Sunteți pe pagina 1din 23

A dynamic model for the vertical interaction of the rail

track and wagon system


Y.Q. Sun
a
, M. Dhanasekar
b,
*
a
Centre for Railway Engineering, Central Queensland University, Rockhampton 4702, Australia
b
Faculty of Engineering and Physical Systems, Central Queensland University, Rockhampton 4702, Australia
Received 10 December 2000
Abstract
With the advent of high-speed trains, there is a renewed interest in the rail trackvehicle interaction studies. As part
of an ongoing investigation of the track system optimisation and fatigue of the track components, a dynamic model is
developed to examine the vertical interaction of the rail track and the wagon system. Wagon with four wheelsets
representing two bogies is modelled as a 10 degree of freedom subsystem, the track is modelled as a four-layer sub-
system and the two subsystems are coupled together via the non-linear Hertz contact mechanism. The current model is
validated using several eld test data and other numerical models reported in the literature by other researchers. 2002
Elsevier Science Ltd. All rights reserved.
Keywords: Rail track; Wagon; Timoshenko beam; Hertz contact; Wheel/rail irregularities; Steady-state responses
1. Introduction
As dierent divisions within railway departments traditionally have been managing the wagon and track
separately, the dynamics of the rail track and wagon are often studied as two relatively independent
problems. For example, the wagon dynamics is examined by assuming the track as either a rigid support or
as an elastic foundation using commercial software such as NUCARS (Blader et al., 1989). The longitudinal,
the lateral and the vertical dynamics of a wagon or a train in motion are examined using the software.
The track dynamics, on the other hand, is investigated either by simplied beams on elastic foundation
approach (Jenkins et al., 1974; Grassie and Cox, 1982; Duy, 1990; Sato, 1977; Newton and Clark, 1979;
Clark et al., 1982; Tunna, 1988; Ilias and M} uuller, 1993; Cai and Raymond, 1994; Ishida et al., 1997) or by
the nite element model of rail track system (Lin and Trerhewey, 1988; Thompson, 1991; Dong et al., 1994;
Luo et al., 1996). Usually these track models have been assumed to be excited either by a single wheel or by
a single bogie with two wheelsets rolling on the rail.
Knothe and Grassie (1993) presented a state-of-the-art review on the modelling of railway track and
vehicletrack interaction. It is generally found that the Euler or Timoshenko beams representing the rail
International Journal of Solids and Structures 39 (2002) 13371359
www.elsevier.com/locate/ijsolstr
*
Corresponding author. Tel.: +61-7-4930-9677; fax: +61-7-4930-9382.
E-mail address: m.dhanasekar@cqu.edu.au (M. Dhanasekar).
0020-7683/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S0020- 7683( 01) 00224- 4
resting on an elastic foundation provides only a limited insight into the dynamic response of various track
components. An improvement to such models is achieved by accounting the discrete spacing of the sleepers.
The discrete support models and the nite element model allow improved prediction of the rail response
and oer the potential for renement by including all conceivable track components as layers.
In the trackvehicle interaction research, several eld experiments have also been reported in the liter-
ature. Two of them were conducted by the British Railways (Jenkins et al., 1974; Newton and Clark, 1979).
Other experimental investigations include those carried out by the Swedish railways (Fermeer and Nielsen,
1995) and the South African railways (Fr oohling et al., 1997).
Jenkins et al. (1974) varied the train speed up to 160 km/h. Defect was induced via dip at the shplate
joints of the rail. Two peak forces due to impact of the wheel and the rail were detected. The rst peak with
very high frequency was classed as P
1
force and the second peak with relatively low frequency was classed
as P
2
force. Newton and Clark (1979) conducted further eld experiment and measured the impact contact
forces when the trains passed over a purpose made indentation on the top surface of the rail. A range of
train speed from 27 to 117 km/h was considered in the investigation. The results of these experiments are
used as a source of validating several numerical models reported in the literature. The model presented in
this paper also has been validated using these results in addition to other results.
Fermeer and Nielsen (1995) reported a full-scale experiment carried out on the West Coast line in Sweden
using a wagon equipped with instrumented wheelsets at speeds up to 275 km/h. Five consecutive sleepers and
one rail instrumented with accelerometers and strain gauges were used. The inuence of wagon speed and
axle load on dynamic responses was studied. It was concluded that the pad stiness and the axle load largely
aected the contact forces due to wheel ats. This test data are also used to validate the model in this paper.
Fr oohling et al. (1997) reported a more detailed controlled eld test conducted in South Africa. The
purpose of the test was to understand the possible detrimental eects caused by the low frequency contact
forces. It was concluded that the track dynamic responses were aected by the vehicle load, the vehicle
speed, the track geometry, the track stiness, and the accumulating trac.
With the increase in the axle load and wagon speed, the cost of damage to track components and de-
railment risks increase substantially. This leads to widespread interests in the investigation of the dynamic
interactions of the rail track and the wagon. More rened analytical models of the rail track and wagon
system have, therefore, started emerging in the literature with the potential to optimise the design pa-
rameters of both the rail track and wagon components that would reduce the dynamic interactions. The
model reported in this paper is classed to this category of investigation. As part of an ongoing research at
the centre for railway engineering (CRE) to optimise the track and wagon components, a model containing
all components of the track and the wagon subsystems has been developed. Although an extensive eld-
testing is desirable to fully validate such a detailed model, limited validation would be possible by com-
paring the wheelrail interface forces predicted by the model with the results reported in the literature. Since
the second approach is much cheaper, as a preliminary phase of the ongoing research, we have validated the
detailed model via the results reported in the literature by other researchers. Should the objective be only to
evaluate the wheelrail interface forces, the detailed model presented in this paper would not have been
necessary. The detailed model, on the other hand, provides an insight into the dissipation of the interface
forces into the components of the track and the wagon. Furthermore the detailed trackwagon model
allows the investigation of the eect of unequal axle loads (anticipated in freight wagons).
This paper describes a track submodel developed using the discrete beam concept. All track components
are assembled exactly as per the conventional ballasted track structure used in the heavy haul railway
network. This submodel comprises of the rails, the fasteners, the pads, the sleepers, the ballast, the sub-
ballast and the subgrade arranged in four layers. The wagon submodel comprises of a rigid car body, two
two-piece bogies and four wheelsets that are connected by secondary and primary suspensions. The wheel
rail contact patch submodel is developed according to the non-linear Hertz theory of contact (Johnson,
1985). Although more sophisticated models are available for the denition of contact (for example, in-
1338 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
cluding friction), from the perspective of the overall dynamics of the trackwagon system, it was decided to
use the simple Hertz theory.
The model reported in this paper could, therefore, be described as a full wagon-four-layer track model
capable of predicting the vertical dynamic interaction when the wagon runs on the track at a steady-state
velocity in the longitudinal direction. The model could simulate the interaction of the wheel and the rail
with and without defects or irregularities (as excitation sources). Two types of excitation sources, namely,
periodic and impulse, are considered in the simulation. The model is capable of predicting the distribution
of the dynamic responses at the wheelrail interface in the downward direction to the track subgrade and in
the upward direction to the car body. The characteristics of the steady-state responses, and impact re-
sponses are presented in this paper.
2. The vertical dynamic model for the rail track and wagon interaction
Prior to presenting our model, we discuss some models that are recently reported in the literature. Cai
and Raymond (1992) reported a track dynamic interaction model consisting of one bogie. The wheelset
model included two unsprung masses, side frame mass and pitch inertia, and primary suspensions. The
track was modelled as a 40-sleeper long discretely supported system of elastic beams representing the rails
and the sleepers. This model was used to examine the dynamic response due to various wheel and rail
defects. It was found that the wheel and rail impact behaviour depends highly on the train speed. It was also
found that a wheel with an irregular prole causes not only a high impact force on itself, but also greatly
increase the impact force on the adjacent wheel.
Dahlberg (1995) reported a theoretical model similar to that of Cai and Raymond (1992) with one bogie
and track with a view to modelling the eld experiments. This model was used to investigate the sensitivity
of the parameters such as the wagon speed, the axle load, the wheel base of a bogie, the defects in rail and
wheel on the dynamic behaviour of the track and wagon components.
Zhai and Sun (1993) presented a detailed model that represented the wagon as two bogies multi-body
system and the track as an innite Euler beam supported on a discretecontinuous elastic foundation
consisting of three layers of rail, sleeper, and ballast. The signicance of mutual dynamic inuence of the
neighbouring wheelsets via the rail and the bogie was determined in the paper.
Ripke and Knothe (1995) developed a model similar to that of Zhai and Sun (1993) but used the
Timoshenko beam formulation to model the rail and sleepers instead of the Euler formulation adopted by
Zhai and Sun (1993). This model was used to investigate the eects of the local defects of the track on the
contact forces.
Our model is pictorially represented in Fig. 1. The track submodel consists of the rail, the pads, the
fasteners, the sleepers, the ballast, the subballast and the subgrade. The rst layer of model consists of
the rail represented as a continuous Timoshenko beam that is discretely supported on the fasteners and the
pads represented by the linear spring and damping elements. The enclosed dash-line box signies that the
mass of the pad and the fasteners are disregarded. The sleeper is represented in the second layer with its
mass and viscoelastic properties (spring and damper enclosed within solid-line box that signies the mass
being included). The third and the fourth layers of the model consist of the ballast and the subballast
respectively. The ballast and the subballast are considered as pyramids for calculating the eective mass,
stiness and damping coecients. The viscoelastic springs and dampers connecting one pyramid to the
other represent the continuity of the ballast and subballast in the longitudinal direction. The subgrade is
modelled as the viscoelastic elements without mass connecting the subballast to the ground.
The wagon submodel consists of the wagon body, the two two-piece bogies and the four wheelsets as
rigid bodies. The spring and the damping elements representing the secondary suspension connect the
wagon body with the two bogies. Similarly, the spring and the damping elements representing the primary
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1339
suspension connect the bogies with the wheelsets. While the car body and bogies are allowed vertical
displacement and in-plane rotation (bounce and pitch motions respectively) the wheelsets are allowed
vertical displacement (bounce motion) only as shown in Fig. 1. The wagon is thus represented by a 10
degree of freedom (DOF) system.
2.1. Rail
Rail is modelled as an innitely continuous (with its vertical deformation and rotation vanishing at both
ends long enough to be considered as innity) Timoshenko beam shown in Fig. 2.
The Timoshenko beam theory (Dym, 1973) expresses the equations for the vertical deection and ro-
tation of the rail at any point under the action of forces as shown in Eq. (1):
Fig. 1. The dynamic model of rail track and wagon system.
1340 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
qA
o
2
w
R
ot
2
GAk
o
2
w
R
ox
2

