FINITE ELEMENT ANALYSIS OF RAILWAY TRACK UNDER VEHICLE DYNAMIC

IMPACT AND LONGITUDINAL LOADS

BY

ZIJIAN ZHANG

THESIS

Submitted in partial fulfillment of the requirements

for the degree of Master of Science in Civil Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2015

Urbana, Illinois

Advisor:

Professor Bassem O. Andrawes

 ABSTRACT

Impact loads caused by flat spots on railcar wheels impose a major maintenance burden

on railroads and can cause severe damage to both railcar and rail track components. In addition,

the increasing tractive power of locomotives leads to significant increase in the longitudinal load

demand in railway tracks. The capacity of longitudinal restraint of existing rail fastening system

and its dependency on track parameters affects the future design of rail fastening system. This

thesis focuses on investigating numerically these two problems using finite element (FE) method.

An FE model with multiple crossties and their accompanying fastening systems is developed and

utilized in this study after being validated using field data. The results of the impact load study

indicate that impact loading consists of direct wheel impact loads and track system vibration

induced impact loads. Both of these impact mechanisms are sensitive to the parameters

considered in this study, including the stiffness of rail pad and the speed of train. It is observed

that rail pad with a moderate stiffness provides the most effective impact attenuation.

Furthermore, impact load is found to increase with increasing train speed.

On the other hand, the study conducted on longitudinal track loads focused on

investigating the effects of wheel acceleration, elastic modulus of clips, rail-to-railpad coefficient

of friction (COF), and crosstie spacing on the distribution of longitudinal force in the rail

fastening system. The FE model results suggest that a coefficient of friction (COF) of 0.65 is

most efficient in maximizing the longitudinal restraint per railseat. Additionally, a crosstie

spacing of 24 inches can lead to a desired distribution of longitudinal force.

ii
 to father and mother,

for always supporting me

致我的父母，

感谢你们无私的支持

iii
 ACKNOWLEDGEMENTS

My gratitude cannot be contained in the pages of this thesis. First, I would like to thank

the United States Department of Transportation (US DOT) Federal Railroad Administration

(FRA) and the National University Rail (NuRail) for funding my position as a graduate research

assistant and this research project.

My most sincere gratitude goes to Professor Bassem O. Andrawes. This research is only

possible under his foreseeing and valuable guidance. There were bumps along the way, but

Professor Andrawes has always been supportive and showed his faith in me. A vast portion of

the success of this research is attributed to the constant help from Professor Andrawes.

Also, it was a great honor to part of the RailTEC research team. John Riley Edwards has

been a great leader of the team. He has facilitated my knowledge on railroad engineering. The

good vibe for research within the RailTEC team is not possible without him. I would also like to

thank Marcus Dersch for always answering my questions and for providing helpful information

whenever I encounter difficulties on my research.

I am also thankful to George Zhe Chen. He provided invaluable help when I first started

learning ABAQUS and building my dynamic model. I would not be able to build a good model

without the guidance from him. My friendship with Sihang Wei is another windfall to me. I will

never forget the days we spent discussing the research work on impact load and all those times

we spent talking about trivia.

I would also like to thank everyone I worked with in the RailTEC group. It has been a

great time hanging out with all of you guys.

iv
 Last but not least, I would like to thank my family. My mom and dad. You have been

giving my unconditional support and love. I would have never come so far with my academic

career without your supports.

v
 TABLE OF CONTENTS

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

1.1. Motivation..................................................................................................................................... 1
1.2. Thesis Scope ................................................................................................................................. 3
1.3. Thesis Outline .............................................................................................................................. 3
CHAPTER 2 BACKGROUND AND LITERATURE REVIEW .......................................................... 5
2.1. Concrete Crosstie Fastening Systems .......................................................................................... 5
2.2. Background on Impact Load ....................................................................................................... 9
2.3. Track Buckling and Track Geometry Irregularity.................................................................... 12
2.4. Longitudinal Loads in Railway Tracks ..................................................................................... 13
CHAPTER 3 FINITE ELEMENT ANALYSIS OF TRACK RESPONSE UNDER WHEEL FLAT
INDUCED IMPACT LOADS .................................................................................................................. 18
3.1. Finite Element Software ABAQUS ........................................................................................... 18
3.2. FE Model Overview ................................................................................................................... 19
3.3. Mesh Sensitivity Analysis........................................................................................................... 22
3.4. Element Type and Mesh Size ..................................................................................................... 25
3.5. Constitutive Relationships ......................................................................................................... 25
3.6. Contact Interactions ................................................................................................................... 28
3.7. Loading Procedure and Boundary Conditions ......................................................................... 30
CHAPTER 4 FINITE ELEMENT MODEL VALIDATION FOR IMPACT LOAD MODEL ....... 33
4.1. Field Test Setup .......................................................................................................................... 33
4.2. Field Validation of FE Model with No Impact Load ................................................................ 35
4.3. Field Validation of FE Model with Impact Load...................................................................... 37
4.3.1. Impact Force ....................................................................................................................... 37
4.3.2. Vertical Strain in the Rail ................................................................................................... 42
CHAPTER 5 FINITE ELEMENT ANALYSIS OF TRACK RESPONSE UNDER
LONGITUDINAL LOADS ...................................................................................................................... 45
5.1. FE Model Overview ................................................................................................................... 45
5.2. Loading Procedure and Boundary Conditions ......................................................................... 48
CHAPTER 6 FINITE ELEMENT MODEL VALIDATION OF LONGITUDINAL LOAD
MODEL ..................................................................................................................................................... 49
6.1. Field Test Setup .......................................................................................................................... 49
6.2. Testing Train Consists ............................................................................................................... 50

vi
 6.3. Field Validation of FE Model.................................................................................................... 51
6.3.1. Longitudinal Displacement of the Railpad ......................................................................... 52
6.3.2. Longitudinal Strain in the Rail............................................................................................ 53
6.3.3. Vertical Strain in the Rail ................................................................................................... 54
CHAPTER 7 PARAMETRIC STUDIES .............................................................................................. 56
7.1. Parametric Study on Impact Load............................................................................................. 56
7.1.1. Railpad Stiffness Results ..................................................................................................... 56
7.1.2. Speed Results ...................................................................................................................... 59
7.2. Parametric Study on Longitudinal Load ................................................................................... 60
7.2.1. Wheel Acceleration Results ................................................................................................. 61
7.2.2. Elastic Modulus of Clips Results ........................................................................................ 64
7.2.3. Coefficient of Friction between Rail and Railpad Results .................................................. 66
7.2.4. Crosstie Spacing Results ..................................................................................................... 68
CHAPTER 8 SUMMARY AND CONCLUSIONS ............................................................................... 71
8.1. Summary..................................................................................................................................... 71
8.2. Conclusions from Impact Load Analysis .................................................................................. 72
8.3. Conclusions from Longitudinal Load Analysis ........................................................................ 73
8.4. Recommendations ...................................................................................................................... 74
8.4.1. Improved Component Design for Optimal Impact Attenuation .......................................... 74
8.4.2. Improved Capacity of Longitudinal Restraint of Rail Fasteners ........................................ 75
8.5. Future Work ............................................................................................................................... 76
REFERENCES .......................................................................................................................................... 77

vii
 CHAPTER 1

INTRODUCTION

1.1. Motivation
The advancement of technology in the manufacture of trains, mainly both higher

operating speed and heavier axle loads is essential to satisfy the growing demand for passenger

transit and freight. This rapid development in modern rail vehicle greatly improves the power

and speed of trains, yet it imposes increasing performance demand on the railway track system.

The heavy freight and high speed passenger trains can cause much higher traffic loads that

exceed the design strength and capacity of current railroads. For example, the deployment of

heavier and faster trains will result in the increase in impact loads and longitudinal forces, which

can potentially compromise the structural reliability of track components. In other words, the

dynamic vehicle loads can be much amplified. Vertical dynamic loads in the railway track can be

attributed to both the vibration interaction between the wheel and rail (Kaewunruen &

Remennikov, 2010) and the irregularities of the wheel and rail (Kumar & Sambasivarao, 2014).

A more severe vertical dynamic load is typically associated with the presence of wheel and rail

irregularities such as wheel flats and insulating rail joints (Nielsen, 2008); the latter is typically

known as the impact load. Previous records in North America have shown that the impact loads

can exceed the static load by 2.5 to 3.5 times, and therefore pose a critical concern regarding the

reliability of track components (Zarembski & Bell, 2002). Impact loads are hazardous to the

track structure as they may cause cracking in the crossties and accelerate the degradation of track

geometry, and thus lead to large maintenance expense (Zarembski & Bell, 2002).

1
 In addition to the impact load, the increasing longitudinal load due to more powerful

trains is another concern for the track structure. With the increasing use of continuously welded

rail (CWR), the longitudinal load is even of more importance. The CWR, unlike conventional

joint rail, allows for the rail segments to form an uninterrupted section. The absence of joints

offers several advantages including increased passenger comfort, reduction in audible noise

caused by train passage, reduced rate of deterioration of track, and reduced dynamic effects

associated with the existence of joints (Carvalho et al., 2013). Nevertheless, the effects of

longitudinal load can be much exacerbated as the allowance for the longitudinal movement of

rail by the rail joints is deprived in the CWR.

The longitudinal load in CWR is commonly induced due to the uniform temperature

change of the rail, tractive effort of the train, and sudden change of stiffness of the supporting

structure (Ruge & Birk, 2007). The latter two, cause localized longitudinal stress along a finite

length of the rail whereas the uniform temperature change induces longitudinal stress along a

much longer section. While all three causes can lead to buckling of rail, longitudinal load due to

the vehicle tractive effort possesses a higher degree of uncertainty as it is unique to every train

consist. Furthermore, it is likely to undergo significant increase as the rail vehicle technology

advances. The resistance to longitudinal load and movement of rail is provided by rail fastening

system which anchors the rail to the crossties (AREMA, 2014). Therefore, the capacity of rail

fastening system for longitudinal resistance is an important reference for the longitudinal

stability of track structure.

In addition to the continuous increase in load demands, the last few decades have

witnessed a shift away from the use of timber crossties to the use of concrete crossties in new

construction of railroads. This shift is primarily due to the scarcity of timber and its increasing

2
price (National Academy of Sciences (U.S.), 1976). However, the current design specifications

practiced in the United States for railway tracks with concrete crossties is derived from empirical

design process primarily developed for timber tie track system (Van Dyk, 2013). As a result, the

service life of track components is likely reduced, leading to more frequent maintenance and thus

higher costs. Therefore, the response of concrete-crosstie track system to complex loading

environments needs to be investigated.

1.2. Thesis Scope

This study investigates numerically the behavior of impact load caused by wheel flat in

concrete crosstie tracks. The study will shed the light on the relationship between the impact

force and track design parameters. In addition, the study examines the distribution and

magnitude of longitudinal force in the concrete crosstie fastening system and their relationships

with track design parameters. The study is carried out numerically using finite element (FE)

analysis. Two detailed 3D FE models were developed, validated using field data, and utilized in

the study.

1.3. Thesis Outline

Chapter 2 provides background on railway track system, in particular concrete crosstie

track which is the focus of this study. It also introduces two major hazards in a concrete crosstie

track, that is, large longitudinal and impact loads. In addition, it also contains information from

literatures on the cause of longitudinal and impact loads and their associated damaging

mechanisms.

Chapter 3 elaborates on the development of the finite element (FE) model for analyzing

track response to dynamical impact load. It discusses the modeled track components and their

3
material properties, contact interaction properties defined for track components, and the loading

procedure used for the FE analysis.

Chapter 4 introduces the validation of the FE model using field data. It introduces the

setup and instrumentation of the field test and the parameters used to validate the FE model.

Chapter 5 introduces the second FE model that was developed to analyze the track

response due to wheel induced longitudinal load.

Chapter 6 presents the results of the validation of the second FE model using field

collected data.

Chapter 7 presents parametric study results from both FE models and discusses the

observations associated with each parameter.

