Fundamentals of Rail Vehicle Dynamics

FUNDAMENTALS OF RAIL VEHICLE DYNAMICS
ADVANCES IN ENGINEERING
Series Editors:
Fai Ma, Department of Mechanical Engineering,
University of California, Berkeley, U.S.A.
Edwin Kreuzer, Department of Mechanics and Ocean Engineering,
Technical University Hamburg-Harburg, Hamburg, Germany
FUNDAMENTALS OF
RAIL VEHICLE DYNAMICS
GUIDANCE AND STABILITY
A.H. WICKENS
Loughborough University, UK
Wickens, A. H.
Fundamentals of rail vehicle dynamics : guidance and stability / A.H. Wickens.
p. cm. -- (Advances in engineering ; 6)
Includes bibliographical references and Index.
ISBN 90-265-1946-X
1. Railroads--Cars--Dynamics. I. Title. II. Advances in engineering (Lisse, Netherlands)
; 6
TF550.W53 2003
625.2015313--dc21
2003045675
Copyright 2003 Swets & Zeitlinger B.V., Lisse, The Netherlands
All rights reserved. No part of this publication or the information contained
herein may be reproduced, stored in a retrieval system, or transmitted in any
form or by any means, electronic, mechanical, by photocopying, recording or
otherwise, without written prior permission from the publishers.
Although all care is taken to ensure the integrity and quality of this publication
and the information herein, no responsibility is assumed by the publishers nor
the author for any damage to property or persons as a result of operation or
use of this publication and/or the information contained herein.
Published by: Swets & Zeitlinger Publishers
www.swets.nl
ISBN 90 265 1946 X
This edition published in the Taylor & Francis e-Library, 2005.
To purchase your own copy of this or any of Taylor & Francis or Routledges
collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.
ISBN 0-203-97099-3 Master e-book ISBN
(Print Edition)
Contents
Arrangement of Book .......................................................................................... ix
Preface ................................................................................................................. xi
1. Basic Concepts.................................................................................................. 1
1.1 Introduction ......................................................................................... 1
1.2 The Railway Wheelset......................................................................... 2
1.3 Creep ................................................................................................... 6
1.4 Stability................................................................................................ 8
1.5 Guidance............................................................................................ 10
1.6 Suspension, Performance and Criteria............................................... 13
1.7 Suspension, Articulation and Curving............................................... 15
References ............................................................................................... 17
2. Equations of Motion ....................................................................................... 19
2.1 Introduction ....................................................................................... 19
2.2 Freedoms and Constraints ................................................................. 19
2.3 Wheel Rail Geometry ........................................................................ 21
2.4 Contact Mechanics ............................................................................ 32
2.4.1 Elasticity and Friction..................................................................... 32
2.4.2 Laws of Friction ............................................................................. 32
2.4.3 Contact Between Wheel and Rail ................................................... 32
2.4.4 Creep .............................................................................................. 33
2.4.4.1 Background ................................................................................. 33
2.4.4.2 Formulation of the Creep Problem.............................................. 34
2.4.4.3 Creep Forces for Small Creepages .............................................. 36
2.4.4.4 Creep Forces for Arbitrary Creepages......................................... 37
2.4.4.5 An Approximate Theory for Arbitrary Creepages....................... 37
2.4.4.6 Heuristic Approximations............................................................ 39
2.4.4.7 Non-Hertzian Effects................................................................... 39
2.4.5 Transient Effects............................................................................. 40
2.5 Creepages .......................................................................................... 40
2.6 Contact Forces ................................................................................... 42
2.7 Kinematics of the Wheelset ............................................................... 44
2.8 Equations of Motion.......................................................................... 46
2.9 Constrained Motion........................................................................... 48
2.10 Equations of Motion for Small Displacements................................ 52
2.11 Equations of Motion for a Two-Axle Vehicle................................. 61
References ............................................................................................... 66
vi CONTENTS
3. Dynamics of the Wheelset .............................................................................. 71
3.1 Introduction ....................................................................................... 71
3.2 The Unrestrained Wheelset ............................................................... 71
3.3 Root Locus for Small Motions of the Restrained Wheelset .............. 76
3.4 Instability and Feedback.................................................................... 81
3.5 Amplitude Dependent Behaviour and Limit Cycles.......................... 84
3.6 Energy Balance.................................................................................. 88
3.7 Dynamic Aspects of Guidance .......................................................... 89
3.8 Alternative Methods of Guidance...................................................... 94
References ............................................................................................. 103
4. Guidance of the Two-Axle Vehicle .............................................................. 107
4.1 Introduction ..................................................................................... 107
4.2 Properties of the Stiffness Matrix.................................................... 109
4.3 Steering on Large Radius Curves .................................................... 111
4.4 Response to Cant Deficiency on Large Radius Curves ................... 116
4.5 The Conflict Between Steering and Stability .................................. 118
4.6 Motion on Sharper Curves............................................................... 121
4.7 Response to Misalignments ............................................................. 127
4.8 Flange Forces and Derailment ......................................................... 130
References ............................................................................................. 131
5. Dynamic Stability of the Two-Axle Vehicle ................................................ 133
5.1 Introduction ..................................................................................... 133
5.2 Equations of Motion........................................................................ 135
5.3 Stiff and Flexible Vehicles .............................................................. 142
5.4 The Flexible Vehicle with Zero Suspension Damping.................... 147
5.5 Damping and the Long Wheelbase Flexible Vehicle ...................... 151
5.6 Stability of the Flexible Vehicle in General .................................... 153
5.7 The Application of Cross-Bracing and Yaw Relaxation ................. 159
5.8 The Stiff Vehicle or Bogie............................................................... 160
5.9 The Three-Piece Bogie .................................................................... 169
References ............................................................................................. 169
6. The Bogie Vehicle ........................................................................................ 173
1 Introduction ........................................................................................ 173
2 Equations of Motion........................................................................... 175
3 Dynamics of the Conventional Bogie Vehicle ................................... 177
4 Steering and Stability of Multi-Axle Vehicles in General.................. 182
5 Steering and Stability of a Generic Bogie Vehicle............................. 185
6 Application to Specific Configurations .............................................. 190
7 Stability of Bogie Vehicles with Steered Wheelsets .......................... 197
8 Simple Bogie Model ........................................................................... 199
9 Stability of Simple Bogie Model ........................................................ 201
References ............................................................................................. 206
CONTENTS vii
7. The Three-Axle Vehicle ............................................................................... 209
7.1 Introduction ..................................................................................... 209
7.2 Steering and Stability of Three-Axle Vehicles................................ 210
7.3 Steering with Unequal Conicities .................................................... 215
7.4 Stability of Vehicle with Uniform Conicity .................................... 217
7.5 Stability with Unequal Conicities.................................................... 225
7.6 Dynamic Response .......................................................................... 230
References ............................................................................................. 232
8. Articulated Vehicles ..................................................................................... 235
8.1 Introduction ..................................................................................... 235
8.2 Steering and Stability....................................................................... 238
8.3 Application to Specific Configurations ........................................... 244
8.4 Stability of an Articulated Three-Axle Vehicle............................... 246
8.5 Stability and Response of an Articulated Four-Axle Vehicle.......... 252
References ............................................................................................. 260
9. Unsymmetric Vehicles.................................................................................. 261
9.1 Introduction ..................................................................................... 261
9.2 Stability Theorems for Rigid and Semi-Rigid Vehicles .................. 263
9.3 Unsymmetric Rigid Vehicle ............................................................ 265
9.4 Steering of a Vehicle with Unsymmetric Inter-Wheelset Structure 267
9.5 Stability of a Two-Axle Articulated Vehicle................................... 272
9.6 The Influence of Elastic Stiffness on Stability ............................... 276
9.7 Applications of Unsymmetry........................................................... 279
References ............................................................................................. 281
Index ................................................................................................................. 283
Arrangement of the book
Sections are numbered serially within each chapter. If reference is made to a section
within the chapter containing the section, the section number is cited as a single
number. Otherwise, a section is identified by two numbers separated by a decimal
point, the first number referring to the chapter in which the section appears, and the
second identifying the section within the chapter.
Equations are numbered serially within each section. If reference is made to an
equation within the section containing the equation, the equation number is cited as
a single number. If reference is made elsewhere in the same chapter then the
equation number is cited as a two-figure number and if reference is made in another
chapter all three numbers-chapter, section and equation are cited.
Figures and tables are numbered by chapter.
Preface
The fundamental method of guidance of the railway vehicle is the coned and flanged
wheelset. Whilst facilitating guidance in curves, coning can give rise to sustained
lateral oscillations, termed hunting. This oscillation induces forces which can cause
damage to both vehicle and track and there can be, at least, discomfort to the pas-
senger and, at worst, the risk of derailment. Inadequate steering on curves can have
similar consequences. This book concentrates on the resulting problem of the con-
flict between guidance and stability and its resolution by proper design of the sus-
pension connecting the wheels and car body of the railway vehicle.
The invention of the wheelset, the progressive development of the bogie and the
various schemes of articulation which have been developed over the years in order
to resolve the design conflict between stability and steering, all predate the theory of
railway vehicle dynamics. Engineering insight brought railway technology a long
way but empirical methods were not adequate once the railway renaissance started
and train speeds increased. A fundamental change in railway technology took place
in which the empirical evolution of railway bogies was replaced by a more scientific
and numerate approach. This approach has been very successful; for example, not
only has stable operation of steel wheel on steel rail vehicles been demonstrated at
speeds of over 500 km/h (more than double the speed of the fastest train fifty years
ago) but the analytical and predictive capabilities now available have stimulated a
rising tide of innovative designs.
The detailed modelling of the dynamics of railway vehicles is made possible by
the several excellent computer packages that are available, which provide suffi-
ciently detailed and validated mathematical models that can be used with confidence
in engineering design and development. These models permit the simulation of the
actual motion on a specified stretch of track so that the performance of a specific
design can be analysed, or a particular incident recreated. Thus, by simulation the
overall performance of a vehicle can be checked. Realism is, of course, essential in
design but equates to complexity, and computer output must be tempered with un-
derstanding and scepticism. It is important, therefore, that fundamental principles
are well understood.
This book is concerned with the fundamental principles of guidance and stabil-
ity, which are a consequence of the mechanics of wheel-rail interaction as embodied
in the equations of motion. For research purposes, where the objective is to achieve
an understanding of an innovative system or a particular problem, simple models
can be very useful and can provide productive insights. Analytical studies which
describe the mechanics of various phenomena by the simplest model possible can be
used to explore new suspension and vehicle design concepts.
much of current practice, though it pre-dates the availability of theory, is the way it
is. Moreover, it becomes clear why many innovations failed in the past.
Because an important consequence of a more analytical approach is to separate
out the dynamic properties of a system from the detailed design of its components
the latter will not be discussed. Moreover, the application of active controls to steer-
ing and ride control (including body tilting) will not be covered. Active systems will
play a large part in the future and those working in the field will require a sound
grounding in passive systems. As the emphasis is on ride quality and guidance, a
frequency range of roughly 0 15 Hz is of prime interest. This makes it possible to
consider that, in general, wheelsets and track (except in the areas of contact) are
rigid and that car bodies are without flexibility. This means that some significant
phenomena are not discussed here. Moreover, simple forms of suspension elements
are assumed. The more straightforward problems of response in the vertical plane,
or in the longitudinal direction are not addressed.
The basic concepts are described in Chapter 1. A detailed discussion of the equa-
tions of motion follows in Chapter 2 in which a compromise has been made between
the mathematical rigour of some investigators and the ad-hoc use of Newtons Laws
of others. In this Chapter, though an engineering approach has been followed, great
reliance has been placed on the careful derivations of Professor de Pater. The fol-
lowing Chapters deal with the single wheelset and then with progressively more
complex configurations of vehicle. If possible, simple analytical results have been
derived as these, if available, provide the best basis for understanding the mechanics
of the systems involved. All numerical results have been obtained using standard
commercially available software for numerical computation.
It is a pleasure to acknowledge the stimulus and help I have received over the
years from colleagues, too numerous to mention here, at British Rail Research,
Loughborough University and through the International Association of Vehicle Sys-
tem Dynamics.
A. H. Wickens
Idridgehay, 2002
Attention will be concentrated on the configuration and parametric design of the
bogie, in relation to steering, dynamic response and stability. Therefore the treat-
ment of the various configurations of vehicle do not simply concentrate on a current
typical set of parameters but attempt to consider the consequences of the complete
range of parameters open to the designer. By this approach, it is possible to see why
PREFACE xii
1
Basic Concepts
1.1 Introduction
The railway train running along a track is one of the most complex dynamical sys-
tems in engineering. It has many degrees of freedom, the interaction between wheel
and rail involves both complex geometry of wheel tread and rail head and non-
conservative forces generated by relative motion in the contact area, and there are
many non-linearities.
The long history of railway engineering provides many practical examples of
dynamical problems which have degraded performance and safety. The two essen-
tial features of operation, running in a train of vehicles and guidance by the track,
cause problems which are unique to railways. Inadequate guidance on curves results
in high lateral forces between wheel and rail, rapid wear of wheels and rails and the
possibility of derailment. Dynamic and static instabilities, and excessive response to
track irregularities and other features of track geometry, can result in poor ride qual-
ity and high stresses and can contribute to derailment. Operation in a train involves
the control of forces acting between the vehicles in the train as the propulsive and
braking forces are varied in response to the train traversing hills and valleys. High
frequency interaction between wheel and rail can lead to damage to the contacting
surfaces and corrugation of the rails, and excessive noise and vibration.
The dynamics of the railway vehicle represents a balance between the forces act-
ing between the wheel and the rail, the inertia forces and the forces exerted by the
suspension and articulation. Of these, the basic characteristics of the wheel-rail in-
terface such as friction, geometry, and the elasticity in the contact area are hardly
under the control of the designer. But the configuration, suspension and forms of
articulation can be varied over a wide range of possibilities, limited mainly by the
degree of complexity considered acceptable for each application. The objective of
suspension design is, therefore, to control the motion of the railway vehicle so that
good ride quality is achieved, at the same time dynamic loads and the tendency to
derail are reduced to acceptable levels, whilst running on track with geometry that is
economically acceptable.
In a complete model of the dynamics of a railway vehicle, the vehicle is consid-
ered to be assembled from wheelsets, car bodies and intermediate structures which
are flexible, and which are connected by components such as springs and dampers.
Similarly, the vehicle is considered to run on a track which has a complex structure
RAIL VEHICLE DYNAMICS 2
with elastic and dissipative properties. Each major component has six rigid body
degrees of freedom plus additional degrees of freedom representing the elastic dis-
tortion of the component. In the latter case, these additional degrees of freedom
might represent a finite element model of the structure or a series of natural modes
of vibration. The track can be modelled as a continuous structure with a moving
interface at the points of contact, where the interaction between wheel and rail is
dependent on the relative motion. This kind of model, with varying assumptions, is
provided by various computer software packages which are used in the engineering
design and analysis of railway vehicles. So, one objective of the study of the dy-
namics of railway vehicles is the development of sufficiently detailed and validated
mathematical models that permit the simulation of the actual motion, on a specified
stretch of line, so that the performance of a specific design can be analysed, or a
particular incident recreated. Thus, by simulation, the overall performance of an
existing or projected vehicle can be checked and design decisions made.
A second objective of the study of railway vehicle dynamics is to develop ana-
lytical or numerical models describing the mechanics of various phenomena by the
simplest model possible. These can be used to explore new suspension and vehicle
concepts and to develop a basis for physical understanding and insight. Ideally, not
only analysis but synthesis is required in which various possibilities for design are
exposed. Simpler models are typically generated by simplifying assumptions and in
this book, concerned with guidance and stability, these are that
the vehicle has a longitudinal plane of symmetry (parallel to the direction
of motion on straight track) making it possible, under certain conditions, to
separate equations governing those motions which are symmetric with re-
spect to the plane of symmetry from those which govern anti-symmetric
motions;
variations in longitudinal motion are not considered so that the vehicle
moves at constant forward speed;
the motions of interest are at low frequencies and, in most cases, flexibility
of components can be neglected.
It is the objective of this chapter to explain the basic concepts of stability and guid-
ance of railway vehicles as a preliminary to more detailed mathematical analysis.
1.2 The Railway Wheelset
The basic unit of a railway vehicle is the wheelset, Figure 1.1. The conventional
wheelset of today has the following features: it consists of two wheels fixed on a
common axle, so that each wheel rotates with a common angular velocity and a con-
stant distance between the two wheels is maintained. Flanges are provided on the
inside edge of the treads and the flange-way clearance allows, typically, 710 mm
of lateral displacement to occur before flange contact. Whilst many wheelsets com-
mence life with purely coned treads, typically coned at 1/20 or 1/40, these treads
wear rapidly in service, so that the treads come to possess curvature in the transverse
direction. Similarly, rails also possess curvature in the transverse direction. All these
BASIC CONCEPTS 3
features contribute to the behaviour of the railway vehicle as a dynamic system, and it
is important to consider their purpose.
The conventional railway wheelset has a long history [1] and seems to have
evolved by a process of trial and error. Naturally, in the pioneering days of the early
railways most attention was concentrated on reducing rolling resistance so that the
useful load that could be hauled by horses could be multiplied. Another major prob-
lem was the lack of strength and resistance to wear of the materials then available.
Moreover, the level of adhesion between rolling wheel and the track was unknown.
As a result, many possibilities were tried. An obvious step was to fit wheels with
cylindrical treads. However, if the wheels are fixed on the axle and the treads are
intended to be cylindrical very slight errors in parallelism would induce large lateral
displacements which would be limited by flange contact. There is no guidance until
flange contact and thus a wheelset with cylindrical treads tends to run in continuous
flange contact. The position of the flange, either inside or outside the rails, was con-
troversial well into the nineteenth century. Nor was there agreement as to whether
the wheels should be rigidly fixed to an axle or free to revolve on the axle, though
the usual practice seemed to be that wheels were fixed to the axle. The play allowed
between wheel flange and rail was initially minimal. In the early 1830s the flange-
way clearance was opened up with the objective of reducing the lateral forces be-
tween wheel and rail.
A further important point is that the geometry of the wheel and rail as it has
evolved is particularly favourable for the method of switching which involves a
minimum of moving parts and only small gaps in the running surfaces of the rails.
It is not known when coning of the wheel tread was first introduced. It would be
natural to provide a smooth curve uniting the flange with the wheel tread, and wear
of the tread would contribute to this. Moreover, once wheels were made of cast
iron, taper was normal foundry practice. The purpose of coning was partly to reduce
the rubbing of the flange on the rail, and partly also to ease the motion of the vehi-
cle in curves.
A wheelset with coned wheels in a curve can maintain a pure rolling motion if it
moves outward and adopts a radial position. Redtenbacher [2] provided the first
theoretical analysis in 1855 which is illustrated in Figure 1.2. From the geometry in
this figure it can be seen that there is a simple geometric relationship between the
Figure 1.1 Railway wheelset.
lateral movement of the wheelset on a curve y, the radius of the curve R, the wheel
radius r
0
, the lateral distance between the points of contact of the wheels with the
rails 2l and the conicity of the wheels in order to sustain pure rolling. In practice a
wheelset can only roll round moderate curves without flange contact, and a more
realistic consideration of curving requires the analysis of the forces acting between
the vehicle and the track.
It can be seen, in broad terms, why the wheelset adopted its present form. If the
flange is on the inside the conicity is positive and as the flange approaches the rail
there will be a strong steering action tending to return the wheelset to the centre of
the track. If the flange is on the outside the conicity is negative and the wheelset
will simply run into the flange and remain in contact as the wheelset moves along
the track. Another factor is the behaviour in sharp curves. If the flange is on the in-
side then the lateral force applied by the rail to the leading wheelset is applied to the
outer wheel and will be combined with an enhanced vertical load. As explained
later, this diminishes the risk of derailment. With outside flanges the lateral force
applied by the rail applied to the inner wheel which has a reduced vertical load and
thus the risk of derailment is increased. These factors can be easily demonstrated
with the aid of model wheelsets [3].
Thus, it can be seen that for small displacements from the centre of straight or
slightly curved track the primary mode of guidance is conicity and it is on sharper
O
R
A
B
C
D
y
l l
R
2
OAB = OCD
( r
0
- y )/( R - l ) = ( r
0
+ y )/( R + l )
y = r
0
l / R
Figure 1.2 Redtenbachers formula for the rolling of a coned wheelset on a curve.
BASIC CONCEPTS 5
curves and switches and crossings that the flanges become the essential mode of
guidance. Though this appears to be a modern view, in 1838 Brunel [4] wrote
The flanges are a necessary precaution but they ought never to touch the rail
and therefore they cannot be said to keep the wheels on the rails. They ought
not to come into action except to meet an accidental, lateral force. A railway
with considerable curves might be travelled over with carriages at any veloc-
ity and with wheels without flanges. The wheels are made conical, the
smaller circumference at the outer edge. The pair of wheels are fixed to the
axle and thus if anything throws the wheels in the slightest degree to one side
the wheel is immediately rolling on a larger circumference than the other and
the tendency to roll back is introduced. The carriage is kept always in the
middle of the track. A beautiful arrangement.
As a concept, this view led to many significant improvements in the design of rail-
way vehicle suspensions in the 20th century.
Coning of the wheel tread was well established by 1821. George Stephenson in
his Observations on Edge and Tram Railways [5] stated that
It must be understood the form of edge railway wheels are conical that is the
outer is rather less than the inner diameter about 3/16 of an inch. Then from a
small irregularity of the railway the wheels may be thrown a little to the right
or a little to the left, when the former happens the right wheel will expose a
larger and the left one a smaller diameter to the bearing surface of the rail
which will cause the latter to loose ground of the former but at the same time
in moving forward it gradually exposes a greater diameter to the rail while
the right one on the contrary is gradually exposing a lesser which will cause
it to loose ground of the left one but will regain it on its progress as has been
described alternately gaining and loosing ground of each other which will
cause the wheels to proceed in an oscillatory but easy motion on the rails.
y = a sint s = Vt
1
2
2
2
2
2
0
2
R
d y
ds
y
V
r l
V R
= = =

