Proceedings of RTD2002

2002 ASME/IEEE Joint Rail Conference

April 23-25, 2002, Washington, DC, USA

RTD2002-1642

MULTIBODY SIMULATION OF A FREIGHT BOGIE WITH FRICTION DAMPERS

N. Bosso, Politecnico di Torino A. Gugliotta, Politecnico di Torino

A. Somà, Politecnico di Torino

ABSTRACT
Freight bogies are often built using friction element instead of NOMENCLATURE
viscous dampers with great advantage regarding cost and X : longitudinal direction (running direction)
manutenibility. On the other hand this technical choice leads to Y : lateral direction (wheelset axle direction)
low performance of the vehicle. Z : vertical direction (gravity direction)
The aim of this work is to evaluate the behaviour of a Y25 V : velocity
freight bogie that is the most diffuse in Europe where friction χ : Friction force/relative velocity linear
dampers are used to reduce vertical and hunting vibrations. coefficient.
The purpose is also to improve a bi-dimensional model of µ : friction coefficient.
friction damper and to test its numerical efficiency on a whole N : Normal load.
vehicle model in a multibody code. θ : auxiliary angle : velocity direction in the
friction plane.
INTRODUCTION α : link inclination.
In this work will be analysed the dynamical behaviour of a l : link lenght.
freight railway vehicle with Y25 bogies using a multibody code. Kx, Ky, Kz : Inner spring stiffness in x,y,z direction.
To be able to model and simulate freight bogies it is necessary Kx2, Ky2, Kz2 : Outer spring stiffness in x,y,z direction.
to introduce a numerically stable and efficient model of the FxL, FyL, FzL : Left Spring force in x,y,z direction.
friction elements, which always are employed in this kind of Fxo, Fyo, Fzo : Outer Springs force in x,y,z direction.
vehicles. M : Bogie mass (Matlab Model).
In fact the simulation of friction phenomena lead to the ψ : Wheelset yaw angle.
introduction of a discontinuity in the model which is often cause ψ0 : theoretical wheelset yaw angle (curve radial
of numerical instability when using a multibody code. direction).
Therefore in this paper will be defined a frictional model Cij : Kalker coefficients.
thorough an analytical continuous function able to improve the g* : Slant inclination for the pivots slant.
numerical stability of the simulation. g+ : Slant inclination for the bogie slant.
This model has been then extended to simulate a bi-dimensional Q0 : Nominal Load.
friction surface. Once introduced in the vehicle model has been Q1 : Wheel unloading due to the bogie slant.
studied the dynamic behaviour of the vehicle regarding in term Q2 : Wheel unloading due to the pivot slant.
of running stability and curving to verify the response of the Qanc : Wheel unloading due to non compensated
friction model. Furthermore has been evidenced that the acceleration.
methods usually adopted for the investigations on vehicles with Ymax : Maximum lateral force allowed by the track.
viscous dampers are not adequate when friction dampers are Anc : non-compensated lateral acceleration.
used.

1 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1. FRICTION MODEL lateral direction (see fig. 3), so it is necessary to extend the
The representation of the friction forces following the friction force formulation to two degrees of freedom.
classical Coulomb’s law. Often this formulation lead to
numerical simulation problem, due to the discontinuity
introduced by the friction force behaviour.

Ff
Friction force Ff,z

Ff,y Vy Y

Relative velocity
V
Vz

C o n tinuos function

If-E l s e f u n c t i o n
Z

Figure 1: Friction force patterns. Figure 3: Friction force components in the ZY plane.

To avoid this problem the friction force has been modelled The two components of the vertical and lateral velocity are
with a continuos function of the relative velocity between the z& and y& ; the absolute value of the relative velocity is:
friction surface described by the following equation:
V ⋅χ V = z& 2 + y& 2 (2)
Ff = (1)
2
V ⋅χ 
1 +   The components of the velocity are z& = V ⋅ cos(θ ) and
 N ⋅µ  y& = V ⋅ sin(θ ) , while the components of the friction force in
the z and in the y direction are:
limV →0 F f = V ⋅ χ
limV →±∞ F f = ± N ⋅ µ
V ⋅ χ ⋅ cos(θ ) z& ⋅ χ
Where Ff is the friction force, V is the relative velocity. The F f , Z = F f ⋅ cos (θ ) = = (3)
2 2
χ parameters represent the angle between the velocity axis and V ⋅χ  V ⋅χ 
1 +   1 +  
the friction force curve around the origin, µ is the kinetic  N ⋅µ   N ⋅µ 
friction coefficient, N is the force normal to the friction
surfaces.
V ⋅ χ ⋅ sin (θ ) y& ⋅ χ
F f , Z = F f ⋅ sin (θ ) = = (4)
2 2
V ⋅χ  V ⋅χ 
1 +   1 +  
 N ⋅µ   N ⋅µ 
A