Chapter 8 includes the summary and main conclusions of the study as well as brief

recommendations for future work.

4
 CHAPTER 2

BACKGROUND AND LITERATURE REVIEW

2.1. Concrete Crosstie Fastening Systems

A railway track is a sophisticated system comprising multiple components, and the

strength of each track component is liable for the structural integrity of the whole track system.

The connection between the rail and concrete crossties is provided by the fastening system which

consists of multiple track components (Esveld, 2001). The functions of a concrete crosstie

fastening system include providing and maintaining various degrees of gauge restraint, transfer

of vertical, lateral, and longitudinal loads from rail to crosstie, load or impact attenuation, and

electrical isolation (AREMA, 2014). There are two major types of fastening systems for concrete

crosstie tracks: screw systems and clip systems (Kerr, 2003). Compared to clip systems, screw

systems possess certain disadvantages as they require more maintenance and the clamping force

produced can vary among railseats (Kerr, 2003). In North America, four types of clip fastening

systems are commonly used with concrete crossties. As illustrated in Figure 2.1, these include

the e-clip, the Safelok I and III, and the Fastclip (Kerr, 2003).

5
 Fastclip
E-Clip

Safelok I Safelok III

Figure 2.1. Concrete Crosstie Fastening Systems (adapted from CXT Concrete Crossties)

The fastening system considered in this study is the Safelok I system. It includes a pair of

spring clips, insulators and shoulders and a rail pad assembly. The spring clips exert clamping

forces on the rail which control longitudinal rail movement due to thermal and tractive forces

and minimize rail gap in the event of a rail break (AREMA, 2014). The insulators serve to

prevent interference with signal system and deterioration of the fastening system through

electrical leakage (AREMA, 2014). The shoulders is a cast-in component that provides

anchorage points within crossties for rail fastening systems and other miscellaneous components

(AREMA, 2014). The rail pad assembly consists of a rail pad and an abrasion plate and is

6
installed between the rail and concrete crossties to reduce impact and vibration effects on the

track structure and minimize railseat deterioration (AREMA, 2014).

Figure 2.2 illustrates the fastening components of the Safelok I fastening system. The

shoulder is designed to hold the clip to the tie and is cast-in during the manufacturing process of

crosstie (Romero et al., 2010). The pair of legs (stems) at the bottom of the shoulder bond with

the concrete crosstie, producing resistance to pull-out and ensuring stability of the connected

fastening components (do Carmo, 2014). The spring clip is driven into the shoulder and exerts a

clamping force by its elastic deformation. The clamping force per railseat produced by spring

clips can range from 4,400 lbs to 5,800 lbs for the common types of concrete crosstie fastening

system in North America (Romero et al., 2010). The Safelok I fastening system can provide

4,800 lbs clamping force (Romero et al., 2010). The insulators on the field and gauge sides have

different geometries. They are commonly made from nylon 6/6 which possesses high mechanical

strength and good abrasion and high temperature resistance (do Carmo, 2014). The railpad is also

made from polyurethane. As polyurethane is a soft material, it is expected to provide good

attenuation for contact interaction between the rail and crosstie caused by dynamic impact. The

abrasion plate, on the other hand, protects the railseat against mechanical wear and prevents

sawing action from the longitudinal rail movements (Hay, 1982).

7
 Spring Clip Cast-in Shoulder

Field-side Insulator Gauge-side Insulator

Rail Pad Abrasion Plate

Figure 2.2. Safelok I Fastening System Components

8
2.2. Background on Impact Load
One of the hazards of impact loads to track system is the damage to track components,

predominantly being the cracking of crossties. Kaewunruen and Remennikov (2010) suggested

that the train-track interaction and its resulting resonance in track components were the primary

causes of large impact forces. Compared to the dynamic force due to general dynamic wheel-rail

interactions, which has low magnitude but high cycle, the impact force has much higher

magnitude but lower cycle and is more damaging to prestressed concrete crossties (Ye et al.,

1994; Wang, 1996; Wakui & Okuda, 1999; Gustavson, 2002; Stevens & Dux, 2004). Even

though the high magnitude impact load has been recognized as damaging to concrete crossties,

the failure of a concrete crosstie is more likely to occur due to cumulative damage rather than

due to a single occurrence of impact load (Kaewunruen, 2007; Kaewunruen & Remennikov,

2009). Therefore, it is crucial to assess the threshold of the value of impact load that can result in

cumulative damage of concrete crosstie.

In addition to impact load in perfectly supported tracks, uneven support conditions for

concrete crossties in the field is not a rare occurrence especially after the reworking of ballast.

And the presence of voids under concrete crossties is likely to ease the excitation of vibration

modes of concrete crossties. In such events, impact loads are believed to be excessively large.

One of the primary causes of dynamic impact loads in a railway track is the rail wheel

flat spots. The formation of a wheel flat is generally attributed to the braking of a train. Braking

slows the rotation of wheels and can cause the wheels to slide along the rails so that the train is

decelerated by the friction between the wheels and rails. However, the sliding can wear off a

portion of the wheel treads and result in a flat spot (Dukkipati & Dong, 1999). The size and

shape of wheel flats, as well as wheel load and train speed are the factors that affect the impact

load; and certain combinations of these factors can result in large impact loads that may cause

9
serious damage to track structures (Bian et al., 2013). In addition, as the magnitude of impact

loads can be much higher than the cyclic loads from the passage of wheels, and the track

components are prone to accelerated degradation under the effects of impact loads.

The determination of impact factor is typically highly simplified in practice. The

American Railway Engineering and Maintenance-of-Way Association (AREMA) assumes an

impact factor of 2.0 for the flexural design of concrete crossties. However, data acquired by

wheel impact load detectors (WILD) on Amtrak at Edgewood, Maryland and elsewhere on

freight railroads indicated that impact factors could easily exceed the 2.0 design value (Van Dyk

et al., 2014). There were also instances of loads exceeding the 2.0 design value observed from

passenger car wheels. These data suggest that the design impact factor be revisited to

accommodate for wheel loads that exceed the current value of 2.0.

In addition, many studies in the past have presented different variations of formulas used

to determine impact factors (Doyle, 1980; Van Dyk et al., 2014). While most formulas only

relate the impact factor to train speed; other more comprehensive formulas also incorporated

vehicle-related parameters, including wheel diameter, vehicle unsprung mass, etc.; and track-

related parameters, including track modulus, cant radius, etc. (Van Dyk et al., 2014). However,

most of the equations were empirically derived and independent of the types of rail and wheel

defects, thus they shall not be used to predict impact loads caused by wheel flat.

In a study by Dukkupati and Dong (1999), an FE model was employed to investigate the

characteristics of impact loads due to wheel flats and other defects. The railway track was

modeled as a Timoshenko beam on discrete supports and the wheel flat was approximated by a

haversine defect and grounded into the rail as an equivalent of a flat spot on the wheel. The use

of a haversine defect on the rail might not represent the wheel defect at high speed as the

10
researchers realized that the wheel flew over the rail defect and resulted in small impact loads at

high speeds. Also, as the FE model was in two-dimensions (2D) and linear constitutive relation

was assumed for all components, the results may not capture the realistic nonlinear behaviors of

the railway track system. Therefore, a three-dimensional (3D) FE model is needed to better

investigate the characteristics of impact loads caused by wheel flats.

Bian et al. (2013) developed a 3D FE model which simulated the rolling of a railcar

wheel on a three-crosstie railway track system. The FE model was used to simulate various

magnitudes of impact loads at different speeds and at different static wheel loads. A symmetric

boundary condition was enforced at the track centerline as only half of the track was modeled.

Therefore, the FE model was only representative of symmetrical impact loading on both rails.

However, as impact loading is typically asymmetrical, that is, only occurs on one rail, this FE

model is incompetent in simulating realistic railway track behavior under impact loading. In

addition, this FE model only included three railseats, which was fewer than the five railseats

suggested by AREMA (2014) as the longitudinal distribution length for vertical wheel load.

An improved 3D model of a wheel-track system is presented in order to study the effects

of impact loads due to wheel flats. It highlights a full railway track consisting of two rails, which

allows the wheel flat to be presented on only one rail. In addition, a longer length of track was

modeled, so that the distribution of railseat load can be correctly simulated. Field experimental

data was used to validate the FE model. The FE modeling results are used for the investigation of

the behaviors of impact loading with different system parameters which include the rail pad

stiffness and train speed.

11
2.3. Track Buckling and Track Geometry Irregularity
Track longitudinal loads cause significant concerns related to the lateral stability of the

rail. In other words, excessive longitudinal loads could cause lateral buckling of the track. Track

buckling is the formation of large lateral misalignments in CWR; it often results in catastrophic

derailments. Buckling of the track is typically due the combined effect of three major factors:

high compressive forces, weakened track conditions, and vehicle loads (Volpe, 2014).

High compressive forces are a result of a temperature raise above the rail neutral

temperature of rail. The rail neutral temperature refers to the state wherein the rail is stress free in

the longitudinal direction. It is typically the temperature at which the track is constructed.

Therefore, if the ambient temperature rises above the neutral temperature, the rail is subjected to

thermal expansion which then induces compressive stress. The problem associated with high

thermal load is often addressed by installing the rail at a controlled temperature of 95-110 °F. As

the temperature in the rail can reach 130-150 °F in some field conditions, the high neutral

temperature prevents the rail from experiencing excessively high buckling forces (Volpe, 2014).

Weakened track conditions refer to reduced track lateral resistance as well as permanent

track misalignment. Track lateral resistance is the reaction offered by the ballast against lateral

movement. The track lateral resistance can be reduced by 40-70% due to surfacing and tamping

of ballast, causing the rail to become buckling prone (Kish, 2011). In addition to the impaired

track lateral resistance, track lateral misalignment is another cause for track buckling. As

illustrated in Figure 2.3, for a perfectly aligned tangent track, a snap-through buckling

mechanism prevails with increasing compressive stress in the rail. In other words, it is essentially

the introduction of a small lateral deflection that causes the rail to snap-through from the

perfectly aligned and unbuckled state to the buckling prone state. Once the misalignment is

present in the track, its lateral resistance is significantly reduced (Figure 2.3).

12
Figure 2.3. Peak and Limit Resistance Impacts on Buckling Temperature (adapted from Kish,

2011)

In tangent tracks, vehicle loads are often resulted from slack action, braking of trains, and

rolling friction of trains. Slack action refers to the amount of free movement of one railcar before

its motion is transmitted to a coupled car (UMKC School of Law, 1945). It is a source of track

buckling as the tractive effort produced by a locomotive is at its greatest if a slack is present

between the locomotive and trailing cars. The braking and rolling friction of trains cause track

buckling in a similar way as they both exert localized longitudinal force in the rail. In curved

tracks, buckling can also be triggered as large lateral force can be exerted on the rail (Williams et

al., 2014).

2.4. Longitudinal Loads in Railway Tracks

Longitudinal forces in a railway track are produced by the tractive effort and dynamic

braking of a train (Srinivas et al., 2011). As railway traffic demands exhibit a rapid increase over

the past few decades in terms of their tonnages and speed of trains, more powerful locomotives

have been built to meet the surging demands; the latest model of diesel-electric locomotive is

13
capable of producing twice as much tractive force as its predecessors (Foutch et al., 2006). As a

result, the longitudinal forces imparted from the tractive effort are increased significantly. The

longitudinal forces transfer from the rail to the crossties through the fastening system. In other

words, the fastening system serves as the connection between the rail and the crossties, thus

anchors the rail against its longitudinal movement due to forces in the longitudinal direction

(Srinivas et al., 2011). Therefore, the increase in the wheel-induced longitudinal forces can

impose a higher load demand in the fastening system.