o/
R
ox

N
s
i1
F
RSi
dx x
i

4
j1
P
WRj
dx x
j

qI
o
2
/
R
ot
2
GAk
ow
R
ox
/
R
EI
o
2
/
R
ox
2
0
_

_
1
where w
R
is the vertical deection of the rail, /
R
is the rotation of the rail, q is the rail density, A is the area
of the rail cross-section, G is the shear modulus of the rail, E is the Youngs modulus of the rail material, I is
the second moment of area of the rail section, k is the Timoshenko shear coecient, F
RSi
is the reaction
force between the rail and the ith sleeper, P
WRj
is the contact force between the jth wheel and the rail, dx is
the Dirac delta function, x
i
is the position of the ith sleeper, x
j
is the position of the jth wheel, and N
s
is the
number of sleepers considered. The subscript i is used for the sleeper count and j for the wheel count.
The vertical deection w
R
and rotation /
R
of the rail are obtained using modal superposition as given in
Eq. (2):
w
R


N
c
h1
N
w
h; xW
h
t
/
R


N
c
h1
N
/
h; xU
h
t
_

_
2
where N
w
h; x and N
/
h; x are the hth mode shape functions of the vertical deection and rotation re-
spectively of the rail, W
h
t and U
h
t are the hth mode time coecients of the rail vertical deection and
rotation respectively of the rail, N
c
is the number of modes considered and x represents the linear coordinate
along the length of the rail beam.
By substituting Eq. (2) into Eq. (1), we modify the partial dierential equation (1) into ordinary dif-
ferential equation shown in Eq. (3) below. This transformation facilitated the application of the Newmark-
b method to solve the equations.
d
2
W
h
dt
2

Gk
q
Ph
L
_ _
2
W
h

A
I
_
Gk
q
Ph
L
_ _
U
h

N
s
i1
F
RSi
N
w
h; x
i

4
j1
P
WRj
N
w
h; x
j
h 1; 2; . . . ; N
c

d
2
U
h
dt
2

GAk
qI

E
q
Ph
L
_ _
2
_ _
U
h

A
I
_
Gk
q
Ph
L
_ _
W
h
0
_

_
3
in which L is the length of the rail considered, and the reaction force between the rail and the ith sleeper F
RSi
is expressed as in Eq. (4):
F
RSi
C
pi
C
fi

N
c
m1
N
w
m; x
i

_
WW
m
K
pi
K
fi

N
c
m1
N
w
m; x
i
W
m
C
pi
C
fi
_ ww
si
K
pi
K
fi
w
si
4
where w
si
is the vertical displacement of the ith sleeper. C
pi
, K
pi
and C
fi
, K
fi
are the damping and stiness
coecients of the ith pad and the ith fastener respectively.
Fig. 2. Timoshenko beam model of the rail.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1341
In Eq. (3), the contact force P
WRj
between the jth wheel and the rail is determined by the non-linear Hertz
contact theory and is given in Eq. (5),
P
WRj
t
C
H
w
wj
t w
r
x
pj
; t w
d
t
_ _
3=2
if w
wj
t w
r
x
pj
; t w
d
t > 0
0 if w
wj
t w
r
x
pj
; t w
d
t < 0
_
5
where w
wj
t and w
r
x
pj
; t are the displacements of the wheel and the rail at the jth contact point, w
d
t is
the wheel and/or the rail irregularity function (for example, an out-of-round wheel, rail corrugation or rail
surface geometric irregularity), C
H
is the Hertz contact coecient that can be deduced from Johnson (1985)
as follows:
C
H