Figure 1.3 Derivation of Klingels formula for the kinematic oscillation of a wheelset from
Redtenbachers formula in Figure 1.2.
This is a very clear description of what is now called the kinematic oscillation, as
shown in Figure 1.3.
Thus, if a wheelset is rolling along the track and is displaced slightly to one side,
the wheel on one side is running on a larger radius and the wheel on the other side is
running on a smaller radius. Because the wheels are mounted on a common axle one
wheel will move forward faster than the other because its instantaneous rolling ra-
dius is larger. Hence, if pure rolling is maintained, the wheelset moves back into the
centre of the track a steering action is provided by the coning. However, the
wheelset overshoots the centre of the track and the result is the kinematic oscillation.
In 1883 Klingel gave the first mathematical analysis of the kinematic oscillation
[6] and derived the relationship between the wavelength and the wheelset conicity
, wheel radius r
0
and lateral distance between the contact points between wheels
and rails 2l as
= 2 (r
0
l/)
1/2
This simple formula follows purely from the geometry of Figure 1.3, and is consis-
tent with Redtenbachers formula for the wheelset in a curve. Since distance along
the track s = Vt where V is the forward speed and t is time, Klingels formula shows
that, as the speed is increased, so will the frequency of the kinematic oscillation.
Very little else can be deduced about the dynamical behaviour of railway vehicles
which must come from a consideration of the forces acting.
1.3 Creep
Pure rolling rarely takes place, and wheels and rails are not rigid. The normal load between
wheel and rail causes local elastic deformation and an area of contact, the contact patch, is
formed. In the case where the surfaces of the wheels and rails are smooth and have con-
stant curvature in the vicinity of the contact patch, Hertz [7] showed that the contact patch
was elliptical in shape, and the distribution of normal pressure between wheel and rail over
the contact patch is semi-ellipsoidal.
If a longitudinal force is applied to the wheel, so that it is braked, a deviation from the
pure rolling motion occurs. The deviation in relative velocity divided by the forward speed
of the wheel is referred to as the longitudinal creepage. Similarly, lateral creepage is de-
fined as the (incremental) relative lateral velocity divided by the forward speed. In addition,
relative angular motion between wheel and rail about the normal to the contact patch is
referred to as spin. If the longitudinal creepage is small, it is accommodated by elastic
strains in the vicinity of the contact patch. As the wheel rotates, unstrained material enters
the contact patch at its leading edge. As the material moves through the contact patch, the
relative velocity between the wheel and rail equals the rate of change of strain so that the
surfaces are locked together. The magnitude of the resulting longitudinal tangential stress
increases linearly with distance from the leading edge. Similarly, lateral creepage gives rise
to lateral tangential stresses. Both longitudinal and lateral creepage therefore generate
forces which are directly proportional to the corresponding creepage. When there is spin,
the pattern of elastic strain is more complicated. In this case, as the material moves
BASIC CONCEPTS 7
through the contact region the relative velocity between wheel and rail is di-
rectly
proportional to the distance from the centre of the contact region and therefore the
strain field becomes curved. As a consequence, a lateral force is generated (the cou-
ple about the common normal is small and may be safely neglected).
As the creepages and spin increase, the tangential stresses increase, and where
these stresses exceed the normal pressure multiplied by the coefficient of friction,
slipping takes place. The result is that the area of adhesion at the front of the contact
patch in which the surfaces are locked together progressively reduces as the creep-
age increases, Figure 1.4. The relationship between the creep force and creepage is
then as shown in Figure 1.5(a). For sufficiently large creepage, slipping takes place
over the whole contact patch and the creep force is equal to the normal force multi-
plied by the coefficient of friction.
If both longitudinal and lateral creep occur simultaneously then for small creep-
ages the creep forces can be superposed, but for larger creepages in the area of slip-
ping the tangential stresses act in a direction opposite to the local resultant relative
velocity. The result is that then all the creep forces are influenced by both lateral and
longitudinal and lateral creepages and spin. Though the lateral force is proportional
to spin for small values of the spin, for large values of the spin slipping takes place
over a large part of the contact patch and the lateral force reduces to zero. The rela-
tionship between lateral force and spin is therefore as shown in Figure 1.5(b).
It was Carter's [8] introduction of the creep mechanism into the theory of lateral
dynamics that was the crucial step in developing a realistic model of the wheelset.
locked
region
slip
x
=
z
(a)
(b)
(c)
locked slip
Figure 1.4 The contact patch between wheel and rail (a) elevation showing locked region of
adhesion at leading edge and region of slip at trailing edge (b) normal pressure
z
and
tangential traction applied by wheel to rail
x
(c) contact patch in plan view.
In the light of the creep theory the action of conicity can be considered as follows
If a wheelset is rolling along a track and is displaced laterally, the rolling radii
are different on the two wheels. Noting that the wheels are fixed on a common axle,
The tread velocities are fractionally different and so longitudinal creep is generated.
The corresponding creep forces are equivalent to a couple which is proportional to
the difference in rolling radii or conicity, and which tends to steer the wheelset back
into the centre of the track. This is the basic guidance mechanism of the wheelset. In
addition, when the wheelset is yawed, a lateral creep force is generated. In effect,
this coupling between the lateral displacement and yaw of the wheelset represents a
form of feedback, and this introduces the possibility of dynamic instability.
1.4 Stability
An important feature of a railway vehicle is that, in addition to the vertical suspension
connecting the wheelsets to the vehicle body, there are lateral and longitudinal springs
and dampers. This is illustrated by the plan-view of a two-axle vehicle shown in Fig-
ure 1.6. The purpose of the plan-view suspension is to stabilise the tendency of the
wheelsets to oscillate and to facilitate the motion of the vehicle in curves.
Because of the action of the creep forces the motion of a rolling wheelset incorpo-
rated in a vehicle is significantly different from Klingels description. At low forward
speeds successive overshoots decrease in magnitude as the vehicle moves along the
track, eventually to travel in a straight line down the centre of the track. The vehicle is
dynamically stable, because following a slight disturbance it returns to its original
path. At high forward speeds successive overshoots grow as the vehicle moves along
the track, for in this case it is dynamically unstable.
0
0.015
0
15000
1
-T
1
(N)
0 6
0
8000

3
-T
2
(N)
(a)
(b)
Figure 1.5 (a) Variation of longitudinal creep force T
1
with longitudinal creepage
1
showing
the limiting value T
3
where = 0.3 is the coefficient of friction and T
3
= 39000 is the
normal force (b) Lateral creep force T
2
as function of spin
3
(zero longitudinal and lat-
eral creep).
BASIC CONCEPTS 9
The mechanism of this instability may be appreciated by considering the simple
vehicle of Figure 1.6 in the special case where the vehicle body has very high iner-
tia and is assumed to move forward at constant speed but does not undergo any lat-
eral motion.
At any forward speed the tendency of the wheelset is to oscillate at the frequency
of the kinematic oscillation. Since this frequency is proportional to speed, then at
low speeds the inertia forces will be small and the main component of the resultant
force acting on the wheelset is the restoring force provided by the springs connect-
ing the wheelsets to the vehicle body. In order to balance this force, creep must be
developed and this will cause a progressive reduction in lateral displacement as the
wheelset pursues its oscillatory path.
At high speeds the inertia forces will dominate, as the frequency is correspond-
ingly high. In this case, creep must be developed which will cause a progressive
increase in the lateral displacement of the wheelset during its lateral oscillation. It
follows that there is a speed at which the successive overshoots neither grow nor
decay, the wheelset then, and only then, tracing out a sinusoidal path. Klingels so-
lution for pure rolling then emerges as a special case of the dynamics of the wheel-
set.
In general, a railway vehicle is stable at low speeds so that following a distur-
bance the vehicle, together with its wheelsets, will undergo a decaying oscillation,
and the vehicle will return to the centre of the track. As the speed is increased, the
decay rate of the oscillation is reduced. In most cases, at a sufficiently high speed
the oscillations following a disturbance grow and eventually lead to a limit cycle
oscillation where the amplitude is limited by either contact of the flanges with the
rails or slipping of the wheels on the rails. The energy required to sustain this oscil-
lation clearly comes from the energy of the forward motion.
This fully developed limit cycle oscillation is generally termed hunting. The
lowest vehicle speed at which sustained oscillations can occur is known as the criti-
cal speed. Early measurements of hunting, made in the 1960s, showed that two-
axle freight vehicles of then current design had critical speeds as low as 30 km/h [9]
Figure 1.6 Plan view of two-axle railway vehicle showing lateral and longitudinal suspen-
sion springs.
and this was a contributory cause of many derailments. Similarly a typical passenger
bogie vehicle of the same era had a critical speed of about 110 km/h which though
not dangerous was the cause of bad riding [10]. In practice, large lateral forces can
be experienced by a hunting vehicle, as indicated in Figure 1.7. The first successful
theoretical prediction and experiment on the track is described in [11]. The danger
posed by hunting at high speeds was appreciated by Matsudaira who was the first to
study the effects of the suspension on stability [12], and was graphically demon-
strated by the lateral distortion of the track during high speed trials of a locomotive
by French Railways in 1955 [13].
Earlier, the Sevenoaks accident in which a locomotive was derailed at high speed
was explained by Carter [14] as a form of static instability in which the wheelbase
buckled under the action of the creep forces.
Thus the guidance offered by the coning of the wheels is the source of potential
instability. This is the fundamental conflict in the design of the running gear of a
railway vehicle.
1.5 Guidance
Guidance is the ability of a vehicle to follow the geometric layout of the track. Ride
is the ability of a vehicle to minimise the dynamic response, in terms of stresses and
accelerations, to the layout of the track. The actual track layout will consist of the
design layout, largely determined by geographical and operational factors and super-
imposed irregularities or lack of accuracy of real track.
The most important aspect of guidance is the behaviour of vehicles in curves.
The application of Redtenbachers formula shows that a wheelset will only be able to
move outwards to the rolling line if either the radius of curvature or the flangeway
Hunting
Limit Cycle
V
H
V
c
Figure 1.7 Typical variation of lateral force H acting on wheelset as a function of forward
speed V for a two-axle vehicle subject to instability.
BASIC CONCEPTS 11
clearance is sufficiently large. Moreover, as discussed above, for stability the wheel-
sets in a vehicle must be constrained in some way. Thus in many curves, the wheel-
sets are not able to take up a radial position, and typically the attitude of a two-axle
vehicle in plan view is as shown in Figure 1.8. Because a wheelset is constrained by
the longitudinal and lateral stiffnesses connecting it to the rest of the vehicle, it bal-
ances a yaw couple applied to it by the suspension by moving further out in a radial
direction so as to generate equal and opposite longitudinal creep forces, and it will
balance a lateral force by yawing further.
In sharp curves the flange will be in contact with the rail. In the case of the lead-
ing wheelset of a vehicle, the flange force will be reacted by the creep forces gener-
ated by the wheels creeping laterally towards the inside of the curve, Figure 1.8.
The first essentially correct description of curving was given by Mackenzie [15] in
1883. His discussion was based on sliding friction, and neglects coning, so that it is
most appropriate for sharp curves, where guidance is provided by the flanges.
Whilst new wheel profiles are often purely coned on the tread, usually to an angle
of 1:20, treads wear rapidly and assume a hollow form, Figure 1.9. For a worn profile
the variation of the difference in rolling radius between the right hand and left hand
wheels as the wheelset is displaced laterally can be very complicated. A useful con-
cept is the equivalent conicity which is the linearised slope of the difference in
S
S
Y
C
Figure 1.8 Two-axle vehicle in curve showing the relative magnitude of the forces acting
between wheel and rail together with the forces applied to the wheelsets by the suspen-
sion S, the centrifugal force C and horizontal component of the normal force between
wheel and rail Y.
Figure 1.9 Geometry for typical worn wheel rolling on worn rail, showing the move-
ment of the contact point as the wheel is displaced laterally.
(a)
(b)
Figure 1.10 Normal and lateral tangential forces acting on wheelset (a) in central position (b)
in laterally displaced position, illustrating the gravitational stiffness effect.
BASIC CONCEPTS 13
rolling radius plotted against lateral displacement. In practice, the equivalent conic-
ity of a worn wheelset is much greater than that of a purely coned wheelset. This
increased conicity enables a wheelset to roll round a curve of much smaller radius
than is possible for a coned wheel.
The influence of the hollow wheel tread gives rise to additional, significant
forces acting between wheel and rail. When the wheelset is rolling on equal radii in
the central position, the contact plane is inclined to the horizontal at a small angle as
shown in Figure 1.10(a). When the wheelset is displaced laterally, contact is made at
new points and the inclination of the contact planes is changed; on one wheel it is
increased and on the other it is reduced, as shown in Figure 1.10(b). As the normal
reactions between wheel and rail, which support the weight carried by the wheelset
are similarly inclined, it is found that there is a lateral resultant when these normal
reactions are resolved horizontally. This lateral restoring force is proportional to
lateral displacement and is referred to as the gravitational stiffness.
A second effect due to the hollow wheel tread arises from the lateral force gener-
ated by spin [16]. As the inclination of the contact plane changes with lateral dis-
placement, the resolved component of the angular velocity of the wheelset due to its
rolling motion, taken about an axis normal to the contact plane, gives rise to changes
in spin creepage. This generates a lateral force which is proportional to lateral dis-
placement and which is of the opposite sign to that of the gravitational stiffness
force. For small displacements, typically the total contact stiffness (gravitational +
spin) is reduced to about 20% of the gravitational stiffness, and is consequently
rather small. However, as mentioned above, the lateral force due to spin becomes
small for the large values of spin achieved in flange contact, and there is a large re-
storing force due to flange contact.
Heumann [17] suggested that profiles approximating to the fully worn should be
used rather than the purely coned treads then standard. After re-profiling to a coned
tread, tyre profiles tend to wear rapidly so that the running tread normally in contact
with the rail head is worn to a uniform profile. This profile then tends to remain sta-
ble during further use, and is largely independent of the original profile and of the
tyre steel. Similarly, rail head profiles are developed which also tend to remain sta-
ble after the initial period of wear is over. These results suggested that vehicles
should be designed so as to operate with these naturally worn profiles, as it is only
with these profiles that any long-term stability of the wheel-rail geometrical parame-
ters occurs. Moreover, a considerable reduction in the amount of wear is possible by
providing new rails and wheels with an approximation to worn profiles at the outset
and this has now become common practice in many countries.
1.6 Suspension, Performance and Criteria
For satisfactory performance, a railway vehicle must meet certain criteria. The most
fundamental of these is concerned with the possibility of derailment. Figure 1.11
shows a yawed wheelset in flange contact with a rail. There are initially two points of
contact between wheel and rail, one on the wheel tread and the other on the flange.
The latter point of contact lies ahead of the former. The onset of flange-climbing
derailment occurs when the vertical load Q is carried entirely by the point of contact
on the flange, and so the derailment limit is defined by the minimum value of the
lateral reaction Y. If the component of the tangential force in the transverse vertical
plane is denoted by F, and N is the normal reaction, the balance of forces in Figure
1.12 shows that
Y
Q
N F
N F
<
+
tan
tan
The minimum value of Y/Q occurs when F is a maximum so that by the laws of
friction F cannot exceed N where is the coefficient of friction, hence
Y
Q
=
+
tan
tan

