THE MOVING ELEMENT METHOD AND ITS APPLICATION IN

HIGH-SPEED TRAIN-TRACK SYSTEM

LOW WIL LIAM

DEPARTMENT OF CIVIL & ENVIRONMENTAL ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2020
THE MOVING ELEMENT METHOD AND ITS APPLICATION IN

HIGH-SPEED TRAIN-TRACK SYSTEM

LOW WIL LIAM

A THESIS SUBMITTED FOR THE DEGREE OF BACHELOR OF

ENGINEERING

DEPARTMENT OF CIVIL & ENVIRONMENTAL ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

Acknowledgement

Firstly, I wish to express my sincere thanks to A/Prof. Ang Kok Keng, supervisor of my final

year project. I am extremely thankful to him for sharing his valuable guidance and

encouragement despite his busy schedule.

I am also grateful to A/Prof. Lei Tuo from Chang’an University, China. I am grateful for his

selfless knowledge sharing and guidance on my final year project when he was a visiting

researcher in National University of Singapore. I am also grateful of his online guidance and

continuous supports even after he finished his research period in National University of

Singapore and went back to China.

Besides, I wish to express my gratitude to Dr. Dai Jian who is currently doing research work

at Norwegian University of Science and Technology for his advices and comments on my

final year project.

I also place on record, my sense of gratitude to one and all, who directly or indirectly, have

lent their hand in this venture.

Table of Contents

1 Summary ............................................................................................................................ 1
2 List of symbols and Nomenclature .................................................................................... 3
3 List of tables....................................................................................................................... 6
4 Lift of figures ..................................................................................................................... 7
5 Track Model and Moving Element Method....................................................................... 9
5.1 Introduction ................................................................................................................. 9
5.2 Assumption of the model .......................................................................................... 11
5.3 Mathematical modelling ............................................................................................ 12
5.3.1 Slab track model (foundation of the model) ...................................................... 13
5.3.2 Vehicle model (superstructure on top of foundation) ........................................ 16
5.3.3 Wheel-rail contact model (the interaction condition between foundation and
superstructure).................................................................................................................. 18
5.4 Moving Element Method (MEM) ............................................................................. 19
5.5 Validation of the MEM application on the models ................................................... 29
5.5.1 Literature review ................................................................................................ 29
5.5.2 Differences in assumptions between models ..................................................... 30
5.5.3 Parameters used for Numerical model verification ........................................... 30
5.5.4 Validation and discussion .................................................................................. 32
5.6 Conclusion................................................................................................................. 33
6 Track irregularity ............................................................................................................. 34
6.1 Introduction to track irregularity ............................................................................... 34
6.2 Types of track irregularities ...................................................................................... 36
6.2.1 Vertical (longitudinal level) irregularity ............................................................ 37
6.2.2 Cross level irregularity ....................................................................................... 38
6.2.3 Rail gauge irregularity ....................................................................................... 38
6.2.4 Alignment irregularity ....................................................................................... 39
6.2.5 Twist irregularity ............................................................................................... 39
6.3 Power spectrum of track irregularity ......................................................................... 40
6.3.1 German track irregularity power spectrum ........................................................ 40
6.3.2 Chinese track irregularity power spectrum ........................................................ 44
6.3.3 Comparison between German and Chinese PSD ............................................... 46
6.4 Numerical simulation method of track irregularities random profile ....................... 47
6.4.1 White noise filtering method and quadratic filtering method ............................ 47
 6.4.2 Trigonometric series method ............................................................................. 48
6.4.3 Inverse Fast Fourier Transformation (IFFT)...................................................... 51
6.4.4 Validation and discussion .................................................................................. 57
6.5 Conclusion................................................................................................................. 60
7 Numerical study ............................................................................................................... 61
7.1 Introduction ............................................................................................................... 61
7.2 Result analysis aspect ................................................................................................ 62
7.2.1 Structural aspect: Displacement of rail, track slab and concrete base ............... 62
7.2.2 Safety aspect: Wheel-load reduction rate .......................................................... 63
7.3 Result simulation and discussion .............................................................................. 65
7.3.1 Effects caused by varying stiffness of discrete pads (𝑘1) ................................. 66
7.3.2 Effects caused by varying damping coefficient of discrete pads (𝑐1)............... 68
7.3.3 Effects caused by varying stiffness of CA mortar layer (𝑘2) ........................... 69
7.3.4 Effects caused by varying damping coefficient of CA mortar layer (𝑐2) ......... 71
7.3.5 Effects caused by varying stiffness of subgrade (𝑘3) ....................................... 72
7.3.6 Effects caused by varying damping coefficient of subgrade (𝑐3) ..................... 74
7.4 Conclusion................................................................................................................. 75
8 Conclusion of the thesis ................................................................................................... 77
9 References ........................................................................................................................ 78

2
1 Summary
This thesis introduced a numerical method to analyze dynamic response of slab track system

during the operation of the high-speed rail.

The first part of the thesis introduced the model of train-track system and Moving element

method to analyze it. The train-track system was modelled as a 6 layers structure with 3

infinite beams with Timoshenko beam behavior and discrete rail pads. The basic idea of

Moving Element method was by changing the mass, stiffness and damping properties of the

track to simulate the movement of the train although the train is simulated to be stationary at

the center of the model. To achieve such intention, a converted coordinate moving with the

vehicle was introduced to simulate the stationary position of the train with respect to the

origin of the converted coordinates. The stiffness, mass and damping matrices of the moving

element on the slab track are derived by using the standard Galerkin’s method. The train is

simplified as a multi-rigid-body model with 10 degrees of freedom. The wheel-rail

interaction was assumed under nonlinear Hertz model where the random vertical

irregularities profile was applied. The validation of the numerical model and method was

verified with other similar study.

The second part of this thesis introduced the vertical irregularity profile of the track and the

methods to generate it from power spectral density (PSD). As the track irregularity is one of

the factors that affect the train and track system performance and maintenance intervals but

inevitable due to construction technology, settlement and wear, it was important to analyze

the effect of vertical irregularity of the track to the train and track system especially when the

irregularity along the train was random. The method adopted was inverse Fast Fourier

Transform (IFFT) as it denoted the samples with high accuracy with original PSD. The

samples simulated under IFFT method was verified with other similar study as well.

1
The last part of this thesis was the numerical analysis of the effect of the properties of the

track to the dynamic responses of the track under operation. By varying the stiffness and

damping properties of the intermediate layer in the model suggested, the vertical

displacement of the track and wheel load reduction rate of the result was analyze as the

vertical displacement of the track might induced excessive dynamic response from the train

operation while wheel load reduction rate was one of the derailment factor.

2
2 List of symbols and Nomenclature
𝜌𝑎 Density of rail
𝜌𝑏 Density of track slab
𝜌𝑐 Density of concrete base
𝐴𝑎 Cross-sectional area of rail
𝐴𝑏 Cross-sectional area of track slab
𝐴𝑐 Cross-sectional area of concrete base
𝐼𝑎 Moment of inertia of rail
𝐼𝑏 Moment of inertia of track slab
𝐼𝑐 Moment of inertia of concrete base
𝐸𝑎 Young’s modulus of rail
𝐸𝑏 Young’s modulus of track slab
𝐸𝑐 Young’s modulus of concrete base
𝐺𝑎 Shear modulus of rail
𝐺𝑏 Shear modulus of track slab
𝐺𝑐 Shear modulus of concrete base
𝑘𝑎 Timoshenko shear correction coefficient of rail
𝑘𝑏 Timoshenko shear correction coefficient of track slab
𝑘𝑐 Timoshenko shear correction coefficient of concrete base
𝑘1 Stiffness of discrete pads
𝑘2 Stiffness of concrete-asphalt mortar layer
𝑘3 Stiffness of subgrade
𝑐1 Damping coefficient of discrete pads
𝑐2 Damping coefficient of concrete-asphalt mortar layer
𝑐3 Damping coefficient of subgrade
𝑦𝑎 Vertical displacements of the rail
𝑦𝑏 Vertical displacements of the track slab
𝑦𝑐 Vertical displacements of the concrete base
𝜑𝑎 Bending rotation of the rail
𝜑𝑏 Bending rotation of the track slab
𝜑𝑐 Bending rotation of the concrete base
𝐹𝑗 Hertz normal contact force at the point between rail and 𝑗 𝑡ℎ train wheel set (𝑗 =
1,2,3,4)
𝐿𝑠 spacing between two adjacent pads along the track
𝑋𝑗 travel distance of the 𝑗𝑡ℎ train wheel set (𝑗 = 1,2,3,4)
δ(g) Dirac-delta function
𝑚𝑣 Mass of coach body
𝑚𝑏𝑔 Mass of bogies
𝑚𝑤 Mass of wheels sets
𝐽𝑣 Moment of inertia of coach body
𝐽𝑏𝑔 Moment of inertia of bogies
𝑘𝑝 Stiffness of primary suspension system
𝑘𝑠 Stiffness of secondary suspension system
𝑐𝑝 Damping coefficient of primary suspension system
𝑐𝑠 Damping coefficient of secondary suspension system
𝑦𝑣 Vertical displacement at the centroids of the coach body
𝑦𝑏𝑟 Vertical displacement at the centroids of the rear bogie

3
𝑦𝑏𝑓 Vertical displacement at the centroids of the front bogie
𝑦𝑤𝑗 Vertical displacement at the centroids of the wheel sets. (𝑗 = 1,2,3,4)
𝜃𝑣 Pitching rotations at the centroids of the coach body
𝜃𝑏𝑟 Pitching rotations at the centroids of the rear bogie
𝜃𝑏𝑓 Pitching rotations at the centroids of the front bogie
𝑙1 Half distance between the center of bogies
𝑙2 Half distance between the center of wheel sets in each bogies
𝑀𝑈 Total mass matrices of the train.
𝐶𝑈 Damping matrices of the train
𝐾𝑈 Stiffness matrices of the train
𝑍𝑈 Displacement vectors of the train
𝐹𝑈 Force vectors of the train
G Wheel-rail contact coefficient
𝑦𝑎𝑗 Vertical displacement of the rail at the contact point
𝑦𝑡 Vertical irregularity of the rail at the contact point
𝑟 A coordinate system depends on velocity and travel time of a train
𝑉 Velocity of the train
𝑡 Travel time of the train
𝑥 Fixed 𝑥 -axis
𝑦𝑖𝑎 Vertical displacement of rail on the left end of the element. (𝑖 = 1,2,3,4)
𝑦𝑖𝑏 Vertical displacement of track slab on the left end of the element. (𝑖 = 1,2,3,4)
𝑦𝑖𝑐 Vertical displacement of concrete base on the left end of the element. (𝑖 =
1,2,3,4)
𝑦𝑗𝑎 Vertical displacement of rail on the right end of the element. (𝑗 = 1,2,3,4)
𝑦𝑗𝑏 Vertical displacement of track slab on the right end of the element. (𝑗 =
1,2,3,4)
𝑦𝑗𝑐 Vertical displacement of concrete base on the right end of the element. (𝑗 =
1,2,3,4)
𝜑𝑖𝑎 Bending rotation of the rail on the left end of the element. (𝑖 = 1,2,3,4)
𝜑𝑖𝑏 Bending rotation of the track slab on the left end of the element. (𝑖 = 1,2,3,4)
𝜑𝑖𝑐 Bending rotation of the concrete base on the left end of the element. (𝑖 =
1,2,3,4)
𝜑𝑗𝑎 Bending rotation of the rail on the right end of the element. (𝑗 = 1,2,3,4)
𝜑𝑗𝑏 Bending rotation of the track slab on the right end of the element. (𝑗 = 1,2,3,4)
𝜑𝑗𝑐 Bending rotation of the concrete base on the right end of the element. (𝑗 =
1,2,3,4)
𝑁𝑎𝑦 Vectors of shape function for the vertical displacement of rail nodes
𝑁𝑏𝑦 Vectors of shape function for the vertical displacement of track slab nodes
𝑁𝑐𝑦 Vectors of shape function for the vertical displacement of concrete base nodes
𝑁𝑎𝜑 Vectors of shape function for the rotations of rail nodes
𝑁𝑏𝜑 Vectors of shape function for the rotations of track slab nodes
𝑁𝑐𝜑 Vectors of shape function for the rotations of concrete base nodes
𝑍𝐿 Displacement vectors of the track
𝐹𝐿 force vectors of the track
𝑧𝑟 overall vertical irregularity of the interested track
𝑧𝐿 Vertical irregularity of the left rail on the track
𝑧𝑅 Vertical irregularity of the right rail on the track
𝑧𝑐 Cross-level irregularity of the interested track

4
 𝑌𝑔 Rail gauge irregularity
𝑌𝑡 Theoretical distance between left and right rail
𝑌𝑎𝑡 Actual distance between left and right rail
𝑌𝑎 Alignment irregularity of the interest track
𝑌𝐿 horizontal distance between actual position of left rail to the center line of
theoretical track
𝑌𝑅 horizontal distance between actual position of right rail to the center line of
theoretical track
𝑆𝑣 (𝛺) Vertical irregularity power spectral density function
𝑆𝑎 (𝛺) Alignment irregularity power spectral density function
𝑆𝑐 (𝛺) Cross-level irregularity power spectral density function
𝛺 Spatial angular frequency of track irregularity
𝛺𝑐 One of the cutoff frequencies suggested by German
𝛺𝑟 One of the cutoff frequencies suggested by German
𝛺𝑠 One of the cutoff frequencies suggested by German
𝐴𝑣 One of the roughness coefficients suggested by German
𝐴𝑎 One of the roughness coefficients suggested by German
𝑏 Half of the rollover distance from left to right of the train
𝑆𝑔 (𝛺) Gauge irregularity power spectral density function
𝐴𝑔 One of the roughness coefficients suggest by Lei.
𝑓 Spatial frequency of track irregularity
𝐴 One of the coefficients which defers in different segment of spatial frequency
suggested by Chinese.
𝑘 One of the coefficients which defers in different segment of spatial frequency
suggested by Chinese.
𝑥𝑖 Random sequence where 𝑖 = 1, 3, 5, ⋯ , 𝑛 − 1
𝑥𝑠 Time sequence where 𝑠 = 0, 1, ⋯ , (𝑁 − 1)
𝑇 Recorded length
∆ Time interval
𝜏 Latency to correlation function
𝑅𝑥𝑥 & 𝑅𝑟 Autocorrelation function of time sequence 𝑥𝑠
𝑆𝑥𝑥 (𝑘) Power spectral density function of time sequence 𝑥𝑠
𝑋(𝑘) Spectral density of time sequence 𝑥𝑠
𝑆𝑣𝑡 (𝑓𝑡 ) Vertical irregularity power spectral density function with time-domain.
𝑓𝑡 Time frequency
∆𝑃 Wheel load reduction
𝑃𝑘 Wheel load at different wheels. (𝑘 = 1,2)
̅
𝑃 Total wheel load of the two wheels
𝑝𝑚𝑎𝑥,𝑛 Maximum contact force obtained from the contact point profile in 𝑛𝑡ℎ
simulation (𝑛 = 1,2,3)
𝑝𝑚𝑖𝑛,𝑛 Minimum contact force obtained from the contact point profile in 𝑛𝑡ℎ
simulation (𝑛 = 1,2,3)
MEM Moving element method
FEM Finite element method
CA Concrete-asphalt
PSD Power spectral density

5
3 List of tables
Table 5.1: Coefficients used in track slab model. .................................................................... 13

Table 5.2: Label used in illustration of moving element method ............................................ 22

Table 5.3: Parameters for Chinese high-speed train CRH3 (Wang, n.d.) ................................ 31

Table 5.4: Parameters for CRTSII slab track (Wang, n.d.)...................................................... 31

Table 6.1: Suggested coefficients of German PSDs (Xia, Zhang, & Guo, 2018). .................. 41

Table 6.2: Suggested coefficient for Chinese PSD ................................................................. 44

Table 6.3: Segmentation points for Chinese PSD. ................................................................... 44

Table 6.4: Conversion factor corresponding to Chinese PSD percentile................................. 45

Table 7.1: Maximum and minimum contact force examples under different simulation ........ 65

6
4 Lift of figures
Figure 5.1: Illustration of train-track system suggested........................................................... 12

Figure 5.2: Illustration of Slab track model ............................................................................. 13

Figure 5.3: Illustration of vehicle model.................................................................................. 16

Figure 5.4: Illustration of moving element method ................................................................. 22

Figure 5.5: Illustration of train model suggested by Wang (Wang, n.d.). ............................... 29

Figure 5.6: Comparison of rail vertical displacement between Wang's result and analysis

result. ........................................................................................................................................ 32

Figure 5.7: Comparison of wheel-rail contact force between Wang's result and analysis result.

