Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden

A DYNAMIC MODEL FOR ANALYSIS OF DAMAGE OF RAILWAY SWITCHES Valéri Markine*, Michael Steenbergen* and Ivan Shevtsov# *

Section of Road and Railway Engineering Faculty of Civil Engineering and Geosciences Delft University of Technology Stevinweg 1, 2628 CN, Delft, The Netherlands #

ProRail Utrecht, The Netherlands e-mail: [email protected]

Abstract A railway switch (turnout) is a very important element of the railway infrastructure. Due to the discontinuity in the rail geometry high dynamic amplification of the wheel loads occurs in the crossing nose. These dynamic forces can severely damage the turnout structure. Especially the high-frequency impact loads (the so-called P1 forces) are responsible for RCF damage on the crossing nose. The RCF damage can be reduced by decreasing the high-frequency dynamic forces in the crossing nose. In the present study the relationship between the elastic properties of the turnout supporting structure (such as the rail pads, under sleeper pads and ballast mats) and the dynamic forces in the crossing nose has been investigated. The relation between the wheel/rail geometry and the dynamic forces in the crossing nose has been investigated as well. The dynamic interaction between the railway vehicle and track structure has been analysed numerically using DARTS_NL software (TU Delft). The performance of the turnout has been assessed using numerical simulations in which a railway vehicle (the ICE locomotive) was running through the turnout at 140 km/h. In this simulation only the vertical dynamic forces in the crossing point have been considered: lateral behaviour was disregarded. The results of the parameter analysis have shown that the wheel/rail geometry has significant effect on the dynamic forces in the switch crossing. The numerical results have demonstrated that by varying the elastic properties of the supporting track structure the forces in the crossing nose can significantly be reduced. It was also shown that by varying substructure elasticity the dynamic forces on other track components such as sleepers and ballast can be reduced as well.

1

INTRODUCTION

Railway switch (turnout) is one of the important elements of the railway infrastructure, which enables trains to be guided from one track to another at a railway junction as shown in Figure 1. In this figure it can clearly be seen that at the location of the crossing nose the rail geometry (the inner rail) is discontinuous. Due to such discontinuities the switches (and especially the crossing nose) experience impact loads from the wheels of passing trains, which initiate various types of damage in the railway switch. Statistical evidence shows that failures in switches and crossings cause major operational disturbances in a railway network.

1

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden

Diverg ing dir ec

tion

Ma in

dir e

cti on

Crossing nose

Figure 1 Railway switch and crossing nose

The train and switch interaction has been investigated in a number of research papers which has been published recently [1]-[5]. In [1] two alternative multibody system models for analysis of the dynamic interaction between a train and a standard turnout were presented and compared. Cyclic deformations in the crossing nose of a railway turnout due to repeated wheel passages using a finite element model were studied in [3]. The effects of high resilience under-sleeper mats on track stability and ground-borne noise in railway switches were investigated in [4]. In [5] the relationship between velocity of a vehicle passing through the turnout and the maximum vertical impact forces in the crossing were analysed. In this paper only damage of the crossing nose is considered. This type of damage usually originates from high frequency impact loads, the so-called P1 forces, caused by the contact point jump on a wheel tread at the transition from the wing rail to the crossing nose. The P1 and P2 forces are typical for the dynamic wheel behaviour in the presence of short-waves irregularities [6]. An example of such behaviour due to a dipped rail joint is schematically shown in Figure 2. Similar forces acting at the transition from the wing rail to the crossing nose where the wheel/rail contact point experiences a jump on the wheel tread. Since the P1 and P2 forces act in different frequency regions (Figure 2) the effect on the components of the track structure that they have is also different. Such cyclic high-frequency impact loads (P1) cause very local plastic deformation and work hardening in the rail, until the material reaches the ratchetting regime with crack initiation and propagation. This process manifests itself as severe rolling contact fatigue (RCF) damage of the crossing nose (Figure 3). This type of rail damage has significant impact on the life span of the switch structure, essentially influenced by service conditions (type of traffic, vehicles speed, traffic frequency, climatic conditions) and track conditions (affected by dynamic forces and track parameters). P = Dynamic amplification Pst 6

P1

5 P2

4 3 2 1

time [ms] 0

2

4

6

8

10 12 14 16 18 20 22

Figure 2 Dynamic wheel force due to dipped rail joint (schematic representation)

Figure 3 RCF damage in crossing nose (1:15)