1
This derailment criterion is due to Nadal, but like many succeeding studies, its
analysis was based on suspect assumptions. However, the application of the laws of
creep by Gilchrist and Brickle [18] has shown that Nadals formula is correct for the
most pessimistic case when the angle of attack is large and the longitudinal creep on
the flange is small. It will be apparent that derailment is most likely when a large
lateral force occurs simultaneously with a reduced vertical load on a wheel, typi-
cally in a curve with a significant vertical irregularity, or a high degree of track
twist.
Another mode of failure occurs when the lateral forces imposed by the vehicle
are sufficient to shift the track laterally. An empirical criterion for this has been
given by Prudhomme [19] and is
F
F
N
Y
Q
Q
a
b
Figure 1.12 Flange contact when wheelset is yawed at angle with rail, showing first point
of contact on tread a and second point of contact with flange b.
BASIC CONCEPTS 15
Y = 10 + W/3
where W is the axle-load, and both Y and Ware measured in kN.
These are extreme conditions. More generally, the curving ability of a railway
vehicle is considered good if the angle of yaw of each wheelset is small, flange con-
tact is avoided on all but sharp curves, the lateral forces between wheel and rail are
low and the energy expended in wheel rail contact is small. A measure of curving
performance that is often used in practice is the degree of flange wear experienced
in service and empirical relationships have been established between rates of wear
and the energy expended in the contact area.
After the essential stability and curving requirements are met, it is considerations
of ride quality which dominate the detailed design of railway vehicle suspension
systems. The layout of curves is defined by the maximum cant of the track, and ve-
hicle speed, so that the lateral acceleration applied to passengers is within accept-
able limits. Moreover, the length of the transition between straight and curved track,
and in some cases the shape, is determined by limits on the time rate of change of
cant deficiency.
So far, the dynamics of railway vehicles running on perfectly aligned straight or
curved track has been considered. A further important problem is that of the dy-
namic response of vehicles to track irregularities. Track irregularities arise from
specific features such as switch and crossing work or from the continuous distribu-
tion of roughness which must be controlled by maintenance of the track. In the lat-
ter case the irregularities may be considered as randomly distributed. The dynamic
response, in terms of the acceleration level on the car body, to this stochastic input
will characterise the ride quality of the vehicle. Detailed international standards for
passenger comfort criteria exist which define frequency weighting characteristics
describing human response to vertical and lateral vibration, which are used in the
specification and assessment of vehicles.
The excursions of the car body on the suspension must remain within the struc-
tural clearance gauge of the route on which the vehicle is to run.
1.7 Suspension, Articulation and Curving
It can be seen that a vehicle with a perfect suspension would be stable at all opera-
tional speeds, would negotiate curves by minimising the forces acting between
wheel and rail and in traversing irregular track would minimise the acceleration lev-
els in the car body and the stresses applied to both vehicle and track structures.
These requirements are usually conflicting and compromise, informed by analysis,
is required in design.
Not only are the parameters that are associated with wheel-rail contact, both
geometrical and frictional, not under the control of the designer or operator, but
they are not known exactly and can vary over a wide range. It follows that practical
designs must be very robust in relation to such parameters. On the other hand, there
is enormous scope for the design of the suspension system in terms of the way in
which the wheelsets and car bodies in a train are connected.
It has long been the objective of vehicle design to incorporate wheel and steering
arrangements that permit a vehicle to follow a chosen path or a track by a motion
which involves pure rolling of the wheels, apart from the necessary transmission of
traction forces and the reaction of centrifugal force. In the first place pure rolling
might be achieved by a choice of configuration of the wheels and the way in which
they are articulated. Such rolling motions may or may not be statically or dynami-
cally stable. Additionally the introduction of creep, in response to inertia and sus-
pension forces may stabilise or destabilise the system. It follows that there are many
possible mechanisms of guidance and stability which can be considered. There have
been many attempts to provide an alternative to the railway wheelset and reference
will be made to some of these in Chapter 3. However, in using the conventional
railway wheelset, there are many ways to improve performance in curves by making
a vehicle more flexible in plan view, thus encouraging the axles to take up a more-
or-less radial position in curves. It will be shown later that a two-axle vehicle that is
capable of radial steering on a uniform curve will be dynamically unstable at all
speeds so that the design of a two-axle vehicle requires a compromise between sta-
bility and curving. This is the subject of Chapters 4 and 5.
As discussed in Chapters 6 and 7 for a vehicle with three or more axles it is pos-
sible to arrange the suspension so that radial steering and dynamic stability are both
achieved. One approach is to provide elastic or rigid linkages directly between
wheelsets in a vehicle. This can be referred to as self-steering as the vehicle body is
not involved. Alternatively, a linkage system can be provided which allows the
wheelsets to take up a radial position but provides stabilising elastic restraint from
the vehicle body. This is so-called forced steering as it can be considered that the
vehicle body imposes a radial position on the wheelsets.
There are many designs in which there is articulation of the vehicle bodies of a
vehicle or train. Articulation, in the present context, describes an arrangement in
which the relative motion between the vehicle bodies is used to influence the stabil-
ity and guidance of the vehicle. In many cases the interaction between the vehicles
in a train is minimised by the form of coupling between the vehicles, so that longi-
tudinal forces can be transmitted between car bodies, but the coupler is capable of
transmitting little or no lateral force or yaw couple. In this case it is a good approxi-
mation to treat each vehicle as if it were isolated and the lateral dynamics of each
vehicle can be considered to be largely independent of that of the rest of the train. In
an articulated vehicle the connections between vehicles form an essential part of the
running gear. A common design feature is to link the relative angle between vehicle
bodies to yaw of the wheelsets. As is discussed in Chapter 8, such designs improve
curving performance and other aspects of vehicle design but can exhibit a wide
spectrum of various hunting instabilities.
All the configurations discussed so far have been symmetric fore-and-aft. Un-
symmetric configurations make it possible, in principle, to achieve a better com-
promise between curving and dynamic stability, at least in one direction of motion.
But additional forms of instability can occur as discussed in Chapter 9.
The equations of motion are fundamental for all configurations of vehicle, and
the derivation of these for a wheelset and a simple two-axle vehicle is discussed in
Chapter 2.
BASIC CONCEPTS 17
References
1. Wickens, A.H.: The dynamics of railway vehicles-from Stephenson To Carter.
Proc. I. Mech. E. 212, Part F (1998), pp. 209-217.
Gilchrist, A.O.: The long road to solution of the railway hunting and curving
problems. Proc. I. Mech. E. 212, Part F (1998), pp. 219-226.
2. Redtenbacher, F.J.: Die Gesetze des Locomotiv-Baues. Verlag von Friedrich
Bassermann, Mannheim, 1855, p. 22.
3. Wickens, A.H.: Dynamics and the advanced passenger train. Speaking of Science
1977, Proceedings of The Royal Institution of Great Britain, 50 (1978), pp. 33-65.
4. Vaughan, A.: Isambard Kingdom Brunel Engineering Knight Errant. John
Murray, London, 1992, p. 102.
5. Dendy Marshall, C.F.A.: History of British Railways Down to the Year 1830. Ox-
ford University Press, Oxford, 1938, p. 165.
6. Klingel.: Uber den Lauf der Eisenbahnwagen auf Gerarder Bahn. Organ Fortsch.
Eisenb-wes. 38 (1883), pp. 113-123.
7. Timoshenko, S.P.: A History of the Strength of Materials, Mcgraw-Hill, New
York, 1953, p. 348.
8. Carter, F.W.: The electric locomotive. Proc. Inst. Civ. Engs. 221, 1916, pp. 221-
252.
9. Pooley, R.A.: Assessment of the critical speeds of various types of four-wheeled
vehicles. British Railways Research Department Report E557, 1965.
10. King, B.L.: The measurement of the mode of hunting of a coach fitted with stan-
dard double-bolster bogies. British Railways Research Department Report E439,
1963.
11. Gilchrist, A. O., Hobbs, A.E.W., King, B.L. and Washby, V.: The riding of two
particular designs of four wheeled vehicle. Proc. I. Mech. E. 180 (1965), pp. 99-
113.
12. Matsudaira, T.: Hunting problem of high-speed railway vehicles with special
reference to bogie design for the New Tokaido Line. Proc. I. Mech. E. 180 (1965),
pp. 58-66.
13. Knothe, K. and Bohm, F.: History of stability of railway and road vehicles. Ve-
hicle System Dynamics, 31 (1999), pp. 283-323.
14. Carter, F.W.: The running of locomotives, with reference to their tendency to
derail. Inst. Civil Engs, Selected Engineering Paper, No. 81, 1930.
15. Mackenzie, J.: Resistance on railway curves as an element of danger. Proc. Inst.
Civ. Engs. 74 (1883), pp. 1-57.
16. Johnson, K.L.: Effect of spin upon the rolling motion of an elastic sphere on a
plane. Trans. A. S. M.E. Ser. E, 80 (1958), pp. 332-338.
17. Heumann, H.: Zur Frage des Radreifen-Umrisses. Organ Fortschr. Eisenb.-wes.
89 (1934), pp. 336-342.
18. Gilchrist, A.O. and Brickle, B.V.: A re-examination of the proneness to derail-
ment of a railway wheelset. J. Mech. Eng. Sci. 18 (1976), pp. 131-141.
19. Birmann, F.: Theoretical and experimental solutions of track problems for high
speeds. Monthly Bulletin of the International Railway Congress Association 45
(1968), pp. 391-460.
2
Equations of Motion
2.1 Introduction
The basic physical phenomena involved in the dynamics of railway vehicles have
been described in Chapter 1. Equations of motion governing the stability and dy-
namic response of vehicles will now be derived which encompass the essential fea-
tures of the wheel-rail geometry, the frictional forces acting between wheel and rail
and the elastic and damping forces generated by the suspension.
As attention will be confined to the dynamics at low frequencies, the wheelset
and track are assumed to be rigid apart from local elasticity in the contact patch be-
tween wheel and rail, and the contributions of the local deflections near the contact
patch to the overall motion of the wheelset are neglected. The wheelset, which is
assumed to be axisymmetric about the axle centreline, is considered to be con-
strained to run along the track at constant speed. The track is arbitrarily curved in
plan view and may be canted.
The kinematics of the wheelset is considered first, and this is followed by a dis-
cussion of wheel rail geometry. An evaluation of the creep forces acting between
wheel and rail makes it possible to formulate equations of motion of a freely run-
ning wheelset. This is followed by the derivation of the equations of motion of a
complete two-axle vehicle in which the action of the suspension is taken into ac-
count.
2.2 Freedoms and Constraints
The track possesses curvature in a horizontal plane with radius R
0
, cant or cross-
level
0
, and can be displaced locally through a lateral displacement y
0
, Figure 2.1.
R
0
,
0
, and y
0
vary with the distance s along the track. The wheelset reference frame
Oxyz is attached to the centreline of the undistorted track, Figure 2.1, and moves
along the track at the speed of the vehicle V. Thus the irregularities y
0
are measured
from this centreline. The origin of this set of axes is located at the centre of mass of
the wheelset when the wheelset is central on the track. Ox lies along the tangent to
the track centreline, Oy lies along the wheelset axle centreline when the wheelset is
central and lying in the radial direction on the curve, and Oz is mutually perpendicu-
lar. The coordinates X, Y and azimuth of the origin of the frame Oxyz, Figure 2.1,
20 RAIL VEHICLE DYNAMICS
with reference to an axis system fixed in the earth are given by
dX/ds = cos (1)
dY/ds = sin (2)
d/ds = 1/R
0
(3)
A second set of axes O*x*y*z* has origin at the centre of mass of the wheelset.
O*y* coincides with the axle centreline, O*x* is perpendicular to O*y* and O*z* is
mutually perpendicular.
To specify the orientation of the frame O*x*y*z* it will be convenient to select
successive rotations, yaw , and roll , about the carried axes Oz*, Ox*. Thus the
rotations are taken about the position the axes have taken following the previous
rotation. The rotation of the wheelset about the carried axis Oy* is denoted by .
The displacements of the wheelset centre of mass O* relative to Oxyz are denoted
X
Y
y*
y
O*
O
x*
x
z
z*
y
y*
O
G
0
Figure 2.1 Wheelset axis systems and coordinates.
EQUATIONS OF MOTION 21
by the vector x
0
with components u
x
, u
y
and u
z
, but as O* is in the plane yOz and
both O and O* move forward at constant speed V
x
0
= [ 0 u
y
u
z
] (4)
and the longitudinal position along the track s = Vt. Thus the position and orient-
ation of a wheelset can be defined in terms of the six variables, s, the lateral and
vertical displacements u
y
and u
z
and three rotations, yaw , roll , and .
As the area of contact between wheel and rail is small compared with the dimen-
sions of the track contact between the wheelset and the rails can be considered to
take place ordinarily at two points. As will be discussed later, two point contact
gives rise to two constraint equations which makes it possible to eliminate two of
the above coordinates. It will be convenient to eliminate the vertical displacement
and roll angle of the wheelset as independent coordinates, so that they become sim-
ply dependent functions of lateral displacement and yaw. As the vehicle speed is
constrained to be constant, the system has three degrees of freedom.
2.3 Wheel Rail Geometry
As discussed in Chapter 1 the most important geometrical characteristics of the
wheel rail geometry are (a) the variation of rolling radius with lateral displacement
as this governs the conicity effect and (b) the variation of the slope at the contact
point with lateral displacement as this governs the gravitational stiffness effect. For
a typical wheel and rail combination, Figure 2.2, both profiles have curvature which
varies continuously across the rail head and wheel tread and are defined by

w
= f (
w
)
r
= g (
r
) (1)
r
Figure 2.2 Typical wheel and rail profiles relative to the rail coordinates
r
,
r
(mm).
where
w
,
w
are the wheel coordinates and
r
,
r
are the rail coordinates, the pro-
files being the same for the right-hand and left-hand wheels, Figure 2.3.
In order to derive equations of motion, the position of the contact points, and the
slopes and curvatures at these contact points, as functions of the wheelset lateral
displacement and yaw are required. It will be assumed that the cross-sectional ge-
ometry of the wheel-rail system does not vary with distance along the track. It fol-
lows that the cross-sectional geometry is independent of s and . The wheels and
rails will be assumed to be rigid in so far as their mutual geometry is concerned.
When the wheelset is in the central position on the track, and is not yawed, the angle
made by the contact plane with the horizontal is
0
and the tread circles of the wheels
have the same radius r
0
. When the wheelset is displaced laterally, the angles made
between the contact planes and the axle centreline at the new points of contact are
wr
and
wl
. Similarly, the radii of the tread circles become r
r
and r
l.
The position of the
contact points is determined by noting that the wheel and rail contact points must
occupy the same position in space, the angles made by the contact planes at the
points of contact must be the same for wheel and rail, and the contacting bodies can-
not penetrate each other. As the angle of yaw of a wheelset is small in most ordinary
circumstances, the two-dimensional case where the influence of yaw is neglected
will be considered. Consider the right hand wheels and rails, Figure 2.3. When the
r
central
displaced
u
y
-y
0
-r
0
wr
B
rr
A
C
rr
wr
-u
z
-l
Figure 2.3 Wheel rail geometry (right-hand). A is the origin of the rail axes
r
,
r
, B is the
origin of the wheel axes
w
,
w
, so that A and B are coincident when the wheelset is cen-
tral. When the wheelset is displaced, contact takes place at C.
wheelset is in the central position the contact point is A. The centreline of the track
is displaced laterally by y
0
from the reference axis from which u
y
is measured. When
the wheelset centre of mass is displaced laterally through a distance u
y
from the ref-
erence axis the wheelset rotates about a longitudinal axis through a small angle
and so the lateral displacement at the contact point is u
y
- y
0
- r
0
, and contact is
made at a new point C. If the lateral movement of the contact point on the right
hand wheel is
wr
and that on the rail is
rr
then
u
y
- y
0
- r
0
-
wr
+
rr
= 0 (2)

Similarly, considering the vertical movement of the wheelset
wr
at the right hand
rail
u
z
+ l +
wr
-
rr
= 0 (3)

Also, if if
rr
denotes the angle between the rail axes and the contact plane

-
wr
+
rr
= 0 (4)

The three corresponding equations for the left hand wheel are

u
y
- y
0
- r
0
+
wl
-
rl
= 0 (5)

u
z
- l +
wl
-
rl
= 0 (6)