.................................................................................................................................................. 32

Figure 6.1: Illustration of track irregularity types. ................................................................... 36

Figure 6.2: German Low Interference PSD. ............................................................................ 43

Figure 6.3: German High Interference PSD. ........................................................................... 43

Figure 6.4: Chinese PSD at 25%, 63.2% and 90%. ................................................................. 46

Figure 6.5: Comparison between German PSDs and Chinese PSDs ....................................... 46

Figure 6.6: Illustration of bilateral spectrum (Chen, & Zhai, 1999) ........................................ 54

Figure 6.7: Simulated sample from German Low Interference PSD ....................................... 57

Figure 6.8: Simulated sample from German High Interference PSD ...................................... 57

Figure 6.9: Gao's result from German Low Interference PSD (Gao, 2015). ........................... 57

Figure 6.10: Gao's result from German High Interference PSD (Gao, 2015). ........................ 57

Figure 6.11: Simulated sample from Chinese PSD at 63.2% .................................................. 58

Figure 6.12: Simulated sample from Chinese PSD at 90% ..................................................... 58

Figure 6.13: Gao's sample from Chinese PSD at 63.2% (Gao, 2015). .................................... 58

Figure 6.14: Gao's sample from Chinese PSD at 90% (Gao, 2015). ....................................... 58

Figure 6.15: Comparison between PSD from simulated result and original German PSD ..... 59

7
Figure 6.16: Comparison between PSD from simulated result and original Chinese PSD ..... 59

Figure 7.1: Rail displacement profile example from 1st simulation......................................... 62

Figure 7.2: Rail displacement profile example from 2nd simulation........................................ 62

Figure 7.3: Rail displacement profile example from 3rd simulation ........................................ 63

Figure 7.4: Wheel-rail contact force profile from 1st simulation. ............................................ 64

Figure 7.5: Wheel-rail contact force profile from 2nd simulation ............................................ 64

Figure 7.6: Wheel-rail contact force profile from 3rd simulation............................................. 65

Figure 7.7: Vertical displacements under different Stiffness of pads ...................................... 66

Figure 7.8: Wheel load reduction rate under different Stiffness of pads ................................. 66

Figure 7.9: Vertical displacements under different damping coefficient of pads .................... 68

Figure 7.10: Wheel load reduction rate under different damping coefficient of pads ............. 69

Figure 7.11: vertical displacements under different Stiffness of CA mortar layer .................. 70

Figure 7.12: Wheel load reduction rate under different Stiffness of CA mortar layer ............ 70

Figure 7.13: Vertical displacements under different damping coefficient of CA mortar layer

.................................................................................................................................................. 71

Figure 7.14: Wheel load reduction rate under different damping coefficient of CA mortar

layer.......................................................................................................................................... 72

Figure 7.15: Vertical displacements under different Stiffness of the subgrade ....................... 73

Figure 7.16: Wheel load reduction rate under different Stiffness of the subgrade .................. 73

Figure 7.17: Vertical displacements under different damping coefficient of the subgrade ..... 74

Figure 7.18: Wheel load reduction rate under different damping coefficient of the subgrade 75

8
5 Track Model and Moving Element Method

5.1 Introduction

Slab track has become the most popular track used in high-speed rail construction as it has been

proved to have more advantages over the traditional ballasted track (Gautier, 2015). The

simplest method for train-track problems is to simplify the model into a moving load on an

infinite beam on an elastic foundation. Therefore, the methods based on this assumption such

as Laplace transform method (Frýba, 1999), Fourier transform (Kargarnovin, & Younesian,

2004), Fast Fourier Transform (Kim, 2005), mode superposition (Cai, Cheung, & Chan, 1988),

and spectral element method (Azizi, Saadatpour, & Mahzoon, 2012) have been used to analyze

train-track dynamic responses. However, those methods were only applicable to linear elastic

problem and simple cases.

One of the most popular method to analyze train-track system was the finite element method

(FEM). Under this method, slab track has been assumed into many different models, such as a

double beams with Euler-Bernoulli beam behavior (Esmaeli, Mohammadzadeh, & Mehrali,

2016), and three beams with Euler-Bernoulli beam behavior (Lei, & Zhang, 2011). However,

FEM required a global fixed coordinate system to form the structural element matrices and it

will be impractical to apply it on train-track system especially when train operating speed is

high. The reason was the faster the vehicle moves, the smaller computation time step and

element mesh required to obtain accurate result. Furthermore, the length of the track model

must have sufficient length to simulate the movement of high-speed rail to avoid inaccuracy of

the result caused by the boundaries. Hence, FEM was considered inefficient to analyze train-

track system as it involved huge computation cost.

9
Therefore, the moving element method was suggested to analyze train-track system (Koh, Ong,

Chua, & Feng, 2003). The method suggested the train remain stationary on the rail at the center

of the track model while the structural element of the track ‘flow’ with the train that was

supposed to be moving. As compared to FEM, MEM have less computation cost. There are

many studies done by applying MEM such as study on responses of high-speed rail system for

abrupt change of the foundation stiffness (Ang, & Dai, 2013), studies on responses of ballasted

track due to discretely supported high-speed rail system (Dai, Ang, Tran, Luong, & Jiang,

2017) and unsupported sleepers (Dai, Ang, Tran, Luong, & Jiang, 2018), study on responses

of three layer beam with Euler-Bernoulli beam behavior and continuous supported (Lei, &

Wang, 2013). Since Dai’s study suggested that rail with discrete supports produced more severe

vibration than rail supported continuously on ballasted track under high operating train speed,

it was meaningful to analyze the responses of three-layer beam model with discrete supported

rail and continuously support layer underneath. Furthermore, many studies have shown that the

Timoshenko beam model has better descriptions on rail vibrations than Euler-Bernoulli model

especially under high frequency vibration scenario (Knothe, Strzyzakowski, & Willner,

1994)(Xu, Zhai, Wang, & Wang, 2003)(Blanco, Alonso, Kari, Gil-Negrete, & Giménez, 2018).

Therefore, Timoshenko beam model was applied in this thesis as well to study the responses

of the train track system with discretely supported rail.

In conclusion, the train-track system was modelled with a 10 degree of freedom train, nonlinear

relationship of wheel-rail contact, discretely supported rail with Timoshenko beam behavior,

and continuously supported track slab and concrete base with Timoshenko beam behavior,

followed by analysis with MEM.

10
5.2 Assumption of the model

Before constructing the numerical model, the following assumptions on train and track system

was made:

I. The vertical dynamic behaviors of train and track system were the only concern here.

II. As the vehicle load was assumed to be distributed evenly on two rails, only one side of

the rail with half of vehicle load will be modelled.

I. The numerical model will be tested by using the CRTSII slab track which consisted of

rail, pads, track slab, CA mortar layer, concrete base and subgrade (Lei, Wu, & Zhang,

2016). The vehicle used in the model will be Chinese high-speed train CRH3 (Lei, Wu,

& Zhang, 2016).

II. The rail is considered as an infinite elastic beam with Timoshenko beam behavior

supported by discrete pads with elastic and damping properties.

III. The track slab and concrete base are also considered as infinite elastic beam with

Timoshenko beam behavior supported by continuous CA mortar layer and subgrade

respectively. The elastic and damping properties of CA mortar layer and subgrade will

be considered in the model.

11
5.3 Mathematical modelling

The train-track system was shown in Figure 5.1.

Figure 5.1: Illustration of train-track system suggested.

This model comprised three different model:

1. Slab track model (foundation of the model)

2. Vehicle model (superstructure on top of foundation)

3. Wheel-rail contact model (the interaction condition between foundation and

superstructure)

12
5.3.1 Slab track model (foundation of the model)

Figure 5.2: Illustration of Slab track model

According to Figure 5.2, the Rail, Track slab and Concrete base were modelled as Timoshenko

beams. They are interconnected by Discrete pads, CA mortar layer and Subgrade which have

different stiffness and damping coefficient. Furthermore, the rail is supported by discrete pads

where the distance between pads is 625mm while Track slab and Concrete base were

continuously support by CA mortar layer and subgrade respectively.

The label used in Figure 5.2 was according the different properties of the component shown in

Table 5.1.

Table 5.1: Coefficients used in track slab model.

Rail Track Concrete Discrete CA Subgrade

slab base pads mortar
layer
Densities 𝜌𝑎 𝜌𝑏 𝜌𝑐 - - -
Cross-sectional area 𝐴𝑎 𝐴𝑏 𝐴𝑐 - - -
Moments of inertia 𝐼𝑎 𝐼𝑏 𝐼𝑐 - - -
Young’s modulus 𝐸𝑎 𝐸𝑏 𝐸𝑐 - - -
Shear modulus 𝐺𝑎 𝐺𝑏 𝐺𝑐 - - -
Timoshenko shear 𝑘𝑎 𝑘𝑏 𝑘𝑐 - - -
correction
coefficients
Stiffness - - - 𝑘1 𝑘2 𝑘3
Damping coefficient - - - 𝑐1 𝑐2 𝑐3

13
The equations of motions were obtained for the rail supported by discrete pads as shown in Eq.

(5.1) and (5.2).

𝜕 2 𝑦𝑎 (𝑥, 𝑡) 𝜕 2 𝑦𝑎 (𝑥, 𝑡) 𝜕𝜑𝑎 (𝑥, 𝑡)

ρ𝑎 𝐴𝑎 2
− 𝑘 𝐴 𝐺
𝑎 𝑎 𝑎 2
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎
𝜕𝑡 𝜕𝑥 𝜕𝑥
𝑛
𝜕𝑦𝑎 (𝑥, 𝑡) 𝜕𝑦𝑏 (𝑥, 𝑡)
+ ∑ 𝑐1 ( − ) × δ(𝑥 − i𝐿𝑠 )
𝜕𝑡 𝜕𝑡
𝑖=1

𝑛 (5.1)
+ ∑ 𝑘1 (𝑦𝑎 (𝑥, 𝑡) − 𝑦𝑏 (𝑥, 𝑡)) × δ(𝑥 − i𝐿𝑠 )
𝑖=1

= − ∑ 𝐹𝑗 𝛿(𝑥 − 𝑋𝑗 (𝑡))
𝑗=1

𝜕 2 𝜑𝑎 (𝑥, 𝑡) 𝜕 2 𝜑𝑎 (𝑥, 𝑡) 𝜕𝑦𝑎 (𝑥, 𝑡)

ρ𝑎 𝐼𝑎 − E 𝑎 𝐼𝑎 − 𝑘𝑎 𝐴𝑎 𝐺𝑎
𝜕𝑡 2 𝜕𝑥 2 𝜕𝑥 (5.2)
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎 𝜑𝑎 (𝑥, 𝑡) = 0

Where

𝑦𝑎 and 𝑦𝑏 represent the vertical displacements of the rail beam and track slab respectively.

𝜑𝑎 represent the bending rotation of the rail.

𝐹𝑗 represent the contact force between rail and 𝑗 𝑡ℎ train wheel set

𝐿𝑠 represent the spacing between two adjacent pads along the track

𝑋𝑗 represent the travel distance of the 𝑗 𝑡ℎ train wheel set

δ(g) represent the Dirac-delta function.

14
The equations of motions were obtained for the track slab supported by CA mortar layer as

shown in Eqs. (5.3) and (5.4).

𝜕 2 𝑦𝑏 (𝑥, 𝑡) 𝜕 2 𝑦𝑏 (𝑥, 𝑡) 𝜕𝜑𝑏 (𝑥, 𝑡)

ρ𝑏 𝐴𝑏 − 𝑘𝑏 𝐴𝑏 𝐺𝑏 + 𝑘𝑏 𝐴𝑏 𝐺𝑏
𝜕𝑡 2 𝜕𝑥 2 𝜕𝑥
𝑛
𝜕𝑦𝑏 (𝑥, 𝑡) 𝜕𝑦𝑎 (𝑥, 𝑡)
+ ∑ 𝑐1 ( − ) × δ(𝑥 − i𝐿𝑠 )
𝜕𝑡 𝜕𝑡
𝑖=1

𝑛
(5.3)
+ ∑ 𝑘1 (𝑦𝑏 (𝑥, 𝑡) − 𝑦𝑎 (𝑥, 𝑡)) × δ(𝑥 − i𝐿𝑠 )
𝑖=1

𝜕𝑦𝑏 (𝑥, 𝑡) 𝜕𝑦𝑐 (𝑥, 𝑡)

+ 𝑐2 ( − ) + 𝑘2 (𝑦𝑏 (𝑥, 𝑡)
𝜕𝑡 𝜕𝑡

− 𝑦𝑐 (𝑥, 𝑡)) = 0

𝜕 2 𝜑𝑏 (𝑥, 𝑡) 𝜕 2 𝜑𝑏 (𝑥, 𝑡) 𝜕𝑦𝑏 (𝑥, 𝑡)

ρ𝑏 𝐼𝑏 − E 𝑏 𝐼𝑏 − 𝑘𝑏 𝐴𝑏 𝐺𝑏
𝜕𝑡 2 𝜕𝑥 2 𝜕𝑥 (5.4)
+ 𝑘𝑏 𝐴𝑏 𝐺𝑏 𝜑𝑏 (𝑥, 𝑡) = 0

Where

𝑦𝑐 represent the vertical displacements of the concrete base.

𝜑𝑏 represent the bending rotation of the track slab.

The equations of the motions were obtained for the concrete base supported by subgrade as

shown in Eq. (5.5) & (5.6).

𝜕 2 𝑦𝑐 (𝑥, 𝑡) 𝜕 2 𝑦𝑐 (𝑥, 𝑡) 𝜕𝜑𝑐 (𝑥, 𝑡)

ρ𝑐 𝐴𝑐 2
− 𝑘 𝐴 𝐺
𝑐 𝑐 𝑐 2
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐
𝜕𝑡 𝜕𝑥 𝜕𝑥

𝜕𝑦𝑐 (𝑥, 𝑡) 𝜕𝑦𝑏 (𝑥, 𝑡)

+ 𝑐2 ( − ) + 𝑘2 (𝑦𝑐 (𝑥, 𝑡) (5.5)
𝜕𝑡 𝜕𝑡

𝜕𝑦𝑐 (𝑥, 𝑡)
− 𝑦𝑏 (𝑥, 𝑡)) + 𝑐3 + 𝑘3 𝑦𝑐 (𝑥, 𝑡) = 0
𝜕𝑡

15
 𝜕 2 𝜑𝑐 (𝑥, 𝑡) 𝜕 2 𝜑𝑐 (𝑥, 𝑡) 𝜕𝑦𝑐 (𝑥, 𝑡)
ρ𝑐 𝐼𝑐 − E 𝑐 𝐼𝑐 − 𝑘𝑐 𝐴𝑐 𝐺𝑐
𝜕𝑡 2 𝜕𝑥 2 𝜕𝑥 (5.6)
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐 𝜑𝑐 (𝑥, 𝑡) = 0

Where

𝜑𝑐 represent the bending rotation of the concrete base.

5.3.2 Vehicle model (superstructure on top of foundation)

Figure 5.3: Illustration of vehicle model

According to Figure 5.3, the simplified model of CRH3 train consist of coach body, 2sets of

bogies with 2 sets of wheels attached on each bogie. Consider the 2 connections between coach

body, 4 connections between bogies and wheels and 4 connections between wheels and rail,

the simplified train model has 10 degrees of freedom. The label used in the figure was

according the different properties of the component shown in the following tables.

