) established a model for the analysis of train-track-bridge

dynamic interactions which also accounts for the continuity problem. Varandas et al. (2011)

and Varandas (2013) confirmed the main cause for the higher displacement amplitudes

24
registered on the transition zone was the existence of a group of consecutive hanging

crossties. A nonlinear model (see Figure 2.7) was presented for the interaction between

force and displacement of springs and also took into account the unloaded position of track

and ballast (including hanging distances). The model can simulate when the wheels moving

onto the trough existing after the culvert as a lever effect that will lift the rail/crossties

system before the culvert. Yang et al. (2013) studied the dynamic responses of bridge-

approach embankment transition section of HSR by establishing a spatial dynamic

numerical model (see Figure 2.8) of vehicle-track-subgrade coupling system based on the

vehicle-track coupling dynamic theory.

Figure 2.7 Train and Track Interaction Model (Varandas, 2013).

25
 Figure 2.8 Vertical dynamic analysis model of vehicle-track-subgrade coupled system
(Yang et al., 2013).

Indraratna et al. (2019) pointed out that a mathematical model based on beams on

elastic foundations was applied extensively in practice. The approach modeled the rail as

Euler-Bernoulli beam or Timoshenko beam laying on a Winkler foundation. Czyczula et

al. (2017) showed that both Euler-Bernoulli beam and Timoshenko beam gave almost the

same results.

2.5.2 Numerical modeling approaches

Compared to the advantages of modeling, laboratory scale experiments and field

testing are not commonly possible or the choice for studying behavior of a railway track

system due to long time commitments and often large efforts/costs associated. Yet

analytical models, such as the methods discussed in the previous section, are not always

accurate for the specific details and different geometries, which makes numerical modeling

a viable alternative for studying the track system under train loading. GEOTRACK (Chang

et al. 1980) is a layered elastic system program which has been validated and widely used

26
for track structural analysis. In GEOTRACK, stresses and deformations are calculated as

a function of multi-axle loads, properties of rails and ties, properties of substructure layers,

and geometry details of ties and underlying layers (Li and Selig, 1998). KENTRACK

(Huang et al. 1984) is a finite element-based trackbed structural design program that can

be utilized to analyze responses of granular ballast trackbed as well as asphalt trackbed and

slab track. Of course, these models also come with certain assumptions and several

limitations such that in some situations they may not be appropriate. As a limitation, for

instance, they do not consider dynamic response behavior of the track system.

To study the time-dependent behavior of track under dynamic loading, often the

Finite Element Method (FEM) has been adopted and used by researchers as the numerical

modeling method of choice (Feng 2011, Smith et al. 2006). Feng (2011) used FEM to

simulate different track systems with various analytical formulations and complexities,

such as beam on discrete support model, track with ballast mass model, and rail on crosstie

on continuum model. For individual cases, the displacements of trackbed were evaluated.

Numerical solutions indicated that railpad stiffness considered in the model had a major

effect on the resonance frequency. Wang and Valeri (2018) also applied FEM method for

short-term analysis on how differential settlement influences dynamic responses in

transition zones. In the analysis, subgrade, ballast and crossties were modeled as 3D elastic

elements, and the bridge was modeled as a two layer structure with support layer and

concrete slab, and all of these elements had linear isotropic elastic material properties.

Since the model validation approved to provide accurate results, the FEM model was

combined with an empirical settlement model to establish a long-term analysis, which

presents that the settled area would gradually propagate from locations close to the bridge

27
to farther locations. Wang and Valeri (2019) improved the numerical model by

incorporating rails and crossties as beam elements, fasteners as spring elements, ballast as

3D solid elements, vehicles as mass-spring system. To obtain more accurate results and

detailed ballast stress analysis, they modeled rails as Hughes-Liu beam elements, crossties

as 3D elastic bodies, and these elements had the linear isotropic elastic material properties.

Meanwhile, the contact between ballast and crossties was modeled as nonlinear with

employing the penalty method and considering the ballast stress determining the permanent

settlement in ballast.

At track transitions where hanging ties are common (see Figure 2.9), tracks

sometimes experience impact loading conditions and nonlinear track components.

However, only a few numerical models include these features (Smith et al. 2006,

Kaewunruen et al. 2011, Giner et al. 2012, Wang et al. 2015). Smith et al. (2006) conducted

a parametric study using a finite difference program FLAC3D, considering the effects of

velocity and stiffness of substructure materials. The limitation of this model is that the train

load was not distributed as in reality through the rail and ties. Kaewunruen et al. (2011)

used a nonlinear solver in STRAND7 to conduct numerical FEM simulations of track and

focused on hanging tie problem due to lack of support or excessive track settlement. The

model was able to evaluate the effect of voids on the dynamic responses of concrete crosstie.

The negative dynamic bending moments were enlarged by hanging ties and thus could

cause ballast to wear out sooner. Yin and Wei (2013) developed an FEM of railway bridge

transition zone using ABAQUS commercial software. The model included vehicle

modeling, bridge modeling, as well as open track modeling. Track irregularity was also

included in the model. More recently, a dynamic FEM model using explicit integration of

28
the track transition zone was developed (Wang et al., 2015) to model transition zones with

differential settlement or hanging ties. This model was validated through field

measurements on a ballast track. Kouroussis et al. (2015) studied the influence of rolling

stock dynamics on ground-borne vibration levels in a numerical way. To accurately model

the rail unevenness and other irregularity on the rail/wheel surface, this study first

constructed a numerical model with a flexible track on which the vehicles traverse at a

constant speed, then using a two-step approach to simulate ground wave propagation. In

this model, multibody formalism was used for vehicle modeling, while 2D approach was

preferred, the rail was modeled by a Euler-Bernoulli beam, the rail masses were modeled

using Timoshenko formulation, and the vertical wheel/rail contact forces calculation used

Hertzian theory. When Kouroussis et al. (2017) studied railway vibration, a combined

numerical/experimental prediction method was used. The vehicle/track/foundation model

consisted of a classical multibody vehicle coupled to a finite element/lumped mass track

model. The additional experimental setup was applied to help solve the complexities in

numerical modeling caused by wheel impact local defects. The 3D dynamic model

established by Koch and Hudacsek (2017) used the PLAXIS 3D Dynamic package. Besides

modeling the rail as beam element, applying standard fixities and absorbent boundaries,

and setting certain values of material properties of crossties and rails, PLAXIS 3D used a

time-force signal to define the dynamic loads. There was a multiplier for each single

dynamic point load. But the model with different structure types and soil conditions should

be implemented with cyclic degradation capability to approve it can produce predictions

on the long-term track behavior. Wilk (2017) also developed a numerical model for LS-

DYNA analyses that simulates a train truck entering an open deck bridge. The tie-ballast

29
gaps are considered in this model. Simulation results show that the differential stiffness

between bridge deck and transition zone leads to 20% increase in dynamic wheel load.

Also, evenly increasing the ballast settlement in the transition zone produced tie-ballast

gaps and increased tie loads by over 100%; while unevenly increasing the ballast settlement

in the transition zone also produced tie-ballast gaps and concentrated the wheel load on the

most well supported ties.

Figure 2.9 Degradation of Track Transition Zone (Wang et al., 2015).

2.6 Remedial Measures to Mitigate Differential Movement at Track

Although the problem of “bump” development at highway bridge approaches has

been extensively studied (Zaman et al. 1991, Briaud et al. 1997, White et al. 2005, and

Puppala et al. 2009), only limited research studies have focused on mitigating the

differential movement at railway track transitions. These few studies investigating track

transition problems have primarily focused on preselected mitigation techniques and

presented test section and parametric analysis results on the effectiveness of these remedial

measures (Sasaoka et al. 2005, Read and Li 2006, and Nicks 2009).

Several different remedial measures have been suggested by researchers to mitigate

differential movement problems at track transitions. Nicks (2009) divided the remedial

measures aimed at mitigating bump development at railway bridges into the following

30
interrelated categories: (1) reduce approach settlement, (2) decrease modulus on bridge

deck, (3) increase modulus on approach track, (4) reduce ballast wear and movement, and

(5) increase damping on the bridge deck. Kerr and Moroney (1993) concluded that most

problems at track transitions arise from rapid changes in the vertical acceleration of wheels

and cars in the transition zone. Accordingly, all remedial measures were recommended to

aim at reducing the train vertical acceleration at the transition zones, and could be divided

into the following three categories (Kerr and Moroney 1993): (a) smoothing the track

stiffness (often represented as ‘k’) distribution on the “soft” side of the transition; (b)

smoothing the transition by increasing the bending stiffness of the rail-tie structure on the

“soft” side, in close vicinity of the transition point; and (c) reducing the vertical stiffness

on the “stiff” side of the transition. Remedial measures under category (a) include use of

oversized ties, reduced tie spacing, ballast reinforcement using geogrids, hot-mix-asphalt

(HMA) underlayment, and use of approach slabs. On the other hand, the most commonly

known method under category (b) was developed by the German Federal Railways (DB)

and involves attaching four extra rails (two inside and two outside the running rails) to the

cross ties (Kerr and Bathurst 2001). Finally, the primary approach in category (c) involves

the installation of tie pads and/or ballast mats on ballast-deck bridges to reduce the track

stiffness on the “stiff” side of a transition point. Besides HMA underlayment, according to

Moale et al. (2016), Geoweb can be adopted as an alternative as it can reduce cost while

increase the roadbed strength and provide good drainage.

Chawla and Shahu (2016) found that the geogrid reinforcement under ballast layer

could significantly mitigate tie displacement only at low values of subgrade modulus and

effective subgrade shear strength parameters. It was equally effective at mitigating tie

31
displacement within the sub-ballast thickness from 450 to 1000 mm. The geotextiles at

subgrade surface could facilitate quick in-plane drainage and dissipation of pore water

pressure which results in lower mud pumping potential.

Using analytical procedures, Kerr and Moroney (1993) engineered pad stiffness to

“match” the track running over the bridge to the approach track. Accordingly, a later study

(Kerr and Bathurst 2001) installed these “matched pads” on three different open-deck

bridges near Chester, PA, Catlett, VA and Philadelphia, PA in the US. Field test results

subsequent to the installations indicated significant improvements in track geometry near

the bridge abutments. Sasaoka and Davis (2005) tried different methods to alter the track

stiffness and damping characteristics on bridge approaches. Installing ties made of different

materials, they reported that plastic ties on a concrete span ballasted-deck bridge effectively

reduced the stiffness difference at track transitions. Through parametric analyses using the

GEOTRACK multi-layered elastic program, they concluded that subgrade improvement in

the approach and altering tie pad properties on the bridge deck were the most effective

methods to minimize track stiffness differences at bridge approaches. Similarly, from

dynamic analyses using NUCARS, they concluded that providing extra dampers on the

bridge deck could improve the impact attenuation at the transition by up to 30%. Li and

Davis (2005) concluded that remedies intended to strengthen the subgrade were not

effective for sites where ballast/subballast layers were primarily responsible for the

differential movement. In such cases, mitigation techniques such as rubber pads under the

concrete ties, or rubber mats on the concrete bridge deck had to be used to reduce the track

stiffness and enhance the damping characteristics.

32
 Rose and Anderson (2006) presented asphalt underlayment trackbeds as an

effective method for improving the performance of track transitions at tunnels, bridge

approaches, special trackwork like crossing diamonds, crossovers and turn-outs, as well as

at rail/highway at-grade crossings. Placement of a thicker HMA underlayment adjacent to

the bridge and a thinner section close to the existing all-granular trackbed reportedly

improved performances of both open-deck and ballast-deck bridges. Rose and Anderson

(2006) reported four bridge approaches that were rehabilitated using this technique along

a Kentucky mainline with over 50 million gross tons (MGT) annual tonnage and a line

speed of 80-96 km/h. Over five years since the renewal of these approaches, no resurfacing

was needed to correct track geometry.

Johnson and Arzamendi (2015) used an innovative material, Lightweight Cellular

Concrete Fill (LCCF) with ground improvement to mitigate railroad bridge approach

settlement. Their project was conducted over San Diego River with a double track bridge,

the maximum settlement with normal weight fill is 4.8’’ while it’s only 1.9’’ with LCCF.

The maximum settlement with normal weight fill and compaction grouting was 1.4’’ while

it was only 0.5’’ with LCCF and compaction grouting. Therefore, lightweight fill is proven

useful for settlement mitigation although innovative materials can take significant time and

effort for approval.

Apart from the above listed remedial measures, researchers have also suggested

converting open-deck bridges to ballast-deck ones (Hyslip et al. 2009), constructing

approach slabs (Sharpe et al. 2002), and applying chemical grouting (Woodward et al. 2007,

Hyslip et al. 2009) and stone blowing (Chrismer 1990, McMichael and McNaughton 2003)

as alternatives to mitigate track transition problems. Bridge approach slabs have been used

33
as a semi-structural method to ease the transition from approach embankments to the fixed

bridge structure (Sharpe et al. 2002). However, concrete approach slabs require good

connection at the abutment and good support conditions away from the abutment to be

effective so they are not being widely used. Hyslip et al. (2009) proposed chemical grouting

as a conceptual solution for bridge transition improvement.

Background and previous research study details related to chemical grouting, under

tie pads and stone blowing applications are listed next in Sections 2.6.1 to 2.6.3.

2.6.1 Chemical grouting

Chemical grouting, also known as polyurethane stabilization (see Figure 2.10), has

been successfully used in the past to reduce excessive ballast vibrations and long-term

permanent deformations. Polyurethanes are extremely large complex molecules, also

known as polymers, produced by combining a large number of simpler molecules called

monomers. The lower the closed cell content in polyurethanes, the more the material acts

like a flexible foam. In the case of a rigid-foam, where mechanical properties such as high

strength and stiffness are intended, for example, for ballast reinforcement, a high content

closed-cell foam is ideally preferred. When it is reacted properly, rigid-foam polyurethanes

can have a largely closed-foam structure (Szycher, 1999). In terms of maintenance

schedule and easiness, polyurethane stabilization method is economically feasible due to

its negligible curing period and less interruption of traffic. Usually, no track shutdown is

required. Woodward et al. (2007) highlighted that when properly applied, chemical grouts

had the ability to bind the ballast particles together, thus reducing excessive vibrations and

particle migration. Moreover, the polyurethane chemical grouting application reportedly

does not adversely affect the drainage ability of the ballast layer.

34
 Figure 2.10 Polyurethane Grouting of Ballast.

Laboratory testing and evaluation of polyurethane stabilized ballast has been

conducted extensively. Dersch et al. (2010) evaluated the effectiveness of elastomer

polyurethane coating of ballast, with the use of urethane only 1-2% by weight of dry ballast

aggregate, through a direct shear box test. Polyurethane coated ballast samples formed rigid

blocks and resulted in two to three fold increases in the shear strength while the amount of

ballast breakdown during testing decreased due to less particle movement and reorientation

under shear loading (Dersch et al. 2010 and Boler 2012). Boler (2012) further studied

ballast aggregates treated with polyurethane by the use of an image-aided particle shape

modeling and particle packing simulations using the Discrete Element Method (DEM) and

showed that shear strength increases of the urethane treated ballast assembly was

significantly influenced by the bond strengths of the particle contacts. Keene and Edil

(2012) tested specimens of polyurethane stabilized ballast under repeated loading. The

35
results showed that polyurethane stabilized ballast had good resistance to accumulation of

plastic strain compared to untreated substructure materials. Also, from the flexural strength

tests conducted on polyurethane stabilized ballast, strength properties (approximately 938

kPa) were close to that of cement stabilized soils. However, flexural modulus of

polyurethane stabilized ballast was much lower than that of cement stabilized soils. The

bonding and tensile capacity of polyurethane stabilized ballast are believed to provide

longer life cycles and reduced maintenance requirements.

Besides laboratory testing, field testing and full-scale testing were also conducted

by researchers. Woodward et al. (2012) evaluated the polyurethane ballast reinforcement

technique through laboratory tests under cyclic loading and then based on the good

performance from experiments, through a case study in Grovehill Tunnel, U.K. With

proper injection of polyurethane to achieve desired stiffness, strength, and penetration

depths, gauge clearance issues on ballasted railway track could be maintained. Full scale

testing was conducted by Kennedy et al. (2013) to compare the traditional ballast and

polyurethane reinforced ballast behavior trends for up to a maximum of 500,000 load

cycles at different loading levels. Significant reduction in track settlement was achieved

with the polyurethane reinforcement. Polyurethane treated ballast railway track gave

almost a similar performance to slab track with low track settlement.

Recently, Warren (2015) used an expanding rigid polyurethane foam to reinforce

an in-service railroad track in the State of Illinois, where track settlement and progressive

shear failure were found to be predominant on site. Though field settlements were not

measured, their study mainly focused on life-cycle cost analysis and the analyses related

to maintenance cycles of both the stabilized ballasted track and traditional track. Their

36
results showed that over a 10-year period, polyurethane injection could save them $1,000-

$29,000 for maintenance.

Mishra et al. (2016) conducted an experiment to decrease differential settlement at

two bridge approaches along Amtrak’s NEC near Chester, Pennsylvania. According to

Figure 2.11, the peak transient deformation recorded in the ballast layer decreased

significantly after chemical grouting. It could be concluded that chemical grouting proved

to be effective as a remedial measure to reduce ballast transient deformation, improve

support conditions under tie and reduce excessive vibrations and particle migration within

the ballast layer.

Figure 2.11 Effect of Chemical Grouting on Ballast Transient Deformation (Mishra et al.
2016).

37
2.6.2 Under tie pads

To maintain a smooth transitional stiffness along the track between regular track

and bridge deck and to slow down the track deterioration at the bridge approach, one of the

most commonly studied techniques is to use Under Tie Pads (UTPs), also known as Under

Crosstie Pads or Under Sleeper Pads (Lundqvist et al., 2006). UTPs are resilient pads

attached to the bottom surface of ties to provide an elastic layer between tie and the ballast.

They have been used in track structures for almost 25 years. According to Lundqvist et al.

(2006), rail-tie-ballast contact force variations can be minimized by controlling stiffness

using UTPs.

UTPs have been tested in laboratory by many researchers and practitioners (Le Pen

et al. 2018, Grabe et al. 2015). It is believed that with the use of UTPs, the contact area

between tie and ballast can be increased for concrete ties from around 3% to 10-35%

depending on the pad category. France and Austria utilizations of under tie pads proved

that tamping frequency can be lowered and therefore maintenance cost is greatly reduced

(Fimor 2015). Grabe et al. (2015) conducted full tie and half tie tests of UTPs under various

loading conditions such as static, dynamic, and cyclic loading and used tactile surface

sensors to measure the pressure under the tie. The results show that the use of UTPs can

provide a reduction in settlement by approximately 27-44%, increase of contact area by

around 10-33%, reduction of contact pressure by 70%, and reduction of ballast breakdown

by 10-40% (Grabe et al. 2015).

Field installations of UTPs have generally proven to be successful. Li and Davis

(2005) conducted preliminary investigations on the benefits of using rubber pads under

concrete ties in a section of test track at the Transportation Technology Center (TTC) in

38
Pueblo, Colorado. The changing of track modulus and damping showed that UTPs not only

reduced track stiffness, but also increased track damping. UTPs could also reduce impacts

between ballast particles and concrete ties or bridge surfaces. Two different types of rubber

pads used as UTPs were tested by Sasaoka and Davis (2005). The pads were adhered to

the bottom of concrete ties and installed on both a concrete span bridge and a steel beam

span bridge. They concluded that both types of pads were successful at lowering the bridge

modulus to below that of the approach track as well as providing damping for the bridge

structure. Sasaoka and Davis (2005) suggested using durable material with long service

life as UTPs since it had to resist abrasive action of the ballast and exposure to elements

such as water and oxygen.

Full-scale field tests were conducted in Switzerland to investigate the influence of

UTPs (Schneider et al. 2011). The UTP-installed track typically generated higher rail and

crosstie accelerations, but lower UTP strains. UTPs could also slow down geometry

degradation. Recently, Li and Maal (2015) reported performance of UTPs installed on a

ballast deck concrete tie bridge back in 2007 and subjected to heavy axle load traffic since

then. The bridge did not need any localized repair or major track maintenance since 2007

although maintenance efforts were required on a quarterly basis on this bridge before.

Moreover, high amplitude vibrations were significantly reduced. Pen et al. (2015) used

geophones (velocity transducers) to measure track substructure responses in a UK site to

conclude the potential benefits of UTPs such as to: (1) increase the number and area of

contacts; (2) reduce the rate of plastic settlement; (3) reduce the support stiffness and spread

the load along a greater length of track; and (4) and add in a consistent increment to the

track deflection and reduce support stiffness variation.

39
 UTPs usually consist of two layers: one elastic spring layer and a geotextile to

protect the spring layer. To study the properties and effectiveness of under tie pads

numerical models have been developed. For example, a parametric study was conducted

by Johansson et al. (2008) with different combinations of dynamic stiffness properties of

rail pads, under tie pads, and ballast materials and the results were analyzed through

simulated responses using two numerical models. Schematic drawing of one of the

analytical models, known as GroundVib, is shown in Figure 2.12. Results showed that

under tie pad stiffness only influenced the lower part of the frequency spectrum (lower than

250 Hz). Higher under tie pad stiffness was found to lead to lower rail bending moment,

and lower vertical rail displacement and velocity. No clear relationship between under tie

pad stiffness and rail acceleration was found. However, higher under tie pad stiffness would

also lead to higher load on tie from the ballast interface.

Figure 2.12 Schematic Analytical Model GroundVib (Johansson et al., 2008).

40
 Witt (2008) examined the influence of UTPs using the FEM. Three types of UTPs

(stiff, medium, and soft) were used. Results showed that stiff UTPs did not influence

transition area load. The peak load caused by hanging crosstie could even increase when

using stiff UTPs as compared to using no pads. Similarly, when using soft UTPs, the force

on transition area did not exhibit large difference with no pad areas. The best results came

from using medium stiffness UTPs. Small variations of contact forces occurred and the

peak caused by the transition area disappeared. Insa et al. (2011) also performed

simulations to study the effectiveness of using UTPs. Obtained results showed that UTPs

contributed significantly to stiffness reduction by lowering the vertical stiffness of track on

bridge. Ribeiro et al. (2015) used FEM modeling to show that soft UTPs would amplify

rail displacement, tie accelerations, and reduce abrupt variations in track vertical stiffness.

Despite many of the successful installations of UTPs in different countries and field

applications, some researchers emphasize the importance of more accurate knowledge still

needed to understand behavior of UTPs for practical implementation of transition designs

since the variations in stiffness and quality are crucial to track settlement (Witt 2008,

Ribeiro et al. 2015). Paixao et al. (2015) reported their findings of two transitions

monitored which had similar box culverts. UTPs were added in one of the transitions.

Comparison results indicated that the UTPs influenced dynamic behavior, increasing

vertical flexibility and amplifying both rail displacements and crosstie accelerations. The

installation of UTPs, in contrary to the original expectation, did not provide a smoothing

stiffness of transition after all. Hence, Paixao et al. (2015) indicated that UTPs had a

significant influence on track dynamic behavior that was yet to be carefully studied and

41
well understood to design properly with its resilient properties as well as for the geometry

requirements and other arrangements along the transition zones to achieve expectations.

2.6.3 Stone blowing

Stone blowing, also referred to as ‘Stoneblowing’ or ‘Stone Injection’, is one of the

potential remedial measures at bridge approaches. It consists of the addition of stone (or

ballast) to the surface of existing ballast such that any gap between tie bottom and ballast

layer is closed. Originally developed by British Railways, the process of stone blowing

involves the following steps (Selig and Waters, 1994): (1) The geometry of the existing

track is measured; (2) the precise track lift required at each crosstie to restore it to an

acceptable geometry is calculated; (3) the volume of stone that needs to be blown beneath

the crosstie to achieve such a lift is deduced from the known relationship between volume

of added stone and residual lift; and (4) the track is applied stone blowing. For stone

blowing, the target elevation during stone injection is always set “higher” than the desired

elevation so that enough space is provided to have the stones adequately distributed under

the tie. This leads to a smooth transition between the bridge approach and the bridge deck.

This is because the freshly injected stones underneath the ties usually undergo some

amount of settlement under train loading. The stone matrix tends to attain a “stable”

configuration through particle rearrangement and packing. It is therefore common for the

track profile to gradually “move down” with loading immediately after stone blowing until

a stable configuration is attained. The procedure can be illustrated as follows in Figure 2.13.

42
 Figure 2.13 Stone Blowing Procedure (Selig and Waters, 1994).

According to Figure 2.14 from Mishra et al. (2016), the peak transient deformation

recorded in the ballast layer decreased significantly after stone blowing. The conclusion

was that stone blowing could be proved to be effective as a remedial measure for reducing

ballast transient deformation, and it could also effectively reduce excessive rail lift-off

often caused by inadequate support conditions under ties.

43
 Figure 2.14 Effect of Stone Blowing on Ballast Transient Deformation (Mishra et al.
2016).

2.7 Summary

Track geometry problems at railway transitions have been well identified. All

design standards including European Rail Research Institute (ERRI), UIC code 719,

Design Standards for Railway Structures and Commentary, and AREMA involve the

designs of transition zones (Muramoto 2013). Although railway track transitions include

tunnels, special track work, and highway/rail at-grade crossings, the most common

transition problems are often encountered at bridge approaches. Due to drastic differences

in substructure and loading conditions, the tracks on a bridge deck undergo significantly

small deformations under loading compared to the approach tracks. This sudden change in

track deformation behavior at the transition point results in extreme dynamic loading

conditions and ultimately leads to rapid deterioration of the track and bridge structural

components.

44
 This chapter first provided an introduction to railway track systems with an

explanation of track transitions zones and special emphasis on bridge approaches. Previous

research studies that focused on track transition zones were reviewed to identify major root

causes of the rough train ride and other cause-effect relations. Field instrumentation efforts

for monitoring railway track transitions with the intention to identify main causes were

summarized and practical solutions adopted by researchers and practitioners were listed as

remedial measures. Existing analytical and numerical models established for railway track

modeling under either static or dynamic loading conditions were also reviewed for

effectiveness. In the end, mitigation solutions were presented, and three chosen remedial

measures identified in the current research effort, i.e., chemical (polyurethane) grouting,

under tie pads, and stone blowing, were discussed in detail.

45
CHAPTER 3: QUANTIFICATION OF RAILROAD BRIDGE
APPROACH PERFORMANCE

The first task in this PhD study involved identification of problematic track bridge

approach locations that experienced differential movement problems for instrumentation

and performance monitoring, and conducting field instrumentation at selected locations.

By acquiring data from instrumented bridge approaches, ground truth validation cases were

established for the dynamic track model developed in this study. This chapter describes

these ground truth field studies and also deals with statistical analyses conducted on the

collected field data to quantify track transient response and performance trends at the

studied bridge approaches.

3.1 Field Instrumentation

3.1.1 Instrumentation location selection

Approximately 18 miles south of Philadelphia near Chester, Pennsylvania, a

problematic portion of Amtrak’s Northeast Corridor (NEC) comprises 8 to 10 closely-

spaced bridges with recurring differential movement problems at the bridge-embankment

interfaces. The NEC is primarily a high-speed passenger railroad with occasional freight

traffic. On the NEC, the high-speed Acela Express passenger train operates up to a

maximum speed of 241 km/h (150 mph). This segment of NEC near Chester consists a

total of four tracks. Two of the tracks are maintained for high-speed passenger trains,

typically operating at 177 km/h (110 mph). The current research focused on monitoring the

track response and performance data due to Acela passenger trains operating on the 2nd and

3rd main lines. Acela passenger train weights or wheel loads, were relatively consistent

46
over time periods and the train geometry data were known. Thus, ACELA train loading

related instrumentation data were used to validate the proposed track-train-bridge model.

Historical geometry data were obtained from Amtrak spanning a 60-month period

from January 2005 to January 2010. In June 2012, Ground Penetrating Radar (GPR)

scanning along the track was conducted to identify substructure features. The analyses of

track geometry data and GPR data led to three worst bridge approach locations as far as

recurring differential movement problems were concerned, including bridges over Upland,

Madison, and Caldwell Streets. These three locations with frequent maintenance needs

were selected for field instrumentation, shown in Figure 3.1.

Figure 3.1 Three Instrumentation Locations on Google Map.

Figure 3.2 represents ground penetrating radar (GPR) as well as geometry data for

Upland St. location. Note that GPR and geometry data for Madison St. and Caldwell St.

locations can be found in Appendix A. From Figure 3.2, significant deviations in the

vertical profile are clearly identified, shown in red color. It is indicated that a “bump” exists

47
at a near bridge location about 15 ft. from the north abutment. At 60 ft. from north abutment

in the open track location, the vertical profile becomes smoother as regular track is situated

away from transition zones. Therefore, the instrumentation locations were chosen at 15 ft.