2

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden To reduce the negative effect of these dynamic loads several elastic elements are present in the railway switch structure such as the rail pads and ballast. Recently, some new elements such as resilient under-base-plate supports on sleepers, under-sleeper pads (USP) and ballast mats have been introduced in the switch design in order to reduce deterioration of the switch structure. It should be noted that due to the variable rail cross-section and variable length of the sleepers the vertical track stiffness (as it is ‘experienced’ by a passing axle) is varying along the switch. The dynamic interaction between train and switch have been analysed numerically. The finite element models of a switch with concrete sleepers on ballast and two modifications of this structure, one with the USP and one with ballast mats installed under the crossing nose have been developed using the house software DARTS_NL (TU Delft). The models are described in Chapter 2. Since the high dynamic wheel forces are to a large extent responsible for the RCF damage of crossing nose, one can try to reduce the RCF damage by reducing the dynamic wheel forces. This means that both mass and elasticity of track elements, as well as their distribution throughout the vertical cross-section, contribute in this stiffness [8], [9]. One of the goals of the research presented here was establishing the relationship between the dynamic wheel forces acting in the crossing nose and the elastic properties of the elements of the switch structure. The effect of the rail geometry on the level of the dynamic forces has been investigated as well. The results of these investigations are presented in Chapter 3.

2

DYNAMIC MODEL OF A SWITCH

The dynamic forces acting on the crossing point are analysed here numerically. The numerical models of the switch structure are created using the house software DARTS_NL developed at TU Delft (The Netherlands). DARTS_NL is specialised Finite Element software for analysis of the dynamic railway vehicle – track interaction wherein various track structures (classical and slab track, viaduct, etc.) can be modelled. The library of railway vehicles implemented in this software consists of a number of conventional and high-speed trains. The software has been successfully used for various railway applications such as optimisation of a slab track [10], [11] identification of dynamic properties of track components [12], assessment of various high-speed track structures [13], [14], analysis of the dynamic forces due to bad welds [9] etc. The main parts of the software and the developed models of the turnouts are described below. 2.1

DARTS_NL software

The full 3-D dynamic analysis of vehicle-track interaction accounting track flexibility can be computationally expensive. In order to reduce the computational efforts numerical models in DARTS_NL are restricted to two dimensions (the vertical and longitudinal directions) and linear material behaviour. A track is modelled using a series of alternating hard and soft layers. Each hard layer consists of Timoshenko beam elements while the elastic layers are represented by distributed spring and damper combinations (Kelvin elements) as shown in Figure 4. Depending on the properties assigned to the finite elements these layers represent various components of the track structure such as rails, rail pads, sleepers, ballast, concrete slabs, subgrade, foundation piles, etc. Figure 5 shows a model of railway vehicle on classical track (with ballast) developed in DARTS_NL. It should be noted that due to symmetry of a track in the vertical plane only half of track (containing one rail) is modelled in DARTS_NL. The external load acting on the model is then also reduced to the half of the vehicle weight (wheel load instead of axle load). In Figure 5 a basic railway vehicle implemented in DARTS_NL is shown as well. Each vehicle is modelled as a mass-spring system that consists of four wheels, two bogies and one car body. The wheels, bogies and car body are modelled as rigid bodies connected to each other by the primary and secondary suspensions (Kelvin elements). The contact forces between the wheels and rail are modelled using the non-linear Hertzian spring with the stiffness [15]:

3

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden

KH 

3

6 E 2 P Rw Rr

(1)

4(1  2 ) 2

where P is the static wheel load Rw and Rr are the radii of the wheel and rail profile (the lateral cross-section) E is the Young modulus of the wheel and rail material  is the Poisson coefficient.

The basic vehicles can be combined in a series so that the dynamic behaviour of a whole train can be simulated. By default, one conventional and two high-speed train models are available in DARTS_NL. Car body

x  Secondary suspension

Bogie

1

Figure 4 Schematic representation of multi-layer track model (DARTS_NL)

2

z Primary suspension Rail pad

Wheelset

Sleeper

Hertzian spring

Ballast

3

4

Figure 5 DARTS_NL model of railway vehicle on classical track (with ballast)

Proper representation of the track geometry in the numerical model is important for realistic simulation of the vehicle-track interaction. Containing both long and short wave irregularities the rail geometry represents the main source of the dynamic excitations in the vehicle-track system. In DARTS_NL the vertical rail geometry can be defined either as a periodic function or as a numeric data profile obtained from measurements. Obviously, the latter rail geometry representation gives more accurate results. The dynamic analysis is performed in the time domain following the concept of the displacement method [16], [17]. The main steps in the numerical procedure are:  

