ADHESION MODELS FOR WHEEL/RAIL CONTACT IN

RAILWAYS
B. ALLOTTA, M. MALVEZZI, P. TONI
Department of Energetics “Sergio Stecco”, Section of Applied Mechanics, University of Florence,
via Santa Marta n. 3, 50139 Firenze – ITALY; e-mail: [email protected], [email protected],
[email protected]

SUMMARY
The forces and moments, which occur in the wheel/rail contact regions, have a dominant effect on the vehicle dynamics.
Conventionally, the horizontal force exchanged between rail and wheel is expressed as the product of the vertical force
acting on the wheel and the so-called adhesion coefficient. When the adhesion conditions between wheels and rail are
degraded (for example in case of rain, fog, ice, dead leaves etc.) and the vehicle is accelerating or braking (i.e. if its
acceleration is different from zero) pure rolling condition do not hold anymore, and macroscopic slips occurs on one or
more of the wheels. Two similar phenomena occur during acceleration (in locomotives’ wheels) and braking (in all
vehicles’ wheels), called slippage and skidding. This paper deals with the problem of identification of the parameters of
adhesion suitable for describing wheel/rail interaction in trains, during braking. The adhesion in the wheel/rail contact
area depends on a large number of factors; the first of these variables to be taken into consideration is slip. This work is
devoted to find a model which correlates slip and adhesion coefficient.

Keywords: railways, wheel/rail contact, adherence, friction coefficient, and identification.

1 INTRODUCTION contact zone between rail and wheel, varies in a random

manner. The research of a law that relates it to other
The contact between wheel and rail, for its fundamental
directly measurable variables (train speed, wheel
importance in railways, has been studied since the very
velocity etc.) is a difficult task.In this paper we report on
beginning of railways. The experiences done by railway
some studies being done in collaboration between the
companies during the first half of the 20th century have
University of Florence and TRENITALIA S.p.A. with
supplied simple formulations based on friction
the ambitious goal of obtaining a reliable model of
coefficients.
adhesion to be used in an hardware-in-the-loop
Then, the growing speed has driven engineers' attention simulator devoted to the test of wheel slip protection
towards vehicle running stability which is influenced by systems (WSP) and odometry systems.
the forces exchanged between wheel and rail. The
In Sec. 2 the state of the art in the formulation of the
contact problem has thus been more deeply investigated,
adhesion coefficient is summarised. In Sec. 3 we discuss
by introducing more complex mathematical models.
some trends in the development of new test rigs devoted
By increasing the tangential force starting from zero, to the testing of railway subsystems which, in our
and fixed the remaining relevant parameters, two opinion, justify the present research. In Sec. 4 we
different phases can be distinguished in the wheel-rail describe the simplified model of an isolated vehicle used
contact phenomenon. In the first phase, named pseudo- to infer, by means of the method described in Sec. 5,
sliding, due to elastic deformation of the two bodies, in adhesion data from existing results of running tests. In
the contact area, there is a first zone where relative slip Sec. 6 we discuss the obtained results
occurs, and a second characterised by adhesion. In this
phase contact forces may generate running instability: in
2 STATE OF THE ART
design practice the incipient instability is evaluated by Due to the complexity of the phenomenon, with constant
linearising the exchanged forces as in [1][2]. The environmental conditions, it is commonly accepted to
dimension of the two zones depends upon the normal express adhesion as a function of sliding speed i.e. the