(near bridge) and 60 ft. (open track) for monitoring track performance. Note that the “bump”

shown in Figure 3.2 is more obvious at bridge “entrance” where the train moves from the

approach onto the bridge deck, rather than exiting the bridge.

Figure 3.2 (a) GPR Scans, (b) Mid-Chord Offset Data, (c) Space Curve at Upland Street
Bridge Approach.

3.1.2 Instrumentation details

The instrumentation of Amtrak NEC bridge approaches took place in July-August

of 2012. Multidepth Deflectometers (MDDs) were selected to install and monitor the

movement of individual track substructure layers. As stated earlier, the MDD technology

was first developed in South Africa in the early 1980s to measure individual layer

48
deformations in highway pavements (De Beer et al., 1989). MDDs typically consist of up

to six linear variable differential transformers (LVDTs) installed vertically at preselected

depths to measure the displacements of individual substructure layers with respect to a

fixed anchor buried deep in the ground (Scullion et al. 1989). The use of MDDs to monitor

railway track performance has been extensively pursued in South Africa (Grabe and Shaw

2010, Priest et al. 2010, Vorster and Grabe 2013) as well as in the U.S. (Sussmann and

Selig 1998, Bilow and Li 2005).

Figure 3.3 illustrates the location of each MDD module typically placed at layer

interfaces in the track substructure. Note that the LVDT #1 was installed at the bottom of

the crosstie. The rest of the LVDTs (#2 - #5) were installed at track substructure layer

interfaces. The anchor is placed sufficiently deep in the ground (10 ft. below the crosstie)

which is assumed to be rigid bottom. Therefore, the assumption of calculating deformation

in each layer is that the anchor does not move with time under train loading. According to

the results from geometry data and GPR data, a total of six MDD strings were installed at

the Amtrak NEC bridge approaches, two at each selected bridge approach. MDDs were

installed at 15ft. and 60ft. from the North abutment at Upland Street location; 12ft. and

60ft. from South abutment at Madison Street location to monitor both near bridge and open

track locations, respectively. At Caldwell Street location, two MDDs were installed at 80

ft. from the south abutment on the left and right ends of the same time to monitor open

track performance.

49
 Figure 3.3 Schematic of Track Substructure Profile with MDD Modules Placed at
Individual Layer Interfaces

Figure 3.4(a) shows a photo of the individual LVDT modules used in the MDD

system. The LVDTs were of 4-mm full range measurement and specially developed for

MDD usage. Figure 3.4(b) illustrates the independent anchoring system of MDDs. The

inner core of bottom-most LVDT (No.5) is mounted directly to the bottom anchor. The

core for LVDT module No. 4 is directly mounted on module No.5. The same pattern is

followed for the rest of LVDTs. Hence, except for the bottom-most MDD module, all other

MDD modules have movable anchors. The deflection measured by each LVDT represents

the deflection in that particular layer. Note that as LVDT module No.1 is attached to the

bottom of crosstie, LVDT 1 measures the total deflection of ballast and crosstie.

50
 LVDT # 1
d1
Core # 1

LVDT # 2
d2
Core # 2

LVDT # 3
d3
Core # 3

LVDT # 4

d4 Core # 4

LVDT # 5

Core # 5

d5

(a) (b)

Figure 3.4 (a) Photo of LVDT Module and (b) Independent Anchoring MDD System.

The drilling process before installation of MDDs was carried out in small

increments of 75-100 mm. Compressed air and high capacity vacuum cleaner were used to

clean the drilled hole. One of the primary tasks during the drilling process was to identify

the locations of track substructure layer interfaces. This would facilitate the mounting of

individual LVDT modules at the layer interfaces to measure the deformations of individual

track substructure layers. Through visual inspection of soil samples in the field, layer

interfaces were established upon noticing significant differences in the material type

removed from the drilled hole. Figure 3.5 shows the track substructure layer types and

51
thickness encountered at the Upland Street instrumentation location (sketches for Madison

and Caldwell Street are presented in Appendix A).

(a) (b)

Figure 3.5 Substructure Layer Profile for a) Upland Street 15 ft. from the North
Abutment b) Upland Street 60 ft. from the North Abutment

Besides the MDDs, strain gauges were also mounted on the rail at the instrumented

sites to measure vertical wheel loads applied during the passage of a train. Dual-element

350 Ohm shear gauges were welded on the rail at the neutral axis. After installation, a

calibration frame was used to correlate applied vertical load levels to the voltages induced

in the strain gauges by use of a Wheatstone bridge circuit. Figure 3.6 shows the photos of

the strain gauges during and after installation. Note that strain gauges were placed on two

rail sections, named as Type I strain gauge and Type II strain gauge. The Type I strain

gauge measured wheel load on rail on top of crosstie; while the Type II strain gauge

measured load on rail in between crossties.

52
 Figure 3.6 Photos of Strain Gauges Welded on Rail in the Field.

3.2 Exploratory Analyses of Field Data at Bridge Approaches

In statistics, exploratory data analysis represents an approach to analyze data by

summarizing the main characteristics, often with visual methods and observations. In this

section, exploratory analysis is conducted on field data including loadings measured by

strain gauges and vertical deformations measured by MDDs.

3.2.1 Field loading conditions

At the instrumented Amtrak NEC bridge approaches, the predominant train traffic

is Acela Express passenger train, operating at 177 km/h (110 mph). Acela Express trains

consist of two locomotives, one at each end of the train and six passenger cars in between.

Figure 3.7 shows the bogie, axle, and wheel spacing for the Acela Express trains. As shown

in the figure, wheel spacing for Acela locomotive is approximately 2.8 m; and wheel

spacing for Acela passenger car is approximately 3.0 m.

53
 Figure 3.7 ACELA Locomotive and Passenger Car Axle Spacings.

3.2.2 Track settlement trends

Monitoring the structural response and performance trends at bridge approaches

involved periodic data acquisition from the instrumented locations to record permanent

deformations of track substructure layers. Voltage changes registered by each LVDT

represent movements in that corresponding substructure layer. Figure 3.8 presents the

measured track substructure settlement trends from 2012 to 2015 at Upland Street

instrumentation location. Each marker on the figure represents one data collection. It is

noticed that the top layer (LVDT No.1) at near bridge location (15 ft. from abutment)

experienced significantly higher permanent deformation or settlement over time compared

to that at open track location (60 ft. from abutment). Maintenance was more frequently

carried out at near bridge locations to maintain proper vertical profile. Layers 2 to 5 show

small settlement accumulations over time at both locations.

54
 (a)

(b)

Figure 3.8 Layer Settlement Trends at Upland Street Location (a) Open Track (60 ft.
from abutment) (b) Near Bridge (15 ft. from abutment)

55
3.2.3 Transient responses under train loading

Besides periodic measurement of settlements within each track substructure layer,

transient responses at instrumentation locations were also measured to quantify the elastic

bounce back behavior of the substructure layers under loading. The transient layer

deformations and wheel load data collected in the field are presented next. The Upland

Street location was chosen as an example to analyze the dynamic responses at railroad

bridge approaches. Detailed transient response data measured at Madison and Caldwell

street locations can be found in Appendix B.

Within 3 years from 2012 to 2015, seven data collection trips were organized at

Amtrak NEC bridge approach locations at the Upland Street site. Figure 3.9 shows the

transient vertical deformation time history recorded by LVDT modules under a passing

train at open track location (60 ft. from abutment). The recorded deformations

corresponded to the 8-car Acela Express passenger train, including one locomotive at each

end of the train and six regular passenger cars. A total of 32 peaks were observed from the

time history. These peaks corresponding to the passage of each wheel over the

instrumented tie are quite distinguishable. Note that the locomotives created higher

deformations compared to passenger cars since locomotives are heavier. Moreover, it is

important to note that at the open track location, the transient responses collected at

different times are similar in magnitude. The maximum vertical transient deformation

registered under Acela Express train was approximately 1.7 mm.

56
 Figure 3.9 Upland St. Open Track Transient Vertical Deformations with Train Passage

Figure 3.10 shows the transient load time history recorded by strain gauges under

a passing train at the open track location (60 ft. from abutment). Note that the strain gauge

circuit installed on rail crib location clearly registered 32 peaks, corresponding to the 32

wheels in an Acela Express train operating along the NEC with two locomotives (one each

at the front and rear ends of the train) and six passenger cars. As expected, the locomotives

register higher load levels. At the open track location, the transient loadings collected at

different times are similar in magnitude. The highest measured load level at the open track

location is approximately 135 kN.

57
 Figure 3.10 Upland St. Open Track Transient Loads on Crib with Train Passage

Besides the open track location, field data were also collected at the near bridge

location (15 ft. from abutment) for performance measurement. Figure 3.11 shows the

vertical transient deformation time history from 2012 to 2015 the at Upland Street near

bridge location. Unlike the open track data, the near bridge data showed great variability

in magnitudes of transient deformations collected over time. The maximum vertical

deformation continued to increase from August 2012 (2.5 mm) to July 1st 2014 (8 mm).

Next, the maximum vertical deformation suddenly decreased to 1.6 mm on July 22nd of

2014, then up to 4 mm in May 2015. This sudden decrease in deformation was due to

tamping maintenance activity conducted at the Upland Street location. In addition, a clear

trend of increasing dynamic response was observed with time at the near bridge location.

From visual inspection, it is noticeable that open track and near bridge locations have

significantly different response trends under train passage.

58
 Figure 3.11 Upland St. near Bridge Transient Vertical Deformations with Train Passage

Similarly, Figure 3.12 shows the transient load time history recorded by strain

gauges under a passing train at the near bridge location. There are 32 peaks, corresponding

to the 32 wheels in an Acela Express train. The transient loads collected at different times

at the near bridge location are similar in magnitude. The highest load level is approximately

150 kN, which is higher than that recorded at the open track location (same train passage).

This is due to the fact that strain gauge captures both static and dynamic loading effects.

Though static loading of the train contributes the same amount of magnitude to the open

track and near bridge locations, dynamic impact load coming from the wheel-rail impact,

irregularity, etc. is higher at the near bridge location. Considering the relatively stable

magnitude in load levels from 2012 to 2015, it is anticipated that the increasing trend of

transient deformation (Figure 3.11) is mainly associated with change in track substructure

properties (weaker support over time). This finding is confirmed by previously presented

measured settlement data, indicating that the substructure layers develop larger permanent

59
settlement over time at near bridge locations, creating a gap between layers which caused

larger transient displacements under train passage.

Figure 3.12 Upland St. near Bridge -Transient Loads on Crib with Train Passage

3.3 Statistical Analyses of Structural Performance Trends at Bridge Approaches

In addition to the exploratory analyses of field transient deformation and loading

data over time, as illustrated in Figures 3.9-3.12, statistical analyses were also conducted

for (i) determining the maximum/peak deformation and (ii) estimating the 95% confidence

interval (CI) of the field data collected. Again, Upland Street field data are presented as an

example herein. Statistical analyses of transient responses for Madison and Caldwell street

bridge approaches can be found in Appendix B.

3.3.1 Peak transient deformation at Upland Street

The peak transient deformation corresponding to locomotive wheel passage was

determined from each dataset collected and graphed in Figure 3.13. Each marker represents

one dataset collected in the field. The dashed blue line with the triangle markers illustrates

60
the peak transient deformation over time at the near bridge location. The solid orange line

with the round markers shows the peak transient deformation over time at the open track

location. The dotted lines show the 95% CI for the two locations, respectively.

From Figure 3.13 the peak deformation under locomotive passage at the open track

location is nearly constant from 2012 until 2015. On the contrary, the peak deformation

under locomotive passage at the near bridge location is changing over time, and thus the

deviation is larger, leading to a larger range for confidence interval. The statistics of the

field data, including mean, standard deviation and 95% CI, of the peak transient

deformation at both locations are summarized in Table 3.1. Based on the statistical analysis,

it is concluded that the vertical deformation behavior at the near bridge and open track

locations are statistically different because the mean of one location does not fall within

the 95% confidence interval of the other location. This implies the difference in dynamic

responses at these two locations was not due to random error.

9
Nearbridge
Jul 1-14
8 Opentrack
7
Peak Displacement(mm)

Jun-13
6

5
Jan-15
4 Jul 22-14 May-15
Jan-13
Aug-12
3

2
Aug-12 May-15
1 Jan-13 Jun-13 Jul 22-14 Jan-15
Jul 1-14
0
4/1/12 10/18/12 5/6/13 11/22/13 6/10/14 12/27/14 7/15/15
Time

Figure 3.13 Upland St. Peak Vertical Deformation over Time (2012-2015)

61
 Table 3.1 Statistics of Peak Vertical Deformation (2012-2015)

Open Track (mm) Near Bridge (mm)

Mean 1.63 4.03

Standard Deviation 0.07 2.10

95% CI High End 1.71 5.98

95% CI Low End 1.55 2.09

3.3.2 Peak transient load measured on top of crosstie

Besides the deformation, the loadings on track structure were also recorded and

analyzed statistically. Note that strain gauges were mounted on two rail sections to measure

two types of loading. One is located on rail where there is crosstie support underneath

(Type I strain gauge). The other measures the force on the crib of the rail, which equals to

the dynamic loading on top of rail (Type II strain gauge). The peak load corresponding to

locomotive wheel passage was selected from the dataset and graphed with time in Figure

3.14. The dashed blue line with the triangle markers illustrates the peak load measured by

Type I strain gauge over time at the near bridge location. Solid orange line with round

markers represents the peak load measured by Type I strain gauge at the open track location.

The dotted lines are the high and low limits of the 95% CI for the two locations,

respectively.

The loading time history does not reveal clear trends as in the case of the

deformation time history (see Figure 3.13). The maintenance activities occurred in July

2014 at the near bridge location is associated with a decrease in force under locomotive

passage. The statistics of the peak force measured by Type I strain gauge are summarized

62
in Table 3.2. The results indicate a statistically significant difference in the rail load on top

of crosstie at two locations.

140
Aug-12 Jul 1-14
Jun-13
120 Jan-13 Jan-15
May-15
Peak Force on Tie (kN)

100
Jul 22-14
80

60
Aug-12 Jul 22-14 May-15
Jan-13 Jun-13 Jan-15
Jul 1-14
40
Nearbridge
20 Opentrack

0
4/1/12 10/18/12 5/6/13 11/22/13 6/10/14 12/27/14 7/15/15
Time

Figure 3.14 Upland St. Peak Force on Tie of Locomotives over Time (2012-2015)

Table 3.2 Statistics of Peak Vertical Load Measured on Top of Tie (2012-2015)

Open Track (kN) Near Bridge (kN)

Mean 63.6 109

Standard Deviation 3.85 13.2

95% CI High End 67.9 121

95% CI Low End 59.3 96.6

3.3.3 Peak transient load measured on rail crib

Similarly, the data collected by Type II strain gauges (mounted on crib of the rail)

were analyzed and graphed in Figure 3.15 which shows the peak load corresponding to the

63
locomotive wheel passage measured by Type II strain gauges. Again, the loading time

history does not reveal a clear trend as in the case of the deformation time history (see

Figure 3.13). The maintenance activity occurred in July 2014 at the near bridge location

did not cause any significant change in loading behavior. The statistics in Table 3.3 indicate

a significant difference in the loading on rail between crossties at two locations.

180 Aug-12 Jul 1-14

160 Jan-13 Jun-13 Jul 22-14 Jan-15 May-15
140
Peak Force on Crib (kN)

120 Aug-12 Jun-13

Jan-13
Jul 22-14 Jan-15 May-15
100

80

60

40 Nearbridge
20 Opentrack
0
4/1/12 10/18/12 5/6/13 11/22/13 6/10/14 12/27/14 7/15/15
Time

Figure 3.15 Upland St. Peak Force on Crib over Time (2012-2015)

Table 3.3 Statistics of Peak Vertical Load Measured on Rail Crib (2012-2015)

Open Track (kN) Near Bridge (kN)

Mean 132 148

Standard Deviation 6.24 9.6

95% CI High End 139 157

95% CI Low End 126 139

64
3.4 Track Substructure Modulus Backcalculation

Previous analyses showed that the open track and near bridge locations were

statistically different from each other as far as transient deformation and loading trends

were concerned. Moreover, the change in track substructure properties over time was

noticeable, especially at near bridge locations. To quantify the substructure properties and

to gain better understanding of the support conditions, transient response measurements

obtained from the three instrumented NEC bridge approaches of Amtrak passenger lines

near Chester, Pennsylvania, USA were used to analyze properties of different layers as well

as the influence of the tie support conditions at these bridge approaches.

3.4.1 GEOTRACK software program

GEOTRACK (Chang et al. 1980), a multi-layer elastic software program, was used

as a tool for estimating track substructure layer moduli from the field data. It allows the

calculation of displacements in ballast, subballast and subgrade layers as a function of axle

loads, properties of rails, ties, ballast and underlying layers and geometry of the track

structure (Li and Selig, 1998). With the measured displacements and axle loads at the

instrumented sites, substructure layer properties can be backcalculated. The basic

assumptions and features of GEOTRACK program are listed in Table 3.4. Figure 3.16

shows a schematic of track structure commonly analyzed by the GEOTRACK program.

65
Table 3.4 Features and Assumptions of the GEOTRACK Program (Chang et al., 1980)

• 3-dimensional (3D) multilayer

General • Up to 5 substructure layers

• Only vertical loading considered

• Linear elastic beams

• Spans eleven ties

Rail
• Free to rotate at ends, and at each tie

• Linear spring connection between rail and tie

• Linear elastic beams

Ties • Supported at 10-equally spaced circular

locations by the underlying ballast

66
Figure 3.16 Track Elements in GEOTRACK Model (Reproduced from Chang et al. 1980)

3.4.2 Backcalculation procedure and assumptions

The first step during the analysis of a given track structure using GEOTRACK

involves the selection of relevant input parameters for different track components. Some

of the input parameters such as rail, tie and fastener properties, etc., remain constant across

different sites, whereas other input parameters such as substructure layer properties, etc.,

change from one site to another. Table 3.5 lists typical parameters used in GEOTRACK

analyses. Note that K1 and K2 are coefficients for the calculation of stress-dependent soil

moduli. As the soil moduli is simplified to constant in the assumption, K1 and K2 are set

to zero.

The wheel loads corresponding to leading wheels of the trailing locomotive

measured by the strain gauge circuits were used as inputs in the GEOTRACK software.

Peak transient deformations recorded by the individual LVDTs corresponding to the time,

when the leading wheel of the trailing locomotive was directly on top of the instrumented

67
tie, were used as the substructure layer deformations. Position of the second wheel load

was determined based on the axle spacing (obtained from the locomotive manufacturers)

and tie spacing patterns at the instrumented bridge approaches. For instances when the

second wheel location was between two ties, standard force and moment balance methods

were used to assign representative vertical load values to individual ties.

Table 3.5 Typical Track Parameters used in GEOTRACK Analyses

Track variable Value

(1) Rail Properties

Spacing (mm) 1510

Cross sectional area (mm2) 8594

E (MPa) 2.07E+05

Weight (kg/m) 67.46

I (mm4) 3.90E+07

Fastener or sleeper pad stiffness (kN/m) 1.20E+06

(2) Sleeper Properties

Length (mm) 2591

Center to center sleeper spacing (mm) 609.6

Sleeper width at base of sleeper (mm) 274.3

Sleeper width (mm) 228.6

Sleeper height (mm) 177.8

E (MPa) 2.07E+04

Sleeper weight (kg) 386

68
 Table 3.5 (Continued)

No. of sleeper segments having centers between the rails 6

Moment of Inertia I (mm4) 2.42E+08

(3) Material Properties

Unit weight (kN/m3) 19.5, 19.5, 20.5, 20.5, 20.5

Poisson's ratio 0.35,0.35,0.4,0.4,0.4

Modulus (MPa) Needs to be backcalculated

K1 & K2 0

K0 3, 3, 0.7, 0.7, 0.7

Thickness (mm) Determine from MDD system

(4) Wheel Load

Tie number 1,2,6

Magnitude (tonnes) Calculate from strain gauges

In the analyses, the two peak loads were assigned on three ties. When the first load

was assigned to the first tie, and the second load was 2,800 mm away from the first load

per the axle spacing based on axle spacing. The allocated load on the 5th tie and 6th tie can

thus be determined from the force and moment balance equations. An example of detailed

load assignment calculation is given below.

Assume that the two peak loads were recorded as 134 kN and 126 kN, respectively.

Crosstie spacing on site is 609.6 mm. These two peak loads need to be assigned on three

crossties. First load was assigned to the first tie (134 kN). According to force and moment

equilibrium equations, the reaction force caused by the second load can be determined.

69
Note that the axle spacing is 2,800 mm. Hence, the number of ties between the first load

and the second load is calculated by:

2800 𝑚𝑚
= 4.59
609.6 𝑚𝑚

Therefore, the second load lies between 5th and 6th crosstie.

The distance from 5th tie for the second load is:

(4.59 − 4) × 609.6 𝑚𝑚 = 361.6 𝑚𝑚

The distance from 6th tie for the second load is:

609.6 𝑚𝑚 − 361.6 𝑚𝑚 = 248 𝑚𝑚

The allocated load on the 5th tie can be determined from the following equilibrium equation:

𝑀𝑜𝑚𝑒𝑛𝑡 (𝑀) = 0

𝐿𝑜𝑎𝑑 𝑜𝑛 5𝑡ℎ 𝑇𝑖𝑒 × 609.6 = 𝑠𝑒𝑐𝑜𝑛𝑑 𝑙𝑜𝑎𝑑 × 248

248
𝐿𝑜𝑎𝑑 𝑜𝑛 5𝑡ℎ 𝑇𝑖𝑒 = 126𝑘𝑁 × = 51.26𝑘𝑁
609.6

Similarly, the allocated load on the 6th tie is calculated as:

𝐿𝑜𝑎𝑑 𝑜𝑛 6𝑡ℎ 𝑇𝑖𝑒 = 126𝑘𝑁 − 51.26𝑘𝑁 = 74.74𝑘𝑁

After determining the loading condition, the moduli of individual track substructure

layers were then determined by trial and error. A set of seed moduli was first assigned to

the track substructure layers and GEOTRACK was used to calculate deflections at the

depths corresponding to the locations of individual MDD modules. Based on comparisons

between the calculated and measured deflection values at each depth, adjustments were

70
made to the individual layer modulus values. For example, if the calculated deflections

were larger than those measured in the field, the modulus values assigned to the track

substructure layers had to be increased for the next iteration. This process was repeated

until the deflections measured in the field and those predicted using GEOTRACK matched

with a tolerance level of less than 5%. Some of the simplifying assumptions made during

such iterative estimation of track substructure layer moduli using GEOTRACK are:

1. The ballast and subballast layers at the Madison Street 12-ft. and Caldwell Street

west locations were combined into one layer for the analyses;

2. The two silty clay layers (layers 3 and 4) at the Madison Street open-track (60 ft.

from the south abutment) location were combined into one layer;

3. The sandy loam and sand layers (layers 3 and 4) at the Upland Street near-bridge

location (15 ft. from the north abutment) were combined into one layer.

3.4.3 Backcalculation results before eliminating tie gap

Note that transient deformations recorded by the top-most LVDT in an MDD

system comprises contributions from (1) movement of the tie before it encounters the

underlying ballast layer, and (2) movement of the tie-ballast system as a unit. Accordingly,

taking the transient displacements recorded by the top-most LVDT as representative of the

ballast layer deformation can lead to serious over-estimation of movements within the

ballast layer if tie gap exists. This in turn leads to significant under-estimation of the

backcalculated ballast layer modulus values.

To more clearly indicate this, the first set of iterative substructure layer modulus

estimations were carried out without eliminating the tie gap contributions from the LVDT

No.1 measurements. The resulting backcalculated layer modulus values for the Madison,

71
Caldwell, and Upland Street locations are presented in Table 3.6. As seen from Table 3.6,

assuming no gap underneath the tie leads to the estimation of significantly low ballast layer

modulus values (mostly close to 20 MPa). These values are significantly lower than typical

ballast layer modulus values reported in the literature (Sussmann and Selig, 2000), and

therefore need to be revised based on elimination of tie-gap contributions from the LVDT-

1 transient deformations. This in turn specified the importance of considering tie gap in

modeling process rather than assuming a full support condition for track superstructure.

Table 3.6 Backcalculated Track Substructure Layer Moduli Values before Eliminating
Tie Gap

(a) Madison Street Location

Madison 12ft. (MPa) Madison
60ft. (MPa)
Layer Nov Jan June Layer Nov Jan June
2012 2013 2013 2012 2013 2013
Ballast 21 14 17 Ballast 29 28 25
Fouled Ballast 21 14 17 Fouled Ballast 61 68 63
+ Hard Pan

Hard Pan 77 51 56 Silty Clay 36 39 33

Grey Sandy 34 24 26 Silty Clay 40 40 40
Loam + Cinder
Brown Silty 42 39 41 Fat Clay 59 58 59
Clay

72
 Table 3.6 (Continued)

(b) Caldwell Street Location

Caldwell West (MPa) Caldwell East (MPa)

Layer Nov Jan June Layer Nov Jan June
2012 2013 2013 2012 2013 2013
Ballast 20.5 21 23.5 Ballast 22.5 20 30
Fouled 20.5 21 23.5 Fouled Ballast + 59 40 48
Ballast Hard Pan

Brown 116 121.5 127 Possible HMA 72 68 74

Silty Sand Layer (25 mm) +
Moist Sand
Silty Clay 40 42 45 Silty Clay 30.5 27 30

Moist Soft 86.5 92 108 Fat Clay 63.5 55.5 66

Silty Clay
(c) Upland Street Location

Upland 15ft. (MPa) Upland 60ft. (MPa)

Layer Aug Jan June Layer Aug Jan June
2012 2013 2013 2012 2013 2013
Ballast 20 15 5 Ballast 48 51 75
Fouled 40 35 35 Fouled 19 48 51
Ballast Ballast
Sandy Loam 35 28 31 Sandy 33 30 31
Loam
Thin Sand 55 55 55 Clayey Silt 38 31 32
Layer
Sandy Load 85 85 85 Sandy 73 60 70
Loam

3.4.4 Backcalculation results after eliminating tie gap

To calculate the track substructure layer moduli for a more accurate representation

of the field load-deformation behavior of the ballast layer, it was necessary to subtract the

73
tie gap magnitudes from the transient deformations recorded by the top-most LVDT in an

MDD system. Tie gap amounts were estimated by Progressive Load Threshold method.

Detailed progress of determining gap amount can be found in Boler et al. (2018).

Estimated tie gap amounts were subtracted from the peak transient displacements

recorded by the top-most LVDTs in order to calculate the “Corrected Layer 1 Deformation”

values, which can be said to better represent the ballast layer deformation under train

loading. Such corrected layer 1 deformation values for the Madison, Caldwell, and Upland

Street bridge approaches are summarized in Table 3.7. It is seen that near bridge locations

usually exhibit higher tie gap compared with the open track locations, which explains the

larger deformations under train passage in the previous section.

In what follows, Table 3.8 presents the revised track substructure modulus values

estimated using GEOTRACK after the removal of the tie-gap contributions from LVDT-1

measurements. Figures 3.17 – 3.19 clearly indicate that the field measured data and

GEOTRACK predicted data match well using the final set of layer moduli.

74
 (a)

(b)

Figure 3.17 Comparisons of Calculated Deformations with Field Measured Data (a)
Madison 12 ft. near bridge location (b) Madison 60 ft. open track location.

75
 (a)

(b)

Figure 3.18 Comparisons of Calculated Deformations with Field Measured Data (a)
Caldwell West location (b) Caldwell East location.

76
 (a)

(b)

Figure 3.19 Comparisons of Calculated Deformations with Field Measured Data (a)
Upland 15 ft. near bridge location (b) Upland 60 ft. open track location.