Y
2. Y25 MODEL
X

Z In the Y25 bogie the damping of vertical motion is given

Figure 2: Friction surface on the axle-box
In the case of the Y25 bogie the friction surfaces allow the
relative motion on the plane which include the vertical and the
by a mechanical device placed in the primary suspension called
Lenoir-link, that use a portion of the wagon weight to support
the normal load to the friction surfaces, so vertical damping is
also load sensitive.

2 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The secondary suspension is reduced to a centre pivot with • Bogieframe: the bogieframe is connected to ground
a very high stiffness, and the damping of hunting mode is given through a traslational joint, which allow only the vertical
by a couple of sidebearers, two friction surface mounted motion (Z).
outboard of the bogie pivot and preloaded with springs. The • Lenoir-Link (2 bodies): The Lenoir-link has been built as
sidebearers also support the rolling stiffness of the car. two separate parts, one linked to the bogie and the other to
Therefore two kinds of friction elements are present in the the spring holder, both with two revolute joints. The two
vehicle and both are load sensitive. parts are then linked each other with a translational joint
Owing to the high complexity of the Lenoir Link and a single force, which act as a unilateral bumpstop.
substructure, we have reduced it to a transfer function When a force is applied to the spring holder by the spring
introduced between the axle and the axle-box. the two parts of the link are moved away and the bumpstop
In order to test the function we have simulated a single operate such that the force is transferred to the bogie.
axle-box with a detailed model realised with ADAMS (“detailed • Spring holder: The Spring holder keeps the inner spring in
axle-box Model”), then we have introduced the transfer the left side of the axle-box; it is connected to the Lenoir-
function of the Lenoir-link in a Matlab Model comparing the Link as shown above. The Link inclination split the force
result. supplied by the spring in two components in the X-Z plane.
The transfer function is then adopted to create the model of The spring-holder is connected to the left side of the pusher
the entire wagon used to perform the simulations with with a bumpstop so that the force given by the Lenoir-Link
Adams/Rail. in the X direction is transferred to the pusher itself.
• Pusher: the pusher is connected to the bogieframe with a
3.1- ADAMS detailed model of the axle-box traslational joint which allow only the relative motion in
The ADAMS detailed model of the axle-box is described in the X direction. The right side of the pusher is connected to
the following figure. the axle box with a force vector, this element model the
first friction surface. In the X direction the force vector act
Spring as a bumpstop, the value of the X force is then used as the
Bogieframe Lenoir Link Holder Normal force for the bi-dimensional friction force
implemented in the Y and Z direction.
• Axle-Box: The primary friction surface is located in the
left side of the Axle-Box in the Y-Z plane. This surface has
been modeled using four force vectors, one to each vertex
of the surface. Each Force vector is modeled as the one on
the pusher, so that the total friction force is distributed
among the four forces depending on the normal force
acting on each vertex. The vertical load is transferred from
the Axle-Box to the bogie frame thought four springs, the
Axle - Box outer springs have a gap which is closed only in the laden
Friction condition, so that in the tare load condition only the inner
Surfaces Pusher springs support the load. Moreover the right inner spring
LINK DETAIL act thought the Link as described above. The axle box is
connected to the axle with a revolute joint.
Spring
• Axle: The axle in this model is connected to ground with a
Holder
planar joint.
Traslational All the bumpstop elements have been modeled with a
Joint Z stiffness of 1⋅108 N/m and the damping of 1000 N⋅s/m.
Revolute
Joints X
Force
element 3.2 Equivalent model of the Lenoir Link (Matlab)
Bogieframe
The purpose of the analytical model of the axle-box is to
create a transfer function between the bogie and the wheelset.
Figure 4: Axle box detailed model. In the following we will find the equation of the forces
exchanged between bogieframe and axle-box due to the link,
the friction force and the springs.
The model is composed of 7 rigid bodies:

3 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 z = l2 − x2 (7)

Differentiating eq. (6):

dz −x x
= =− = tanα (8)
dx l −x
2 2 z

For small displacement, when ∆x ≅ dx and ∆z ≅ dz, it is

possible to write:
∆z = ∆x ⋅ tan α (9)

A vertical displacement ∆z lead to:

Fx
Fz = k z ⋅ ∆z (10)
FX = FZ tan α = k Z ⋅ ∆z ⋅ tan α (11)

An horizontal displacement ∆x, since ∆z = tanα⋅∆x (eq. 9),

lead to:
Fz = k z ⋅ ∆z = k z ⋅ ∆x ⋅ tan α Fz = kz⋅∆z (12)
Fz
Introducing the equation (5) the equation (12) become:
Dx FX = k z ⋅ ∆x ⋅ tan 2 α (13)
Finally the forces acting between the bogieframe and the
axle-box due to the link coupling are:

αo Dz  Fz = k z ⋅ ∆z + k z ⋅ tan α ⋅ ∆x

 Fx = k z ⋅ tan α ⋅ ∆x + k z ⋅ tan α ⋅ ∆z
2
α
(14)
 F = k ⋅ ∆y
X  y y

To complete the model is required to consider both the

contribution of the friction force, of the left spring and of the
two outer springs.
Z The force normal to the friction surfaces is given by the Fx
force reported above. In the following we will set N=Fx. The
friction forces are given by (the factor 2 is due to the presence
Figure 5: The Lenoir link and the exchanged force of 2 friction surfaces):

The Lenoir link, due to its inclination with the vertical χ ⋅ Vz

direction in the vertical-longitudinal plane, couples the stiffness F fz = 2 ⋅ (15)
in z and x direction.  χ ⋅V 
1 +  
 N ⋅µ 
The vertical and horizontal components of the force, as it
appears in figure n.5 are linked with the relation:
FX = FZ ⋅ tan α χ ⋅ Vy
(5) F fy = 2 ⋅ (16)
 χ ⋅V 
The rotation of the link around the lower hub describe a 1 +  
circumference defined by the following relation:  N ⋅µ 
x2 + z2 = l2 (6)
The left spring contribution is given by:
The vertical (z) displacement can be expressed as:

4 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

  FxL = k x ⋅ ∆x The two models have the same behaviour, so in the model
 of the entire vehicle the Axle-Box force model has been
 FyL = k y ⋅ ∆y (17) implemented. These choice has been adopted since the detailed
 ADAMS Model require a large computation time for the single
 FzL = k z ⋅ ∆z Axle-box and makes ineffective its application to an
The outer springs contribution is given by: ADAMS/Rail Vehicle Model.
 Fzo = 2 ⋅ k z 2 (∆z − ∆z 0 ) if ∆z ≥ ∆z 0
 (18) 3.4 VEHICLE MODEL
 Fzo = 0 if ∆z < ∆z 0
The vehicle model has been modelled using the function
Finally the forces between the bogie and the axle-box are described in par 3.2 to simulate the connection between each
given by: axle-box and the bogie. The function of Fx, Fy, Fz have been
directly introduced in a single Force Vector element. However
has been necessary also to introduce a contribute to simulate the
 FX = k z ⋅ tan 2 (α ) ⋅ ∆x + k z tan (α ) ⋅ ∆z + k x ⋅ ∆x closure of the gap present in the axle-box.

 FY = 2 ⋅ k y ⋅ ∆y + 2 ⋅ F fy (19) This contribute has been neglected during the previous
simulations.