4G
wr

R
e
p
31 m
wr

6
in which G
wr
is the shear modulus, m
wr
is the Poissons ratio, and R
e

rR
p
(r is the rolling radius of wheel,
R q
w
R
t
=q
w
r (q
w
and R
t
are the wheel prole radius and the rail prole radius respectively)).
In Eq. (5) the contact force P
WRj
is calculated based on the relative displacement between the wheel and
the rail at the point of contact x
pj
. This point is easily determined by keeping the angle between the vertical
diameter of the wheel and the axis of the rail as 90 for the non-defect wheels and rails. However, where
defect (in the wheel or the rail) is encountered, the angle between the diameter of the wheel drawn through
the point of contact and the axis of the rail varies from 90. The exact point of contact is determined in such
cases by dividing the contact length obtained from static Hertz analysis into smaller segments and checking
each segment for potential contact. A similar approach has been reported by Dong et al. (1994).
2.2. Pads, fasteners and sleepers
Rubber or high-density polyethylene mats that are used as a bearing layer between the rails and the
concrete sleepers are commonly known as pads. Rail fastener connects the rail and the sleeper together. The
elasticity of the fastener is measured by the spring rate, which is the amount of deection proportional to
the clamping force. In the model, both the pads and the fasteners are modelled as the linear springs and
dampers without mass.
Sleeper is the track component that ties the two rails together thereby providing monolithic action to the
track. Sleepers are positioned between the rails and the ballast and are represented in the model by their
mass, stiness and damping properties. The stiness of sleepers is calculated using the inuence coecient
approach by considering the sleepers as beams on elastic foundation proposed by Prollidis (2000). The
track structure has been considered as medium quality for the evaluation of the sleeper stiness. The
damping coecient is then determined based on the values of stiness and mass.
2.3. Ballast and subballast
The ballast ensures damping of the vibrations and distributes the load evenly to the subgrade. The
subballast protects the top surface of the subgrade from penetration of the ballast stone particles, in ad-
dition to, further distributing the load. Ahlbeck et al. (1975) developed the ballast pyramid model based on
the theory of elasticity. The ballastsubballast pyramid model shown in Fig. 3 assumes that the loading and
pressure distribution is uniform throughout the depth. In Fig. 3 the model is divided into the upper and
lower sections, which reects the actual transmission of the loading. Zhai and Sun (1993) dened the vi-
bration of the ballast as a single block based on the observation that the accelerations of the individual
particles in both upper and lower surfaces of the ballast block do not vary signicantly even though such a
conclusion is not universal. The oscillating mass of each ballast block is calculated by multiplying the
1342 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
volume of ballast block with the ballast density. According to Ahlbeck et al. (1975), the stiness of the ith
ballast block K
bl
is:
K
bl

2 tan h
b
L
s
B
s
E
b
ln
L
s
2 tanh
b
H
b
B
s

B
s
2 tan h
b
H
b
L
s

_ _ 7
in which L
s
and B
s
are the eective length and width of the support area of the rail seat, E
b
is the modulus of
elasticity of the ballast (in N/m
2
), h
b
is the internal friction angle of ballast (20 is chosen for ballast as
Ahlbeck et al. (1975) suggested), and H
b
is the height of the ballast.
Similarly, the stiness of the ith subballast K
sb
is:
K
sb

2 tan h
sb
L
s
B
s
E
sb
ln
2 tanh
b
H
b
L
s
2 tan h
sb
H
sb
2 tan h
b
H
b
B
s

2 tanh
b
H
b
B
s
2 tan h
sb
H
sb
2 tanh
b
H
b
L
s

_ _ 8
in which E
sb
is the modulus of elasticity of the subballast (in N/m
2
), h
sb
is the internal friction angle of the
subballast (35 is chosen for subballast), H
sb
is the height of the subballast.
The damping coecients of the ballast and the subballast are determined as 40% of their critical
damping coecients. This damping ratio (40%) is considered realistic for earth structures and is found that
these values are within the range given by Grassie et al. (1982) (for example, the post-tamping and pre-
tamping tracks were 30 and 82 kNs/m respectively).
The oscillating masses of each ballast block M
bl
and subballast block M
sb
are:
M
bl
q
b
L
s
B
s
H
b
tan h
b
L
s
B
s

4
3
H
2
b
tan
2
h
b
9
M
sb
q
sb
L
s
2 tan h
b
B
s
2 tan h
b
H
sb
tan h
sb
L
s
B
s
4 tan h
b

4
3
H
2
sb
tan
2
h
sb
10
The subgrade stiness K
sg
is:
K
sg
E
sg
2 tan h
sb
H
sb
2 tan h
b
H
b
L
s
2 tan h
sb
H
sb
2 tan h
b
H
b
B
s
11
in which E
sg
is the modulus of the subgrade expressed in N/m
3
.
In the longitudinal direction the continuity of the ballast and the subballast are ensured by including
viscoelastic elements (without mass) connecting the blocks of ballast and subballast in their respective
layers. The coecients of these longitudinal springs and dampers were calculated by multiplying the re-
spective vertical stiness and damping coecients by a factor of 0.3. This factor is not sensitive to the
dynamic responses on the interface between the wagon and the track.
Fig. 3. The ballast and subballast pyramid model.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1343
3. Equations of motion
In the simulation model presented here, the wagon is assumed to be under a steady state motion in the
longitudinal direction. The dynamic equations of motions in the vertical direction for the track and wagon
subsystems and the interface are presented in this section.
3.1. The track dynamics
The equations of motion for the ith sleeper, ballast and subballast blocks are established from the basic
dynamic equilibrium concept.
For the ith sleeper:
M
s
ww
si
C
pi
_
C
fi

C
sli
C
bli
C
sli
C
bli
_
_ ww
si
K
pi
_
K
fi

K
sli
K
bli
K
sli
K
bli
_
w
si
C
pi
C
fi

N
c
m1
N
w
m; x
i

_
WW
m
K
pi
K
fi

N
c
m1
N
w
m; x
i
W
m

C
sli
C
bli
C
sli
C
bli
_ ww
bli

K
sli
K
bli
K
sli
K
bli
w
bli
0 12
for the ith ballast block:
M
bl
ww
bli

C
sli
C
bli
C
sli
C
bli
_
C
sbi
2C
jbi
_
_ ww
bli

K
sli
K
bli
K
sli
K
bli
_
K
sbi
2K
jbi
_
w
bli

C
sli
C
bli
C
sli
C
bli
_ ww
si

K
sli
K
bli
K
sli
K
bli
w
si
C
sbi
_ ww
sbi
K
sbi
w
sbi
C
jbi
_ ww
bli1
K
jbi
w
bli1
C
jbi
_ ww
bli1
K
jbi
w
bli1
0 13
for the ith subballast block:
M
sb
ww
sbi
C
sbi
C
sgi
2C
jsbi
_ ww
sbi
K
sbi
K
sgi
2K
jsbi
w
sbi
C
sbi
_ ww
bli
K
sbi
w
bli
C
jsbi
_ ww
sbi1
K
jsbi
w
sbi1
C
jsbi
_ ww
bli1
K
jbi
w
bli1
0 14
in Eqs. (12)(14), M
s
, M
bl
, M
sb
are the masses of the sleeper, the ballast block and the subballast block
respectively; C
sli
, K
sli
, C
bli
, K
bli
and C
sbi
, K
sbi
are the damping and stiness coecients of the ith sleeper, the
ith ballast block and the ith subballast block respectively; C
jbi
, K
jbi
and C
jsbi
, K
jsbi
are the damping and the
stiness coecients between the ith ballast block and its adjacent ballast blocks and between the ith
subballast block and its adjacent subballast blocks respectively, C
sgi
, K
sgi
are the damping and stiness
coecients of the subgrade; w
bli
, w
sbi
are the vertical displacements of the ith ballast and subballast blocks;
w
bli1
, w
bli1
, w
sbi1
, w
sbi1
are the vertical displacements of the adjacent ballast and subballast blocks of
the ith ballast and subballast blocks.
3.2. The wagon dynamics
The wagon body is connected with the two bolsters, which rests on the secondary suspension. As ex-
plained before, the wagon body presents bounce and pitch motions in the vertical plane. Similarly, the side
frames that are the connection structures between the primary and the secondary suspensions have bounce
and pitch motions. The wheelsets have only bounce motion. According to the model shown in Fig. 1, the
equations of motion of the wagon components are deduced as follows:
for the bounce of wagon body:
M
c
ww
c
2C
sc
_ ww
c
2K
sc
w
c
C
sc
_ ww
b1
K
sc
w
b1
C
sc
_ ww
b2
K
sc
w
b2
0 15
1344 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
for the pitch of wagon body:
J
c