77
 Table 3.7 Corrected Layer 1 Deformations – Considering Tie Gap

(a) Madison Street Location

Field Measured
Corrected Layer
Layer 1 Tie Gap
Location 1 Deformation
Deformation Estimated (mm)
(mm)
(mm)
12 FT 0.83962 0.5454 0.29422
Aug 2012
60 FT 1.9038 1.1209 0.7829
12 FT 1.1875 0.9219 0.2656
Nov 2012
60 FT 0.74028 0.5873 0.15298
12 FT 1.5979 1.3731 0.2248
Jan 2013
60 FT 0.88728 0.6085 0.27878
12 FT 1.7133 1.4275 0.2858
June 2013
60 FT 0.96554 0.7637 0.20184

(b) Caldwell Street Location

Field Measured
Corrected Layer
Layer 1 Tie Gap
Location 1 Deformation
Deformation Estimated (mm)
(mm)
(mm)
West 1.52 0.978 0.542
Nov 2012
East 0.976 0.7099 0.2661
West 1.5 0.9252 0.5748
Jan 2013
East 1.04 0.7166 0.3234
West 1.51 1.007 0.503
June 2013
East 0.794 0.618 0.176

78
 Table 3.7 (Continued)
(c) Upland Street Location

Field Measured
Corrected Layer
Layer 1 Tie Gap
Location 1 Deformation
Deformation Estimated (mm)
(mm)
(mm)
15 FT 1.712 1.5286 0.1834
Aug 2012
60 FT 0.4537 0.3435 0.1102
15 FT 2.6261 1.9836 0.6425
Nov 2012
60 FT 0.34659 0.2364 0.11019
15 FT 1.594 1.3242 0.2698
Jan 2013
60 FT 0.40564 0.2819 0.12374
15 FT 4.5513 3.864 0.6873
June 2013
60 FT 0.31634 0.2176 0.09874

Table 3.8 Backcalculated Track Substructure Layer Moduli Values after Eliminating Tie
Gap
(a) Madison Street Location
Madison 12ft. (MPa) Madison 60ft. (MPa)
Layer Aug Nov Jan June Layer Aug Nov Jan June
2012 2012 2013 2013 2012 2012 2013 2013
Ballast 70 94 102 78 Ballast 32 155 96 105
Fouled 70 94 102 78 Fouled 56 80 80 60
Ballast Ballast +
Hard Pan
Hard Pan 37 51 57 40 Silty Clay 35 39 41 34
Grey 41 60 42 39 Silty Clay 35 39 41 34
Sandy
Loam +
Cinder
Brown 36 36 38 35 Fat Clay 54 50 54 49
Silty Clay

79
 Table 3.8 (Continued)
(b) Caldwell Street Location
Caldwell West (MPa) Caldwell East (MPa)
Layer Nov Jan June Layer Nov Jan June
2012 2013 2013 2012 2013 2013
Ballast 52 57 50 Ballast 80 60 125
Fouled 52 57 50 Fouled Ballast + 60 49 50
Ballast Hard Pan

Brown 126 130 130 Possible HMA 65 70 75

Silty Sand Layer (25 mm) +
Moist Sand
Silty Clay 38 39 35 Silty Clay 27 27 27

Moist Soft 80 75 80 Fat Clay 55 50 55

Silty Clay

(c) Upland Street Location

Upland 15ft. (MPa) Upland 60ft. (MPa)
Layer Aug Nov Jan June Layer Aug Nov Jan June
2012 2012 2013 2013 2012 2012 2013 2013
Ballast 153 40 100 34 Ballast 184 205 180 230
Fouled 80 111 87 77 Fouled 19 57 60 54
Ballast Ballast
Sandy 33 43 39 42 Sandy 31 30 31 31
Loam Loam
Thin 33 43 39 42 Thin 37 33 31 32
Sand Sand
Layer Layer
Sandy 123 110 118 111 Sandy 70 70 64 70
Loam Loam

Comparing the estimated track substructure layer modulus values before and after

the consideration of the tie gap magnitudes indicates a significant increase in the estimated

ballast and subballast layer moduli for the latter case, which is more representative of

values reported in the literature. The estimated substructure layer moduli are particularly

80
useful as initial modulus estimates during advanced numerical modeling of railroad track

transitions.

3.5 Summary

This chapter presented field instrumentation procedure and details, followed by

results from advanced analyses of transient response data and vertical wheel load data

collected at the instrumented bridge approaches on the Amtrak Northeast Corridor near

Chester, PA. First of all, three bridge approach locations were selected for the

instrumentation effort based on analyses of GPR and 60-month period of geometry car data.

Field instrumentation was undertaken at these problematic bridge approaches which

experienced recurrent differential movement problems. Multi-Depth Deflectometers

(MDDs) and strain gauges were installed at the selected bridge approaches to quantify

deformations of individual track substructure layers under loading as well as applied wheel

loads captured on rail.

Next, exploratory analyses were conducted using the field data to visualize the time

history of dynamic responses in the track system. This step helps us to gain a better

understanding of the dynamic behavior of railroad track bridge approaches. It is observed

that the vertical wheel loads at near bridge locations were relatively constant and stable

over the three years monitored, however, the deformations at near bridge locations showed

clear trends of increasing for the same monitoring period. This phenomenon indicates

noticeable changes in substructure support conditions over time at the near bridge locations.

In addition, statistical analyses were carried out on the field collected data. Mean,

standard deviation, and 95% confidence interval values were calculated for both the open

track and near bridge locations. It was found that the transient deformations and wheel

81
loads on rail from the two locations were statistically different. Such major differences

were not due to random error in data collection. There is a need to determine the physical

mechanisms that govern at railroad track bridge approaches and evaluate how they

contribute to the behavioral differences.

Finally, a track substructure analysis tool GEOTRACK was utilized in this chapter

to quantify the individual moduli of substructure layers. An iterative process was developed

to backcalculate the layer moduli based on the field measured transient layer deformation

results and wheel loads. The unusual low moduli backcalculated based on LVDT-1 ballast

layer deformation typically indicated an insufficient support from underneath the tie or a

possible gap between the bottom of the tie and the top of the ballast layer. To better quantify

substructure layer properties, a gap forming at the tie-ballast interface was considered at

each instrumented site. After correction of the top most layer deformation due to this

estimated gap, the final backcalculated layer moduli were more reasonable in scale and

coulfd be used in track subtructure analysis.

82
CHAPTER 4: DISCRETELY SUPPORTED TRAIN-TRACK-BRIDGE
MODEL

4.1 Introduction

Field collected transient deformation and wheel load data presented in Chapter 3

are used in this chapter to analyze the dynamic behavior of train-track-bridge coupling

interaction, which is a complex problem that accounts for when the train passes on the track

transition zone from an open track on to bridge deck. To gain a better understanding of the

dynamic behavior, and to quantify important variables according to their relative influences,

the numerical modeling technique has been adopted herein to develop an innovative model

to simulate the dynamic behavior in time domain at track bridge approaches. This chapter

will introduce this 2-D dynamic train-track-bridge model consisting of a coupled train and

track structure, which considers both regular track on substructure and bridge deck.

In this chapter, components of a train-track-bridge model are summarized and

illustrated first. Next, the mathematical derivations of equilibrium equations are presented

for the train-track-bridge system and numerical solution algorithms are discussed. Finally,

nonlinear springs are introduced in the model to better represent the force-displacement

relationship at track bridge approaches with hanging crossties. Accordingly, the previously

adopted numerical solution algorithm is modified to solve the new nonlinear train-track-

bridge system.

4.2 Development of a Train-Track-Bridge Model

Figure 4.1 illustrates the key components of a newly proposed train-track-bridge

model. The train is assumed to travel in the direction of arrow at a constant speed of 𝑣. The

train presented in the study consists of two locomotives at each end of the train (front/rear)

83
and six passenger cars in the middle, similar to the Amtrak’s ACELA Express passenger

train operating over the instrumented bridge approaches, discussed in Chapter 3.

Each vehicle consists of one car body, two bogies, and four wheel-sets. Each

component is assumed to be a nondeformable rigid body that only has vertical movement

or pitch movement. Each vehicle is modeled as a 10-degree-of-freedom system, consisting

of vehicle car body vertical displacement 𝑈F , vehicle car body pitch motion 𝜑F , two bogie

masses vertical displacement 𝑈HIJ , two bogie pitch motions 𝜑HIJ , and four wheel masses

vertical displacement 𝑈K . The vehicle body is linked with bogies through secondary

suspension springs (𝑘LM ) and dampers (𝑐LM ). The bogies are linked with wheels through

primary suspension springs ( 𝑘LN ) and dampers ( 𝑐LN ). The primary and secondary

suspension systems of the train vehicle are assumed to be linear.

Track substructure on the open track side is modeled as an infinite Euler beam

discretely supported by three layers of viscous-elastic springs and dampers, representing

rail pads, ballast, and soil, and two layers of masses, representing crossties and ballast.

Crossties and ballast masses are assumed rigid bodies. The proposed model accounts for

shear stress between ballast particles. While track substructure on the bridge side is

modeled as an infinite Euler beam discretely supported by two layers of viscous-elastic

foundations representing crossties resting directly on top of bridge deck (open deck bridge).

At each time step, the vertical displacement of rail 𝑈O , vertical displacement of crosstie 𝑈P ,

and vertical displacement of ballast masses 𝑈H are the remaining unknowns in time domain

in the track system.

84
85
Figure 4.1 Proposed Train-Track-Bridge Model.
 To summarize, the proposed train-track-bridge model given in Figure 4.1 has the

following assumptions:

1) The train runs at a constant speed;

2) Each vehicle consists of one rigid car body, two rigid bogies, and four rigid wheel-sets;

3) The primary and secondary suspension systems in the vehicle are considered linear;

4) The vehicles are not connected (independent moving) in the train model;

5) The interaction between wheel and rail is assumed to be linked by linear springs in

vertical direction only;

6) No friction or creeping is considered at the wheel and rail interface;

7) All the springs and dampers in track substructure are assumed linear at the beginning;

8) The bridge deck is assumed to be open deck.

To obtain realistic dynamic responses in the track system from the proposed train-

track-bridge model, the model parameters as input properties need to be chosen with

physical relevance and careful attention. Table 4.1 presents the physical meanings of the

notation symbols in the proposed train-track-bridge model. Table 4.1 also lists one set of

representative property values. Some of the track properties, such as crosstie spacing and

rail properties, are selected based on realistic values at instrumented bridge approach

locations at Amtrak NEC discussed in Chapter 3. The other track properties, such as spring

stiffness and damping ratio are all selected from a list of previously published studies which

also established a three-layer track model (Cantero et al., 2016). Vehicle parameters such

as mass of train, bogie and wheel spacings are selected from Amtrak Acela Express

passenger train’s vehicle data documented in research report published by Judge et al.

(2018).

86
Table 4.1 Notations and Values for Train-Track-Bridge Model in This Study

Track Properties
𝐸 Young's modulus of rail 2.07E+05 Mpa
𝐼 second moment of area 3.9E+07 mm4
mT rail mass per unit length 67.46 kg/m
𝑐U rail pad damping ratio 124 kN.s/m
𝑘U rail pad stiffness 7.80E+04 kN/m
𝑚P crosstie mass 386 kg
𝑚H ballast mass 683 kg
𝑐H ballast damping ratio 82 kN.s/m
𝑘H ballast stiffness 1.20E+05 kN/m
𝑡L ballast shear stiffness 7.80E+03 kN/m
𝑐L subgrade damping ratio 300 kN.s/m
𝑘L subgrade stiffness 5.00E+04 kN/m
𝑈O rail displacement -
𝑈P crosstie displacement -
𝑈H ballast displacement -
𝑘VO vehicle-rail stiffness 1.53E+06 kN/m
Vehicle Properties
𝑀F car body mass 6.31E+04 kg
𝐽F car body mass inertia 1.20E+06 kg.m2
𝜑F car body pitch rotation -
𝑈F car body displacement -
cYM secondary suspension damping 70 kN.s/m
𝑘LM secondary suspension stiffness 5.30E+03 kN/m
𝑀HIJZ[ bogie mass 1.20E+03 kg
𝐽HIJZ[ bogie mass inertia 760 Kg.m2
𝜑HIJZ[ bogie pitch rotation -
𝑈HIJZ[ bogie displacement -
𝑐LN primary suspension damping 49 kN.s/m
𝑘LN primary suspension stiffness 2.10E+03 kN/m
𝑚K wheel mass 1.20E+03 kg
𝑈K wheel displacement -
𝑃 wheel-rail interaction force -
𝑎 bogie distance 5.37 m
𝑏 wheel distance 1.42 m

87
4.3 Equations of Motion of the Train-Track-Bridge System

The proposed 2-dimensional model consists of three major components – train,

track, and bridge. To solve for the dynamic responses of these components, equilibrium

equations are derived for each component. Note that the individual vehicle is considered

independent. Hence, one vehicle rather than the whole train (eight vehicles) will be taken

as an example to derive equilibrium equations. In the following sections, the equations of

motion of vehicle, track on embankment side, and track on bridge side are presented in an

orderly fashion and by discussing the assembly process of coupling these components.

4.3.1 Equations of vehicle

Only vertical displacement and pitch motion are considered for the vehicle. The

body diagrams of dynamic force equilibrium for vehicle components are illustrated in

Figure 4.2. For example, the car body is supported by two bogies underneath. It reaches

dynamic equilibrium with the external forces from bogies and self-weight. Similarly, each

bogie is connected with the car body and two wheel-sets. Therefore, the force equilibrium

for the bogie is an equilibrium between self-weight of bogie, contact forces between bogie

and wheel-set 1, contact forces between bogie and wheel-set 2, and contact forces between

car body and bogie. Each wheel-set has self-weight and two external forces, including

contact force between wheel-set and rail and contact force between wheel-set and bogie.

All the forces exerted on wheel-set should reach equilibrium at any given time. Besides the

force equilibrium, the moment equilibrium for the bogie should also be achieved at the

same time.

88
 Figure 4.2 Body Diagram for Vehicle Components.

The unknown dynamic responses in the model i.e. vertical displacement and pitch

rotation angles are in time domain. The system reaches both force and moment equilibrium

at any given time 𝑡 in the analysis period. Accordingly, the equations of motion for each

component in the vehicle are derived as follows.

1) Equations of motion of car body are given by:

𝑀^ 𝑈̈^ + 2𝑐YM 𝑈̇^ + 2𝑘YM 𝑈^ − 𝑐YM b𝑈̇cdefgN + 𝑈̇cdefgM h − 𝑘YM b𝑈cdefgN + 𝑈cdefgM h
(4.1)
= 𝑀F 𝑔

𝐽^ 𝜓̈^ + 2𝑐YM 𝑎M 𝜓̇^ + 2𝑘YM 𝑎M 𝜓^ − 𝑐YM 𝑎b𝑈̇cdeN + 𝑈̇cdeM h − 𝑘YM 𝑎b𝑈cdeN − 𝑈cdeM h
(4.2)
=0

2) Equations of motion of bogie 𝑖 (𝑖 = 1,2) are given by:

𝑀HIJZ[ 𝑈̈cdefgZ + (𝑐YM + 2𝑐YN )𝑈̇cdefgZ + (𝑘YM + 2𝑘YN )𝑈cdefgZ ∓ 𝑐YM b𝑈̇^ + 𝑎𝜓̇^ h

∓ 𝑘YM (𝑈^ + 𝑎𝜓^ ) − 𝑐YN b𝑈̇mN + 𝑈̇mM h − 𝑘YN (𝑈mN + 𝑈mM )

= 𝑀HIJZ[ 𝑔 (4.3)

89
 𝐽^ 𝜓̈cdefgZ + 2𝑐YN 𝑏 M 𝜓̇cdefgZ + 2𝑘YN 𝑏 M 𝜓cdefgZ − 𝑐YN 𝑏b𝑈̇mN − 𝑈̇mM h

− 𝑘YN 𝑏(𝑈mN − 𝑈mM ) = 0 (4.4)

3) Equations of motion of wheel-set 𝑖 (𝑖 = 1,2,3,4) are given by:

𝑀m 𝑈̈mZ + 𝐶YN b𝑈̇mZ − 𝑈̇cdefgN h + 𝑘YN b𝑈mZ − 𝑈cdefgN h − 𝑐YN 𝑏𝜓̇cdefgN

− 𝑘YN 𝑏𝜓cdefgN + 𝑃Z = 𝑀K 𝑔 (4.5)

where 𝑃Z represents the contact force between wheel-set 𝑖 and rail. It is also a function of

time 𝑡 as the other unknowns in the model such as 𝑈KZ , 𝑈HIJZ[Z , etc. For simplicity, the

contact force between wheel-sets and rail are assumed to be simulated by a linear spring

connecting the two components. Hence, at any given time 𝑡 , 𝑃Z (𝑡) = 𝐾VO × Δ𝑑 =

𝐾VO × (𝑈O − 𝑈KZ ), where the magnitude of constant contact stiffness 𝐾VO is selected from

the Yu and Mao’s (2018) stochastic dynamic model. Their model established a 3D train-

track-bridge coupled system with refined wheel/rail interaction. The contact force 𝑃Z (𝑡) is

substituted into Equation 4.5. The new equilibrium equation of wheel-set 𝑖 is given by:

𝑀m 𝑈̈mZ + 𝐶YN b𝑈̇mZ − 𝑈̇cdefgN h + 𝑘YN b𝑈mZ − 𝑈cdefgN h − 𝑐YN 𝑏𝜓̇cdefgN

− 𝑘YN 𝑏𝜓cdefgN + 𝐾VO (𝑈O − 𝑈KZ ) = 𝑀K 𝑔 (4.6)

4.3.2 Equations of track embankment side

In the proposed train-track-bridge model, the rail is supported by discrete crossties

on top of ballast layer for the track embankment side. The rail is modeled as an infinite

Euler beam. The crossties are modeled as rigid masses at a spacing of 2𝑓𝑡. The ballast

layer is modeled as rigid masses at the same crosstie spacing (2𝑓𝑡). At track embankment

side (open track location), it is assumed that all the stiffnesses are linear. Therefore, the rail,

crossties, and ballast masses in the model are connected with linear springs and dampers.

90
Similar to the train component, the track structure components including rail, crossties, and

ballast masses also exhibit force equilibrium at any given time 𝑡.

Figure 4.3 Body Diagrams for Track Components.

Figure 4.3 shows the free body diagrams of track structure components. The

crosstie comes to equilibrium under contact forces between rail and tie, contact forces

between tie and ballast, and self-weight. A portion of the rail is taken as an example to

illustrate the forces exerted on it, including vertical supports from underneath crosstie and

vertical wheel loads from one wheel-set running on top of that portion of the rail. Ballast

particles experience a combination of shear forces between adjacent ballast particles,

contact forces between ballast and subgrade, contact forces between ballast and crosstie,

and self-weight.

Assume that 𝑥 represents the location in special domain along the train moving

direction. Note that the vehicle is running at a constant speed 𝑣, therefore, the wheel-set

contact force 𝑃Z is moving along the rail with a constant speed 𝑣. Detailed equilibrium

equations can be derived as follows.

1) Equation of motion of rail beam is given by:

91
 y t
𝜕 t 𝑈O (𝑥, 𝑡) 𝜕 M 𝑈O (𝑥, 𝑡)
𝐸𝐼 + 𝜌O + v 𝑎w (𝑡)𝛿(𝑥 − 𝑥w ) = v 𝑃Z 𝛿(𝑥 − 𝑣𝑡 − 𝑥Z ) (4.7)
𝜕𝑥 t 𝜕𝑡 M
N ZzN

Equation of motion of rail-crosstie reaction is given by:

𝜕𝑈O (𝑥w , 𝑡) 𝜕𝑈P (𝑥w , 𝑡)

𝑎w (𝑡) = 𝑐U { − | + 𝑘U (𝑈O (𝑥w , 𝑡) − 𝑈P (𝑥w , 𝑡)) (4.8)
𝜕𝑡 𝜕𝑡

2) Equation of motion of crosstie is given by:

𝜕 M 𝑈P (𝑥w , 𝑡)
𝑚P = 𝑎w (𝑡) − 𝑏w (𝑡) (4.9)
𝜕𝑡 M

Equation of motion of crosstie-ballast reaction is given by:

𝜕𝑈P (𝑥w , 𝑡) 𝜕𝑈H (𝑥w , 𝑡)

𝑏w (𝑡) = 𝑐H { − | + 𝑘H (𝑈P (𝑥w , 𝑡) − 𝑈H (𝑥w , 𝑡)) (4.10)
𝜕𝑡 𝜕𝑡

3) Equation of motion of ballast mass is given by:

𝜕 M 𝑈H (𝑥w , 𝑡) 𝜕𝑈H (𝑥w , 𝑡)

𝑚H M
= 𝑏w (𝑡) − 𝑐L − 𝑘L 𝑈H (𝑥w , 𝑡) + 𝑇w − 𝑇w}N (4.11)
𝜕𝑡 𝜕𝑡

Equation of motion of the shear force between ballast mass is given by:

𝑇w = 𝑡L (𝑈H (𝑥w~N , 𝑡) − 𝑈H (𝑥w , 𝑡)) (4.12)

4.3.3 Equations of track bridge side

Similar to the open track embankment side, the rail is also modeled as an infinite

Euler beam for the bridge side (the Euler beam is extending from open track side to the

bridge side). In addition, since the field instrumented bridge is an open deck bridge (no

ballast), ballast layer is not considered on bridge deck in the proposed train-track-bridge

model. Therefore, on the bridge side in the model, the rail is supported by discrete crossties

directly sitting on top of a rigid bridge deck. The crossties are modeled as rigid masses at

92
a spacing of 2𝑓𝑡 between each crosstie. At first, since no hanging ties are considered, the

rail, crossties, and nondeformable bridge deck are assumed to be connected by linear

springs and dampers. Bridge deck is assumed non-deformable, which is an approximation

for bridge abutment in the field. Realistically, if the bridge deck spans long distance, it

would undergo large deformation under dynamic loading. However, the proposed model

focuses on the dynamic responses at open track location and near bridge location.

Therefore, the bridge deck behavior is simplified. The track structure components

including rail and crossties reach the state of force equilibrium at any given time 𝑡.

1) Equation of motion of rail beam is given by:

• t
𝜕 t 𝑈O (𝑥, 𝑡) 𝜕 M 𝑈O (𝑥, 𝑡)
𝐸𝐼 + 𝜌O + v 𝑎w (𝑡)𝛿(𝑥 − 𝑥w ) = v 𝑃Z 𝛿(𝑥 − 𝑣𝑡 − 𝑥Z ) (4.7)
𝜕𝑥 t 𝜕𝑡 M
y}N ZzN

Equation of motion of rail-crosstie reaction is given by:

𝜕𝑈O (𝑥w , 𝑡) 𝜕𝑈P (𝑥w , 𝑡)

𝑎w (𝑡) = 𝑐U { − | + 𝑘U (𝑈O (𝑥w , 𝑡) − 𝑈P (𝑥w , 𝑡)) (4.8)
𝜕𝑡 𝜕𝑡

2) Equation of motion of crosstie is given by:

𝜕 M 𝑈P 𝜕𝑈P (𝑥w , 𝑡)
𝑚P M
= 𝑎w (𝑡) − 𝑘𝑈P (𝑥w , 𝑡) − 𝑐 (4.13)
𝜕𝑡 𝜕𝑡

The above equilibrium equations 4.1 – 4.13 for the vehicle, track, and bridge system

can be solved numerically with reasonable time steps and initial conditions. In solving

these equations, the partial differential equations of motion of the rail (Equation 4.7) need

to be facilitated first using a reasonable shape function. This process is explained in detail

in the following section.

93
4.3.4 Weak formulation of rail beam equation

To solve the 4th order partial differential equation of motion of the rail beam, the

weak formulation is developed with utilization of rail mode shape function. The admissible

shape function is assumed as:

𝑘𝜋𝑥
𝜑€ (𝑥) = 𝑠𝑖𝑛 ( ) (4.14)
𝑙

Therefore, the dynamic rail response at location 𝑥, time 𝑡 is described as:

„

𝑈O (𝑥, 𝑡) ≈ v 𝜑€ (𝑥)𝑞€ (𝑡) (4.15)

€zN

where

𝑘 is the mode number;

𝑙 is the length of the rail;

𝐾 is the total number of modes.

Next, weak formulations are performed on the rail beam equation of motion to

obtain a series of 2nd order ordinary differential equations. Equation 4.7 is transformed as:

‡ y
𝜕 t 𝑈O (𝑥, 𝑡) 𝜕 M 𝑈O (𝑥, 𝑡)
… †𝐸𝐼 + 𝜌O + v 𝑎w (𝑡)𝛿(𝑥 − 𝑥w )
𝜕𝑥 t 𝜕𝑡 M
ˆ N

− v 𝑃𝛿(𝑥 − 𝑣𝑡 − 𝑥‰ )Š 𝑤Z 𝑑𝑥 = 0
‰zN (4.16)

where

𝑤Z = 𝜑Z (𝑥) (4.17)

By applying partial integration to Equation 4.16, the series of equations can be

described in the form of matrix. The matrix describing equations of motion of rail beam

can be simplified as follows:

94
 [𝑀N ]{𝑞̈ } + [𝐶N ]{𝑞̇ } + [𝐾N ]{𝑞 } + [𝐶M ]{𝑈Ṗ } + [𝐾M ]{𝑈P } = {𝐹 } (4.18)

where
‡ (4.19)
[𝑀N ]Z‰ = … 𝜌O 𝜑Z (𝑥)𝜑‰ (𝑥)𝑑𝑥
ˆ

[𝐶N ]Z‰ = v 𝑐U 𝜑Z (𝑥w )𝜑‰ (𝑥w ) (4.20)

wzN

‡ y

[𝐾N ]Z‰ = … v 𝑘U 𝜑Z (𝑥w )𝜑‰ (𝑥w ) + 𝐸𝐼𝜑Z‘‘ (𝑥)𝜑‰‘‘ (𝑥)𝑑𝑥 (4.21)

ˆ wzN

[𝐶M ]Z‰ = −𝑐U 𝜑Z b𝑥‰ h (4.22)

[𝐾M ]Z‰ = −𝑘U 𝜑Z b𝑥‰ h (4.23)

{𝐹Z } = v 𝑃KZ 𝜑Z (𝑥KZ ) (4.24)

KZzN

4.3.5 Coupling of vehicle and track structure

The equations of motion for vehicle components, crossties and ballast masses are

all used to assemble the final matrix. To be concise, one car of the train, consisting of one

car body, two bogies, and four wheel-sets were considered here to illustrate the basic form

of the final matrix.