 FZ = 2 ⋅ k z ⋅ ∆z + k z ⋅ tan (α ) ⋅ ∆x + Fzo + 2 ⋅ F fz
Bogie

The system analyzed with Matlab is described by the

following equations:
F2

M ⋅ &x& = − Fx F1 Fs Fn X

M ⋅ &y& = − Fy (20) F2

M ⋅ &z& = − Fz

Pusher
3.3 Comparison Friction Surfaces
Axle-box
Z
The two Model described before have been compared. The
free response of the system to a to a vertical and lateral impulse Figure n. 6: Axle-box Longitudinal bumpstops.
has been investigated.
The maximum value of displacements is reported on table 1 The forces indicated as F1 and F2 on figure 6 have been
and 2. added to the force Fx and to the Normal Force, their behaviour
The simulations are made in the tare condition. is the same of a bumpstop with a linear stiffness of 10E8 N/m.
The impulse is supplied by the Vertical load itself for the The secondary suspension is made by a centre pivot with a
vertical motion, the lateral impulse is supplied instead with an very high stiffness (see table 4) in each translational direction,
initial velocity of 0.1 m/s. which allow the tree rotation. The roll torsional stiffness is
supported by two sidebearers (one for each bogie) placed at a
ADAMS Matlab distance of 850 mm from the pivot in the lateral direction and
Max. Vertical displacement [m] 0.0391 0.0387 preloaded each with the 31% of the tare load.
Mean Vertical displacement [m] 0.0202 0.0203
Frequency [Hz] 3.34 3.33 sidebearer Centre –pivot
Table 1- Vertical Motion

ADAMS Matlab
Max. Longitudinal displacement [mm] 0.071 0.070
Frequency [Hz] 2.96 2.99
Table 2 – Lateral Motion

The simulation are performed using the following

parameters:
ADAMS : GSTIFF / digit=7
Matlab : Ode15s / abs. error =1E - 6 / rel. Error = 1E - 8

5 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Figure 7 – Transversal section of the bogie. bogie. This second event is shown in fig. 9 and is indicated as
bogie slant.
The sidebearers also supply the Yaw damping acting as The third effect we have considered is the one due to the
friction elements. The same model described on par. 2 as been lateral acceleration non-compensated by the cant (anc) which,
adopted and implemented on a Force Vector, which has a linear for freight vehicles, can reach a value of 0.6 m/s2 during
stiffness in the vertical direction and a friction damper in the curving. The effect of anc is an unloading of all the wheels of
XY plane. The Normal force is the one exchanged by the the inner side of the curve and a loading of the wheel of the
stiffness in the Z direction. outer side.
In the following tables 3 and 4 are reported the inertial and The simulations have been made using the Test-Rigs of
stiffness data of the model. ADAMS/Rail 9.1.1. The given superelevation has been chosen
according to the reports ORE B55 RP 6 / RP 8 which stand a
limit value for the maximum grade that may be found in the
Inertial data Mass Jxx Jyy Jzz
2 2 2 railway for the bogie slant (g+) and for the slant between the
Body [Kg] [Kg m ] [Kg m ] [Kg m ]
pivots (g*) :
Wagon (tare) 12400 8.4E4 1.3E6 1.28E6 15 5
Wagon (Laden) 172400 9.7E5 1.5E7 1.4E7 g* = + 2.0 =4 0 00 g + = 7.0 − = 8 0 00
2070 1400 2100 2400
2a * 2a +
Bogie frame
Wheelset 1225 750 140 750
Where 2a* is the distance between the two pivots and 2a+
Table 3: Inertial Data. is the distance between the two axles of a bogie. These
inclinations bring the following vertical displacement values:
Stiffness Kx Ky Kz Dz* = 36.8 mm for the slant between the pivots.
[N/m] [N/m] [N/m] Dz+ = 14.4 mm for the bogie slant.
Primary suspension 500000 500000 175000 Next we denote each wheel of the vehicle as “ij” (e.g. 12)
Internal spring 655000 where “i” is the number of the wheelset to whom the wheel
Secondary suspension belong and “j” is the side (1 = right, 2 = left).
Centre-pivot 10E8 10E8 10E8 The following figure shows the way we lift/lower the rigs:
SIDE BEARERS 380000 580000
Table 4: Stiffness.

4 - SIMULATIONS

In the following sections are reported all the simulation

performed on the Vehicle Model using ADAMS/Rail.

