Wear 265 (2008) 15491555

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

The dynamic wheelrail contact stresses for wagon on various tracks

Fujie Xia , Colin Cole, Peter Wolfs
Rail CRE, Faculty of Sciences, Engineering and Health, Central Queensland University, Rockhampton, QLD 4702, Australia

a r t i c l e

i n f o

Article history:
Accepted 16 January 2008
Available online 3 June 2008
Keywords:
Wheelrail contact stress
Creepages
Railway wagon system
Two-dimensional dry friction contact

a b s t r a c t
The maximal stress and tangential surface forces at the wheelrail contact elliptic area are affected by
the wheelrail contact dynamic load and creepages. Dynamic wheel load is related to the wagon dynamic
system, track and wheelrail interaction. Creepages are related to the motion of wheelset and wheelrail
contact parameters. The paper presents an analysis of the effects of creepages on wheelrail contact forces.
A complete Australia wagon with three-piece bogies was modelled and various tracks were selected for the
simulation of dynamic wheelrail contact stresses. The results show that the maximal normal wheelrail
contact stresses is under 1600 MPa in a range of conditions of typical normal operation.
2008 Fujie Xia. Published by Elsevier B.V. All rights reserved.

1. Introduction
Wheelrail contact is always a hot topic for railway vehicle
dynamics researchers and wheeltrack maintenance engineers.
The knowledge of dynamic wheelrail contact stress is useful in
assessing strength and fatigue life, wear of the wheel prole and
rail head, and as criteria to optimize new proles of wheel and rail.
The wheelrail stresses can be divided into surface contact
stresses, subsurface stress and strain eld. For a general case with
elastoplastic material and arbitrary geometries and loads, nite element method (FEM) or boundary element method (BEM) is often
used to get a good approximate solution [1,2]. FEM and BEM both
require considerable numerical effort and this limits the applications.
Considering elastic materials of the wheel and the rail and
rolling contact under vertical loading, Hertzian theory provides
the normal pressure distribution and Kalkers theory gives tangential surface contact forces (creep forces) [3]. As the surface forces
including normal stress and tangential creep forces can be calculated, the subsurface stress and stain eld can be determined. The
present paper investigates parameters affecting wheelrail contact
stresses and wheelrail rolling contact creepages using simulation
of dynamic wheelrail contact stresses with a complete wagon
model and various track conditions.
As the railway wagon system includes several nonlinear connections besides the wheelrail contact it is argued that the stickslip
motions of these nonlinearities and will affect the wheelrail
contact reaction forces. The complete wagon model developed

therefore includes nonlinear modelling of the wedge contact with

bolster and side frame, side frame contact with adaptersboth
belong to two-dimensional dry friction contact.
2. Stresses at contact surface
When a wheel and rail are brought into contact under the action
of the wheel load, the area of contact and the normal pressure distribution are usually expressed as half elliptic by Hertzian theory.
For a purely normal wheel load without tangential traction, the
state of stress at the surface is nearly hydrostatic on an elliptical
area with semi-axes a and b. Due to the wagon motion the effective wheel load is the sum of dynamic and static load components.
The wheel load will produce wheelrail rolling contact stress on the
contact area where the position of the contact area varies according
to lateral and yaw displacement of wheelset.
If only the maximum stresses in the rolling direction (x-axis)
are considered, the two-dimensional model of an innite cylinder
subjected to normal and tangential loading is often used for simplied analysis. In the case of full sliding, the stresses at the contact
surface due to both the pressure and the tangential traction can be
written as [4]:


x = p0


z = p0

x2
x
2 ,
a
a2

x2
,
a2


Corresponding author.
E-mail address: f.xia@cqu.edu.au (F. Xia).

xz = p0

0043-1648/$ see front matter 2008 Fujie Xia. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.wear.2008.01.035

(1)

(2)
x2
a2

(3)

1550

F. Xia et al. / Wear 265 (2008) 15491555

Outside the area of contact (|x/a| > 1),  z ,  xz are zero and the normal
stress in longitudinal direction is

x = 2p0

x
sign
a

xx
a

(4)

For the state of plane strain, the normal stress in lateral direction is
always
y = v(x + z )