[𝑀V ]“𝑈̈V ” + [𝐶V ]“𝑈̇V ” + [𝐾V ]{𝑈V } = {𝐹V } (4.25)

[𝑀M ]“𝑈̈P ” + [𝐶• ]{𝑞̇ } + [𝐾• ]{𝑞 } + [𝐶t ]{𝑈Ṗ } + [𝐾t ]{𝑈P } + [𝐶– ]{𝑈Ḣ } + [𝐾– ]{𝑈H }

= {0} (4.26)

[𝑀• ]“𝑈̈H ” + [𝐶— ]“𝑈̇H ” + [𝐾H ]{𝑈H } + [𝐶˜ ]{𝑈Ṗ } + [𝐾˜ ]{𝑈P } = {0} (4.27)

where

95
 š
[𝑈V ] = “𝑢F , 𝑢HIJZ[N , 𝑢HIJZ[M , 𝑢KN , 𝑢KM , 𝑢K• , 𝑢Kt , 𝜑F , 𝜑HIJZ[N , 𝜑HIJZ[M ” (4.28)

[𝑀V ] = 𝑑𝑖𝑎𝑔(𝑚F , 𝑚HIJZ[N , 𝑚HIJZ[M , 𝑚KN , 𝑚KM , 𝑚K• , 𝑚Kt , 𝐽F , 𝐽HIJZ[N , 𝐽HIJZ[M ) (4.29)

𝑚F 𝑔, 𝑚HIJZ[N 𝑔, 𝑚HIJZ[N 𝑔, 𝑚KN 𝑔 − 𝑃N , 𝑚KM 𝑔 − 𝑃M , 𝑚K• 𝑔 − 𝑃• , š

{𝐹V } = › œ (4.30)
𝑚Kt 𝑔 − 𝑃t , 0,0,0

2𝑐FM −𝑐FM −𝑐FM 0 0 0 0 0 0 0

⎡ ⎤
𝑐FM + 2𝑐FN 0 −𝑐FN −𝑐FN 0 0 −𝑐FM 𝑎 0 0
⎢ ⎥
⎢ 𝑐FM + 2𝑐FN 0 0 −𝑐FN −𝑐FN 𝑐FM 𝑎 0 0 ⎥
⎢ 𝑐FN 0 0 0 0 −𝑐FN 𝑏 0 ⎥
⎢ 𝑐FN 0 0 0 𝑐FN 𝑏 0 ⎥
[𝐶V ] = ⎢
𝑐FN 0 0 0 −𝑐FN 𝑏 ⎥
(4.31)
⎢ 𝑠𝑦𝑚 𝑐FN 0 0 𝑐FN 𝑏 ⎥
⎢ ⎥
⎢ 2𝑐FM 𝑎M 0 0 ⎥
⎢ 2𝑐FN 𝑏M 0 ⎥
⎣ 2𝑐FN 𝑏M ⎦

2𝑘FM −𝑘FM −𝑘FM 0 0 0 0 0 0 0

⎡ ⎤
𝑘FM + 2𝑘FN 0 −𝑘FN −𝑘FN 0 0 −𝑘FM 𝑎 0 0
⎢ ⎥
⎢ 𝑘FM + 2𝑘FN 0 0 −𝑘FN −𝑘FN 𝑘FM 𝑎 0 0 ⎥
⎢ 𝑘FN 0 0 0 0 −𝑘FN 𝑏 0 ⎥
⎢ 𝑘FN 0 0 0 𝑘FN 𝑏 0 ⎥
[𝐾V ] = ⎢
𝑘FN 0 0 0 −𝑘FN 𝑏 ⎥
(4.32)
⎢ 𝑠𝑦𝑚 𝑘FN 0 0 𝑘FN 𝑏 ⎥
⎢ ⎥
⎢ 2𝑘FM 𝑎M 0 0 ⎥
⎢ 2𝑘FN 𝑏M 0 ⎥
⎣ 2𝑘FN 𝑏M ⎦

[𝑀M ] = 𝑑𝑖𝑎𝑔(𝑚PN , … , 𝑚Pw ) (4.33)

[𝐶• ]w€ = −𝑐Uw 𝜑€ (𝑥w ) (4.34)

[𝐾• ]w€ = −𝑘Uw 𝜑€ (𝑥w ) (4.35)

[𝐶t ] = 𝑑𝑖𝑎𝑔b𝑐UN + 𝑐HN , … , 𝑐Uw + 𝑐Hw h (4.36)

[𝐾t ] = 𝑑𝑖𝑎𝑔b𝑘UN + 𝑘HN , … , 𝑘Uw + 𝑘Hw h (4.37)

[𝐶– ] = [𝐶˜ ] = −𝑑𝑖𝑎𝑔(𝑐HN , … , 𝑐Hw ) (4.38)

[𝐾– ] = [𝐾˜ ] = −𝑑𝑖𝑎𝑔(𝑘HN , … , 𝑘Hw ) (4.39)

[𝑀• ] = 𝑑𝑖𝑎𝑔(𝑚HN , … , 𝑚Hw ) (4.40)

[𝐶— ] = 𝑑𝑖𝑎𝑔(𝑐LN + 𝑐HN , … , 𝑐Lw + 𝑐Hw ) (4.41)

96
 𝑡L −𝑡L 0 … 0 0
⎡−𝑡 2𝑡 −𝑡L 0 ⎤
L L … 0
⎢ ⎥
0 −𝑡 0 0 ⎥
[𝐾— ] = 𝑑𝑖𝑎𝑔(𝑘LN + 𝑘HN , … , 𝑘Lw + 𝑘Hw ) + ⎢ L … (4.42)
⎢ 0 0 −𝑡L 0 ⎥
⎢ 0 … 0 −𝑡L 2𝑡L −𝑡L ⎥
⎣ 0 0 … 0 −𝑡L 𝑡L ⎦

In solving all the above equations, the general equations of motion in standard

matrix form is presented in Equation 4.43, consisting of generalized mass, damping,

stiffness matrices, and force vector.

[𝑀]“𝑋̈ ” + [𝐶 ]“𝑋̇” + [𝐾]{𝑋 } = {𝐹 } (4.43)

where

{𝑋} denotes the generalized displacement vector containing displacement of car body,

bogies, wheel-sets, rail, crossties, and ballast masses, as well as the pitch angle of car body

and bogies at any given time 𝑡;

{𝑋} = {𝑈VN , … 𝑈VZ , 𝑞N , … 𝑞€ , 𝑈PN , … , 𝑈P¦ , 𝑈HN , … 𝑈Hw }š

[𝑀] denotes the generalized mass matrix;

𝑀V 0 0 0
0 𝑀N 0 0
[𝑀] = § ¨
0 0 𝑀M 0
0 0 0 𝑀•

[𝐶 ] denotes the generalized damping matrix;

𝐶V 0 0 0
0 𝐶N 𝐶M 0
[𝐶 ] = § ¨
0 𝐶• 𝐶t 𝐶–
0 0 𝐶˜ 𝐶—

[𝐾 ] denotes the generalized stiffness matrix;

𝐾V ′ 𝐾VO 0 0
[𝐾] = §𝐾VO 𝐾N ′ 𝐾M 0
¨
0 𝐾• 𝐾t 𝐾–
0 0 𝐾˜ 𝐾—

97
{𝐹 } denotes the generalized force vector;

{𝐹 } = {𝐹V‘ , 0,0,0}š

Note that 𝐾V‘ , 𝐾N‘ , 𝐾VO , 𝐹V ′ are modified matrix according to the vehicle-rail contact force 𝑃Z .

4.4 Numerical Integration of Train-Track-Bridge Model using Newmark’s Method

Numerical integration method can be utilized to solve the generalized matrix

presented in section 4.3. It is important to note that the generalized stiffness matrix needs

to be updated in each iteration to simulate the effect of moving vehicle (location of vehicle-

rail contact force 𝑃Z is changing).

There are several widely used numerical integration methods for solving the

generalized matrix. For instance, Fourier transformation is one method to solve linear

equations of motion. The Fourier transformation of the substitute pseudo-periodic loading

leads to harmonic components 𝑃(𝑖𝑤). The frequency domain harmonic responses are

simply the product of the harmonic components by the receptance Frequency Response

Function (FRF) 𝐻(𝑖𝑤). The total response is the summation of the harmonic responses by

inverse Fourier transformation.

The Fourier transformation has been adopted here to solve a simplified model,

simulating one wheel on the track substructure moving from open track location to bridge

deck, as a base for evaluating numerical integration effectiveness. The detailed derivation

and numerical results for Fourier transformation integration can be found in Appendix C.

Despite its popularity in numerical integration and effectiveness in solving

simplified track-bridge model, one obvious disadvantage is that Fourier transformation

method is restricted to linear systems. A system is considered nonlinear when damping or

stiffness properties vary in time. In this situation, the calculation of the response of

98
nonlinear system can be done only by step by step direct integration. In the track-bed

system, the suspension system between crossties and ballast particles can be nonlinear if

the crossties are not properly supported: no contact force at the beginning of loading until

the crosstie touches with the top of ballast layer.

In Chapter 3, it was stated that field monitoring data revealed unsupported crossties

which are commonly observed at railroad bridge approaches. Therefore, to adequately

consider the nonlinear nature caused by unsupported crossties, it is important to adopt a

method that can be applied to both linear and nonlinear systems.

Direct time integration method is applicable to linear as well as nonlinear systems.

Hence, it is a general method to calculate response of a dynamic system under arbitrary

loading. Many methods exist for the direct integration of the equations of motion. Herein,

Newmark’s method has been chosen to solve the equations of motion in the track bed

system. Newmark’s method, also called Newmark-𝛽 method, is developed by Nathan

Newmark (1959) for use in structural dynamics, based on finite difference method.

In this section, the algorithm for solving linear track system shown in Equation 4.43

is presented.

[𝑀]“𝑋̈ ” + [𝐶 ]“𝑋̇” + [𝐾]{𝑋 } = {𝐹 } (4.45)

Table 4.2 illustrates the initialization and integration steps in the algorithm. Average

acceleration method is utilized with parameters 𝛾 = 0.5 , 𝛽 = 0.25 . This method is

unconditionally stable and is always converging in the end. Therefore, no time step

restriction applies. The time step of ∆𝑡 = 0.001 can be selected in this case to account for

a compromised accurate loading function and computational time. The initial conditions

are assumed at stationary.

99
 Next, the algorithm is implemented in MATLAB for numerical solutions. The

MATLAB code can be found in Appendix D.

Table 4.2 Algorithm for Numerical Integration using Newmark’s Method at Linear
System (Paultre, 2013)

Initialization of the variables

1. Initial condition: 𝒖𝟎 = 𝟎, 𝒖𝟎̇ = 𝟎, 𝒖𝟎̈ = 𝟎, 𝒕𝟎 = 𝟎

2. Average acceleration method: 𝜸 = 𝟎. 𝟓, 𝜷 = 𝟎. 𝟐𝟓
3. Integration time step: ∆𝒕 = 𝟎. 𝟎𝟎𝟏
Preliminary calculations

1. Calculate the integration constants:

𝟏 𝜸 𝟏 𝟏
𝒂𝟎 = ; 𝒂 𝟏 = ; 𝒂 𝟐 = ; 𝒂 𝟑 = − 𝟏;
𝜷∆𝒕𝟐 𝜷∆𝒕 𝜷∆𝒕 𝟐𝜷

𝜸 𝜸
𝒂𝟒 = − 𝟏; 𝒂𝟓 = { − 𝟏| ∆𝒕; 𝒂𝟔 = (𝟏 − 𝜸)∆𝒕; 𝒂𝟕 = 𝜸∆𝒕
𝜷 𝟐𝜷

¼ = 𝒌 + 𝒂𝟎 𝒎 + 𝒂𝟏 𝒄
2. Calculate effective stiffness: 𝒌
Integrate step by step

for 𝒏 = 𝟎, 𝟏, 𝟐, … , 𝒕𝒅 /∆𝒕

1. Increment time: 𝒕𝒏}𝟏 = 𝒕𝒏 + ∆𝒕 = (𝒏 + 𝟏)∆𝒕

à at time 𝒕𝒏}𝟏 :
2. Calculate effective force matrix 𝒑
à 𝒏}𝟏 = 𝒑𝒏}𝟏 + (𝒂𝟎 𝒖𝒏 + 𝒂𝟐 𝒖𝒏̇ + 𝒂𝟑 𝒖𝒏̈ )𝒎 + (𝒂𝟏 𝒖𝒏 + 𝒂𝟒 𝒖𝒏̇ + 𝒂𝟓 𝒖𝒏̈ )𝒄
𝒑

3. Calculate displacement matrix at time 𝒕𝒏}𝟏 :

¼
à 𝒏}𝟏 /𝒌
𝒖𝒏}𝟏 = 𝒑

4. Calculate acceleration and velocity matrix at time 𝒕𝒏}𝟏 :

𝒖̈ 𝒏}𝟏 = 𝒂𝟎 (𝒖𝒏}𝟏 − 𝒖𝒏 ) − 𝒂𝟐 𝒖𝒏̇ − 𝒂𝟑 𝒖𝒏̈ ; 𝒖̇ 𝒏}𝟏 = 𝒖𝒏̇ + 𝒂𝟔 𝒖̈ 𝒏 + 𝒂𝟕 𝒖̈ 𝒏}𝟏

100
4.5 Consideration of Hanging Crossties at Bridge Approach

As discussed in the previous section, the ballast layer stiffness (𝑘Hć‡ÄLP ) and

damping ratio (𝑐Hć‡ÄLP ) are assumed to be linear at open track locations. To clarify any

misunderstanding, in this Chapter, “open track location” refers to a location far from bridge

abutment and is assumed not to be affected by the discontinuity in transition zone; while

“near bridge location” refers to a location that is close to bridge abutment and may develop

transition problems. The linear track system at open track locations can be solved by the

Newmark’s algorithm presented in Table 4.2.

However, very often at near bridge locations, hanging crossties exist in the track.

These hanging crossties commonly lack adequate support from the underneath ballast layer.

In Chapter 3, Table 3.7 lists the typical magnitudes of gaps observed between instrumented

crossties and top of ballast layer in Amtrak’s Northeast Corridor lines near Chester, PA.

For these inadequately supported crossties, the force-displacement relationship

cannot be assumed to be linear elastic anymore. Instead, it is supposed to be nonlinear due

to the gap between tie and ballast. For simplicity, the proposed nonlinear train-track-bridge

model assumes bi-linear force-displacement relationship. More specifically, it is assumed

that at the beginning of loading, there is no force between crosstie and ballast until the gap

(𝑢J ) is closed. Once the distance between crosstie and ballast reaches the magnitude of gap,

the stiffness of springs between crosstie and ballast becomes linear (𝑘′Hć‡ÄLP ). Note that

the tangent of stiffness from open track and near bridge locations might be different:

𝑘Hć‡ÄLP ≠ 𝑘′Hć‡ÄLP . It is assumed that damping coefficient 𝑐′Hć‡ÄLP at near bridge is

constant in the analysis. Figure 4.4 illustrates the force-displacement relationship at (a)

open track location with linear stiffness and (b) near bridge location with nonlinear stiffness.

101
The gap between crosstie and ballast is represented by 𝑢J . Note that in the model, it is

flexible to choose the locations of the gap since it is discretely supported. The amount of

gap underneath each crosstie can also be assigned arbitrarily based on realistic situation.

(a) (b)

Figure 4.4 Linear and Nonlinear Dynamic Systems: (a) Linear Stiffness at Open Track
Location, (b) Nonlinear Stiffness at Near Bridge Location.

Based on the Newmark’s method for the linear system presented in Section 4.4, a

modified numerical algorithm is proposed for the nonlinear system. The track system at

near bridge locations may develop nonlinear stiffness that depends on the displacements in

time. Reduction of the time step ∆𝑡 leads to reduction in the error within a time step. Note

that in solving the nonlinear system, the tangent stiffness needs to be evaluated at each

iteration, which would considerably increase the calculation time. Before the gap between

102
tie and ballast closes, ballast layer stiffness remains small (avoid zero for computational

purpose). Once the gap closes, ballast layer stiffness becomes constant. Therefore, an

algorithm that generalizes the aforementioned linear Newmark’s integration method is

presented in Table 4.3 and utilized to solve the nonlinear track system proposed for near

bridge location.

Table 4.3 Algorithm for Numerical Integration using Newmark’s Method for Nonlinear
System (Bilinear)

Initialization of the variables

1. Initial condition: 𝒖𝟎 = 𝟎, 𝒖𝟎̇ = 𝟎, 𝒖𝟎̈ = 𝟎, 𝒕𝟎 = 𝟎

2. Average acceleration method: 𝜸 = 𝟎. 𝟓, 𝜷 = 𝟎. 𝟐𝟓
3. Integration time step: ∆𝒕 = 𝟎. 𝟎𝟎𝟏
Preliminary calculations

4. Calculate the integration constants:

𝟏 𝜸 𝟏 𝟏
𝒂𝟎 =; 𝒂 𝟏 = ; 𝒂 𝟐 = ; 𝒂 𝟑 = − 𝟏;
𝜷∆𝒕𝟐 𝜷∆𝒕 𝜷∆𝒕 𝟐𝜷
𝜸 𝜸
𝒂𝟒 = − 𝟏; 𝒂𝟓 = { − 𝟏| ∆𝒕; 𝒂𝟔 = (𝟏 − 𝜸)∆𝒕; 𝒂𝟕 = 𝜸∆𝒕
𝜷 𝟐𝜷

Integrate step by step

𝒕
for 𝒏 = 𝟎, 𝟏, 𝟐, … , ∆𝒕𝒅 :

5. Increment time: 𝒕𝒏}𝟏 = 𝒕𝒏 + ∆𝒕 = (𝒏 + 𝟏)∆𝒕

¼𝐧 = 𝒌𝐧 + 𝒂𝟎 𝒎 + 𝒂𝟏 𝒄 (check if gap closes)
6. Calculate effective stiffness: 𝒌

à at time 𝒕𝒏}𝟏 :
7. Calculate effective force matrix 𝒑
à 𝒏}𝟏 = 𝒑𝒏}𝟏 + (𝒂𝟎 𝒖𝒏 + 𝒂𝟐 𝒖𝒏̇ + 𝒂𝟑 𝒖𝒏̈ )𝒎 + (𝒂𝟏 𝒖𝒏 + 𝒂𝟒 𝒖𝒏̇ + 𝒂𝟓 𝒖𝒏̈ )𝒄
𝒑
8. Calculate displacement matrix at time 𝒕𝒏}𝟏 :
𝒖𝒏}𝟏 = 𝒑 Ã 𝒏}𝟏 /𝒌¼
9. Calculate acceleration and velocity matrix at time 𝒕𝒏}𝟏 :
𝒖̈ 𝒏}𝟏 = 𝒂𝟎 (𝒖𝒏}𝟏 − 𝒖𝒏 ) − 𝒂𝟐 𝒖𝒏̇ − 𝒂𝟑 𝒖𝒏̈ ; 𝒖̇ 𝒏}𝟏 = 𝒖𝒏̇ + 𝒂𝟔 𝒖̈ 𝒏 + 𝒂𝟕 𝒖̈ 𝒏}𝟏

103
4.6 Summary

This chapter presented the formulation and numerical integration algorithms of a

train-track-bridge model that simulates the vertical dynamic responses when a train vehicle

moves along the rail track structure at bridge approaches. At first, the various components

of the train-track-bridge model was illustrated and explained in detail. Assumptions in the

proposed model were summarized. Next the equations of motion of all the components of

the model were derived assuming that the system reaches dynamic equilibrium at every

time step 𝑡. These equations of motion were then combined to formulate a generalized

matrix.

Newmark’s numerical integration method was chosen to solve the set of ordinary

differential equations for the track system because it can be applied to both a linear system

(open track location) as well as a nonlinear system (near bridge location) with a tie-ballast

gap. The force-displacement relationship between crossties and ballast masses at both open

track and near bridge locations were presented. The spring stiffness was assumed to be

nonlinear at near bridge location. Accordingly, Newmark’s method was modified to

evaluate the tangent stiffness at every step to solve for the nonlinear system within tolerable

error.

The proposed train-track-bridge model can be applied to simulate the dynamic

behavior of a track system, including vertical displacements/velocities/accelerations of

vehicle, rail, crosstie, and ballast masses at different locations along the track, etc. Once all

the parameters (stiffness and damping coefficients) in the model are calibrated with field

data, this model can be utilized to predict and analyze the in-track behavior, especially with

104
proper consideration of nonlinear properties at near bridge locations, and to quantify and

better understand the track transition behavior at bridge approaches.

105
CHAPTER 5: FIELD VALIDATION OF TRACK MODEL

In Chapter 4, a vehicle-track-bridge model was developed to analyze bridge

approach behavior under complex dynamic train loads, and the algorithms for solving the

proposed numerical model were presented. In this chapter, the developed model is

calibrated and validated with field instrumentation data collected from both open track and

near bridge locations. The field instrumentation data are chosen carefully to select track

model input parameters. The validation process for open track location is conducted with

linear model parameters. Then, the near bridge location is studied with consideration of

nonlinear stiffness matrix.

Passage of an ACELA Express passenger train is simulated at the field measured

speed and car weights. The nonlinear force-displacement relationship is approximated by

a simplified bi-linear relationship depending on the field measured gap size between the

hanging crosstie and ballast layer. The validated vehicle-track-bridge model is

demonstrated to effectively simulate and analyze the dynamic responses of track structure

under moving train load. The model is presented as an analytical tool to provide insight

into optimizing track structure designs for improved track performance at bridge

approaches.

5.1 Field Instrumentation Data

As stated in Chapter 3, field instrumentation data including both collected transient

deformations and wheel loads were available at different times and locations. Data

interpretations for track properties were conducted and general trends were clearly

demonstrated in Chapters 3 and 4. This chapter intends to utilize the data collected at the

Upland Street location to calibrate the input parameters in the proposed track model.

106
Therefore, the field data hereafter in this chapter refers to data collected at Upland Street

open track and near bridge locations.

In the field near Chester, PA, the ACELA Express passenger train operates at a

speed of 110 mph when passing the Upland Street location. It consists of eight cars in total,

two locomotives at each end of the train and six passenger cars in the middle. The passenger

car weight is approximately 139 kips. Figure 5.1 shows the ACELA Express axle

configuration. The dimensions are in meters. Input parameters for vehicle part in the

proposed model are selected according to the above mentioned field loading patterns

(speed, weight, axle location).

Figure 5.1 ACELA Express Axle Configuration (Chrismer, 2014)

Transient deformation data with train passage were collected in the field at two

locations: open track location (60 ft. from bridge abutment) and near bridge location (15 ft.

from bridge abutment). Figure 5.2 shows the field deformation data collected during train

passage in January 2015. Note that the January 2015 data were taken as an example to

illustrate the large displacements at near bridge and contrast the trends between open track

and near bridge locations. Such large movements at the bridge approach would lead to

107
accelerated track geometry degradation. Instead of only choosing one set of data (e.g.

January 2015) for the validation purpose, the confidence intervals of field data calculated

from different times were also used to calibrate the input parameters in the proposed track

model. In Chapter 3, the 95% confidence intervals (CI) were calculated and documented.

At the open track location, the peak displacement caused by locomotive is between 1.55

mm and 1.71 mm (95% CI); the peak displacement caused by passenger car is between

1.14 mm and 1.28 mm (95% CI). At the near bridge location, the peak displacement by

locomotive is between 2.10 mm and 5.98 mm (95% CI); the peak displacement by

passenger car is between 1.55 mm and 5.51 mm (95% CI).

Figure 5.2 Transient Deformations Measured in the Field at Upland Street Location –
January 2015.

108
5.2 Model Validation for Open Track Location

To eliminate the boundary effects of vehicle entering and leaving the track structure,

the number of crossties on the embankment side is assumed to be 500 and the number of

crossties on the bridge side is also assumed to be 500. Crossties are located at a spacing of

0.6096 m (2 ft.). In the field open track location, the data were collected at 60 ft. from the

bridge abutment, approximately 30 crossties away from the bridge deck. Figure 5.3

illustrates the corresponding observation point in the model.

The locomotive in this case consisted of one car body of 45,000 kg, two bogies of

1,200 kg and four wheels of 1,200 kg, representing a self-weight of 90,000 kg (198 kips)

car. The passenger car consists of one car body of 31,550 kg, two bogies of 1,200 kg, and

four wheels of 1,200 kg, representing a self-weight of 63,100 kg (139 kips) car. The vehicle

is moving at a speed of 50 m/s. The rail unit weight is 67.46 kg/m. The flexural rigidity of

rail beam is 8.073 𝑀𝑁 ∙ 𝑚M . The concrete crosstie mass is 386 kg. Equivalent ballast mass

is 683 kg. The stiffness and damping ratio values are determined by trial and error within

a reasonable range: Rail pad damping coefficient of 124 𝑘𝑁 ∙ 𝑠/𝑚M ; Rail pad stiffness of

78 𝑀𝑁/𝑚M ; Ballast damping coefficient of 82 𝑘𝑁 ∙ 𝑠/𝑚M ; Ballast stiffness of 120 𝑀𝑁/𝑚M ;

Ballast shear stiffness of 7.8 𝑀𝑁/𝑚M ; Subgrade damping coefficient of 300 𝑘𝑁 ∙ 𝑠/𝑚M ;

and Subgrade stiffness of 50 𝑀𝑁/𝑚M .

109
 Figure 5.3 Sketch of Response Observation Point in the Model – Open Track.

With the ACELA Express passenger train moving from embankment side to bridge

side in the model, the structural responses including transient deformation, reaction force,

velocity and acceleration can be obtained from the model. Transient deformations obtained

from the numerical model are used to validate with field collected data. Figure 5.4 shows

the comparisons between model results and field data. Figure 5.4 (a) clearly shows that the

results from the proposed model match well with the field data collected in January 2015.

The model simulation results demonstrate 32 peaks, which correspond to 8 cars with one

passage of ACELA passenger train. The front and rear cars register higher displacements

due to heavier weight of the locomotives. Figure 5.4 (b) compares the confidence intervals

of field data and model results. It can be seen that the maximum peak displacement value

from the model falls into the 95% confidence interval calculated from the maximum peak

displacement value from the field data collected between August 2012 and May 2015.

110
 (a)

(b)
Figure 5.4 Comparisons of Model Results and Field Data – Open Track.

111
5.3 Model Prediction for Open Track Location

The comparisons between field data and model results imply that responses at the

open track location can be accurately reflected and simulated by the proposed vehicle-

track-bridge model. Therefore, the validated vehicle-track-bridge model can be utilized to

analyze other critical responses under the train passage.

Figure 5.5 shows the transient deformations at the open track location at different

substructure depths, including top of rail, top of crosstie and ballast layer. In this simplified

three-layer linear track structure system, ballast layer contributes approximately half of the

total vertical displacement.

Figure 5.5 Transient Deformations in Different Substructure Layers with Train Passage –
Open Track.

112
 The graphs shown in Figure 5.6 present the (a) rail-crosstie reaction force and (b)

crosstie-ballast reaction force with train passage at the open track location. Note that the

rail-crosstie reaction and the crosstie-ballast reaction forces are similar in both trend and

magnitude. This confirms that the force is correctly simulated in the model since at the

open track location the crossties are in good contact with the underlying ballast particles

and thus the external force exerted on crossties can be passed onto the ballast mass

adequately. Chapter 3 lists that the average measured wheel load on rail crib in the field at

the open track location is around 132 kN for locomotive and 91 kN for passenger car. In

model simulations, the crosstie reaction force is approximately 40 kN for locomotive and

29 kN for passenger car. Hence, it is found that each crosstie takes up to 30%-32% of the

train wheel load.

113
 (a) Rail-Tie Reaction Force

(b) Tie-Ballast Reaction Force

Figure 5.6 Reaction Forces with Train Passage – Open Track.

114
 Besides the reaction force, the structural vibration velocities and accelerations can

also be obtained from the model simulation. Figure 5.7 shows the time history of vibration

velocities of ballast mass, crosstie and rail, respectively. The peaks represent each wheel-

set of the ACELA Express passenger train. Note that the vibration velocities of different

components in the system have similar magnitudes in the order of 10~M m/s. Furthermore,

ballast mass registers the lowest vibration velocity in the system of 0.027 m/s; crosstie

mass shows slightly higher velocity of 0.039 m/s; and rail mass implies the highest velocity

of 0.058 m/s. The time history of vibration accelerations of the track system are graphed in

Figure 5.8. The vibration accelerations of different components are all below one standard

gravitational constant 𝑔 = 9.8 𝑚/𝑠 M . The peak accelerations at the open track location are

approximately 2 𝑚/𝑠 M for ballast mass; 3 𝑚/𝑠 M for crosstie; and 4.5 𝑚/𝑠 M for rail mass.

Figure 5.7 Structural Vibration Velocities with Train Passage – Open Track.

115
 Figure 5.8 Structural Vibration Accelerations with Train Passage – Open Track.

5.4 Model Validation for Near Bridge Location

After successfully calibrating and matching the model results with the open track

location field data, further validation is conducted for the near bridge location. The model

input parameters are kept the same as those for the open track location except for the

observation point in the model and the hanging crosstie gap induced nonlinear stiffness

assignment.

At the field instrumented near bridge location, data were collected at 15 ft. from the

bridge abutment, approximately 7 to 8 crossties away from the bridge deck. Therefore,

crosstie No.493 is chosen as the observation point in the model as shown in Figure 5.9.

116
The time history of structural responses at the observation point are calculated to simulate

dynamic response behavior of the track structure at the near bridge location.

Figure 5.9 Sketch of Response Observation Point in the Model – Near Bridge.

In addition to the changing of observation point, nonlinear ballast stiffness

representing hanging crossties should also be properly simulated. Hanging crossties were

observed in the field at the near bridge location, as explained in Chapter 3, where there

were gaps between these hanging crossties and the underlying ballast particles. At the

instrumented Upland Street location, the gap was found to increase over time when the

track lacked proper maintenance. For calibration and validation purposes, a typical gap of

1.5 mm developed in January 2013 is selected in the model simulation. Such gaps are

created underneath crossties No. 485 to No.500 in the model. Before the gap closes, it is

assumed that no force exists between crosstie and ballast mass; as soon as the gap fully

closes during the loading process, the ballast stiffness abruptly increases to the same as in

117
open track locations – 120 𝑀𝑁/𝑚M . Please note that for simplicity, it is assumed that the

ballast damping ratio remains constant at the near bridge location. Using the nonlinear

algorithm presented in Chapter 4, numerical solutions from the model can be obtained.

To validate the track model, transient deformations predicted by the model

simulations are compared to the field measured transient deformations. Figure 5.10 shows

the comparisons between the model predictions and the field data. The model simulation

results clearly demonstrate 32 peaks, which correspond to 8 cars with one passage of

ACELA passenger train. The front and rear cars register slightly higher displacements due

to heavier self-weights of the locomotives. Due to the existence of gap, the peak

displacement of a moving locomotive can be as high as 2.6 mm. If the gap continues to

increase, it is anticipated that the transient deformation can reach to a significantly high

value at the near bridge location.