Assembling the mass M , damping C and stiffness K matrixes and the vector of the external forces f Generation of the equations of motion   Cu  Ku  f Mu

(2)

 

 Solution of the equations of motion yielding displacement vector u and acceleration u  by cutting off the high (Optional) filtering of the obtained displacements vector u and acceleration u frequency contributions  Calculation of responses based on the (filtered) displacements u . The responses comprise of  stresses, forces and bending moments in the rail and rail supports  accelerations of the vehicle  dynamic wheel forces  contact forces

Details over the numerical method implemented in DARTS_NL software can be found in [16]. 2.2

Turnout model

In the present study the BWG turnout with the crossing angle 1:9 has been considered (Figure 6). Usually, trains passing a turnout in the main direction have much higher speed than the ones moving in the diverging direction. Therefore, the vertical dynamic wheel forces in the crossing nose in the first case are much higher. Thus, the situation with a train passing the switch only in the main direction has been considered here. Because DARTS_NL uses symmetry of a track only through rail on the inner part of the switch has been modelled.

4

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden

Crossing Figure 6 Switch BWG E54 1:9 Rail The cross-sectional shape of the through rail is varying along the turnout. In order to resist the high dynamic vertical forces the bending stiffness of the rail used in the crossing is much higher. Therefore in the numerical model of the switch the rail with variable cross-section has been used. The rail cross-sections for the crossing have been obtained from the drawings (Figure 7). Outside the crossing the UIC54 rail profile has been used. Sleeper width [m] 2.5 3

6

2 1.5 3

6

1 0.5 3

6

0

Figure 7 Rail cross-sections in crossing

0

50

100

150

200

250 300 Element

350

400

450

500

Figure 8 Sleeper width in switch model

Rail support The length of the sleepers (track lateral direction) used in the crossing varies (Figure 6) and therefore the vertical support stiffness of the through rail is not homogeneous. In order to account for that, the sleepers in the numerical model have also variable length as it is shown in Figure 8. In this figure (and further in this paper) the green rectangular and the vertical red line located on the horizontal axel correspond to the crossing and the crossing nose respectively. Since the equivalent ballast stiffness in the model is calculated automatically proportional to the support area of the sleepers the rail support is not homogeneous. In order to investigate the effect of the rail support elasticity on the dynamic forces in the crossing nose three models of the turnout has been developed, namely   

Turnout on ballast bed Turnout on ballast bed with USP Turnout on ballast bed with ballast mats

These models are schematically shown respectively in Figure 9, Figure 10 and Figure 11.

Ballast (kb, cb)

USP (kusp, cusp)

Rail pad (kp, cp)

USP (musp)

Sleeper (ms)

Rail (E, I, A, r)

Substructure

Figure 9 DRATS_NL model of switch on ballast bed

5

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden Rail (E, I, A, r ) Rail pad (kp, cp) Sleeper (ms) USP (kusp, cusp) USP (musp) Ballast (kb, cb) Substructure

Figure 10 DRATS_NL model of switch on ballast bed with USP

Rail (E, I, A, r ) Rail pad (kp cp) Sleeper (ms) Ballast (kb cb) Ballast (mb) Ballast mats (km cm) Substructure

Figure 11 DRATS_NL model of turnout on ballast bed with ballast mats It should be noted that in this study additional elasticity (softer pads, USP and ballast mats) has been applied only under the crossing. The reason for that was that such an adjustment for the existing turnouts is much easier to implement only under the crossing than over the whole length of the turnout. The total length of each model was approximately 50 m while the length of one element in the model was equal to 0.1 m. In the numerical simulations the model of a locomotive of the ICE train has been used. The axle load of this vehicle is approximately 82 kN. The train velocity in the simulation was 140 km/h. In each simulation the integration time was 0.61 s whereas the integration step was equal to 0.001 s. 2.3

Vertical track geometry

The vertical track geometry to a large extent determines the magnitude of the dynamic forces acting on the switch. As it was mentioned earlier, in the location of the crossing nose the railway wheel passing the switch encounters a discontinuity in the rail geometry. This discontinuity is responsible for the amplification of the dynamic forces acting on the crossing nose, which in turn causes the crossing nose damage. Since in DARTS_NL only the vertical rail geometry is considered, some simplification of the 3-D rail geometry is required. The simplified method to obtain the vertical rail geometry in the crossing point used here is briefly described below. When the wheel passing the crossing the following three stages can be recognised:   

The wheel is travelling on the wing rail approaching the crossing nose as shown in Figure 12a. The contact between wheel and rail is a single-point one. The rail touches the crossing nose and the wheel load is to be transported from the wing rail to the crossing nose (Figure 12b and Figure 12c). On this transition stage the wheel contacts both the wing rail and the nose. The wheel leaves the wing rail and rolls only over the crossing nose followed by the through rail (Figure 12d).