4.1 - Slant Test

Railway vehicles running on curved tracks are subject to Figure 8: Slant between the pivots.
the altimetrical differences between the two rails imposed by
the cant angle. For this reason, vehicles with very high torsional
stiffness have serious trouble to cross these kinds of defects,
cause the vertical load on some wheels may decrease with
derailment risk.
In our work we have considered three different events
which may cause the wheel unloading and which usually occur
together during curving.
The first event is caused by a slant between the two rail
extended to the entire side of a wagon, and that lead to a
superelevation of only one side (left/right) of a bogie respect to Figure 9: Slant of the bogie.
the second bogie. We indicate this events as slant between the
pivots (fig. 8). Usually the skill of a vehicle to cross a slant is checked
The second event is a slant between the rails with a shorter thought static test, however we thought it right to made dynamic
extension, which is applied to a single bogie lifting the first test (simulated) so to keep in account the contribution of the
wheel and pulling down the second of only one side of the friction force supported by the Lenoir-Link which increase the
wheel unloading.

6 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1 m/s2, Q1 is the load after the application of the bogie slant, Q2
V=0.15 m/s
the load due to the application of the slant between the pivots.
V=0.15 m/s DQ shows the load difference of each case respect the
V=0.1 m/s
V=0.05 m/s reference load.
Quasi-static The report ORE B55 RP 8 fix a limit of 0.8 (80%) to the
maximum admissible DQ/Q0 ratio. The total DQ/Q0 ratio is
obtained summing the contribution of Qanc, Q1 and Q2.
It is necessary to point out that the bogie in the tare case is
very close to the limit imposed by the Norms. In the Laden
condition the situation is quite better. The main reason of the
behaviour of the bogie during the Slant simulations may be
found in a very high value of the torsional stiffness of the bogie
primary suspension in the vertical direction. This is clear from
the bad loading distribution due to the bogie slant, while the
Figure 10: Effect of the lowering velocity on the wheel slant between the pivots is not critical.
unloading.
Wheel 41 42 31 32 21 22 11 12
On Fig. 10 is shown the load on the wheel 41 while the rig Q0
111.7 111.8 111.8 111.8 111.8 111.8 111.8 111.8
where this wheel lay is lowered using different values of the [KN]
velocity. The first curve is obtained thought a quasi-static Qanc
89.2 134.2 89.3 134.2 89.3 134.2 89.3 134.2
analysis. Is clear that over a certain velocity arise dynamic [KN]
effects, therefore we have performed all the simulations using a DQanc
22.5 -22.5 22.5 -22.5 22.5 -22.5 22.5 -22.5
velocity of 0.1 m/s, in this way we consider a quantity of the [KN]
DQanc/Q0
friction force effort to the unloading without introducing 20.1 -20.1 20.1 -20.1 20.1 -20.1 20.1 -20.1
[%]
relevant dynamic contributions.
Q1
109.3 114.1 109.5 114.1 105.8 117.8 105.9 117.7
[KN]
Wheel 41 42 31 32 21 22 11 12 DQ1
2.4 -2.3 2.4 -2.3 5.9 -6.0 5.9 -6.0
Q0 [KN]
26.5 26.5 26.5 26.5 26.5 26.5 26.5 26.5
[KN] DQ1/Q0
2.1 -2.1 2.1 -2.1 5.3 -5.4 5.3 -5.3
Qanc [%]
23.0 30.0 23.0 30.0 23.0 30.0 23.0 30.0
[KN] Q2
95.9 126.7 127.9 96.5 111.8 111.8 111.8 111.8
DQanc [KN]
3.5 -3.5 3.5 -3.5 3.5 -3.5 3.5 -3.5 DQ2
[KN] 15.9 -14.9 -16.2 15.3 0.0 0.0 0 0
[KN]
DQanc/Q0
13.2 -13.2 13.2 -13.2 13.2 -13.2 13.2 -13.2 DQ2/Q0
[%] 14.2 -13.4 -14.5 13.7 0.0 0.0 0.0 0.0
[%]
Q1 DQ/Qtot
24.9 28.1 24.9 28.1 28.5 24.5 28.5 24.6 36.5 -35.6 7.8 -8.5 25.5 -25.5 25.4 -25.4
[KN] [%]
DQ1 Table 6: Slant – Laden.
1.6 -1.6 1.6 -1.6 -1.9 1.9 -1.9 1.9
[KN]
DQ1/Q0
6.0 -6.1 6.1 -6.0 -7.4 7.4 -7.3 7.4
[%] 4.2 - Riding stability
Q2
13.2 40.0 40.2 13.3 26.5 26.5 26.5 26.5
[KN] As already say, the vehicle has a number of friction
DQ2 elements in the primary and in the secondary suspension. The
13.3 -13.5 -13.6 13.2 0.0 0.0 0.0 0.0
[KN] presence of these elements, which moreover act a big influence
DQ2/Q0 on the stability of the vehicle, make inadequate to find the
50.2 -51.0 -51.4 49.8 0.0 0.0 0.0 0.0
[%] critical speed trough a series of eigenvalue analysis, cause the
DQ/QTot heavy non linearity of the friction elements model.
69.5 -70.3 -32.1 30.5 5.9 -5.8 5.9 -5.9
[%] Therefore the simulations to find the critical speed have
Table 5: Slant –Tare. been made trough transient non-linear analysis on a straight
On the following Table 1, 2 are shown the unloading of track, giving a lateral impulse to the wheelset to be able to
each wheel in the tare and in the laden condition. excite the hunting motion.
Q0 is the reference load acting on each wheel, Qanc the
load measured after the superposition of a lateral acceleration of