//
c
2C
sc
L
2
c
_
//
c
2K
sc
L
2
c
/
c
C
sc
L
c
_ ww
b1
K
sc
L
c
w
b1
C
sc
L
c
_ ww
b2
K
sc
L
c
w
b2
0 16
for the bounce of the rear side frame:
M
b1
ww
b1
C
sc
2C
pr
_ ww
b1
K
sc
2K
pr
w
b1
C
sc
_ ww
c
L
c
_
//
c
K
sc
w
c
L
c
/ C
pr
_ ww
w1
K
pr
w
w1
C
pr
_ ww
w2
K
pr
w
w2
0 17
for the pitch of the rear side frame:
J
b1

//
b1
2C
pr
L
2
w
_
//
b1
2K
pr
L
2
w
/
b1
C
pr
L
w
_ ww
w1
K
pr
L
w
w
w1
C
pr
L
w
_ ww
w2
K
pr
L
w
w
w2
0 18
for the bounce of the front side frame:
M
b2
ww
b2
C
sc
2C
pr
_ ww
b2
K
sc
2K
pr
w
b2
C
sc
_ ww
c
L
c
_
//
c
K
sc
w
c
L
c
/ C
pr
_ ww
w3
K
pr
w
w3
C
pr
_ ww
w4
K
pr
w
w4
0 19
for the pitch of the front side frame:
J
b2

//
b2
2C
pr
L
2
w
_
//
b2
2K
pr
L
2
w
/
b2
C
pr
L
w
_ ww
w3
K
pr
L
w
w
w3
C
pr
L
w
_ ww
w4
K
pr
L
w
w
w4
0 20
for the bounce of rear wheelset:
M
w
ww
w1
C
pr
_ ww
w1
K
pr
w
w1
C
pr
_ ww
b1
L
w
_
//
b1
K
pr
w
b1
L
w
/
b1
P
WR1
0 21
for the bounce of the third wheelset:
M
w
ww
w2
C
pr
_ ww
w2
K
pr
w
w2
C
pr
_ ww
b1
L
w
_
//
b1
K
pr
w
b1
L
w
/
b1
P
WR2
0 22
for the bounce of the second wheelset:
M
w
ww
w3
C
pr
_ ww
w3
K
pr
w
w3
C
pr
_ ww
b2
L
w
_
//
b2
K
pr
w
b2
L
w
/
b2
P
WR3
0 23
for the bounce of the leading wheelset:
M
w
ww
w4
C
pr
_ ww
w4
K
pr
w
w4
C
pr
_ ww
b2
L
w
_
//
b2
K
pr
w
b2
L
w
/
b2
P
WR4
0 24
in the Eqs. (15)(24), M
c
, M
b1
, M
b2
, M
w
are the masses of the wagon body, front bogie, rear bogie and the
four wheelsets respectively; K
sc
, C
sc
, K
pr
, C
pr
are the stiness and damping coecients of the secondary and
the primary suspensions; w
c
, /
c
are the vertical displacement and rotation of the wagon body; w
b1
, /
b1
, w
b2
,
/
b2
are the vertical displacement and rotation of the front bogie and the rear bogie and w
w1
, w
w2
, w
w3
, w
w4
are the vertical displacements of the four wheelsets.
3.3. The dynamics of the interface
The equations of the complete system are obtained by assembling the above equations in a matrix form
as shown in Eqs. (25a) and (25b) for the wagon and rail track respectively.
M
W
f ww
W
g C
W
f _ ww
W
g K
W
w
W
f g F
W
f g 25a
M
T
f ww
T
g C
T
f _ ww
T
g K
T
w
T
f g F
T
f g 25b
where
fw
W
g fw
c
/
c
w
b1
/
b1
w
b2
/
b2
w
w1
w
w2
w
w3
w
w4
g
T
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1345
fF
W
g f0 0 0 0 0 0 P
WR1
P
WR2
P
WR3
P
WR4
g
T
is the force vector that is consisted of the contact forces
between the wheels and the rail;
fw
T
g fW
1
. . . W
N
c
U
1
. . . U
N
c
w
s1
. . . w
sN
s
w
bl1
. . . w
blN
s
w
sb1
. . . w
sbN
s
g
T
fF
T
g f
~
FF
1
. . .
~
FF
h
. . .
~
FF
N
c
0 . . . 0g
T
is the force vector that includes the reaction forces between the rail and the
sleepers and the contact forces between the wheels and the rail, in which
~
FF
h

N
s
i1
F
RSi
N
w
h; x
i

4
j1
P
WRj
N
w
h; x
j
h 1; 2; . . . ; N
c
:
3.4. Wheel/rail irregularities
Irregularities in wheel and/or rail generate sharp peak responses in the trackwagon system. Some ir-
regularities cause periodic excitation whilst others cause non-periodic or localised excitation dened as
impulse excitation in this paper. The periodic irregularities include the rail corrugations, the out-of-round
wheels or the rounded at wheels, and the non-periodic irregularities include the indentation on the rail-
head due to the spalling or the defect of welded joint and the dipped-joint.
The periodic irregularities are represented by cosine functions. Table 1 shows some cases of the periodic
excitation sources and the corresponding expression w
d
t. When the excitation source is non-periodic and
L
d
62