6

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden

a. b. Figure 12 Stages of wheel passing crossing nose process

c.

d.

The smoother the wheel transition from the wing rail to the crossing nose the lower the dynamic amplification of the wheel forces. The geometrical properties of the wheel and rail (wing rail and crossing nose) determine the wheel transition in the crossing. In Figure 13 the longitudinal cross-section of the crossing nose is shown. The height of the nose is gradually increasing from zero to the level of the through rail which is achieved in the position IV in Figure 13. Obviously, the dynamic force is higher when the wheel enters the transition zone in position II than when the contact between wheel and crossing point occurs in the position IV since the vertical jump of the contact point is larger in this case. That is why, in order to reduce the damage on the crossing nose the wing rails with increased height are used in practice. By doing so the beginning of the transition zone is shifted farther from the beginning of the nose. Switch geometry [mm] 0 -2 -4

Figure 13 Crossing nose geometry (longitudinal direction)

-6

Crossing point -8

Crossing -10

0

5

10

15

20 25 30 Distance [m][mm] Switch geometry

35

40

45

50

-7.5 -8 -8.5 -9 -9.5 -10 28.05

Figure 14 Local vertical rail geometry based on visual inspection of switch

28.1 28.15

28.2 28.25 28.3 28.35 Distance [m]

28.4 28.45

28.5

Figure 15 Combined vertical rail geometry of switch used in numerical simulations

Therefore, the vertical rail geometry in the crossing depends on the following two parameters:  

The beginning of the wheel load transition zone (a) The length of the transition zone (b)

Together with the crossing nose geometry these parameters define the magnitude of the vertical jump (Hd) of the wheel-rail contact point. All these parameters are shown in Figure 14 where the transition zone can clearly be seen. This figure also shows how the vertical rail geometry, assuming unworn wheel and rail (crossing nose and wing rail), can be obtained based on the measurements of the crossing. In the case shown in Figure 14 the following values of the parameters have been obtained: a  22.2 cm, b  6.0 cm, Hd  0.38mm . Based on these data the local vertical rail geometry has been obtained as shown in Figure 14. In addition to the local rail geometry in the crossing, the rail geometry over 10 m (5 m before and after the crossing point) has been measured as well. Since the measured irregularities were within the required limits the maximum allowable value equal to 8 mm has been chosen. The local and global vertical rail geometries of the switch were than

7

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden combined in the rail profile shown in Figure 15 which has been used in the numerical simulations described below. 2.4

Reference switch design

As a reference design the 54 E1 1:9 BWG switch installed on a ballast bed (Figure 9) has been considered. In this design the relatively hard EVA rail pads are used. The mechanical properties of the turnout components are collected in Table 1. Type

E N/m2

Poisson coefficient

Stiffness MN/m

Damping kNs/m

Mass kg/m3

Rail

UIC-54, crossing - 54 E1 1:9 BWG

2.1e11

0.3

-

-

7850

Rail pad

EVA

-

-

3032

29

-

Sleeper

B70

3.85e10

0.2

-

-

2400

-

120

48

-

Track components

Ballast Quality‘good’ Table 1 Parameters of reference turnout model 2.5

Design assessment criteria

A number of numerical simulations with the above described models have been performed in order to analyse the influence of various elastic components of the three types of the turnout structures (section 2.2) on the dynamic forces in the crossing point. Performance of each switch design has been assessed using the following criteria and responses of the train-track system:   

Damage of the crossing point estimated by the maximum dynamic wheel force (P1 force) Damage of the sleepers estimated by the maximum force acting on the sleepers (Fsl) Damage of the ballast bed estimated by the maximum force acting on the ballast bed (Fb).