and tangential forces. With growing longitudinal forces difference between the vehicle speed and the tangential
the adhesion zone decreases till a limit situation where velocity of the wheel at the point of contact [3][4]:
sliding occurs on the whole contact area. This is the
∆v = v – R ω , (1)
second phase that may occur during braking and
where ∆v is the relative speed, v is the vehicle speed, ω
acceleration phases. In such running situations, the
is the angular wheel velocity and R is the wheel contact
tangential forces depend upon the adhesion coefficient,
radius.
which is a function of several parameters (normal load,
sliding speed, temperature of the two bodies, contact In railway practice, instead considering µ as a function
geometry, weather conditions, presence of rain, snow, of relative speed, it is preferred to consider µ as a
dead leaves, etc.) and the dependency on some of them function of the train speed and the relative slip[5][6][7],
may not be easy to express analytically. given by the following expression:
The adhesion coefficient, which relates the vertical force v − Rω (2)
on the wheel to the tangential force applied in the υ=
v
The relative slip varies from 0 (when v = Rω, i.e. when interpreting experimental results of braking/traction tests
pure rolling conditions are satisfied) and 1 (ω = 0, i.e. by means of the model in Fig. 1 is not trivial. The
the wheel is blocked). problem is even more difficult to solve if "real-time"
realistic simulators, including the wheel/rail interaction
model in poor adhesion conditions, are needed. Such
real-time realistic simulators have to be developed so to
test in laboratory "runs" the behaviour of on board
devices which are part of advanced train control and
monitoring systems, thus reducing the cost of line tests.
Among such train running control and monitoring
systems we mention here the Italian system named
SCMT [8] and the European one named ERTMS [9].
The latter will become a standard which the various
railway companies of the European countries are
currently testing (e.g. in the TRENITALIA-ALSTOM
test site along the high speed Arezzo - Florence line)
and intend to gradually introduce in all the railway
network by the end of 2010, with the aim of increasing
safety and service effectiveness. The on board devices
of such systems perform in real-time odometric
calculations to estimate travelled distance and train
speed [10][11], mainly based on information coming
from sensors located on some of the axles of the train
(two, usually belonging to a locomotive). In this
Figure 1: Adhesion coefficient/relative slip curve framework, the availability of reliable models to
from [5]. describe exchanged forces in braking and traction
Some different theoretical sliding/adhesion curves have phases, in addition to the possibility of estimating the
been proposed. For example, SCNF (France) [5], [6], kinematic variables of the rolling stock in all the running
[7] proposed a curve that relates the relative slip to the conditions, including sliding and skidding, is of key
adhesion coefficient; its qualitative behavior is shown in importance. By the way, the availability of test rigs for
Fig. 1. The curve presents two maximums, the first one on board devices, featuring a real-time model of a
is for small relative slip values (1.5%), while the second vehicle (or train) and the wheel-rail interaction model
one is for relative slip values in the range between 5 and can be used, after validation, also for certification trials
25 %. of wheel slip protection systems (WSP) commonly used
in railway braking apparatus [12], as shown in Fig. 2.
This curve is obtained for a given value of train speed
(constant). As the train speed increases, the curve moves
to lower adhesion values, and the maximum value
moves to higher slip value. The adhesion/slip curve evaluation system velocity control of axes