The current design standard practiced in the United States by AREMA provides a single-

tie and single-rail pullout test as the reference for the design and manufacture of rail fasteners

(Figure 2.4). It states a threshold of 0.20 inch for the longitudinal displacement of the rail as an

increasing longitudinal force, up to 2.4 kips, is applied to the rail and held for 15 minutes; and an

additional longitudinal displacement of 0.01 inch cannot be exceeded after the force is held for

three minutes. A similar test is described in the European standard (EN 2002), except for that the

longitudinal load is increased until slippage occurs in the rail. As is conducted on a single railseat,

the test specified in AREMA (2014) only accounts for a longitudinal force that is not greater than

2.4 kips per railseat. The value of 2.4 kips, according to AREMA (2014), is sufficient

considering general service based on experience, but there are locations where excessive

longitudinal forces are present. In addition, the test only applies to fasteners on crossties with 24-

inch spacing. Therefore, the longitudinal load imparted in the railseat under various service

conditions needs to be investigated.

14
Figure 2.4. Fastener Longitudinal Restraint Test (AREMA, 2014)

In a field experimentation conducted by Srinivas et al. (2011), longitudinal force in a

bridge track was investigated. The field experimentation concerns the scenarios that include the

accelerating and braking of a train as well as the train running at uniform speeds. Based on the

results of the experimentation, among the three testing scenarios, the highest longitudinal force is

imparted in the rail under the passage of an accelerating train. Furthermore, a much more

significant longitudinal force is induced by the locomotive wheel compared to the trailing car

wheels (Srinivas et al., 2011). As illustrated in Figure 2.5, a locomotive wheel is driven by the

15
rotation of the axle while a trailing car wheel rolls due to the friction force between the wheel

and the rail. For an accelerating train, part of the tractive forces is distributed as the friction

forces between the trailing car wheels and the rail. However, as the number of the locomotive

wheels is typically less than that of the trailing car wheels, larger longitudinal forces are

produced by locomotive wheels. Therefore, an accelerating locomotive wheel is of most

importance to be investigated concerning the longitudinal force imparted in the fastening system.

Figure 2.5. Difference in Driving Mechanisms between a Trailing Car and a Locomotive Wheel

In addition to the field experimentation, the mechanical behaviors of a railway track

system can also be studied using FE analysis. In a study presented by Nguyen et al. (2011), a 2D

and a 3D FE model were both developed to investigate the response of a railway track under

high speed vertical dynamic loading. The 2D FE model utilizes Timoshenko beam elements and

spring-dampers and allows for significant savings on the computational cost. However, the 2D

model lacks the ability to simulate the behaviors of the fastening components for which the 3D

FE model is able to compensate. In another study by Chen et al. (2014), a 3D FE model of a

track system with detailed fastening components is developed to study the response of the

fastening system under combined static vertical and lateral wheel loads. It allows for the

investigation of the behaviors of every fastening component. However, the model only considers

16
static loading scenarios which can compromise the accuracy of the results considering the

dynamic effects. As few FE model has been developed for investigating the behaviors of

longitudinal force in the railway track system, a 3D dynamic FE model is needed in order to gain

insights into this subject.

In addition to the FE model discussed in the previous section, another 3D FE model of a

wheel-track system is presented; it focuses on investigating the track response under wheel

dynamic longitudinal load. It highlights an 86-foot railway track consisting of detailed 3D

fastening components. The extensive length of the track system ensures that little longitudinal

force is present in the fastening system at the boundary locations. Field experimental data was

also used to validate the FE model. The FE modeling results are then used for the investigation

of the behaviors of longitudinal force in the fastening system with different system parameters

which include the accelerating rate of the wheel, the elastic modulus of the clips, the COF

between the rail and the railpads, and the spacing of the crossties.

17
 CHAPTER 3

FINITE ELEMENT ANALYSIS OF TRACK RESPONSE UNDER WHEEL

FLAT INDUCED IMPACT LOADS

3.1. Finite Element Software ABAQUS

The commercial finite element software, ABAQUS was utilized in this study. ABAQUS

is capable of conducting both static and dynamic simulation and has significant applications

typically in a wide range of industries. The software provides both pre-processing and post-

processing capabilities. The pre-processing user interface offered the visualization of component

and assembly geometry and provided a variety of mesh generation routines (Dassault Systemes

Simulia Corp., 2013). As parallelization is allowed, the computational efficiency can be greatly

enhanced by taking the advantage of multi-core processors. The software also offers the

capability to visualize analysis results, which significantly facilitates the process of post-

processing.

In addition to the powerful pre- and post-processing user interfaces, ABAQUS provides

capabilities of analysis on a wide range of engineering problems. It is able to conduct basic and

advanced linear analysis as well as nonlinear analysis such as the response of structure due to

earthquake loading (Dassault Systemes Simulia Corp., 2013). Furthermore, its capabilities are

extended to solving multiphysics problems including structural acoustics, thermal-electrical,

electromagnetics, and smoothed particle hydrodynamics (Dassault Systemes Simulia Corp.,

2013). As the modeled track components can potentially undergo plastic deformation, the

nonlinear analysis capability provided in ABAQUS allows for satisfactory prediction of stress-

strain behaviors of track components.

18
 Materials of different stress-strain behaviors need to be considered for the modeling of

railway track, and thus requires the FE software to integrate sufficient material models. An

extensive library of material models is provided by ABAQUS; it can be used to simulate the

realistic behavior of traditional engineering materials such as metals and rubber as well as

composite and biological materials (Dassault Systemes Simulia Corp., 2013).

As a railway track is a sophisticated system comprising multiple components with a large

number of contacting interfaces, the realistic behavior of modeled contact interactions

contributes significantly to the validity of numerical solutions. ABAQUS provides

comprehensive capabilities in modeling contact interactions, including the ability to model

interactions between deformable bodies, rigid bodies, and self-contact. Additionally, ABAQUS

allows for automatic detection of contact between different bodies, which dramatically reduces

the time required to define contact in complex assemblies such as the railway track system

considered in this study (Dassault Systemes Simulia Corp., 2013).

3.2. FE Model Overview

In order to gain insight into the effects of impact loads caused by wheel flats on railway

tracks, a FE model was developed using ABAQUS. The FE model simulated the rolling of two

railcar wheels of an axle, one of which was with a flat spot, on a finite length of a railway track.

Figure 3.1 shows the wheels and the cross-section of the track system in the FE model.

19
Figure 3.1. Schematic View of the FE Model

The FE model included a tangent track with two railcar wheels, that is, one wheel on each

rail. Field experimental results indicated that vertical loadings from the adjacent axles had almost

no effect on the track segment under the influence of the flat wheel. In addition, according to

Kerr (2003), the distribution of railseat pressure for concrete crosstie track extended less than 7

feet to both sides of wheel load application, which was less than the axle spacing of the

passenger coach in the field test. Therefore, incorporating one railcar wheel on each rail was

deemed a reasonable assumption. In the modeled track system, 136 RE rails, Safelok I fastening

systems and concrete crossties were used. As illustrated in Figure 3.1, the Safelok I fastening

system included a rail clip, shoulder, rail pad assembly (i.e. rail pad and abrasion frame), and

insulator. The detailed FE model of each track component was described in greater detail in

Chen et al. (2014). The wheel was modeled as a narrow flange railcar wheel with 1:40 tapered

wheel tread, and the geometries of the wheel web was simplified from curved to flat surfaces. In

addition, a rectangular area of flat surface, as shown in Figure 3.2, was created on the tread of

one wheel to simulate a wheel flat. The size of the wheel flat, measured tangentially along the

wheel perimeter, was specified to be 2 inches. The geometry of track, wheel, and flat spot are

consistent with the measurements from field experimentation (details are presented in Section

4.1) for FE model validation.

20
Figure 3.2. Flat Spot on the Wheel Tread

The modeled track system, as shown in Figure 3.3, consisted of 19 sets of crossties and

fastening systems with uniform crosstie spacing of 24 inches, summing to 38 feet of track. The

track was composed of three parts: a 6-foot segment at the left end, a 14-foot segment at the

center, and an 18-foot segment at the right end. The end segments of the track were provided so

that, as the impact hit the center of the track, the ends of the rails were not affected by the rail

flexural bending under wheel load (Selig & Waters, 1994). In other words, the two end segments

served as boundary conditions to the center segment.

Figure 3.3. FE Model Track Overview (Longitudinal View)

21
3.3. Mesh Sensitivity Analysis
Before element types and sizes were determined, a mesh sensitivity analysis was

performed. Mesh densities on track components, except the wheel and rail, were directly

referenced from the FE model developed by Chen et al. (2014). Because the mesh density at the

contact interface between the wheel and rail has a significantly effect on the accuracy of vertical

contact force between the two contact bodies during dynamic simulation, a mesh sensitivity

analysis was performed to determine the optimal mesh density on the wheel-rail contact interface.

The model was simplified such that only the center segment of rail and wheel were included as

the accuracy of interaction force between the wheel and rail was only affected by these two

components. The boundary condition was further simplified as the base of rail was fixed. Mesh

refinement was only implemented on the perimeter of wheel and the top center of rail to

minimize the number of elements and resulting computational time. Convergence was achieved

for vertical displacement of the wheel under a static wheel load. Three trials were conducted with

three numbers of elements: 21,619, 38,123, and 109,798 (Figure 3.4). It was found that

increasing the number of elements by 76.3%, from 21,619 to 38,123, led to an 83.0% decrease in

the vertical displacement of wheel. Further increasing the number of elements by 188%, from

38,123 to 109,798, only result in a 32.1% decrease in the vertical displacement of wheel. The

results indicated a convergence in the last trial.

22
Figure 3.4. Mesh Sensitivity Analysis for Wheel-Rail Contact Interface

With the optimal number of elements on the rail-wheel contact interface determined, the

sensitivity of the rail deformation with respect to number of elements was also investigated for

vertical strain on the neutral axis of the rail. As the optimal mesh density on the top center of rail

was already determined, mesh refinement was only performed on the areas of rail web and base.

Figure 3.5 shows the relationship between the vertical strain in the rail and number of elements.

Figure 3.5. Mesh Sensitivity Analysis for Rail Web and Base

23
 Based on the results of mesh sensitivity analysis, extremely refined mesh was required to

achieve an accurate numerical solution for wheel-rail contact force. However, as large number of

elements from using dense mesh posed considerably higher computational costs, it was only

implemented in the 14-foot segment of rail located in the center of track (Figure 3.3). Coarse

mesh was implemented on the rail segments at the two ends of track as they were not directly

subjected to the wheel load. Figure 3.6 shows the final mesh on the center and end rail segments.

Figure 3.6. Final Mesh on the (a) Center Rail Segment and (b) End Rail Segment

On the center segment of rail, the top center of rail contained the most refined mesh as

shown in Error! Reference source not found.a and Error! Reference source not found.b, and

the technique of mesh transition was utilized to ensure smooth transition among elements with

different sizes. Similarly, on the wheel, refined mesh was applied on the perimeter of the wheel

with transition to coarse mesh in the rest of the body (Error! Reference source not found.c).

24
Figure 3.7. (a) Cross-section View of the Center Rail; (b) Top view of the Center Rail; (c) Wheel;
(d) Wheel-rail Contact Interface

3.4. Element Type and Mesh Size

All track components were modeled using four-node tetrahedron and eight-node

hexahedron 3D deformable solids except for the prestressing strands which were modeled using

two-node linear beam elements in the crossties. Because the element size at the contact interface

between the wheel and rail is expected to have a significant effect on the accuracy of the output

vertical contact force between the two contact bodies during dynamic simulation, refined

elements were used on the railhead for the 14-foot rail segment. Elements were used for mesh

transition from the refined contact interface to the rest of the rail (Error! Reference source not

found.). Similarly, refined elements were implemented on the perimeter of the wheel tread with

a transition to coarser elements towards the center of the wheel.

3.5. Constitutive Relationships

The material property of concrete was defined using concrete damaged plasticity model

that considered two failure mechanisms; tensile cracking and compressive crushing. Under

25
uniaxial tensile loading, concrete exhibited linear-elastic stress-strain relationship until the

cracking stress was reached, and, thereafter, strain-softening behavior started to take place. An

additional phase, strain-hardening, was present between linear-elastic and strain-softening phases

when concrete was under uniaxial compressive loading. The two damage parameters, dt and dc as

shown in Figure 3.8, characterized concrete unloading stiffness and were not defined in the

model as cyclic loading was not included in the model (Chen et al., 2014). The important

variables used to define the constitutive behavior of concrete are summarized in Table 3.1.