2ra a
2
p
, (where L
d
is the wavelength of irregularity, r is the rolling radius of wheel, a is the wave
depth of the irregularity), the wheel and the rail will not be in contact with the trough of the irregularity.
For small L
d
, when the at wheel runs on the perfect rail or the perfect wheel runs at the defective rail, the
instantaneous rotating centre of the wheel suddenly moves down or up, inducing a vertical impact velocity.
In this situation, we call these excitation sources as the impulse excitation sources. Various impact velocities
have been deduced to simulate small wheel ats, the short length of rail shelling and spalling, and the
defects from rail welded joints. These impact velocities are shown in Table 2.
3.5. Solution technique
Eqs. (25a) and (25b) is solved using the Newmark-b method; A similar numerical integration method has
also been used by Zhai (1996). For the case without the excitation, the term w
d
t in the expression of
Table 1
Harmonic excitation sources
Name and geometry Expression
Out-of-round wheel W
d
t a1 cos Xt=2
X
2pV
L
d
0 6t 6L
d
=V
Indentation on rail surface awave length of irregularity
L
d
wave depth of irregularity
(for out-of-round wheel, L
d
is the length of arc)
1346 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
contact force shown in Eq. (5) is disregarded. Some typical functions for w
d
t have been discussed in the
above section. The ow chart for the solution technique is schematically presented in Fig. 4.
4. Validation of the model
4.1. Dynamic response without defects
Example 1: To validate the model illustrated in this paper its predictions of the responses are compared
with the responses reported by Dong et al. (1994) for the case where there is no excitation source (in other
words perfect wheels running on the defect free surface of the rail). The purpose of investigating the steady-
state responses without defects is to obtain the characteristics of the responses, and more importantly to
determine the natural frequencies of the coupled wagon and track system.
Dong et al. (1994) modelled a problem of one perfect wheel travelling at a constant speed (148 km/h)
over a defect-free rail surface by just one wheel rolling on a track with just two layers. To predict the
responses reported by Dong et al. (1994), it was required to simplify the model reported in this paper.
Accordingly, Eqs. (7)(11), (13) and (14) were removed from the track subsystem and Eqs. (15)(20) and
(22)(24) were removed from the wagon subsystem. Eq. (21) was simplied as the rolling wheel was as-
signed only the static wheel load. This has substantially reduced the overall complexity of the model
presented in this paper. The resulting one wheel-two layer track system contained 341 equations of motion.
This corresponds to 100 sleeper long track. The parameters obtained from Dong et al. (1994) used in the
execution of the simplied current model is shown in Table 3.
The rail and wheel displacements at the contact point and the contact force factor (the contact force
divided by the static wheel load) evaluated by the current model and that of Dong et al. (1994) are shown in
Fig. 5(a) and (b) respectively. Both gures show that the prediction of the responses by the current model is
in good agreement to the responses reported by Dong et al. (1994). The period of vibration predicted by
both models is almost the same (0.0192 s), which reects that the period is obtained by dividing the sleeper
spacing by the wagon speed. The dierence between the displacement of the wheel and the rail at any
dened time interval is also almost exactly same (0.1 mm) as that of Dong et al. (1994). However, whilst the
trough displacement of both the rail and the wheel during the entire period of simulation compare very well
with that of Dong et al. (1994), the peak displacement vary. The situation seems to be just reverse for the
Table 2
Impulse excitation sources
Name and geometry Impulse velocity
Raise on welding joint V
0
V
2H
r