7 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The criterion to consider as unstable the ride to a certain
speed was to watch if the wheelset oscillations led by the
impulse were damped or not.

250 Friction 0.4 0.3 0.1 0.05 0 Linear

200 Coefficient
Critical speed [m/s]

150
χ Critical speed [m/s] 25
3500000 65 59 37 32 24
100
750000 58 49 35 31 24
50
Table 8: Effect of the friction coefficient – Tare.
0
0 20 40 60 80 100 120
Finally the effect of the wear of the rails and of the wheels
Lateral impulse [KN]
is considered, using different values of the equivalent conicity.
In the following table are reported the results for the Laden
Figure 11: Effect of the impulse force on the critical condition (which is the more critical).
speed (Laden)
Eq. Conicity [rad] 0.05 0.2 0.35
However is known [5], [9], that the impulse value may
Critical speed [m/s] 69 61 57
have a direct influence on the critical speed. Therefore several
Table 9: Effect of the equivalent conicity – Laden.
simulations have been performed changing the impulse force to
find a value over which no more critical speed increment are
Failure of a sidebearers (anti-yaw dampers)
detected.
The figure 11 reports this limit for the laden vehicle which
Since on freight vehicles the maintenance is made seldom,
is of 100 KN, for the tare condition the value is 25 KN, these
in this section has been considered the effect of a failure to one
value were always used in the following for the simulations.
of the sidebearers (on the 4 installed).
The contact model adopted was the one defined by the
Two failure mode are considered:
Johnson linear Theory, which is implemented in ADAMS/Rail
as the Level 2a contact model. Have been used the creepage • Failure mode 1: heavy reduction of the friction coefficient
coefficients calculated for the profiles UIC60/S1002 with an due to wear, presence of oil or ice on the friction surfaces.
equivalent conicity of 0.2 rad. This mode is simulated adopting a friction coefficient of
Several simulation are performed changing the anti-yaw 0.01 in the damper.
friction dampers parameters (such as the friction coefficient), • Failure mode 2: block of a sidebearer, event that may arise
while regarding the damping supplied by the Lenoir-Link no after a long period of inactivity caused by the formation of
variations are made since we have detected that this element rust. This mode is simulated with a friction coefficient of
have no influence on the critical speed. 0.7.
In the following tables is shown the effect of the variation
of the friction coefficient and of the χ parameters, (which Critical Speed [m/s]
represents the starting inclination of the characteristic damping Load Failure mode 1: Failure mode 2 : No failure
force – relative velocity) in both the tare and the laden µ=0.01 µ=0.7
condition. Tare 53 57 65
Laden 47 49 61
Friction 0.4 0.3 0.1 0.05 0 Linear Table 10: effect of the failure of a sidebearers.
Coefficient
Curving stability.
χ Critical speed [m/s] 23.55
3500000 61 54 30 26 22.5
The vehicle in exam is made such that the torsional roll
750000 53 44 29 25 22.5
stiffness is supplied by the same element used as anti – yaw
Table 7: Effect of the friction coefficient – Laden. dampers. For this reason during curving, due to the lateral non-
compensated accelerations, the normal load acting on the
From the obtained values is clear that the inclination χ is as friction surfaces of the sidebeares has a large variation between
much important as bigger is the friction coefficient. the two side of the vehicle. To keep in account this effect a
Furthermore the result obtained with an eigenvalue analysis simulation has been performed using a straight track but
(Linear in the table) are near to the ones obtained in the imposing on the vehicle a lateral acceleration of 1 m/s2.
transient simulation without anti-yaw dampers (friction
coefficient =0).