Reduction of the forces on ballast is important to prevent (slow down) degradation of the overall switch geometry. As it was shown in [18] the overall switch geometry has also influence on the dynamic wheel forces. Other responses that have been collected to assess the turnout design (safety, durability of components etc.) were:     

Maximum (lower frequency) dynamic wheel force (P2 force) Maximum deformation of the rail pads (Dp) Maximum displacements of the rails (Ur) Maximum displacements of the sleepers (Usl) Maximum stress on ballast bed (Sb)

It should be noted that the above mentioned responses have been analysed only for the crossing area where the changes in the turnout design have been applied.

3

NUMERICAL RESULTS

In order to reduce the damage of the crossing nose the effect of the rail geometry and track elasticity on the dynamic wheel forces has been investigated. The dynamic performance of a switch has been estimated using the criteria and dynamic responses described in the previous section. The numerical results are presented and discussed below. 3.1

Rail geometry

Firstly, the effect of the vertical rail geometry on the dynamic forces in the crossing nose has been investigated.

8

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden The vertical rail geometry in a switch consists of the following components:  

Global or long wave rail geometry, such as the one measured over 10 m (described in Section 2.3) Local rail geometry determined by the trajectory of the wheel as it is passing the crossing

The global rail geometry is to a large extent defined by the quality of the ballast bed while the wheel trajectory depends on the geometry of the wheel profile and of the rail in the crossing area. In the study on influence of the rail geometry the simulations are performed with the reference switch model described in Section 2.4. Influence of global rail geometry First of all the effect of the long wave geometry on the dynamic forces has been investigated. The vertical long wave geometry measured over the distance of 10 m has the maximum deviation of 8 mm. Two simulations have been performed, namely one with only local rail geometry (Figure 14) and one with the combined rail geometry (Figure 15) wherein the deviation over 10 m length were included. The time history of the dynamic force calculated for the first wheel of the train in the simulations with local and combine rail geometry profile are shown in Figure 16 and Figure 17 respectively. In these figures the dynamic forces P1 and P2 introduced earlier (Figure 2) can clearly be seen. Dynamic wheel force [kN]

Dynamic wheel force [kN]

300

300

250

P1=254.03 kN, d=28.4167m

250 P1=216.92 kN, d=28.4167m

200

200

150

150

100

P2=109.39 kN, d=28.6111m

100

P2=91.479 kN, d=28.6111m

50

50

0

0

-50

-50

20

25

30 Distance [m]

35

40

Figure 16 Dynamic wheel force (reference model), local rail (crossing) geometry only

20

25

30 Distance [m]

35

40

Figure 17 Dynamic wheel force (reference model), combined rail geometry (variant v00)

In Figure 17 the disturbances in the wheel force at the distance of approximately 23 m can be observed which are absent in Figure 16. This distance correspond to the beginning of the long wave deviation introduced in the combined profile (Figure 15). As a result the dynamic forces P1 and P2 in case of the combined rail geometry are respectively 18% and 20% higher. The increase of the P1 force (as the high frequency force) leads to increase of RCF damage in the crossing nose. The higher P2 force results in faster deterioration of ballast and global switch geometry as a result which the dynamic wheel forces become even higher. That is why maintaining the switch geometry is very important to reduce the damage in the switch crossings. From these results it is also concluded that the global rail geometry has significant effect on the dynamic forces and should be taken into account in the numerical simulations. This effect increases for the higher train velocities. Thus, the combined vertical rail geometry (Figure 15) has been used in the numerical simulations presented in the subsequent sections. Effect of local rail geometry The defects in the geometry of the wheel and rail such as worn wheels, worn wing rails or wrongly repaired (welded) crossing nose can strongly affect the trajectory of the wheel when it passes the switch crossing. The possible consequences of the deteriorated wheel and rail geometry for the wheel trajectory (and the local rail geometry in the switch crossing) are:   

The distance to the transition zone (the distance a in Figure 14) becomes shorter as a result of which the two-point contact (Figure 12b) occurred earlier The length of the transition zone (the length b in Figure 14) reduces The vertical jump of the wheel and the contact point (Hd in Figure 14) increases

In order to establish qualitative relationship between the wheel and rail geometrical defects and the abovementioned consequences to the vertical wheel trajectory (the vertical rail geometry in the numerical