1=2
rwheel radius
V wagon spped
Dipped-joint V
0
a
1
a
2
V
a
1
, a
2
dip angles of joint
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1347
contact force factor shown in Fig. 5(b). The dierence of the maximum amplitudes of the two results is
about 5.6% and 1.95% respectively for the displacement and contact force factor. It is found that the
amplitudes of the dynamic responses of both the wheel and rail displacements and the contact force at the
contact point are sensitive to the damping coecients of both the pad and the ballast. From Table 3 it could
be seen that the damping coecients of the pad and the ballast are calculated indirectly from the data
provided in the paper and perhaps these values dier to the actual value. This might partly explain the
dierence.
4.2. Dynamic responses with defects
Example 2: Most wheel ats are created by wheel slide during the application of break. Newton and
Clark (1979) published a good set of data obtained from controlled eld experiment for a long indention
(150 mm length and 2.15 mm depth) on the top surface of the rail to simulate a wheel at.
We have used these eld experimental data for validation of the dynamic model of wagon and track
interaction reported in this paper. All the equations from Eqs. (3) to (24) are therefore used in the simu-
Fig. 4. Flow chart for solution technique.
1348 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
lation. The total number of equations of motion solved for the simulation was 550. Once again, as in
Example 1, the track used was 100 sleeper long. Due to the limited parameters given by Newton and Clark
(1979), some parameters were extracted from a commonly known wagon and track system in Australia
(McClanachan, 1999; Zhang et al., 1998). All parameters used in the simulation are listed in Table 4.
Although our detailed model was used in the evaluation of the impact response of the track system, it was
of general interest to feel for the static mean stiness of the track. For this purpose the static mean stiness
of the track was calculated as 26.9 MN/m from the sleeper, ballast, the subballast and the subgrade stiness
values (Table 4). The static mean stiness of 26.9 MN/m is within an acceptable range of values provided in
Grassie (1992) and Knothe and Grassie (1993).
The dynamic response of the whole system was simulated when the wagon travelled at a constant speed
of 117 km/h and passed through the indentation on the rail top surface. Fig. 6(a) shows the comparison of
the wheel and rail contact force factor calculated using the current model with the eld experimental data
obtained by Newton and Clark (1979). It can be seen that when the wheel touches the indentation, the
contact force reduces to zero. This really means that the wheel and the rail separates for a while; when they
meets again, a large peak force is induced between the wheel and the rail, and the impact is accompanied by
the obvious oscillation as the natural frequency of the wagon and track system. However, the oscillation
decays very fast. This phenomenon is predicted by both the experimental data and the simulation results.
For the large peak force, the magnitude of the current model has a certain deviation compared with the
measured value by Newton and Clark (1979) and the error is about 18.4%. However, the duration of this
peak force is almost the same for both results.
The contact force is sensitive to the damping coecients of the pad, and the sleeper. As these values are
only assumed or calculated (as shown in Table 4), there is a possibility of them being dierent to that
Table 3
All parameters for Example 1: Dong et al. (1994)
Notation Parameter Value
Wagon subsystem
F
s
Static wheel load
a
82 kN
M
w
Wheel mass
a
500 kg
Track subsystem
m
r
Rail mass per meter
a
56 kg/m
A
r
Rail cross-section area
a
7:17 10
3
m
2
E
r
Elastic modulus of rail
a
2:07 10
11
N/m
2
G
r
Shear modulus of rail
a
8:1 10
10
N/m
2
I
r
Rail second moment of area
a
2:35 10
5
m
4
k Timoshenko shear coecient
a
0.34
C
p
Pad damping
b
21.8 kNs/m
K
p
Pad stiness
a
200 MN/m
m
s
Sleeper mass
b
50 kg
S
l
Sleeper spacing
a
0.79 m
C
b
Ballast damping
b
21.8 kNs/m
K
b
Ballast stiness
a
31.6 MN/m
Interface subsystem
C
H
Hertz spring constant
a
1:0 10
11
N/m
3=2
a
The values given by Dong et al. (1994).
b
The value is calculated based on the related parameters given by Dong et al. (1994). For example, the damping coecients of both
pad and ballast are equally given by multiplying the foundation damping per unit length (27.6 kNs/m
2
) with sleeper spacing; the
sleeper mass is obtained by subtracting the rail mass per meter from the track mass per unit length (119 kg/m), and then multiplying the
sleeper spacing.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1349
of the actual. Perhaps this explains the dierence in the peak magnitude of the eld test result and that of
the current model.
Fig. 6(b) shows the time histories of the displacements of the wheel and the rail at the contact point and
the rail irregularity. The rail prole is also shown in the gure. It can be seen that without the contact force,
the rail jumps up due to its elasticity while the wheel moves down due to gravity. When the wheel and the
rail meets again, a large contact force is induced. Fig. 6(c) shows the time histories of displacements of the
sleeper, the ballast and the subballast. It can be seen that the obvious dynamic responses occur with a time
lag due to their damping eect.
Fig. 5. (a) Comparison of the displacements of the wheel and the rail at the point of contact with Dong et al. (1994) and (b) comparison
of the contact force factor with Dong et al. (1994).
1350 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
The time histories of the secondary and the primary suspension force factor (the force divided by the
static wheel load) are shown in Fig. 6(d). It can be seen that the secondary suspension force almost remains
unchanged during the wheel impact on the rail while the primary suspension force drops due to the sep-
aration of the wheel and the rail and peaks when the wheel and the rail meets again. It can also be seen that
the adjacent primary suspension force has a similar response. Such periodic loading due to a large wheel
Table 4
Basic parameters for Example 2: Newton and Clark (1979)
Notation Parameter Value
Wagon system
M
c
Wagon body mass (loaded)
a
58 400 kg
J
c
Wagon mass moment of inertia about Y axis
a
617 282 kg m
2
M
b
Bogie mass
a
3600 kg
J
b
Bogie mass moment of inertia about Y axis
a
1801 kg m
2
K
sc
Secondary suspension stiness
a
2555 kN/m
C
sc
Secondary suspension damping
a
30 kNs/m
K
pr
Primary suspension stiness
b
6500 kN/m
C
pr
Primary suspension damping
b
10 kNs/m
L
c
Distance between two bogie Y -direction centrelines
a
10.36 m
L
w
Wheelset base
a
1.675 m
r Wheel radius
c
0.50 m
Track subsystem
m
r
Rail mass per meter
c
56 kg/m
A
r
Rail cross-section area
c
7:17 10
3
m
2
E
r
Elastic modulus of rail
c
2:07 10
11
N/m
2
G
r
Shear modulus of rail
c
8:1 10
10
N/m
2
I
r
Rail second moment of area
c
2:35 10
5
m
4
r Rail prole radius on top
a
0.30 m
k Timoshenko shear coecient
c
0.34
m Poissons ratio of rail
a
0.27
C
p
Pad damping
b
70 kNs/m
K
p
Pad stiness
c
200 MN/m
C
f
Fastener damping
b
0.4 kNs/m
K
f
Fastener stiness
a
1 MN/m
m
s
Sleeper mass
d
50 kg
S
l
Sleeper spacing
c
0.79 m
C
s
Sleeper damping
b
50 kNs/m
K
s
Sleeper stiness
b
79 MN/m
R
t
Rail top prole radius
a
0.30 m
L
s
Eective length of rail seat support area
a
0.18 m
B
s
Eective width of rail seat support area
a
0.164 m
H
b
Ballast height
a
0.30 m
H
sb
Subballast height
a
0.15 m
E
b
Elastic modulus of ballast
a
130 MN/m
2
E
sb
Elastic modulus of subballast
a
200 MN/m
2
E
s
Elastic modulus of subgrade
a
65 MN/m
3
Interface subsystem
C
H
Hertz spring constant
c
1:0 10
11
N/m
3=2
a
Assumed values based on a wagon and track system in Australia (M.J. McClanachan, Y.J. Zhang). The wagon body mass is
adjusted to have a static wheel force of 82 kN.
b
Assumed value (V.A. Prollidis, M.J. McClanachan, Y.J. Zhang).
c
The values given by Newton and Clark (1979).
d
The value is calculated based on the related parameters given by Newton and Clark (1979). Sleeper mass is obtained by subtracting
the rail mass per meter from the track mass per unit length (119 kg/m), and then multiplying the sleeper spacing.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1351
ats would certainly cause fatigue failure of the primary suspension. Similar conclusion has been made by
Ahlbeck and Hadden (1985) with reference to bearing fatigue.
Example 3: Fermeer and Nielsen (1995) reported their results of the eld testing on the study of vertical
interaction between the wagon and the track. The experiments were carried out with wheel ats on the