(5)

with Poissons ratio . The sign in Eqs. (1), (3) and (5) is different for
driving and braking. The upper sign refers to the wheel in the case
of braking and to the rail in the case of driving, whereas the lower
sign refers to the wheel in the case of driving and to the rail in the
case of braking.
The maximum pressure is given by p0 = (3N)/(2ab), where N is
the normal load, a, b are the semi-axis dimensions of the contact
ellipse and  is friction coefcient.
For three-dimensional elastic bodies in rolling contact without
sliding friction, neglecting the effect of friction on normal pressure, then the normal stress and tangential surface forces can be
determined by Kalkers theory according to the calculated contact creepages and spin. Generally, these are adequate to estimate
fatigue and predict wear.
3. Factors for affecting the wheelrail contact stresses
For given materials of wheel and rail, and the proles of wheel
and rail, the factors affecting wheel rail contact stresses are only
two: normal load Nd and creepages. The former depends on vehicle
system, track structure and wheelrail interaction, the creepages
involve running speed and running state of the vehicle. Using a
rened vehicle model will provide more accurate normal load and
creepages. Modelling of the wagon will be discussed in Section 4.
Here we need to discuss the creepages calculation.
At present, there exist several formulae to calculate lateral
wheelrail contact creepages in literature. They can be divided into
two families. The main terms of one can be expressed as [5,6]:

yj
= () cos (j ),

j = 1, 2

(6)

The main terms of the other can be written as [79]:

yj =

()
cos (j )

(7)

where (*) stands for other terms of the lateral creepage, yj . The
difference between them can be expressed with their ratio:
yj

yj

1
=
cos2 (j )

Wtxj
V

Selecting the origin of coordinate system at the mass centre of

the wheelset with the x-axis positive forward, y-axis positive to
the right and z-axis positive downward, then the total velocity and
angular velocity of a wheelset can be expressed as [7,8]:
vw =

yj =

Wtyj
V

sj =

nj
V

ws =

where Wtxj , Wtyj stand for the relative longitudinal and lateral velocities between a wheel and rail, nj is the relative angular velocity
between a wheel and rail and V is the actual velocity of the wheelset.

 w + V
ro

T
(10)

z w

y w

w V
ro

Vw
ro

T
(11)

where x w , y w , z w stands for the velocity of the parasitic motion

of wheelset in longitudinal, lateral and vertical directions, respectively. w , w and  w , w are the rotational angles and rotational
velocities of the wheelset in roll and yaw direction, respectively.
w
stands for the perturbation of the angular speed of a wheelset and
ro is the wheel radius when wheelset is in its central position. bj
is the distance from mass centre of wheelset to wheelrail contact
point and R stands for the radius of the curvature of curved track.
We suppose the track to be rigid so the rail velocity is zero at the
wheelrail contact point. Then the velocity at the contact point is

w rcj
vcj = vw +

(12)

where
stands for screw symmetrical matrix:

w Vw

w =

r
 o


Vw
ro

0
 w + V
ro

ro

w V
ro
 w + V
ro

(13)

and

[ b
rcj

bj rj w

bj w + rj ]

(14)

Neglect the high order terms and then the resulting velocity vector
is

(9)

V + x w Vbj
R

and

(8)

where j = 1 stands for right-hand side wheelrail contact point and

j = 2 left-hand side.
Fig. 1 shows the increase of the ratio with the wheelrail contact
angle.
The difference in lateral creepage calculated by the two different
methods will increase with increase of contact angle, . For example, if the contact angle is 45 then the different is up to 2. Why
does there exists the two different formulae for the determination
of lateral creepage? The following analysis could be helpful.
The creepages are dened by
xj =

Fig. 1. The change of the ration of lateral creepages with two different methods.

vcj =

x w + rj
w bj

rj V

ro
y w rj  w V
z w bj  w bj V
ro

+
w
w

V Vbj
R

(15)

F. Xia et al. / Wear 265 (2008) 15491555

1551

The components of the vcj project onto the contact plane are

Wtxj
Wtyj
Wnj

1
0
0 cos j
0 sin j

0
sin j
cos j

x w + rj
w bj

rj V

ro
y w rj  w V
z w bj  w bj V

V Vbj

w
w

(16)

ro
If high order terms are neglected then the components become
Wtxj = x w + rj
w bj

rj V

ro

V Vbj

Wtyj

= (y w rj  w V

w )cos j

(17)

z w bj  w bj V

ro

sin j
(18)

Wnj = sin j (y w rj  w V

w) +

z w bj  w bj V
ro


w

cos j
(19)