Figure 3.8. Stress-strain Relation of Concrete in (a) Tension and (b) Compression (Dassault
Systemes Simulia Corp., 2013)

In the field, a track substructure is composed of multiple layers that include ballast,

subballast, and subgrade. Given that the substructure is largely made up of discrete particles,

they can be modeled using discrete element method which is able to capture the realistic

response of the track substructure (Huang & Tutumluer, 2011). However, as the computational

cost imposed by discrete element method was high and the focus of this study was not on the

behavior of the track substructure, it was simplified as a single layer of supporting block. The

26
material property incorporated in the FE model was in accordance with field data obtained from

the testing track at TTCI in Pueblo, CO (Grasse, 2013). Based on the measurement results, the

hyperelastic material model was defined for the substructure (Figure 3.9). Under compression,

the supporting block exhibited a linear stress-strain relation up to a strain value of -0.083%,

preceding a nonlinear stress-strain relation thereafter. In the range of nonlinear stress-strain

relation, when the stress was increased, the elastic modulus increased as the supporting block

hardened.

Figure 3.9. Stress and Strain Relation for Track Substructure under Compression

In the FE model, the plastic behavior of rail steel was neglected as the plastic response of

the rail was not expected in this research work. Therefore a linear-elastic stress-strain

relationship was employed to describe the material behavior of rail. Table 3.1 summarizes major

material properties associated with each track component. The field-side insulator is made of two

materials; thus it has two material properties listed.

27
Table 3.1 - Material Properties for Each Track Component
Component Density Young's Poisson's Yield Ultimate Ultimate
(lbs/in3) Modulus Ratio Strength Strength Strain
(psi) (psi) (psi)
Abrasion frame 0.0412 4.40E+05 0.35 1.20E+04 N/A N/A
Ballast 1 N/A 0.4 N/A N/A N/A
Clip 0.2836 2.30E+07 0.29 1.83E+05 2.02E+05 0.05
Crosstie 0.08333 4.35E+06 0.2 3.53E+03 7.00E+03 0.00143
Field-side Insulator 0.0412 4.40E+05 0.35 1.20E+04 N/A N/A
0.2836 2.45E+07 0.3 4.50E+04 6.50E+04 0.01
Gauge-side Insulator 0.0412 4.40E+05 0.35 1.20E+04 N/A N/A
Rail 0.29 3.00E+07 0.3 N/A N/A N/A
Rail pad 0.0368 7.50E+03 0.394 5.20E+03 N/A N/A
Shoulder 0.2836 2.45E+07 0.3 4.50E+04 6.50E+04 0.01
Strand 0.29 3.24E+07 0.3 2.55E+05 N/A N/A
Wheel 0.29 3.00E+07 0.3 N/A N/A N/A

3.6. Contact Interactions

Contact interactions between track components were formulated using surface-to-surface

contact discretization, and a master and a slave surface were defined for each contact pair. This

contact formulation method prevents large and undetected penetrations of nodes on master

surface into slave surface, providing more accurate stress and strain results compared to other

methods (Dassault Systemes Simulia Corp., 2013). The basic Coulomb friction model with the

penalty friction formulation was used to simulate the frictional force response at the contact

interface. The maximum allowable frictional stress is related to contact pressure by COF

between contacting bodies (Dassault Systemes Simulia Corp., 2013). The COFs of the contact

pairs in the model were determined from literature (Yamaguchi, 1990; Stachowiak & Batchelor,

2005) and based on a series of large-scale abrasion tests conducted at UIUC (Kernes et al., 2012).

Table 3.2 summarizes the values of COFs used in the FE model.

28
Table 3.2. Coefficient of Friction (COF) Input used in the FE Model
Frictional Interaction COF
Rail pad-rail interface 0.3
Frame-concrete interface 0.3
Insulator-rail interface 0.15
Insulator-clip interface 0.15
Insulator-shoulder interface 0.15
Shoulder-clip interface 0.5
Crosstie-ballast interface 0.7
Wheel-rail interface 0.0

The wheel-rail contact interaction was simplified from what may occur in reality with

either one or two contact patches might present between wheel and rail. The contact patches may

include the contact between the wheel tread and rail and/or the contact between the wheel flange

and rail (Anderson et al., 2004). However, according to Dukkipati and Dong (1999), wheel-rail

friction force has little influence on the impact load. In addition, as the behavior of track system

was studied in the vertical plane for this research, the contact between the wheel flange and rail

was neglected. Therefore, frictionless contact was defined for the interaction between the wheel

and rail.

Contact interaction between the legs of a shoulder and concrete crosstie involves contacts

of relatively more complex geometries and was difficult to simulate using conventional contact

formulation methods. As the relative movement between the shoulder-insert and concrete

crosstie is expected to be negligible, the constraint feature “embedded region” in ABAQUS

provides a convenient approach in modeling the interaction. The elements of shoulder inserts

were defined to be embedded in concrete crosstie, and the translational degrees of freedom of

nodes on shoulder inserts were constrained by that on concrete crosstie, well representing the

29
bonding between the two components prior to the occurrence of cracking in concrete crosstie

(Dassault Systemes Simulia Corp., 2013).

In reality, some amount of bond slippage occurs between prestressing strands and

concrete as the concrete crosstie undergoes bending. However, slippage of prestressing strands is

not likely to have significant effects on the behavior of fastening components, hence it is

neglected for this analysis. Therefore, the “embedded region” constraint that allows no bond-

slippage between hosting and embedded regions was used to model the interaction between

prestressing strands and concrete crosstie.

No axle was applied in the FE model to connect the two wheels; instead, a reference node

was created at the centroid of the wheel to simulate an axle, and the “rigid body” constraint was

used to bind the translational and rotational degrees of freedom of all nodes on the wheel with

that of the reference node (Dassault Systemes Simulia Corp., 2013). In other words, the rotation

and translation of the reference node were transmitted to the wheel.

3.7. Loading Procedure and Boundary Conditions

The FE analysis consisted of two phases. The first phase was the static loading phase that

served to stabilize the track system in a static sense before a dynamic simulation. The static

analysis phase included prestressing concrete crossties with strands, clamping clips onto the rail,

applying gravity loads to the system, and applying the wheel load. Following the manufacturer’s

specification, the tensile capacity of all the prestressing strands was 8.75 kips/strand. A

prestressing force of 7 kips was applied to each of the 20 strands embedded in each concrete

crosstie, which was 80% of the tensile strength of a strand. When the prestressing force was

released, the deformation of strands would engage concrete with compressive force. In the FE

model, the assembly of clips was initiated in the same step. Pressures were first applied to lift up

30
the toes of clips over the insulators. With the pressures applied, clips were then displaced

towards the rail and the clip inserts were socketed into the grooves on the shoulders. As the tips

of clips were directly placed over insulators, lifting pressure was decreased and the clips, then,

clamped onto the insulators. The following step was to apply gravity loads to the superstructure

components of the track system to simulate the resistance to upward deflection resulted from the

vertical wheel load. In addition, the wheel load was applied to the reference point of the wheel.

Figure 3.10 illustrates the loading and boundary conditions in each step.

The next phase was the dynamic simulation as shown in Figure 3.10, Steps 7 – 9.

ABAQUS provides two types of integration schemes for dynamic simulation: explicit and

implicit time integration schemes. An explicit dynamic analysis is computationally efficient for

analyzing large models with relatively short dynamic response times and for analyzing events or

processes that are extremely discontinuous. It takes small time increments and is typically chosen

for transient time dynamic analysis. In contrast, an implicit dynamic analysis usually gives

acceptable and/or more stable solutions with time increments typically larger than explicit

schemes by one to two orders of magnitude (Dassault Systemes Simulia Corp., 2013). As the

total dynamic step running time was expected to be relatively long, explicit schemes are less

computationally efficient. Therefore, implicit schemes were selected for the dynamic simulation.

The increment time around time of impact was refined such that the transient effects could be

simulated accurately. In the three dynamic analysis steps, a rotational speed and a translational

speed were both applied to the reference nodes at the center of the wheels. The total dynamic

simulation time was deemed sufficient as the wheel-rail interaction force could reach dynamical

equilibrium prior to impact and the wheel could keep rolling some distance after the impact.

31
Figure 3.10. Sequence of Application of Loadings and Boundary Conditions in the FE Model

32
 CHAPTER 4

FINITE ELEMENT MODEL VALIDATION FOR IMPACT LOAD MODEL

4.1. Field Test Setup

Field experiments were conducted by researchers of University of Illinois Rail

Transportation and Engineering Center (RailTEC) at the Transportation Technology Center, Inc.

(TTCI) in Pueblo, CO, USA. The results from the testing on a tangent track section were used for

model validation. For this segment, 15 new concrete crossties were installed and tamped prior to

experimentation, and strain gauges were installed on the rail to record the dynamic wheel loads

and rail behavior (Figure 4.1).

Figure 4.1. The Instrumented Track Segment at TTCI

To quantify the actual vertical wheel loads entering the rail head, strain gauges were

installed on both sides of rail web. As shown in Figure 4.2, Chevron patterns which consisted of

two gauges placed perpendicular to each other were installed 6 inches apart and centered in the

crib (the portion between two crossties) at the rail’s neutral axis on both sides of rail.

33
Figure 4.2. Strain Gauge Pattern for Vertical Wheel Load Measurement

By wiring strain gauges into a Wheatstone bridge, the difference of the averaged shear

strain at the left and right side of the applied load can be output. Referring to Euler-Bernoulli

beam theory, the concentrated applied load can be calculated by combining the shear

deformation with the cross-sectional properties of the rail. For accuracy, the strain gauge bridges

were calibrated under static loads prior to the dynamic testing.

Strain gauges were also installed in the vertical direction above the railseats to examine

the rail behavior under the wheel loads (Figure 4.3). For the instrumented locations, three gauges

were placed in a line on both sides of the rail. To eliminate the effect of lateral wheel loads, the

average value measured from the front and back sides of rail was used to compare with the

modeling results.

34
Figure 4.3. Locations of Vertical Strain Gauges Placed Above Railseat

4.2. Field Validation of FE Model with No Impact Load

Component–level models and single-crosstie FE models were previously calibrated with

field and laboratory experimental results (Chen et al., 2014). Therefore, only system level

validation was performed for this FE model. The FE model was first validated with the absence

of an impact load; two recorded time histories were compared with the FE model: vertical strains

in the rail web and vertical internal strains of the concrete crosstie read from embedment strain

gauges.

The validation based on the vertical strains in the rail web ensured that the vertical

reaction forces transmitted from the wheel to the rail was physically making sense. Figure 4.4

presents the comparisons between the field and numerical results. A section of signal containing

eight peaks is shown, in which the first two peaks are caused by the last two axles of a car, the

middle four peaks are due to all the four axles of the car following, and the last two peaks are

from the first two axles of the third car. Due to the rail roughness and other geometry

imperfections, slight fluctuation was observed in the field data. Because of the limitation of the

length of the track in the finite element model, short during of signal was recorded. To make an

appropriate comparison, the peak of finite element output is aligned to the first peak of the field

35
data. Comparing the output from the finite element model with the field data, the shapes of the

two curves exhibited good agreement. Both curves show that the vertical strain in the rail web

decreases first and reaches a minimum value (maximum compression), then starts to increase

after the wheel passage. The FE model gave a maximum compressive strain value of 150 micro

strain which was 4.7% higher than the 143 micro strain measured in the field. Although the

upper half of the signal from the finite element result looks like narrower than that from the field

test, it could be explained by the fact that only one axle was modeled in the finite element

analysis, which neglected the affect from the adjacent axles. Therefore, a good agreement

between field and numerical results for vertical strain in rail web was achieved.