depends also on the condition of the contact area, if this partial vehicle/track
angular velocities
angular
velocities
dynamic simulator
references
+ angular velocities
condition is degraded (for example in case of rain, fog (pumping, pitch, control
servomotors
and adherence)
-
etc.) the curve moves to lower values. braking tachometers
forces

During a braking, the train speed decreases from an braking system

initial value to a final one. The adhesion coefficient does simulator

(or dummy brake)
WSP system
not vary on a curve like the one shown in Fig. 1, but on
a “surface”, because both relative slip and train speed WSP system
to be tested
angular velocity
sensors
vary. (including valves)

In order to fully describe the braking performance the

braking performance of a given train, several families of
curves as the one shown must be known, to account not Figure 2: Test rig to evaluate performances of WSP
only of varying train speed, but also for different systems.
adhesion conditions.
If the adhesion coefficient is related to the absolute slip The wheel/rail interaction model is currently
[3] [4], the curve does not depend upon speed and a investigated by European railway companies within their
unique curve (for given adhesion conditions) is found. technical co-ordination committees [8], with various
proposal, not yet assessed nor satisfactory, being
3 PERSPECTIVE discussed.

The wheel-rail interaction model based on the adhesion Even the multibody dynamic simulation software
coefficient represents a framework for a qualitative currently appointed as the best for railway problems
interpretation of the phenomenon. However, the task of (ADAMS Rail) features a simple Coulomb friction
model not adequate to describe exchanged forces in real The horizontal acceleration is &x& = −d , the equilibrium
running situations. equation can then be written as follow:
The experimental evaluation of wheel-rail friction is c .
− mi d = Qi − Ti − pi (5)
almost impossible to perform by using conventional test r
rigs, for the mechanics of contact between railway The rotation equilibrium equation of each axles
wheels and rollers used to simulate the presence of rail (indicated by the index i) can be written as follow:
is an unacceptable deformation of phenomenological
reality. On the one hand, rollers of so big diameter to be J iω& i = Ti R − M f = Ti R − cpi
a good approximation of (zero) rail curvature should be
so in practice not realizable. On the other hand, the J iω& i c (6)
Ti = + pi
instrumentation of a vehicle devoted to perform line R R
tests aimed to study the wheel-rail contact is quite The term ω& i represent the angular acceleration of the
complex and expensive. A trade-off must be searched wheel.
between the execution of a limited and targeted number
The horizontal vehicle translation equilibrium equation,
of line tests, trying to simplify the instrumentation of the
obtained neglecting the aerodynamics and internal
test vehicle by means of an accurate a priori analysis,
resistances, is:
and laboratory investigation, including theoretical study
and analysis of experimental data. For the above Md = ∑i =1 (Qi − F fi )
4
(7)
mentioned reasons, the importance of the knowledge of
from the equations (3) and (7) the following expression
the adhesion coefficient as a function of operational
is obtained:
parameters is now and will certainly be in the next future
of key importance for the new traffic control and 4  c c 
Md = ∑i =1  Ti + p i − mi d − p i  =
management systems that railway companies want to  r r  (8)
introduce in Europe, for various reasons, including:
4  J ω& c 
to improve the performances and reliability of ∑i =1  i + pi − mi d 
 R R 
wheel slide protection systems (during braking);
The horizontal and vertical components of the resulting
to improve the performance and reliability of anti
force exchanged between the rail and each wheel are
skid traction control systems (during traction);
related by the following expression
to define design specifications for on board systems
devoted to modify the way to drive trains, Ti = µi N i (9)
increasing both railway transport effectiveness and The adhesion coefficient can be evaluated from equation
safety (6) and (9).
to build simulators to test new on-board systems in
order to perform their evaluation with reduction of 1  J iω& c  (10)
µi =  + pi 
expensive line tests. Ni  R R 
4 MODEL DESCRIPTION
Each axles is characterized by:
mass mi
moment of inertia Ji
wheel radius R
distance between the center of the wheel and the
application point of braking force: r
On each axles a braking moment M fi = cp i is applied,

c is a constant; Figure 3: Scheme of the forces acting on the wheel

pi is the pressure in the brake cylinders.
The single axle (indicated by the index i) translation
5 ADHESION COEFFICIENT
equilibrium equation can be written as follow:
IDENTIFICATION FROM
mi &x& = Qi − Ti − F fi (3) EXPERIMENTAL DATA
Qi is the horizontal force transmitted from the vehicle to In the experimental tests used for the identification the
the wheel, Ti is the horizontal force exchanged between following data were available (Fig. 3):
the wheel and the rail, and F fi is the braking force, given Absolute train speed, measured using an optical
by the following expression: sensor;
Wheel velocity, measured by encoder-type sensors
M fi c
F fi = = pi . (4) placed on the wheel axes;
r r Oil pressure in the brake cylinders.
 50
train speed N 12 M t ⋅ g
40 1 s t wheel velocity N1 = N 2 = + + Ma ⋅ g , (12.1)
2 2
[m /s]

30

N M ⋅g
+ Ma ⋅ g , (12.2)
speed

20
N 3 = N 4 = 34 + t
10
2 2
0
25 30 35 40 45 50 55 60 65 where Mt and Ma are respectively the mass of the bogie
tim e [s ]
and of the axles. The pull up angle were evaluated using
4
the following expression:
[bar]

3
Mc ⋅h
ϑ=− d. (13)
pressure

2 2k 2 l 2
1
Where k2 is the stiffness of the secondary suspension For
25 30 35 40 45
tim e [s ]
50 55 60 65
each sample, the coefficient c defined in the previous
section can be found from the following expression
(obtained from equation (8)):
Figure 4: Train speed, 1st axle velocity and brake
cylinder pressure during a braking with degraded 1 4
M tot d − ∑ J iω& i
adhesion conditions cm = R i =1 (14)
1 4
Each information was sampled with a sampling period ∑ pi
R i =1
of 2.4 ms. The experimental tests were conducted using the c coefficient during a braking is represented on
a single vehicle. Fig.6, as it can be seen, its variation is “small”, so we
From wheel velocity information, the angular speed and supposed it to be a constant, whose value is the mean of
acceleration were evaluated. The accelerations were the computed ones.
evaluated using the finite difference method. The
acceleration of the train was obtained from its speed, by 3000
derivation (using the finite difference method). c
ev aluated
c
mean