There are two concepts: one considers the relative velocity in normal direction is not zero [5,6] and in that way with Eq. (9) the nal
creepages are
xj =

yj =

x w + rj
w bj

V
(y w rj  w V

rj
ro

+1

w )cos j

(y w rj  w V

bj

(20)

(z w bj  w bj V

w /ro )sin j

w )cos j

ro

sin j
cos j

(y w rj  w V

w)

(22)

Substituting the above Eq. in (18) we get

yj =

y w rj  w V

Ty
y + C
= 
Ty
y + C

(26)

where C = cC23 /C22 and c = ab depend on the radii a and b of contact ellipse; C22 , C23 are Kalker coefcients which depend on the
value a/b; the spin is dened by j = (sin( j ))/rj with rj being rolling
radius.
If we take the contact parameters as c = 4 mm, and C22 2.715,
C23 0.720,  = 0.1, then the change of the ratio of Ty /Ty with the
contact angle and C is shown in Fig. 2.
For the contact angle being less than 10 the difference between
the two lateral creep forces formulae is not large but with the
increases of the contact angle the difference increases quickly. For
the 40 contact angle the ratio of two lateral creepage forces reaches
1.7.
4. Wagon model

The other one Eq. (7) considers the normal velocity, Wn , is zero
[79] and then the lateral creepage can be obtained below.
From Wn = 0, Eq. (19) yields
w

The ratio of Ty /Ty will be

(21)

z w bj  w bj V

Fig. 2. The ratio of lateral creep forces with two different creepage formulae.

V cos j

(23)

we take the symbol + for j = 1 and for j = 2 in all the equations

in this section.
Which one should be selected? It is noted that for the creepages calculation we assume that for the wheelrail contact there
is no separation or penetration. Based on this formula (23) will be
selected in our wagon model. For the case of separation, the normal contact force is zero and there is no tangential contact forces,
in which case the creepages are not needed. It is also acknowledged
that it may be argued that the contact between wheel and rail can
be modelled by a Hertzian spring and penetration is permitted. In
that case, the rigid assumption of wheel and rail does not apply.
This will be a topic in our future investigation.
From Kalkers theory [3] the tangential wheelrail contact force
in lateral direction can be expressed as
Ty = Gc 2 (C22 y + cC23 )

(24)

Ty = Gc 2 (C22 y + cC23 )

(25)

A wagon with three-piece bogies was selected as it is widely

used in Australia. Roughly speaking, a wagon consists of 11 bodies: one wagon car body, two bolsters, four side frames and four
wheelsets. Each body in space has six degrees of freedom so there
are a total 66 degrees of freedom for a wagon system. As the connection between wagon car body and two trucks is through two
centre bowls, there are at least two constraints in the vertical direction. This kind of constraint also exists between side frames and
adapters (wheelsets), Fig. 3 gives eight constraints in vertical direction. If these constraints are included in our wagon model then
the system differential algebraic equations (DAE) must be used to
describe the system [10]. As an alternative, the constraints can be
replaced by spring connections with suitable stiffness as shown in
Fig. 3 [11]. In this way, the mathematical equations of the wagon
system become a simple set of ordinary differential equation (ODE).
This system of equations is much easier to solve. In this paper, we
will solve the system of ODEs.
The equation of wagon system can be written as
[M]X = Fn + Ft + Fw + Fg + Fc + Fd + Fs + Ff ,

(27)

The symbols in equation are below: M, system mass matrix; Fn ,

normal wheelrail contact force vector; Ft , tangential wheelrail
contact force vector; Fw , weight vector; Fg , gyroscopic force vector;
Fc , centrifugal force vector; Fd , damping force vector; Fs , spring force
vector; Ff , friction force vector.
To solve the system, rstly, the kinematical wheelrail contact
parameters were calculated prior to simulation by the program

1552

F. Xia et al. / Wear 265 (2008) 15491555

Fig. 5. Description of effective friction in Vampire.

and the breakout force is

Fig. 3. Contact between side frame and adapter.