Figure 4.4. Vertical Strain in Rail Web with no Impact Load

In addition to the vertical strain in the rail web, longitudinal strain measured 2 inches

below the surface of the concrete crosstie at the railseat was also calibrated against field data.

Figure 4.5 illustrates the comparisons between the field results and the numerical solutions.

Similar to Figure 4.4, the finite element signal was aligned with the first peak of the field data.
36
Despite the fluctuation in the field data caused by imperfect field conditions, the curves exhibited

similar trend. A maximum compressive strain was observed during the passage of a wheel,

indicating the measurement area was subjected to compression. The maximum compressive

strain value measured from the field was 84 micro strain and was 5.2% higher than the numerical

solution, hence good agreement was shown.

Figure 4.5. Embedment Strain in Concrete Crosstie with no Impact Load

4.3. Field Validation of FE Model with Impact Load

4.3.1. Impact Force

In the field experimentation, a wheel with a flat spot on its tread was placed on the third

axle of the 9th car of the freight consist. The diameter of the flat spot was measured as

approximately 2 inches (Figure 4.6). For the FE model validation, the train speeds in the field

experimentation and FE model were both 45 mph.

37
Figure 4.6. Flat Spot on the Wheel

The dynamic wheel loads as well as the impact loads due to train passage were measured

with the strain gauge bridges (Figure 4.2). Figure 4.7 presents the time history of the four axles

of the 9th car rolling over the instrumented track section. It should be noted that the calibration

of the bridges were conducted under various static loads applied immediately above the center of

two Chevron patterns. Therefore, the history of the measured vertical wheel loads is only valid at

the center point of each peak. In Figure 4.7, the measured dynamic vertical loads due to the

second, third and fourth wheel are all about 34 kips. Due to the existence of the flat spot located

at the first wheel, the first peak in Figure 4.7 splits up into two much narrower peaks, which

exceed the magnitude of the other readings by a factor of two. The separated peaks could be

explained by the vibrations of rail and wheel aroused by the large impact load caused by the flat

pattern.

38
Figure 4.7. Time History of the Vertical Loading Output from a Strain Gauge Bridge

For comparison, in the FE model, three cases of impact loading caused by the flat spot

were examined: the flat spot hitting directly above the centerline of a railseat, 2 inches before

approaching the railseat, and immediately above the centerline of a crib. The monitored wheel

load history for the three cases is shown in Figure 4.8. Because Figure 4.8 presents the actual

wheel-rail contact force output from the model directly, it shouldn’t be used to compare with the

shape shown in Figure 4.7. From the wheel load history, it could be found, regardless of the flat

spot hitting location, the shape of the impact load is very similar. The magnitude of the impact

load recorded from the model is approximately 74 kips, which is about 1% lower than the peak

of the impact load measured from the field. In addition, split peaks were also observed in the

numerical results, which confirmed the vibration of the rail and wheel as a result of the impact.

39
Figure 4.8. Wheel Load History Output from FE Model

In all three cases, the interaction force exhibited an abrupt decrease to zero prior to the

occurrence of wheel impact, which indicated a loss of contact between the wheel and the rail. As

illustrated in Figure 4.9, this observation can be attributed to the sudden change of contact from

the intact wheel tread to a flat spot, which causes the wheel to lose contact with the rail for a

transient time. In addition, the impact loadings were superimposed by multiple peak loads which

were caused by different mechanisms. The first peak load was a result of the colliding of wheel

and rail as shown in Figure 4.9, hence referred to as the direct impact load. The peak loads that

followed were from the vibrations of the track system under the influence of wheel impact, and

thus were the vibration-induced impact loads.

40
Figure 4.9. Changes of Wheel-rail Contact with a Flat Wheel

As shown in Figure 4.8, the magnitudes of impact loading were similar and not

significantly affected by the location of impact. When the impact occurred directly above the

centerline of the railseat, the impact loading was 74 kips, which slightly increased to 77 kips

when the location of impact was above the centerline of the crib. In addition, an impact loading

of 72 kips was observed as impact location was moved 2 inches away from the railseat. However,

a considerable difference was observed for the patterns of wheel-rail interaction force. In the

case where the impact occurred at the center of a crib, the fluctuations of loading history were

found to be stronger, indicating a larger extent of vibrations in the track system. In addition, the

maximum impact load was the vibration-induced load as opposed to the other two cases in which

the direct impact load was larger. It indicates that loads caused by the vibrations of the track

system have a potential to surpass the loads produced by the impact of the wheel on the rail, and

therefore, both the direct impact load and vibration-induced impact load need to be attenuated in

order to mitigate the damage from impact loadings.

41
4.3.2. Vertical Strain in the Rail
Vertical rail strains were also used to validate the FE model. In the same manner used for

wheel load validation, the flat spot hit the rail at three locations in the FE model. Figure 4.10 –

Figure 4.12 present the comparison of the field and numerical results under these three loading

conditions. It has been validated that the maximum vertical rail strain due to round wheel is

about 150 micro strain in compression (Figure 4.2). When the flat wheel hit immediately above

the centerline of railseat (Figure 4.10), due to the superposition of regular wheel load and impact,

the magnitude of strain was raised up to about 170 micro strain in compression. This was

demonstrated by both FE and field testing results. Similar to the wheel load, split peaks were

found, followed by smaller fluctuations. When the flat spot hit 2 inches ahead of the railseat

(Figure 4.11), the magnitude of the first peak was decreased, followed with a high peak with a

magnitude of about 150 micro strain in compression. The reduction in strain following the first

peak could be explained by the short duration of disengagement of wheel and rail when the flat

spot rolled directly above the rail, which has weakened the effect of wheel load. The second peak

was due to the re-engagement of the wheel and rail which could be seen as the source of the

impact. When the flat spot hit the center of the crib beyond the location of strain measurement

(Figure 4.12), a secondary peak was recorded next to the main peak due to the round wheel.

Because the impact occurred 2 inches away from the strain measurement, its influence was not as

strong as the first and second cases. The maximum strain due to the impact corresponding to the

third case was found to be 75 micro strain in compression.

42
Figure 4.10. Vertical Rail Strain due to Flat Spot Hit Directly Above the Centerline of Railseat

Figure 4.11. Vertical Rail Strain due to Flat Spot Hit 2 inches Ahead of the Centerline of Railseat

43
Figure 4.12. Vertical Rail Strain due to Flat Spot Hit Directly Above the Centerline of Rail Crib

44
 CHAPTER 5

FINITE ELEMENT ANALYSIS OF TRACK RESPONSE UNDER

LONGITUDINAL LOADS

5.1. FE Model Overview

In order to gain insight into the behavior of the fastening system under dynamic

longitudinal wheel load, another FE model was developed also using ABAQUS (Dassault

Systemes Simulia Corp., 2013). The FE model simulated the rolling of a locomotive wheel on a

finite length of a railway track. As mentioned in Section 2.4, the maximum longitudinal wheel

load is imparted from an accelerating locomotive wheel, thus the FE model only considered a

locomotive wheel. Figure 5.1 shows the wheel and the cross-section of the track system in the FE

model.

Figure 5.1. Schematic View of the FE Model

45
 The FE model included a tangent track with a locomotive wheel. As the tangent railway

track was symmetric about its center, the FE model only concerned half of the track system,

which significantly reduced the computational cost. The wheel-rail contact interaction was

simplified from what may occur in reality where either one or two contact patches might present

between wheel and rail. The contact patches may include the contact between the wheel tread

and rail and/or the contact between the wheel flange and rail (Anderson et al., 2004). However,

the wheel flange and rail contact was assumed to have little effect on track longitudinal response,

thus the wheel is modified such that the flange was neglected compared to the FE model

discussed in Chapter 3. As illustrated in Figure 5.1, the wheel was modeled with a circular plate

which represented the wheel tread and web. In addition, same track components were included in

the FE model.

Compared to the other FE model (Section 3.2), the length of track was extended as shown

in Figure 5.2. The modeled track consisted of 43 sets of crossties and fastening systems also with

uniform crosstie spacing of 24 inches, summing to 86 feet of track. The track was also composed

of three parts: a 36-foot segment at the left end, a 14-foot segment at the center, and another 36-

foot segment the right end. The length of the center segment was the same as the other FE model

whereas the end segments were lengthened. As the rail is much stiffer longitudinally than

vertically, longitudinal force remains more significant than vertical force beyond the region

affected by vertical wheel loads (Rhodes & Coats, 2008). Therefore, the extended length of the

end segments allowed for the fastening system located at the boundaries to experience little

longitudinal force.

46
Figure 5.2. FE Model Track Overview (Longitudinal View)

The components in the FE model used the same types of elements and mesh as that

presented in Section 3.4. In addition, same materials properties were also applied for the track

components as summarized in Table 3.1.

Same COFs were used for contact pairs as discussed in Section 3.6 except for the

interaction between the wheel and rail. Frictionless contact interaction was assumed in the other

FE model (Section 3.6) as the effect of wheel-rail friction was considered insignificant on the

behavior of impact load. On the contrary, as the frictional interaction between the wheel and rail

was the cause of longitudinal force imparted in the rail, the assumption could no longer hold

reasonable. The COF between the wheel and rail is a variable with high degree of nonlinearity

and is dependent on a variety of field variables, including temperature, humidity, speed, etc.

Based on the findings of an experimental study on the adhesion behavior between wheel and rail,

the COF ranges from 0.4 to 0.5 under dry conditions; and the value decreases with increasing

speed (Wang et al., 2011). As a simplification, a COF of 0.5 between the wheel and rail was

used in the FE model.

47
5.2. Loading Procedure and Boundary Conditions
The FE analysis consisted of two phases. The first phase was the static loading phase

which was the same as the static loading phase discussed in Section 3.7. The detailed sequence

of application of loading and boundary conditions is shown in Figure 5.3. The dynamic

simulation followed the static loading phase. As the implicit time integration schemes allowed

for larger increments of time than explicit schemes and the total dynamic step running time was

expected to be relatively long, the dynamic simulation step again utilized the implicit schemes.

To conform to the acceleration of the passenger train in the field experimentation, a rotational

acceleration of 0.65 radians/sec2, equivalent to a translational acceleration of 4.81 inch/sec2, was

applied to the reference node at the center of the wheel in the dynamic simulation step in the FE

model. The total step time was deemed sufficiently long for the wheel to pass the center segment

of rail such that enough data points could be collected.

Figure 5.3. Sequence of Application of Loadings and Boundary Conditions in the FE Model

48
 CHAPTER 6

FINITE ELEMENT MODEL VALIDATION OF LONGITUDINAL LOAD

MODEL

6.1. Field Test Setup

Field experimentation data was obtained from the same field experiments conducted at

TTCI in Pueblo, CO, USA (Section 4.1). In addition to the vertical strain gauges that were

already discussed, strain gauges were also installed on the rail web to measure the longitudinal

strain in the rail under the passage of the testing train. The layout of the longitudinal strain

gauges also formed chevron patterns except that the two perpendicular strain gauges were

aligned vertically and longitudinally. And the strain gauges were installed on the neutral axis of

the rail and at the center of the crib.

In addition, linear potentiometers were installed to measure the longitudinal displacement

of railpad under the influence of longitudinal wheel loads. Figure 6.1 shows the location of the

linear potentiometer; it was placed at the edge of the crosstie such that the tip of it was in contact

with the railpad. Readings recorded from the linear potentiometer that had positive values

indicated that the railpad displaced along the same direction the train was traveling and vice

versa.

49
Figure 6.1. Linear Potentiometer Placement in the Field Experimentation

6.2. Testing Train Consists

The field experimentation was conducted using two types of trains: a passenger train and

a freight train. The passenger train consisted of an EMD four-axle unit and nine four-axle

passenger coaches. The locomotive weighed 259.6 kips, resulting in an average wheel load of

32.45 kips. The weight of each passenger coach was approximately 87 kips. The freight train

consisted of the same locomotive and 10 four-axle freight cars. Except for an empty car that

weighed 44 kips, each three of the rest nine cars had the same weights; and the three different

weights were 263, 286, and 315 kips. The arrangements of cars and their weight correspondence

are shown in details in Figure 6.2.