The absolute accelerations was filtered using a digital 2500

first order filter with a rise time of 2 s (i.e. a cut-off

frequency of 0.5 Hz), while the angular accelerations 2000
c c oeffic ient [Nm /bar]

were filtered using a filter with a cut-off frequency of

2 Hz. 1500

1000

500

0
46 48 50 52 54 56 58 60 62 64 66
tim e [s ]

Figure 6: Coefficient that relates the pressure in the

brake cylinders to the braking moment during a braking
test.
Figure 5: Scheme of the vehicle geometry
For each sample and for each axles the following values
The normal force acting on each bogie was evaluated were evaluated:
using the following expression, in which the dynamic
terms were neglected and the train deceleration were the adhesion coefficient µ (using equation (10));
supposed constant: the absolute speed v;
Mc ⋅ g Mc ⋅h , the absolute slip ∆v;
N 12 = − ⋅d (11.1) the relative slip ν = ∆v/v;
2 2l
the absolute slip derivative;
M ⋅ g Mc ⋅h , (11.2)
N 34 = c + ⋅d the relative slip derivative.
2 2l
where Mc is the case mass, g is the gravity acceleration, The adhesion coefficient and the absolute speed during a
d is the train deceleration, l and h are respectively the braking test is shown in Fig. 6.
horizontal distance between connection points between
the case and the bogies and the vertical distance between
the same points and the case center of gravity (as shown
in Fig. 4). The normal force acting on each wheel was
nearly:
 10
Some results are shown on Fig. 9 and discussed in the
abs olute sliding [m /s ] following section.
5 0.2
experimental data
0.18 evaluated curve

0 0.16
25 30 35 40 45 50 55 60 65
tim e [s] 0.14
0.2
0.12
0.15
0.1
es timated

µ
0.1
0.08

0.05
µ

0.06

0 0.04
25 30 35 40 45 50 55 60 65
0.02

Figure 7: Absolute slip and adhesion coefficient during 0

1 2 3 4 5 6 7
a braking with degraded adhesion conditions absolute sliding [m/s]

Figure 9: Absolute slip/adhesion curve- experimental

The adhesion coefficient, during a loss and a recovery of
data and interpolating curve
adhesion of an axles, has been evaluated. The
identification zone has been chosen in order to: The same analysis was repeated using relative slip
The train speed can be approximately considered instead of absolute one. Some results are shown in Fig.
constant; 10. As it can be seen, in this case the adhesion
The derivative of the relative slip doesn’t change coefficient depend also upon train speed, in particular,
its sign. as the speed increase, the adhesion coefficient decreases.
Some results are shown in Fig. 8.
0.2

0.18
0.12
vmean=28.5514+/-4.0488 m/s 0.16
v = 5 m/s
0.14
0.1
vmean=31.8834+/-1.0021 m/s
0.12 v = 10 m/s

0.08
µ

0.1
v = 30 m/s
0.08
0.06
µ

0.06

0.04 0.04

vmean=32.5598+/-0.78411 m/s 0.02

0.02
vmean=30.0385+/-3.1732 m/s 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
relative slip