FB = Fn ,
WRKIN [8] to form the wheelrail contact table which includes the
static wheel normal force as a function of the lateral and yaw of the
wheelsets. The wheelrail contact parameter table is then lookedup during the simulation. The effective normal wheel force was
determined by
Fnd =

2/3
(Fn0

3/2
2/3
+ Kh qd )

(28)

where qd is dynamic penetration, Fn0 stands for static wheel load

and Kh is Hertzian spring stiffness. The dimensions of the contact
ellipse are then given by
ad = a0

 F 1/3
nd

Fn0

bd =b0

 F 1/3
nd

Fn0

ad bd =a0 b0

 F 2/3
nd

Fn0

(29)

where a0 , b0 is the dimension of elliptic contact area to the static

wheel load, Fn0 . With the known contact dimension the tangential
wheelrail contact force can be determined by Kalker theory-based
formulae, e.g., SHEs formulae [12].
As a three-piece bogie uses a friction wedge damper system to
damp out the vibration of the system, the phenomenon of contacting with friction should be reasonably described. Due to the lateral
and vertical motion of a bolster relative to side frames the friction on the surfaces of the wedge is two-dimensional. According to
friction nature, the relative motion between two contact bodies is
stickslip motion. The same case exists for the side frame contacting
with adapters.
There are several ways to describe friction. The friction element
used in VAMPIRE is described as shown in Fig. 4 [11]. The series
force is
Fsc = Fks + Fcs ,

(31)

Finally the effective friction can be expressed as (see Fig. 5)

Ff =

Fsc , |Fsc | FB
,
FB sgn(d), |Fsc | > FB

where d stands for resultant relative displacement.

It is known that static and kinetic friction coefcients are different, for dry friction the static friction coefcient is normally larger
than kinetic friction coefcient. As a consequence, the kinetic friction force is usually less than static friction force. Using the friction
element in VAMPIRE the kinetic friction force can only be equal to
the breakout force (static friction force). The condition of the absolute value of the force FSC being greater than the absolute value
of the breakout force prevents accurate modelling, because when
the relative velocity is larger than zero (or a certain small value)
then the friction force should be determined by kinetic friction

conditions (i.e a constant less than FB , but not necessarily ks x + cs x).

There is another way to treat this problem using a continuous
friction model by varying friction coefcient, e.g.,
 = (k + (1 k )sec h[1 vr ])tanh[2 vr ]

(33)

where 1,2 are positive coefcients. This technique approximates

the friction force near zero relative speed and at vr = 0 the friction
is assumed to be zero. That is not usually true because the friction force is generally not zero when the connection is stuck as we
discussed above.
Another method of modelling the friction is called the rened
method. For the rened method, we take the system shown in Fig. 6

(30)

Fig. 4. Friction element used in Vampire.

(32)

Fig. 6. A two DOF system.

F. Xia et al. / Wear 265 (2008) 15491555

1553

Fig. 7. The friction force description for stick and slip modes of the new friction element.

as an example, the relation between the slip motion and the friction
force is determined by
Fk = Nk sgn(x 1 x 2 ).

(34)

In this case, the system has two degrees of freedom. For the stick
situation, that is, when the relative velocity between m1 and m2 is
zero, the friction force is then determined by [8,11]
Fs

1
=
[m2 F1 + m1 F2 ].
m1 + m2

(35)

The system has one degree of freedom during the stick mode.
The switch conditions governing the transition from stick mode to
slip motion are
x 1 x = 0

(m2 F1 + m1 F2 ) (m1 + m2 )Ns .

(36)

If the above conditions are true the motion of the system is stuck,
and vice versa.
This method will cause a very complicated formulism for the
railway wagon system [8].
In the present paper, we take the method used in VAMPIRE as
a basis, to develop another suitable two-dimensional friction element [13]. The principle can be shortly expressed as (see Fig. 7)


Ff =

ks d, ks d Ffs and |vr | vr ,

Ffs ,
ks d > Ffs and |vr | vr ,
Ffk ,
|vr | > vr .

(37)

where Ffs = Ns sign(vr ) stands for static friction force; Ffk =
Nk sign(vr ) is kinetic friction force; d stands for resultant relative displacement and vr is relative velocity and vr stands for a
small value of relative velocity for numerical analysis requirement.
Static and kinetic friction coefcients are represented by s and k ,
respectively.
For the two-dimensional case the components of friction forces
in x and y directions from Eq. (37) are

Ffx =

ks dx ,

Ffs dx ks d Ffs and |vrs | vr ,

, k d > F and |v | v ,
s
rs
r
fs

Ffk vrx , |vrs | > vr .