9

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden simulations) a detailed 3-D model of the wheel and switch is required. Since such a model was not available at this stage of the research only the relationships between the changes in the local vertical rail geometry in the switch crossing (the wheel trajectory) and the dynamic forces has been studied here. For that reason the distance to the transition zone and the length of the transition zone have been varied as it shown in Table 2. In this table the numerical results of this simulations are presented as well. The graphical results are shown in Figure 18 and Figure 19. The dynamic wheel forces for the reference switch design (variant v00) are shown in Figure 17. Wiel/rail local geometry variation Train speed 140 km/h Variant v00 v40 v41

a [cm] 12 6 12

b [cm] 6 6 3

Hd [cm] 0,38 1,33 0,854

P1 [kN] 254,03 359,65 256,78

P2 [kN] 109,39 140,86 114,96

Table 2 Results of local rail geometry variation From these results it can be seen that in both cases the changes in the wheel trajectory, which in fact correspond to some defects in the wheel/rail geometry, have resulted in increase of the dynamic forces as compared to the reference design. It should noted that by increasing the distances a and b the vertical jump of the contact point Hd increases as well which ultimately resulted in the amplification of the dynamic wheel forces. This means that the level of the dynamic forces in the switch is determined by the magnitude of the vertical jump of the wheel. Dynamic wheel force [kN]

Dynamic wheel force [kN] 400

400

P1=359.65 kN, t=0.29s

300

300

200

200

P1=256.78 kN, t=0.291s

P2=140.86 kN, t=0.295s

P2=114.96 kN, t=0.295s

100

100 0

0

-100

-100 -200

-200 0

0.1

0.2

0.3 0.4 time [s]

0.5

0.6

0

0.7

Figure 18 Dynamic wheel force, shorter distance to transition zone (variant v40)

0.1

0.2

0.3 0.4 time [s]

0.5

0.6

0.7

Figure 19 Dynamic wheel force, shorter transition zone (variant 41)

Based on the results the following conclusions on the relationships between the vertical trajectory of the wheel and the dynamic forces in the switch crossing:   

3.2

The vertical wheel trajectory and therefore the wheel/rail geometry has significant effect on the dynamic forces in the crossing nose. The magnitude of the vertical jump of the contact point in the switch crossing determines the level of the dynamic forces. By improving wheel and rail (crossing nose an wing rails) geometry the wheel trajectory can be improved which results in the reduction of the dynamic forces and ultimately in reduction of the damage in the crossing nose. In order to establish the relations between the wheel/rail geometry and dynamic forces in switch crossing the relation between the wheel rail geometry and the vertical wheel trajectory (the vertical rail geometry in the numerical simulation) should be investigated. This can be done using the 3-D model of switch. Variation of track elasticity on different levels

In order to reduce the dynamic forces in the crossing nose the stiffness of the rail support can be adjusted. The extra elasticity can be added on different levels in the rail supporting structure. In this research the following possible solutions have been considered:   

Adjustment of the elastic properties of the rail pads Application of under sleeper pads Application of ballast mats

10

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden In order to determine in which level the extra elasticity should be applied the following variations in the turnout design have been considered:   

Turnout with the soft rail pads (variant v01) Turnout with the USP (variant v20) Turnout with ballast mats (variant v30)

The numerical models described in Section 2.2 have been used here. The parameters of the reference turnout design are given in Table 1 whereas the parameters of the additional elements of the turnout used in the other two models are given in Table 1. The damping parameters of the USP and ballast mats have been taken the same Elastisity pa ra m eter varia tio n Train s peed 140 km /h K pad Cp ad V aria nt [MN/m ] [kNs/m ] v00 v01 v20 v30

3032 40 3032 3032

M sl [kg /m ]

Kusp

Kbm

[kN/m 3]

[kN/m 3]

P 1 [kN]

P 2 [kN]

120 120 120 120

0.05 -

0.03

254.03 201.63 260.39 253.95

109.39 101.03 93.59 109.46

29 29 29 29

Table 1 Results of track elasticity variation as for ballast. In all these simulations the changes have been applied in the location of the crossing. The structural responses have been compared with the ones obtained with the reference turnout design (variant v00). The model parameters together with the numerical results of the simulations are given in Table 1 whereas the graphical results are shown in Figure 20. In this figure the time histories of the wheel forces (the first wheel of the vehicle) for different turnout designs are shown. The magnitude and the frequency of the P1 and P2 forces are shown as well. Obviously, the frequencies of the P2 forces are lower than the frequencies of the P1 forces. Therefore, the P2 forces can only affect the lower frequency components of the track such as sleepers and ballast while the P1 forces have major effect on the high frequency components such as the rail. (as it will be shown later on in the parameter analysis) . By analysing the results of the simulation (Table 1) it can be concluded that application of the softer rail pads results in reduction of both P1 and P2 forces, especially the P1 force has significantly been reduced (21% reduction). That means that the use of softer pads can be favourable from the point of view of reduction of Dynamic wheel force [kN]