8 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The result are shown in the following table: contact method). This last method allow 4 degree of freedom to
the wheelset (yaw and spin rotation, lateral and longitudinal
displacement) and gives results for the Y force for each wheel
Critical Speed [m/s] separately as shown in table 13. In the following table are
Load
Curve anc = 1 m/s2 Straight reported the maximum value of the Y/Q ratio for each wheel
Tare 40 65 found during the simulation.
Laden 29 61
Table 11: Effect of non-compensated acceleration LOAD Curve Speed ANC Y/Q
during curving on critical speed. radius
2
[m] [m/s] [m/s ] [/]
The running of the vehicle subject to a high value of lateral Tare 200 7.5 0.6 0.15
acceleration lead to a drastic reduction of the critical speed. Tare 100 11 0.6 0.20
This effect may be explained by the fact that one of the two Tare 60 6 0.6 0.43
sidebearers is locked by a high normal load while the one on the Laden 60 6 0.6 0.13
other side is almost unloaded and ineffective. Table 12: Maximum Y/Q value obtained with the Level
1 contact model.
4.3 Curving
Both the result are similar and furthermore the Level I
The evaluation of the vehicle behaviour during curving has method is conservative.
been made with both the determination of the derailment safety
ratio (Y/Q), of the wheelset yaw angle and verifying the Level I
Level III Level III Level III
maximum lateral force (Ripage). left right sum
Wheelset 1 0.12 0.37 -0.26 0.11
Derailment safety ratio - Y/Q Wheelset 2 0.09 0.05 0.02 0.07
Wheelset 3 0.15 0.36 -0.26 0.10
The Y/Q ratio has been evaluated running the vehicle on Wheelset 4 0.08 0.05 0.02 0.07
curve with different radius to a speed such that the ANC value Table 13: Differences between level I and III
was of about 0.6 m/s2 that is the maximum allowed for freight derailment factor
vehicles in Europe.
We choose to use non-canted tracks because often on small Wheelset yaw angle
radius curve (e.g. 60 m) the track is not canted in the reality.
Furthermore in non-canted tracks the desired value of ANC is These last simulations are performed using the Rail Level
reachable at lower speed (far from the critical speed), so that is III contact model, with the Fastsim 2 algorithm. The results are
possible to evaluate the curving behaviour of the vehicle reported only for the first wheelset in the running direction that
without superposition of dynamic effect and of slant effects shows the worst performance.
which have been considered separately.
On Table 1 are shown the Y/Q value obtained for our Wheelset Curve Yaw angles vrs. Track
vehicle on different curve tracks using the simpler contact 1 ref. frame[mrad]
model implemented in ADAMS/Rail indicated as “Level 1” Load Radius Curvature Steady- Peak Ψ/Ψ0
contact model. This contact model simulate the wheel-rail Condition (Ψ0) state(Ψ)
[m] [mrad] [mrad] [mrad] [%]
contact with a revolute joint applied between the track centre-
line and the wheelset, therefore only two degree of freedom are Laden 200 5 4.1 4.1 82 %
allowed to the wheelset: the yaw rotation and the longitudinal Laden 400 2.5 0.9 1.4 36 %
displacement. Laden 1000 1 0.3 0.7 30 %
Obviously the Laden vehicle has a much better Tare 200 5 4.9 4.9 98 %
performance respect the tare vehicle. Table 14: Level III curving simulation - Wheelset Yaw
However even in the tare case the limit ratio, which is fixed angle.
to 0.8 (depends on the shape of the wheel profile, we use S1002 The table 14 reports the curvature of the considered curve
profile) is respected with a good margin also in the 60 m curves, Ψ0 that is the yaw inclination which should affect the wheelset if
that are the minimum allowed for this vehicle. no additional constrains were present on it. Since the primary
For a curve radius of 200 m has been made the comparison suspension has a very high stiffness, the wheelset is retained to
with the Y/Q values obtained with the simplified contact model assume this inclination.
(“Level 1”) and with a contact model base on the Fastsim The difference between the theoretical angle Ψ0 and the
Algorithm (implemented in ADAMS/Rail in the “Level III” effective wheelset angle is indicated as “Yaw angles vrs. Track