Figure 5.10 (a) shows the general trends of transient deformations in terms of peak

location and peak magnitude under loading that are comparable with the field data.

However, the deformations predicted between wheel axles do not match very well with the

field measured data. The physical meaning of this “delayed bouncing” is that in the

simplified physical relationship between displacement and force for hanging crossties,

there is a little reaction force observed when the differential movement between crosstie

and ballast mass is small (𝑘~0). Hence, there is not sufficient upward direction force

exerted on crosstie for the bounce back movement. While in the field, even though there is

no upward force from the underlying ballast, the fastening system will lead the crosstie to

move back to its original position.

118
 Figure 5.10 (b) compares the confidence intervals of field data and model results.

The upper and lower boundaries of 95% confidence intervals are graphed for locomotive

passage and passenger car passage, respectively. Due to the increase in the gap amount

with years, the peak displacement may vary significantly. Therefore, the higher and lower

boundaries of the confidence intervals are relatively far away from each other compared

with the open track case. It is noticed that maximum peak displacement value from the

model still falls into the 95% confidence interval calculated from the maximum peak

displacement value from the field data collected between August 2012 and May 2015.

The proposed model is considered to have the predicton ability for the general

trends and peak responses although it was somewhat unable to predict the bouncing

movement. Considering that the main purpose of the model is to detect the critical

responses, it is believed that the developed model has been successfully validated with field

measurements at the near bridge location.

119
 (a)

(b)

Figure 5.10 Comparisons of Model Results and Field Data – Near Bridge.

120
5.5 Model Prediction for Near Bridge Location

The validated vehicle-track-bridge model can be used to predict other critical

responses, such as deformation at different substructure depths, vibration velocities and

vibration accelerations, under the train passage at near bridge locations.

Figure 5.11 shows the transient deformations at the near bridge location at different

substructure depths, including top of rail, top of crosstie and ballast layer. Recall that with

the exact same input parameters at the open track location, the ballast layer contributes

approximately half of the total vertical deformation. While at the near bridge location, the

ballast layer only contributes 1/3 of the total vertical deformation. Ballast mass itself

undergoes similar amount of deformation. Hanging crossties experience much larger

deformations under the same loading as a result of inadequate support from underlying

ballast layer. These results further explain that in the three-layered track system consisting

of rail, crosstie and ballast, hanging crossties control the total vertical deformation under

train passage. It is of vital importance to mitigate these tie-ballast gaps that occur at the

near bridge to maintain a stable settlement behavior.

121
 Figure 5.11 Transient Deformations within Different Substructure Layers due to Train
Passage – Near Bridge.

Figure 5.12 shows the predicted rail-crosstie reaction force and the crosstie-ballast

reaction force obtained from the vehicle-track-model. Clear loading patterns can be

observed in the time history. However, the nonlinear model at the near bridge location

registers more “noise” when compared to the open track location. From the field measured

data, the average peak locomotive wheel load on rail crib is approximately 148 kN while

the average peak wheel load on rail crib due to passenger car is approximately 100 kN.

From the numerical model predictions, the peak rail-crosstie reaction forces corresponding

to locomotive and passenger car are found to be around 100 kN and 80 kN, respectively.

Even though the impact wheel load on crib is only slightly higher than that at the open

track location, the reaction force on crossties takes up to 67%-80% of the total impact force

at the near bridge location compared with 30% of the total force at the open track location.

122
Such high impact forces would lead to accelerated degradation of both rail and crossties in

the field.

In addition to the rail-tie reaction force, tie-ballast reaction force also exhibits quite

different behavior as compared to the open track location. At the open track location, the

rail-tie reaction and tie-ballast reaction forces are comparable in both trends and magnitude.

However, at the near bridge location, it is noticed that the tie-ballast reaction force is

considerably smaller than the rail-tie reaction force. The difference between the rail-tie

reaction force and tie-ballast reaction force is the total external force on crosstie, which

may possibly cause large vibration accelerations.

123
 (a) Rail-Tie Reaction Force

(b) Tie-Ballast Reaction Force

Figure 5.12 Reaction Forces with Train Passage – Near Bridge.

124
 Figure 5.13 shows the time history of vibration velocities of ballast mass, crosstie

and rail, respectively. Figure 5.14 illustrates the time history of vibration accelerations of

these components. The peaks in the time history are corresponding to the passage of

wheelsets of ACELA Express passenger train. The vibration velocity and vibration

acceleration are good indicators of track structure stability.

From Figure 5.13 it can be seen that vibration velocities at the near bridge location

are within the range of (−0.5 ~ + 0.5) 𝑚/𝑠. The magnitude is almost ten times of that at

the open track location. Larger vibration velocity primarily comes from inadequate support

of the hanging crossties. Rail registers the smallest positive peak vibration velocity of 0.15

𝑚/𝑠. Ballast mass vibrates with a slightly higher velocity of 0.2 𝑚/𝑠. The peak vibration

velocity of crosstie is approximately 0.26 𝑚/𝑠.

Figure 5.14 clearly demonstrates that the track system accelerations at the near

bridge location are significantly higher than those obtained at the open track location. The

standard gravity constant 𝑔 = 9.8 𝑚/𝑠 M is used as the measurement unit. Recall that the

accelerations at the open track location were all below 1𝑔. While the accelerations at the

near bridge location are all above 10𝑔. In fact , the peak vibration acceleration of rail is

about 15 𝑔 under ACELA Express passenger train. Ballast masses experience larger

accelerations under loading with a peak vibration acceleration of 20𝑔. And the crosstie

exhibits the highest vibration acceleration of up to 40𝑔. These high acceleration values

were also reported in a recent related study by Wilk (2017) on track responses with hanging

ties. According to Wilk (2017), tie-ballast impact forces are observed when an unsupported

tie impacts the ballast during train loading within a short duration and the accelerations range

from about 20 to 40g, depending on the magnitude of the tie-ballast gap. Figure 4.10 shows an

125
example of a tie-ballast gap caused behavior, in which the impact produces a 25 to 30g load

for each pass of a double-train truck (back truck and front truck of next car). Higher

accelerations in the track system can lead to component degradation including ballast

breakdown more easily. In conclusion, the track model prediction results imply a greater

risk of material degradation at the near bridge location.

Figure 5.13 Structural Vibration Velocities with Train Passage – Near Bridge.

126
 Figure 5.14 Structural Vibration Accelerations with Train Passage – Near Bridge.

5.6 Sensitivity Analyses on Train Speed and Weight

5.6.1 Speed effect

The validated train-track-bridge model is utilized to study train speed effect on

track responses. ACELA Express passenger trains moving at 50 mph and 90 mph were

studied using the model to predict and compare responses to those obtained with an

operating speed of 110 mph at the NEC bridge approaches. Figures 5.15 and 5.16 present

the predicted total vertical displacement and vibration responses at the open track and near

bridge locations, respectively. For both locations, the vertical displacements in general do

not increase with the increasing train speed. At the open track location, the vibration

velocities and accelerations increase significantly as the train speed increases; however, at

locations with tie gap, the vibration responses are dominated by the tie gap and are not

sensitive to train speed.

127
 1.8 1.8 1.8

1.6 1.6 1.6

Vertical Displacement (mm) 1.4 1.4 1.4

Vertical Displacement (mm)

Vertical Displacement (mm)

1.2 1.2 1.2

1 1 1

0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

0 0 0

-0.2 -0.2 -0.2

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
Time (sec) Time (sec) Time (sec)

(a) (b) (c)

Vibration Velocity (m/s) Vibration Velocity (m/s) Vibration Velocity (m/s)

0.05 0.05 0.05

Ballast
Ballast

Ballast

0 0 0

-0.05 -0.05 -0.05

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

0.05 0.05 0.05

Crosstie
Crosstie

Crosstie

0 0 0

-0.05 -0.05 -0.05

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

0.05 0.05 0.05

Rail
Rail

Rail

0 0 0

-0.05 -0.05 -0.05

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (sec) Time (sec) Time (sec)

(d) (e) (f)

Vibration Accleration (m/s/s) Vibration Accleration (m/s/s) Vibration Accleration (m/s/s)
5 5 5
Ballast

Ballast

Ballast

0 0 0

-5 -5 -5
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

5 5 5
Crosstie
Crosstie

Crosstie

0 0 0

-5 -5 -5
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

5 5 5
Rail
Rail

Rail

0 0 0

-5 -5 -5

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (sec) Time (sec) Time (sec)

(g) (h) (i)

Figure 5.15 Track Vertical Displacements Predicted at Train Speeds of (a) 50 mph, (b) 90
mph and (c) 110 mph at the Open Track Location; Track Vertical Vibration Velocities
Predicted at Train Speeds of (d) 50 mph, (e) 90 mph and (f) 110 mph at the Open Track
Location; Track Vertical Vibration Accelerations Predicted at Train Speeds of (g) 50
mph, (h) 90 mph and (i) 110 mph at the Open Track Location.

128
 3 3 3

2.5 2.5 2.5

Vertical Displacement (mm)

Vertical Displacement (mm)

2

Vertical Displacement (mm)

2 2

1.5 1.5 1.5

1 1 1

0.5
0.5 0.5

0
0 0

-0.5
0 1 2 3 4 5 6 7 8 9 10 -0.5 -0.5
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (sec)
Time (sec) Time (sec)

(a) (b) (c)

Vibration Velocity (m/s) Vibration Velocity (m/s) Vibration Velocity (m/s)
0.5 0.5 0.5
Ballast

Ballast
Ballast

0 0 0

-0.5 -0.5 -0.5

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

0.5 0.5 0.5

Crosstie

Crosstie
Crosstie

0 0 0

-0.5 -0.5 -0.5

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

0.5 0.5 0.5

Rail
Rail

Rail
0 0 0

-0.5 -0.5 -0.5

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (sec) Time (sec) Time (sec)

(d) (e) (f)

Vibration Accleration (m/s/s) Vibration Accleration (m/s/s) Vibration Accleration (m/s/s)
500 500 500
Ballast
Ballast

Ballast

0 0 0

-500 -500 -500

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

500 500 500

Crosstie

Crosstie

Crosstie

0 0 0

-500 -500 -500

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

500 500 500

Rail

Rail

0
Rail

0 0

-500 -500
0 1 2 3 4 5 6 7 8 9 10 -500
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (sec) Time (sec) Time (sec)

(g) (h) (i)

Figure 5.16 Track Vertical Displacements Predicted at Train Speeds of (a) 50 mph, (b) 90
mph and (c) 110 mph at the Near Bridge Location; Track Vertical Vibration Velocities
Predicted at Train Speeds of (d) 50 mph, (e) 90 mph and (f) 110 mph at the Near Bridge
Location; Track Vertical Vibration Accelerations Predicted at Train Speeds of (g) 50
mph, (h) 90 mph and (i) 110 mph at the Near Bridge Location.

129
5.6.2 Weight effect

ACELA Express passenger train used previously for model validation with the

locomotive car body weighing 31,550 kg and vehicle car body weighing 45,000 kg is used

again in this section. To quantify the train weight effect on vertical displacement, vibration

velocity, and vibration acceleration responses, the locomotive weight is increased to 52,800

kg and the vehicle car body weight is increased to 57,664 kg, thus simulating a freight train

weight configuration. Note that the purpose of this analysis is to evaluate the train weight

effect on track responses at bridge approaches. Therefore, few other simplification

assumptions are made: (i) The train geometry remains the same at both weights for the

passenger and freight conditions; (ii) the bogie is assumed to be 2,400 kg and the wheel is

assumed to be 1,200 kg at both freight train weight and passenger train weight levels; and

(iii) the train speed remains the same as 110 mph.

Figure 5.17 presents the track responses predicted at the open track location with

different train weights. The vertical responses generated by heavier train weight (freight)

is higher than those generated by lighter train weight (passenger). A locomotive of 31,550

kg registers a peak displacement of 1.6 mm while a locomotive of 52,800 kg registers a

peak displacement of 1.8 mm. Similar trends are found for vibration responses at the open

track location. At the near bridge location (see Figure 5.18), the predicted vertical

displacements due to the heavier train are much higher than those due to the typical

ACELA Express passenger train, however, the vertical vibration velocities and vibration

accelerations only show slight increases for the heavier freight train.

130
 1.8 1.8

1.6 1.6

1.4 1.4
Vertical Displacement (mm)

Vertical Displacement (mm)

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-0.2 -0.2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(a) (b)

Vibration Velocity (m/s) Vibration Velocity (m/s)

0.05 0.05
Ballast

Ballast
0 0

-0.05 -0.05

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0.05 0.05
Crosstie

Crosstie
0 0

-0.05 -0.05

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0.05 0.05
Rail

Rail

0 0

-0.05 -0.05

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(c) (d)

Vibration Accleration (m/s/s) Vibration Accleration (m/s/s)

10 10
Ballast

Ballast

0 0

-10 -10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10 10
Crosstie

Crosstie

0 0

-10 -10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10 10
Rail

Rail

0 0

-10 -10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(e) (f)

Figure 5.17 Track Vertical Responses for the Heavier Freight Train (a), (c), (e) Compared
to those for ACELA Express Train (b), (d), (f) at the Open Track Location.

131
 3 3

2.5 2.5

Vertical Displacement (mm)

2 2

Vertical Displacement (mm)

1.5 1.5

1 1

0.5 0.5

0 0

-0.5 -0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(a) (b)

Vibration Velocity (m/s) Vibration Velocity (m/s)

0.5 0.5
Ballast

Ballast
0 0

-0.5 -0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0.5 0.5
Crosstie

Crosstie
0 0

-0.5 -0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0.5 0.5
Rail

Rail

0 0

-0.5 -0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(c) (d)

Vibration Accleration (m/s/s) Vibration Accleration (m/s/s)

500 500
Ballast

Ballast

0 0

-500 -500
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

500 500
Crosstie

Crosstie

0 0

-500 -500
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

500 500
Rail

Rail

0 0

-500 -500
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(e) (f)

Figure 5.18 Track Vertical Responses for the Heavier Freight Train (a), (c), (e) Compared
to those for ACELA Express Train (b), (d), (f) at the Near Bridge Location.

132
5.7 Summary

In this chapter, the validation process of the developed vehicle-track-bridge model

with field measured data was presented. The near bridge location at Upland Street was 15

ft. away from the abutment. The open track location was 60 ft. away from the bridge

abutment. Input parameters of the model were carefully chosen so that the model

predictions could match with the field data. Calculated 95% confidence intervals were also

utilized to evaluate the effectiveness of peak responses obtained from the model. For both

locations, the model predictions successfully matched with the field data collected.

This chapter also investigated the model simulation results including reaction

forces, vibration velocities, and vibration accelerations. Analyses show that near bridge

location experiences: (i) larger vertical deformations, especially for crosstie; (ii) larger

impact forces on rail, crosstie, and ballast layer; (iii) larger vibration velocities; and finally,

(iv) larger vibration accelerations.

Furthermore, this chapter conducted a parametric study to investigate effects of

train speed and train weight on track dynamic track responses. Results indicated that train

speed does not have a significant influence on the vertical displacements predicted at both

the open track and the near bridge locations. The predicted vibration velocities and

accelerations, on the other hand, increased with the train speed increasing at the open track

location, but this pheonomenon was not obvious at the near bridge location. Vertical

displacements predicted under a heavier freight train were higher at both the open track

and the near bridge locations.

It is important to note that rail irregularity was not considered in the developed

vehicle-track-bridge model, which could lead to an underestimation of the track system

133
responses. Future studies are recommended to include rail irregularity and its effects on the

predicted dynamic responses.

134
CHAPTER 6: EVALUATION OF MITIGATION METHODS

In this chapter, the validated vehicle-track-bridge model is used to evaluate the

effectiveness of several mitigation methods, which are often adopted to improve the

dynamic performance behavior of track structures at railroad track bridge approaches.

Suggestions on ballasted track designs for newly constructed track transitions as well as

mitigation methods for existing track bridge approaches are provided based on the

evaluation results. To achieve this goal, various combinations of mitigation methods will

be carried out using the validated model. Possible methods include changing of crosstie

spacing, improving substructure properties (rail-pad, ballast etc.), installation of ballast on

open bridge decks, and installation of under tie pads.

6.1 Introduction

There have been several studies proposing in-field mitigation methods for existing

tracks aimed at eliminating the differential movement at railroad track bridge approaches.

Most of the evaluation criteria of proposed mitigation methods highly depend on in-field

performance trends. Field applications of such mitigation methods are usually expensive

and time consuming. Moreover, if the solution conducted is not effective as assumed, it

can result in a waste of resources such as labor and time. Therefore, it is of vital importance

to evaluate these possible mitigation methods for effectiveness potentially with a realistic

track model before any field implementation. This chapter intends to utilize the vehicle-

track-bridge model to conduct parametric studies and evaluate the effectiveness of possible

mitigation methods.

From the structural perspective, there are four major components in the track

system, namely rail, crosstie, ballast, and subgrade. Usually, the rail beam properties are

135
consistent across the track transition zone as the rail beam spreads over a long distance in

the field. Crossties, ballast, subgrade (open track side) and pads between the major

components, on the contrary, are adjustable along the train moving direction. Therefore,

the mitigation plans are focused on modifying the properties of aforementioned track

components for providing a smoother transition between the track bed and bridge deck.

To achieve this ultimate goal, individual effects of rail pad properties, crosstie

properties, and ballast and subgrade properties on the bridge approach dynamic responses

need to be assessed. Based on such a sensitivity study, two mitigation plans are proposed

to be studied in this chapter. The first one is installation of under tie pads at near bridge

locations, and it can be achieved by modifying the damping ratio and stiffness inputs in the

vehicle-track-bridge model. The other proposed mitigation method is installing a ballast

layer on an open bridge deck. This can be simulated by adding a layer of ballast mass and

spring and dampers in the current validated model.

6.2 Sensitivity Study of Selected Track Properties

The effects of changing selected track properties on the dynamic responses have

been studied with the Train-Track-Bridge model. The selected track parameters include:

(1) rail pad stiffness, (2) crosstie spacing, (3) ballast stiffness, and (4) subgrade stiffness.

Table 6.1 lists the default values for these track parameters in the track model. Sections

6.2.1 to 6.2.4 summarize the detailed findings of sensitivity analyses.

136
 Table 6.1 Default Parameters in the Track Model

rail pad stiffness 7.80E+04 kN/m

ballast stiffness 2.00E+05 kN/m

subgrade stiffness 5.00E+04 kN/m

crosstie spacing 0.6096 m

6.2.1 Effect of rail pad stiffness

Within a typical rail pad stiffness range, six values (between 10 𝑡𝑜 1000 𝑀𝑁 ⁄𝑚 /

𝑚) were selected to run the Train-Track-Bridge model at both near bridge and open track

locations. Figures 6.1 and 6.2 show the simulated effects of the rail pad stiffness on the

magnitude of peak transient deformation and peak reaction force between rail and crosstie

at both locations. The results show a clear negative correlation between rail pad stiffness

and vertical deformation, and a clear positive correlation between rail pad stiffness and

rail-tie reaction force. From the perspective of minimizing vertical displacement and

lowering the rail-tie reaction force, the current train-track-bridge model appears to yield an

optimum rail pad stiffness value. It can be seen that when the rail pad stiffness is lower

than 100 𝑀𝑁⁄𝑚 /𝑚, the transient vertical deformation increases rapidly with decreasing

rail pad stiffness, at both near bridge and open track locations. The simulation results also

indicate that at open track location, rail pad stiffness of approximately 100 𝑀𝑁⁄𝑚 /𝑚

could be a turning point for the magnitude of rail-tie reaction force, after which, the

increase of rail-tie reaction force gradually slows down with increase of rail pad stiffness.

While at near bridge location, the rail-tie reaction force continues to increase at a high rate

with increasing rail pad stiffness until stiffness reaches 500 𝑀𝑁⁄𝑚 /𝑚.

137
 Figure 6.1 Effect of Rail-Pad Stiffness on Displacement

Figure 6.2 Effect of Rail-Pad Stiffness on Rail-Tie Reaction Force

138
6.2.2 Effect of crosstie spacing

The sensitivity study of crosstie spacing is carried out via the numerical Train-

Track-Bridge model. Crosstie spacing refers to the distance between two adjacent crossties.

In the field testing and original model setup, the distance between instrumented concrete

crossties has been fixed at 0.6096 m (2 ft.). In practice, it is anticipated that larger spacing

leads to higher deformation as the track stiffness could be lowered. From the simulation

results, it is confirmed that the vertical deformation increases with increasing crosstie

spacing (Figure 6.3). The near bridge location results indicate that by decreasing the

crosstie spacing from 0.6 m to 0.4 m, the vertical displacement decreases by around 10%.

The open track location results indicate a more significant effect from the same amount of

change in crosstie spacing. There is an approximate 20% decrease in the vertical

displacement with the smaller crosstie spacing.

Figure 6.3 Effect of Crosstie Spacing on Vertical Displacement

139
 The magnitude of crosstie spacing also affects greatly the rail-tie reaction force.

The relationship of crosstie spacing and rail-tie reaction force at both open track and near

bridge locations is presented in Figure 6.4. The rail-tie reaction force increases when the

spacing is increased from 0.4 m to 0.6096 m. Specifically, the maximum value of rail-tie

reaction force is up to 39 kN at open track location with a crosstie spacing of 0.6096 m,

whereas the maximum value is only 28 kN with a crosstie spacing of 0.4 m. Similarly, at

near bridge location, the value of rail-tie reaction force can be as high as 92 kN with the

current crosstie spacing value of 0.6096 m, while it decreases to 79 kN when the crosstie

spacing is lowered to 0.4 m.

The sensitivity study of crosstie spacing indicates the potential of decreasing

crosstie spacing in a track system design to mitigate the high deformation and reaction

force responses at near bridge locations.

Figure 6.4 Effect of Crosstie Spacing on Rail-Tie Reaction Force

140
6.2.3 Effect of ballast stiffness

Five ballast stiffness values were selected to conduct sensitivity study through the

Train-Track-Bridge model. The lowest simulated ballast stiffness value is 10 𝑀𝑁⁄𝑚 /𝑚

and the highest value is 1000 𝑀𝑁⁄𝑚 /𝑚. Figures 6.5 and 6.6 illustrate the effect of ballast

stiffness on vertical deformation and rail-tie reaction force, respectively. It is observed that

ballast stiffness has a major impact on the total track system displacement. By increasing

ballast stiffness, the maximum vertical deformation in the track system decreases from 4.3

mm to 1.3 mm at open track location, and from 4.3 mm to 2.3 mm at near bridge location.

Note that the ballast stiffness has a more significant effect on open track location, it is

because the near bridge location encounters a gap between crosstie and ballast layer, thus

the decrease of ballast stiffness will not have a huge impact on the original “larger”

deformation.

Similarly, in Figure 6.6, it can be seen that the rail-tie reaction force increases when

the ballast stiffness increases. Maximum incurred reaction force at open track location is

around 34 kN for a given ballast stiffness of 10 𝑀𝑁⁄𝑚 /𝑚 . If the ballast stiffness

increases to 1000 𝑀𝑁⁄𝑚 /𝑚 , the maximum calculated reaction force at open track

location is 42 kN. At near bridge location, the rail-tie reaction force increases from 34 kN

to 280 kN as the ballast stiffness increases. Interestingly, even though the ballast stiffness

has a smaller impact on deformation at near bridge location, it has a much larger influence

on reaction force in the track system at near bridge location.

Therefore, the magnitude of ballast stiffness should be chosen carefully in track

transition design process to generate an optimal track system. If the ballast stiffness is too

small, the track system yields a high deformation, which is detrimental to the long-term

141
track geometry; on the other hand, if the ballast stiffness is too large, the track system

withstands high forces on track components, which is harmful to the longevity of structure.

Figure 6.5 Effect of Ballast Stiffness on Vertical Displacement

Figure 6.6 Effect of Ballast Stiffness on Rail-Tie Reaction Force

142
6.2.4 Effect of subgrade stiffness

The effect of subgrade stiffness was studied through the Train-Track-Bridge model

by altering the subgrade stiffness parameter in the simulations. The simulations were

conducted for four subgrade stiffness input values, including 10, 50, 100, and 1000

𝑀𝑁⁄𝑚 /𝑚.

The calculated track structural responses, i.e., maximum vertical deformation and

maximum rail-tie reaction force, are graphed in Figures 6.7 and 6.8, respectively. The

vertical deformations are reduced rapidly with increasing subgrade stiffness until

100 𝑀𝑁⁄𝑚 /𝑚, followed by a slower rate of decrease. In a similar manner, the reaction

force increases rapidly with increasing subgrade stiffness until 100 𝑀𝑁⁄𝑚 /𝑚, and slows

down when the subgrade stiffness continues to increase.

Furthermore, the influence of subgrade stiffness on vertical deformation is different

at the two examined locations. At open track location, the maximum calculated vertical

deformation increases by 2.8 mm with subgrade stiffness decreasing from 1000 𝑀𝑁 ⁄𝑚 /

𝑚, which is about 3 times the original deformation (0.9 mm). At near bridge location, the

increase in deformation is approximately 2.5 mm, which is slightly higher than the original

deformation (2.2 mm). The effects of subgrade stiffness on rail-tie reaction force are quite

comparable at both locations, increasing by approximately 7-8 kN for the range of

increases in subgrade stiffness, regardless of open track or near bridge location.

In addition to the previous findings, it is important to note that the effect of subgrade

stiffness on vertical deformation is noticeably larger than that on reaction force. This

indicates that track system deformation is more sensitive to subgrade stiffness. And the

rail-tie reaction force in the system is not affected significantly by the subgrade stiffness.

143
Figure 6.7 Effect of Subgrade Stiffness on Vertical Displacement

Figure 6.8 Effect of Subgrade Stiffness on Rail-Tie Reaction Force

144
6.3 Proposed Mitigation Plan I – Installation of Ballast on Bridge Deck

6.3.1 Converting to ballasted deck bridge

Traditionally railroad bridges are open deck, which means there are no ballast

materials placed on the bridge deck to support the track system. Researchers have

suggested converting open-deck bridges to ballasted deck ones to mitigate track transition

problems (Hyslip et al. 2009). According to the previous studies, the alternatives of

effective mitigation plans should ease the transition from approach embankments to the

fixed bridge structure. It is important to test whether installation of ballast layer on bridge

deck could provide such an easier transition. Hence, converting an open deck bridge to a

ballasted deck bridge was selected as one of the mitigation plans for reducing differential

movement at bridge approaches. Figure 6.9 illustrates the original track structure on bridge

and the proposed ballasted deck bridge.

The proposed plan is to install the same quality of ballast materials on the bridge.

Therefore, all the parameters related to ballast (ballast density, ballast stiffness, and ballast

damping coefficient) are chosen the same as those of the ballast on the embankment side.

145
 (a) Current Open Deck Bridge

(b) Proposed Ballasted Deck Bridge

Figure 6.9 Illustration of Converting Open Deck Bridge to Ballasted Deck Bridge.

6.3.2 Effectiveness of ballasted deck bridge

The effectiveness of installation of ballast layer on the bridge deck is evaluated by

conducting a parametric study with the Train-Track-Bridge model. As shown in Figure 6.9,

ballast masses are added at the bridge side. The gap between tie and ballast at bridge

approach is approximately 1.5 mm. The spring stiffness underneath the ballast mass is still

the same as the previous stiffness on top of bridge deck, which is 100 times higher than the

subgrade stiffness on the embankment side to accommodate the effect of stiff bridge. The

observation point is at the 10th crosstie away from the bridge abutment. Track responses at

146
this observation point were calculated by the model and analyzed for before and after

converting an open deck bridge to ballast deck bridge. The reaction forces predicted in the

track structure upon train passage are illustrated in Figure 6.10.

(a) (b)

(c) (d)

Figure 6.10 Calculated Reaction Forces (a) Rail-Tie before (b) Rail-Tie after (c) Tie-
Ballast before (d) Tie-Ballast after Converting to Ballasted Deck.

From Figure 6.10, it can be seen that installation of ballast layer on bridge deck can

slightly reduce the reaction forces in the track structure. The peak force between rail and

147
crosstie decreases from over 90 kN to 80 kN; and the peak extensive force between crosstie

and ballast mass decreases from 50 kN to 30 kN. Therefore, even though the gap still exists

at near bridge locations, the resultant impact loads between track components can be

reduced due to converting an open deck bridge to a ballasted deck one.