+
wl
-
rl
= 0

(7)
In addition, the slopes at the contact points are given by
tan
wr
= d
wr
/d
wr
(8)
tan
wl
= d
wl
/d
wl
(9)
tan
rr
= d
rr
/d
rr
(10)
tan
rl
= d
rl
/d
rl
(11)
For profiles specified by (1), equations (1) to (11) can be solved to yield the contact
positions and slopes, the vertical displacement and roll angle as functions of u
y
.
Measuring equipment has been developed to measure wheel and rail profiles with
the necessary accuracy, [1, 2, 3]. The solution has been implemented in computer
programs and typically uses a Newton-Raphson iterative procedure to solve the
nonlinear algebraic equations. The first step is to determine the points of contact
when the wheelset is central. This then defines a new origin for the wheel rail geo-
metric data. Then equations (1) to (11) are solved for the contact points. Once the
contact points have been established the various geometrical characteristics can be
determined. Detailed discussions of the analytical aspects of wheel rail geometry
have been given by de Pater [4] and Yang [5].
As an example, for the wheelset and rail combination shown in Figure 2.2 the
variation of , u
z
and the location of the contact points is shown in Figure 2.4. Fig-
ure 2.5 shows the variation of rolling radius, contact angle and the transverse curva-
tures of the rail and wheel with lateral displacement. The examples shown refer to
worn wheel and rail profiles and in this case yield quite smooth characteristics.
However, in practice, many rail and wheel profile combinations yield characteristics
with major discontinuities.
Though, as the wheelset is displaced laterally, the rotation and vertical dis-
placement u
z
are small it will be seen later that their derivatives with respect to u
y
play an important part in the equations of motion. Expressions for these derivatives
will now be derived.
-10 0 10
-1
0
1
-10 0 10
-0.4
-0.2
0
u
z
(mm)
10
-10 0
-10
0
10
r
(mm)
-10 0 10
-20
0
20
10
u
y
(mm)
-10 0 10
0
2
(mr)
l
r
w
(mm)
r
-10 0
0
1
2
r
(mm)
w
(mm)
u
y
(mm)
r
l
r
l
Figure 2.4 Roll angle , vertical displacement u
z
and location of contact points as a function
of lateral displacement for the wheel rail combination of Figure 2.2.
Figure 2.5 Variation of rolling radius, contact slope, and radii of curvature with lateral dis-
placement for the wheel rail combination of Figure 2.2.
I
u
y
z
u
y
l - r
l
tan
wl
l - r
r
tan
wr
r
l
r
r
r
l
tan
wl r
r
tan
wr
Figure 2.6 Tilting and vertical displacement of wheelset due to lateral displacement. Wheel-
set rotates about instantaneous centre I.
r (mm)
r
r
l
l
R
r
(mm)
R
w
(mm)
r
r
l
l
du
du
l
d
du
d
du
d
du
z
y y
rr
y
wr
y
+ =

=
d
d
d
du
d
d
d
du
rr
rr
rr
y
wr
wr
wr
y
and using (8) and (9)

= tan tan
rr
rr
y
wr
wr
y
d
du
d
du
and as is small
=
tan

rr
rr
y
wr
y
d
du
d
du
and differentiating (2) and substituting
=
tan
rr
y
r
d
du
1
0
(12)
Similarly
du
du
l
d
du
r
d
du
z
y y
rl
y
=
tan 1
0
(13)

Hence from (12) and (13)

y
=d/du
y
= - ( tan
rl
+ tan
rr
)/(2l - r
0
tan
rr
- r
0
tan
lrl
) (14)
z
y
= du
z
/du
y
= l ( tan
rl
- tan
rr
)/(2l - r
0
tan
rr
- r
0
tan
rl
)
(15)
which also follow from the geometry of Figure 2.6. For small displacements from
the central position

y
= - /l ( 1 - r
0
0
/ l ) z
y
= -u
y
/l ( 1 - r
0
0
/ l ) (16)
where
=
0
= (
rr
-
rl
)l/2u
y
(17)
is the parameter that determines the change of inclination of the normal force
between wheel and rail as the wheelset is displaced laterally and therefore influences
the gravitational stiffness. is the roll parameter.
Figure 2.7 shows the variation of the geometrical derivatives
y
and z
y
.
Numerical values of the parameters for the wheel rail combination of Figure 2.2 are
Differentiating (3) with respect to u
y
r
0
= 0.4500 m,
0
= 0.0493 and l = 0.7452 m.
If u
y
*
is the lateral displacement of the wheelset in the plane of the original points
of contact, then from Figure 2.6
u
y
*
= 2lu
y
/(2l - r
0
tan
wr
- r
0
tan
wl
) (18)
because the wheelset is rotating about the point I. For small displacements this be-
comes
u
y
*
= u
y
/( 1 - r
0
0
/ l ) (19)
From (19) the difference between the lateral displacement of the wheelset at the
axle and at the contact points for the profiles of Figure 2.2 is only 3%.
For small displacements equations (1)-(11) have a simple solution. In the vicinity of
the point of contact R
w
and R
r
are the radii of curvature of the wheel tread and rail
head respectively. When the wheelset is displaced laterally through a distance u
y
it
can be seen from Figure 2.3 that the lateral displacement of the contact point on the
rail is, approximately,
rr
= R
r
(
rr
-
0
) (20)
to the first order in
0
and
rr
The lateral displacement of the contact point on the
wheel is
wr
= R
w
(
wr
-
0
) (21)
Substituting in (2) from (20), (21) and (4) and noting from (16) that =
y
u
y
, yields
for the right hand side
-10 0 10
-0.4
-0.2
0
0.2
0.4
u
y
(mm)
z
y
-10 0 10
-0.6
-0.4
-0.2
0
y
(1/m)
u
y
(mm)
Figure 2.7 Variation of derivatives z
y
and
y
with lateral displacement for the wheel rail
combination of Figure 2.2.
rr
,
rl
=
0

0
u
y
/l (22)
where, since to first order (1+ r
0
0
/ l)
-1
= (1 - r
0
0
/ l)
0
= l ( 1 + R
w
0
/l)/ (R
w
- R
r
)(1 - r
0
0
/ l) (23)
and
wr
,
wl
=
0

0
*
u
y
/l (24)
where
0
*
= l ( 1 + R
r
0
/l)/ (R
w
- R
r
)(1 - r
0
0
/ l) (25)
For example, for the wheel rail combination of Figure 2.2,
0
= 6.423 and
0
=
6.372. Note that
0
-
0
= . For conical wheels, R

w
,
0
= 0 and
0
=
0
/( 1 - r
0
0
/ l ) (26)
For profiled or worn wheels R
w
0
/ l << 1 and R
r
0
/ l << 1. For example, for the
wheel rail combination of Figure 2.2, R
w
0
/ l = 0.01795 and R
r
0
/ l = 0.00989.
Hence, in this case
0
=
0
*
= l / ( R
w
- R
r
)( 1 - r
0
0
/ l ) (27)
Reference to Figure 2.3 shows that
wr
= - R
w
(cos
wr
- cos
0
)
= - R
w
(
0
2
-
wr
2
)/2 (28)
approximately, so that substituting from equations (22) and (24), to first order
r
r
, r
l
= r
0

0
u
y
(29)
where
0
is the effective conicity
0
=
0
R
w
( 1 + R
r
0
/l)/ ( R
w
- R
r
)( 1 - r
0
0
/ l ) (30)
For example, for the wheel rail combination of Figure 2.2,
0
= 0.1144. For conical
wheels
0
=
0
( 1 + R
r
0
/l) /( 1 - r
0
0
/ l ) (31)
and for profiled or worn wheels where R
w
0
/ l << 1 and R
r
0
/ l <<1,
0
=
0
R
w
/ (R
w
- R
r
)(1 - r
0
0
/ l) (32)
as first derived by Heumann [6].
For purely coned wheels the vertical displacement of the wheelset is negligible.
Considering therefore the case of profiled or worn wheels so that
0
=
0
*
rr
= - R
r
(cos
rr
- cos
0
)
=- R
r
(
0
2
-
rr
2
)/2 (33)
approximately, so that substituting (28) and (33) into (22) and (24) and then adding
(3) and (6) the vertical displacement of the centre of mass is found to be
u
z
= - ( R
w
- R
r
)
0
2
u
y
2
/2l
2
(34)
consistent with (16). For completeness, to this could be added the small vertical
displacement which occurs when the wheelset yaws through a small angle , which
is
0
l
2
, so that the total vertical displacement is
u
z
= - ( R
wr
- R
rr
)
2
y
2
/2l
2
+
0
l
2
(35)
The above analysis was first given in [7] and was subsequently elaborated by Blader
[8] and Joly [9]. A very precise consideration of the linearisation of the wheel rail
geometry has been given by de Pater [4]. For real wheel and rail profiles linearisation
has severe limitations and whilst the above approximate approach is adequate to il-
lustrate the nature of the problem, in practice a more precise numerical analysis is
necessary. This is illustrated by the plots of the variation of the difference in rolling
-10 0 10
-4
-2
0
2
4
u
y
(mm)
r
r
- r
l
(mm)
-10 0 10
-0.2
-0.1
0
0.1
0.2
r
-
l
u
y
(mm)
Figure 2.8 Variation of rolling radius difference and contact slope difference with lateral
displacement for the wheel rail combination of Figure 2.2.
radii and in contact angles with lateral displacement of the wheelset shown in Fig-
ure 2.8 for the wheel rail combination of Figure 2.2.
Thus the wheel rail geometry is essentially non-linear. An equivalent conicity
can be defined in terms of the difference of rolling radii on opposite wheels when
the wheelset is displaced laterally [10]. It is dependent on the amplitude of the lat-
eral displacement in a wheelset oscillation and is approximately equal to the mean
slope out to the amplitude in question. Similarly, equivalent contact slope and roll
parameters can be defined so that
r
r
-r
l
= 2(a)u
y
(36)

r
-
l
= 2(a)u
y
/l (37)
= - (a)u
y
/l (38)
A number of more refined approaches to quasi-linearisation have been developed.
The describing function method assumes that the motion is approximately sinusoi-
dal. Hence a sinusoidal variation of the lateral displacement of the wheelset is as-
sumed
u
y
= asint (39)
Then the resulting rolling radius difference can be expanded as a Fourier series, in
which it will ordinarily be sufficient to take only the first term. Thus, the following
amplitude dependent describing functions can be defined
(a) = (1/a) ( ) sin r r sds
r l
0
2
(40)

(a) = (1/a)
sin sds
0
2
(42)
0 2 4 6
0
0.05
0.1
0.15
0.2
0.25
0 2 4 6
0
5
10

a (mm) a (mm)
Figure 2.9 Describing functions for the wheel rail combination of Figure 2.2.
(a) = (1/a) ( ) sin
r l
sds
0
2
(41)
In practice, the values of can vary over a wide range of values, 0.05 < < 0.5 and,
as will be seen later, it is this variation that imposes difficulties on suspension de-
sign rather than the detailed shape of the rolling radius difference graph. Various
practical aspects of equivalent conicity have been discussed by Pearce [11].
The above discussion generally neglects the effect of wheelset yaw on the wheel-
rail geometry, which is a realistic assumption except in the case of flange contact at
large angles of wheelset yaw, such as occurs during flange climbing derailment.
Three-dimensional geometry analyses have been given by Cooperrider and Law [3],
Hauschild [12], Duffek [13], Yang [5], de Pater [14] and Muller [15]. The major
effect of yaw at large angles of attack is to provide an additional torque on the
wheelset, see Section 5.
Many wheel-rail combinations experience contact at two points on one wheel for
certain values of the lateral wheelset displacement. This commonly occurs, for ex-
ample, when contact is made between the throat of the flange and the gauge corner
of the rail, Figure 2.10. If the wheels and rails are considered to be rigid, as was
done in the case of single-point contact, discontinuities occur in the geometric char-
acteristics such as the rolling radius difference and slope difference graphs. This is
revealed by singular solutions of equations (2)-(7). The mathematical aspects of
two-point contact in this case have been considered by Yang [16]. However, in this
case the distribution of forces between the points of contact depends on the elastic-
ity in the contact areas [17,18].
Figure 2.10 Two point contact between wheel and rail.
2.4 Contact Mechanics
2.4.1 Elasticity and friction
So far, the size of the contact patch has been assumed to be small compared with
the dimensions of the wheelset. However, the forces acting between wheel and rail
depend on their elastic interaction in the vicinity of the contact patch, as well as the
effects of friction. A comprehensive treatment of contact mechanics is given in [19].
2.4.2 Laws of friction
Laws of friction were postulated by Leonardo da Vinci, Amontons and Coulomb.
The nature of frictional phenomena is complex and simple laws cannot be supposed
to be rigorously true under all conditions. Nevertheless, rail and wheel surfaces are
sufficiently smooth that it can be usually assumed that the laws of friction are
obeyed at each point in the contact patch, so the tangential friction traction
t
acting
at a point
(i) opposes the direction of the relative motion that would occur if the friction did
not exist, or opposes the relative motion if it occurs;
(ii) is independent of the position in the area of contact;
(iii) has magnitude always just sufficient to prevent relative motion at the point un-
der consideration, provided that
t
<
z
where is the coefficient of limiting fric-
tion;
(iv) is proportional to the normal pressure
z
between the bodies when relative mo-
tion takes place so that
t
=
z
.
It is assumed that is independent of the relative velocity.
2.4.3 Contact between wheel and rail
When an elastic body, such as a wheel, is pressed against another elastic body, such
as a rail, so that a normal load is transmitted, a contact area is formed. As the elastic
deformation in the vicinity of the contact area is small its effect on the geometrical
analysis of Section 3 can be neglected. Then, assuming that the curvatures of wheel
and rail are constant in the vicinity of the contact patch, that the contact patch is small
compared with the radii of curvature and the dimensions of the wheel and rails, the
contacting bodies can be represented by elastic half-spaces and their shape can be ap-
proximated by quadratic surfaces. Usually it can be assumed that the material proper-
ties of wheel and rail are the same and in this case it can be shown that the tangential
tractions do not affect the normal pressures acting between the bodies. Then with
these assumptions, for the case where the wheels and rails are smooth, the dimensions
of the contact area can be obtained from the theory of Hertz [20], which is described
by Love [21] and Timoshenko and Goodier [22].
In this case the contact area is elliptical and the normal pressure distribution is semi-
ellipsoidal. The shape and orientation of the contact ellipse depends only on the transverse
radius of curvature of the wheel tread at the point of contact R
w
(as calculated geometri-
cally, measured positive if the wheel profile is hollow), the radius of curvature of the rail
head at the point of contact R
r
, and the wheel radius r. Then if N is the normal force acting
in the contact area the major and minor semi-axes of the contact area, a and b are given by
where E is Youngs modulus, is Poissons ratio, and m and n are functions tabulated
in [22] as a function of where
cos = (1/r + 1/R
r
- 1/R
w
)/(1/r - 1/R
r
+ 1/R
w
)
Here if cos is positive then a lies across the rail, and if cos is negative then a
lies along the rail. Table 2 gives some example results for the wheel-rail geometry of
Figure 2.2 for the cases where the wheelset is central or displaced laterally by
6 mm. In this Table and henceforth a and b are the semi-axes of the contact area in
the forward and lateral directions. The variation of the shape of the contact area with
wheelset displacement is noteworthy and, as illustrated in Table 2.1, in some wheel
and rail combinations contact can take place over a large area in the flange throat
and corner of the rail. The assumptions of the Hertzian theory are then invalid and
the calculation of normal pressures and tangential tractions must be considered si-
multaneously. Methods of analysis have been developed for this case by Nayak and
Paul [23] and by Kalker [24].
2.4.4 Creep
2.4.4.1 Background
As discussed in Chapter 1, Carter introduced the fundamental concept of creep in
relation to the rolling wheel. The relationship between the creepage and the creep
forces has been investigated for a variety of practical situations. Carter [25] gave a
solution to the creep problem for the two-dimensional case of two long cylinders,
pressed together by a normal force, and transmitting a tangential force across the
contact strip. Similar results were obtained by Poritsky [26], and in the discussion of
[26], Cain [27] pointed out that the region of adhesion must lie at the leading edge
of the contact area. A three-dimensional case was solved approximately by Johnson
[28] who considered an elastic sphere rolling on an elastic plane. This solution was
based on the assumption that the area of adhesion is circular and tangential to the
Table 2.1 Contact ellipses for geometry of Figure 2.2 N = 39240 N.
y mm -6 0 6
r mm 449.59 450 452.42
R
r
mm 336.50 149.55 55.935
R
w
mm 784.93 271.28 86.43
m 1.099 1.110 1.461
n 0.9174 0.9082 0.7275
deg. 82.28 81.42 61.34
a mm 4.67 5.14 5.75
b mm 5.59 4.20 2.86
(a/m)
3
= (b/n)
3
= 3N(1 -
2
)/E(1/r + 1/R
r
- 1/R
w
) (1)
area of contact (which is also circular) at the leading edge. Good agreement with
experiment was obtained. The influence of spin about an axis normal to the contact
area was first studied by Johnson [29]. The general case where the contact area is
elliptical was considered by Haines and Ollerton [30] who confined their attention
to creep in the direction of motion and assume that Carter's two-dimensional stress
distribution holds in strips parallel to the direction of motion. A general theory for
the elliptical contact area, based on similar assumptions to those made in [28], was
developed by Vermeulen and Johnson [31], yielding the relationship between creep-
age and tangential forces for arbitrary values of the semi-axes of the contact area. De
Pater [32] initiated the complete solution of the problem by considering the case
where the contact area is circular, and derived solutions for both small and large
creepages, without making assumptions about the shape of the area of adhesion.
However, this analysis was confined to the case where Poisson's ratio was zero;
Kalker [33] gave a complete analytical treatment for the case in which Poisson's
ratio is not zero. The agreement between these theoretical results and the experimen-
tal results of Johnson [28] is very good. Kalker gave a full solution of the general
three-dimensional case in [34] covering the case of arbitrary creepage and spin, and
subsequently gave simpler approximate solution methods [35]. Kalkers theory is
described in [36].
2.4.4.2 Formulation of the creep problem
A local set of axes is defined with origin at the centre of the contact patch, with the
axis Oz normal to the contact area, and fixed in the contact patch so that the material
forming the contacting surfaces moves backward in the direction -Ox at speed V. If
tangential tractions are applied by the wheel to the rail in the contact patch elastic
strains result which cause a departure from the pure rolling motion, which is meas-
ured in terms of the longitudinal creepage
1
= (V
x
w
V
x
r
)/ V (4)
the lateral creepage
2
= (V
y
w
- V
y
r
)/ V (5)
and the spin