(38)

vrs

Ffy =

ks dy ,

Ffs dy ks d Ffs and |vrs | vr ,

, k d > F and |v | v ,
s
rs
r
fs
d
Ffk vy
|vrs | > vr .
,
vrs

(39)

where d, vrs are dened by

d =

dx2 + dy2 ,

vrs =

2 + v2
vrx
ry

(40)

and for this case Ffs = Ns and Ffk = Nk .

In the same way, the friction torque acting on joint can be written
as

Tf =

k , k  Tfs and |r | r ,
k  > Tfs and |r | r ,
Tfs ,
Tfk ,
|r | > r .

(41)

where k stands for torsion spring stiffness connecting two relative

rotation bodies;  is relative rotation angle; r stands for relative
rotation velocity; r is a small value of relative rotation velocity
required numerically; Tfs stands for static friction torque and Tfk for
kinetic friction torque.
The wagon model has 66 degrees of freedom and the model is
implemented with C++, so we call it the C66 model.
5. Simulation results
A hopper wagon is used for this investigation. Some parameters
of the wagon are listed below.
Semi-spacing of truck
Half axle spacing of wheelset
Lateral semi-spacing of primary suspension
Wheel radius
Semi-spacing of side support
Truck distance
Wedge static friction coefcient
Static load on wedge friction surface
Car body mass (empty/loaded)
Car body roll inertia (empty/loaded)
Car body pitch inertia
Car body yaw inertia
Side frame mass
Side frame roll inertia
Side frame pitch inertia
Side frame yaw inertia
Bolster mass
Bolster roll inertia
Bolster pitch inertia
Bolster yaw inertia
Wheelset mass
Wheelset roll and yaw inertia
Wheelset pitch inertia

5.18 m
0.838 m
0.8 m
0.425 m
0.616 m
14.820 m
0.4
20.0 kN
8.1/66.10 Mg
10.4/85.58 Mg m2
79.3/647.18 Mg m2
80.0/652.98 Mg m2
0.447 Mg
0.101 Mg m2
0.1156 Mg m2
0.1156 Mg m2
0.465 Mg
0.175 Mg m2
0.115 Mg m2
0.176 Mg m2
1.12 Mg
0.4201 Mg m2
0.1 Mg m2

The combination of the proles of wheel and rail is ASLW3/AS60.

In order to verify the C66 model we did comparison of the results
between C66 and Vampire in many cases, Fig. 8 shows the case of
vertical sinusoidal track irregularity.

1554

F. Xia et al. / Wear 265 (2008) 15491555

Table 1
Simulation cases
No.

Track irregularity

Empty or loaded car

Speed (m/s)

Radius of curve (m)

Elevation of
outside rail (mm)

Maximal normal
stress level (MPa)

og. creep
force (kN)

ateral creep
force (kN)

1
2
3
4
5
6
7
8
9
10
11
12

Class 6
Class 5
Class 4
Class 6
Class 5
Class 4
None
None
None
None
None
None

Empty
Empty
Empty
Loaded
Loaded
Loaded
Empty
Empty
Empty
Loaded
Loaded
Loaded

20
20
16
20
20
20
20
20
20
20
20
20

200
500
1000
200
500
1000

0
0
0
0
0
0
120
120
100
120
120
100

630730
550760
450800
10401095
10201120
10001135
5001000
650665
675700
9001400
9601120
10201090

3 to 3
4 to 5
4 to 6
7 to 7
11 to 11
15 to 15
1 to 4.5
04
03.5
5 to 25
1 to 19
3 to 13

1 to 2
3 to 4
5 to 5
5 to 5
10 to 10
10 to 10
0.8 to 4.5
0.8 to 3.8
0.8 to 3.5
5 to 20
3 to 18
010

Fig. 9. wheelrail contact stresses of leading wheelset for empty wagon on class 5
track.

Fig. 8. Results comparison of C66 and VAMPIRE. Top: vertical wheelrail contact
force of leading wheelset. Bottom: maximal contact stress of leading wheelset.
Excitation: vertical track irregularity with z = 10 sin(2s/20) mm, running velocity
V = 20 m/s. Empty wagon is used.
Fig. 10. RHS wheelrail contact creep forces of leading wheelset for empty wagon
on class 5 track.