0
0 1 2 3 4 5
absolute slip - m/s
6 7 8 9 Figure 10: Relative slip/adhesion curve – experimental
data.
Figure 8: Absolute slip and adhesion coefficient during
a braking with degraded adhesion conditions the solid 6 RESULTS
lines represent loss of adhesion phases, while dashed
lines represent recovery of adhesion phases From the available experimental tests the coefficient that
relates the pressure in the brake cylinders to the braking
To evaluate the slip/adhesion curve, a grid where each moment (that depends on the friction of the brake shoes
element is characterized by a fixed absolute speed and a and the adhesion coefficient between the rail and the
fixed absolute slip were defined. The size of the grid is wheel were evaluated. The coefficient that relates the
40×40, the speed varies from 0 to 40 m/s, while the pressure on the brake cylinders to the braking force is
absolute slip varies from 0 to 10 m/s. The difference about constant during a braking. It has been also verified
between the reference speed of two adjacent elements that the coefficient is nearly constant in different tests.
was 1 m/s and the difference between the reference slip The analysis on the adhesion coefficient as function of
of two adjacent elements was 0.25 m/s. In each element the relative slip confirms these theoretical results:
the speed and the slip could then be considered constant.
The adhesion coefficient depends also on absolute
For each element of the grid the corresponding µ values train speed, in particular, as the speed increase, the
were found and stored in a vector. A 3D matrix was adhesion coefficient decrease;
obtained. The mean and the variance of each vector was The adhesion coefficient is smaller if the derivative
computed and displayed on a 3D plot. The obtained of the slip is positive (i.e. the axles is losing
surface was then sectioned, in order to examine the adhesion).
adhesion coefficient dependence on absolute speed and The analysis on adhesion coefficient as function of the
slip. absolute slip permits to conclude that:
 The dependence on absolute train speed is not yet [4] M.W. Winterling, E. Tuinmann, W. Deleroi,
visible; “Simulation of Drive Line Dynamics of Light Rail
The adhesion coefficient decreases as the absolute Vehicles”, International Conference on Simulation, 30
slip increase. Sept – 2 Oct 1998, Conference Publication n. 457, IEE,
1998.
7 CONCLUSION AND FUTURE WORK
[5] Office for Research and Experiments of the
This analysis was conducted using braking tests whose International Union of Railways, "Adhesion during
objective was the type approval of anti-skid devices. In braking, and anti-skid devices," September 1985.
these tests, even if the adhesion conditions were [6] M. Boiteux, "Le problème de l’adhérence en
artificially degraded, the slip was controlled by these freinage," Revue générale des chemins de fer, 59-72,
devices and was maintained in a narrow band. So only a February 1986.
part of the adhesion coefficient/absolute slip and [7] M. Boiteux, M. Cadier, J. Kling, W. Kunnes,
adhesion coefficient/relative slip curves was identified. "Adhérence en freinage et anti-enrayeurs," Document
Better results could be obtained if tests in which the technique DT257 (B164), Office de Recherches et
sliding varied in a wider range (i.e. the relative slip d’Essays de L’Union Internationale des Chemins de fer,
varied from 0 to 1) were available. July 1992.
8 ACKNOWLEDGEMENTS [8] TRENITALIA S.p.A. Unità Tecnologie Materiale
Rotabile "SCMT Progetto dell'algoritmo per il calcolo
The authors want to thank Mr. Paolo Presciani and Mr. della velocità istantanea del treno e dello spazio
Gianluca Cocci of UTMR – Trenitalia S.p.A. for percorso," UTMR/DT.S.PS, 31/10/2000.
supporting them with useful discussions and for [9] ERTMS-SRS, Version 4a dated 14.07.97
supplying them experimental data.
[10] B. Allotta, M. Malvezzi, G. Cocci, P. Presciani,
"Train Speed Estimation from Wheel Velocity
9 REFERENCES Measurements," Proc. of Int. Conf. on Railway Traction
[1] J. J. Kalker, "Three-dimensional elastic bodies in Systems, Capri, Italy, 15-17 May, 2001.
rolling contact" Dordrecht [etc.], Kluwer academic [11] M. Malvezzi, P. Toni, B. Allotta, V. Colla, "Train
publishers, c1990. speed and position evaluation from wheel velocity
[2] Rao V. Dukkipati, Joseph R. Amyot, " Computer measurements," Proc. of 2001 IEEE/ASME Int. Conf.
Aided simulation in Railway Dynamics", Marcel on Advanced Intelligent Mechatronics, Como, Italy, 8–
Dekker, Inc. New York and Basel,1988. 11 July 2001.
[3] K. Ohishi, Y. Ogawa, I. Miyashita, S. Yasukawa, [12] John Tunley, Stephen Kent, "The Evaluation of
”Anti Slip Re-adhesion Control for Electric Motor Wheel Slide Protection Equipment", AEA Technology
Coach Based on Force Control Using Disturbance Rail, 1997.
Observer” Transaction on IEEJ, vol. 120-D, N. 3, pp.
382-389, 2000.