3
=(
z
w
-
z
r
)/V (6)
where V
x
w
and V
y
r
are the rigid body velocities of the wheel in the Ox, Oy directions,
V
x
r
, V
y
r
are the rigid body velocities of the contact point of the rail, the mean veloc-
ity of the wheel along the rail is V = (V
x
w
+ V
x
r
)/ 2, and
z
w
and
z
r
are the angu-
lar velocities of the wheel and rail about the Oz axis.
In the plane of the contact area, the relative velocities vary and are
v
x
= v
x
w
- v
x
r
(7)
v
y
= v
y
w
- v
y
r
(8)
Similarly, the relative displacements in the plane of the contact area will be denoted
by u
x
and u
y
in this Section and are given by
u
x
= u
x
w
- u
x
r
(9)
u
y
= u
y
w
- u
y
r
(10)
A particle will experience a change in velocity through moving to a position where
the displacement has a different value, so that at time t + t the particle which was
originally at x is at x + Vt. Therefore the change in displacement in time t is u
x
(x
+ Vt, t + t) - u
x
(x, t) and hence the rate of change of the relative displacement is
Vu
x
/x - u
x
/t. To this must be added the contribution of creepage from (4), (5)
and (6) so that
v
x
= V(
1
-
3
y) + Vu
x
/x - u
x
/t (11)
v
y
= V(
2
+
3
x) + Vu
y
/x - u
y
/t (12)
The normal pressure acting at any point in the contact area is
z
, given by the
theory of Hertz, and there are tangential tractions
x
and
y
, with resultant
t
, ap-
plied by the wheel to the rail
t
= (
x
2
+
y
2
)
1/2
(13)
The contact area consists of a region in which there is adhesion and a region in
which there is slipping. In the former v
x
= v
y
= 0 and if is the coefficient of limit-
ing friction
|
t
|
z
(14)
and in the region in which there is slip
|
t
| =
z
(15)
and the direction of the resultant traction opposes the slip velocity
x
/ v
x
=
y
/ v
y
=-
t
/ ( v
x
2
+ v
y
2
)
1/2
(16)
A particle entering the contact area is initially unstrained, so that at the leading edge
the tangential tractions must be zero
x
( x
l
, y) = 0 (17)
y
( x
l
, y) = 0 (18)
where
x
l
= a{1 - (y/b)
2
}
1/2
(19)
defines the leading edge.
2.4.4.3 Creep forces for small creepages
For small values of the creepages and spin, there is adhesion over the complete con-
tact area and there is a linear relationship between the creep forces and the creep-
ages. This is of the form
T
1
= - f
11
1
(20)
T
2
= - f
22
2
- f
23
3
(21)
M
3
= f
23
2
- f
33
3
(22)
The moment M
3
can ordinarily be neglected. The general linear case of a three di-
mensional wheel rolling on a three dimensional rail has been analysed by Kalker
[37]. In this case the f
ij
are given by
f
11
= Gc
2
C
11
f
22
= Gc
2
C
22
f
23
= Gc
3
C
23
(23)
where c
2
= ab, G is the elastic modulus of rigidity and the coefficients C
ij
are tabu-
lated in [37]. Table 2.2 gives some example results for the wheel-rail geometry of
Figure 2.2 for the cases where the wheelset is central or displaced laterally by
6 mm. The longitudinal and lateral creep coefficients f
11
and f
22
are similar in size
and assuming their equality is sometimes a useful approximation.
Table 2.2 Example linear creep coefficients N = 39240 N.
y mm -6 0 6
a/b 0.835 1.222 2.014
C
11
3.95 4.33 5.10
C
22
3.46 3.95 4.90
C
23
1.29 1.72 2.64
f
11
MN 8.20 7.44 6.67
f
22
MN 7.19 6.79 6.40
f
23
kNm 13.7 13.7 14.0
2.4.4.4 Creep forces for arbitrary creepages
Kalker [34] has solved equations (7)-(18) in the case of steady motion using the ba-
sic equations of the theory of elasticity to relate the displacement differences u
x
, u
y
to the tractions
x
,
y
. The creep forces are functions of the creepages, the ellipticity
a/b of the contact patch, the normal force N and the coefficient of friction. Kalker
has developed computer programs and tables of results which are in non-
dimensional form [34].
A set of representative results of these calculations is shown in Figure 2.11. In
the absence of spin, and constant lateral creep, as the longitudinal creep increases
the longitudinal creep force eventually achieves the limiting value of N, whilst the
lateral creep force is reduced. An analogous variation occurs for lateral creep. In the
presence of spin, the lateral force is increased at the expense of a reduction in the
longitudinal force. The significant contribution of spin to the lateral force is to be
noted. For small values of the spin the lateral force is proportional to the spin, par-
ticularly at low creepages, but as the spin increases the lateral force reaches a
maximum and then reduces in value at large values of spin. As might be expected
from considerations of symmetry, the lateral creep force is affected equally by posi-
tive or negative values of longitudinal creepage and the longitudinal creep force is
zero if the longitudinal creepage is zero. At very large values of the creepages and
spin, there is pure sliding and the forces correspond to the laws of friction discussed
above.
Experimental measurement of creep forces carried out by Hobbs [38], Illing-
worth [39] and Brickle [40] suggest that Kalkers results are verified when the con-
tacting surfaces are clean. If the surfaces are contaminated the situation is less clear,
the most relevant experiments being those of Pearce and Rose [41]. It seems likely
that the coefficient of friction is affected by surface contamination but the creep co-
efficients are not. Measurements of the resultant forces acting on wheelsets exerting
large tractive efforts are not consistent with the theory as it stands. Knothe et al [42]
point out three main deviations: (a) measurements show large dispersion in the
tractive force-creepage characteristic (b) the maximum value of the tractive force
occurs at a higher value of the creepage than theory predicts (c) beyond this value
of the creepage the measured tractive force decreases. This latter effect can be ex-
plained by assuming that the coefficient of friction depends on the surface tempera-
ture and hence on the sliding velocity. An extension of Carters theory to cater for
this effect shows that the traction forces at large creepages not only depend on the
creepages but also on vehicle speed [43].
2.4.4.5 An approximate theory for arbitrary creepages
An alternative to the complete numerical solution of the problem is the approximate
and faster approach given by Kalkers simplified theory [35] which is implemented
in the computer program FASTSIM. If the flexibility in the contact area was iso-
tropic then
u
x
= f
x
u
y
= f
y
(24)
then substitution in (11) and (12) would give for the area of adhesion

x
x f
y
f
=
1 3
(25)
2
0.01
0
20
1
-T
1
(kN)
3
=0
3
=0.6
2
= 0.001
-T
2
(kN)
0.01 0
10
3
=0
3
=0.6
1
= 0.001
0 0.01
20
2
-T
2
(kN)
3
=0
3
=0.6
2
0
0.01
-T
1
(kN)
3
=0
3
=0.6
0
1
20
3
-T
1
(kN)
2
=0
2
=0.001
1
= 0.002
10
0 1
3
-T
2
(kN)
2
=0
2
=0.001
20
Figure 2.11 Typical relationships between creep forces and creepages and spin. N =39240 N,
= 0.3, r
0
= 0.45 m, R
r
= 149 mm, R
w
= 271 mm.

y
x f
x
f
= +
2 3
(26)
These equations are approximated by

x
x f
y
f
=
1
1
3
3
(27)

y
x f
x
f
= +
2
2
3
3
(28)
where f
1
, f
2
and f
3
are determined so as to obtain results which agree with linear the-
ory for small creepages. This leads to
f
1
= 8a/3C
11
G f
2
= 8a/3C
22
G f a a b C G
3 23
4 = / / (29)
Equations (27) and (28) can then be integrated, subject to the boundary conditions
(13-19).
2.4.4.6 Heuristic approximations
The analysis by Vermeulen and Johnson [31] is valid for arbitrary creepages but not
for spin, and represented a simple way of accounting for creep saturation. It was,
therefore, useful for dynamics analysis. Shen et al [44] proposed an extension to the
method of Vermeulen and Johnson which includes the effects of spin. If the resul-
tant of the creep forces calculated from (20) and (21) is T
L
then the effect of creep
saturation is represented by
if T 3
T N
T
N
T
N
T
N
L L L
=

1
3
1
27
2 3
(30)
if T > 3 T N = (31)
and then the longitudinal and lateral forces from (20) and (21) are reduced by the
ratio T/T
L
.
Another approximation has been given by Polach [45].
2.4.4.7 Non-Hertzian effects
Kalker [24] has extended the simplified theory of rolling contact to cover this case
which can occur particularly in the case of worn wheels and rails. The case of non-
elliptical contact areas has also been considered by Knothe and Le-The Hung [46].
In the case of two-point contact, the complicated non-Hertzian contact patches can
be approximated by one or more elliptic patches [18]. A further assumption of the
Hertz theory is that the contacting surfaces are perfectly smooth. Nayak and Paul
[23, 47] have given an alternative theory which assumes that the friction in the con-
tact area arises from the deformation of friction junctions and thus exploits the
rough surface theory of Greenwood [48]. An analysis by Bucher [49] has shown
that the longitudinal creep coefficient is reduced when the wheel and rail surfaces
are very rough.
2.4.5 Transient effects
Finally, the above analysis considers that steady-state conditions have been estab-
lished and transient effects have been neglected. Newland [50] and Knothe and
Gross-Thebing [51] have considered the calculation of the complex transfer func-
tions for the case of periodically varying creepages. Their results show that the
creep forces deviate significantly from the steady state case only when the wave-
length of the motion approaches the diameter of the contact patch, a situation un-
usual in vehicle dynamics. A crude representation of the transient behaviour for
small creepages is given by
{1 + (2a/V)
d
dt
}T
1
= - f
11
1
(32)
{1 + (2a/V)
d
dt
}T
2
= - f
22
2
- f
23
3
(33)
which will suffice later for a preliminary examination of the effect.
2.5 Creepages
The longitudinal velocity components of the wheel relative to the origin of a coor-
dinate system O123 fixed in the contact patch are
V
1r
w
= r
r
- l

(1)
.
01
r
r
- l
u
y
*
+ r
r
02
u
y
.
Figure 2.12 Velocity components of wheelset at contact point on right hand wheel.
V
1l
w
= r
l
+ l

(2)
where =

(= -V/r
0
when the wheelset is in the central position), so that r
r
is the
forward velocity at the tread in the plane of the wheel. The velocity in the Oy direc-
tion of the wheelset axes is r
r
. The velocity of the wheel normal to the contact
patches must be zero, and so the velocity components of the wheel in the plane of
the wheel at the contact points are, Figure 2.12
V
2r
w
= (
*
u
y +
r
r
) sec
r
(3)
V
2l
w
= (
*
u
y
+ r
l
) sec
l
(4)
The angular velocities about the normal to the contact patch are

3r
w
= sin
r
+

cos
r
(5)

3l
w
= sin
l
+

cos
l
(6)
The forward velocity of the wheelset is measured at the centre of mass and Figure
2.13 shows that the angular velocity of rotation due to the curve is V/(R
0
- r
0
sin
0
).
Hence the longitudinal velocity of the right hand and left hand rails relative to the
contact patch will be V(R
0
lcos
0
)/(R
0
- r
0
sin
0
) or since
0
is small
V
1r
r
= - V (1 - l/R
0
) (7)
V
1l
r
= - V (1 + l/R
0
) (8)
l
l
V/(R
0
- rsin
0
)
0
R
0
- lcos
0
R
0
Figure 2.13 Velocity components in curve.
The other velocity components of the rail relative to the contact area axis system are
V
2r
r
= 0 (9)
V
2l
r
= 0 (10)

3r
r
= -Vcos
r
/R
0
(11)

3l
r
= Vcos
l
/R
0
(12)
Substituting into equations (4.4), (4.5) and (4.6) yields the following expressions for
the creepages
1r
= r
r
/V + 1 - l V
/ - l/R
0
(13)
1l
= r
l
/V + 1 + l V
/ + l/R
0
(14)
2r
= (
*
u
y
+ r
r
) sec
r
/V (15)
2l
= (
*
u
y
+ r
l
) sec
l
/V (16)
3r
= (/V)sin
r
+ (

/ V )cos
r
+ cos
r
/R
0
(17)

3l
= (/V)sin
l
+ (

/ V )cos
l
+ cos
l
/R
0
(18)
2.6 Contact Forces
The forces acting in each contact area, in the contact axis system, are obtained by
evaluation of the creep forces as described in Section 4 using equations (5.13) to
(5.18) to evaluate the creepages. From Figure 2.14 these forces referred to the
wheelset axis system are
T
xr
= T
1r
(1)
T
yr
= T
2r
cos
r
+ T
3r
sin
r
(2)
T
zr
= T
3r
cos
r
- T
2r
sin
r
(3)
T
xl
= T
1l
(4)
T
yl
= T
2l
cos
l
- T
3l
sin
l
(5)
T
zl
= T
3l
cos
l
+ T
2l
sin
l
(6)
The resultant creep forces acting on the wheelset in wheelset axes are derived by
assuming that the forces act on the wheelset at the positions of the contact areas
when the wheelset is in the central position (with one exception discussed below).
They are then given by
T
x
= T
xr
+ T
xl
(7)
T
y
= T
yr
+ T
yl
(8)
T
z
= T
zr
+ T
zl
(9)
M
x
= T
zr
l - T
zl
l - T
yr
r
r
- T
yl
r
l
(10)
M
y
= ( T
xr
+ T
yr
T
zr
tan
r
)r
r
+ ( T
xl
+ T
yl
+ T
zl
tan
l
)r
l
(11)
M
z
= T
xl
l - T
xr
l (12)
T
1r
T
1l
T
y
T
3r
T
x
M
z
T
y
T
2r
T
3l
T
2l
M
x
T
z
02
03
03
02
01 01
Figure 2.14 Definition of axes and creep forces.
The terms T
zr
tan
r
and T
zl
tan
l
in equation (11) arise from the longitudinal
shift of the contact point from vertically below the wheelset axis when the wheelset
is yawed, Figure 2.15. The importance of this term was pointed out by Gilchrist and
Brickle [53] who refer to Matsui [54] and Bodecker [55].
2.7 Kinematics of the Wheelset
Referring to the axis systems defined in Section 2.2, the vector velocity of the
wheelset with respect to fixed axes instantaneously aligned with Oxyz is v where
v =
a
v +
b
v (1)
the velocity of O is
a
v = [ V 0 0 ]
T
(2)
and the velocity of O* relative to O is
b
v = [ u u u
x y z
]
T
(3)
The angular velocity vector of the wheelset with respect to Oxyz is
T
xl
r
l
tan
l
T
zl
T
yl
r
l
tan
l
r
l
T
xl
Figure 2.15 Effect of longitudinal shift of contact point.
=
a
+
b
+
c
(4)

where
a
is the angular velocity vector of the reference frame Oxyz with respect to
Oxyz (due to track geometry and vehicle forward speed),
b
is the angular velocity
vector of O*x*y*z* with respect to Oxyz, and
c
is the angular velocity vector of
the wheelset due to wheelset rotation about its axis of revolution. Therefore
a
= [

0
Vsin
0
/R
0
Vcos
0
/R
0
]
T
(5)
b
= [

0

]
T
(6)
and if the angular velocity vector of the wheelset about its axis is
d
where
d
= [ 0

0 ]
T
(7)
then
c
= G
d
(8)

where G is the rotation matrix ( G is orthogonal, G
T
G = I
33
the 3 3 unit matrix),
the elements of which depend on the relative angular orientation of the two frames.
By carrying out successive rotations of the axes it is found that
G =
cos cos sin sin sin

sin cos cos sin cos
sin cos

0
(9)
If the rotations and are small the order in which the rotations take place is
immaterial and (9) reduces to
=
1 0
1
0 1
G (10)
As the variation in the rotational speed of the wheelset will usually be small
= - V/r
0
+ (11)
where is small. Then to first order
= [

0
+

+ V/r
0
, Vsin
0
/R
0
- V/r
0
+ , Vcos
0
/R
0
+

- V/r
0
]
T
= [
x
z
]
T
(12)

2.8 Equations of Motion
With the axis systems defined as above, the wheelset reference axes are not fixed in
the wheelset and are themselves moving with respect to inertial space. The formula-
tion of the equations of motion therefore require special care and their logical deri-
vation has been considered by de Pater [4] and Schiehlen [56]. The derivation given
here follows that of de Pater whose paper should be consulted for rigorous detail,
and applies the principles of linear and angular momentum referring the forces and
moments, and the linear and angular displacements to the wheelset reference axis
system Oxyz.
Assuming that the displacements u
i
and their derivatives are small, the vector
equations of motion if the axes are moving are
p +
a
p = X (1)
h +
a
h = L (2)
where

(the tilde will be used to indicate the skew symmetric matrix corresponding to a vec-
tor quantity with the elements arranged as above). In (1) and (2)
[ ]
p m u V u u
x y z
T
= + (4)
[ ]
h
x y z
T
= (5)
where is defined with respect to Oxyz and therefore varies with and . If I* is
defined with respect to O*x*y*z* then
I
I
I
I
y
* =
0 0
0 0
0 0
(6)
so that, to first order,
a
a a
a a
a a
0
0
0
3 2
3 1
2 1
(3)
= =
GI G
I I I
I I I I I
I I I
T
y
y y y
y
*
( )
( ) ( )
( )
0
0
(7)
Evaluation of (1) and (2), neglecting second order terms, then leads to the equations
of motion of the unconstrained wheelset written in the form
Ar F r ( ) + = 0 (8)
where
r = [ u
x
u
y
u
z
]
(9)
and
A = diag [ m m m I I
y
I ] (10)
The forces F(r) are given by
F(r) =
c
F +
g
F +
w
F +
f
F +
s
F+
n
F (11)
where
c
F are the centrifugal forces
c
F = [ 0 mV
2
cos
0
/R
0
- mV
2
sin
0
/R
0
0
I
y
V (d/dt)(sin
0
/R
0
) IV(d/dt)(cos
0
/R
0
) ]
T
(12)
g
F are the gyroscopic forces
g
F = [ 0 0 0 I
y
V (