50
Figure 6.2. Car Weights of Passenger and Freight Train Consists

As aforementioned, a locomotive wheel was expected to produce much higher

longitudinal load than a railcar wheel. Therefore, in order to validate the FE model, data recorded

during the passage of locomotive wheels was used.

6.3. Field Validation of FE Model

As component–level models and single-crosstie FE models were previously calibrated by

Chen et al. (2014), validation was again only performed on the system level for this FE model.

The FE model was validated by comparing recorded time histories of three properties: vertical

and longitudinal strains in the rail and longitudinal displacement of the railpad. The validation

based on the vertical and longitudinal strains of the rail web ensured that the vertical and

longitudinal forces transmitted from the wheel to the rail were physically making sense. And

validating the longitudinal displacement of the railpad in the FE model assured the longitudinal

force transferred from the rail to the railpad was a close approximation of the field condition.

In the FE model, the wheel accelerated from stationary; while, in the field

experimentation, the train accelerated past the instrumented section with some initial speed. In

other words, the testing data and the modeling data had different time scales. Therefore, side-by-

51
side comparisons were made and the emphasis was on the peak values observed and the shapes

of the curves.

6.3.1. Longitudinal Displacement of the Railpad

Based on field and numerical results (Figure 6.3), the railpad underwent a positive

displacement followed by a negative displacement. As the rail and railpad were in contact, the

moving direction of railpad always conformed to that of the rail. Therefore, the positive

displacement was attributed to the positive longitudinal elongation of rail as a result of Poisson’s

effect under the influence of vertical wheel load as the wheel approaches (Figure 6.4). Similarly,

the negative displacement was a result of both the negative longitudinal elongation of rail, as

illustrated in Figure 6.4, and movement of rail along the negative direction due to wheel-induced

friction force. As shown in Figure 6.3, the shape of the modeling data well resembled the field

data. Field data showed a maximum positive displacement of 0.000922 inch compared to

0.000867 inch from the numerical result, resulting in a mere 5.96% difference. The maximum

negative displacement in the FE model was -0.000421 inch and was 8.45% less than the field

data of -0.00046 inch. The percentage differences were considerably small; therefore, a good

agreement between the field and modeling data was realized for the longitudinal displacement of

the railpad.

52
Figure 6.3. Comparison between Field Data (left) and Modeling Data (right) for Railpad
Longitudinal Displacement

Figure 6.4. Longitudinal Force at the Rail-railpad Interface due to Vertical Wheel Load

6.3.2. Longitudinal Strain in the Rail

Figure 6.5 shows the comparison between the field data and numerical data for the

longitudinal strain in the rail. In both curves, a minimum value (maximum compression) was

proceeded and followed by two positive peak values (maximum tension). The field data gave a
53
maximum compressive strain value that was 6.23% greater than the numerical result. In addition,

the FE model gave two maximum tensile strain values that were 20.5% and 13.5% less than the

field data, respectively. Therefore, a considerably good comparison was observed between the

field data and the numerical result for the longitudinal strain in the rail.

Figure 6.5. Comparison between Test Data (left) and Modeling Data (right) for Longitudinal
Strain in Rail Web

6.3.3. Vertical Strain in the Rail

The comparison of the vertical strain in the rail is shown in Figure 6.6. The shapes of the

two curves exhibited similarities to a high extend. Both the field and the modeling data decreased

and reached a maximum compressive strain as the wheel approached, indicating a compressive

load in the rail, and started to increase after the passage of the wheel. The FE model gave a

maximum compressive strain value of -0.142 millistrain which was 2.16% different from -0.139

millistrain recorded from the field. Therefore, the comparisons between the numerical solutions

and the field data demonstrated that the FE model was capable of representing longitudinal

behaviors of the railway track structure under a dynamic wheel loading; thus, the validated FE

model could be used for further analysis.

54
Figure 6.6. Comparison between Field Data (left) and Modeling Data (right) for Vertical Strain

in Rail

55
 CHAPTER 7

PARAMETRIC STUDIES

A series of parametric studies were conducted using each of the validated FE model to

investigate the behaviors of track system under impact and longitudinal loads independently.

7.1. Parametric Study on Impact Load

For the investigation of track response to impact loads due to wheel flats, two groups of

parameters were considered, including the stiffness of rail pad and the speed of train. The

stiffness of railpad was related to the vertical modulus and vertical vibration frequency of track

system, which potentially affected the response of track to impact loads. In other words, the

impact factor, which was the ratio between the maximum impact force and nominal wheel load,

could change with the railpad stiffness. The wheel speed, on the other hand, had a direct effect

on the rotation-induced impact force as shown in Figure 4.9c, and thus could also affect the

behaviors of impact load.

7.1.1. Railpad Stiffness Results

According to Harper (1996) and Hepburn (1982), the stiffness of rail pad ranges from

731 kips/inch to 7,320 kips/inch. Three additional cases including 731 kips/inch, 1,924 kips/inch,

and 2,747 kips/inch were simulated to compare with the previous case in which a stiffness of

1,370 kips/inch was used as per the manufacturer. The time history of the wheel-rail interaction

force corresponding to each case is shown in Figure 7.1.

56
Figure 7.1. Time Histories of Wheel-rail Interaction Force with Rail Pads of Four Different
Stiffness

It can be realized that softer rail pads were able to attenuate vibration-induced peak loads

yet failed to reduce the direct impact load. Compared to stiff pads, soft pads allowed for larger

vertical rail deflection, causing the rail to bend more. As the flat spot passed over the rail, the

loss of contact between the rail and wheel gained the rail an upward action (Figure 4.9b). The

larger rail deflection allowed by soft rail pads would result in a higher rail unbend acceleration.

Therefore, a higher force could be incurred when the wheel and rail collided (Figure 4.9c). In

contrast to the increase in the direct impact load, vibration-induced peak loads were reduced as

softer rail pads attenuated the vibrational interactions between the rail and concrete crosstie.

When the stiffness was increased to 2,747 kips/inch, an exact opposite behavior of impact loads

to the case with softer pads was observed. Stiffer rail pads were able to attenuate the direct

impact load; however, the vibration-induced peak loads increased significantly.

57
 In order to quantify the effect of rail pad stiffness on the impact factor, that is, the ratio

between the maximum impact load and nominal wheel load, their relationship is illustrated in

Figure 7.2. Rail pads with a stiffness of 1,370 kips/inch provided best impact attenuation with an

impact factor of 2.35. Even though softer pads reduced vibration-induced peak loads, the impact

factor was increased to 2.56. Stiffer rail pads further increased the value to 2.53 and 2.84.

Figure 7.2. Effect of Rail Pad Stiffness on Impact Attenuation

According to Dean et al. (1983) who did a single tie impact test which utilized a drop

hammer to produce impact loading, softer pads led to significant reductions in impact factor.

Dukkipati and Dong (1999) reached the same conclusion using FE analysis and the stiffness

values used in their study were 1,142 kips/inch and 4,854 kips/inch. Their observations agree

with the right (ascending) branch of the curve in Figure 7.2. With the results gained from this

58
study, it was made aware that the selection of rail pad with small stiffness should be made with

extra caution.

7.1.2. Speed Results

In addition to the rail pad stiffness, the speed of a train can also have an effect on impact

factor. To compare with the previous case in which the wheel had a speed of 45 mph, three

higher speeds of 90 mph, 120 mph, and 225 mph were simulated. As presented in Figure 7.3,

impact factor kept increasing with increasing train speed. As illustrated in Figure 4.9c, the

wheel-rail impact was essentially the collision between the edge of wheel flat and rail. In other

words, the impact force was resulted from both rotation-induced force and rail unbend force. A

higher train speed directly increased the force induced by wheel rotation, and thus, led to higher

impact load at the time of wheel-rail impact.

Figure 7.3. Effect of Speed on Impact Factor

59
 In addition, vibration-induced impact loads diminished at high train speeds. Figure 7.4

presented the time history of wheel-rail interaction force when the speed of train was 120 mph.

The profile of vibration-induced impact load was not as conspicuous as at a lower speed (Figure

7.1). Its magnitude was only 54 kips and was 29% lower than when the speed was 45 mph. The

loss of wheel-rail contact prior to impact was much shorter at high speeds, which averted most of

the unbend action of the rail (Figure 4.9). Consequently, the vibration of the track was alleviated,

hence fiddling loads from track vibration.

Figure 7.4. Time History of Wheel-rail Interaction Force at 120 mph Train Speed

7.2. Parametric Study on Longitudinal Load

Four groups of parameters were investigated for their effects on the distribution of

longitudinal loads at the rail and railpad interface. The parameters included the acceleration of

wheel, the elastic modulus of clips, the COF between rail and railpad, and the spacing of

crossties. A higher acceleration of wheel led to a greater wheel-rail longitudinal force and could

affect both the distribution and maximum longitudinal force at the rail-railpad interfaces. The

60
modulus of elasticity of clips determined the clamping force that was produced, and a greater

clamping force was likely to increase the longitudinal load capacity of a railseat. Similarly, the

COF between the rail and railpad determined the maximum longitudinal force that could be

experienced by a railseat. In other words, if the COF increased, the longitudinal load capacity of

a railseat increased as well. Different crosstie spacing could change the distribution of vertical

wheel load at railseats, which would further affect the distribution of longitudinal forces.

A series of parametric studies were conducted using the validated FE model. The

parameters include the acceleration of wheel, elastic modulus of the clips, COF between the rail

and railpads, and spacing between crossties. After the model was validated with an acceleration

input of 4.81 inch/sec2, 30 inch/sec2 was applied to all cases other than the group of wheel

acceleration in the parametric study. The higher acceleration was deemed to represent a more

critical case as it could exert greater longitudinal force in the rail.

7.2.1. Wheel Acceleration Results

The acceleration of a train is highly dependent on the power of its locomotives, the total

weight of the train, and the maximum adhesion between the locomotive wheels and rail. The

acceleration of the train in the field experimentation was 4.81 inch/sec2. In order to investigate

the cases where the longitudinal force between the wheel and rail was larger, additional

accelerations of 12.5, 20, and 30 inch/sec2 were simulated.

61
Figure 7.5. Distribution of Longitudinal Load for Rail-to-Railpad Interfaces at the Center Seven
Crossties at Wheel Accelerations of (a) 4.81 inch/sec2, (b) 12.5 inch/sec2, (c) 20 inch/sec2, and (d)
30 inch/sec2

Figure 7.5 compares the percent distributions of longitudinal load between the rail and

railpads for the center seven railseats for the four different cases of wheel accelerations. The

railseat numbered zero is the 22nd railseat located at the center of the modeled track (Figure 3.3).

It could be observed that the longitudinal force was more concentrated to the railseats in the

vicinity of the wheel at low acceleration. As wheel acceleration increased, the distribution

62
became less skewed towards the center railseat. Higher accelerations caused the rail to displace

more in its longitudinal direction, imparting more longitudinal force to the far railseats.

The values of percent distribution were negative at the railseats where the railpad exerted

longitudinal force opposite to the direction the wheel traveled on the rail. This observation can be

explained by Figure 6.4. Before the wheel reached a railseat, the longitudinal displacement of the

railpad was positive, hence a positive longitudinal force on the railpad from the rail. Therefore,

the longitudinal force exerted on the bottom of the rail by the railpad was negative. As shown in

Figure 7.5, the percent distribution of the negative forces decreases with increasing wheel

accelerations. The reason was that higher wheel accelerations caused the rail to displace more

towards the negative direction, therewith reduced the positive displacement resulted from the

effect of vertical bending.