We use the power spectral density (PSD) formulae of Federal America Railroad (FAR) to generate track irregularity from
track class 46 [5]. For the page limitation here we only provide
the results for the empty wagon on class 5 track as shown in
Figs. 9 and 10.
More simulation results are shown in Table 1. As the dynamic
wheelrail contact stresses vary with time, only the ranges of the
minimal and maximal values are presented.

According to BR standard criteria for wagon the normal stress

should be less than 1600 MPa. From the results in Table 1 all values
of normal stresses are under the criteria level.
6. Conclusion
The rolling contact stresses are dependent on the wagon dynamics as illustrated by the different stress levels resulting from the

F. Xia et al. / Wear 265 (2008) 15491555

wagon running on various tracks. The normal stress levels appear to

be acceptable for class 46 tracks. Rougher track, for example, class
2 or 3 may result in normal stress levels being exceeded. Further
investigation is planned.
Wheelrail rolling creepages play a key role to determine creep
forces (tangential surface forces). It is noted that different results
will be obtained depending on the choice of creepage formula. The
differences in results increase with increasing wheelrail contact
angle. The results show that for the contact angle less than 10 the
difference between the two lateral creepage formulae on the effect
of lateral creep force is not obvious, but with the increase of contact
angle the difference increases quickly. For the 40 contact angle the
ratio of two lateral creepage forces reaches 1.7.
A 66 DOF wagon system dynamic model was developed with
a typical Australia hopper wagon to simulate wheelrail contact
stresses on various tracks. The model included stick and slip modes
of the two-dimensional dry friction behaviour on the surfaces of
wedge damping systems and axle adapters so that the rail contact
stresses calculated include the effects of nonlinear wagon dynamics.
Twelve cases for wheelrail contact stresses were simulated. The
results show that the wheelrail maximal normal contact stresses
of Australia wagon for the track surface irregularities simulated are
under the limitation (1600 MPa) in all the cases.
Acknowledgments
The paper is supported by the Cooperative Research Centre
for Railway Engineering and Technologies of Australia (Rail CRC)
project 147. The work is also obtained support from the Centre for

1555

Railway Engineering (CRE) at Central Queensland University and

Faculty of Engineering and Physical Systems, Central Queensland
University.
References
[1] M. Akama, T. Mori, Boundary element analysis of surface initiated rolling
contact fatigue cracks in wheel/rail contact systems, Wear 253 (2002)
3541.
[2] M. Pao, et al., Distribution of contact pressure in wheelrail contact area, Wear
253 (2002) 256274.
[3] J.J. Kalker, Three-dimensional Elastic Bodies in Rolling Contact, Kluwer Academic Publishers, 1990.
[4] M. Ertz, K. Knothe, Thermal stress and shakedown in wheel/rail contact, Arch.
Appl. Mech. 72 (2003) 715729.
[5] V.K. Garg, R.V. Dukkipati, Dynamics of Railway Vehicle Systems, Academic
Press, 1984.
[6] G. Chen, W.M. Zhai, A new wheel/rail spatially dynamic coupling model and its
verication, Vehicle System Dynamics, vol. 41, no. 2, pp. 301322.
[7] J.A. Elkins, Prediction of wheel/rail interaction: the state-of-the-art, Vehicle
Syst. Dyn. 20 (Suppl.) (1991).
[8] F. Xia, The Dynamics of the Three-piece-freight Truck, Ph.D. Thesis, Informatics and mathematical modelling, the Technical University of Denmark,
2002.
[9] A.D. De Pater, P. Meijers, I.Y. Shevtsov, Simulation of the motion of a railway vehicle along curved tracks, Part I. General theory, Report no. LTM 1196, Laboratory
for Engineering Mechanics, Delft University of Technology, 1999.
[10] E. Eich-Soellner, C. Fuhrer, Numerical Methods in Multibody Dynamics, B.G.
Teubner, Stuttgart, 1998.
[11] P. Klauser, Modeling three-piece truck using VAMPIRE, in: VAMPIRE North
American User Conference, Chicago, IL, 2003.
[12] Z.Y. Shen, J.K. Hedrick, J.A. Elkins, A comparison of alternative creep-force
models for rail vehicle dynamic analysis, in: Proceedings of the Eighth IAVSD
Symposium, MIT, Cambridge, MA, 1984.
[13] F. Xia, C. Cole, P. Wolfs, The effects of the friction between frame and adapters
on the performance of three-piece truck, in: Proceedings of the ASME/Joint Rail
Conference, Atlanta, GA, America, April 46, 2006.