+ Vcos
0
/R
0
)/r
0
0 -I
y
V (

+

0
)/r
0
]
T
(13)
In (12) and (13) it is assumed that V/R
0
is a first order quantity.
w
F are the gravita-
tional forces
w
F = [ 0 - mgsin
0
- mgcos
0
0 0 0 ]
T
(14)
f
F are the creep forces dependent on the terms in equations (6.1)-(6.6) in
T
1r
, T
2r
, T
1l
, T
2l
only and
n
F are the normal forces dependent on the terms in equa-
tions (6.1)- (6.6) in T
3r
and T
3l
only, so that
f
F +
n
F = [ - T
x
- T
y
- T
z
- M
x
- M
y
- M
z
]
(15)
s
F are the forces applied by the rest of the vehicle and which are necessary to sus-
tain the motion of the wheelset along the track
s
F = [ P
x
P
y
P
z
L
(16)
Written out in full, these equations are
mu
x
- T
x
+ P
x
= 0 (17)
mu
y
- T
y
+ mV
2
cos
0
/R
0
- mgsin
0
+ P
y
= 0 (18)
mu
z
- T
z
- mV
2
sin
0
/R
0
- mgcos
0
+ P
z
= 0 (19)
I
x

+ I
x

0
- M
x
- I
y

(

+ Vcos
0
/R
0
) + L

= 0 (20)
I
y

- M
y
+ I
z
V(d/dt)(sin
0
/R
0
) + L

= 0 (21)
I
z

+ I
y

(

+

0
) - M
z
+ I
z
V(d/dt)(cos
0
/R
0
) + L

= 0 (22)
2.9 Constrained Motion
The actual motion of the wheelset is constrained. The motion of the wheelset is as-
sumed to be constrained to be at constant speed V so that u V
x
= and relative to the
wheelset frame of reference u
x
= 0. Furthermore, the wheelset is constrained to re-
main in contact with the track at the two points of contact. As
n
F are the normal
forces dependent on the terms in equations (6.1)-(6.6) in T
3r
and T
3l
only, from equa-
tions (6.7)-(6.12) it is possible to write
n
F = HT (1)
where T = [T
3r
T
3l
]
T
. The normal forces T
3r
and T
3l
are given by the theory of
Hertz, and may be expressed as an equivalent linear stiffness k
h
so that if the normal
deflections at the contact points on the right and left hand sides respectively are e =
[ e
3r
e
3l
]
T
.
T = k e (2)
where k = diag[ k
hr
k
hl
]

and
e = H
r (3)
Substituting (3) and (2) into (1) and (8.8) results in the following form of the equa-
tions of motion
A r + F
0
(r) + HkH
T
r = 0 (4)
where F
0
=
c
F +
g
F +
w
F +
f
F +
s
F. One approach to the solution of the equa-
tions of motion is to solve these equations directly. (This may be particularly appro-
priate in the case of two-point contact. For methods of computation in this case see
[57], [58] and [59], for example). However, as the Hertz stiffness is large, the nor-
mal deflections will be small in comparison with the displacements associated with
the gross motion of the wheelset. Therefore, a new set of generalised coordinates q
,
q
y
and q

is defined which represent motions in which there is no normal displace-
ment of the contact area, so that
r = Jq (5)
or
where z
y
and
y
are dependent on q
y
through the specified wheel-rail geometry as
expressed by equations (3.14) and (3.15). Since u = 0 for arbitrary values of q, on
substitution from (5) into (3), to first order
H
T
J = 0
23
(6)
where 0
23
is the 2 3 null matrix. Therefore premultiplying (8.8) by J
T
the motion of
the system is governed by the set of three equations of motion
J
T
AJ q + J
T
F = 0 (7)
Hence the equations of motion become
I
y q
- R
+ I
z
V(d/dt)(sin
0
/R
0
) + L
= 0 (8)
{m( 1 + z
y
2
)

+ I
x
y
2
} q
y
+ I
x
0
+ mV
2
cos
0
/R
0
- mgsin
0
- z
y
mV
2
sin
0
/R
0
- z
y
mgcos
0
-
y
I
y
q
( q
+ Vcos
0
/R
0
) - R
y
+ P
y
+ P
z
z
y
+ L
y
= 0 (9)
I
z
q
+ I
y
q
(
y
q
y
+

0
) + I
z
V(d/dt)(cos
0
/R
0
) - R

+ L

= 0 (10)
where
u
u
u z
x
y
z y
y
0 0 0
0 1 0
0 0
0 0
1 0 0
0 0 1
q
q
q
y

R
y
= T
y
+ T
z
z
y
+ M
x
y
(11)
R
= M
y
(12)
R
= M
z
(13)
are the generalised creep forces. Substituting into equation (11) from equations (6.8)
to (6.10) and (3.14) and (3.15) and neglecting terms of third order in contact angles
R
y
= T
2r
cos
r
+ T
2l
cos
l
- z
y
T
2r
sin
r
+ z
y
T
2l
sin
l
-
y
lT
2r
sin
r

-
y
lT
2l
sin
l
-
y
r
r
T
2r
cos
r
-
y
r
l
T
2l
cos
l
(14)
which is independent of T
3r
and T
3l
to first order, as expected, as the normal reac-
tions do no work, the normal relative velocities being zero at the contact points.
Similarly, substituting in equations (6.11) from equations (6.1) to (6.6)
M
y
= (T
1r
+ T
2r
/cos
r
)r
r
+ (T
1l
+ T
2l
/cos
l
)r
l
(15)
T
3r
and T
3l
can be determined by substituting
y y z
u z u

= and

=
y y
u (where z
y
and
y
are given by (3.14-15)) into (8.19) and (8.20) respectively, and eliminating u
y
using (8.18). P
y
, P
z
, L
x
and L
z
are forces and moments transmitted by the suspen-
sion and are a function of the relative motion between the wheelset and the rest of
the vehicle. The evaluation of these forces requires consideration of a complete
vehicle which is considered later. If the applied braking or tractive torque L
is zero
then the solution of the coupled equations (8)-(10) will yield the wheelset rotational
speed

, and (8.17) will give the value of the longitudinal force necessary to sus-
tain the motion at constant speed V. Otherwise both
and V can be specified and

(8) will give the torque, and (8.17) will give the longitudinal force, necessary to sus-
tain the steady values of V and
. The various steps in the solution of the equations

of motion are shown in Figure 2.16.
Consideration of practical values shows that a number of small terms can be ne-
glected in the equations of motion (8)-(10). In (9) as z
y
and
y
are small {m( 1 + z
y
2
)
+ I
x
y
2
} can be replaced by m.
y
q
y
is small compared with unity and can be ne-
glected though the gyroscopic terms will be retained for future reference. It will be
adequately accurate to put cos
0
= 1 and sin
0
=
0
. The term z
y
mgcos
0
= z
y
mg
represents a restoring force generated by the gravitational stiffness and arises
from the raising of the centre of mass as the wheelset is displaced laterally.
The rate of working of the various forces acting on an unrestrained wheelset when
running on straight level track is obtained by multiplying equations (8-10) respec-
tively by q
, q
y
, and q
, and multiplying (8.17) by V and adding, giving

Figure 2.16 Steps in the solution of the equations of motion (heuristic creep model).
GEOMETRY
Equs (3.1-11)
u
y
y

0
R
0
y
0
, r r R
r
R
w
CREEPAGES
Equs (5.13-18)
CONTACT AREAS
Equs (4.1)
COEFFICIENTS
Equs (4.23)
NORMAL FORCES
Equs (8.19-20)
CREEP FORCES
Equs (4.20-22, 30-31)
FORCES
Equs (6.7-12)
EQUATIONS OF MOTION
Equs (9.8-10)
PATH
Equs (2.1-3)
f
ij
T
1
T
2
T
1
T
2
Q
y
M
y
M
z
X Y

a b
T
3
u
y
y
d
dt
T U R q R q R q L q P V
G y y x
( ) + + + =

0 (17)
where the kinetic energy T of the system is
T m z I q mV Iq I q
y x y y y
= + + + + +
1
2
1
1
2
1
2
1
2
2 2 2
2
2
2
{ ( ) }

(18)
the potential energy U
G
is
U mgz q
G y y
= (19)
Substituting for R
y
, R
, and R
from equations (12)-(14), (6.1)-(6.12) and (5.13)-

(5.18)
d
dt
T U VT VT VT VT
G r r l l r r l l
( ) +
1 1 1 1 2 2 2 2
+ + = L q P V
x
0 (20)
This follows from the definition of the creepages, equations (5.1318).
2.10 Equations of Motion for Small Displacements
Much of importance can be discussed using a linearised form of the equations of
motion. These may be derived either by assuming that the motions are of very small
amplitude, as in this section, or, more realistically, by adopting describing functions
as discussed in Section 3.
To the first order in q
y
= =
/ V r
0
(1)
so that the equations of motion for the linearised system will have two generalised
coordinates q
y
and q
. For simplicity, q
y
and q
will be represented by y and

henceforth. The corresponding creepages are given by equations (5.13) - (5.18), if
0
,
r
and
l
are small and r
0
<< R
0
, as

1r
= - r
r
/r
0
+ 1 - l V
/ - l/R
0
(2)
1l
= - r
l
/r
0
+ 1 + l V
/ + l/R
0
(3)
2r
= y /V - (4)
2l
= y /V - (5)

3r
=
r
/r
0
+

/ V + 1/R
0
(6)

3l
=
l
/r
0
+

/ V + 1/R
0
(7)
The above equations show that when the wheelset is central, on straight track,
1r
=
1l
=
2r
=
2l
= 0 (8)
but
3r
= -
3l
= -
0
/ r
0
(9)
Substituting from equations (4. 20) - (4.21) there are corresponding forces
T
1r
= T
1l
= 0 (10)
T
2r
= - T
2l
= - f
23
3r
= f
23
0
/ r
0
(11)
In addition,
T
3r
= T
3l
= - N
0
(12)
so that N
0
is the normal force when the wheelset is central. As pointed out by de
Pater [55], because of this steady-state spin there will be a first-order variation of the
lateral creep force due to the change in the normal force as the wheelset is displaced,
Figure 2.17. Note that by symmetry N
r
= - N
l
= N. De Pater [4] and Yang [60]
have carried out the analysis for arbitrary values of
0
. Equations (4.20)-(4.22) must
be extended and to first order
T
1r
= - f
11
1r
(13)

0
f
23
/r
0
N
N
0
f
23
/r
0
N
0
0
f
23
/r
0
N
0
f
23
/r
0
N
0
N
N
0
N
0
Figure 2.17 Variation of the lateral creep force due to the change in the normal force as the
wheelset is displaced.
T
1l
= - f
11
1l
(14)
T
2r
= - f
22
2r
- f
23
3r
+
0
f
N
23
N/r
0
(15)
T
2l
= - f
22
2l
- f
23
3l
+
f
N
23
N/r
0
(16)
where the f
ij
are evaluated with the wheelset central. From equation (4.23) it can be
seen that f
23
is proportional to c
3
and hence from (4.1)
f
N
23
= f
23
/N
0
(17)
The variation of normal force with lateral displacement of the wheelset is found
from the equation of rolling moments, (8.20). The rolling moment of the creep
forces is given by equation (6.10). The creep forces are given by equations (13)-(16)
and the creepages and spins by (2)-(7). Using the geometric parameters from (3.24)
and (3.30) then
M
x
= 2f
22
r
0
( y /V - ) + 2f
23
r
0
(-
0
*
y/ r
0
l +

/V + 1/R
0
)
+ 2N
0
(
0
0
y +
0
*
r
0
y/l) - 2Nl(1 + f
23
0
/N
0
l) (18)
where it is assumed that
0
l << r
0
and
0
r
0
<< l and second order terms in the dis-
placements are neglected. Substituting from (18) into (8.20) and using (1) yields
N = {-I
x
- I
x
0
- I
y
V(

+ V/R
0
) /r
0
+ 2f
22
r
0
( y /V - ) +
2N
0
(
0
0
y +
0
*
r
0
y/l) - L

+ 2f
23
r
0
(-
0
*
y/ r
0
l +

/V + 1/R
0
)}/ 2l(1 + 2f
23
/N
0
) (19)

The generalised creep forces are given by (9.11-13) and to first order are
R
y
= - 2f
22
( y /V - ) - 2f
23
(-
0
*
y/ r
0
l +

/V + 1/R
0
) + 2N
0
f
23
/N
0
r
0
(20)
R
= - 2f
11
(
0
yl/ r
0
+ l
2
/V + l
2
/R
0
) (21)
Substituting from (20) and (21) into (9.9) and (9.10) the equations of motion take
the form
[ As
2
+ (B/V + GV)s + C + E
g
+ E] q = Q (22)
where q = [ y ]
T
, s is the operator d/dt and Q represents the externally applied
forces. A is the inertia matrix, B and C are the creep damping and stiffness matrices,
G is the gyroscopic matrix, E
g
is the gravitational stiffness matrix and E is the elas-
tic suspension stiffness matrix. A, B, E
g
and E are symmetric matrices but C is not
symmetric indicating that the system is non-conservative. The contribution of the
variation of the lateral creep force due to change in the normal force is, for typical
values of the parameters, negligible except in the case of G where G
12
is multiplied
by a factor

=
0
( 1 - f
23
/N
0
r
0
) (23)
If, moreover, the small term M
3
from equation (4.22) is retained, then the system
matrices adopt the simple form
A
m
I
=
0
0
(24)
B
f f
f f l
=
2 2
2 2
22 23
23 11
2
(25)
C
f r l f
f l r f
=

2 2
2 2
23 0 0 22
11 0 0 23
*
/
/
(26)
G
I r l
I r l
y
y
=
0
0
0
0
/
/
(27)
E
W l
Wl
g
=
0
0
0
0
*
/
(28)
and representing the action of a lateral spring of stiffness k
y
and a yaw spring of
stiffness k

E
k
k
y
=
0
0
(29)
The generalised applied forces are given by
Q
mV R mg I V R r f R
I V r I VsR f l R
y
y z
=
+ + +

2
0 0
2
0 0 23 0
0 0 0 11
2
0
2
2
/ / /
/ /

(30)
Written out in full, the equations of motion are

Q K V l f I y l r V I V f r ly f
Q l r V I V f f K V y f y m
z y
y y y
= + + +
= + + +
/ 2 ) / / 2 ( / 2
) / / 2 ( 2 / 2
2
11 0 0 23 0 0 11
0 23 22 22

where
K
y
= k
y
+ (2N
0
0
*
/l)( 1 - f
23
/N
0
r
0
) (33)
K

= k

+ 2N
0
l (-
0
+ f
23
/N
0
l) (34)
In the absence of any external forces, premultiplying (22) by q
T
yields the equation
of energy balance
d
dt
T U U U q Bq V q Cq
G F
T T
( ) / + + + + + =
0 (35)
where
T q Aq
T
=
1
2
(36)
U q Eq
T
=
1
2
(37)
U q E q
G
T
g
=
1
2
(38)
U q C C q
F
T T
= +
1
4
( ) (39)
The skew symmetric matrix
(31)
(32)
C C C
T
=
1
2
( ) (40)
clearly plays an important role in the stability of the system. The kinetic and poten-
tial energy may be increased or decreased depending on whether or not the dissipa-
tion represented by the damping coefficients is greater than the energy added by the
skew-symmetric stiffnesses. It should be noted, however, that this equation of en-
ergy balance (35) does not take into consideration the forces that are necessary to
sustain the forward motion of the wheelset and is not the complete energy balance.
It does not indicate the source of energy which distinguishes the non-conservative
system from a conservative system. In order to maintain a constant forward speed V
work must be done on the system by a corresponding longitudinal force P
x
, or trac-
tive torque L
. Moreover, + q Bq q Cq
T T
is not the total rate of working of the
creep forces. This is given by
2F = V(f
22
2r
2
+ f
22
2l
2
+ f
11
1r
2
+ f
11
1l
2
) (41)
so that F is a dissipation function for the creep forces, analogous to Rayleighs dis-
sipation function. Note that f
23
does not appear in this equation. Substituting into
(9.20) from (13)-(16) then yields the corresponding version of the energy balance
equation
d
dt
T U U F
G
( ) + + + 2 = L V r P V
x
/
0
0 (42)
Reference to Figure 2.18 which shows the forces acting on the wheelset including
the forces necessary to sustain the motion shows that to the second order in the gen-
eralised coordinates
P f y V
x
= 2
22
( / ) (43)
L f y
y
r V

= + 2
11
0
(

) (44)
Substitution of (43) and (44) into (42) shows that (42) and (35) are equivalent and
thus the source of energy which promotes the active behaviour of the system is the
tractive effort applied either as a longitudinal force or as a driving or braking
torque.
The discussion of energy balance demonstrates the active and nonconservative
nature of the system which is indicated by the assymmetry of the terms coupling lat-
eral displacement of the wheelset with yaw, C
12
and C
21
, which arise from the creep
forces acting in conjunction with the conicity of the wheels. Stiffnesses also arise
from forces acting in the contact area and comprise gravitational stiffnesses in the
matrix E
g
(E
g11
is generated by the tilting of the normal reaction as the wheelset is
displaced) and terms arising from the lateral creep force generated by spin. The gy-
roscopic couplings G
12
and G
21
are due to the tilting of the angular momentum vec-
tor as the wheelset is displaced laterally. G
12
is modified because of the first-order
variation of the lateral creep force due to the change in the normal force as the
wheelset is displaced, as discussed above. For a wide range of practical parameters
f
23
/N
0
r
0
= 4/5 approximately, so that =
0
/5, and G
12
= G
21
/5 approximately.
The total contact stiffness
C
11
+ E
g11
= (2N
0
0
*
/l)( 1 - f
23
/N
0
r
0
) (45)
so that four-fifths of the gravitational stiffness effect is counteracted by the lateral
force due to spin. Similarly,
C
22
+ E
g22
= 2N
0
( -
0
+ f
23
/N
0
l) (46)
so the stiffness is positive though very small.
These considerations can be illustrated by a representative set of parameters for a
restrained wheelset shown in Table 2.3. Parameters for the wheel rail geometry are
taken from Figure 2.5 and the creep coefficients are from Table 2.2 and are consis-
tent with Figure 2.2 and the axle-load.
A significant feature of these parameters is that they give system creep coeffi-
cients B
11
= C
12
= 13.6 MN, B
22
= 8.26 MNm and C
21
= 2.82 MNm, which are large
T
1r
r
r
T
1l
T
2r
T
2l
r
l
T
2l
T
2r
P
x
L
Figure 2.18 Energy balance for wheelset.