Coach Bogies Wheels sets Primary Secondary

body suspension suspension
system system
Mass 𝑚𝑣 𝑚𝑏𝑔 𝑚𝑤 - -
Moment of 𝐽𝑣 𝐽𝑏𝑔 - - -
inertia
Stiffness - - - 𝑘𝑝 𝑘𝑠
Damping - - - 𝑐𝑝 𝑐𝑠
coefficient

16
Eqs. (5.7) to (5.16) were obtained for the vehicle under Newton’s second law of motion.

𝑚𝑣 𝑦̈𝑣 + 𝑐𝑠 (2𝑦̇𝑣 − 𝑦̇ 𝑏𝑟 − 𝑦̇ 𝑏𝑓 ) + 𝑘𝑠 (2𝑦𝑣 − 𝑦𝑏𝑟 − 𝑦𝑏𝑓 ) = −𝑚𝑐 𝑔 (5.7)

𝐽𝑣 𝜃̈𝑣 + 𝑐𝑠 𝑙1 (𝑦̇ 𝑏𝑟 + 2𝜃̇𝑣 𝑙1 − 𝑦̇ 𝑏𝑓 ) + 𝑘𝑠 𝑙1 (𝑦𝑏𝑟 + 2𝜃𝑣 𝑙1 − 𝑦𝑏𝑓 ) = 0 (5.8)

𝑚𝑏𝑔 𝑦̈ 𝑏𝑟 + 𝑐𝑠 (𝑦̇ 𝑏𝑟 + 𝜃̇𝑣 𝑙1 − 𝑦̇𝑣 ) + 𝑐𝑝 (2𝑦̇ 𝑏𝑟 − 𝑦̇ 𝑤1 − 𝑦̇ 𝑤2 )

+ 𝑘𝑠 (𝑦𝑏𝑟 + 𝜃𝑣 𝑙1 − 𝑦𝑣 ) + 𝑘𝑝 (2𝑦𝑏𝑟 − 𝑦𝑤1 − 𝑦𝑤2 ) (5.9)

= −𝑚𝑏𝑔 𝑔

𝐽𝑏𝑓 𝜃̈𝑏𝑟 + 𝑐𝑝 𝑙2 (𝑦̇ 𝑤1 + 2𝜃̇𝑣 𝑙1 − 𝑦̇ 𝑤2 ) + 𝑘𝑝 𝑙2 (𝑦𝑤1 − 2𝜃𝑏𝑟 𝑙2 − 𝑦𝑤2 ) = 0 (5.10)

𝑚𝑏𝑔 𝑦̈ 𝑏𝑓 + 𝑐𝑠 (𝑦̇ 𝑏𝑓 − 𝜃̇𝑣 𝑙1 − 𝑦̇𝑣 ) + 𝑐𝑝 (2𝑦̇ 𝑏𝑓 − 𝑦̇ 𝑤3 − 𝑦̇ 𝑤4 )

+ 𝑘𝑠 (𝑦𝑏𝑓 − 𝜃𝑣 𝑙1 − 𝑦𝑣 ) + 𝑘𝑝 (2𝑦𝑏𝑓 − 𝑦𝑤3 − 𝑦𝑤4 ) (5.11)

= −𝑚𝑏𝑔 𝑔

𝐽𝑏 𝜃̈𝑏𝑓 + 𝑐𝑝 𝑙2 (𝑦̇ 𝑤3 + 2𝜃̇𝑏𝑟 𝑙2 − 𝑦̇ 𝑤4 ) + 𝑘𝑝 𝑙2 (𝑦𝑤3 + 2𝜃𝑏𝑓 𝑙2 − 𝑦𝑤4 ) = 0 (5.12)

𝑚𝑤1 𝑦̈ 𝑤1 + 𝑐𝑝 (𝑦̇ 𝑤1 + 𝜃̇𝑏𝑟 𝑙2 − 𝑦̇ 𝑏𝑟 ) + 𝑘𝑝 (𝑦𝑤1 + 𝜃𝑏𝑟 𝑙2 − 𝑦𝑏𝑟 )

(5.13)
= 𝐹1 − 𝑚𝑤1 𝑔

𝑚𝑤2 𝑦̈ 𝑤2 + 𝑐𝑝 (𝑦̇ 𝑤2 − 𝜃̇𝑏𝑟 𝑙2 − 𝑦̇ 𝑏𝑟 ) + 𝑘𝑝 (𝑦𝑤2 − 𝜃𝑏𝑟 𝑙2 − 𝑦𝑏𝑟 )

(5.14)
= 𝐹2 − 𝑚𝑤2 𝑔

𝑚𝑤3 𝑦̈ 𝑤3 + 𝑐𝑝 (𝑦̇ 𝑤3 + 𝜃̇𝑏𝑟 𝑙2 − 𝑦̇ 𝑏𝑟 ) + 𝑘𝑝 (𝑦𝑤3 + 𝜃𝑏𝑟 𝑙2 − 𝑦𝑏𝑓 )

(5.15)
= 𝐹3 − 𝑚𝑤3 𝑔

𝑚𝑤4 𝑦̈ 𝑤4 + 𝑐𝑝 (𝑦̇ 𝑤4 − 𝜃̇𝑏𝑟 𝑙2 − 𝑦̇ 𝑏𝑟 ) + 𝑘𝑝 (𝑦𝑤4 − 𝜃𝑏𝑟 𝑙2 − 𝑦𝑏𝑟 )

(5.16)
= 𝐹4 − 𝑚𝑤4

Where

𝑦𝑣 , 𝑦𝑏𝑟 , 𝑦𝑏𝑓 , and 𝑦𝑤𝑗 (j = 1,2,3,4) represent the vertical displacement at the centroids of the

coach body, rear bogie, front bogie and wheel sets respectively.

17
𝜃𝑣 , 𝜃𝑏𝑟 , and 𝜃𝑏𝑓 represent the pitching rotations at the centroids of the coach body, rear bogie,

and front bogie respectively.

2𝑙1 and 2𝑙2 represent the distance between the center of bogies and distance between wheel

sets in each bogie respectively.

By using Eqs. (5.7) to (5.16), the dynamic equation of the moving train can be deduced as Eq.

(5.17)

𝑀𝑈 𝑍̈𝑈 + 𝐶𝑈 𝑍̇𝑈 + 𝐾𝑈 𝑍𝑈 = 𝐹𝑈 (5.17)

Where

𝑀𝑈 represent the total mass matrices of the train.

𝐶𝑈 represent the damping matrices of the train.

𝐾𝑈 represent the stiffness matrices of the train.

𝑍𝑈 represent the displacement vectors of the train.

𝐹𝑈 represent the force vectors of the train.

5.3.3 Wheel-rail contact model (the interaction condition between foundation

and superstructure)

Nonlinear Hertz model was adopted to compute the normal contact force between wheel sets

and rail. Eq. (5.18) was deduced. The condition set below was to accommodate the jumping

wheel sets situation where there is no tension force between wheel sets and rail practically.

3
1 2
𝐹𝑗 = {[𝐺 (𝑦𝑎𝑗 + 𝑦𝑡 − 𝑦𝑤𝑗 )] , 𝑤ℎ𝑒𝑟𝑒 𝑦𝑎𝑗 + 𝑦𝑡 − 𝑦𝑤𝑗 ≥ 0 (5.18)
0 , 𝑤ℎ𝑒𝑟𝑒 𝑦𝑎𝑗 + 𝑦𝑡 − 𝑦𝑤𝑗 < 0

18
Where

𝐹𝑗 represents the Hertz normal contact force between the 𝑗 𝑡ℎ train wheel sets and rail at the

contact point.

G represents the wheel-rail contact coefficient. (4.57𝑅 −0.149 × 10−8 for wheels with cone tread,

3.86𝑅 −0.115 × 10−8 for wheels with worn tread, where R represents the radii of the wheels)

𝑦𝑎𝑗 represents the displacement of the rail at the contact point.

𝑦𝑡 represents the vertical irregularity of the rail at the contact point.

𝑦𝑤𝑗 represents the vertical displacement at the centroids of the wheel sets. (𝑗 = 1,2,3,4).

5.4 Moving Element Method (MEM)

To imitate the moving train on the rail without modelling a long rail in finite element model, a

moving coordinate in relation of time was deduce as shown in Eq. (5.19).

𝑟 =𝑥−𝑉 (5.19)

Where

𝑟 represents a coordinate system depends on velocity and travel time of a train

𝑉 represents the velocity of the train

𝑡 represents the travel time of the train

𝑥 represents the fixed 𝑥 -axis

The midpoint of the train model was taken as the origin of the x-axis. When the train moves at

the velocity 𝑉, the track will move at the same velocity as train. When the train travelled at a

19
 𝐿 𝐿
velocity 𝑉 for 𝑡 seconds, the position of the track is [𝑉𝑡 − , 𝑉𝑡 + ] in 𝑥 - coordinate and
2 2

𝐿 𝐿
remain [− 2 , 2] in the 𝑟 -coordinate according to Eq. (5.19).

By applying chain rule to Eq. (5.19), the equation of rail motions expressed in Eq. (5.1) and

Eq. (5.2) could be redefined as Eq. (5.20) and Eq. (5.21).

𝜕 2 𝑦𝑎 𝜕𝑦𝑎 𝜕 2 𝑦𝑎 𝜕 2 𝑦𝑎 𝜕𝜑𝑎
ρ𝑎 𝐴𝑎 (𝑉 2 2
− 2𝑉 + 2
) − 𝑘 𝐴 𝐺
𝑎 𝑎 𝑎 2
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎
𝜕𝑟 𝜕𝑟𝜕𝑡 𝜕𝑡 𝜕𝑟 𝜕𝑟
𝑛
𝜕𝑦𝑎 𝜕𝑦𝑎 𝜕𝑦𝑏 𝜕𝑦𝑏
+ ∑ 𝑐1 ( −𝑉 − +𝑉 )
𝜕𝑡 𝜕𝑟 𝜕𝑡 𝜕𝑟
𝑖=1

× δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 ) (5.20)

𝑛

+ ∑ 𝑘1 (𝑦𝑎 − 𝑦𝑏 ) × δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 )

𝑖=1

= − ∑ 𝐹𝑗 𝛿(𝑟 − 𝑅𝑗 )
𝑗=1

𝜕 2 𝜑𝑎 𝜕𝜑𝑎 𝜕 2 𝜑𝑎 𝜕 2 𝜑𝑎 𝜕𝑦𝑎
ρ𝑎 𝐼𝑎 (𝑉 2 2
− 2𝑉 + 2
) − E 𝐼
𝑎 𝑎 2
− 𝑘𝑎 𝐴𝑎 𝐺𝑎
𝜕𝑟 𝜕𝑟𝜕𝑡 𝜕𝑡 𝜕𝑟 𝜕𝑟 (5.21)
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎 𝜑𝑎 = 0

The equation of track slab and concrete base motions expressed in Eqs. (5.3) to (5.6) could also

be redefine as Eqs. (5.22) to (5.25).

20
 2
𝜕 2 𝑦𝑏 𝜕𝑦𝑏 𝜕 2 𝑦𝑏 𝜕 2 𝑦𝑏 𝜕𝜑𝑏
ρ𝑏 𝐴𝑏 (𝑉 − 2𝑉 + ) − 𝑘𝑏 𝐴𝑏 𝐺𝑏 + 𝑘𝑏 𝐴𝑏 𝐺𝑏
𝜕𝑟 2 𝜕𝑟𝜕𝑡 𝜕𝑡 2 𝜕𝑥 2 𝜕𝑥
𝑛
𝜕𝑦𝑏 𝜕𝑦𝑏 𝜕𝑦𝑎 𝜕𝑦𝑎
+ ∑ 𝑐1 ( −𝑉 − +𝑉 )
𝜕𝑡 𝜕𝑟 𝜕𝑡 𝜕𝑟
𝑖=1

× δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 )
(5.22)
𝑛

+ ∑ 𝑘1 (𝑦𝑏 − 𝑦𝑎 ) × δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 )

𝑖=1

𝜕𝑦𝑏 𝜕𝑦𝑏 𝜕𝑦𝑐 𝜕𝑦𝑐

+ 𝑐2 ( −𝑉 − +𝑉 ) + 𝑘2 (𝑦𝑏 − 𝑦𝑐 )
𝜕𝑡 𝜕𝑟 𝜕𝑡 𝜕𝑟

= 0

𝜕 2 𝜑𝑏 𝜕𝜑𝑏 𝜕 2 𝜑𝑏 𝜕 2 𝜑𝑏 𝜕𝑦𝑏
ρ𝑏 𝐼𝑏 (𝑉 2 − 2𝑉 + ) − E 𝑏 𝐼𝑏 − 𝑘𝑏 𝐴𝑏 𝐺𝑏
𝜕𝑟 2 𝜕𝑟𝜕𝑡 𝜕𝑡 2 𝜕𝑟 2 𝜕𝑟
(5.23)
+ 𝑘𝑏 𝐴𝑏 𝐺𝑏 𝜑𝑏 = 0

𝜕 2 𝜑𝑐 𝜕𝜑𝑐 𝜕 2 𝜑𝑐 𝜕 2 𝑦𝑐 𝜕𝜑𝑐
ρ𝑐 𝐴𝑐 (𝑉 2 2
− 2𝑉 + 2
) − 𝑘 𝐴 𝐺
𝑐 𝑐 𝑐 2
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐
𝜕𝑟 𝜕𝑟𝜕𝑡 𝜕𝑡 𝜕𝑟 𝜕𝑟

𝜕𝑦𝑐 𝜕𝑦𝑐 𝜕𝑦𝑏 𝜕𝑦𝑏 (5.24)

+ 𝑐2 ( −𝑉 − +𝑉 ) + 𝑘2 (𝑦𝑐 − 𝑦𝑏 )
𝜕𝑡 𝜕𝑟 𝜕𝑡 𝜕𝑟

𝜕𝑦𝑐 𝜕𝑦𝑐
+ 𝑐3 ( −𝑉 ) + 𝑘3 𝑦𝑐 = 0
𝜕𝑡 𝜕𝑟

𝜕 2 𝜑𝑐
2
𝜕𝜑𝑐 𝜕 2 𝜑𝑐 𝜕 2 𝜑𝑐 𝜕𝑦𝑐
ρ𝑐 𝐼𝑐 (𝑉 − 2𝑉 + ) − E 𝑐 𝐼𝑐 − 𝑘𝑐 𝐴𝑐 𝐺𝑐
𝜕𝑟 2 𝜕𝑟𝜕𝑡 𝜕𝑡 2 𝜕𝑟 2 𝜕𝑟 (5.25)
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐 𝜑𝑐 = 0

21
 Figure 5.4: Illustration of moving element method

To simulate the movement of track, a track element model shown in Figure 5.4 above was

established from the slab track model. The two ends of the element were subjected to vertical

displacement and rotation as shown in Figure 5.4.

The label used in Figure 5.4 was shown in Table 5.2.

Table 5.2: Label used in illustration of moving element method

Rail Track slab Concrete base

𝑖 𝑖
Vertical displacement (Left end of the element) 𝑦𝑎 𝑦𝑏 𝑦𝑐𝑖
Vertical displacement (Right end of the element) 𝑗 𝑗 𝑗
𝑦𝑎 𝑦𝑏 𝑦𝑐
Bending rotation (Left end of the element) 𝜑𝑎𝑖 𝜑𝑏𝑖 𝜑𝑐𝑖
Bending rotation (Left end of the element) 𝑗 𝑗 𝑗
𝜑𝑎 𝜑𝑏 𝜑𝑐
𝑗
For example, the vertical displacements of the two ends of the rail element were 𝑦𝑎𝑖 and 𝑦𝑎 at

node 1 and node 4 respectively while the rotations of the two ends of the rail element were 𝜑𝑎𝑖

𝑗
and 𝜑𝑎 at node 1 and node 4 respectively.

The discrete pads were modelled as discrete viscoelastic support while CA mortar layer and

subgrade were modelled as continuous viscoelastic coefficients. As the slab track model was

modelled to move under moving element method, the only moving component in the element

is the discrete pads where the direction is opposite to the direction of the moving train.

22
Therefore, the position of the discrete pads will vary with time, resulting in the need of stiffness

updating constantly.

Therefore, the nodal displacement vector of one element was defined as Eq. (5.26).

𝑗 𝑗 𝑗 𝑗 𝑗 𝑗 𝑇
𝑦 𝑒 = {𝑦𝑎𝑖 𝜑𝑎𝑖 𝑦𝑏𝑖 𝜑𝑏𝑖 𝑦𝑐𝑖 𝜑𝑐𝑖 𝑦𝑎 𝜑𝑎 𝑦𝑏 𝜑𝑏 𝑦𝑐 𝜑𝑐 } (5.26)

Therefore, the displacements of the rail, track slab and concrete base were expressed as Eq.

(5.27) and (5.28) by applying interpolation function.

𝑦𝑎 = 𝑁𝑎𝑦 𝑦 𝑒 , 𝑦𝑏 = 𝑁𝑏𝑦 𝑦 𝑒 , 𝑦𝑐 = 𝑁𝑐𝑦 𝑦 𝑒 (5.27)

𝜑𝑎 = 𝑁𝑎𝜑 𝑦 𝑒 , 𝜑𝑏 = 𝑁𝑏𝜑 𝑦 𝑒 , 𝜑𝑐 = 𝑁𝑐𝜑 𝑦 𝑒 (5.28)

Where

𝑁𝑎𝑦 , 𝑁𝑏𝑦 and 𝑁𝑐𝑦 represent the vectors of shape function for the vertical displacement of the

nodes.

𝑁𝑎𝜑 , 𝑁𝑏𝜑 and 𝑁𝑐𝜑 represent the vectors of shape function for the rotations of the nodes.

The vectors were calculated as shown in Eq. (5.29) to (5.34) by using the super convergent

locking-free interdependent interpolation elements with cubic polynomial shape function 𝑁𝑦𝑗

and 𝑁𝜑𝑗 (j = 1,2,3,4) proposed by Reddy (Reddy, 1997).