Figure 6.11 demonstrates the calculated transient displacement time history before

and after converting an open deck to a ballasted deck bridge. Interestingly, the installation

of ballast materials on bridge deck does not help to mitigate the high vertical deformations

at near bridge location. On the contrary, the displacements are slightly increased as a result

of the proposed mitigation method. Therefore, ballast on bridge providing a slightly softer

track structure at bridge side will not benefit reducing deformations at near bridge location

where tie gaps may exist. An additional effort may be needed to handle such a problem.

(a) (b)

Figure 6.11 Calculated Vertical Deformation (a) before and (b) after Installation of
Ballast on Bridge Deck.

Besides the transient deformation and force time histories, ballast and crosstie

vibration characteristics are also analyzed using the track model. Figure 6.12 (a) and (b)

148
show the ballast vibration velocities before and after converting to a ballasted deck bridge.

It is noted that peak ballast vibration velocity is greatly reduced from 0.22 m/s to 0.15 m/s.

Moreover, crosstie vibration velocity after the proposed mitigation method has decreased

from 0.25 m/s to 0.2 m/s as shown in Figure 6.12 (c) and (d). Similarly, the ballast and

crosstie vibration accelerations are also calculated and graphed in Figure 6.12 (e) to (h).

Before any mitigation methods, the peak ballast vibration acceleration can be as high as 20

g (gravity coefficient), while the peak crosstie vibration acceleration can reach 30 g, as a

result of hanging tie. Installation of ballast on a bridge deck can help reduce the ballast

acceleration to 15 g, and the crosstie acceleration to 20 g. Reduced vibration responses and

impact loads indicate that ballasted deck bridges can better maintain the track geometry

and slow down the rate of track geometry deterioration.

The aforementioned numerical analyses demonstrate that most of the dynamic

performance indexes of the track are improved by the proposed mitigation method –

converting an open deck bridge to ballasted deck. However, installation of ballast on bridge

deck will not help to mitigate the existing gap and the gap related large deformation and

differential movement issue at near bridge location. Hence, additional methods, for

example, stone blowing at hanging tie locations are needed to treat the problem. Such a

stone blowing application was carried out successfully at the Madison Street bridge

approach location (Boler et al., 2019).

149
 (a) (b)

(c) (d)

Figure 6.12 Calculated Dynamic Responses (a), (c), (e), (g) before and (b), (d), (f), (h)
after Converting Open Deck Bridge to Ballasted Deck Bridge.

150
Figure 6.12 (Continued).

(e) (f)

(g) (h)

6.4 Proposed Mitigation Plan II – Installation of Under Tie Pads (UTPs)

6.4.1 Installation of UTPs

Under Tie Pads (UTPs) are essentially elastic pads placed underneath the crossties

on top of the ballast. The UTPs have been found to reduce the level of vibration caused in

the ballast layers under train loading, and also increase the damping effect of the track

substructure. Moreover, UTPs can reduce the influence of varying track stiffness on the

151
wheel/rail contact force, and as a result distribute the load of the train over a wider area

within the ballast layer. Several successful research and implementation efforts involving

the use of UTPs were discussed in Chapter 2 of this dissertation. Owing to their ability to

facilitate better load transfer onto the ballast layer, UTPs were selected as one of the

remedial measures in the Northeast Corridor field study (Tutumluer et al. 2016). Figure

6.13 shows a schematic layout of the bridge approaches and the locations where individual

remedial measures were applied (Tutumluer et al. 2016).

Figure 6.13 Schematic Layout of Selected Bridge Approaches and Relative Locations of
Remedial Measures (Tutumluer et al. 2016).

A block of 30 crossties with UTPs attached to bottom of the ties were installed on

the Upland Street Bridge south approach. Tracks No. 1 and 4 were kept active during the

installation but No. 2 and No. 3 were out of service to complete the installation. A pre-

constructed concrete tie track panel was brought on site for installation. The installation

was performed by the Amtrak work crews and involved removing the old track, installing

the new track panel, unloading ballast and compacting the track substructure, surfacing and

aligning the track, and finally thermite welding the rail joints. Work was completed in a

weekend by midnight on Saturday 30 August 2014, well within the track outage limits.

152
 According to the manufacturer Pandrol-CDM Track / Novitec, the installed UTPs

were made from polyurethane elastomers manufactured in France from recycled materials

(ground tires), with a shore hardness of 65 to 75. These pads had on average 10% void

content to allow better ballast fit. Based on track geometry records at the Upland Street

south approach (Track No. 2), it was decided that installing 30 new ties with UTPs would

be satisfactory to reduce the train-induced vibrations and subsequent ballast migration.

UTPs with desired engineering properties were shipped to the Amtrak Wilmington Yard

facility. To ensure full adherence of the pads to the bottom of the newly fabricated concrete

ties, the surface was prepared to eliminate all fines or dust. A special adhesive was placed

on the bottom of the crosstie, then the pads were glued and set to dry for 8 to 12 hours. The

tie bottoms were placed mirroring each other with an insulation material (thin carpet) in

order to use the weight of the tie to press down pad and concrete binding. The elastomers

were cut and installed with half an inch offset from the edge of the tie.

The first step in the installation process was cutting the rail at four locations to

remove the 30-crosstie panel from Track No. 2. The new track panel was approximately

60 ft. in length and constructed with 30 concrete ties and 136# RE rail. It was placed on

the east side of the bridge ensuring that 24-in. spacing was standardized between ties. A

120-ton capacity Kirow crane was used to remove the old track panel. Once the space was

cleared, a front-end loader excavated the old ballast from Track No. 2. Ballast was removed

to a depth of 27 inches from the top of rail (TOR), coinciding with the layer of asphalt that

had previously been placed under the tracks. A small amount of ballast was left as a layer

to accommodate the new panel. Next, the new track panel was installed, the rail was cut in

153
order to have staggered welds and finally, the joint bars were fastened. A photograph of

the tie panel after installation is shown in Figure 6.14.

Figure 6.14 Installation of a New Track Panel Having 30 New Concrete Crossties with
Under Tie Pads at the Upland Street Bridge Approach Site near Chester PA.

6.4.2 Effectiveness of UTPs in field

To test the effectiveness of installing under tie pads, track geometry data and

transient track response data were analyzed. Figure 6.15 shows the space curve and running

roughness data for the Upland Street south approach (Track 2) immediately before and

after installation of the track panel with the UTP mounted crossties. As shown in the figure,

installation of the UTPs in late August 2014 resulted in an immediate increase in the

running roughness and is represented by an upward shift in the space curve. This can be

attributed to uneven resurfacing of the track immediately after installation of the track

panels. However, note that the space curve attains a stable configuration within two weeks

after installation of the track panel with UTP mounted ties. A reduction in the running

154
roughness was observed in December 2014. No significant shift in the space curve, or

increase in the roughness is observed till November 2015. This indicates that the track

attained a “stable” configuration after an initial “shakedown,” and installation of the UTPs

led to a relatively consistent track geometry profile.

Track 2 Upland
1.0 100
UTPs installed 8/29/14

Roughness RP 62 60' window (mm2)

RIght Profile Space Curve (inches)

80
0.5
Space Curve

60

0.0
Roughness
40

-0.5
20

-1.0 0
6/1/14 8/1/14 10/1/14 12/1/14 2/1/15 4/1/15 6/1/15 8/1/15 10/1/15 12/1/15

Figure 6.15 Space Curve and Running Roughness Data for the Upland Street Bridge
Approach Showing the Effect of UTP Installation.

A separate instrumentation effort was carried out in August 2015 to measure the

transient response of the track panel under train loading. Note that this instrumentation

effort was carried out eleven (11) months after installation of the track panel with UTP

mounted ties, and hence can be used as an indicator of long-term performance of the UTPs.

Figure 6.16 shows a schematic of the instrumentation layer to measure the transient

response under train loading for this particular bridge approach. As shown in the figure,

the instrumentation was mounted on the 10th and 11th ties from the bridge abutment.

155
Considering a 24-inch tie spacing, these ties are located at distances of approximately 20

and 22 ft. from the bridge abutment, respectively.

Figure 6.16 Schematic of Instrumentation Layout to Measure Transient Response of Ties

Fitted with Under Tie Pads at the the Upland Street Bridge Approach.

Note that the instrumentation effort for UTPs are different from the other locations

mentioned in previous chapters as the transient deformation herein only measures the

ballast layer deformation. Figure 6.17 presents load and displacement time histories

measured at this bridge approach during the passage of an Acela Express train. As shown

in the figure, peak transient displacements of approximately 1.5 mm were measured for

both Tie No. 10 and Tie No. 11. Considering that this measurement was done almost a year

after installation of the track panel with UTP mounted ties, the low peak transient

displacement magnitudes indicate adequate long-term performance of UTPs as a remedial

156
measure. No negative transient deformations were recorded for either tie. This indicates

adequate support conditions underneath both ties existed even one year after the track panel

installation.

Figure 6.17 Force and Displacement Time Histories Measured at the Upland Street
Bridge Approach.

6.4.3 Model predictions of track system responses with UTPs

To evaluate the effectiveness of installing Under Tie Pads (UTPs) on the dynamic

behavior from a mathematical perspective, a parametric study was also conducted using

the Train-Track-Bridge model. At first, the original structural responses at Tie #10 (away

from abutment) were calculated as a baseline. The tie gap value is the same as used in the

previous analysis, which is approximately 1.5 mm. Next, some track parameters were

modified to simulate adding 30 UTPs at near bridge location. Newly calculated track

structural responses were compared with the original ones before installation of UTPs.

157
 Several assumptions were made for estimating the condition of track system with

UTPs. Under Tie Pads have been previously studied to show their effects on the track

system properties by decreasing the stiffness and increasing the damping ratio. Therefore,

it is assumed that at the locations with UTP installation, the ballast stiffness will decrease

from 120 𝑀𝑁⁄𝑚 /𝑚 to 50 𝑀𝑁⁄𝑚 /𝑚 and ballast damping ratio will increase from 0.082

𝑀𝑁. 𝑠⁄𝑚 /𝑚 to 0.15 𝑀𝑁. 𝑠⁄𝑚 /𝑚. Moreover, since the thickness of UTPs are usually

over 1 mm, it is assumed that the gap between crosstie and ballast layer will be closed.

Based on the aforementioned assumptions and the layout information, numerical

simulations were carried out. The observation point was set at 10 crossties away from the

bridge abutment, the same as in the field condition. Figure 6.18 shows the reaction forces

that occurred in the track system before and after UTP installation. Note that since the

thickness of under tie pads would decrease the amount of gap between tie and ballast, the

dynamic reaction forces at near bridge location can be dramatically mitigated.

158
 (a) (b)

(c) (d)

Figure 6.18 Calculated Reaction Forces (a) Rail-Tie before (b) Rail-Tie after (c) Tie-
Ballast before (d) Tie-Ballast after Under Tie Pads (UTPs) Installation.

Figure 6.19 shows the vertical deformation at top of rail, crosstie, and ballast layer

before and after UTP installation at the same location. Before conducting any mitigation

methods, the top of rail vertical deformation upon train passage can be as high as 2.5 mm,

majorly due to the hanging crosstie at near bridge location. With installation of the 30 UTPs

at these problematic locations, the total vertical deformation is now lowered to 2 mm, even

with lower track stiffness (UTP lowers stiffness).

159
 (a) (b)

Figure 6.19 Calculated Vertical Deformation (a) before and (b) after Under Tie Pad
(UTP) Installation.

Taking a closer look into the dynamic behavior of the track system, velocities and

accelerations of the track components are calculated in the model. Figure 6.20 (a) and (b)

illustrates the ballast velocities underneath the 10th crosstie from bridge abutment before

and after UTP installation. Note that the axis limit differs in the two subplots in order to

show the velocity after UTP clearly. The maximum ballast velocity under train passage

decreases from 0.2 m/s to 0.02 m/s. This is mainly attributed to the (i) mitigation of hanging

crosstie condition and (ii) increased damping coefficient due to UTP installation. Figure

6.20 (c) and (d) demonstrates the velocity time histories of the observed crosstie. The

vibration velocity of crosstie can also be controlled via increasing damping coefficient and

mitigating gap between crosstie and ballast. Figure 6.20 (e) to (h) show the ballast and

crosstie vibration accelerations before and after UTPs, respectively. As acceleration is

derivative of velocity, it follows the similar trend of the vibration velocity. Simulation

results indicate that installation of UTPs can potentially help reduce the ballast and crosstie

160
accelerations tremendously from 20 g to 0.2 g (g is gravity constant). Track vibration

acceleration is an important indicator to monitor track degradation and deterioration status.

Therefore, maintaining a relatively low acceleration response shows that installation of

UTPs at near bridge location is beneficial to the track system.

(a) (b)

(c) (d)

Figure 6.20 Calculated Dynamic Responses (a), (c), (e), (g) before and (b), (d), (f), (h)
after Under Tie Pad (UTP) Installation.

161
 Figure 6.20 (Continued).

(e) (f)

(g) (h)

6.5 Ride Quality Analysis

In this section, the validated train-track-bridge model is adopted to study the train

ride quality. Ride quality, also refers to as riding comfort, is a key factor in the designs of

high-speed passenger trains. According to research studies, the riding comfort can usually

be evaluated by the vertical acceleration of the car body and this value should be limited

to a certain maximum to ensure adequate riding comfort (Youcef et al., 2013; Ziyaeifar,

2008; Kargarnovin et al., 2005; Yean and Yeong, 2003; Jong et al., 1999). To study the
162
train ride quality at bridge approaches, ACELA Express passenger train is simulated to

move from the open track onto the bridge deck. The time history of car body acceleration

at the regular operating speed of 110 mph is graphed. In addition, four different speeds

from 22 mph to 200 mph have also been studied to investigate the speed effect. The number

of ties on track and bridge was limited 500 as in the previous modeling cases. To eliminate

any influence of boundary conditions on car body acceleration, the number of ties on the

open track side was increased to 600, 800, 1000, 1200, and 1500 for the 22, 67, 110, 156,

and 200 mph cases, respectively. In accordance, the simulation results obtained are graphed

in Figure 6.21. Because track irregularity is not considered in the model, car body

acceleration at open track side is almost negligible. When the train arrives at a bridge

approach where there are gaps underneath the crossties, the car body acceleration

immediately increases. Similarly, when the train arrives at a transition point between the

bridge deck and open track, the car body acceleration has another significant jump and then

gradually decreases. Figure 6.22 shows the peak car body acceleration varying with the

train speed. Results show that with the increase of train speed, the peak car body

acceleration also increases. When the train operates at 110 mph, approximate peak car body

acceleration at the near bridge location with hanging ties is around 0.11 m/s/s. When the

operating speed increases to 200 mph (high-speed train), the car body acceleration becomes

0.15 m/s/s. Note that in the field the anticipated car body acceleration will be higher due to

track irregularity. At this stage, the train-track-bridge model has been shown to indicate

riding comfort with idealized examples. Once the track irregularity is introduced, the

vertical car body acceleration can be used to measure the train ride quality quantitatively.

163
 0.15 0.15

Embankment Bridge Embankment Bridge

0.1 0.1
Car Body Acceleration (m/s/s)

Car Body Acceleration (m/s/s)

0.05 0.05

0 0

-0.05 -0.05

-0.1 -0.1

Tie gap starts Tie gap starts

-0.15 -0.15

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(a) (b)
0.15 0.15

Embankment Bridge Embankment Bridge

0.1 0.1
Car Body Acceleration (m/s/s)

0.05 Car Body Acceleration (m/s/s) 0.05

0 0

-0.05 -0.05

-0.1 -0.1

Tie gap starts Tie gap starts

-0.15 -0.15

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec) Time (sec)

(c) (d)
0.15

Embankment Bridge
0.1
Car Body Acceleration (m/s/s)

0.05

-0.05

-0.1

Tie gap starts

-0.15

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Time (sec)

(e)

Figure 6.21 Car Body Acceleration at Train Speeds of (a) 10 m/s (~22 mph), (b) 30 m/s
(~67 mph), (c) 50 m/s (~110 mph), (d) 70 m/s (~156 mph), and (e) 90 m/s (~200 mph).

164
 0.16

Acceleratopm (m/s/s)
0.12

0.08

0.04

0
0 50 100 150 200 250
Train speed (mph)

Figure 6.22 Peak Car Body Accelerations at Different Train Speeds.

6.6 Summary

This chapter was intended to provide some insights for future design and

maintenance activities. First, parametric studies were conducted using the Train-Track-

Bridge model by changin rail pad stiffness, ballast stiffness, subgrade stiffness, and crosstie

spacing. Simulation results show that increasing rail pad stiffness, ballast stiffness, and

subgrade stiffness at near bridge approach is effective to decrease the level of settlement at

track bridge approaches. However, increasing the system stiffness could imply higher

reaction force between track components. Therefore, an appropriate value of stiffness

should be selected for ballasted track design to gain optimal system responses. In addition,

results from the model indicate that crosstie spacing is crucial to both vertical deflection

and reaction force. Smaller spacing than typical 2-ft. spacing can reduce the vertical

deformation as well as reaction force between track components. Hence, constructing the

165
bridge approach with gradually decreased crosstie spacing could help to prevent high

dynamic responses from occurring in the track system.

Next, this chapter presented details on the selection of two mitigation methods for

the problem of differential movement at the instrumented railroad track transition sites at

the Amtrak Northeast Corridor lines. The studied remedial measures were (1) installation

of ballast layer on open deck bridge and (2) installation of a track panel having new

crossties with Under Tie Pads (UTPs). They were selected based on the review of published

literature and site inspection for effectiveness at the Amtrak Northeast Corridor. The model

simulation results indicated that installation of ballast layer on bridge deck could help

mitigate differential movement problems and reduce impact forces between track

components, and vibration velocities and accelerations of crosstie and ballast. Therefore,

it is a good remedial application to maintain the current good track performance and slow

down track geometry deterioration in the future. However, it will not help to mitigate the

high vertical deformation that may occur at near bridge locations. If the track in the field

is already experiencing hanging tie issues, more effective remedial measures, such as stone

blowing, are recommended to mitigate the problem.

Field instrumentation data were collected after installation of a new track panel

having 30 concrete ties with UTPs at the Upland Street bridge approach on Amtrak’s

Northeast Corridor line. The field data showed that after one year of UTP installation, the

original shakedown process of newly added ballast was fully completed. Transient

displacement data indicated good crosstie and ballast contact at the near bridge location.

To further examine the dynamic behavior such as velocities and accelerations of track

components, numerical simulations of UTP installation were conducted. Model predictions

166
proved the installation of UTPs to be an effective remedial measure because UTP installed

ties exhibited reduced (1) impact loads and reaction forces, (2) vertical deformations, and

(3) vibration velocity and acceleration responses due to train loading.

Lastly, the proposed train-track-bridge model was adopted to study ride quality.

Vertical car body acceleration was chosen as an indicator for high-speed train riding

comfort. Results show that at open track location, the car body acceleration was negligible;

while at near bridge location, car body acceleration increased at two locations, the starting

point of hanging crossties and bridge deck/open track transition point.

In this chapter, the model simulation results provided some insights into possible

solutions to the differential movement problem at track transition zones, i.e. bridge

approaches specifically. The developed vehicle-track-bridge model has been proven for its

prediciotn ability by conducting parametric studies and detsailed analyses of what if

scenarios to help gain deeper understanding of the track system dynamic responses at

bridge approaches. Further, the studied mitigation methods through the model simulaions

can be utilized as future design baseline.

167
CHAPTER 7: CONCLUSIONS AND RECOMMENDATIONS

This PhD study was aimed to develop a novel train-track-bridge model to

effectively simulate and quantify the dynamic response and performance behavior of track

structures at the transition zone of a bridge approach under moving train loads. In

accordance, the research work undertaken focused on the statistical field data analysis,

analytical model derivation, algorithmic developments, as well as experimental validation

of the developed track model in order to provide improved bridge approach designs and

track maintenance and rehabilitation practices. The objective was to better understand

governing mechanisms of bridge approach problems that occur near bridge abutments and

utilize the train-track-bridge model to help with designing new bridge approaches and

selecting mitigation methods. This chapter summarizes the main research findings and

contributions of this thesis study and provides suggestions for future work.

7.1 Research Findings

Chapter 3 summarized statistical data analyses of track transient response data

collected at the instrumented bridge approaches at Amtrak’s Northeast Corroidor high

speed passenger lines near Chester, Pennsylvania, including calculated means, standard

deviations, and 95% confidence intervals. Chapter 4 presented the developed ballasted

train-track-bridge model and solving scheme for both open track (linear) and near bridge

(nonlinear) locations. Chapter 5 presented the validation of the simulation results from the

developed ballasted train-track-bridge with the field data. The validation results showed

that the deformations obtained from the track model successfully matched with the

deformation data collected from the field instrumentation. Chapter 6 discussed various

168
mitigation methods applicable to track bridge approaches including changing rail pad

stiffness, ballast stiffness, subgrade stiffness, and crosstie spacing. Furthermore, Chapter 6

presented details on the selection of two mitigation methods for remedying the problem of

differential movement at the instrumented railroad track transition sites at the Amtrak’s

Northeast Corridor, namely, (1) converting an open deck bridge to ballasted deck bridge

and (2) installation of a new track panel having concrete ties with Under Tie Pads (UTPs).

Based on the study, the following findings are drawn concerning the field

performance, track model development, and remedial measures at bridge approaches:

- At near bridge locations, transient vertical deformations measured showed an

increasing trend with time, while the vertical wheel loads were relatively stable,

indicating noticeable negative changes which were clearly occurring in track

substructure support conditions with time.

- 95% confidence interval analyses showed significant statistical differences between

dynamic responses measured at open track and near bridge locations; much higher

vertical deformations and vibrations velocities and accelrations were taking place

at near bridge locations.

- Tie gaps existed at both open track and near bridge locations. The amount of tie

gap at near bridge location could be as high as 3.9 mm at the investigated Upland

Street bridge approach site. Whereas, the amount of tie gap at open track location,

60 ft away from the Upland Street bridge abutment, was approximately 0.2-0.3 mm.

- Hanging tie condition at near bridge location could be succesfully simulated as a

nonlinear spring between crosstie and ballast masses in the novel train-track-bridge

model.

169
- Nonlinear properties could be properly accounted for in the developed track model

by implementing Newton’s method together with Newmark’s numerical integration

scheme.

- Track model predictions indicated that at open track locations, each crosstie took

approximately 30%-32% of the train wheel load. On the contrast, crossties might

take up to 67% of the train wheel load at near bridge locations.

- The amount of track settlement at near bridge locations could be decreased by

properly increasing rail pad stiffness, ballast stiffness, and subgrade stiffness.

- Decreasing crosstie spacing was shown to reduce the vertical deformation as well

as reaction forces between track components, which could prevent high dynamic

responses from occurring in track system.

- Installation of a ballast layer on an open deck bridge could help to reduce impact

forces between track components, vibration velocities and accelerations of crosstie

and ballast. Therefore, it would be considered beneficial to slow down track

geometry degradation. However, converting an open deck bridge to a ballasted deck

would not help to mitigate the high vertical deformations often occurred at near

bridge locations. If the track in field already suffered from major hanging tie issues,

such as existing impact forces and ballast-tie gaps under train loading, additional

remedial methods, such as stone blowing, should be employed to mitigate the

problem.

- Field instrumentation data indicated good crosstie and ballast contact at near bridge

locations almost one year after the installation of a new track panel having 30

concrete crossties with UTPs. Numerical simulation of UTP installation also proves

170
 installation of UTPs to be effective remedial measures because UTP installed ties

exhibited reduced (1) impact loads and reaction forces, (2) vertical deformations,

and (3) vibration velocity and acceleration responses due to train loading.

7.2 Recommendations for Future Research

This dissertation has provided the formulation of a novel train-track-bridge model

that could be used to test the effectiveness of proposed mitigation methods for solving

transition zone problems as well as improve future track designs at bridge approaches.

Many questions remain open regarding more detailed structural analysis and field

performance prediction. There are several important directions for future research.

One important question concerns the development of three-dimensional train-track-

bridge model to include lateral forces and derailment potential analyses. Upon

improvement of the existing model, one can also take track irregularity into consideration

and establish tamping and other track maintenance schedules for maintaining proper track

geometry and other rehabilitation consideration for deteriorated track.

Another important question concerns optimal choices of track input properties for

overall track design and maintenance guidance. For example, if necessary, the track design

could be improved and guided by optimization of the numerical train-track-bridge model

outputs given a certain track condition, such as train speed, subgrade modulus, etc.

Finally, it would be useful to collect field geometry data over a long period of time

and combine it with the numerical model for the transient force and displacement field data

collection. The combined permanent and transient deformation data can be analyzed using

171
machine learning techniques to predict important indexes of track structure, such as

permanent deformation or settlement, differential movement, potential of derailment, etc.

172
REFERENCES

[1] Banimahd, M., Woodward, P. K., Kennedy, J., Medero, G. M. Behavior of train–

track interaction in stiffness transitions. In Proceedings of the Institution of Civil

Engineers-Transport 2012; 165(3): 205-214.

[2] Basu, D., Kameswara Rao., N. S. V. Analytical solutions for Euler–Bernoulli beam

on viscoelastic foundation subjected to moving load. International Journal for

Numerical and Analytical Methods in Geomechanics 2013; 37(8):945-960.

[3] Bian, X., Chen, Y., Hu, T. Numerical simulation of high-speed train induced

ground vibrations using 2.5 D finite element approach. Science in China Series G:

Physics, Mechanics and Astronomy 2008; 51(6): 632-650.

[4] Bilow, D., Li, D. Concrete slab track test on the high tonnage loop at the

Transportation Technology Center. In Proceedings of the 2005 AREMA Annual

Conference; 2005.

[5] Biondi, B., Muscolino, G., Sofi, A. A substructure approach for the dynamic

analysis of train–track–bridge system. Computers & structures 2005; 83(28): 2271-

2281.

[6] Boler, H. (2012). On the shear strength of polyurethane coated railroad ballast.

Master Thesis. University of Illinois at Urbana-Champaign, U.S.

[7] Boler, H., Mishra, D., Hou, W., & Tutumluer, E. (2018). Understanding track

substructure behavior: Field instrumentation data analysis and development of

numerical models. Transportation Geotechnics, 17, 109-121.

[8] Boler, H., Mishra, D., Tutumluer, E., Chrismer, S., & Hyslip, J. P. (2019). Stone

blowing as a remedial measure to mitigate differential movement problems at

173
 railroad bridge approaches. Proceedings of the Institution of Mechanical Engineers,

Part F: Journal of Rail and Rapid Transit, 233(1), 63-72.

[9] Briaud, J. L., James, R. W., & Hoffman, S. B. (1997). Settlement of bridge

approaches: (the bump at the end of the bridge) (Vol. 234). Transportation Research

Board.

[10] Briaud, J. L., Nicks, J. E., Smith, B. J. The bump at the end of the railway bridge,

Final Report, Texas A&M University; 2006.

[11] Cantero, D., Arvidsson, T., OBrien, E., & Karoumi, R. (2016). Train–track–bridge

modelling and review of parameters. Structure and Infrastructure Engineering,

12(9), 1051-1064.

[12] Chang, C. S., Selig, E. T., Adegoke, C. W. Geotrack model for railroad truck

performance. Journal of the Geotechnical Engineering Division 1980; 106(11):

1201-1218.

[13] Chawla, S., & Shahu, J. T. (2016). Reinforcement and mud-pumping benefits of

geosynthetics in railway tracks: numerical analysis. Geotextiles and

Geomembranes, 44(3), 344-357.

[14] Chrismer, S. Track surfacing with conventional tamping and stone injection.

Association of American Railroads Research Report No. R-719, AAR Technical

Center, Chicago; 1990.

[15] Coelho, B., Priest, J., Holscher, P., Powrie, W. Monitoring of transition zones in

railways. Engineering Technics Press; 2009.

174
[16] Coelho, B., Hölscher, P., Priest, J., Powrie, W., & Barends, F. (2011). An

assessment of transition zone performance. Proceedings of the Institution of

Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 225(2), 129-139.

[17] Czyczula, W., Koziol, P., Blaszkiewicz, D. On the Equivalence between Static and

Dynamic Railway Track Response and on the Euler-Bernoulli and Timoshenko

Beams Analogy. Shock and Vibration 2017; 2017: 2701715.

[18] Davis, S. C., Boundy, R. Transportation Energy Data Book. Oak Ridge National

Laboratory; 2018.