Figure 7.6. Relationship between Acceleration and Maximum Friction Force at Rail-railpad
Interface

63
 In addition to the percent distribution of the longitudinal force between the rail and

railpad, the relationship between its maximum value and the wheel acceleration was also

investigated. As shown in Figure 7.6, the maximum longitudinal force at the rail-to-railpad

interface increases linearly with wheel acceleration. However, the percent increase in the

maximum longitudinal force was only 54% while the acceleration increased by more than five

times. It agreed with Figure 7.5 that the percent distribution decreased at the center railseat even

though the force transmitted to it increased; and this behavior could be attributed to the high

rigidity of the rail in the longitudinal direction. In other words, the major portion of the

longitudinal force produced by the wheel at high accelerations was distributed to the far railseats.

7.2.2. Elastic Modulus of Clips Results

The modulus of elasticity of the clips has a direct effect on the clamping force applied on

the rail which is related to the capacity of longitudinal restraint a railseat has. Four elastic moduli;

20,000,000, 23,000,000, 26,000,000, and 29,000,000 psi; were simulated in the FE model. Figure

7.7 compares for the percent distributions of longitudinal load between the rail and railpads for

the center seven railseats for the four different cases of the elastic moduli of clips. The difference

in the distribution of longitudinal forces was not inconspicuous among the four cases. A modulus

of elasticity of 26,000,000 psi lead to the highest percent distribution of longitudinal force at the

center railseat, but only surpassed the lowest value by a wee difference of 0.34 percent points.

Therefore, the elastic modulus of clips had little effect of the distribution of longitudinal forces

among railseats.

64
Figure 7.7. Distribution of Longitudinal Load for Rail-to-Railpad Interfaces at the Center Seven
Crossties with Clip Elastic Modulus of (a) 20,000,000 psi, (b) 23,000,000 psi, (c) 26,000,000 psi,
and (d) 29,000,000 psi

The relationship between the maximum longitudinal force at the rail-railpad interface and

modulus of elasticity of clips is illustrated in Figure 7.8. Similar to the observations for the

distribution of longitudinal forces, there was almost no changes in the maximum force as the

elastic modulus of clips was varied. Additionally, Figure 7.8 suggested that the design of clips

could be too conservative as, for this particular train, reducing the elastic modulus from the

65
design value of 23,000,000 psi to 20,000,000 psi would barely deprive the capacity of

longitudinal restraint of fastening system.

Figure 7.8. Relationship between Clip Elastic Modulus and Maximum Friction Force at Rail-
railpad Interface

7.2.3. Coefficient of Friction between Rail and Railpad Results

The COF for the contact pair of rail and railpad has a direct effect on the longitudinal

restraint a railseat can provide. The COF for the contact between the rail and railpad has a range

between 0.12 and 1.5 (Bely et al., 1982; Friedrich, 1986; Yamaguchi, 1990). A total of four

COFs; 0.15, 0.30, 0.65, and 1.0 were simulated in the FE model. A COF of 0.15 considered a

wet surface condition whereas a COF of 0.30 considered a dry surface condition (Yamaguchi,

1990). The values of 0.65 and 1.0 were included to account for the introduction of various

degrees of roughness on the surface of railpads (Bely et al., 1982). As shown in Figure 7.9, the

percent distribution of longitudinal force at the center railseat exhibited a considerable increase

with COF up to 0.65. No significant change was observed as the COF was raised from 0.65 to

66
1.0. Therefore, considering the additional efforts required to produce a rougher surface, a COF of

0.65 was deemed most efficient for providing longitudinal restraint for a railseat.

Figure 7.9. Distribution of Longitudinal Load for Rail-to-Railpad Interfaces at the Center Seven
Crossties with Rail-to-Railpad COF of (a) 0.15, (b) 0.30, (c) 0.65, and (d) 1.0

The relationship between the maximum longitudinal force at the rail-to-railpad interface

and COF between the rail and railpads is presented in Figure 7.10. The maximum force increased

with COF; nonetheless, the rate of increase slowed down at high COFs. From a COF of 0.15 to

0.30, the maximum force increased by 20% whereas a mere 3.7% increase in the maximum force

67
was obtained as the COF increased from 0.65 to 1.0. The observations in Figure 7.10 agreed with

those in Figure 7.9 as the gains in the capacity of longitudinal restraint at a railseat was not cost-

effective beyond the COF of 0.65.

Figure 7.10. Relationship between Rail-to-Railpad COF and Maximum Friction Force at Rail-
railpad Interface

7.2.4. Crosstie Spacing Results

For concrete crossties, the center-to-center spacing of crossties in typical practice ranges

between 20 inches and 30 inches (AREMA, 2014). The crosstie spacing of the test track in the

field experimentation was 24 inches which fell in the middle of the code-specified range. Two

additional cases, 20- and 30-inch spacing, were investigated in the FE model to reveal the effects

of crosstie spacing on the distribution of longitudinal force.

For crosstie spacing of 20 and 24 inches, the distributions of longitudinal force were

similar (Figure 7.11). Small increase was observed at the center railseat while the adjacent

railseats underwent small decrease in percent distribution. Nevertheless, a significant increase,

68
3.86 percent points, only took place at the center railseat as the spacing increased from 24 to 30

inches whereas the distributions at other railseats remained almost the same. The sole increase at

the center railseat is deemed related to the increase in vertical loads experienced by the railseats

in the vicinity of the wheel due to larger crosstie spacing. Therefore, considering the three

crosstie spacing simulated in the FE model, the 24-inch spacing ensured the most uniform

distribution of longitudinal force without requiring too small of a spacing.

Figure 7.11. Distribution of Longitudinal Load for Rail-to-Railpad Interfaces at the Center Seven
Crossties with Crosstie Spacing of (a) 20 inches, (b) 24 inches, and (c) 30 inches

69
 Based on Figure 7.12, the maximum longitudinal force resembled an exponential growth

as the crossties were spaced further apart. In other words, increasing the spacing of crossties

could potentially impose a severe increase in the demand for longitudinal restraint of fastening

system. Therefore, the relationship between the maximum rail-to-railpad longitudinal force and

crosstie spacing suggested the significance of imposing an upper limit on the spacing of crossties.

Figure 7.12. Relationship between the Spacing of Crossties and Maximum Friction Force at Rail-
railpad Interface

70
 CHAPTER 8

SUMMARY AND CONCLUSIONS

8.1. Summary
FE analysis was employed to investigate the numerical behaviors of railway track system

under the influence of rail vehicle dynamic impact load and longitudinal load independently.

Two FE models were developed and validated to study each of the two loading scenarios.

Parametric studies were conducted using the FE models in an effort to understand the effects of

selected system parameters on the behaviors of impact loading and distribution of longitudinal

load.

Component-level and single-crosstie calibration were previously performed using

experimental and laboratory results by Chen et al. (2014). System-level calibration for each of

the two FE models discussed in the pages of the thesis were conducted using experimentation

data obtained from field experiments conducted at TTCI in Pueblo, CO. For the FE model on the

behavior of impact load due to wheel flat, vertical strain in the rail and internal longitudinal

strain in the concrete crosstie obtained from the FE model were compared with the field

experimental results with and without the presence of wheel impact independently. Considerable

agreements were realized for both the shapes and maximum/minimum values between the

numerical and experimental results. For the FE model on the distribution of longitudinal load,

comparisons were made for the vertical and longitudinal strains in rail and the longitudinal

displacement of railpad between the numerical and experimental results. It was found that the

modeling results were able to predict the field behaviors and resemble the experimental results to

a high extent. Therefore, both of the FE models were successfully validated and could be used to

71
study the behaviors of track system under the influence of impact load and longitudinal load

independently.

Two groups of parameters, including the stiffness of railpad and the speed of wheel, were

varied for the numerical analysis on impact load. For the numerical analysis on the distribution

of longitudinal load, the acceleration of wheel, the elastic modulus of railpad, the COF between

the rail and railpad, and the spacing of crossties were investigated for their relationship with

longitudinal loads in the rail fastening system.

8.2. Conclusions from Impact Load Analysis

Four different stiffness, between 731 kips/inch and 2,747 kips/inch, were investigated for

the effect of rail pad stiffness on impact attenuation. The results indicated rail pad with an

intermediate stiffness, 1,370 kips/inch in this case, performed best in attenuating impact load.

Rail pad with a lower stiffness, 731 kips/inch, led to an 8.9% increase in the impact factor. An

increase of 7.7% in impact factor was also observed for rail pad with a higher stiffness of 1,924

kips/inch. To investigate the relationship between train speed and impact factor, four different

speeds, ranging from 45 mph to 225 mph, were simulated in the FE model. It was realized that

the increase in impact factor was more significant at lower speeds as opposed to higher speeds.

Increasing train speed by twice, from 45 mph to 90 mph, the impact factor was increased by 34%.

In comparison, a mere 13% increase was resulted as the train speed was raised by 88%, from 120

mph to 225 mph.

Some additional conclusions can be made based on the results of the FE analysis:

 Impact loading consists of two mechanisms, direct wheel impact and vibration-induced

impact;

72
  The location of wheel impact has an effect on the behavior of impact loading, that is,

more vibrations are caused in the track system when impact hits the centerline of a crib

and thus more severe vibration-induced impact loads, however, this effect is found to be

insignificant;

 Direct impact load increases with decreasing rail pad stiffness while vibration-induced

impact load increases as rail pad becomes stiffer and an optimized rail pad stiffness exists

to reach the highest attenuation of impact loads;

 Impact load increases with increasing train speed.

 Vibration-induced impact load diminishes as train speed increases.

8.3. Conclusions from Longitudinal Load Analysis

Based on the results of parametric study from the model of the distribution of

longitudinal load, it can be concluded that:

 As the acceleration of wheel increases, the distribution of longitudinal force becomes less

skewed as the maximum percent distribution decreases and a larger portion of the

longitudinal force is distributed to the far railseats beyond the center seven.

 The maximum longitudinal force between the rail and railpad increases linearly with

wheel acceleration.

 The modulus of elasticity of clips has little effect on the distribution of longitudinal force,

suggesting the potential surplus in the elastic modulus of clips regarding the demand for

longitudinal restraint.

 The increase in COF between the rail and railpad causes a larger portion of the

longitudinal force to converge to the center railseat, essentially increasing the capacity of

73
 longitudinal restraint of fastening system. Additionally, a COF of 0.65 is deemed most

efficient in maximizing the longitudinal restraint provided at a railseat.

 Significant increase in the distribution of longitudinal force at the center railseat takes

place as the spacing of crossties increases beyond 24 inches, suggesting that 24-inch

spacing for crossties allows for most uniform distribution of longitudinal force without

requiring too small of a spacing.

 The maximum longitudinal force between the rail and railpad increases exponentially

with the spacing of crossties.

8.4. Recommendations

8.4.1. Improved Component Design for Optimal Impact Attenuation

Based on the FE analysis for the relationship between impact load due to wheel flat and

stiffness of railpad, it was realized that an intermediate stiffness, 1370 kips/inch, led to the most

desired impact attenuation result at a speed of 45 mph. For the modeled railpad, the stiffness

corresponded to an elastic modulus of 7,500 psi which happened to be the design value of the

railpad. The attenuation of impact was achieved by attenuating both direct and vibration-induced

impact loads. It was also observed that with increasing stiffness of railpads, vibration-induced

impact load increased whereas direct impact load decreased. As vibration-induced impact load

diminished at higher speeds, the attenuation of impact load would become the sole effort to

reduce direct impact load. It was thus likely that increasing the modulus of elasticity of the

railpad to be greater than 7,500 psi would reduce the resulting impact load. However, stiff

railpads would result in high vertical track stiffness, leading to increased needs on track

maintenance due to wear and fatigue of the track components caused by the augmented dynamic

forces on the rail (Real et al., 2012). Real et al. (2012) also pointed out that excessive vertical
74
track stiffness would induce higher accelerations of the train, reducing the passenger comfort.

Therefore, in the design process, considerations are needed to balance between increasing the

stiffness of railpad and limiting vertical accelerations of train in order to minimize the cost of

maintenance.