m = 1250 kg I
z
= 700 kgm
2
I
y
= 250 kgm
2
W = 78.48 kN
k
y
= 0.23 MN/m k
= 2.5 MNm/rad c
y
= c
= 0
r
0
= 0.45 m l = 0.7452 m
0
= 0.1174
0
= 6.423
0
= 0.0493 = 0.0508
f
11
= 7.44 MN f
22
= 6.79 MN f
23
= 13.7 kN
relative to the other elements in the system matrices. It was assumed in the above
derivation that m >>
0
2
I
x
/l
2
and
0
r
0
/l << 1. For the above parameters
0
2
I
z
/l
2
=
3.064 and
0
r
0
/l = 0.0298 thus confirming the validity of this assumption. The con-
tact stiffnesses C
11
+ E
g11
= 0.151 MN/m and C
22
+ E
g22
= 0.0245 MNm which,
in this case, are small compared with k
.
In view of the relative magnitudes of the system coefficients it is a useful ap-
proximation to neglect f
23
, E
g
and G giving the simplified equations of motion
Table 2.3 Parameters for an elastically restrained wheelset.
+
1/(ms
2
+ 2f
22
s/V + k
y
)
Y
y
1/(Is
2
+ 2f
11
l
2
s/V + k
)
-2f
22
2f
11
l/r
0

-
-
+
A
Figure 2.19 Feedback representation of wheelset.
m y + 2f
22
( y /V - ) + k
y
y = 0 (47)
2f
11
ly/r
0
+ I
z

+ 2f
11
l
2
/V + k
= 2f
11
l/R
0
(48)
the similarity to a closed loop feedback control system being indicated in Figure
2.19. It is this form of the equations of motion that were originally derived by Carter
as discussed in Chapter 1.
For certain calculations, it is useful to replace s/V by D. D therefore represents
the operation of differentiation with respect to distance travelled along the track. At
low speeds, so that the terms in V
2
may be neglected, the equations are path depend-
ent rather than time dependent and the equations of motion become
[BD + C + E
g
+ E] q = Q (49)
In many applications it is advantageous to cast the equations of motion into the
equivalent first-order or state-space form. For future reference these are
[ ]
x q q
T
= (50)
x = Ax + Y
(51)
where the system matrix A is
where 0
22
and I
22
are the 2 x 2 null and unit matrices and
Y
T
= [ 0
12
A
-1
Q ] (53)
where 0
12
is the 1 2 null matrix.
In many wheelset motions the creepages are often small and it is then reasonable
to express the creep forces in terms of the linear relationship between creep forces
and creepages. This leaves only the geometry as the significant nonlinearity and the
rolling radius difference, the slope difference and the roll angle may be expressed as
linear functions of the lateral displacement y in terms of the describing functions ,
and as discussed in Section 3. The equations may then be described as quasi-
linear as the coefficients will be dependent on a the amplitude of the lateral dis-
placement of the wheelset. It is obvious that the equations so derived will be of the
same form as (47) and (48).
A
+ + +
=

) / ( ) (
1 1
22 22
GV V B A E E C A
I O
g
(52)
Figure 2.20 Generalised coordinates for two-axle vehicle.
2.11 Equations of Motion for a Two-Axle Vehicle
Consider an idealised two-axle vehicle with a body and two wheelsets, Figure 2.20.
The wheelsets and car body are assumed to be rigid and connected by mass-less
elastic structures and mass-less damper elements. The vehicle is symmetric about a
longitudinal plane of symmetry. The car body has six degrees of freedom, longitudi-
nal x
b
, lateral y
b
and vertical translation z
b
, roll
b
, pitch
b
and yaw
b
and, as de-
fined above, each wheelset has three degrees of freedom, lateral translation, rotation
about the axle centreline and yaw. Consequently, the motion of the vehicle is de-
fined by the following set of twelve generalised coordinates
q =[ y
1
1
x
b
y
b
z
b
b
y
2
2
]
T
(1)
For the wheelsets the generalised coordinates y
i
and
i
are measured with respect to
the centreline of the track and the normal to the centreline of the track in the plane
containing the axle when the wheelset is central. For the leading wheelset, the origin
of these coordinates is defined by X
1
, Y
1
and azimuth
1
of the wheelset, Figure
2.21. With reference to fixed axes these are given by
s1
X
1
= cos
1
ds (2)
0
s1
Y
1
= sin
1
ds (3)
0
s1
1
= (1/R
1
) ds (4)
0
Similarly, X
2
, Y
2
and
2
for the trailing wheelset are given by integrals similar to (2),
(3) and (4) with a limit of integration s
2
= Vt - 2h. For the vehicle body, the origin of
the axis system is defined by the intersection of O
1
O
2
and the normal to the track
centreline, Figure 2.21.
The arrangement of the suspension for this simple two-axle vehicle is shown in
Figure 2.22. For ease of exposition the dampers will be assumed to be located simi-
larly to the springs. Adaptation for the more general case is obvious. In terms of the
chosen generalised coordinates the compressive strains in the springs are given by
the equation of compatibility
= a q (5)
where
= [
x1
y1
z1
x2
y2
z2
2
]

(6)
=
0 1 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 1 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 1 0 0 0 0 0 0 1 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 1
0 0 0 0 0 0 0 0 1 0 0 0
h d
h d
a (7)
The forces exerted by the suspension on the car body are given by
P = kaq + ca q (8)
f(X, Y)
2
O
2
O
1
O
b
h
b
X
2
X
1
Y
2
Y
1
O
Figure 2.21 Path coordinates for two-axle vehicle.
where
P = [P
x1
P
y1
P
z1
L
1
L
1
P
x2
P
y1
P
z1
L
2
L
2
] (9)
and k is a matrix of component stiffnesses
k = diag [k
x
k
y
k
z
k
k
x
k
y
k
z
k
] (10)
c is a matrix of component dampings
c = diag [c
x
c
y
c
z
c
c
x
c
y
c
z
c
] (11)
The suspension is strained when the generalised coordinates are zero in the chosen
axis system. Firstly, it is necessary to account for the fact that the vertical suspen-
sion is loaded when the vehicle is running in the central undisturbed position on the
track, and this initially deflected vertical displacement will define the vertical origin
of the axis system for the car body. The vertical suspension forces P
z10
and P
z20
when z
b
= 0 are given by
P
z10
= P
z20
= - (m
b
V
2
sin
0
/R
0
+ m
b
gcos
0
)/2 (12)
For simplicity, if the vertical stiffness is assumed to be linear then the initial vertical
deflection will be given by
z
0
= { m
b
V
2
sin
0
/R
0
+ m
b
gcos
0
}/2k
z
(13)
Secondly, the yaw suspension is strained when the vehicle is on curved track and the
generalised coordinates are zero due to the choice of axes. Yaw moments are exerted
k
y
/2 k
y
/2
h h
k
k
y
/2
k
y
/2
k
y
k
d
Figure 2.22 Arrangement of suspension for simple two-axle vehicle.
on the car body as follows
L
10
= k
(
1
-
b
) (14)
L
20
= k
(
2
-
b
) (15)
and these must be added to the forces defined by equation (8). It will be noted that
the origin of the axis system for the car body is laterally displaced by h
2
/2R
0
from
the centreline of the track.
The equations of motion will then consist of two sets of equations of the form of
(9.8) to (9.10) for the wheelsets, together with the six equations for the car body.
m x P P
b b x x
=
1 2
0 (16)
m x m V R m g P P
b b b y y
cos / sin + =
2
0 0 0 1 2
0 (17)
m z m V R m g P P
b b b z z
sin / cos + =
2
0 0 0 1 2
0 (18)
I
xb b
- L
1
- L
2
+ dP
y1
+ dP
y2
= 0 (19)
I
yb b
+ hP
z1
hP
z2
dP
x1
dP
x2
= 0 (20)
I
zb b
+ I
zb
V(d/dt)(1/R
0b
) - L
1
- L
2
hP
y1
+ hP
y2
= 0 (21)
The equation (9.9) for the wheelsets will have the terms

- z
y
mV
2
sin
0
/R
0
- z
y
mgcos
0
+ P
z
z
y

replaced by
- z
y
(m + m
b
/2){V
2
sin
0
/R
0
+ gcos
0
}

by virtue of (12), so that the gravitational stiffness is proportional to axle-load.
The generalised forces arising from the suspension corresponding to the coordi-
nates of equation (1) are given by the application of the principle of virtual work and
are
Q
s
= - a
T
P (22)
he equations of motion will be of the form
A q + F( q , q, V) + D( q - q
0
) + E( q - q
0
) = Q (23)
where q
0
represents the straining of the suspension due to the choice of axes and A is
the inertia matrix
A = diag [m I I
y
m
b
m
b
m
b
I
xb
I
yb
I
zb
m I I
y
] (24)
and E and D are the suspension elastic stiffness and damping matrices obtained by
substitution of (8) into (22) so that
E = a
T
ka D = a
T
ca (25)
In simple discussions of stability and guidance it is frequently possible, and is very
convenient, to make use of the fact that a vehicle possesses a longitudinal plane of
symmetry. Small disturbances of the car body which are symmetric about the plane
of symmetry, namely x
b
, z
b
and
b
, will not introduce variations in the lateral forces
and yaw moments acting on the wheelsets. The symmetric degrees of freedom
therefore do not couple with the antisymmetric degrees of freedom. It is the latter
motions which describe the stability and guidance of the vehicle. Then, the motion
of the vehicle is defined by the following set of nine generalised coordinates
q =[ y
1
1
y
b
b
y
2
2
]
T
(26)
As far as the wheelset is concerned, for small displacements the angular rotation
/ = V r
0
, and then the motion of the wheelset is determined by the two general-
ised coordinates y and which describe motions which are antisymmetric about the
plane of symmetry. Then

q =[ y
1
1
y
b
b
y
2
2
]
T
(27)
Of course, these approximations are not necessary when numerical solutions of the
full nonlinear equations of motion are carried out but simplification of the equations
is a valuable aid to an appreciation of system behaviour. Consistent with (27), E is
then given by
E
k k k d k h
k k
k k k d k
k d k k d k d k k d
k h k k k h k h k
k k d k h k
k k
y y y y
y y y y
y y y y
y y y
y y y y
=