instrumented wheelset, which were articially grounded with an initial circumferential length of 40 mm and
a depth of 0.35 mm. Such length is the limit value for the at wheel service in the Swedish railways. The
main objective of the experiments was to examine the inuence of the vehicle speed, the axle load, and the
rail pad stiness on the dynamic responses of the vehicle and that of the track.
These eld experimental data due to a small wheel at (relative to the previous example) were used for
the validation of the current model. The same number of equations of motion (550), and the number of
sleepers (100) as in the previous example was used. Only a few parameters are reported by Fermeer and
Nielsen (1995), and most other parameters were extracted based on the commonly known wagon and track
system in Australia (McClanachan, 1999; Zhang et al., 1998). The parameters used in the simulation are
listed in Table 5. Similar to Example 2, the static mean stiness of the track used in the current example was
Fig. 6. (a) Comparison of the contact force factor predicted by the current model for Newton and Clark (1979) experimental results, (b)
the displacements of the wheel and the rail at the contact point predicted by the current model for Newton and Clark (1979) ex-
periments, (c) the displacements of the sleeper, the ballast and the subballast predicted by the current model for Newton and Clark
(1979) experiments and (d) the secondary and primary suspension forces predicted by the current model for Newton and Clark (1979)
experiments.
1352 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
also calculated from the sleeper, ballast, the subballast and the subgrade stiness values (Table 5) as 28.8
MN/m. The static mean stiness of 28.8 MN/m is within an acceptable range of values provided in Grassie
(1992) and Knothe and Grassie (1993).
The dynamic response of the whole system was simulated when the wagon travelled at a constant speed
of 70 km/h with the at wheel.
Table 5
Basic parameters for Example 3: Fermeer and Nielsen (1995)
Notation Parameter Value
Wagon system
M
c
Wagon body mass (loaded)
a
80 600 kg
J
c
Wagon mass moment of inertia about Y axis
a
726 462 kg m
2
M
b
Bogie mass
a
3600 kg
J
b
Bogie mass moment of inertia about Y axis
a
1801 kg m
2
K
sc
Secondary suspension stiness
a
2555 kN/m
C
sc
Secondary suspension damping
a
30 kNs/m
K
pr
Primary suspension stiness
b
6500 kN/m
C
pr
Primary suspension damping
b
10 kNs/m
L
c
Distance between two bogie Y -direction centrelines
a
10.36 m
L
w
Wheelset base
a
1.675 m
r Wheel radius
c
0.475 m
Track subsystem
m
r
Rail mass per meter
c
60 kg/m
A
r
Rail cross-section area
a
7:77 10
3
m
2
E
r
Elastic modulus of rail
a
2:07 10
11
N/m
2
G
r
Shear modulus of rail
a
8:1 10
10
N/m
2
I
r
Rail second moment of area
a
2:94 10
5
m
4
r Rail prole radius on top
a
0.30 m
k Timoshenko shear coecient
a
0.34
m Poissons ratio of rail
a
0.27
C
p
Pad damping
c
45 kNs/m
K
p
Pad stiness
c
140 MN/m
C
f
Fastener damping
b
0.4 kNs/m
K
f
Fastener stiness
a
1 MN/m
m
s
Sleeper mass
a
270 kg
S
l
Sleeper spacing
a
0.685 m
C
s
Sleeper damping
b
130 kNs/m
K
s
Sleeper stiness
b
98 MN/m
R
t
Rail top prole radius
a
0.30 m
L
s
Eective length of rail seat support area
a
0.18 m
B
s
Eective width of rail seat support area
a
0.164 m
H
b
Ballast height
a
0.30 m
H
sb
Subballast height
a
0.15 m
E
b
Elastic modulus of ballast
a
130 MN/m
2
E
sb
Elastic modulus of subballast
a
200 MN/m
2
E
s
Modulus of subgrade
a
65 MN/m
3
Interface subsystem
C
H
Hertz spring constant
d
0:87 10
11
N/m
3=2
a
Assumed values based on a wagon and track system in Australia (M.J. McClanachan, Y.J. Zhang). The wagon body mass is
adjusted to have a static wheel force of 106 kN.
b
Assumed values (V.A. Prollidis, M.J. McClanachan, Y.J. Zhang).
c
The values given by Fermeer and Nielsen (1995).
d
The value is calculated using formula reported in this paper.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1353
Fig. 7. (a1) Theoretical prediction of contact force by Fermeer and Nielsen (1995), (a2) experimental prediction of contact force by
Fermeer and Nielsen (1995), (a3) theoretical prediction of contact force by current model. (b) The displacements of the wheel and the
rail at the contact point predicted by the current model for Fermeer and Nielsen (1995) experiments.
1354 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
Fig. 7(a) shows the comparison of the measured contact force from Fermeer and Nielsen (1995) and the
calculated contact forces using the current model. The contact forces calculated by Fermeer and Nielsen
(1995) is also provided in the gure. To avoid congestion and improve clarity, Fig. 7(a) is presented as three
graphs in Fig. 7(a1), Fig. 7(a2) and Fig. 7(a3) respectively. It can be seen from Fig. 7(a1)(a3), that the
numerical results predicted by our model are in reasonable agreement with the measured responses although
the experimental and predicted frequencies dier especially in the non-peak region. However, it may be
stated that amongst the results of the two models provided in Fig. 7(a1) and (a3), the result of our model
compares better with the experimental result especially at the peak region of the response curve. Both nu-
merical models predict lower level of vibration in the non-peak region relative to the experimental results.
The rst positive peak force of the calculated and the measured results is about 1.5 times of the static
wheel load. The (larger) dierence in the second positive peak force may be attributed to the values of
damping coecient used for several components in the absence of the actual data (Table 5).
The displacement time histories of the wheel and the rail at the contact point calculated in this paper is
presented in Fig. 7(b), which exhibits the continuous penetration of the wheel into the rail. During impact
the maximum magnitudes of the displacement increase by about 11% of the average displacement.
Example 4: This is an example from the impulse source, which is available from Zhai (1996) (written in
Chinese). Zhai (1996) reported the experimental data in order to verify a model developed by him. The
impulse excitation source was from a dipped-joint with the total angle a
1
a
2
0:02 radian (a
1
, a
2
are the
dip angles of joints).
The full model with 550 equations of motion (100 sleeper long track) was used in the simulation. Due to
incomplete parameters given by Zhai (1996), we assumed the missing data based on the commonly known
wagon and track system used in Australia (McClanachan, 1999; Zhang et al., 1998). The data used in the
simulation are presented in Table 6. Similar to Examples 2 and 3, the static mean stiness of the track used
in the current example was also calculated from the sleeper, ballast, the subballast and the subgrade stiness
values (Table 5) as 28.8 MN/m. The static mean stiness of 28.8 MN/m is within an acceptable range of
values provided in Grassie (1992) and Knothe and Grassie (1993).
The dynamic response was calculated when the wagon travelled at a speed of 48.7 km/h and passed
through this dipped-joint.
Fig. 8(a) shows the comparison of the contact forces from the experimental data and that calculated data
by Zhai (1996) with the results simulated in this paper. The displacements of the wheel and the rail at the
contact point are also reported in this paper. From Fig. 8(a), the magnitude of the peak force and its du-
ration are in good agreement with the experimental and calculated data by Zhai (1996), and the peak force is
about twice as the static wheel load in this situation. It should be noted that the frequencies calculated from
the current model is in better agreement with the experimental data than the calculated values of Zhai (1996)
especially in the non-peak region. Fig. 8(b) shows the displacement of the rail and the wheel at the contact
point. In this situation, the wheel and the rail are always in contact with each other without any separation.
5. Conclusions
A model for simulating the dynamic interaction of the rail track and wagon system for steady state travel
of the wagon has been presented. The wheels of the wagon and the rail may or may not have defects. The
current model has been validated using four sets of data reported in the literature by several authors.
The model reported in this paper is capable of predicting the dynamic responses of both the wagon and
the rail track components. The model is also capable of examining the inuence of the properties of the rail
track and the wagon components on the impact forces and other dynamic responses of the rail track and
wagon system. It is possible to achieve design optimisation for the improved performance of wagon and