Dynamic wheel force [kN] 300

300 P1=254.03 kN, Fr=250Hz

250

250

200

200

150

150 P2=109.39 kN, Fr=62.5Hz

100

50

0

0 0.22

0.24

0.26

0.28

0.3 0.32 time [s]

0.34

0.36

P2=100.56 kN, Fr=45.4545Hz

100

50

-50 0.2

P1=201.63 kN, Fr=250Hz

0.38

-50 0.2

0.4

0.22

0.24

0.26

a. Reference design (v00)

0.3 0.32 time [s]

0.34

0.36

0.38

0.4

b. Model with soft rail pads (v01)

Dynamic wheel force [kN]

Dynamic wheel force [kN]

300

300 P1=260.39 kN, Fr=250Hz

250

P1=253.95 kN, Fr=250Hz

250

200

200

150

150

100

50

0

0 0.22

0.24

0.26

0.28

0.3 0.32 time [s]

0.34

0.36

0.38

P2=109.46 kN, Fr=62.5Hz

100

P2=91.992 kN, Fr=26.3158Hz

50

-50 0.2

0.28

-50 0.2

0.4

0.22

b. Model with USP (v20) Figure 20 Dynamic wheel forces, track elasticity variation on various levels

11

0.24

0.26

0.28

0.3 0.32 time [s]

0.34

0.36

0.38

0.4

e. Model with ballast mats (v30)

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden damage to the track. The results also show that the application of the USP has positive effect on P2 forces while slightly increasing the P1 forces. It should be noted that not only magnitude but also the frequency of the P2 force has been reduced. Therefore, USP can be considered as a good remedy to reduce damage to the sleepers and ballast. However, in order to prevent the increase of damage to the rails one might consider using the USP in combination with the softer rail pads. From the results of the simulation it can be seen that the ballast mats have almost no effect on the dynamic wheel forces. Based on these results it has been concluded that in general application of additional elasticity has a positive effect on the level of the dynamic wheel forces acting on the crossing nose. Depending on the level on which the elasticity is added different effects on these forces can be achieved. 3.3

Parameter analysis

The goal of the parameter analysis was to identify the major parameters of the turnout design that allow improvement of the turnout performance and reduction of damage of crossing nose. Based on the results of the previous section the number of the structural elements to be considered in the parameter analysis has been reduced to the ones located around the sleeper level such as the rail pads, sleeper, USP etc. The results of this study are presented in the extended version of this paper

4

CONCLUSIONS

In the present study the relationship between the elastic properties of the turnout substructure and the RCF damage on the crossing points has been investigated. The dynamic interaction between the railway vehicle and track structure has been analysed numerically using DARTS_NL software (TU Delft). The BWG turnout 1:9 with a ballast supporting structure has been modelled in this software. Two alternative designs of this turnout equipped with under sleeper pads and ballast mats have been modelled as well. To model the rail geometry in crossing point a procedure based on the results of visual inspection of turnouts has been developed. The results of this study have shown that the dynamic forces are strongly affected by the geometrical properties of the wheel and the rail (crossing nose and wing rail). By controlling the wheel and rail geometry the performance of the turnout can significantly be improved. The global (overall) geometry of the crossing that depends on the ballast quality also affects the dynamic forces in the crossing nose. Proper and timely maintenance of the vertical switch geometry can reduce the dynamic forces and ultimately damage in the switch crossings. The RCF damage on the crossing point is to a large extent caused by the high frequency impact dynamic forces (so-called P1). The results of the parameter analysis have demonstrated that by varying the elastic properties of the turnout substructure the forces on the crossing point can significantly be reduced. In the present paper variation of the elastic properties of the following track components has been considered: rail pads, under sleeper pads and ballast mats. The numerical results have shown that the elastic properties of the rail pads have the biggest influence on the wheel impact forces while adding ballast mats had almost no effect on the dynamic wheel forces. Based on the results of this study it can be concluded that adding extra elasticity in the turnout design has positive effect on the reduction of dynamic interaction forces. Limitations of increasing elasticity are imposed by the resulting increased vertical displacements of structural components such as rails and sleepers.