[19] Dersch, M. S., Tutumluer, E., Peeler, C. T., & Bower, D. K. (2010). Polyurethane

coating of railroad ballast aggregate for improved performance. In 2010 Joint Rail

Conference (pp. 337-342). American Society of Mechanical Engineers.

[20] De Beer, M., Horak, E., Visser, A. T. The multidepth deflectometer (MDD) system

for determining the effective elastic moduli of pavement layers. In Nondestructive

Testing of Pavements and Backcalculation of Moduli. ASTM International; 1989.

[21] Feng, H. (2011). 3D-models of railway track for dynamic analysis. Master Thesis.

Royal Institute of Technology, Sweden.

[22] FIMOR (2015). Fimor products for track superstructure. In Railway Engineering

13th International Conference & Exhibition. Edinburgh, U.K.

[23] Frohling, R. D., Scheffel, H., Ebersöhn, W. The vertical dynamic response of a rail

vehicle caused by track stiffness variations along the track. Vehicle System

Dynamics 1996; 25(S1): 175-187.

[24] Giner, I. G., López Pita, A., Vieira Chaves, E. W., & Rivas Álvarez, A. M. (2012).

Design of embankment–structure transitions for railway infrastructure. In

175
 Proceedings of the Institution of Civil Engineers-Transport (Vol. 165, No. 1, pp.

27-37). Thomas Telford Ltd.

[25] Gräbe, P. J., Shaw, F. J. Design life prediction of a heavy haul track foundation.

Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and

Rapid Transit 2010; 224(5): 337-344.

[26] Gräbe, P. J., Mtshotana, B.F, Sebati, M.M., & Thunemann, E.Q. (2015). The effects

of undersleeper pads on ballast contact pressure, settlement and breakdown.

Transportation Research Board 94th Annual Meeting. Washington, D.C.

[27] Hayano, K., Ishii, K., & Muramoto, K. (2013). Effects of ballast thickness and tie-

tamper repair on settlement characteristics of railway ballasted tracks. Railway

Technical Research Institute. Geotechnical Pavement Research in Japan II, final

report WG4.

[28] Hayano,K., Hoshi, S., & Kumara, J.J. (2013). Soil mass density test on railway

ballasts with using a 3D digital camera. Railway Technical Research Institute.

Geotechnical Pavement Research in Japan II, final report WG4.

[29] Holm, G., Andréasson, B., Bengtsson, P. E., Bodare, A., Eriksson, H. Mitigation of

track and ground vibrations by high speed trains at Ledsgard, Sweden. Swedish

Deep Stabilization Research Centre Report 2002; 10.

[30] Huang, H., Shen, S., Tutumluer, E. Sandwich model to evaluate railroad asphalt

trackbed performance under moving loads. Transportation Research Record:

Journal of the Transportation Research Board 2009; (2117): 57-65.

176
[31] Huang, H., Gao, Y., Stoffels, S. Fully coupled three-dimensional train-track-soil

model for high-speed rail. Transportation Research Record: Journal of the

Transportation Research Board 2014; (2448): 87-93.

[32] Hyslip, J. P., Li, D., McDaniel, C. R. Railway bridge transition case study. In

Bearing Capacity of Roads, Railways and Airfields, 8th International Conference,

Champaign, Illinois (BCR2A'09); 2009.

[33] Indraratna. B., Sajjad, M., Ngo, T., el at. Improved performance of ballasted tracks

at transition zones: A review of experimental and modelling approaches.

Transportation Geotechnics 2019; 21: 100260.

[34] Insa, R., Salvador, P., Inarejos, J., & Roda, A. (2012). Analysis of the influence of

under sleeper pads on the railway vehicle/track dynamic interaction in transition

zones. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of

Rail and Rapid Transit, 226(4), 409-420.

[35] Johansson, A., Nielsen, J. C., Bolmsvik, R., Karlström, A., & Lundén, R. (2008).

Under sleeper pads—Influence on dynamic train–track interaction. Wear, 265(9),

1479-1487.

[36] Jong, D. Y., Yeong, B. Y. and Shyh, R. K. Impact response of high speed rail

bridges and riding comfort of rail cars. Engineering Structures 1999; 21(9): 836-

844.

[37] Johnson, N. and Arzamendi, M. Lightweight Cellular Concrete Fill to Mitigate

Railroad Bridge approach Settlement. In: Western Bridge Engineer’s Seminar

Practical Solutions To Bridge Engineering Challenges. 2015.

177
[38] Judge, A., Ho, C., Huang, H., Hyslip, J., & Gao, Y. (2018). Field Investigation and

Modeling of Track Substructure Performance Under Trains Moving at Critical

Speed (No. DOT/FRA/ORD-18/21).

[39] Kaewunruen, S., & Tang, T. (2019). Idealisations of Dynamic Modelling for

Railway Ballast in Flood Conditions. Applied Sciences, 9(9), 1785.

[40] Kaewunruen, S., Remennikov, A., Aikawa, A. A numerical study to evaluate

dynamic responses of voided concrete railway sleepers to impact loading 2011.

[41] Kalker, J. J. Discretely supported rails subjected to transient loads. Vehicle System

Dynamics 1996; 25(1): 71-88.

[42] Kargarnovin, M.H., Younesian D., Thompson, D. and Jones, C. Ride comfort of

high-speed trains travelling over railway bridges. Vehicle System Dynamics 2005;

43(3): 173-197.

[43] Keene, A., & Edil, T. B. (2012). Mitigating Ballast Fouling Impact and Enhancing

Rail Freight Capacity. National Center for Freight & Infrastructure Research &

Education, University of Wisconsin-Madison.

[44] Kennedy, J., Woodward, P. K., Medero, G., & Banimahd, M. (2013). Reducing

railway track settlement using three-dimensional polyurethane polymer

reinforcement of the ballast. Construction and Building Materials, 44, 615-625.

[45] Kerr, Arnold D., and Brian E. Moroney. "Track transition problems and remedies."

Bulletin 1993; (742): 267-298.

[46] Kerr, A., Bathurst, L. (2001). Upgrading track transitions for high-speed service.

Report No. DOT/FRA/RDV-02/05, Federal Railroad Administration, US DOT.

178
[47] Koch, E., Hudacsek, P. 3D Dynamic Modeling of Transition Zones. World

Academy of Science, Engineering and Technology 2017; 11(9): 1310-1316.

[48] Kouroussis, G., Vogiatzis, K., Connolly, D. A combined numerical/experimental

prediction method for urban railway vibration. Soil Dynamics and Earthquake

Engineering 2017; 97: 377-386

[49] Kouroussis, G., Vogiatzis, K., Connolly, D. Modelling the Environmental Effects

of Railway Vibrations from Different Types of Rolling Stock: A Numerical Study.

Shock and Vibration 2015: 142807.

[50] Krylov, V. V. Generation of ground vibrations by superfast trains. Applied

Acoustics 1995; 44(2): 149-164.

[51] Lee, H. P. Dynamic response of a Timoshenko beam on a Winkler foundation

subjected to a moving mass. Applied Acoustics 1998; 55(3): 203-215.

[52] Lei, X., Noda, N. A. Analyses of dynamic response of vehicle and track coupling

system with random irregularity of track vertical profile. Journal of Sound and

Vibration 2002; 258(1): 147-165.

[53] Lei, X., & Rose, J. G. (2008). Track vibration analysis for railways with mixed

passenger and freight traffic. Proceedings of the Institution of Mechanical

Engineers, Part F: Journal of Rail and Rapid Transit, 222(4), 413-421.

[54] Le Pen, L., Watson, G., Hudson, A., & Powrie, W. (2018). Behaviour of under

sleeper pads at switches and crossings–Field measurements. Proceedings of the

Institution of Mechanical Engineers, Part F: Journal of rail and rapid transit, 232(4),

1049-1063.

179
[55] Li, X., Ekh, M., Nielsen, J. Three-dimensional modelling of differential railway

track settlement using a cycle domain constitutive model. International Journal for

Numerical and Analytical Methods in Geomechanics 2016; 40(12): 1758-1770.

[56] Li, D., & Maal, L. (2015). Heavy axle load revenue service bridge approach

problems and remedies. In 2015 Joint Rail Conference (pp. V001T01A020-

V001T01A020). American Society of Mechanical Engineers.

[57] Li, D., Selig, E. T. Method for railroad track foundation design I: Development.

Journal of Geotechnical and Geoenvironmental Engineering 1998; 124(4): 316-322.

[58] Li, D., Davis, D. Transition of railroad bridge approaches. Journal of Geotechnical

Engineering 2005; 131(11):1392-1398.

[59] Ling, L., Xiao, X., Xiong, J., el at. A 3D model for coupling dynamics analysis of

high-speed train/track system. Journal of Zhejiang University SCIENCE A 2014;

15(12): 964-983.

[60] Lundqvist, A. N. D. R. E. A. S., Larsson, R. I. K. A. R. D., & Dahlberg, T. O. R. E.

(2006). Influence of railway track stiffness variations on wheel/rail contact force.

Track for High-Speed Railways, Porto, Portugal.

[61] McMichael, P., & McNaughton, A. (2003). The stoneblower-delivering the

promise: development, testing and operation of a new track maintenance system.

In TRB 2003 Annual Meeting CD-ROM.

[62] Mise, K., Kunii, S. A theory for the forced vibrations of a railway bridge under the

action of moving loads. The Quarterly Journal of Mechanics and Applied

Mathematics 1956; 9(2): 195-206.

180
[63] Mishra, D., Tutumluer, E., Stark, T. D., Hyslip, J. P., Chrismer, S. M., Tomas, M.

Investigation of differential movement at railroad bridge approaches through

geotechnical instrumentation. Journal of Zhejiang University SCIENCE A 2012;

13(11): 814-824.

[64] Mishra, D., Tutumluer, E., Boler, H., Hyslip, J., Sussmann, T. Railroad track

transitions with multidepth deflectometers and strain gauges. Transportation

Research Record: Journal of the Transportation Research Board 2014; (2448): 105-

114.

[65] Mishra, D., Boler, H., Tutumluer, E., Hyslip, J. P. Effectiveness of Chemical

Grouting and Stone Blowing as Remedial Measures to Mitigate Differential

Movement at Railroad Track Transitions. In 2016 Joint Rail Conference. American

Society of Mechanical Engineers 2016: V001T01A030-V001T01A030.

[66] Moale, C., Wilk, S., Smith, D., el at. Design and Performance of Three Remediated

Bridge Approaches. AREMA; 2016

[67] Muramoto, K. Geotechnical challenges for railway tracks & substructures

including transition zone. Railway Technical Research Institute. Geotechnical

Pavement Research in Japan II, Final Report WG4 2013.

[68] Muramoto, K., Nakamura, T., Sakurai, T. A study of the effect of track irregularity

prevention methods for the transition zone between different track structures.

Quarterly Report of RTRI 2012; 53(4): 211-215.

[69] Namura, A., Suzuki, T. Evaluation of countermeasures against differential

settlement at track transitions. Quarterly Report of RTRI 2007; 48(3): 176-182.

181
[70] Nicks, J. The bump at the end of the railway bridge. PhD Thesis, Texas A&M

University, College Station, TX; 2009.

[71] Nielsen, J. C., Oscarsson, J. Simulation of dynamic train–track interaction with

state-dependent track properties. Journal of sound and vibration 2004; 275(3): 515-

532.

[72] Özceylan, E. A decision support system to compare the transportation modes in

logistics. International Journal of Lean Thinking 2010; 1(1): 58-83.

[73] Paixão, A., Fortunato, E., Calçada, R. Design and construction of backfills for

railway track transition zones. Proceedings of the Institution of Mechanical

Engineers, Part F: Journal of Rail and Rapid Transit 2015; 229(1): 58-70.

[74] Paixão, A., Alves Ribeiro, C., Pinto, N., Fortunato, E., & Calçada, R. (2015). On

the use of under sleeper pads in transition zones at railway underpasses:

experimental field testing. Structure and Infrastructure Engineering, 11(2), 112-128.

[75] Paultre, P. (2013). Dynamics of structures. John Wiley & Sons.

[76] Pen, L.L., Abadi, T., Hudson, A., Zervos, A., Powrie, W. (2015). The use of under

sleeper pads to improve the performance of ballasted railway track at switches and

crossings: a case study.

[77] Plotkin, D., Davis, D. Bridge approaches and track stiffness. Final Report,

DOT/FRA/ORD-08-01, Federal Railroad Administration; 2008.

[78] Priest, J. A., Powrie, W., Yang, L., Grabe, P. J., Clayton, C. R. I. Measurements of

transient ground movements below a ballasted railway line. Géotechnique 2010;

60(9): 667-677.

182
[79] Puppala, A. J., Saride, S., Archeewa, E., Hoyos, L. R., Nazarian, S.

Recommendations for design, construction, and maintenance of bridge approach

slabs: Synthesis report 2009: 6022-1.

[80] Read, D., Li, D. Design of track transitions. Transit cooperative research program:

research results digest 79. Federal Transit Administration; 2006.

[81] Ribeiro, Alves. C., Paixão, A., Fortunato, E., & Calçada, R. (2015). Under sleeper

pads in transition zones at railway underpasses: numerical modelling and

experimental validation. Structure and Infrastructure Engineering, 11(11), 1432-

1449.

[82] Rose, J., Anderson, J. Long-term performance of asphalt underlayment trackbeds

for special trackbed applications. In Proceedings, American Railway Engineering

and Maintenance-of-Way Association (AREMA) Annual Conference and

Exposition; 2006.

[83] Sasaoka, C. D., Davis, D. Implementing track transition solutions for heavy axle

load service. In Proceedings, AREMA 2005 Annual Conference, Chicago, IL; 2005.

[84] Sasaoka, C., Davis, D., Koch, K., Reiff, R., & GeMeiner, W. (2005). Implementing

track transition solutions. Technology Digest TD-05-001, Transportation

Technology Center Inc., Pueblo, CO.

[85] Sañudo, R., Dell'Olio, L., Casado, J. A., Carrascal, I. A., Diego, S. Track transitions

in railways: A review. Construction and Building Materials 2016; 112: 140-157.

[86] Sakurai, T. , Nakamura, T. , and Muramoto, K. Cyclic Loading Test with Full-Scale

Model for Pre-Stressed Ballast Track. Geotechnical Pavement Research in Japan II,

Final Report WG4. Railway Technical Research Institute, Tokyo, 2013, pp. 175–

183
 178. Reproduced in Geotechnical Pavement Research in Japan II—Final Report by

Japanese Local Task Force Committee for ISSMGE Technical Committee on

Transportation Geotechnics (TC202) and distributed at the 18th International

Conference on Soil Mechanics and Geotechnical Engineering (ISSMGE) in Paris,

Sept. 2–6, 2013, pp. 175–178.

[87] Selig, E. T., & Li, D. (1994). Track modulus: Its meaning and factors influencing

it. Transportation Research Record, (1470).

[88] Selig, E. T., & Waters, J. M. (1994). Track geotechnology and substructure

management. Thomas Telford.

[89] Schneider, P., Bolmsvik, R., & Nielsen, J. C. (2011). In situ performance of a

ballasted railway track with under sleeper pads. Proceedings of the Institution of

Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 225(3), 299-309.

[90] Scullion, T., Briggs, R. C., Lytton, R. L. Using the multidepth deflectometer to

verify modulus backcalculation procedures. In Nondestructive Testing of

Pavements and Backcalculation of Moduli. ASTM International; 1989.

[91] Sharpe, P., Armitage, R. J., Heggie, W. G., & Rogers, A. (2002, July). Innovative

design of transition zones. In Proceedings of the International Conference Railway

Engineering.

[92] Silva, É. A., Pokropski, D., You, R., & Kaewunruen, S. (2017). Comparison of

structural design methods for railway composites and plastic sleepers and

bearers. Australian journal of structural engineering, 18(3), 160-177.

184
[93] Smith, M. E., Bengtsson, P. E., & Holm, G. (2006). Three-dimensional analyses of

transition zones at railway bridges. Numerical methods in geotechnical engineering,

237-242.

[94] Stark, T. D., Wilk, S. T., Thompson, H. B., & Sussmann, T. R. (2015, June). Effect

of Unsupported Ties at Transition Zones. In Proceedings of railway engineering-

2015 conference.

[95] Sussmann, T., Selig, E. Evaluation of increased axle loading on northeast corridor

track substructure. Test Report for Amtrak (National Railroad Passenger

Corporation), Ernest T. Selig Inc., Hadley, MA; 1998.

[96] Szycher, M. (Ed.). (1999). Szycher's handbook of polyurethanes. CRC press.

[97] Tanabe, M., Wakui, H., Matsumoto, N., Okuda, H., Sogabe, M., Komiya, S.

Computational model of a Shinkansen train running on the railway structure and

the industrial applications. Journal of Materials Processing Technology 2003;

140(1): 705-710.

[98] Tutumluer, E., Stark, T. D., Mishra, D., Hyslip, J. P. Investigation and mitigation

of differential movement at railway transitions for US high speed passenger rail

and joint passenger/freight corridors. In 2012 Joint Rail Conference. American

Society of Mechanical Engineers; 2012: 75-84.

[99] Tutumluer, E., Mishra, D., Boler, H., Hyslip, J.P., Thomas, M. Dynamic Responses

of Bridge Approach Concrete Ties after the Applications of Polyurethane Injection,

Stone Blowing and Under-Tie Pad Remedial Measures. In International Crosstie

and Fastening System Symposium. June 2016.

185
[100] Varandas, J. N., Hölscher, P., Silva, M. A. Dynamic behaviour of railway tracks on

transitions zones. Computers & structures 2011; 89(13): 1468-1479.

[101] Varandas, J. Long-term behaviour of railway transitions under dynamic loading.

Portugal: Faculdade de Cieˆncias e Tecnologia da Universidade Nova de Lisboa;

2013. Ph.D. Thesis.

[102] Vorster, D. , Gräbe, P. J. The effect of axle load on track and foundation resilient

deformation under heavy haul conditions. India: New Delhi, In 10th international

heavy haul association conference; 2013.

[103] Wang, H., Markine, V. L., Shevtsov, I. Y., Dollevoet, R. Analysis of the dynamic

behaviour of a railway track in transition zones with differential settlement. In 2015

Joint Rail Conference. American Society of Mechanical Engineers; 2015:

V001T01A024-V001T01A024.

[104] Wang, H., Markine, V. Methodology for the comprehensive analysis of railway

transition zones. Computers and Geotechnics 2018; 99: 64-79

[105] Wang, H., Markine, V. Dynamic behaviour of the track in transitions zones

considering the differential settlement. Journal of Sound and Vsibration 2019; 459:

114863.

[106] Warren, B.J. (2015). Field application of expanding rigid polyurethane stabilization

of railway track substructure. Master Thesis. University of Wisconsin-Madison,

U.S.

[107] Wilk, S. T. (2017). Mitigation of differential movements at railroad bridge

transition zones (Doctoral dissertation, University of Illinois at Urbana-

Champaign).

186
[108] Witt, S. (2008). The influence of under sleeper pads on railway track dynamics.

Sweden.

[109] White, D., Sritharan, S., Suleiman, M., Mekkawy, M., & Chetlur, S. (2005).

Identification of the Best Practices for Design, Construction, and Repair of Bridge

Approaches. Final Report, Iowa DOT Project TR-481, Center for Transportation

Research and Education, Iowa State University.

[110] Woodward, P.K., Thompson, D., Banimahd, M. Geocomposite technology:

reducing railway maintenance. Proceedings of the ICE-Transport 2007; 160(3):

109-115.

[111] Woodward, P. K., Kennedy, J., Medero, G. M., & Banimahd, M. (2012).

Maintaining absolute clearances in ballasted railway tracks using in situ three-

dimensional polyurethane GeoComposites. In Proceedings of the Institution of

Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 226(3), 257-271.

[112] Yang, Y. B, Yau, J. D., Hsu, L. C. Vibration of simple beams due to trains moving

at high speeds. Engineering structures 1997; 19(11): 936-944.

[113] Yang, C., Sun, H., Zhang, J. Dynamic responses of bridge-approach embankment

transition section of HSR. Journal of Central South University 2013; 20(10): 2839-

2839.

[114] Yean, S. W. and Yeong, B. Y. Steady-state response and riding comfort of trains

moving over a series of simply supported bridges. Engineering Structures 2003;

25(2): 251-265.

187
[115] Youcef, K., Sabiha, T., Mostaf, D., Ali, D. and Bachi, M.. Dynamic analysis of

train-bridge system and riding comfort of trains with rail irregularities. Journal of

Mechanical Science and Technology 2013;27(4): 951-962.

[116] Yin, C., Wei, B. Numerical simulation of a bridge-subgrade transition zone due to

moving vehicle in Shuohuang heavy haul railway. Journal of Vibroengineering

2013, 15(2).

[117] Yu, Z. W., & Mao, J. F. (2018). A stochastic dynamic model of train-track-bridge

coupled system based on probability density evolution method. Applied

Mathematical Modelling, 59, 205-232.

[118] Zaman, M., Gopalasingam, & A., Laguros, J. (1991). Consolidation settlement of

bridge approach foundation. Journal of Geotechnical Engineering, 117(2), 219-240.

[119] Zhai, W., Sun, X. A detailed model for investigating vertical interaction between

railway vehicle and track. Vehicle system dynamics 1994; 23(S1): 603-615.

[120] Zhai, W., Cai, Z. Dynamic interaction between a lumped mass vehicle and a

discretely supported continuous rail track. Computers & structures1997; 63(5):

987-997.

[121] Zhai, W., He, Z., Song, X. Prediction of high-speed train induced ground vibration

based on train-track-ground system model. Earthquake Engineering and

Engineering Vibration 2010, 9(4): 545-554.

[122] Zhai, W., Xia, H., Cai, C., Gao, M., Li, X., Guo, X., Wang, K. High-speed train–

track–bridge dynamic interactions–Part I: theoretical model and numerical

simulation. International Journal of Rail Transportation 2013; 1(1-2): 3-24.

188
[123] Ziyaeifar, M. Vibration control in train–bridge–track systems. Vehicle System

Dynamics 2008;46(5): 355-384.

[124] 2017 Infrastructure Report Card. Retrieved from: chrome-

extension://oemmndcbldboiebfnladdacbdfmadadm/https://www.infrastructurerepo

rtcard.org/wp-content/uploads/2017/01/Rail-Final.pdf

189
APPENDIX A : FIELD GPR DATA AND LAYER PROFILES

A.1 Field Ground Penetrating Radar (GPR) Data

Figure A.1 (a) GPR Scans, (b) Mid-Chord Offset (MCO) Data, and (c) Space Curve for
Track 2 at Madison Street Bridge Approach.

190
Figure A.2 (a) GPR Scans, (b) Mid-Chord Offset (MCO) Data, and (c) Space Curve for
Track 2 at Caldwell Street Bridge Approach.

191
A.2 Substructure Layer Profiles

Figure A.3 Substructure Layer Profile for Madison Street 12 ft. from the South Abutment
(Track 2).

Figure A.4 Substructure Layer Profile for Madison Street 60 ft. from the South Abutment
(Track 2).

192
Figure A.5 Substructure Layer Profile for Caldwell Street 80 ft. from the South Abutment
(Track 3; West End of Tie).

Figure A.6 Substructure Layer Profile for Caldwell Street 80 ft. from the South Abutment
(Track 3; East End of Tie).

193
APPENDIX B : FIELD MONITORED TRANSIENT DATA

B.1 Madison Street Field Transient Data

Figure B.1 Madison Open Track -Transient Vertical Deformation with Train Passage.

Figure B.2 Madison Open Track -Transient Wheel Load on Crib with Train Passage.

194
 Figure B.3 Madison Open Track -Transient Wheel Load on Tie with Train Passage.

Figure B.4 Madison Near Bridge -Transient Vertical Deformation with Train Passage.

195
Figure B.5 Madison Near Bridge -Transient Wheel Load on Crib with Train Passage.

Figure B.6 Madison Near Bridge -Transient Wheel Load on Tie with Train Passage.

196
 3.5

Peak Transient Deformation (mm)

3
Nearbridge_CI_high

Opentrack_CI_high
2.5
Nearbridge_CI_low

2 Opentrack_CI_low

Nearbridge
1.5 Opentrack

1
4/1/2012 10/18/2012 5/6/2013 11/22/2013 6/10/2014 12/27/2014 7/15/2015
Time

Figure B.7 Madison Street -Peak Vertical Deformation over Time (2012-2015).

Table B.1 Madison Street - Statistics of Peak Vertical Deformation (2012-2015)

Open Track (mm) Near Bridge (mm)

Mean 2.070 2.441

Standard Deviation 0.478 0.427

95% CI High End 2.571 3.210

95% CI Low End 1.568 1.671

197
 100

90 Nearbridge_CI_high
Peak Wheel Load on Tie (kN)

Opentrack_CI_high
80
Nearbridge_CI_low
70 Opentrack_CI_low

Nearbridge
60
Opentrack
50

40
9/14/2011 10/18/2012 11/22/2013 12/27/2014 1/31/2016

Time

Figure B.8 Madison Street-Peak Wheel Load on Tie over Time (2012-2015)

Table B.2 Madison Street - Statistics of Peak Vertical Load Measured on Tie

(2012-2015)

Open Track (kN) Near Bridge (kN)

Mean 81.600 70.073

Standard Deviation 14.710 3.342

95% CI High End 97.038 74.704

95% CI Low End 66.162 65.442

198
 160

Peak Wheel Load on Crib (kN) 150

140

130
Nearbridge
120
Opentrack
110 Nearbridge_CI_high
100 Opentrack_CI_high
90 Nearbridge_CI_low

80 Opentrack_CI_low

70
4/1/2012 10/18/2012 5/6/2013 11/22/2013 6/10/2014 12/27/2014 7/15/2015
Time

Figure B.9 Madison Street -Peak Wheel Load on Crib over Time (2012-2015)

Table B.3 Madison Street - Statistics of Peak Vertical Load Measured on Rail Crib

(2012-2015)

Open Track (kN) Near Bridge (kN)

Mean 119.293 128.54

Standard Deviation 20.378 16.252

95% CI High End 140.679 148.72

95% CI Low End 97.908 108.36

199
B.2 Caldwell Street Field Transient Data

Figure B.10 West of Caldwell Open Track - Transient Vertical Deformation with Train
Passage.

Figure B.11 West of Caldwell Open Track - Transient Wheel Load on Crib with Train
Passage.

200
Figure B.12 West of Caldwell Open Track - Transient Wheel Load on Tie with Train
Passage.

Figure B.13 East of Caldwell Open Track - Transient Vertical Deformation with Train
Passage.

201
Figure B.14 East of Caldwell Open Track - Transient Wheel Load on Crib with Train
Passage.

Figure B.15 East of Caldwell Open Track - Transient Wheel Load on Tie with Train
Passage

202
 Aug-12_1
2.5
Aug-12_2
Peak Transient Deformation (mm)
2.3
East of Open Track
West of Open Track
2.1
East_CI_high
West_CI_high
1.9
Aug-12_1 Aug-12_2 East_CI_low
West_CI_low
1.7

1.5
4/1/2012 10/18/2012 5/6/2013 11/22/2013 6/10/2014 12/27/2014 7/15/2015
Time

Figure B.16 Caldwell Street - Peak Vertical Deformation over Time (2012-2015)

Table B.4 Caldwell Street -Statistics of Peak Vertical Deformation (2012-2015)

West (mm) East (mm)

Mean 1.92 2.2

Standard Deviation 0.16 0.15

95% CI High End 2.08 2.36

95% CI Low End 1.75 2.05

203
 120
Aug-12_1
Peak Wheel Load on Tie (kN) Aug-12_2
110

100
East of Open Track
Aug-12_2 West of Open Track
90
East_CI_high
80 Aug-12_1 West_CI_high
East_CI_low
70 West_CI_low

60
4/1/2012 10/18/2012 5/6/2013 11/22/2013 6/10/2014 12/27/2014 7/15/2015

Time

Figure B.17 Caldwell Street -Peak Wheel Load on Tie over Time (2012-2015)

Table B.5 Caldwell Street - Statistics of Peak Vertical Load Measured on Tie
(2012-2015)

West (kN) East (kN)

Mean 88.84 100.11

Standard Deviation 12.06 8.93

95% CI High End 101.49 109.48

95% CI Low End 76.18 90.74

204
 160
Aug-12_1

Aug-12_2

Peak Wheel Load on Crib (kN)

150

140
East of Open Track
130 West of Open Track
East_CI_high
Aug-12_2
120 West_CI_high
Aug-12_1
East_CI_low
110
West_CI_low

100
4/1/2012 10/18/2012 5/6/2013 11/22/2013 6/10/2014 12/27/2014 7/15/2015
Time

Figure B.18 Caldwell Street -Peak Wheel Load on Crib over Time (2012-2015)

Table B.6 Caldwell Street - Statistics of Peak Vertical Load Measured on Rail Crib
(2012-2015)

West (kN) East (kN)

Mean 127.8 136.52

Standard Deviation 7.81 10.49

95% CI High End 136.0 147.52

95% CI Low End 119.6 125.51

205
B.3 Field Monitored Settlement Data

(a)

(b)

Figure B.19 (a) Layer Settlements at Madison Street 12 ft. from South Abutment (b)
Layer Settlements at Madison Street 60 ft. from South Abutment.