track using this model.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1355
In addition, the model can be easily reduced to a one wheeltrack model or one bogie-track model due to
its exibility of removing a few equations as demonstrated in Example 1. This can provide with the means
to investigate the eectiveness of various models for the vertical interaction of rail track and wagon system.
Table 6
Basic parameters for Example 4: Zhai (1996)
Notation Parameter Value
Wagon system
M
c
Wagon body mass (loaded)
a
75 300 kg
J
c
Wagon mass moment of inertia about Y axis
a
706 454 kg m
2
M
b
Bogie mass
b
3823 kg
J
b
Bogie mass moment of inertia about Y axis
a
1801 kg m
2
K
sc
Secondary suspension stiness
b
2300 kN/m
C
sc
Secondary suspension damping
a
30 kNs/m
K
pr
Primary suspension stiness
b
3740 kN/m
C
pr
Primary suspension damping
c
10 kNs/m
L
c
Distance between two bogie Y -direction centrelines
b
9 m
L
w
Wheelset base
b
2.9 m
r Wheel radius
b
0.625 m
Track subsystem
m
r
Rail mass per meter
b
60 kg/m
A
r
Rail cross-section area
a
7:77 10
3
m
2
E
r
Elastic modulus of rail
a
2:07 10
11
N/m
2
G
r
Shear modulus of rail
a
8:1 10
10
N/m
2
I
r
Rail second moment of area
a
2:94 10
5
m
4
r Rail prole radius on top
a
0.30 m
k Timoshenko shear coecient
a
0.34
m Poissons ratio of rail
a
0.27
C
p
Pad damping
c
70 kNs/m
K
p
Pad stiness
b
44 MN/m
C
f
Fastener damping
c
0.4 kNs/m
K
f
Fastener stiness
a
1 MN/m
m
s
Sleeper mass
b
251 kg
S
l
Sleeper spacing
b
0.544 m
C
s
Sleeper damping
c
130 kNs/m
K
s
Sleeper stiness
c
98 MN/m
R
t
Rail top prole radius
a
0.30 m
L
s
Eective length of rail seat support area
a
0.18 m
B
s
Eective width of rail seat support area
a
0.164 m
H
b
Ballast height
a
0.30 m
H
sb
Subballast height
a
0.15 m
E
b
Elastic modulus of ballast
a
130 MN/m
2
E
sb
Elastic modulus of subballast
a
200 MN/m
2
E
s
Modulus of subgrade
a
65 MN/m
3
Interface subsystem
C
H
Hertz spring constant
d
0:92 10
11
N/m
3=2
a
Assumed values based on a wagon and track system in Australia (M.J. McClanachan, Y.J. Zhang). The wagon body mass is
adjusted to have a static wheel force of about 100 kN.
b
The values given by Zhai (1996).
c
Assumed values (V.A. Prollidis, M.J. McClanachan, Y.J. Zhang).
d
The value is calculated using formula reported in this paper.
1356 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
From the examples reported in this paper it could be concluded that exact match between the eld data
and model would be possible only when all data, in particular the damping coecients of various com-
ponents, are precisely known. The examples also indicate that
(i) even though the periodic defect is more serious than the impulse excitation (due to the formers po-
tential to cause fatigue failure), the impulse excitation produces much higher impact forces (as a percent-
age of static loading);
Fig. 8. (a) Comparison of the contact force predicted by the current model for Zhai (1996) experiments (results of Zhai (1996) model
are also shown) and (b) comparison of the contact force predicted by the current model for Zhai (1996) experiments (results of Zhai
(1996) model is also shown).
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1357
(ii) the secondary suspension is not aected by the wheelrail impact forces and the eect of impact on
primary suspension is also marginal.
Acknowledgements
Central Queensland University and Queensland Rail nancially supported this research via a scholarship
awarded to the rst author. The support and encouragement of Professor Dudley Roach, Director of the
Centre for Railway Engineering and the assistance of Mr. Brain Hagaman, and Mr. Tim McSweeney of the
Queensland Rail are thankfully acknowledged.
References
Ahlbeck, D.R., Hadden, J.A., 1985. Measurement and prediction of impact loads from worn railroad wheel and rail surface proles.
ASME J. Engng. Industry. 107, 197205.
Ahlbeck, D.R., Meacham, H.C., Prause, R.H., 1975. The development of analytical models for railroad track dynamics. In: Proc.
Symp. on Railroad Track Mech. Pergamon Press, Oxford, pp. 239263.
Blader, F.B., Elkins, J.A., Wilson, N.G., Klauser, P.E., 1989. Development and validation of a general railroad vehicle dynamics
simulation (NUCARS). In: Proc. IEEE/ASME Joint Railroad Conf. Philadelphia, Pennsylvania, pp. 3946.
Cai, A., Raymond, G.P., 1994. Modelling the dynamic response of railway track to wheel/rail impact loading. Struct. Eng. Mech. 2 (1),
95112.
Cai, A., Raymond, G.P., 1992. Theoretical model for dynamic wheel/rail and track interaction. In: International Wheelset Congress,
Sydney, Australia, pp. 127131.
Clark, R.A., Dean, P.A., Elkins, J.A., Newton, S.G., 1982. An investigation into the dynamic eects of railway vehicles running on
corrugated rails. J. Mech. Eng. Sci. 24 (2), 6576.
Dahlberg, T., 1995. Vertical dynamic train/track interaction verifying a theoretical model by full-scale experiments. Vehicle Syst. Dyn.
Supplement 24, 4557.
Dong, R.G., Sankar, S., Dukkipati, R.V., 1994. A nite element model of railway track and its application to the wheel at problem.
Proc. Inst. Mech. Eng. 208, 6172.
Dym, C.L., 1973. Solid Mechanics: A Variational Approach. New York, McGraw-Hill.
Duy, D.G., 1990. The response of an innite railroad track to a moving vibrating mass. Trans. ASME J. Appl. Mech. 57, 6673.
Fermeer, M., Nielsen, J.C.O., 1995. Vertical interaction between train and track with soft and sti railpadsfull-scale experiments and
theory. Proc. Inst. Mech. Eng. 209, 3947.
Fr oohling, R.D., Tomas, M., Ebers oohn, W., 1997. Low frequency dynamic vehicle/track interaction: Instrumentation and measurement.
In: Proc. Sixth Int. Heavy Haul Conf., Cape Town, South Africa, pp. 462474.
Grassie, S.L., 1992. Dynamic models of the track and their uses. In: Kalkar, J.J., Cannon, D.F., Ohringer, O. (Eds.), Rail Quality and
Maintenance for Modern Railway Operation, Int. Conf. Delft, pp. 185202.
Grassie, S.L., Cox, S.J., 1982. The dynamic response of railway track to high frequency vertical excitation. J. Mech. Eng. Sci. 24 (2),
7790.
Grassie, S.L., Gregory, R.W., Harrison, D., Johnson, K.L., 1982. The dynamic response of railway track to high frequency vertical
excitation. J. Mech. Eng. Sci. 24, 7790.
Ilias, H., M} uuller, S., 1993. A discretecontinuous track-model for wheelsets rolling over shot wavelength sinusoidal rail irregularities.
In: Proc. 13th IAVSD Symp., pp. 221233.
Ishida, M., Miura, S., Kono, A., 1997. The inuence of track stiness on track dynamic behaviour. QR of RTRI 38 (3), 129134.
Jenkins, H.H., Stephenson, J.E., Clayton, G.A., Morland, G.W., Lyon, D., 1974. The eect of track and vehicle parameters on wheel/
rail vertical dynamic forces. Railway Eng. J. 3 (1), 216.
Johnson, K.L., 1985. Contact Mechanics. Cambridge University Press, Cambridge, MA.
Knothe, K.L., Grassie, S.L., 1993. Modelling of railway track and vehicle/track interaction at high frequencies. Vehicle Syst. Dyn. 22
(1993), 209262.
Lin, Y.H., Trerhewey, M.W., 1988. Finite element analysis of elastic beams subjected to moving dynamic loads. J. Sound Vib. 136 (2),
323342.
1358 Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359
Luo, Y., Yin, H., Hua, C., 1996. The dynamic response of railway ballast to the action of trains moving at dierent speeds. Proc. Inst.
Mech. Eng. 210, 95101.
McClanachan, M.J., 1999. Investigation of extreme wagon dynamics in central queensland coal trains, Master Thesis. Central
Queensland University, Australia.
Newton, S.G., Clark, R.A., 1979. An investigation into the dynamic eects on the track of wheelats on railway wagon. J. Mech. Eng.
Sci. 21 (4), 287297.
Ripke, B., Knothe, K., 1995. Simulation of high frequency wagon-track interactions. Vehicle Syst. Dyn. Supplement 24, 7285.
Sato, Y., 1977. In: High frequency track vibration and characteristics. Permanent Way, pp. 18.
Thompson, D.J., 1991. Theoretical modelling of wheel-rail noise generation. Proc. Inst. Mech. Eng. Part F 205, 137149.
Tunna, J.M., 1988. Wheel/rail force due to wheel irregularities. Proc. 9th Int. Wheelset Congress, Montreal, paper 6-2.
Prollidis, V.A., 2000. Railway Engineering. Printed in Great Britain at the University Press, Cambridge.
Zhai, W., Sun, X., 1993. A detailed model for investigating interaction between railway vehicle and track. In: Proc. 13th IAVSD
Symp., pp. 603614.
Zhai, W., 1996. Locomotive-and-track system coupling dynamics and its application to the study of locomotive performance. J. China
Railway Sci. (in Chinese) 17 (2), 5873.
Zhang, Y.J., Murray, M., Ferreira, L., 1998. A mechanistic approach for estimation of track modulus. In: Proc. Conf. Railway Eng.,
pp. 914.
Y.Q. Sun, M. Dhanasekar / International Journal of Solids and Structures 39 (2002) 13371359 1359

S-ar putea să vă placă și