8.4.2. Improved Capacity of Longitudinal Restraint of Rail Fasteners

As previously discussed, the distribution of longitudinal load at the locations of railseat

exhibited insignificant changes regardless of the modulus of elasticity of clips. Additionally, the

maximum rail-to-railpad friction force throughout the passage of the wheel exhibited little

change with the elastic modulus of clips. The observations suggested that the design clamping

force was more than sufficient regarding the longitudinal restraint it provided at a railseat. In the

modeled track, the design modulus of elasticity of clips was 23,000,000 psi. Based on the

modeling results, reducing the modulus of elasticity by 13%, from the design value to

20,000,000 psi, led to almost no reduction in the capacity of longitudinal restraint of the railseat.

Therefore, from the mere perspective of longitudinal restraint, rail clips is allowed to be

manufactured using less stiff material, which can lead to lower manufacturing costs and less

effort for field installation.

In addition to the modulus of elasticity of clips, the COF between the rail and railpad was

another parameter that, upon modifications, could increase the capacity of longitudinal restraint

of the railseat. Previous abrasion tests by Kernes et al. (2012) suggested a COF of 0.3 between

the rail and railpad under dry field conditions. However, the COF could decrease to as low as

0.15 if moist entered the contact interface (Yamaguchi, 1990; Stachowiak & Batchelor, 2005).

And according to the modeling results, significant loss in the capacity of longitudinal restraint

was resulted as the COF decreased from 0.3 to 0.15. The maximum friction force experienced at

75
a railseat underwent a 17% reduction. Furthermore, a COF of 0.65 improved the capacity by 11%

compared to the case in which a COF of 0.3 was used. However, no significant further

improvement was exhibited as the COF increased from 0.65 to 1.0. The observations thus

indicated a considerable enhancement to the capacity of longitudinal restraint of the railseat by

increasing the design COF between the rail and railpad from 0.3 to 0.65; and it could be achieved

by roughening the surface of railpad (Bely et al., 1982).

8.5. Future Work

The parameters considered in this study for investigating the behavior of impact load due

to wheel flat included only the speed of train and the stiffness of railpad. In order to provide a

more comprehensive understanding about impact loads, other parameters such as the wheel load

applied and thickness of railpad can be included. Additionally, as concrete crossties are more

prone to excitation with the presence of gap between the ballast and crosstie, assessing the effect

of gaps can be a significantly complementary the study.

In this study, longitudinal load was considered solely with the absence of other

longitudinal stress. As aforementioned, buckling due to longitudinal stress can be a combined

effect of thermal stress, vehicle loads, and reduced track lateral resistance. Therefore, in addition

to vehicle-induced longitudinal force, applying thermal load as well as an initial lateral

misalignment in the modeled track allows for the assessment of the relationship between track

parameters and track buckling stress and temperature.

76
 REFERENCES

American Railway Engineering and Maintenance-of-Way Association (AREMA). (2014).

AREMA Manual for Railway Engineering. Lanham, MD 29706: American Railway
Engineering and Maintenance-of-Way Association.

Anderson, E., Berg, M., & Stichel, S. (2004). Rail Vehicle Dynamics. KTH Railway Technology.

Bely, V. A., Sviridenok, A. I., & Petrokovets, M. I. (1982). Friction and Wear in Polymer-Based
Materials. Elsevier.

Bian, J., Gu, Y., & Murray, M. (2013). Numerical Study of Impact Forces on Railway Sleepers
under Wheel Flat. Advances in Structural Engineering, 127-134.

Carvalho, J., Delgado, J., Calcada, R., & Delgado, R. (2013). A New Methodology for
Evaluating the Safe Temperature in Continuously Welded Rail Tracks. International
Journal of Structural Stability and Dynamics.

Chen, Z., Shin, M., Andrawes, B., & Edwards, J. (2014). Parametric study on damage and load
demand of prestressed concrete crosstie and fastening systems. Engineering Failure
Analysis, 49-61.

Dassault Systemes Simulia Corp. (2013). ABAQUS Analysis User's Manual, Version 6.13.

Dean, F., Harrison, H., Prause, R. H., & Tuten, J. M. (1983). Investigation of the Effects of Tie
Pad Stiffness on the Impact Loading of Concrete Ties in the Northeast Corridor.
Washington, D.C.: Federal Railroad Administration, Office of Research and
Development.

do Carmo, T. B. (2014). Multifaceted Approach for the Analysis of Rail Pad Assembly Response
(M.S. Thesis). Urbana, IL 61801: University of Illinois at Urbana-Champaign.

Doyle, N. F. (1980). Railway Track Design: A Review of Current Practice. BHP Melbourne
Research 301 Laboratories, Bureau of Transport Economics.

Dukkipati, R., & Dong, R. (1999). Impact Loads due to Wheel Flats and Shells. Vehicle System
Dynamics, 1-22.

Esveld, C. (2001). Modern Railway Track. Zaltbommel, the Netherlands: MRT Productions.

Foutch, D. A., Kim, T., Otter, D. E., & Doe, B. E. (2006). Investigation of Longitudinal Forces
in a Concrete Railroad Trestle. Journal of Bridge Engineering, 11(5), 618-625.

Friedrich, K. (1986). Friction and wear of polymer composites. Amsterdam: Elsevier.

Grasse, J. S. (2013). Field Test Program of the Concrete Crosstie and Fastening System (M.S.
Thesis). Urbana, IL 61801: University of Illinois at Urbana-Champaign.

77
Gustavson, R. (2002). Structural behaviour of concrete railway sleepers (PhD Thesis). Sweden:
Department of Structural Engineering, Chalmers University of Technology.

Harper, C. (1996). Handbook of plastics, elastomers, and composites (3rd ed.). New York:
McGraw-Hill.

Hay, W. W. (1982). Railroad Engineering (2nd Edition). New York, New York: John Wiley &
Sons, Inc.

Hepburn, C. (1982). Polyurethane elastomers. London: Applied Science.

Huang, H., & Tutumluer, E. (2011). Discrete Element Modeling for Fouled Railroad Ballast.
Construction and Building Materials, 3306-3312.

Kaewunruen. (2007). Experimental and numerical studies for evaluating dynamic behaviour of
prestressed concrete sleepers subject to severe impact loading (PhD Thesis). Australia:
School of Civil, Mining, and Environmental Engineering.

Kaewunruen, S., & Remennikov, A. M. (2007). Investigation of free vibrations of voided

concrete sleepers in railway track system. Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit, (pp. 495-507).

Kaewunruen, S., & Remennikov, A. M. (2009). Progressive impact behaviour of prestressed

concrete sleepers. Engineering Structures, 2460-2473.

Kaewunruen, S., & Remennikov, A. M. (2010). Dynamic Properties of Railway Track and Its
Components: Recent Findings and Future Research Direction. Insight: Non-destructive
Testing & Condition Monitoring, 20-22.

Kernes, R. G., Edwards, J. R., Dersch, M. S., Lange, D. A., & Barkan, C. P. (2012).
Investigation of the Dynamic Frictional Properties of a Concrete Crosstie Rail Seat and
Pad and Its Effect on Rail Seat Deterioration (RSD). Proceedings of the Transportation
Research Board 91st Annual Meeting. Washington, D.C.

Kerr, A. (2003). Fundamentals of railway track engineering. Omaha, NE: Simmons-Boardman

Books.

Kish, A. (2011). On the Fundamentals of Track Lateral Resistance. Proceedings of the AREMA
2011 Annual Conference. Minneapolis, MN: AREMA.

Kumar, D. K., & Sambasivarao, K. (2014). Static and Dynamic Analysis of Railway Track.
International Journal of Engineering Research and General Science, 662-671.

LBFoster. (n.d.). Fastening Systems. Retrieved from CXT Concrete Crossties:

http://www.cxtconcreteties.com/fastening_systems.asp#mstto=

National Academy of Sciences (U.S.). (1976). Report of the National Academy of Sciences.
National Research Council.

78
Nguyen, K., Goicolea, J. M., & Galbadon, F. (2011). Dynamic Analysis of High Speed Railway
Track Loads on Ballasted Track. Proceedings of the Fifth International Symposium on
Environmental Vibration. Chengdu, China.

Nielsen, J. C. (2008). High Frequency Vertical Wheel-rail Contact Forces - Validation of a

Prediction Model by Field Testing. Wear, 1465-1471.

Railway Applications - Track - Test Methods for Fastening System Part 1: Determination of
longitudinal rail restraint. (2002). In Standard EN1314601:2002. Brussels: Comitée
Européen de Normalistion (CEN).

Real, J. I., Gomez, L., Montalban, L., & Real, T. (2012). Study of the influence of geometrical
and mechanical parameters on ballasted railway tracks design. Journal of Mechanical
Science and Technology, 2837-2844.

Rodes, D., & Coats, B. (2008). Resistance to rail creep – what do rail fastenings really have to do?
Proceedings of the AREMA 2008 Annual Conference. Salt Lake City, UT.

Romero, M. J., Edwards, J. R., Barkan, C. P., Wilson, B., & Mediavilla, J. (2010).
Advancements in Fastening System Design for North America Concrete Crosstie in
Heavy-Haul Service. Proceedings of the AREMA 2010 Annual Conference & Exposition.
Orlando, FL: AREMA.

Ruge, P., & Birk, C. (2007). Longitudinal forces in continuously welded rail on bridgedecks due
to nonlinear track-bridge interaction. Computers and Structures, 458-475.

Selig, E., & Waters, J. (1994). Track geotechnology and substructure management. London: T.
Telford.

Stachowiak, G., & Batchelor, A. (2005). Engineering tribology (3rd ed.). Amsterdam: Elsevier
Butterworth-Heinemann.

Stevens, N. J., & Dux, P. F. (2004). World Intellectual Property Organisation, International
Bureau Patent No. WO2004018772 A1.

UMKC School of Law. (1945). SOUTHERN PAC. CO. v. STATE OF ARIZ. EX REL. SULLIVAN.
Retrieved from umkc.edu:
http://law2.umkc.edu/faculty/projects/ftrials/conlaw/sopacific.html

Van Dyk, B. J., Dersch, M. S., Edwards, J. R., Ruppert, C. J., & Barkan, C. P. (2014). Evaluation
of Dynamic and Impact Wheel Load Factors and Their Application for Design.
Transportation Research Board 93rd Annual Meeting. Washington D.C.

Van Dyk, B. J., Rapp, C. T., Dersch, M. S., Edwards, J. R., Ruppert, C. J., & Barkan, C. P.
(2013). Evaluation of Existing Loading Environment in North America for Improved
Concrete Sleepers and Fastening Systems. Proceedings of the 2013 World Congress of
Railway Research. Sydney, Australia.

79
Volpe, the National Transportation System Center. (2014). Track Buckling Research. Volpe, the
National Transportation System Center.

Wakui, H., & Okuda, H. (1999). A study on limit-state design for prestressed concrete sleepers.
Proceedings of JSCE (pp. 1-25). JSCE.

Wang, N. (1996). Resistance of concrete railroad ties to impact loading (PhD Thesis). Canada:
University of British Columbia.

Wang, W. J., Shen, P., Song, J. H., Liu, Q. Y., & Jin, X. S. (2011). Experimental Study on
Adhesion Behavior of Wheel/Rail under Dry and Water Conditions. Wear, 2699-2705.

Williams, B. A., Edwards, J. R., Kernes, R. G., & Barkan, C. P. (2014). Analysis of the Lateral
Load Path in Concrete Crosstie Fastening Systems. Proceedings of the 2014 Joint Rail
Conference. Colorado Springs, CO: JRC2014.

Yamaguchi, Y. (1990). Tribology of plastic materials: Their characteristics and applications to

sliding components. Amsterdam: Elsevier.

Ye, X., Wang, N., & Mindess, S. (1994). Effect of loading rate and support conditions on the
mode of failure of prestressed concrete railroad ties subjected to impact loading. Cement
and Conrete Research, 1286-1298.

Zarembski, A., & Bell, J. G. (2002, October). Limiting the Effects of High-speed Dynamic
Forces on Track Structure. TR News, 25-26.

80