𝑁𝑎𝑦 = [𝑁𝑦1 𝑁𝑦2 0 0 0 0 𝑁𝑦3 𝑁𝑦4 0 0 0 0] (5.29)

𝑁𝑎𝜑 = [𝑁𝜑1 𝑁𝜑2 0 0 0 0 𝑁𝜑3 𝑁𝜑4 0 0 0 0] (5.30)

𝑁𝑏𝑦 = [0 0 𝑁𝑦1 𝑁𝑦2 0 0 0 0 𝑁𝑦3 𝑁𝑦4 0 0] (5.31)

𝑁𝑏𝜑 = [0 0 𝑁𝜑1 𝑁𝜑2 0 0 0 0 𝑁𝜑3 𝑁𝜑4 0 0] (5.32)

𝑁𝑐𝑦 = [0 0 0 0 𝑁𝑦1 𝑁𝑦2 0 0 0 0 𝑁𝑦3 𝑁𝑦4 ] (5.33)

𝑁𝑐𝜑 = [0 0 0 0 𝑁𝜑1 𝑁𝜑2 0 0 0 0 𝑁𝜑3 𝑁𝜑4 ] (5.34)

23
A weighting function W was multiplied with Eq. (5.20) and Eq. (5.21) and integrated over a

typical length l, lending to the weak forms as Eq. (5.35) and (5.36).

𝑙
𝜕 2 𝑦𝑎 𝜕𝑦𝑎 𝜕 2 𝑦𝑎 𝜕 2 𝑦𝑎
∫ 𝑊(𝑟) {ρ𝑎 𝐴𝑎 (𝑉 2 − 2𝑉 + ) − 𝑘 𝐴 𝐺
𝑎 𝑎 𝑎
𝜕𝑟 2 𝜕𝑟𝜕𝑡 𝜕𝑡 2 𝜕𝑟 2
0

𝜕𝜑𝑎
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎
𝜕𝑟
𝑛
𝜕𝑦𝑎 𝜕𝑦𝑎 𝜕𝑦𝑏 𝜕𝑦𝑏
+ ∑ 𝑐1 ( −𝑉 − +𝑉 ) × δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 )
𝜕𝑡 𝜕𝑟 𝜕𝑡 𝜕𝑟 (5.35)
𝑖=1

+ ∑ 𝑘1 (𝑦𝑎 − 𝑦𝑏 ) × δ(𝑟 + 𝑉𝑡 − i𝐿𝑠 )

𝑖=1

+ ∑ 𝐹𝑗 𝛿(𝑟 − 𝑅𝑗 )} 𝑑𝑟 = 0
𝑗=1

𝑙
𝜕 2 𝜑𝑎 𝜕𝜑𝑎 𝜕 2 𝜑𝑎 𝜕 2 𝜑𝑎
∫ 𝑊(𝑟) {ρ𝑎 𝐼𝑎 (𝑉 2 − 2𝑉 + ) − E 𝐼
𝑎 𝑎
𝜕𝑟 2 𝜕𝑟𝜕𝑡 𝜕𝑡 2 𝜕𝑟 2
0
(5.36)
𝜕𝑦𝑎
− 𝑘𝑎 𝐴𝑎 𝐺𝑎 + 𝑘𝑎 𝐴𝑎 𝐺𝑎 𝜑𝑎 } 𝑑𝑟 = 0
𝜕𝑟

The Galerkin’s method was applied later with the shape function 𝑁𝑎𝑦 and 𝑁𝑎𝜑 to obtain the

element mass, damping coefficient, and stiffness matrices of the first rail beam of the moving

track element as shown in Eqs. (5.37) to (5.39).

𝑙 𝑙

𝑀𝑎𝑒 = ρ𝑎 𝐴𝑎 ∫ 𝑁𝑎𝑦
𝑇 𝑇
𝑁𝑎𝑦 𝑑𝑟 + ρ𝑎 𝐼𝑎 ∫ 𝑁𝑎𝜑 𝑁𝑎𝜑 𝑑𝑟 (5.37)
0 0

𝑙 𝑙

𝐶𝑎𝑒 = −2ρ𝑎 𝐴𝑎 𝑉 ∫ 𝑁𝑎𝑦

𝑇 𝑇
𝑁𝑎𝑦,𝑟 𝑑𝑟 − 2ρ𝑎 𝐼𝑎 𝑉 ∫ 𝑁𝑎𝜑 𝑁𝑎𝜑,𝑟 𝑑𝑟
0 0 (5.38)

𝑇 𝑇
+ 𝑐1 𝑁𝑎𝑦 𝑁𝑎𝑦 δ(𝑆𝑗 ) − 𝑐1 𝑁𝑎𝑦 𝑁𝑏𝑦 δ(𝑆𝑗 )

24
 𝑙 𝑙

𝐾𝑎𝑒 = ρ𝑎 𝐴𝑎 𝑉 2 ∫ 𝑁𝑎𝑦
𝑇
𝑁𝑎𝑦,𝑟𝑟 𝑑𝑟 + ρ𝑎 𝐼𝑎 𝑉 2 ∫ 𝑁𝑎𝜑
𝑇
𝑁𝑎𝜑,𝑟𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
− E𝑎 𝐼𝑎 ∫ 𝑁𝑎𝜑 𝑁𝑎𝜑,𝑟𝑟 𝑑𝑟 − 𝑘𝑎 𝐴𝑎 𝐺𝑎 ∫ 𝑁𝑎𝑦 𝑁𝑎𝑦,𝑟𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎 ∫ 𝑁𝑎𝑦 𝑁𝑎𝜑,𝑟 𝑑𝑟 − 𝑘𝑎 𝐴𝑎 𝐺𝑎 ∫ 𝑁𝑎𝜑 𝑁𝑎𝑦,𝑟 𝑑𝑟 (5.39)
0 0

𝑙
𝑇 𝑇
+ 𝑘𝑎 𝐴𝑎 𝐺𝑎 ∫ 𝑁𝑎𝜑 𝑁𝑎𝜑 𝑑𝑟 − 𝑐1 𝑉𝑁𝑎𝑦 𝑁𝑎𝑦,𝑟 δ(𝑆𝑗 )
0

𝑇 𝑇
+ 𝑐1 𝑉𝑁𝑎𝑦 𝑁𝑏𝑦,𝑟 δ(𝑆𝑗 ) + 𝑘1 𝑁𝑎𝑦 𝑁𝑎𝑦 δ(𝑆𝑗 )

𝑇
− 𝑘1 𝑁𝑎𝑦 𝑁𝑏𝑦 δ(𝑆𝑗 )

Where

( )( ),𝑟 and( )( ),𝑟𝑟 represent the first and second partial derivatives with respect to 𝑟

respectively.

The terms containing the Dirac-delta function δ(𝑆𝑗 ) are used to accommodate the effect of the

positions of discrete pads.

Similarly, the Eqn. (5.22) and Eqn. (5.23) were used to deduce the track slab element matrices

as shown in Eqs. (5.40) to (5.42).

𝑙 𝑙

𝑀𝑏𝑒 𝑇
= ρ𝑏 𝐴𝑏 ∫ 𝑁𝑏𝑦 𝑇
𝑁𝑏𝑦 𝑑𝑟 + ρ𝑏 𝐼𝑏 ∫ 𝑁𝑏𝜑 𝑁𝑏𝜑 𝑑𝑟 (5.40)
0 0

𝑙 𝑙

𝐶𝑏𝑒 = −2ρ𝑏 𝐴𝑏 𝑉 ∫ 𝑁𝑏𝑦

𝑇 𝑇
𝑁𝑏𝑦,𝑟 𝑑𝑟 − 2ρ𝑏 𝐼𝑏 𝑉 ∫ 𝑁𝑏𝜑 𝑇
𝑁𝑏𝜑,𝑟 𝑑𝑟 + 𝑐1 𝑁𝑏𝑦 𝑁𝑏𝑦 δ(𝑆𝑗 )
0 0
(5.41)
𝑙 𝑙
𝑇 𝑇 𝑇
− 𝑐1 𝑁𝑏𝑦 𝑁𝑎𝑦 δ(𝑆𝑗 ) + 𝑐2 ∫ 𝑁𝑏𝑦 𝑁𝑏𝑦 𝑑𝑟 − 𝑐2 ∫ 𝑁𝑏𝑦 𝑁𝑐𝑦 𝑑𝑟
0 0

25
 𝑙 𝑙
𝑇 𝑇
𝐾𝑎𝑒 = ρ𝑏 𝐴𝑏 𝑉 2 ∫ 𝑁𝑏𝑦 𝑁𝑏𝑦,𝑟𝑟 𝑑𝑟 + ρ𝑏 𝐼𝑏 𝑉 2 ∫ 𝑁𝑏𝜑 𝑁𝑏𝜑,𝑟𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
− E𝑏 𝐼𝑏 ∫ 𝑁𝑏𝜑 𝑁𝑏𝜑,𝑟𝑟 𝑑𝑟 − 𝑘𝑏 𝐴𝑏 𝐺𝑏 ∫ 𝑁𝑏𝑦 𝑁𝑏𝑦,𝑟𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
+ 𝑘𝑏 𝐴𝑏 𝐺𝑏 ∫ 𝑁𝑏𝑦 𝑁𝑏𝜑,𝑟 𝑑𝑟 − 𝑘𝑏 𝐴𝑏 𝐺𝑏 ∫ 𝑁𝑏𝜑 𝑁𝑏𝑦,𝑟 𝑑𝑟
0 0

𝑙
𝑇 𝑇
+ 𝑘𝑏 𝐴𝑏 𝐺𝑏 ∫ 𝑁𝑏𝜑 𝑁𝑏𝜑 𝑑𝑟 + 𝑐1 𝑉𝑁𝑏𝑦 𝑁𝑎𝑦,𝑟 δ(𝑆𝑗 )
0 (5.42)
𝑇 𝑇
− 𝑐1 𝑉𝑁𝑏𝑦 𝑁𝑏𝑦,𝑟 δ(𝑆𝑗 ) + 𝑘1 𝑁𝑏𝑦 𝑁𝑏𝑦 δ(𝑆𝑗 )

𝑙
𝑇 𝑇
− 𝑘1 𝑁𝑏𝑦 𝑁𝑎𝑦 δ(𝑆𝑗 ) − 𝑐2 𝑉 ∫ 𝑁𝑏𝑦 𝑁𝑏𝑦,𝑟 𝑑𝑟
0

𝑙 𝑙
𝑇 𝑇
+ 𝑐2 𝑉 ∫ 𝑁𝑏𝑦 𝑁𝑐𝑦,𝑟 𝑑𝑟 + 𝑘2 ∫ 𝑁𝑏𝑦 𝑁𝑏𝑦 𝑑𝑟
0 0

𝑙
𝑇
− 𝑘2 ∫ 𝑁𝑏𝑦 𝑁𝑐𝑦 𝑑𝑟
0

Similarly, the Eqn. (5.24) and Eqn. (5.25) were used to deduce the concrete base element

matrices as shown in Eqs (5.43) to (5.45).

𝑙 𝑙

𝑀𝑐𝑒 = ρ𝑐 𝐴𝑐 ∫ 𝑁𝑐𝑦
𝑇 𝑇
𝑁𝑐𝑦 𝑑𝑟 + ρ𝑐 𝐼𝑐 ∫ 𝑁𝑐𝜑 𝑁𝑐𝜑 𝑑𝑟 (5.43)
0 0

𝑙 𝑙

𝐶𝑏𝑒 = −2ρ𝑐 𝐴𝑐 𝑉 ∫ 𝑁𝑐𝑦

𝑇 𝑇
𝑁𝑐𝑦,𝑟 𝑑𝑟 − 2ρ𝑐 𝐼𝑐 𝑉 ∫ 𝑁𝑐𝜑 𝑁𝑐𝜑,𝑟 𝑑𝑟
0 0
(5.44)
𝑙 𝑙 𝑙
𝑇 𝑇 𝑇
+ 𝑐2 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦 𝑑𝑟 − 𝑐2 ∫ 𝑁𝑐𝑦 𝑁𝑏𝑦 𝑑𝑟 + 𝑐3 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦 𝑑𝑟
0 0 0

26
 𝑙 𝑙

𝐾𝑐𝑒 = ρ𝑐 𝐴𝑐 𝑉 2 ∫ 𝑁𝑐𝑦
𝑇
𝑁𝑐𝑦,𝑟𝑟 𝑑𝑟 + ρ𝑐 𝐼𝑐 𝑉 2 ∫ 𝑁𝑐𝜑
𝑇
𝑁𝑐𝜑,𝑟𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
− E𝑐 𝐼𝑐 ∫ 𝑁𝑐𝜑 𝑁 𝑏 𝑑𝑟 − 𝑘𝑐 𝐴𝑐 𝐺𝑐 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦,𝑟𝑟 𝑑𝑟
𝑐𝜑,𝑟𝑟
0 0

𝑙 𝑙
𝑇 𝑇
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐 ∫ 𝑁𝑐𝑦 𝑁𝑐𝜑,𝑟 𝑑𝑟 − 𝑘𝑐 𝐴𝑐 𝐺𝑐 ∫ 𝑁𝑐𝜑 𝑁𝑐𝑦,𝑟 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
+ 𝑘𝑐 𝐴𝑐 𝐺𝑐 ∫ 𝑁𝑐𝜑 𝑁𝑐𝜑 𝑑𝑟 − 𝑐2 𝑉 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦,𝑟 𝑑𝑟 (5.45)
0 0

𝑙 𝑙
𝑇 𝑇
+ 𝑐2 𝑉 ∫ 𝑁𝑐𝑦 𝑁𝑏𝑦,𝑟 𝑑𝑟 + 𝑘2 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦 𝑑𝑟
0 0

𝑙 𝑙
𝑇 𝑇
− 𝑘2 ∫ 𝑁𝑐𝑦 𝑁𝑏𝑦 𝑑𝑟 − 𝑐3 𝑉 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦,𝑟 𝑑𝑟
0 0

𝑙
𝑇
+ 𝑘3 ∫ 𝑁𝑐𝑦 𝑁𝑐𝑦 𝑑𝑟
0

The mass, damping and stiffness matrices of a typical moving track element were deduced by

assembling all 9 equations from Eqs. (5.37) to (5.45) as shown in Eqs. (5.46) to (5.48).

𝑀𝑒 = 𝑀𝑎𝑒 + 𝑀𝑏𝑒 + 𝑀𝑐𝑒 (5.46)

𝐶 𝑒 = 𝐶𝑎𝑒 + 𝐶𝑏𝑒 + 𝐶𝑐𝑒 (5.47)

𝐾 𝑒 = 𝐾𝑎𝑒 + 𝐾𝑏𝑒 + 𝐾𝑐𝑒 (5.48)

Finally, the total mass matrix 𝑀𝐿 , damping 𝐶𝐿 and stiffness matrix 𝐾𝐿 of the slab track could

be deduced by using ‘set-in-right-position’ rule. The dynamic equation of the track can be

deduced as shown in Eq. (5.48).

𝑀𝐿 𝑍̈𝐿 + 𝐶𝐿 𝑍̇𝐿 + 𝐾𝐿 𝑍𝐿 = 𝐹𝐿 (5.49)

27
Where

𝑍𝐿 represent the displacement vectors of the track.

𝐹𝐿 represent the force vectors of the track.

The Eq. (5.17) and Eq. (5.49) were combined into Eq. (5.50) to analyze the entire model

developed earlier.

𝑀𝑈 0 𝑍̈𝑈 𝐶 0 𝑍̇𝑈 𝐾 0 𝑍𝑈 𝐹
[ ]{ } + [ 𝑈 ]{ } + [ 𝑈 ] { } = { 𝑈} (5.50)
0 𝑀𝐿 𝑍̈𝐿 0 𝐶𝐿 𝑍̇𝐿 0 𝐾𝐿 𝑍𝐿 𝐹𝐿

The following steps were used to solve Eqn. (50).

1. Eq. (5.13) to (5.16) in Eq. (5.17) were substituted into Eqn. (5.49) to eliminate 𝐹𝑗 .

2. Eq. (5.18) was substituted into Eq. (5.13) to (5.16) to eliminate the negative 𝐹𝑗 in Eq.

(0.17).

3. Eq. (5.50) was now converted to a coupled nonlinear system without 𝐹𝑗 as shown in Eq.

(5.51).

𝑓1 (𝑍̈𝑈 , 𝑍̇𝑈 , 𝑍𝑈 , 𝑍̈𝐿 , 𝑍̇𝐿 , 𝑍𝐿 , 𝑐1 ) = 0

{ ⋯ (5.51)
𝑓𝑛 (𝑍̈𝑈 , 𝑍̇𝑈 , 𝑍𝑈 , 𝑍̈𝐿 , 𝑍̇𝐿 , 𝑍𝐿 , 𝑐𝑛 ) = 0

Where

𝑓1 , …, 𝑓𝑛 represent each equation in Eqn. (5.50).

𝑐1, …, 𝑐𝑛 represent constant term in each equation.

4. Eqn. (5.51) was linearized by using Newton-Raphson’s scheme at each iteration.

5. Newmark’s integration method was used at each time step.

28
5.5 Validation of the MEM application on the models

The verification of the numerical proposed is done by comparing the result between analysis

and a published result (Wang, n.d.).