REFERENCES [1] Kassa, E., Andersson, C. and Nielsen, J.C.O. (2006) Simulation of dynamic interaction between train and

12

Presented on the 21st International Symposium on Dynamics of Vehicles on Roads and Tracks 17-21 August 2009, Stockholm, Sweden railway turnout, Vehicle System Dynamics, March 2006, Vol. 44, No.3, pp. 247-258. [2] Loy, H. (2007) Bedding optimization in turnouts, European Railway Review, 2007, Issue 6. [3] Wiest, M., Daves, W., Fischer, F.D. and Ossberger, H. (2008) Deformation and damage of a crossing nose due to wheels passages. Wear, Vol. 265, pp. 1431-1438. [4] Carels, P., Ophalffens, K., Beelen, H., Mys, J. and Schillemans, L. (2004) On the effects of high resilient undersleeper mats in turn-outs built in main line ballasted track, on ballast degradation, track stability and ground borne noise levels. In Proceedings of the 8th International Workshop on Railway Noise, Buxton, UK, 8-11 September, 2004. [5] Andersson, C. and Dahlberg, T. (2000) Load impacts at railway turnout crossing, Vehicle System Dynamics, 2000, vol. 33, SUPPL, pp. 131 – 142. [6] Jenkins, H.H., Stephenson, J.E., Clayton, G.A., Morland, G.W. and Lyon, D. (1974) The effect of track and vehicle parameters on wheel/rail vertical dynamic forces. Railway Engineering Journal, January, 1974. [7] Johansson, A., Nielsen, J.C.O., Bolmsvik, R., Karlström, A. (2008) Under sleeper pads – Influence on dynamic train-track interaction. Wear, Vol. 265, pp.1479-1487. [8] Steenbergen, M.J.M.M. (2008) Quantification of dynamic wheel-rail contact forces at short rail irregularities and application to measured rail welds. Journal of Sound and Vibration, 2008, 312, pp. 606-629. [9] Steenbergen, M.J.M.M. (2008) Wheel-rail interaction at short-wave irregularities. PhD Thesis, Delft University of Technology. [10] Markine, V.L., A.P. de Man, Toropov, V.V., Esveld, C. (2000) Optimization of Railway Structure Using Multipoint Approximations Based on Response Surface Fitting (MARS). Proceedings of the 8th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization, Long Beach, California, 6-8 September 2000 (CD Proceedings). AIAA Paper 2000-4795 ISBN 1-56347-462-X [11] Markine, V.L., Zwarthoed, J.M., C. Esveld, C. (2001) Use Of Numerical Optimisation In Railway Slab Track Design. In. O.M. Querin (Ed.): Engineering Design Optimization Product and Process Improvement. Proceedings of the 3rd ASMO UK / ISSMO conference, Harrogate, North Yorkshire, UK, 9th –10th July 2001. ISBN: 0-85316-219-0 (for text version), ISBN: 0-85316-222-0 (for CD-ROM version) [12] Markine, V.L., de Man, A.P., Esveld, C. (2003) Identification of dynamic properties of a railway track. Proceedings of the IUTAM Symposium on Field Analyses for Determination of Material Parameters Experimental and Numerical Aspects, held in Abisko National Park, Kiruna, Sweden, July 31 – August 4, 2000. Kluwer Academic Publishers. ISBN 1-4020-1283-71 [13] Esveld, C. and Markine, V.L. (2006) Assessment of High-Speed Slab Track Design. European Railway Review, Vol. 12, Issue 6, 2006, pp. 55-62. ISSN 1351 - 1599. [14] Markine, V.L. and Esveld, C. (2007) Assessment of High-Speed Slab Track Design, in Proceedings of the Eleventh International Conference on Civil, Structural and Environmental Engineering Computing, B.H.V. Topping, (Editor), Civil-Comp Press, Stirlingshire, United Kingdom, paper 54, 2007. CDROM ISBN 978-1905088-16-4 [15] Grassie, S.L., (1984) Dynamic modelling of railway track and wheelsets’, Proc. second int. conf. recent advances struct. dyn., Inst. of sound and vibration. research, University of Southampton. [16] Kok, A.W.M. (1995) Lumped pulses and discrete displacements’. PhD thesis, Delft University of Technology, Delft University Press. ISBN 90-407-1118-6 [17] Zienkiewicz, O.C., Taylor, R.L. (1988) The Finite Element Method. 4rd ed., Mc Graw Hill, London. [18] Markine, V. and Steenbergen, M.J.M.M. (2009) Improvement of Turnout Performance: Reduction of Crossing Nose Damage. Research in assignment of ProRail. TU Delft report 7-09-234-02. ISSN 0169-9288

13