9 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ref. Frame” and is reported both for the steady state condition in that is not effective for friction-damped systems. It is
the middle of the curve and for the maximum value (peak) nnecessary instead to execute a transient analysis and to verify
assumed during the simulation. therefore the growing up of the hunting motion from the system
The ratio Ψ/Ψ0 gives a measure of the curving performance time history. In the study of the curve behaviour it has been
of the vehicle, the higher is the ratio the less is the performance. found that the implemented contact models become unstable in
It is shown that as in tare condition on small radius curve the presence of a small radius. Such limitation however, has not
angle is high; this often leads to heavy wear troubles. affected the evaluation of the friction element here presented,
but of course to have a good evaluation of the vehicle’s
Ripage behavior it is necessary to use different contact elements.
The influence of the friction force during curving is not of
The lateral forces between the rail and the wheelset must be great importance respect a vehicle using conventional dampers.
limited within a certain value in order to avoid a failure in the
rail or in the armature.
The limit for the total lateral force (Ripage) [6] can be ACKNOWLEDGMENTS
found as:
Ymax = ∑ (Fyleft + Fy rigth ) = 0.85 ⋅ 1000 + Q  [daN ]
This research has been supported by grant MURST
 3 40%/2000 of the Italian Research Ministry.
Where Q is the maximum vertical load on the wheelset,
therefore a limit of 65150 N is found for the laden vehicle and REFERENCES
of 19750 N for the tare wagon.
The results reported in table 15 show for the wheelset 1 [1] Kalker, J.J.: A fast algorithm for the simplified theory of
(which is the first in the running direction and have the worst rolling contact. Vehicle System Dynamics, 11, pp. 1-13, 1982.
behaviour) that the limit value is not reached. [2] Kalker, J.J.: Three-dimensional elastic bodies in rolling
Is to keep in consideration that the superposition of a slant contact. Dordrecht, Kluwer Academic Publishers, 1990.
(not considered in this simulation) could reduce strongly the [3] Pater, A.D. de: The geometrical contact between track and
Ymax limit. wheelset. Vehicle System Dynamics, 17, pp. 127-140, 1988.
[4] Pater, A.D. de: The motion of a railway wheelset supported
Wheelset 1 Curve Lateral Force by a pair of rollers as compared with the motion of such a
Load Radius Left Right Sum Ymax wheelset along a tangent track. Delft University of Technology,
Condition wheel wheel Report No. 1012, 1993.
[m] [N] [N] [N] [N] [5] Polach O. : A Fast Wheel-Rail Forces Calculation
Laden 200 43800 29500 14300 65150 Computer Code. Vehicle System Dynamics Supplement 33, pp.
Laden 400 10300 350 9950 65150 728-739, 1999.
Laden 1000 7900 3100 4800 65150 [6] N. Bosso, A. Gugliotta, E. Napoli, A. Somà: “Simulation of
Tare 200 11500 6900 4600 19750 a scaled roller rig”, 5th ADAMS/Rail Users’ Conference;
Table 15: Level III curving simulation - Lateral forces Haarlem, The Netherlands – May 10th –11th , 2000
between rail and wheels [N].

5. FINAL REMARKS

The friction model here introduced has demonstrated to

have a good numerical stability, concurring to carry out all the
numerical simulations necessary to identify the dynamic
performances of the vehicle under examination. Moreover has
been evidenced that, in presence of friction dampers, it is not
possible to adopt the same simulation methodologies used for
vehicles with viscous dampers. As an example in the case of the
simulation of the track slants, the friction force cause an
additional wheel unloading approximately up to the 20% of the
total unloading. Such result it cannot be reached by means of
the normal simulation methodology that consists in a static
simulation. Also in the case of the critical speed determination,
usually is adopted an eigenvalue solution (linearized model)

10 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/08/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use