5.5.1 Literature review

The proposed train-track model in the literature was shown in Figure 5.5.

Figure 5.5: Illustration of train model suggested by Wang (Wang, n.d.).

The assumption made by literature (Wang, n.d.):

1. Only vertical dynamic response was the concern here.

2. The numerical model was built by using the CRTSII slab track which consisted of rail,

pads, track slab, CA mortar layer, concrete base and subgrade. The vehicle used in the

model will be Chinese high-speed train CRH3

3. As the vehicle load was assumed to be distributed evenly on two rails, only one side of

the rail with half of vehicle load will be modelled.

29
 4. The rail was continuously supported by track slab fixed with Vossloh Loarv 3001 rail

fastening and pad with a stiffness kr and damping coefficient cr.

5. The track slab was continuously supported by CA mortar layer with a stiffness kt and

damping coefficient ct.

6. The concrete slab was continuously supported by subgrade with a stiffness kb and

damping coefficient cb.

5.5.2 Differences in assumptions between models

As Wang’s paper analyzed the dynamic response of the slab-track model of high-speed rail

system by using the Moving Element Method, the result obtained in the paper will be a reliable

reference to verify the numerical model proposed in this paper. Furthermore, the train-track

system proposed by Wang was identical the system proposed earlier with the only exception

of pads connection between two system. Wang’s paper suggested a continuously pads while

the system proposed earlier applied a discrete pad between the rail and track slab. In addition,

Wang’s paper assumed the rail, track slab and concrete slab as beam with Euler-Bernoulli beam

properties while the system proposed applied Timoshenko beam theory in analysis. Although

by applying the same parameters for Chinese high-speed train CRH3 and CRTSII slab track

model into the system proposed, the differences mentioned should be noted while comparing

the results between Wang’s paper and calculation.

5.5.3 Parameters used for Numerical model verification

As mentioned, the parameters for Chinese high-speed train CRH3 and CRTSII slab track

shown in Table 5.3 and

Table 5.4 were applied to the system proposed and the results were compared with Wang’s

paper.

30
 Table 5.3: Parameters for Chinese high-speed train CRH3 (Wang, n.d.)

Parameters Value from

reference
Mass of vehicle body 2 mv (kg) 40,000
Mass of bogie 2 mt (kg) 3200
Mass of wheel 2 mwi (kg) 2400
Pitch inertia of vehicle body 2 J v (kg·m2) 5.47×105
Pitch inertia of vehicle body 2 J t (kg·m2) 6800
Stiffness of primary suspension system 2𝑘𝑝 (MN/m) 2.08
Stiffness of secondary suspension system 2𝑘𝑠 (MN/m) 0.8
Damping of primary suspension system 2𝑐𝑝 (kN·s /m) 100
Damping of secondary suspension system 2𝑐𝑠 ( kN·s /m) 120
Wheelbase 2 l1 (m) 2.50
Distance between centers of front and rear bogies 2 l2 (m) 17.375
Stiffness of wheel-rail contract kc (MN/m) 1.325×103
Axial load (kN) 140

Table 5.4: Parameters for CRTSII slab track (Wang, n.d.)

Parameters Value
Rail Pad Distance of ties (m) 0.625
Stiffness of clips and pad 60
(MN·m-1)
Damping of clips and pad 50
(kN·s·m-1)
Length (m) 6.45
Width (m) 2.55
Track slab Height (m) 0.20
Density (kg·m-3) 2500
Young's modulus (MPa) 3.9×104
Cement-asphalt layer (CA Stiffness (MN·m-1) 0.9×103
layer) Damping (kN·s·m-1) 80
Upper bottom width (m) 2.95
Lower bottom width (m) 3.25
Concrete base Height (m) 0.30
Density (kg·m-3) 2500
Young's modulus (MPa) 3.3×104
Subgrade Stiffness (MN·m-1) 65
Damping (kN·s·m-1) 90

As Wang’s paper mentioned the result was obtained with train operating speed of 72km/h and

without track irregularity, the system will be adjusted to this scenario.

31
5.5.4 Validation and discussion

Figure 5.6: Comparison of rail vertical displacement between Wang's result and analysis
result.

Figure 5.7: Comparison of wheel-rail contact force between Wang's result and analysis result.

32
As shown in Figure 5.6 and Figure 5.7, the analysis result was identical to Wang’s result despite

some minor discrepancies observed at the contact point between wheels and rail. The results

could be due to the difference in model where Wang’s suggested continuous pads while the

numerical applied in this paper suggested discrete pads.

5.6 Conclusion

In this chapter, the numerical model of train-track system, consisted of train model, track model

and wheel-rail contact model, was introduced to simulate the actual model of CRTSII slab track

and CRH3 train. The numerical model was solved by MEM suggested and the validation of the

system was done by comparing results with previous study (Wang, n.d.) with same parameter

used. The small difference between the results was justify by the model difference between the

numerical model suggested in this paper and Wang’s model.

33
6 Track irregularity

6.1 Introduction to track irregularity

As high-speed rail track structure consisted of concrete structure, the inconsistency of the

density and elasticity of the supporting structure will inevitably cause the geometrical defects

of track. The type of track irregularity includes vertical (longitudinal level) irregularity, cross

level irregularity, rail gauge irregularity, alignment irregularity and twist irregularity. These

track irregularities were not just affecting the safety and comfort level of the operating train,

but also induced the additional loading and unloading of the wheel-rail contact force which

leads to damages of the operating train.

As suggested by many experimental data from rail transportation system from different

countries, the vibration and wheel-rail contact force of the train will be increased drastically

when the track irregularities profile was unfavorable, even though the overall structural

strength of the track was adequate. It will also lead to higher possibility of derailment and

higher frequency of the train maintenance need. Therefore, the track irregularities were the

main deciding factor of train operating speed. In order to achieve high speed intercity

transportation while prolong the life span of the parts in train and track structure, the track

irregularities must be controlled under certain level. As the high-speed rail transportation will

have shorter interval between trains and higher capacity to transport cargo, the track structure

of high-speed rail transportation system will experience higher stress than normal rail and the

analysis of train-track dynamic responses will be meaningful and beneficial to improve the

efficiency of high-speed rail transportation system. However, the analysis required a reliable

track irregularity profile sample.

34
The two mainstream approaches to assume track irregularity profile were the deterministic

function method and random process method. The deterministic function method is assuming

the irregularity to have a sinusoidal or cosinusoidal profile. The random process method is

assuming the irregularity to have randomized profile which could be generate from the power

spectral density (PSD) of the track. As the occurrence of track irregularities usually due to

inconsistency of concrete structural strength throughout the track, consolidation at random

location of track etc., the track irregularity profile is random realistically and the random

process method was considered more preferable method than the deterministic function method.

The random process method was widely used in many studies such as study on dynamic

responses of vehicle and track system with the track vertical irregularity which considered as

a stationary ergodic Gaussian random process (Lei, & Noda, 2002).

As the random irregularity profile of track can only obtained through instrumentation result

from actual built track, many countries such as China (China National Railway Administration ,

2014), Germany, United States (Yang, Zhai, Gao, Li, Kang, & Liu, 2014) has begun to run

experiment on the actual track and proposed their PSDs of tracks. German PSDs was adopted

in many studies such as study on dynamic responses of slab track system due to track

irregularity under German PSDs (Yang, Gu, Zhou, Yang, Zhou, & Lian, 2015). One notable

study was the study on the comparison of the Chinese PSD at different grade to German PSDs

(Gao, 2015). Other than that, there was no other study on the dynamic responses of train-track

system due to track irregularity under Chinese PSDs existing.

In conclusion, the random process method was applied to generate track irregularity profile

and only German PSDs and Chinese PSDs were further discussed in Chapter 6.3 in this thesis.

German and Chinese power spectrum would be used to verify the numerical model while

Chinese power spectrum would be used for further analysis of dynamic responses of train-track

35
system. As vertical dynamic responses is the only interest here, only vertical irregularity profile

of track would be generated.

6.2 Types of track irregularities

There are five types of track irregularities in total as shown in Figure 6.1 according to a study

done by Ling, Deng, Guan, and Jin (2017).

Figure 6.1: Illustration of track irregularity types.

The five types of the track irregularity were described as following sections (Tian, n.d.).

36
6.2.1 Vertical (longitudinal level) irregularity

The vertical irregularities of the track are the differences of the elevation level of the rail surface

along the track. The reason might be because of construction error, the inconsistency of the

elastic behavior of the discrete pads, CA mortar layer and subgrade, the connections between

components of the track models, soil consolidation, plastic deformation of track slab and

concrete base, etc.

The vertical irregularities of the track are the main reason to cause vertical vibration of the train

and excessive inertial force at the contact point of wheels and rails. The larger the track

irregularity amplitude, changing rate of the amplitude and wavelength of the track irregularities,

the higher the irregularity and the stronger the vibration of the operating train. Short wavelength

with high changing rate of the amplitude will cause the large striking force at the contact point

of wheels and rails.

The vertical irregularity is denoted by the average irregularities of left rail and right rail.

𝑧𝑟 = (𝑧𝐿 + 𝑧𝑅 )/2 (6.1)

Where

𝑧𝑟 represent the overall vertical irregularity of the interested track

𝑧𝐿 represent the vertical irregularity of the left rail on the track

𝑧𝑅 represent the vertical irregularity of the right rail on the track

37
6.2.2 Cross level irregularity

The cross-level irregularity is the level difference between left rail and right rail of the track

along the track, derived from vertical irregularity. Cross level irregularity is the main reason of

train rollover. When the amplitude of level irregularity and occurrence frequency increase and

approaching the natural frequency of train rollover, the transverse vibration of the train body

will occur with higher amplitude. It will result in uneven loading between left and right rail,

potentially lead to derailment, which compromise the safety of high-speed rail operation.

The cross-level irregularity is denoted by the horizontal level difference between left and right

rail.

𝑧𝑐 = 𝑧𝐿 − 𝑧𝑅 (6.2)

Where

𝑧𝑐 represent the cross-level irregularity of the interested track

6.2.3 Rail gauge irregularity

Rail gauge irregularity is the distance inconsistency between two rails along the track. This

type of irregularity will cause the abrasion on the rail surface and wheel, affecting the safety of

high-speed rail operation, even lead to derailment if the distance between two rails are too far.

The rail gauge irregularity is denoted by the difference between the theoretical distance and

actual distance between left and right rail.

𝑌𝑔 = 𝑌𝑡 − 𝑌𝑎𝑡 (6.3)

Where

𝑌𝑔 represent the rail gauge irregularity

𝑌𝑡 and 𝑌𝑎𝑡 represent the theoretical distance and the actual distance between left and right rail.

38
6.2.4 Alignment irregularity

The alignment irregularity is the misalignment of left and right rail, particularly the bending of

the rail along horizontal direction. The reason might be the human error during construction

and maintenance, and the plastic deformation due to high-speed rail operation. The alignment

irregularity might cause the train to move horizontally and produce excessive horizontal force

and torque of rollover, potentially lead to derailment if the derailment coefficient exceeded the

allowable value.

The alignment irregularity is denoted by the error between the center line of theoretical track

and actual track.

𝑌𝑎 = (𝑌𝐿 + 𝑌𝑅 )/2 (6.4)

Where

𝑌𝑎 represent the alignment irregularity of the interest track

𝑌𝐿 and 𝑌𝑅 represent the horizontal distance between actual position of left and right rail to the

center line of theoretical track.

6.2.5 Twist irregularity

The twist irregularity is the rate of change of cross level irregularity within a stretch of track.

For example, the left rail was vertically higher than right rail initially, after a certain distance,

the right rail is now vertically higher than left rail. With twist irregularity, train will prone to

horizontal movement and rollover, even causing the wheel load one of the wheels in the bogie

to be reduced significantly if the twist irregularity condition is severe, potentially lead to

derailment.

39
6.3 Power spectrum of track irregularity

6.3.1 German track irregularity power spectrum

In early 1980s’, Germany has developed three PSD function to address vertical irregularity,

alignment irregularity and cross-level irregularity as shown in Eqs. (6.5) to (6.7) (Xia, Zhang,

& Guo, 2018).

Vertical irregularity:

𝐴𝑣 𝛺𝑐2
𝑆𝑣 (𝛺) = (6.5)
(𝛺 2 + 𝛺𝑟2 )(𝛺 2 + 𝛺𝑐2 )

Alignment irregularity:

𝐴𝑎 𝛺𝑐2
𝑆𝑎 (𝛺) = (6.6)
(𝛺 2 + 𝛺𝑟2 )(𝛺 2 + 𝛺𝑐2 )

Cross-level irregularity:

𝐴𝑣 ∙ 𝑏 −2 ∙ 𝛺𝑐2 ∙ 𝛺 2
𝑆𝑐 (𝛺) =
(𝛺 2 + 𝛺𝑟2 )(𝛺 2 + 𝛺𝑐2 )(𝛺2 + 𝛺𝑠2 ) (6.7)

Where

𝑆𝑣 (𝛺) represents the vertical irregularity PSD function with a unit of 𝑚2 /(𝑟𝑎𝑑/𝑚).

𝑆𝑎 (𝛺) represents the alignment irregularity PSD function with a unit of 𝑚2 /(𝑟𝑎𝑑/𝑚).

𝑆𝑐 (𝛺) represents the cross-level irregularity PSD function with a unit of 1/(𝑟𝑎𝑑/𝑚).

𝛺 represents the spatial frequency of track irregularity with a unit of 𝑟𝑎𝑑/𝑚.

𝛺𝑐 , 𝛺𝑟 , and 𝛺𝑠 represent the cutoff frequency with a unit of 𝑟𝑎𝑑/𝑚.

𝐴𝑣 and 𝐴𝑎 represent the roughness coefficients with a unit of 𝑚2 ∙ (𝑟𝑎𝑑/𝑚).

40
𝑏 represents the half of the rollover distance from left to right of the train, usually is 0.75m

Although the PSD function of rail gauge irregularity was not included by Germany, there is a

guideline where the allowable range of rail gauge irregularity is (-3mm, 3mm). As the rail

gauge irregularity will have a similar expression to cross-level irregularity, an expression to

address rail gauge irregularity was suggested as shown in Eq. (6.8) (Zhai, 2007).

𝐴𝑔 ∙ 𝛺𝑐2 ∙ 𝛺 2
𝑆𝑔 (𝛺) = (6.8)
(𝛺 2 + 𝛺𝑟2 )(𝛺 2 + 𝛺𝑐2 )(𝛺2 + 𝛺𝑠2 )

Where

𝐴𝑔 represent a reference value calculated by using the value within the allowable range of rail

gauge irregularity provided earlier, which is (-3mm, 3mm).

Germany has also suggested two sets of coefficients for two different grade of track, namely

low interference coefficient and high interference coefficient, as shown in Table 6.1.

Table 6.1: Suggested coefficients of German PSDs (Xia, Zhang, & Guo, 2018).

Track grade Low interference High interference

𝛺𝑐 /(𝑟𝑎𝑑/𝑚) 0.8246 0.8246
𝛺𝑟 /(𝑟𝑎𝑑/𝑚) 0.0206 0.0206
𝛺𝑠 /(𝑟𝑎𝑑/𝑚) 0.4380 0.4380
𝐴𝑎 /(𝑚2 ∙ 𝑟𝑎𝑑/𝑚) 2.119 × 10−7 6.125 × 10−7
𝐴𝑣 /(𝑚2 ∙ 𝑟𝑎𝑑/𝑚) 4.032 × 10−7 1.08 × 10−7
𝐴𝑔 /(𝑚2 ∙ 𝑟𝑎𝑑/𝑚) 5.32 × 10−7 1.032 × 10−7

As the vertical irregularity is the only interest, only the German low interference vertical

irregularity PSD and high interference vertical irregularity PSD were presented. However, the

units of German vertical irregularity PSD 𝑆𝑣 (𝛺) was 𝑚2 /(𝑟𝑎𝑑/𝑚),and the unit of the

inputting spatial angular frequency value 𝛺 was 𝑟𝑎𝑑/𝑚. In order to standardize the unit, the

German vertical irregularity PSD 𝑆𝑣 (𝛺) was converted to 𝑆𝑣 (𝑓) with a unit of 𝑚𝑚2 /(1/𝑚),

and the inputting spatial angular frequency value 𝛺 was converted to spatial frequency 𝑓 with

41
a unit of 1/𝑚. The detailed calculation was shown from Eq. (0.9) to (0.12) by adopting

similar conversion method from previous study (Chen, & Zhai, 1999).

The conversion equations were shown in Eq. (6.9) and (6.10).

𝑆𝑣 (𝑓) = 106 ∙ 2𝜋 ∙ 𝑆𝑣 (𝛺) (6.9)

𝛺 = 2𝜋𝑓 (6.10)

Where

𝑓 represent time frequency.

By substituting Eq. (0.10) into Eq. (6.1), the Eq. (0.11) was formed.