206
 (a)

(b)

Figure B.20 Layer Settlements at Caldwell Street 80 ft. from South Abutment (a) West
End of Tie (b) East End of Tie

207
APPENDIX C : FOURIER TRANSFORM BASED MODEL

C.1 Fourier Transform Sample Calculations

Following the work by Kalker (1996) and Huang et al. (2009), Fourier Transform

technique can be utilized to solve for the equations of motion of the system. The Fourier

Transform was first performed from time to frequency domain and from spatial domain to

wave length domain.

Upland Street bridge approach location is used with its measured transient

deformations as an example to compare and validate the predictions by the track transition

model. Please note that these are the preliminary results of the track transition model and

consider only embankment side.

In this case study, the track model included 40 ties spaced at 609.6 mm on the

embankment side and 15 ties spaced at 609.6 mm on the bridge side. The wheel load was

taken as 140,000 𝑁 for Acela Express locomotive moving at a constant speed of 50 m/s

(110 mph), in accordance with the strain gauge measurements at the Upland Street bridge

approach site. Note that in the field instrumentation, MDDs were installed at approximately

60 ft. away from the abutment of the bridge to correspond to the open track location and

15 ft. away from the abutment of the bridge for the bridge approach location. For further

validation of the developed analytical model with the field data, the load was set to start

from tie No.4 to tie No. 50. The observation point on the embankment side was at tie No.

10, which is 30 ties away from the bridge abutment (approximately 60 ft.). Figure C.1

shows a sketch of the loading and response measurement locations in the case study. Note

that the sketch is not to scale.

208
 Figure C.1 Sketch of Loading and Response Measurement Location.

Figure C.2 shows the vertical transient displacement in the model under one

constant moving load on the embankment side. It can be seen that the moving load occurs

on top of the observation point at around 0.07 sec. The rail oscillates when the moving load

is first applied, followed by an increasing trend as the load is approaching the observation

point, and finally quickly dissipating to almost zero due to the natural frequency of rail

itself. The maximum amount of downward deflection of rail predicted by track transition

model is approximately 1.7 mm. The maximum amount of upward deflection predicted by

track transition model is approximately 0.3 mm. The results obtained by Fourier Transform

matches well with field collected transient deformation data.

The results from Fourier Transform Data are used as a reference case for checking

the effectiveness of train-track-bridge model solved by Newmark’s integration method.

209
 -4 Open Track Side - Upland August 2012
x 10
20

15
Total Vertical Displacement (m)

10

-5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (sec)

Figure C.2 Rail Deflection under One Moving Load at Open Track Location.

210
APPENDIX D : CODE FOR TRAIN-TRACK-BRIDGE MODEL

D.1 MATLAB Code for Train-Track-Bridge Model

This section documents the MATLAB code developed for establishing the train-

track-bridge model and solving via Newmark’s integration method numerically.

close all;clear all;

tic

%%%%%%% This is part of Wenting Hou's PhD dissertation

%%%%%%% Chapter 5 Vehicle-Track-Bridge Model Results

%%%%%%% Version 6.5.2019

%%%%%%% ACELA Express Train

%%%%%%% Consisting of 2 locomotives and 6 passenger cars

%%%%%%% With shear component between ballast particles

%%%%%%% With open bridge deck

%%%%%%% Linear stiffness and damping ratios

%% Input parameters

EI=8.073e6; % Rail bending stiffness Pa.m^4

p=67.46; % Rail mass kg/m

Num_tie_track=500; % Number of ties on open deck side

Num_tie_bridge=500; % Number of ties on bridge side

Num_tie = Num_tie_track + Num_tie_bridge;

Num_mode=500; % Number of shape mode

Num_veh = 8;

211
%% Track structure parameters

cp=ones(Num_tie,1)*1.24e5; % Rail pad damping coefficient N.s/m/m, 6.3e4 as Huang,

1.24e5 as Zhai

kp=ones(Num_tie,1)*7.8e7; % Rail pad stiffness N/m/m, 2.8e8 as Huang, 1.2e8 as Zhai,

1.2e9 as original

mt=ones(Num_tie,1)*386; % Tie mass kg

cb=ones(Num_tie_track,1)*8.2e4; % Ballast damping coefficient N.s/m/m, 8.2e4 as Huang,

5.9e4 as Zhai

kb=ones(Num_tie_track,1)*1.2e8; % Ballast stiffness N/m/m, 7e7 as Huang, 2.4e8 as Zhai

mb=ones(Num_tie_track,1)*683; % Ballast mass kg, 300kg as Huang, 683kg as Zhai

ts=ones(Num_tie_track,1)*7.8e6; % Ballast shear stiffness, 7.8e7 as Zhai

ks=ones(Num_tie_track,1)*5e7; % Subgrade stiffness N/m/m on open deck side, 6.5e7 as

Zhai

cs=ones(Num_tie_track,1)*3e5; % Subgrade damping stiffness on open deck side, 3e4 as

Zhai

k_deck=ones(Num_tie_bridge,1)*5e9; % Stiffness N/m/m on bridge

c_deck=ones(Num_tie_bridge,1)*3e5; % Damping ratio on bridge

tiespace=0.6096; % Tie spacing m (~24 inch)

tie(1:Num_tie)=0; % Location of first tie and preallocate size of tie matrix,-2.54 as

Huang

for tt=2:Num_tie % Location of each tie

tie(tt)=tie(tt-1)+tiespace;

end

L=tie(Num_tie); % Total length;

212
%% Vehicle parameters

Velocity=50; % Train speed constant m/sec (~110mph,49.2mps)

Loco = 130000; % Locomotive weight130000

Car = 90000; % Passenger car weight

l1=2.8448; % Axle spacing, in meter

l2=7.8994;

l3=2.8448;

d1 = 6.5; % Axle spacing between loco and car, in meter

d2 = 5.5; % Axle spacing between car and car, in meter

l1_c=2.9972; % Axle spacing, in meter

l2_c=15.1384;

l3_c=2.9972;

pos = zeros(1,Num_veh*4);

pos(1:4) = [0,l1,l1+l2,l1+l2+l3]; % Wheel Positions for leading locomotive

pos(5:8) = pos(4)+[d1,d1+l1_c,d1+l1_c+l2_c,d1+l1_c+l2_c+l3_c]; % Wheel Positions for car 1

pos(9:12) = pos(8)+[d2,d2+l1_c,d2+l1_c+l2_c,d2+l1_c+l2_c+l3_c]; % Wheel Positions for car 2

pos(13:16) = pos(12)+[d2,d2+l1_c,d2+l1_c+l2_c,d2+l1_c+l2_c+l3_c]; % Wheel Positions for car 3

pos(17:20) = pos(16)+[d2,d2+l1_c,d2+l1_c+l2_c,d2+l1_c+l2_c+l3_c]; % Wheel Positions for car 4

pos(21:24) = pos(20)+[d2,d2+l1_c,d2+l1_c+l2_c,d2+l1_c+l2_c+l3_c]; % Wheel Positions for car 5

pos(25:28) = pos(24)+[d2,d2+l1_c,d2+l1_c+l2_c,d2+l1_c+l2_c+l3_c]; % Wheel Positions for car 6

pos(29:32) = pos(28)+[d1,d1+l1,d1+l1+l2,d1+l1+l2+l3]; % Wheel Positions for trailing

locomotive

T=L/Velocity; % Total running time sec

%% Locomotive

213
Mc = 45000; % Mass of Car body in kg 77000,90000

Mbogie1 = 1200; % Mass of Bogie in kg

Mbogie2 = 1200;

Mw1 = 1200; % Mass of Wheel in kg

Mw2 = 1200;

Mw3 = 1200;

Mw4 = 1200;

g = 9.8; %gravity

Jc = 1.2e6; % kg m^2 1.2e5

Jbogie1 = 760; % 3200

Jbogie2 = 760;

Kc1 = 2.14e6; %all 4 below multiply 2

Kc2 = 5.32e6;

Cc1 = 4.9e4;

Cc2 = 7e4;

a = 0.5*(l1+l2); % Distance between car center and bogie center in m

b = 0.5*l3; % Distance between bogie center to wheel in m

R=300;

G = 4.57*R^(-0.149)*10e-8; % Hertz coefficient

%Kvr = 1/G; % Contact stiffness between wheel and rail

Kvr = 1.5314*10^9; % д╔╨д в╤пн 1.2598*10^9

%% Passenger car

Mc_c = 31550; % Mass of Car body in kg 63100

Mbogie1_c = 1200; % Mass of Bogie in kg

Mbogie2_c = 1200;

Mw1_c = 1200; % Mass of Wheel in kg

Mw2_c = 1200;

214
Mw3_c = 1200;

Mw4_c = 1200;

Jc_c = 1.2e6; % kg m^2 1.2e5

Jbogie1_c = 760; % 3200

Jbogie2_c = 760;

Kc1_c = 2.14e6; %all 4 below multiply 2

Kc2_c = 5.32e6;

Cc1_c = 4.9e4;

Cc2_c = 7e4;

a_c = 0.5*(l1_c+l2_c); % Distance between car center and bogie center in m

b_c = 0.5*l3_c; % Distance between bogie center to wheel in m

%% Observation Point

%tob= Num_tie_track-100; % Open Track Location

tob=floor(Num_tie_track/2);

%tob =; % Near Bridge Location

xob=(tob-1)*tiespace;

rob = (tob-0.5)*tiespace; % between ties rail observation

%% Vehicle equation matrix 10dof

%%%%%% Locomotive

%Mass matrix

M=[Mc 0 0 0 0 0 0 0 0 0;

0 Mbogie1 0 0 0 0 0 0 0 0;

0 0 Mbogie2 0 0 0 0 0 0 0;

0 0 0 Mw1 0 0 0 0 0 0;

0 0 0 0 Mw2 0 0 0 0 0;

0 0 0 0 0 Mw3 0 0 0 0;

215
 0 0 0 0 0 0 Mw4 0 0 0;

0 0 0 0 0 0 0 Jc 0 0;

0 0 0 0 0 0 0 0 Jbogie1 0;

0 0 0 0 0 0 0 0 0 Jbogie2];

%Stiffness matrix

K=[ 2*Kc2 -Kc2 -Kc2 0 0 0 0 0 0 0;

-Kc2 Kc2+2*Kc1 0 -Kc1 -Kc1 0 0 -Kc2*a 0 0;

-Kc2 0 Kc2+2*Kc1 0 0 -Kc1 -Kc1 Kc2*a 0 0;

0 -Kc1 0 Kc1+Kvr 0 0 0 0 -Kc1*b 0;

0 -Kc1 0 0 Kc1+Kvr 0 0 0 Kc1*b 0;

0 0 -Kc1 0 0 Kc1+Kvr 0 0 0 -Kc1*b;

0 0 -Kc1 0 0 0 Kc1+Kvr 0 0 Kc1*b;

0 -Kc2*a Kc2*a 0 0 0 0 2*Kc2*a^2 0 0;

0 0 0 -Kc1*b Kc1*b 0 0 0 2*Kc1*b^2 0;

0 0 0 0 0 -Kc1*b Kc1*b 0 0 2*Kc1*b^2];

%Damping matrix

C=[ 2*Cc2 -Cc2 -Cc2 0 0 0 0 0 0 0;

-Cc2 Cc2+2*Cc1 0 -Cc1 -Cc1 0 0 -Cc2*a 0 0;

-Cc2 0 Cc2+2*Cc1 0 0 -Cc1 -Cc1 Cc2*a 0 0;

0 -Cc1 0 Cc1 0 0 0 0 -Cc1*b 0;

0 -Cc1 0 0 Cc1 0 0 0 Cc1*b 0;

0 0 -Cc1 0 0 Cc1 0 0 0 -Cc1*b;

0 0 -Cc1 0 0 0 Cc1 0 0 Cc1*b;

0 -Cc2*a Cc2*a 0 0 0 0 2*Cc2*a^2 0 0;

216
 0 0 0 -Cc1*b Cc1*b 0 0 0 2*Cc1*b^2 0;

0 0 0 0 0 -Cc1*b Cc1*b 0 0 2*Cc1*b^2];

%%%%%%% Passenger Car

%Mass matrix

M_c=[Mc_c 0 0 0 0 0 0 0 0 0;

0 Mbogie1_c 0 0 0 0 0 0 0 0;

0 0 Mbogie2_c 0 0 0 0 0 0 0;

0 0 0 Mw1_c 0 0 0 0 0 0;

0 0 0 0 Mw2_c 0 0 0 0 0;

0 0 0 0 0 Mw3_c 0 0 0 0;

0 0 0 0 0 0 Mw4_c 0 0 0;

0 0 0 0 0 0 0 Jc_c 0 0;

0 0 0 0 0 0 0 0 Jbogie1_c 0;

0 0 0 0 0 0 0 0 0 Jbogie2_c];

%Stiffness matrix

K_c=[ 2*Kc2_c -Kc2_c -Kc2_c 0 0 0 0 0 0 0;

-Kc2_c Kc2_c+2*Kc1_c 0 -Kc1_c -Kc1_c 0 0 -Kc2_c*a_c 0 0;

-Kc2_c 0 Kc2_c+2*Kc1_c 0 0 -Kc1_c -Kc1_c Kc2_c*a_c 0 0;

0 -Kc1_c 0 Kc1_c+Kvr 0 0 0 0 -Kc1_c*b_c 0;

0 -Kc1_c 0 0 Kc1_c+Kvr 0 0 0 Kc1_c*b_c 0;

0 0 -Kc1_c 0 0 Kc1_c+Kvr 0 0 0 -Kc1_c*b_c;

0 0 -Kc1_c 0 0 0 Kc1_c+Kvr 0 0 Kc1_c*b_c;

0 -Kc2_c*a_c Kc2_c*a_c 0 0 0 0 2*Kc2_c*a_c^2 0 0;

0 0 0 -Kc1_c*b_c Kc1_c*b_c 0 0 0 2*Kc1_c*b_c^2 0;

0 0 0 0 0 -Kc1_c*b_c Kc1_c*b_c 0 0 2*Kc1_c*b_c^2];

217
%Damping matrix

C_c=[ 2*Cc2_c -Cc2_c -Cc2_c 0 0 0 0 0 0 0;

-Cc2_c Cc2_c+2*Cc1_c 0 -Cc1_c -Cc1_c 0 0 -Cc2_c*a_c 0 0;

-Cc2_c 0 Cc2_c+2*Cc1_c 0 0 -Cc1_c -Cc1_c Cc2_c*a_c 0 0;

0 -Cc1_c 0 Cc1_c 0 0 0 0 -Cc1_c*b_c 0;

0 -Cc1_c 0 0 Cc1_c 0 0 0 Cc1_c*b_c 0;

0 0 -Cc1_c 0 0 Cc1_c 0 0 0 -Cc1_c*b_c;

0 0 -Cc1_c 0 0 0 Cc1_c 0 0 Cc1_c*b_c;

0 -Cc2_c*a_c Cc2_c*a_c 0 0 0 0 2*Cc2_c*a_c^2 0 0;

0 0 0 -Cc1_c*b_c Cc1_c*b_c 0 0 0 2*Cc1_c*b_c^2 0;

0 0 0 0 0 -Cc1_c*b_c Cc1_c*b_c 0 0 2*Cc1_c*b_c^2];

%% vehicle matrix

M_v = zeros(Num_veh*10,Num_veh*10);

C_v = zeros(Num_veh*10,Num_veh*10);

K_v = zeros(Num_veh*10,Num_veh*10);

for mm=1:Num_veh

if mm == 1|| mm==Num_veh

M_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=M;

else

M_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=M_c;

end

end

for mm=1:Num_veh

if mm == 1|| mm==Num_veh

218
 C_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=C;

else

C_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=C_c;

end

end

for mm=1:Num_veh

if mm == 1|| mm==Num_veh

K_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=K;

else

K_v((mm-1)*10+1:mm*10,(mm-1)*10+1:mm*10)=K_c;

end

end

%% Rail beam equation matrix

M1=diag(p/2*L*ones(Num_mode,1));

C1=zeros(Num_mode,Num_mode);

K1=zeros(Num_mode,Num_mode);

C2=zeros(Num_mode,Num_tie);

K2=zeros(Num_mode,Num_tie);

for i=1:Num_mode

for k=1:Num_mode

for mm=1:Num_tie

C1(i,k)= C1(i,k)+cp(mm,1)* sin(pi*i*tie(mm)/L)*sin(pi*k*tie(mm)/L);

K1(i,k)= K1(i,k)+kp(mm,1)* sin(pi*i*tie(mm)/L)*sin(pi*k*tie(mm)/L);

end

219
 end

for mm=1:Num_tie_track

C2(i,mm)= -cp(mm)* sin (pi*i*tie(mm)/L);

K2(i,mm)= -kp(mm)* sin (pi*i*tie(mm)/L);

end

K1(i,i) = K1(i,i)+0.5*EI*L*(pi*i/L)^4;

end

%% Tie equation matrix

M2=diag(mt);

C3=transpose(C2);

K3=transpose(K2);

C4=diag(cp+[cb;c_deck]);

K4=diag(kp+[kb;k_deck]);

C5_track=-diag(cb);

K5_track=-diag(kb);

C5_bridge=zeros(Num_tie_bridge,Num_tie_track);

K5_bridge=zeros(Num_tie_bridge,Num_tie_track);

C5 = [C5_track;C5_bridge];

K5 = [K5_track;K5_bridge];

%% Ballast equation matrix

M3=diag(mb);

C6=diag(cb+cs);

C7=transpose(C5);

K7=transpose(K5);

K6main=diag(kb+ks+2*ts);

220
K6up = diag(-ts(2:Num_tie_track),1);

K6down = diag(-ts(1:Num_tie_track-1),-1);

K6=K6main+K6up+K6down;

%% Form Rail-Substructure Matrix

Mr=[M1,zeros(Num_mode,Num_tie+Num_tie_track);...

zeros(Num_tie,Num_mode),M2,zeros(Num_tie,Num_tie_track);...

zeros(Num_tie_track,Num_mode+Num_tie),M3];

Cr=[C1,C2,zeros(Num_mode,Num_tie_track);...

C3,C4,C5;...

zeros(Num_tie_track,Num_mode),C7,C6];

Kr=[K1,K2,zeros(Num_mode,Num_tie_track);...

K3,K4,K5;...

zeros(Num_tie_track,Num_mode),K7,K6];

%% Form Final Matrix

MMM = [M_v,zeros(Num_veh*10,Num_mode+Num_tie+Num_tie_track);...

zeros(Num_mode+Num_tie+Num_tie_track ,Num_veh*10),Mr];

CCC = [C_v,zeros(Num_veh*10,Num_mode+Num_tie+Num_tie_track);...

zeros(Num_mode+Num_tie+Num_tie_track ,Num_veh*10),Cr];

KKK = [K_v,zeros(Num_veh*10,Num_mode+Num_tie+Num_tie_track);...

zeros(Num_mode+Num_tie+Num_tie_track ,Num_veh*10),Kr];

F1 = diag (M_v)* g; %Mass of car bogie wheels

for f = 1:Num_veh

F1((f-1)*10+8 : f*10)=0;

end

221
%% Newmark Beta Integration Method

dt = 0.01; % delta t in integration

Gamma=0.5;

Beta=0.25;

a0=1/(Beta* dt^2);

a1=Gamma/(Beta* dt);

a2=1/(Beta* dt);

a3=1/(2*Beta)-1;

a4=Gamma/Beta-1;

a5=dt *(Gamma/Beta-2)/2;

a6=dt *(1-Gamma);

a7=Gamma* dt;

t = 0; % Initiate values at time = 0

nstep = floor(T/dt);

u_u = zeros(Num_veh*10+ Num_mode+ Num_tie + Num_tie_track,nstep);

v_v = zeros(Num_veh*10+ Num_mode+ Num_tie + Num_tie_track,nstep);

a_a = zeros(Num_veh*10+ Num_mode+ Num_tie + Num_tie_track,nstep);

F_F = zeros(Num_veh*10+ Num_mode+ Num_tie + Num_tie_track,nstep);

F_bar=zeros(Num_veh*10+ Num_mode+ Num_tie + Num_tie_track,nstep);

for n=1:nstep

if n == 1

t = t + dt

% Track structure

for k=1:Num_mode

222
 for v = 1:Num_veh

KKK((v-1)*10+4,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-

1)*4+1))/L); %vehicle-rail interaction

KKK((v-1)*10+5,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+2))/L);

KKK((v-1)*10+6,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+3))/L);

KKK((v-1)*10+7,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+4))/L);

KKK(Num_veh*10+k,(v-1)*10+4)=-Kvr*sin(pi*k*(Velocity*t-pos((v-

1)*4+1))/L); %vehicle-rail interaction

KKK(Num_veh*10+k,(v-1)*10+5)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+2))/L);

KKK(Num_veh*10+k,(v-1)*10+6)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+3))/L);

KKK(Num_veh*10+k,(v-1)*10+7)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+4))/L);

end

end

for II = 1:Num_mode

for JJ = 1:Num_mode

for p = 1:Num_veh*4

KKK(Num_veh*10+II,Num_veh*10+JJ) =...

KKK(Num_veh*10+II,Num_veh*10+JJ)+...

Kvr*(sin(pi*II*(Velocity*t-pos(p))/L)*sin(pi*JJ*(Velocity*t-pos(p))/L));

end

end

end

K_bar=KKK; % Effective stiffness matrix - rail

F_F (:,n+1)= [F1;zeros(Num_mode+Num_tie+ Num_tie_track,1)];

F_bar(:,n+1)=F_F(:,n+1);

223
 u_u(:,n+1)=K_bar\F_bar(:,n+1);

else

t = t + dt

KKK = [K_v,zeros(Num_veh*10,Num_mode+Num_tie+Num_tie_track);...

zeros(Num_mode+Num_tie+Num_tie_track ,Num_veh*10),Kr];

% Track structure

for k=1:Num_mode

for v = 1:Num_veh

KKK((v-1)*10+4,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-

1)*4+1))/L); %vehicle-rail interaction

KKK((v-1)*10+5,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+2))/L);

KKK((v-1)*10+6,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+3))/L);

KKK((v-1)*10+7,Num_veh*10+k)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+4))/L);

KKK(Num_veh*10+k,(v-1)*10+4)=-Kvr*sin(pi*k*(Velocity*t-pos((v-

1)*4+1))/L); %vehicle-rail interaction

KKK(Num_veh*10+k,(v-1)*10+5)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+2))/L);

KKK(Num_veh*10+k,(v-1)*10+6)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+3))/L);

KKK(Num_veh*10+k,(v-1)*10+7)=-Kvr*sin(pi*k*(Velocity*t-pos((v-1)*4+4))/L);

end

end

for II = 1:Num_mode

for JJ = 1:Num_mode

for p = 1:Num_veh*4

KKK(Num_veh*10+II,Num_veh*10+JJ) =...

224
 KKK(Num_veh*10+II,Num_veh*10+JJ)+...

Kvr*(sin(pi*II*(Velocity*t-pos(p))/L)*sin(pi*JJ*(Velocity*t-pos(p))/L));

end

end

end

K_bar=KKK+a0*MMM+a1*CCC; % Effective stiffness matrix - rail

F_F (:,n+1)= [F1;zeros(Num_mode+Num_tie+ Num_tie_track,1)];

F_bar(:,n+1)=F_F(:,n+1)+MMM*(a0*u_u(:,n)+a2*v_v(:,n)+a3*a_a(:,n))+... % Effective force

vector

CCC*(a1*u_u(:,n)+a4*v_v(:,n)+a5*a_a(:,n));

u_u(:,n+1)=K_bar\F_bar(:,n+1);

a_a(:,n+1)=a0*(u_u(:,n+1)-u_u(:,n))-a2*v_v(:,n)-a3*a_a(:,n);

v_v(:,n+1)=v_v(:,n)+a6*a_a(:,n)+a7*a_a(:,n+1);

end

end

%% Calculate Rail Displacement from Modal Displacement

%u_u = [vehicle,q,ut,ub]

tspan = (0:dt:T);

Ur = zeros (1,length(tspan));

225
Urdot = zeros (1,length(tspan));

Urdotdot = zeros (1,length(tspan));

Urob = zeros (1,length(tspan));

for k = 1:Num_mode

Ur = (Ur + sin(pi*k*xob/L)*u_u(Num_veh*10+k,:));

Urob= (Urob + sin(pi*k*rob/L)*u_u(Num_veh*10+k,:));

Urdot = (Urdot + sin(pi*k*xob/L)*v_v(Num_veh*10+k,:));

Urdotdot = (Urdotdot + sin(pi*k*xob/L)*a_a(Num_veh*10+k,:));

end

Ut = u_u (Num_veh*10+Num_mode+tob,:); % tie displacement in m

Utdot = v_v (Num_veh*10+Num_mode+tob,:); % tie velocity in m/s

Utdotdot = a_a (Num_veh*10+Num_mode+tob,:); % tie acceleration in m/s/s

Ub = u_u (Num_veh*10+Num_mode+Num_tie+tob,:); % ballast displacement in m

Ubdot = v_v (Num_veh*10+Num_mode+Num_tie+tob,:); % ballast velocity in m/s

Ubdotdot = a_a (Num_veh*10+Num_mode+Num_tie+tob,:); % ballast acceleration in m/s/s

Ucar = u_u (1,:); % car body displacmeent in m

Ucardot = v_v(1,:);

Ucardotdot = a_a (1,:);

Uw1 = u_u (4,:); % Wheelset 1 displacement in m

Uw2 = u_u (5,:);

Uw3 = u_u (6,:);

Uw4 = u_u (7,:);

Uw5 = u_u (14,:);

Ubogie1 = u_u (2,:); % Bogie 1 displacement in m

phi_bogie1 = u_u (9,:); % Bogie 1 rotation

226
Uw1dot = v_v (4,:); % Wheelset 1 velocity in m/s

Ubogie1dot = v_v (2,:); % Bogie 1 velocity in m/s

phi_bogie1dot = v_v (9,:); % Bogie 1 rotation velocity

Uw1dotdot = a_a (4,:); % Wheelset 1 acceleation in m/s/s

% Pwheel =

%% Calculate Force

WRF1= Kvr*(Urob-Uw1); % Wheel-Rail Reaction Force

WRF= WRF1(floor((pos(1)+rob)/Velocity/dt)+1);

RTF = cp(tob) * (Urdot-Utdot)+ kp(tob)* (Ur-Ut); % Rail-Tie Reaction Force in Newton

if tob <= Num_tie_track

TBF = cb(tob) * (Utdot-Ubdot)+ kb(tob)* (Ut-Ub); % Tie-Ballast Reaction Force in Newton

end

%% plot the data

%figure (1) % displacement data

%% rail displacement

% plot(tspan,Ur*10^3,'b') % time vs. rail deflection (mm)

% xlabel('Time (sec)','FontSize',12,'FontName','Arial');

% ylabel('Vertical Displacement (mm)','FontSize',12,...

% 'FontName','Arial');

% %xlim([0,5]);

227
% %ylim([-0.2,1.8]);

% %title('Vertical Displacement')

% hold on

% %tie displacement

% plot(tspan,Ut*10^3,'r')

% hold on

%%

% % ballast displacement

% plot(tspan,Ub*10^3,'k')

% legend('Top of Rail','Top of Tie', 'Ballast')

% set(legend,...

% 'FontSize',12,...

% 'FontName','Arial');

% % xlim([0,5])

% % ylim([-0.2,1.8])

%maxd=max(abs(Ur*10^3));

%% Rail Tie Reaction Force

% figure (2)

% plot(tspan,RTF*10^(-3),'b') % time vs. rail tie reaction force (kN)

% xlabel('Time (sec)','FontSize',12,'FontName','Arial');

% ylabel('Rail-Tie Reaction Force (kN)','FontSize',12,...

% 'FontName','Arial');

% %title('Rail-Tie Reaction Force')

% % xlim([0,5]);

% % ylim([-10,50])

% %maxf=max(abs(RTF))

228