𝐴𝑣 𝛺𝑐2
𝑆𝑣 (𝛺) =
(𝛺 2 + 𝛺𝑟2 )(𝛺2 + 𝛺𝑐2 )

𝐴
(2𝜋 ∙ 2𝜋𝑣 ) (2𝜋𝑓𝑐 )2
= (6.11)
[(2𝜋𝑓)2 + (2𝜋𝑓𝑟 )2 ][(2𝜋𝑓)2 + (2𝜋𝑓𝑐 )2 ]

𝐴
(2𝜋𝑣 ) 𝑓𝑐2
=
2𝜋(𝑓 2 + 𝑓𝑟2 )(𝑓 2 + 𝑓𝑐2 )

By substituting Eq. (6.9) into Eq. (0.11), the converted German vertical irregularity PSD was

produced as shown in Eq. (6.12).

𝐴
(2𝜋𝑣 ) 𝑓𝑐2 ∙ 106
𝑆𝑣 (𝑓) = (6.12)
(𝑓 2 + 𝑓𝑟2 )(𝑓 2 + 𝑓𝑐2 )

By using Eq. (6.12), the German vertical irregularity PSD for low interference track and high

interference track was provided as shown in Figure 6.2 and Figure 6.3.

42
Figure 6.2: German Low Interference PSD.

Figure 6.3: German High Interference PSD.

43
6.3.2 Chinese track irregularity power spectrum

The Chinese track irregularity spectra of high-speed rail slab track was officially released

with various percentiles of irregularities (China National Railway Administration, 2014). The

range of the spatial frequency for the spectra is (0.005(1/m), 0.5(1/m)), which corresponding

to a range from 2m to 200m in wavelength. The PSD function of all types of track

irregularities is expressed with a segmentally fitted power function as Eq. (6.13).

𝐴
𝑆(𝑓) = (6.13)
𝑓𝑘

Where

𝑓 represents the spatial frequency

𝐴 and 𝑘 represent the coefficient which defers in different segment of spatial frequency

corresponding to the average percentile of track irregularity spectra, which is 63.2%. They

could be obtained from Table 6.2 and

Table 6.3.

Table 6.2: Suggested coefficient for Chinese PSD

Irregularity type Rail Gauge Cross-level Alignment Vertical

1st Segment 𝐴 5.4978 × 10−2 3.6148 × 10−3 3.9513 × 10−3 1.0544 × 10−5
𝑘 0.8282 1.7278 1.8670 3.3891
2nd Segment 𝐴 5.0701 × 10−3 4.3685 × 10−2 1.1047 × 10−2 3.5588 × 10−3
𝑘 1.9037 1.0461 1.5354 1.9271
rd
3 Segment 𝐴 1.8778 × 10−4 4.5867 × 10−3 7.5653 × 10−4 1.9784 × 10−2
𝑘 4.5948 2.0939 2.8171 1.3643
4th Segment 𝐴 − − − 3.9488 × 10−4
𝑘 − − − 3.4516

Table 6.3: Segmentation points for Chinese PSD.

Segmentation point Rail Gauge Cross-level Alignment Vertical

Between 1st & 2nd Spatial frequency 0.1090 0.0258 0.0450 0.0187
Segment Spatial wavelength 9.2 38.8 22.2 53.5
Between 2nd & 3rd Spatial frequency 0.2938 0.1163 0.1234 0.0474
Segment Spatial wavelength 3.4 8.6 8.1 21.1
44
 Between 3rd & 4th Spatial frequency − − − 0.1533
Segment Spatial wavelength − − − 6.5
2
Based on assumption that the irregularity spectra satisfied the χ distribution with a degree of

freedom of 2, the coefficient of PSD spectrum to different percentile of track irregularity was

shown in Table 6.4.

Table 6.4: Conversion factor corresponding to Chinese PSD percentile.

Percentile Conversion factor

10.0 0.105
20.0 0.223
25.0 0.288
30.0 0.357
50.0 0.693
60.0 0.916
63.2 1.000
70.0 1.204
75.0 1.386
80.0 1.609
90.0 2.303

As the only concern here is vertical irregularity, the Chinese vertical irregularity PSD at 25%,

63.2% and 90% were shown in Figure 6.4.

45
 Figure 6.4: Chinese PSD at 25%, 63.2% and 90%.

6.3.3 Comparison between German and Chinese PSD

Figure 6.5: Comparison between German PSDs and Chinese PSDs

46
As shown in Figure 6.5, the Chinese PSD were better than German PSD as the Track

irregularity of Chinese track was less than German in terms of vertical irregularity, which

resulted in higher safety and comfort level of passengers under less irregularity excitation.

6.4 Numerical simulation method of track irregularities random

profile

There are mainly 4 numerical simulation method mentioned adopted practically around the

globe to obtain track irregularities profile from power spectra, namely white noise filtering

method, secondary filtering method, triangular series method and inverse fast Fourier

transformation (Tian, n.d.).

6.4.1 White noise filtering method and quadratic filtering method

White noise filtering method is assuming the random process of track irregularity to be a

white noise which fulfilled some conditions and consider it as a linear system input.

Therefore, the output of the method will be a random process sample that fulfilling the

condition of PSD function. Quadratic filtering method is designing wave filter according to

specific power spectrum, filtering the white noise for two times in order to produce a

randomized sample that fitted the power spectrum. The fitting accuracy of the two methods

are related the design of wave filter. In order to improve the accuracy of the output, one paper

suggested a calculation method into wave filter design (Qian, 2000). The detail calculation

method was as followed.

1. A random sequence 𝑢𝑖 was produced and filled with values that were evenly

contributed in the range of (0, 1).

2. Transform 𝑢𝑖 to a normal distributed random sequence 𝑥𝑖 with average value of 0 and

mean-square value of 𝜎.

47
 𝑥𝑖 = 𝜎√− ln 𝑢𝑖 cos(2𝜋𝑢𝑖+1 ) (6.14)

𝑥𝑖+1 = 𝜎√− ln 𝑢𝑖 sin(2𝜋𝑢𝑖+1 ) (6.15)

Where

𝑖 = 1, 3, 5, ⋯ , 𝑛 − 1

𝜋
𝜎 = 2 √2𝑘𝐴𝑣 𝑣∆𝑡

∆𝑡 represents the time sampling interval

3. Random process sampling function 𝜂𝑖 could be produced from 𝑥𝑖 through the filtering

process as shown in Eq. (6.16).

𝜂𝑖 = (𝑎 + 1)𝜂𝑖+1 − 𝑎𝜂𝑖+2 + (1 − 𝑎)𝑥𝑖 (6.16)

Where

𝑎 = e−𝜔𝑐Δ𝑡

As the frequency of the track irregularity randomized sample produced by method above are

wide in range, the sample has to be processed with bandwidth filter to simulate the practical

range of instrumentation result.

6.4.2 Trigonometric series method

By assuming track irregularity to be ergodic random process 𝑥(𝑡), the PSD function is

assumed to be as shown in Eq. (6.17) and (6.18) (Tian, n.d.).

𝐹𝑥 (𝜔, 𝑇) = ∫ 𝑥(𝑡)𝑒 −𝑖𝜔𝑡 𝑑𝑡 (6.17)

0

1
𝑆𝑥 (𝜔) = lim |𝐹𝑥 (𝜔, 𝑇)|2 (6.18)
𝑇→∞ 𝑇

48
The track irregularity random process could be expressed in complex form of Fourier series

as shown in Eq. (6.19) and (6.20).

𝑁−1

𝑥(𝑡) = ∑ 𝐶𝑘 𝑒 −𝑖𝑘𝜔0𝑡 (6.19)

𝑘=0

𝑇
1
𝐶𝑘 = ∫ 𝑥(𝑡)𝑒 −𝑖𝑘𝜔0 𝑡 𝑑𝑡 (6.20)
𝑇
0

The discrete expressions were shown as shown in Eq. (6.21) and (6.22).

𝑁−1
2𝑛𝑘𝜋
𝑥(𝑁∆𝑡) = ∑ 𝐶𝑘 𝑒 −𝑖 𝑁 (6.21)
𝑘=0

𝑁−1
1 2𝑛𝑘𝜋
𝐶𝑘 = ∑ 𝑥(𝑁∆𝑡)𝑒 −𝑖 𝑁 (0.22)
𝑁
𝑛=0

From Eq. (0.17) to (0.20), Eq. (0.23) was deduced

𝑇
1 1
𝐶𝑘 = ∫ 𝑥(𝑡)𝑒 −𝑖𝑘𝜔0𝑡 𝑑𝑡 = 𝐹𝑥 (𝑘𝜔0 , 𝑇) (0.23)
𝑇 𝑇
0

By substituting Eq. (6.21) and (6.22) into Eq. (6.17) and (6.18), Eq. (6.24) and Eq. (6.25).

1
𝑆𝑥 (𝑘𝜔0 ) = lim |𝐹𝑥 (𝑘𝜔0 , 𝑇)|2 = lim 𝑇|𝐶𝑘 |2 (6.24)
𝑇→∞ 𝑇 𝑇→∞

𝑆𝑥 (𝑘𝜔0 )
|𝐶𝑘 | = lim √ (6.25)
𝑇→∞ 𝑇

Trigonometric series method assumes track irregularity as an ergodic Gaussian process, as

express it as triangular series. The following steps elaborate the concept and steps of

triangular series method.

49
Expressing track irregularity into complex form of Fourier series as shown in Eq. (6.26)

𝑥(𝑡) = ∑ 𝐶𝑗 𝑒 𝑖𝑗𝜔0𝑡
𝑗=−𝑘

𝑇 (6.26)
2
1
𝐶𝑗 = ∫ 𝑥(𝑡)𝑒 −𝑖𝑗𝜔0𝑡 𝑑𝑡
𝑇
𝑇
−
2

As 𝐶𝑗 = 𝐶−𝑗 and 𝐶0 = 0, therefore the expression could be expressed as shown in Eq. (6.27).

𝑥(𝑡) = 2 ∑|𝐶𝑘 | cos(𝑗𝜔0 𝑡 + 𝜃𝑗 )

𝑗=1

𝑇 (6.27)
2
1
𝐶𝑗 = ∫ 𝑥(𝑡)𝑒 −𝑖𝑗𝜔0𝑡 𝑑𝑡
𝑇
𝑇
−
2

In order to simulate the random behavior of coefficient 𝐶𝑗 , a random variable 𝜃𝑗 was

introduced and the value are evenly distributed from 0 to 2π. Therefore, the triangular series

expression of track irregularity random process is deduced as shown in Eq. (6.28).

𝑘 𝑗𝜋
√2𝑆𝑥 (𝑘∆𝑡) 𝑗𝑛𝜋 (6.28)
𝑥(𝑛∆𝑇) = ∑ cos ( + 𝜃𝑗 )
𝑘∆𝑡 𝑘
𝑗=1

Where

𝜔0 = 𝜋/(𝑘∆𝑡)

As the calculation method between trigonometric series method and fast Fourier

transformation are different, there will be error between the power spectra of sampling

produced by triangular series method and analytical power spectra.

50
6.4.3 Inverse Fast Fourier Transformation (IFFT)

IFFT method was based on the concept of fast algorithm of power spectra, and also applying

the inverse Fourier transformation, to significantly shorten the simulation duration than

triangular series method (Tian, n.d.). Quadratic filtering method required design of wave

filter according to the specific track irregularity PSD function, therefore it is not efficient and

practical to design wave filter for every unique track irregularity PSD function.

Trigonometric series method and white noise filtering method are assuming track irregularity

as a stationary Gaussian random process, which was considered by a study that not all track

irregularities are stationary random process (Luo, 1982). In fact, the power spectral of track

irregularity was produced by periodic graph method based on signal sample from time

domain, the main calculation method was fast Fourier transformation. With the imperfection

of the first three method introduced, inverse fast Fourier transformation appears to have the

least calculation error, shorter in calculation duration and higher versatility. Therefore, IFFT

method was adopted in this paper. The calculation procedure was suggested by Chen (Chen,

& Zhai, 1999).

6.4.3.1 Estimation of power spectra by adopting Blackman-Turkey (BT) method

An introduction of Blackman-Turkey method was necessary before introducing IFFT

method.

𝑇
1
𝑅𝑥𝑥 (𝜏) = ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡
𝑇
0
(6.29)
𝑁−1
1
𝑅𝑟 = 𝑅𝑥𝑥 (𝜏 = 𝑟∆) = ∑ 𝑥𝑠 𝑥𝑠+𝑟
𝑁
𝑠=0

𝑥𝑠 ≡ 𝑥(𝑠∆), where 𝑥𝑠 represents a time sequence where 𝑠 = 0, 1, ⋯ , (𝑁 − 1)

𝑇 = 𝑁∆, where 𝑇 represents the recorded length and ∆ represents time interval

51
𝜏 = 𝑟∆, where 𝜏 represents the latency to correlation function, which is also discrete value,

and 𝑟 = 0, 1, ⋯ , (𝑁 − 1)

𝑅𝑥𝑥 and 𝑅𝑟 represent the autocorrelation function of time sequence 𝑥𝑠 .

Therefore,

𝑁−1
1 1 1 2𝜋
𝑆𝑥𝑥 (𝑓) ⟺ 𝑅𝑥𝑥 (𝜏) ⟹ 𝑆𝑥𝑥 (𝑘) = 𝑆𝑥𝑥 (𝑓 = 𝑘 ) = ∑ 𝑅𝑟 𝑒 −𝑖𝑘 𝑁 𝑟 (6.30)
𝑇 𝑇 𝑁
𝑟=0

Where

𝑆𝑥𝑥 (𝑘) represents the PSD function of time sequence 𝑥𝑠

Hence,

𝑁−1 𝑁−1
1 1 2𝜋
𝑆𝑥𝑥 (𝑘) = ∑ { ∑ 𝑥𝑠 𝑥𝑠+𝑟 } 𝑒 −𝑖𝑘 𝑁 𝑟
𝑁 𝑁
𝑟=0 𝑠=0
(6.31)
𝑁−1 𝑁−1
1 2𝜋 1 2𝜋
(𝑠+𝑟)
= { ∑ 𝑥𝑠 𝑒 𝑖𝑘 𝑁 𝑠 } { ∑ 𝑥𝑠+𝑟 𝑒 −𝑖𝑘 𝑁 }
𝑁 𝑁
𝑠=0 𝑟=0

By establishing 𝑗 = 𝑟 + 𝑠,

𝑁−1 𝑁−1+𝑠
2𝜋 2𝜋
−𝑖𝑘 (𝑠+𝑟)
∑ 𝑥𝑠+𝑟 𝑒 𝑁 = ∑ 𝑥𝑗 𝑒 −𝑖𝑘 𝑁 𝑗 (6.32)
𝑟=0 𝑗=𝑠

However, the time sequence 𝑥𝑠 was discretely periodized with a periodic period of 𝑁 due to

the discrete Fourier transformation. Thus,

𝑁−1+𝑠 𝑁−1
2𝜋 2𝜋
−𝑖𝑘 𝑗
∑ 𝑥𝑗 𝑒 𝑁 = ∑ 𝑥𝑗 𝑒 −𝑖𝑘 𝑁 𝑗 (6.33)
𝑗=𝑠 𝑗=0

By substituting (6.32) and (6.33) into (6.31),

52
 𝑁−1 𝑁−1
1 2𝜋 1 2𝜋 1
𝑆𝑥𝑥 (𝑘) = { ∑ 𝑥𝑠 𝑒 𝑖𝑘 𝑁 𝑠 } { ∑ 𝑥𝑗 𝑒 −𝑖𝑘 𝑁 𝑗 } = |𝐷𝐹𝑇[𝑥𝑠 ]|2
𝑁 𝑁 𝑁
𝑠=0 𝑗=0
(6.34)
1
= [𝑋 ∗ (𝑘)𝑋(𝑘)]
𝑁2

Where

𝑋(𝑘) represent the spectral density of time sequence 𝑥𝑠 , where 𝑠 = 0, 1, ⋯ , (𝑁 − 1), 𝑘 =

0, 1, ⋯ , (𝑁 − 1).

6.4.3.2 Calculation procedures of inverse fast Fourier transformation

As the existing track irregularity power spectrum are all spatial domain spectrum, which

brought inconvenience in numerical analysis of track irregularity sample, the spatial domain

spectrum should be converted to time domain spectrum. Furthermore, the relationship

between discrete sampling values of PSD and the spectral density of the signal could be

established by the periodic graph method that estimates PSD. Therefore, the track irregularity

excitation function 𝑥(𝑡) under time domain could be obtained if the spectrum density 𝑋(𝑘)

could be obtained by discretely sampling from PSD function, followed by IFFT on 𝑋(𝑘). The

detailed procedure was showed as followed.

1. As the existing track irregularity PSDs are all spatial domain spectrum, it is necessary

to convert it into time domain spectrum. The first step will be standardizing the unit of

PSD and inputting spatial frequency to be 𝑚𝑚2 /(1/𝑚) and 1/𝑚 respectively. This

could be achieved by applying similar calculation method adopted from Eqs. (6.9) to

(6.12). Furthermore, the mean square value of spatial domain PSD or time domain

PSD will be same within the spectrum bandwidth. Therefore, Eq. (6.35) was

suggested by Chen by taking vertical irregularity PSD function as an example (Chen,

Ling, Zhu, Xu, & Liu, 2012).

53
 𝑆𝑣 (𝑓)d𝑓 = 𝑆𝑣𝑡 (𝑓𝑡 )d𝑓𝑡 (6.35)

Where

𝑆𝑣 (𝑓) represents the spatial domain PSD.

𝑆𝑣𝑡 (𝑓𝑣𝑡 ) represents the converted time domain PSD.

𝑓 represents the spatial frequency.

𝑓𝑡 represent the time frequency.

The conversion equation from spatial frequency to time frequency was shown in Eq.

(6.36).

𝑓𝑡
𝑓= (6.36)
𝑉

Where

𝑉 represents the velocity of operating train.

Eq. (6.37) was derived from Eq. (6.35) and Eq. (6.36) as shown.

d𝑓 𝑓𝑡 1
𝑆𝑣𝑡 (𝑓𝑡 ) = 𝑆𝑣 (𝑓𝑣 ) ∙ = 𝑆𝑣 ( ) ∙ (6.37)
d𝑓𝑡 𝑉 𝑉

Hence, the time domain PSD was produced by Eq. (0.37).

2. As the time-domain PSD is a unilateral spectrum, it must be converted in to a bilateral

spectrum 𝑆𝑥 (𝑓) as shown in Figure 6.6.

Figure 6.6: Illustration of bilateral spectrum (Chen, & Zhai, 1999)

54
 By assuming 𝑇𝑠 as a time sequence in simulation and ∆ as the time interval, the

samples of time domain and spatial domain were defined as 𝑁𝑟 = 𝑇𝑠 /∆. As the

common value of 𝑁𝑟 was not exactly equal power of 2 with an integer exponent, it

must be converted to a number happened to be equal to power of 2 with an integer

exponent to maintain the short calculation duration of fast Fourier series. The

sampling interval was defined as ∆𝑓 = 1/(𝑁𝑟 ∆). As a known fact, the PSD estimated

by using periodical graph method will behave periodically and it will be an even

function. As shown in the figure, upper bound and lower bound of cutoff frequency

are defined as 𝑓𝑢 and 𝑓𝑙 respectively, the samples 𝑁𝑓 within effective range of

frequency was equal to (𝑓𝑢 − 𝑓𝑙 )/∆𝑓. By assuming 𝑁0 = 𝑓1 /∆𝑓, the samples where

the corresponding frequencies are within 0 to (𝑁0 − 1) were considered as 0. If 𝑁0 +

𝑁𝑓 < 𝑁𝑟 /2, the samples where the corresponding frequencies are within 𝑁𝑓 and 𝑁𝑟 /2

were considered as 0. If 𝑁0 + 𝑁𝑓 > 𝑁𝑟 /2, the 𝑇𝑠 should be increased to fulfil the

condition 𝑁0 + 𝑁𝑓 < 𝑁𝑟 /2. Therefore, 𝑁𝑟 /2 discrete samples 𝑆𝑥 (𝑓 = 𝑘∆𝑓)were

collected from PSD 𝑆𝑥 (𝑓), where 𝑘 = 0, 1, ⋯ , 𝑁𝑟 /2. Finally, the samples will be

mirrored by vertical line 𝑥 = 𝑁𝑟 /2 to obtained an even function 𝑆𝑥 (𝑓 = 𝑘∆𝑓), where

𝑘 = 0, 1, ⋯ , 𝑁𝑟 /2.

3. Simulated value of time domain sequence could be derived by using Eq. (6.30) and

Eq. (6.34) as shown as Eq. (6.38) (Chen, & Zhai, 1999).

|𝑋(𝑘)| = |𝐷𝐹𝑇[𝑥(𝑛)]| = √𝑁𝑟2 × 𝑆𝑘 (𝑘) = 𝑁𝑟 √𝑆𝑘 (𝑘)

(6.38)
= 𝑁𝑟 √𝑆𝑘 (𝑘∆𝑓)∆𝑓

Where

𝑘 = 0, 1, ⋯ , 𝑁𝑟 − 1

55
4. Eq. (0.38) assigned simulated value 𝑋(𝑘) to sequence 𝑥(𝑛). As the sequence 𝑥(𝑛)

was a random process, the phase spectrum of the sequence should behave randomly.

Therefore, an independent phase series 𝜉𝑛 was assumed that the weighted average to

be 0. Furthermore, as a complex sequence will be produced by applying FFT on a real

sequence, 𝜉𝑛 should be a complex number and |𝜉𝑛 | = 1. Therefore, Eq. (6.39) was

assumed as shown.

𝜉𝑛 = cos Φ𝑛 + sin Φ𝑛 = 𝑒 𝑖Φ𝑛 (6.39)

Where

Φ𝑛 follows normal distribution within 0 and 2𝜋.

Nonetheless, the real parts of 𝑋(𝑘) is an even function which is symmetrical with 𝑥 =

𝑁𝑟 /2, but the imaginary parts of 𝑋(𝑘) is an odd function which is symmetrical with

𝑥 = 𝑁𝑟 /2. Hence, by only obtaining 𝑋(𝑘) where 𝑘 = 0, 1, ⋯ , 𝑁𝑟 /2, 𝑋(𝑘), where 𝑘 =

0, 1, ⋯ , 𝑁𝑟 − 1, could be derived from (6.38) as shown as Eq. (6.40).

𝑋(𝑘) = ξ(𝑘)|𝑋(𝑘)| = ξ(𝑘)𝑁𝑟 √𝑆𝑘 (𝑘∆𝑓)∆𝑓 (6.40)

Where

𝑘 = 0, 1, ⋯ , 𝑁𝑟 /2

5. 𝑥(𝑛) could be obtained by applying IFFT on complex series 𝑋(𝑘) as shown in Eq.

(6.41) (Chen, & Zhai, 1999).

𝑁𝑟 −1
1 2𝜋𝑘
𝑖 𝑛
𝑥(𝑛) = ∑ 𝑋(𝑘)𝑒 𝑁𝑟 (6.41)
𝑁
𝑘=0

Where

𝑛 = 0,1, ⋯ , 𝑁𝑟 − 1

56
6.4.4 Validation and discussion

The irregularities samples from German PSDs and Chinese PSDs were generated and the

amplitude were compared with previous study (Gao, 2015). The samples correctness was also

proved by estimating PSDs from the samples and compare with original PSDs. The samples

produced by German PSDs through IFFT were shown in Figure 6.7 and Figure 6.8.

Figure 6.7: Simulated sample from German Figure 6.8: Simulated sample from German
Low Interference PSD High Interference PSD

Figure 6.9 and Figure 6.10 shown the irregularities samples produced by German PSDs in

Gao’s paper.

Figure 6.9: Gao's result from German Low Figure 6.10: Gao's result from German High
Interference PSD (Gao, 2015). Interference PSD (Gao, 2015).

As the samples is a random process, the only comparable value between two result will be the

amplitudes. The amplitude range of irregularity in the simulated results of German low and

high interference PSD were approximately 10mm to -9mm and 15mm to -15mm respectively.
57
The amplitude range of irregularity in the Gao’s results of German low and high interference

PSD were approximately 10mm to -7.5mm and 15mm to -12.5mm respectively, which were

closed to the simulated results.

The samples produced by Chinese PSD at 63.2% and 90% through IFFT were shown in

Figure 6.11 and Figure 6.12.

Figure 6.11: Simulated sample from Figure 6.12: Simulated sample from Chinese
Chinese PSD at 63.2% PSD at 90%

Figure 6.13 and Figure 6.14 shown the irregularities samples produced by Chinese PSDs at

63.2% and 90% in Gao’s paper.

Figure 6.13: Gao's sample from Chinese Figure 6.14: Gao's sample from Chinese
PSD at 63.2% (Gao, 2015). PSD at 90% (Gao, 2015).

The amplitude range of irregularity in the simulated results of Chinese PSDs at 63.2% and

90% were approximately 4.5mm to -3mm and 7mm to -4.5mm respectively. The amplitude

58
range of irregularity in the Gao’s results of Chinese PSDs at 63.2% and 90% were

approximately 4mm to -2.5mm and 6.5mm to -4mm respectively, which were closed to the

simulated results.

Other than comparing the results with Gao’s paper, the simulation samples correctness was

also proved by estimating PSDs from the samples and compare with original PSDs. Figure

6.15 compared the PSDs estimated from the samples simulated from German low and high

interference with the original PSDs. Figure 6.16 compared the PSDs estimated from the

samples simulated from Chinese PSDs at 63.2% and 90% with the original PSDs.

Figure 6.15: Comparison between PSD from Figure 6.16: Comparison between PSD
simulated result and original German PSD from simulated result and original Chinese
PSD

Figure 6.15 and Figure 6.16 showed that the time-domain conversion of the PSD by the IFFT

method produced a highly accurate samples as the PSDs estimated by the samples produced

were similar to the original PSDs.

59
6.5 Conclusion

In this chapter, track irregularity was introduced as a factor that affect the dynamic responses

of the train-track system. The vertical track irregularity was the focus of this thesis among all

the irregularity type introduced. The two mainstream approaches to assume track irregularity

profile were the deterministic function method and random process method. The

deterministic function method is assuming the irregularity to have a sinusoidal or

cosinusoidal profile. The random process method is assuming the irregularity to have

randomized profile which could be generate from the power spectral density (PSD) of the

track. As track irregularity is a random process, German PSD and Chinese PSD were

introduced as a function to generate irregularity profile by applying IFFT other than all the

methods available. The samples generated were then verified with previous study. PSDs were

also generate by using the samples produced to compare with the original PSDs as a second

verification of the numerical method.

60
7 Numerical study

7.1 Introduction

As the track structure is a complex model consisted of 6 components, it is necessary to study

the effect of every component properties on the dynamic response of the system. For

example, the CA mortar layer in the slab track system was located in between track slab and

concrete layer, which increase the elasticity of track slab decrease vibration during operation

and make adjustment to the vertical irregularity of the track together with rail pads.

Therefore, the quality of CA mortar layer is one of the critical parts in slab track system

design and evaluation.

The simulation model used to produce result was the model suggested in Chapter 0 and 5.4

and parameters applied were the parameters suggested in Chapter 5.5.3. Furthermore, the

vertical track irregularity profile was a random process samples produced by Chinese PSD at

63.2% and the samples was produced by using the IFFT method suggested in Chapter 6.4.3.

The simulated train velocity was 200km/h and the simulated distance travelled was 150m in

order to analyze the effect of each parameters in slab track model on the dynamic response of

the system. As there is a random variable involved in the calculation, 3 samples under same

condition where produced and the average values of the outputs were considered effective.

Furthermore, the vertical irregularity generated for every result was based on Chinese PSD at

63.2%.

All results produced were evaluated in terms of Wheel-load reduction rate and the vertical

displacement of the rail, track slab and concrete base.

61
7.2 Result analysis aspect

Two different aspect were considered during the analysis of the result, which were structural

aspect and safety aspect.

7.2.1 Structural aspect: Displacement of rail, track slab and concrete base

By obtaining the vertical displacement profile of rail, track slab and concrete base at every

time steps, the maximum vertical displacement among all time steps were determined and

evaluated as it represented the most critical scenario for rail, track slab and concrete slab. As

there were three samples under the same conditions were simulated, the average value of

maximum vertical displacement from these three samples was evaluated to prevent bias.

For example, Figure 7.1, Figure 7.2 and Figure 7.3 showed the vertical rail displacement

profiles in three samples as mentioned. The average maximum vertical displacement was

calculated by using the maximum vertical displacement values from the three samples.

Figure 7.1: Rail displacement profile Figure 7.2: Rail displacement profile
example from 1st simulation example from 2nd simulation

62
 Figure 7.3: Rail displacement profile example from 3rd simulation

7.2.2 Safety aspect: Wheel-load reduction rate

As suggested in Lei’s published reference book on high-speed railway track dynamics, there

are four evaluation parameters for train safety, namely derailment coefficient, lateral force,

wheel-load reduction rate and overturning coefficient (Lei, 2017). However, derailment

coefficient and overturning coefficient involved lateral force which was not considered in the

simulation. Therefore, the only meaningful evaluation parameter was wheel-load reduction

rate. The wheel load reduction rate equation was shown in Eq. (7.1) (Lei, 2017).

∆𝑃 𝑃1 − 𝑃2
= (7.1)
𝑃̅ 𝑃1 + 𝑃2

Where

∆𝑃 represents the wheel load reduction

𝑃1 and 𝑃2 represent the wheel load at two wheels

𝑃̅ represents the total wheel load of the two wheels

Lei suggested that the primary and secondary upper limit of wheel load reduction rate were

0.65 and 0.60 respectively to prevent overturning.

63
The wheel load reduction rate was determined to check whether the reduction of wheel load

on one side of the track was too significant that lead to derailment. Therefore, 𝑃1 and 𝑃2 were

considered as the average maximum and minimum contact force at the wheel-rail contact

point from all simulations which were under same conditions as shown in Eq. (7.2) and (7.3)

respectively.

𝑛
1
𝑃1 = ∑ 𝑝𝑚𝑎𝑥,𝑛 (7.2)
𝑛
𝑁=1

𝑛
1
𝑃2 = ∑ 𝑝𝑚𝑖𝑛,𝑛 (7.3)
𝑛
𝑁=1

Where

𝑝𝑚𝑎𝑥,𝑛 and 𝑝𝑚𝑖𝑛,𝑛 represent the maximum and minimum contact force obtained from the

contact point profile in 𝑛𝑡ℎ simulation respectively.

For example, three contact force profiles were simulated as shown in Figure 7.1, Figure 7.2

and Figure 7.3 with maximum and minimum contact force marked with red marker and blue

marker respectively. Table 7.1 showed the value of all maximum and minimum value of

contact force profile.

Figure 7.4: Wheel-rail contact force profile Figure 7.5: Wheel-rail contact force profile
from 1st simulation. from 2nd simulation

64
 Figure 7.6: Wheel-rail contact force profile from 3rd simulation

Table 7.1: Maximum and minimum contact force examples under different simulation

1st simulation 2nd simulation 3rd

simulation
Maximum contact force 𝑝𝑚𝑎𝑥,𝑛 (𝑘𝑁) 74.6642 74.1980 75.1798
Minimum contact force 𝑝𝑚𝑖𝑛,𝑛 (𝑘𝑁) 62.5462 64.3431 63.4428

By substituting the maximum and minimum contact force value in table into Eq. (7.2) and

(7.3), 𝑃1 = 74.6807 𝑘𝑁 and 𝑃2 = 63.4440 𝑘𝑁 for this particular scenario were obtained. By

substituting value of 𝑃1 and 𝑃2 into Eq. (7.1), wheel load reduction rate of 0.0814 was

obtained for this scenario.

7.3 Result simulation and discussion

In this part, the numerical models mentioned in Chapter 0 was simulated to have only one

parameter from Chapter 5.5.3 changed while others remain unchanged to analyze the effect

of that particular parameter on the dynamic response of the train-track model, especially

structure integrity and safety.

65
7.3.1 Effects caused by varying stiffness of discrete pads (𝑘1 )

The result for this section was produced by changing the stiffness of the discrete pads and

leaving other coefficients unchanged. The stiffness varies from 10 MN/m to 100 MN/m

with interval of 20 MN/m. Figure 7.7 showed the vertical displacement of rail, track slab and

concrete base behavior under different discrete pads stiffness respectively. Figure 7.8 showed

the effect on wheel load reduction rate due to the change of pads stiffness.

Figure 7.7: Vertical displacements under different Stiffness of pads

Figure 7.8: Wheel load reduction rate under different Stiffness of pads

66
As shown in Figure 7.7, the rail displacement decreased as the stiffness of discrete pads

increased. Since the larger the rail displacement induced by the operation of high-speed rail

will pose higher instability to the train operation and more damages and maintenance need to

the rail, it should be eliminated as much as possible. Without considering the other potential

effects of increasing stiffness of discrete pads such as cost efficiency, it should be as high as

possible to eliminate the rail displacement. However, the possibility of derailment was

increased when the discrete pads stiffness increased as Figure 7.8 showed the increasing trend

of wheel load reduction rate together with discrete pads stiffness. Furthermore, the change of

discrete pads posed insignificant effects on track slab and concrete base displacement

according to Figure 7.7. Hence, the stiffness of the discrete pads should be controlled at a

range to prevent excessive rail displacement and derailment, while maximizing the

performance of the track structure.

67
7.3.2 Effects caused by varying damping coefficient of discrete pads (𝑐1 )

The result for this section was produced by changing the damping coefficient of the discrete

pads and leaving other coefficients unchanged. The damping coefficient varies from

20 kNs/m to 100 kNs/m with interval of 20 kNs/m. Figure 7.9 showed the vertical

displacement of rail, track slab and concrete base behavior under different discrete pads

damping property respectively. Figure 7.10 showed the effect on wheel load reduction rate

due to the change of pads damping property.

Figure 7.9: Vertical displacements under different damping coefficient of pads

68
 Figure 7.10: Wheel load reduction rate under different damping coefficient of pads

As shown in Figure 7.9, the effect of different damping coefficient of pads on the

displacement of rail, track slab and concrete base was insignificant and could be neglected.

Furthermore, Figure 7.10 showed a trend of decreasing in wheel load reduction rate when the

damping coefficient of pads was increasing, with an abnormal data at pads damping

coefficient of 80 kNs/m. This might be due to the random irregularity profile used. Therefore,

more samples and data should be produced under same condition to enhance the accuracy.

7.3.3 Effects caused by varying stiffness of CA mortar layer (𝑘2 )

The result for this section was produced by changing the stiffness of the CA mortar layer and

leaving other coefficients unchanged. The stiffness varies from 500 MN/m to 1300 MN/m

with interval of 200 MN/m. Figure 7.11 showed the vertical displacement of rail, track slab

and concrete base behavior under different CA mortar layer stiffness respectively. Figure

7.12 showed effect on wheel load reduction rate by changing CA mortar layer stiffness.

69
 Figure 7.11: vertical displacements under different Stiffness of CA mortar layer

Figure 7.12: Wheel load reduction rate under different Stiffness of CA mortar layer

As shown in Figure 7.11, the effect of different stiffness of CA mortar layer was

insignificant. The comparison between Figure 7.11 and Figure 7.7 was also showed that the

change of stiffness in CA mortar layer has less influence to the change of stiffness in pads on

70
rail displacement, while displacement of track slab and concrete base remained uninfluenced.

Figure 7.12 showed a zig-zag behavior of wheel-load reduction rate due to the change of

stiffness in CA mortar layer. As the numerical value was considered small, the zig-zag

behavior might be caused by the random vertical irregularity profile.

7.3.4 Effects caused by varying damping coefficient of CA mortar layer (𝑐2 )

The result for this section was produced by changing the damping coefficient of the CA

mortar layer and leaving other coefficients unchanged. The damping coefficient varies from

40 kNs/m to 120 kNs/m with interval of 20 kNs/m. Figure 7.13 showed the vertical

displacement of rail, track slab and concrete base behavior under different CA mortar layer

damping property respectively. Figure 7.14 showed the effect on wheel load reduction rate

due to the change of CA mortar layer damping property.

Figure 7.13: Vertical displacements under different damping coefficient of CA mortar layer

71
 Figure 7.14: Wheel load reduction rate under different damping coefficient of CA mortar
layer

As shown in Figure 7.13, the effect of different damping coefficient of CA mortar layer on

the displacement of rail, track slab and concrete base was insignificant and could be

neglected same as the effect of different damping coefficient of pads. Besides, Figure 7.14

also showed a zig-zag behavior of wheel-load reduction rate due to the change of damping

coefficient in CA mortar layer. Same assumption made for Figure 7.12 was also applied here.

7.3.5 Effects caused by varying stiffness of subgrade (𝑘3 )

The result for this section was produced by changing the stiffness of the subgrade and leaving

other coefficients unchanged. The stiffness varies from 20 MN/m to 100 MN/m with

interval of 20 MN/m. Figure 7.15 showed the vertical displacement of rail, track slab and

concrete base behavior under different subgrade stiffness respectively. Figure 7.16 showed

the effect on wheel load reduction rate due to the change of subgrade stiffness.

72
 Figure 7.15: Vertical displacements under different Stiffness of the subgrade

Figure 7.16: Wheel load reduction rate under different Stiffness of the subgrade

As shown in Figure 7.15, the displacement of rail, track slab and concrete base were all

decreased when the stiffness of subgrade increased. This can be explained as the subgrade

was the layer supporting all the structure above, so the change of its structure properties

73
might cause the displacement of the superstructure to be affected. Therefore, it should be

suggested that the stiffness of the subgrade should be as high as possible with only

considering the vertical displacement of rail, track slab and concrete base. Figure 7.16 also

showed a zig-zag behavior of wheel-load reduction rate due to the change of stiffness in

subgrade. Therefore, same assumptions made for Figure 7.12 and Figure 7.14 were applied.

7.3.6 Effects caused by varying damping coefficient of subgrade (𝑐3 )

The result for this section was produced by changing the damping coefficient of the subgrade

and leaving other coefficients unchanged. The damping coefficient varies from 50 kNs/m to

130 kNs/m with interval of 20 kNs/m. Figure 7.17 showed the vertical displacement of rail,

track slab and concrete base behavior under different subgrade damping property

respectively. Figure 7.18 showed the effect on wheel load reduction rate due to the change of

subgrade damping property.

Figure 7.17: Vertical displacements under different damping coefficient of the subgrade

74
 Figure 7.18: Wheel load reduction rate under different damping coefficient of the subgrade

As shown in Figure 7.17, the effect of different damping coefficient of subgrade on the

displacement of rail, track slab and concrete base was also insignificant and could be

neglected. Furthermore, Figure 7.18 also showed a zig-zag behavior of wheel-load reduction

rate due to the change of damping coefficient in subgrade. Same assumption was made for

Figure 7.12, Figure 7.14 and Figure 7.16 was also applied here.

7.4 Conclusion

In this chapter, the same train-track properties suggested in Chapter 5 was applied in this

numerical analysis. Furthermore, assumptions of 200km/h as the operating train speed and

150m as the travel distance were made as well. One of the aims of the numerical analysis was

to study the effect of the structural properties of pads, CA mortar layer and subgrade to the

displacement of rail, track slab and concrete base. The wheel load reduction rate under

different structural properties of track was the other aim of this study as well. As three results

75
were generated under the same structural properties for every scenario, the average value of

displacement from three samples was present while the maximum and minimum wheel load

were used to calculate the wheel load reduction rate.

The results showed that the change in subgrade stiffness was the most critical scenario in

Chapter 7.3.5 as it affects the displacement of all three layers interested which were rail, track

slab and concrete base. Therefore, the stiffness of the subgrade should be as high as possible

to eliminate vertical displacement of the entire track structure. The second critical scenario

was the change in pads stiffness in Chapter 7.3.1 as it affected the vertical rail displacement

which was closely related to the stability of the train operation. As the increase of the pads

stiffness will lead to a higher wheel load reduction rate, the stiffness of pads should not be as

high as possible to prevent high wheel load reduction rate that leads to derailment. The

change of CA mortar stiffness, pads damping coefficient, CA mortar layer damping

coefficient and subgrade damping coefficient were not significantly affecting the

displacement of rail, track slab and concrete base. Besides, the wheel load reduction rate from

most of the cases were not following an obvious trend during the change of structural

properties of the track except change of pads stiffness and damping coefficient. Therefore, it

may be due to vertical random irregularity sample of the track was different under the same

scenario and it could be assumed as bias results due to the small samples size of 3.

76
8 Conclusion of the thesis
This thesis has mentioned the application of moving element method on analysis of high-

speed rail train-track system model with discrete pads. The track irregularity was also

introduced as a factor affecting the wheel-rail contact model and affecting the dynamic

responses of entire system. The focus of this thesis was on the effect of track properties such

as stiffness and damping coefficient of intermediate supporting layer to the dynamic

responses of the train-track system with discrete pads.

From the findings in Chapter 7, it can be suggested that the stiffness of the concrete base is

the most critical properties of the model follow by stiffness of discrete pads and CA mortar

layer while the damping properties of pads, CA mortar layer and concrete base were less

important.

As the result obtained was based on a constant train operating speed of 200km/h, the effect of

different properties of track component under different operating train speed on the dynamic

responses of train-track system could be further studied. Furthermore, since only three

samples were generated for each scenario analyzed in this thesis, the number of samples

could be increased to further enhance the accuracy of the result outcomes. Another possible

research direction will be studying the dynamic responses based on the track irregularity

profile generated from different PSD as this thesis applied Chinese PSD at 63.2% to generate

irregularity profile.

77
9 References
Ang, K. K., & Dai, J. (2013). Response analysis of high-speed rail system accounting for

abrupt change of foundation stiffness. Journal of Sound and Vibration, 332(12),

2954–2970. doi: 10.1016/j.jsv.2013.01.005

Azizi, N., Saadatpour, M., & Mahzoon, M. (2012). Using spectral element method for

analyzing continuous beams and bridges subjected to a moving load. Applied

Mathematical Modelling, 36(8), 3580–3592. doi: 10.1016/j.apm.2011.10.019

Blanco, B., Alonso, A., Kari, L., Gil-Negrete, N., & Giménez, J. G. (2018). Implementation

of Timoshenko element local deflection for vertical track modelling. Vehicle System

Dynamics, 57(10), 1421–1444. doi: 10.1080/00423114.2018.1513538

Cai, C., Cheung, Y., & Chan, H. (1988). Dynamic response of infinite continuous beams

subjected to a moving force—An exact method. Journal of Sound and Vibration,

123(3), 461–472. doi: 10.1016/s0022-460x(88)80163-9

Chen, G., & Zhai, W. (1999). Tie Lu Gui Dao Bu Ping Shun Sui Ji Guo Cheng De Shu Zhi

Mo Ni (Numerical simulation of the stochastic process of railway track irregularities).

Journal of Southwest Jiaotong University, 34(2)

Chen, S., Ling, X., Zhu, Z., Xu, X., & Liu, Y. (2012). Analyses of track vertical profile

irregularity spectra. Journal of Earthquake Engineering and Engineering Vibration,

32(5).

China National Railway Administration. (2014). PSD of ballastless track irregularities of

high-speed railway.

78
Dai, J., Ang, K. K., Jiang, D., Luong, V. H., & Tran, M. T. (2018). Dynamic Response of

High-Speed Train-Track System Due to Unsupported Sleepers. International Journal

of Structural Stability and Dynamics, 18(10), 1850122. doi:

10.1142/s0219455418501225

Dai, J., Ang, K. K., Tran, M. T., Luong, V. H., & Jiang, D. (2017). Moving element analysis

of discretely supported high-speed rail systems. Proceedings of the Institution of

Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 232(3), 783–797.

doi: 10.1177/0954409717693147

Esmaeili, M., Mohammadzadeh, S., & Mehrali, M. (2016). Dynamic Response of the

Coupled Vehicle-Floating Slab Track System using Finite Element Method.

International Journal of Transportation Engineering, 4(1), 9-26. doi:

10.22119/ijte.2016.33493

Frýba, L. (1999). Vibration of Solids and Structures under Moving Loads. doi:

10.1680/vosasuml.35393

Gao, J. (2015). Comparative analysis of power spectrum density of ballastless track

irregularities of Chinese hign-speed railway . Journal of Railway Science and

Engineering, 12. doi: 10.19713/j.cnki.43-1423/u.2015.04.001

Gautier, P.-E. (2015). Slab track: Review of existing systems and optimization potentials

including very high speed. Construction and Building Materials, 92, 9–15. doi:

10.1016/j.conbuildmat.2015.03.102

Kargarnovin, M., & Younesian, D. (2004). Dynamics of Timoshenko beams on Pasternak

foundation under moving load. Mechanics Research Communications, 31(6), 713–

723. doi: 10.1016/j.mechrescom.2004.05.002

79
Khajehdezfuly, A. (2019). Effect of rail pad stiffness on the wheel/rail force intensity in a

railway slab track with short-wave irregularity. Proceedings of the Institution of

Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 233(10), 1038–

1049. doi: 10.1177/0954409718825410

Kim, S.-M. (2005). Stability and dynamic response of Rayleigh beam–columns on an elastic

foundation under moving loads of constant amplitude and harmonic variation.

Engineering Structures, 27(6), 869–880. doi: 10.1016/j.engstruct.2005.01.009

Knothe, K., Strzyzakowski, Z., & Willner, K. (1994). Rail Vibrations in the High Frequency

Range. Journal of Sound and Vibration, 169(1), 111–123. doi: 10.1006/jsvi.1994.1009

Koh, C. G., Ong, J. S. Y., Chua, D. K. H., & Feng, J. (2003). Moving element method for

train-track dynamics. International Journal for Numerical Methods in Engineering,

56(11), 1549–1567. doi: 10.1002/nme.624

Lei, X. (2017). High Speed Railway Track Dynamics: Models, Algorithms and Applications.

Singapore: Springer Nature. doi: 10.1007/978-981-10-2039-1

Lei, X., & Noda, N.-A. (2002). Analyses of Dynamic Response of Vehicle And Track

Coupling System With Random Irregularity Of Track Vertical Profile. Journal of

Sound and Vibration, 258(1), 147–165. doi: 10.1006/jsvi.2002.5107

Lei, X., & Wang, J. (2013). Dynamic analysis of the train and slab track coupling system with

finite elements in a moving frame of reference. Journal of Vibration and Control,

20(9), 1301–1317. doi: 10.1177/1077546313480540

Lei, X., & Zhang, B. (2011). Analysis of Dynamic Behavior for Slab Track of High-Speed

Railway Based on Vehicle and Track Elements. Journal of Transportation

Engineering, 137(4), 227–240. doi: 10.1061/(asce)te.1943-5436.0000207

80
Lei, X., Wu, S., & Zhang, B. (2016). Dynamic Analysis of the High Speed Train and Slab

Track Nonlinear Coupling System with the Cross Iteration Algorithm. Journal of

Nonlinear Dynamics, 2016, 1–17. doi: 10.1155/2016/8356160

Lei, Y., Tian, X., Qi, F., Chen, D., & Tian, T. (2016). Vertical Track Irregularity Influence on

the Wheel High-Frequency Vibration in Wheel-Rail System. Mathematical Problems

in Engineering, 2016, 1–13. doi: 10.1155/2016/5082319

Ling, L., Deng, Y., Guan, Q., & Jin, X. (2017). Effect of track irregularities on the dynamic

behavior of a tram vehicle. Journal of Advances in Vehicle Engineering, 3(1), 29–39.

doi: 10.1234/J73

Luo, L. (1982). Gui Dao Sui Ji Gan Rao Han Shu (Track stochastic disturbance function).

Chinese Railway Science, 4.

Qian, X. (2000). Track irregularity simulation in time domain. Journal of the China Railway

Society, 22(4), 94–98.

Reddy, J. (1997). On locking-free shear deformable beam finite elements. Computer Methods

in Applied Mechanics and Engineering, 149(1-4), 113–132. doi: 10.1016/s0045-

7825(97)00075-3

Tian, G. (n.d.). Tie Lu Gui Dao Bu Ping Shun Gong Lu Pu Fen Xi Yu Shu Zhi Mo Ni

(Numerical Analysis and Simulation of Railway Track Irregularity Power Spectral

Density) (Bachelor’s Dissertation, East China Jiaotong Univerisity, Nanchang,

China).

Tian, G., Gao, J., & Zhao, C. (2016). Progress in the Research on the Railway Track

Irregularity Power Spectral Density. Journal of Railway Engineering Society, 33(9),

35–40

81
Wang, J. (n.d.). Moving element method for High-speed Railway ballastless track dynamic

performance analyses (Master’s thesis, East China Jiaotong Univerisity, Nanchang,

China).

Xia, H., Zhang, N., & Guo, W. (2018). Dynamic Interaction of Train-Bridge Systems in

High-Speed Railways. Advances in High-Speed Rail Technology. doi: 10.1007/978-3-

662-54871-4

Xin, T., Wang, P., & Ding, Y. (2019). Effect of Long-Wavelength Track Irregularities on

Vehicle Dynamic Responses. Shock and Vibration, 2019, 1–11. doi:

10.1155/2019/4178065

Xu, Z., Zhai, W., Wang, K., Wang, Q. (2003). Analysis of vehicle-track system vibration:

Comparison between Timoshenko beam and Euler beam track model. Journal of

Earthquake Engineering and Engineering Vibration. 23. 74-79.

Yang, F., Zhai, W., Gao, J., Li, H., Kang, X., & Liu, X. (2014). PSD of ballastless track

irregularities of high-speed railway. SCIENTIA SINICA Technologica, 44(7), 687–

696. doi: 10.1360/n092014-00088

Yang, X., Gu, S., Zhou, S., Yang, J., Zhou, Y., & Lian, S. (2015). Effect of track irregularity

on the dynamic response of a slab track under a high-speed train based on the

composite track element method. Applied Acoustics, 99, 72–84. doi:

10.1016/j.apacoust.2015.05.009

Zhai, W. (2007). Vehicle-track coupled dynamics (3rd ed.). Beijing: Science Press

82