+
+
0 0 0
0 0 0 0 0
0 2 2 0 0
2 2 2 0 0
0 0 2 2
0 0 0
0 0 0 0 0
2
2

(28)
and D is of similar form to E. For a real vehicle with a complex suspension system
with possible flexible structural components E and D (or the nonlinear equivalents)
would obviously be much more complex and depend on design detail.
References
1. Gostling, R.J.: The measurement of real wheel and track profiles and their use in
finding contact conditions, equivalent conicity and equilibrium rolling line. British
Rail Research Technical Note TN DA 22, 1971.
2. Anon.: Geometry of contact between wheelset and track, part 1, methods of
measurement and analysis. ORE Report C116 RP3, 1973.
3. Cooperrider, N.K., Hedrick, J.K., Law, E.H. Kadala,, P.S. and Tuten, J.M.: Ana-
lytical and experimental determination of nonlinear wheel/rail geometric constraints.
Report FRA-O&RD 76-244, US Dept. of Transportation, Washington, 1975.
4. De Pater, A.D.: The motion of a single wheelset along a curved track. Delft Uni-
versity of Technology, Laboratory for Engineering Mechanics Report 1072, 1995.
5. Yang, Guang.: Dynamic Analysis of Railway Wheelsets and Complete vehicle
systems. Delft University of Technology, Faculty for Mechanical Engineering and
Marine Technology, Doctoral Thesis, 1993. pp. 42-50.
6. Heumann, H.: Zur Frage des Radreifen-Umrisses. Organ Fortschr. Eisenb.-wes.
89 (1934), pp. 336-342.
Heumann, H.: Grndzuge der Fhrung der Schienenfahrzuge, Oldenbourg,
Munchen, 1953, p. 133.
7. Wickens, A.H.: The dynamic stability of railway vehicle wheelsets and bogies
having profiled wheels. Int. J. Solids and Structs. 1 (1965), pp. 319-341.
8. Blader, F.B.: Free lateral Oscillations in long freight Trains. Doctoral Thesis,
Queens University, Kingston, Canada, 1972.
9. Joly, R.: Study of the transverse stability of a railway vehicle running at high
speed. Rail International, 3, No. 2 (1972), pp. 83-118.
10. Gilchrist, A.O., Hobbs A.E.W., King, B.L. and Washby, V.: The riding of two
particular designs of four-wheeled railway vehicle, Proc. Instn. Mech Engrs. 180,
Part3F (1965-66), pp. 99-113.
11. Pearce, T.G.: Wheelset guidance-conicity, wheel wear and safety. Proc. Instn.
Mech Engrs. J. Rail and Rapid Transit. 210, Part F (1996), pp. 1-10.
12. Hauschild, W.: Die Kinematik des Rad-Scheine Systems. Institut fr Mechanik,
Technische Universitt Berlin, 1977.
13. Duffek, W.: Contact geometry in wheel rail mechanics. Proc. Symp. Contact
Mechanics and Wear of Rail/Wheel Systems, (Ed.) J. Kalousek et al., University of
Waterloo Press, 1982, pp. 161-179.
14. de Pater, A.D.: The geometric contact between wheel and rail. Vehicle System
Dynamics, 17, No. 3 (1988), pp. 127-140.
15. Mller, C.Th.: Kinematik, Spurfhrungsgeometrie und Fhrungsvermogen der
Eisenbahnrdsatz. Glasers Annalen, 77 (1953), pp. 264-281.
16. Yang (ibid), pp. 16-42.
17. Netter, H. Schupp, G. Rulka, W. and Schroeder, K.: New aspects of contact
modelling and validation within multibody system simulation of railway vehicles. In
L. Palkovics (Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc. 15th
IAVSD Symposium, Budapest, August 1997, pp. 246-269. Swets and Zeitlinger
Publishers, Lisse, 1992.
18. Pascal, J.P.: About multi-Hertzian contact hypothesis and equivalent conicity in
the case of S1002 and UIC60 analytical wheel/rail profiles. Vehicle System Dynam-
ics 22 (1993), pp. 263-275.
19. Johnson, K.L.: Contact Mechanics. Cambridge University Press, Cambridge,
1985.
20. Hertz, J.: Maths. (Crelles J.) 1881.
21. Love, A.E.H.: Theory of Elasticity. 4th Ed., Cambridge University Press, Cam-
bridge, 1927, pp. 193-198.
22. Timoshenko, S. and Goodier, J.N.: Theory of Elasticity. 2nd Ed. McGraw-Hill,
New York, 1951, pp. 377-379.
23. Nayak, P.R. and Paul, I.L.: A new theory of rolling contact. MIT Report DSR-
76109-7, 1968.
24. Kalker, J.J.: A simplified theory for non-Hertzian contact. In: J.K. Hedrick (Ed.):
The Dynamics of Vehicles on Roads and Tracks, Proc. 8th IAVSD Symposium,
Cambridge, Mass., August 1983, pp. 295-302. Swets and Zeitlinger Publishers, Lisse,
1984.
25. Carter, F.W.: On the action of a locomotive driving wheel. Proc. Royal Soc. Ser.
A 112 (1926), pp. 151-157.
26. Poritsky, H.: Stresses and deflections of cylindrical bodies in contact with ap-
plic-ation to contact of gears and of locomotive wheels. J. Applied Mech. Trans.
ASME 72 (1950), pp. 191-201.
27. Cain, B.S.: Discussion of reference 25. J. Applied Mech. Trans. ASME 72
(1950), pp. 465-466.
28. Johnson, K.L.: The effect of tangential contact force upon the rolling motion of
an elastic sphere upon a plane. J. Appl. Mech Trans. ASME 80 (1958), pp. 339-346.
29. Johnson, K.L.: The effect of spin upon the rolling motion of an elastic sphere
upon a plane. J. Appl. Mech Trans. ASME 80 (1958), pp. 332-338.
30. Haines, D.J. and Ollerton, E.: Contact stress distributions on elliptical contact
surfaces subjected to radial and tangential forces. Proc. Inst. Mech. Engrs. 177
(1963), pp. 95-114.
31. Vermeulen, P.J. and Johnson, K.L.: Contact of non-spherical elastic bodies
transmitting tangential forces. J. Appl. Mech Trans. ASME 86 (1964), pp. 338-340.
32. de Pater, A.D.: On the reciprocal pressure between two elastic bodies. Proc.
Symposium on Rolling Contact Phenomena, pp. 29-74. Elsevier, Amsterdam, 1962.
33. Kalker, J.J.: The transmission of force and couple between two elastically simi-
lar rolling spheres. Proc. Kon. Ned. Akad. Wet. Amsterdam, B70 (1964), pp. 135-
177.
34. Kalker, J.J.: On the rolling of two elastic bodies in the presence of dry friction.
Doctoral Thesis, Delft University of Technology, 1967.
35. Kalker, J.J.: Simplified theory of rolling contact. Delft Progress Report No 1,
1973, pp. 1-10. See also: Kalker, J.J.: A fast algorithm for the simplified theory of
rolling contact. Vehicle System Dynamics 11 (1982), pp. 1-13.
36. Kalker, J.J.: Three-dimensional elastic Bodies in Rolling Contact. Kluwer Aca-
demic Publishers, Dordrecht, 1990.
37. Kalker, J.J.: Survey of wheel-rail rolling contact theory. Vehicle System Dynam-
ics 8 (1979), pp. 317-358.
38. Hobbs, A.E.W.: A survey of creep. British Railways Technical Note DYN 52,
1967.
39. Illingworth, R.: The mechanism of railway vehicle excitation by track irregulari-
ties. Doctoral Thesis, University of Oxford, 1973.
40. Brickle, B.V.: The steady state forces and moments on a railway wheelset in-
cluding flange contact conditions. Doctoral Thesis, Loughborough University of
Technology, 1973.
41. Pearce, T.G. and Rose, K.A.: Tangential force-creepage relationships in theory
and practice. Proc. Int. Symp. on Contact Mechanics and Wear of Rail/Wheel Sys-
tems, Vancouver, Canada, 1982.
Pearce, T.G. and Rose, K.A.: Measured force-creepage relationships and their
use in vehicle response calculations. In: O. Nordstsrom (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 9th IAVSD Symposium, Linkoping, June 1985.
Swets and Zeitlinger Publishers, Lisse, 1986, pp. 427-440.
42. Knothe, K., Wille, R. and Zastrau, B.W.: Advanced contact mechanics-road and
rail. Vehicle System Dynamics 35 (2001), pp. 361-407.
43. Ertz, M. and Knothe, K.: Die Temperaturentwickling im Rad-Scheine-Kontakt
und ihre Auswirkungen auf Kraftschluss und Materialverhalten. Tagungsband zur
VDEI Fachtagung Bahnbau 2000, Berlin, September 12-15, 2000, pp. 232-237.
44. Shen, Z.Y., Hedrick, J.K. and Elkins, J.A.: A comparison of alternative creep-
force models for rail vehicle dynamics analysis. In: J.K. Hedrick (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 8th IAVSD Symposium, Cambridge,
Mass., August 1983. Swets and Zeitlinger Publishers, Lisse, 1984, pp. 591-605.
45. Polach, O.: A fast wheel-rail forces calculation computer code. In: R. Frohling
(Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc. 16th IAVSD Symposium,
Pretoria, August 1999. Swets and Zeitlinger Publishers, Lisse, 2000, pp. 728-739.
46. Knothe, K. and Le-The Hung.: Determination of the tangential stresses and the
wear for the wheel-rail rolling contact problem. In: O. Nordstsrom (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 9th IAVSD Symposium, Linkoping,
June 1985. Swets and Zeitlinger Publishers, Lisse, 1986, pp. 264-277.
47. Nayak, P.R.: Surface roughness effects in rolling contact. J. Applied Mech.
Trans. ASME 39, Ser. E, No. 2 (1972), pp. 456-460.
48. Greenwood, J.A. and Tripp, J.H.: The elastic contact of rough surfaces. J. Ap-
plied Mech. Trans. ASME 34, Ser. E (1967), pp. 153-167.
49. Bucher, F.: Normal and tangential contact problem of surfaces with measured
roughness. Paper presented at the 5
th
International Conference on Contact Mechan-
ics and Wear of Rail/Wheel Systems, Tokyo, Japan, 25-28
th
July 2000. To be pub-
lished in Wear, 2001/2002.
50. Newland, D.E.: On the time-dependent spin creep of a railway vehicle. J. Mech.
Eng. Sci. 24, No. 2 (1982), pp. 55-64.
51. Knothe, K., and Gross-Thebing, A.: Derivation of frequency dependent creep
coefficients based on an elastic half- space model. Vehicle System Dynamics. 15, No
3 (1986), pp. 133-153.
53. Gilchrist, A.O. and Brickle, B.V.: A Re-examination of the proneness to derail-
ment of a railway wheelset. J. Mech. Eng. Sci 18, No. 3 (1976), pp. 131-141.
54. Matsui, N.: On the derailment quotient Q/P. Railway Technical Research Insti-
tute, Japanese National Railways, 1966.
55. Boedecker, Chr.: Die Wirkungen zwischen Rad und Schiene und ihre Einflsse
auf den Lauf und den Bewegungswiderstand der Fahrzeuge in den Eisenbahnzgen.
Hahnsche Buchhandlung, Hannover, 1887.
56. Schiehlen, W.: Modelling of complex vehicle systems. In J.K.Hedrick, (Ed.):
Cambridge, Mass., August 1983. Swets and Zeitlinger Publishers, Lisse, 1984, pp.
548-563.
57. Kik, W., Knothe, K. and Steinborn, H.: Theory and numerical results of a
general quasi-static curving algorithm. In: A.H. Wickens (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 7th IAVSD Symposium, Cambridge, August
1983. Swets and Zeitlinger Publishers, Lisse, 1984, pp. 414-426.
58. Piotrowski, J.: A theory of wheelset forces for two point contact between wheel
and rail, Vehicle System Dynamics, 11 (1982), pp. 69-87.
59. Elkins, J.A. et al.: The effect of a restraining rail on the curving behaviour of a
transit vehicle. In: J.K. Hedrick (Ed.): The Dynamics of Vehicles on Roads and
Tracks, Proc. 8th IAVSD Symposium, Cambridge, Mass., August 1983. Swets and
Zeitlinger Publishers, Lisse, 1984, pp. 132-147.
60 Yang (ibid), pp. 90-95.
3
Dynamics of the Wheelset
3.1 Introduction
As the wheelset is the basic guidance element of the railway vehicle it is appropriate
to consider the dynamics of the wheelset as a preliminary to a discussion of the be-
haviour of a complete vehicle. In the simple two-axle vehicle of Figure 2.22, atten-
tion is now concentrated on a single wheelset which is connected by means of a
mass-less suspension to the vehicle body which moves forward at constant velocity
with negligible lateral and vertical motion. The vehicle body is assumed to apply a
constant vertical force through the vertical suspension so that the total axle load is
W. As discussed in Chapter 2, the system has three degrees of freedom, with the
corresponding generalised coordinates lateral translation y of the centre of mass,
angle of yaw about a vertical axis and the rate of rotation of the wheelset . The
importance of this rather simple system lies in the fact that for a wide range of prac-
tical values of the parameters, the body of the vehicle does not strongly participate
in the lateral motions of its wheelsets at high speeds, as will be discussed later, so
that the elastically restrained wheelset exhibits behaviour which is typical of com-
plete vehicles. Moreover, it is a system which lends itself to experimentation using
a roller rig [1-4].
In addition, alternatives to wheelset guidance such as the peg-in-slot system, ex-
hibit some contrasting aspects of the behaviour of guided wheeled systems, and will
be briefly discussed.
3.2 The Unrestrained Wheelset
The kinematic oscillation of a wheelset has already been discussed in Chapter 1.
Before considering the stability and response of an elastically restrained wheelset it
is useful to discuss the behaviour of an isolated wheelset in motion at low speed. If
the solution of the full set of equations of motion, equations (2.9.8)-(2.9.10), is con-
sidered for the unrestrained wheelset in which the suspension stiffnesses k
y
and k
are zero, then the behaviour conforms broadly to the simple model considered in
Chapter 1. Figure 3.1 shows the response of a wheelset with the parameters of Table
2.3 and the wheel-rail geometry of Figure 2.2 to a step change in alignment of the
track of 6 mm at low speed. After an initial transient, there is a lightly damped os-
cillation which is similar to the kinematic oscillation in that the angle of yaw is ap-
proximately 90 out of phase with lateral displacement and it is approximately sinu-
soidal, Figure 3.1(a). The longitudinal creep is small, and as shown in Figure 3.1(b)
the lateral creep forces are mainly generated by spin (which is mainly proportional
to the contact slope through equation (2.5.17) and (2.5.18)) and are, therefore, in
phase with lateral displacement. Figure 3.1(c) shows the spin on each wheel and, as
discussed in Section (2.10), even when the wheelset is in the central position equal
and outward lateral creep forces are generated by spin. Figure 3.1(d) shows that the
resultant lateral force is almost completely reacted by the gravitational stiffness
force. Though the amplitude of the oscillation is relatively large at 6 mm, the peak
creep forces, at 6000 N. are far from saturation ( = 0.3, N
0
= 11772 N.). Figure
3.1(e) and 3.1(f) shows that the variation of the normal forces and rotational speed
is small and occurs at double the frequency.
Accordingly, consider the extent to which purely kinematic motions of a wheel-
set are possible. On straight track equations (2.5.13-18) reduce to, if r
0
0
/l<< 1,

1r
= r
r
/V + 1 - l
/V (1)
1l
= r
l
/V + 1 + l
/V (2)
y (m)
t (s)
T
yr
(N)
T
yl
(N)
3l
3r
s
-1
T
yr
+ T
yl
(N)
z
w
W (N)
(a) (b)
(c) (d)
N - W/2
+ V/r
0
Figure 3.1 Oscillation of unrestrained wheelset in response to a step change in alignment y
0
= 6 mm at V = 10 m/s.
(f) (e)
DYNAMICS OF THE WHEELSET 73
2r
= ( y + r
r
) sec
r
/V (3)
2l
= ( y + r
l
) sec
l
/V (4)

3r
= (/V)sin
r
+ (
/V)cos
r
(5)

3l
= (/V)sin
l
+ (

/V)cos
l
(6)
In view of the near equivalence of the lateral creep force due to spin and the gravita-
tional restoring force their effect will be neglected and equations (5) and (6) dis-
carded. Equating (1) and (2) to zero and adding, the variable rotational velocity of
the wheelset is found to be
=
+
2V
r r
r l
( )
(7)
Equating (3) and (4) to zero, adding and substituting from (7)
y
V
= 0
(8)
Equating (1) and (2) to zero, subtracting and substituting from (8)
( )
( )
y
V r r
r r l
r l
r l
+
+
=
2
0
(9)
describing a periodic motion with wavelength independent of speed. The numerical
solution of (7), (8) and (9) for the response to a step change in alignment of the track
of 6 mm at low speed for the wheel-rail geometry of Figure 2.2 yields a time-history
for y, and closely similar to that in Figure 3.1(a) and 3.1(f).
Consider motion at small amplitudes, and replace s/V by D so that D represents
the operation of differentiation with respect to distance travelled along the track.
Then substitution of the equivalent conicity from (2.2.36) into (9) yields
( D
2
+ /r
0
l ) y = 0 (10)
indicating that the motion consists of a single undamped oscillatory constituent of
constant wavelength, with circular frequency
= V ( /r
0
l )
1/2
(11)
the frequency given by Klingels formula discussed in Chapter 1. The mode shape
of this oscillation is, if the amplitude is a,
y = a sint = a ( /r
0
l )
1/2
cost (12)
Exactly these latter results can, of course, be recovered from the simplified linear-
ised equations of motion (2.10.47-48). As the wheelset is unrestrained, the suspen-
sion stiffnesses k
y
and k
are zero, and in terms of the operator D, these equations

reduce to
(mV
2
D
2
+ 2f
22
D)y - 2f
22
= 0 (13)
2f
11
ly/r
0
+ (I
z
V
2
D
2
+ 2f
11
l
2
D) = 0 (14)
It can be seen that at low speeds, since the terms in V
2
can be neglected, equations
(13) and (14) become
2f
22
Dy - 2f
22
= 0 (15)
2f
11
ly/r
0
+ 2f
11
l
2
D = 0 (16)
Elimination of y and then gives (10). This recovers the solution first given by
Carter [5].
As discussed in Chapter 1, closely related to the kinematic oscillation is the be-
haviour of a wheelset on curved track, with zero cant deficiency. For steady motion
on curved track equations (2.5.13-18) reduce to

1r
= r
r
/V + 1 - l/R
0
(17)
1l
= r
l
/V + 1 + l/R
0
(18)
2r
= r
r
sec
r
/V (19)
2l
= r
l
sec
l
/V (20)
3r
= (/V)sin
r
+ (1/R
0
)cos
r
(21)

3l
= (/V)sin
l
+ (1/R
0
)cos
l
(22)
As before, neglecting the effects of spin, rolling round the curve can occur if the
wheelset moves outward so that the longitudinal creep is zero and adopts a radial
position = 0 so that the lateral creep is zero. Adding (17) and (18) the rotational
velocity of the wheelset is given by (7) and subtracting (17) and (18) yields
( )
( )
r r
r r
l
R
r l
r l
+
=
0
(23)
which determines the lateral displacement of the wheelset for zero longitudinal and
lateral creepage. Thus, in response to track curvature the unconstrained wheelset
attempts to roll, without creep, round the curve adopting a radial position which
causes the two rolling circles to form cross-sections of a cone having its apex at the
centre of curvature of the track. On substitution from (2.2.36), (23) reduces to Red-
tenbachers formula discussed in Chapter 1.
Redtenbachers formula can, of course, be recovered from the simplified linear-
ised equations of motion (2.10.47-48). The suspension stiffnesses k
y
and k
are zero
so that for steady motion in a uniform curve with zero cant deficiency these equa-
tions reduce to
- 2f
22
= 0 (24)
2f
11
ly/r
0
= - 2f
11
l/R
0
(25)
giving the expected results.
For the wheel and rail profiles of Figure 2.2, the application of equations (7) and (23)
yield the results shown in Figure 3.2. On a very large radius curve, say R
0
> 1000m.,
there is agreement with Redtenbachers formula as might be expected. It is notewor-
thy that rolling can occur within the flangeway clearance down to relatively small
radius curves. On sharper curves, even if the wheelset is able to take up a perfectly
radial position, the lateral displacement is limited by the available flangeway clear-
ance. However, the worn wheel and rail combination is capable of supporting motion
without lateral and longitudinal creep down to relatively low radii; for example, on a
radius of 150 m the lateral displacement of the wheelset is 7.28 mm (just within the
flangeway clearance). The reduction in rotational velocity is negligible.
It can also be seen from the equations of motion that if a lateral force, such as
0 1000 2000
-10
-8
-6
-4
-2
0
R
0
(m)
y (mm)
0 1000 2000
0
0.05
0.1
0.15
0.2
V/r
0
+ (s
-1
)
R
0
(m)
Figure 3.2 Lateral displacement and change in rotational velocity for wheelset with geometry
of Figure 2.2 on curve with zero lateral and longitudinal creep, equations (7) and (23).
that arising from centrifugal force or cant deficiency, is applied to a wheelset then
the wheelset yaws in order to generate a reacting lateral force.
Thus, as discussed in Chapter 1, it can be seen that for small displacements from
the centre of the track the primary mode of guidance is provided by conicity. It is on
sharp curves and switches and crossings that the flanges become the essential mode
of guidance. As already discussed, as a wheelset is displaced laterally the normal
reaction forces change direction and give a lateral resultant force.
3.3 Root Locus for Small Motions of the Restrained Wheelset
Consider the wheelset elastically restrained by a lateral spring k
y
and a yaw spring
k
,

both connected to ground. For motions of small amplitude, the linear equations
of motion in first order form, equations (2.10.51), are valid and, in the absence of
forcing, the motions can be represented by the trial solution
x(t) = Xe
st
(1)
where X is a column matrix with elements independent of t. Substitution of (1) into
(2.10.51) gives
sX = A X (2)
The condition that equation (2) will have non-zero solutions is
det (A - s) = 0 (3)
so that for a system with N degrees of freedom there will be 2N eigenvalues s
p
which satisfy (3). For asymptotic stability, in which the displacements of the system
eventually decay to zero irrespective of the initial conditions, it is necessary that the
eigenvalues, if real be negative, and if complex should have negative real parts. A
real positive eigenvalue indicates divergence or static instability. In this case an ini-
tial displacement given to the system can result in growing displacements. A com-
plex eigenvalue possessing a positive real part indicates oscillatory instability. In
this case an initial displacement given to the system can result in growing oscilla-
tions. The requirement that the eigenvalues s
p
, if real must be negative, and if com-
plex must have negative real parts is a necessary condition for stability, but in gen-
eral it is not sufficient. For if there are equal imaginary eigenvalues, there may be a
general solution of the form
X = (a
1
+ a
2
t )e
st
(4)
in which case the solution grows indefinitely with time and is unstable. However, if
the system is conservative a
2
is zero, and the system is not unstable [6, p.50]. Exact
repetition of the eigenvalues is unlikely to be encountered in the case of the railway
wheelset and will not be considered here.
Corresponding to each eigenvalue s
p
there will be a solution to (2) so that the
eigenvector X
p
satisfies
s
p
X
p
= A X
p
(5)
and the p solutions can be consolidated into the single equation
U = A U (6)
where U is formed from the eigenvectors X
p
and is a diagonal matrix with ele-
ments consisting of the eigenvalues, U and being ordered consistently. By virtue
of equation (2.10.50) the structure of U has the form in which the pth column is
U
u
u s
p
p
p p
=
(7)
where u
p
is, of course, an eigenvector of the original second order equations,
(2.10.22). For further details of the eigenvalue problem reference may be made to
[6] and [7].
In the case of the wheelset (3) can be expanded to obtain the characteristic poly-
nomial
p
4
s
4
+ p
3
s
3
+ p
2
s
2
+ p
1
s + p
0
= 0 (8)
where
p
4
= mI
p
3
= 2( mf
11
l
2
+ If
22
)/V
p
2
= m K
+ I K
y
+ 4f
11
f
22
l
2
/V
2
+ ( 2f
23
/V - I
y
V/r
0
l )( 2f
23
/V - I
y
V/r
0
l )
p
1
= 2f
22
K
/V + 2f
11
l
2
K
y
/V - 4f
23
( f
22
+ f
11
l/r
0
)/V + 2 I
y
V( f
22
+ f
11
l/r
0
)/r
0
l
p
0
= K
y
K
+ 4f
22
f
11
l/r
0
Of prime interest is the way in which the eigenvalues vary with speed V. At low
speeds there is one complex or oscillatory root A and two large roots B and C which
are real. Figure 3.3(a) shows a typical root locus as V is varied, and Figure 3.3 (b)
shows the variation of the imaginary parts with V. The imaginary part of the oscilla-
tory root, branch A, is closely equal to the frequency of the kinematic oscillation,
and the real part is negative the damping is positive. However, as the speed in-
creases, the damping reaches a maximum and then decreases until it becomes nega-
tive beyond a speed V
B
. The system is unstable at all speeds V > V
B
. Thus, small
motions will not remain small, and the behaviour of the system will then have to be
examined using the nonlinear equations of motion, equations (2.9.8)-(2.9.10). This
is done below. At higher speeds, the two large real roots at B and C coalesce to form
a complex pair shown as D in Figure 3.3, which remains over-critically damped.
It is necessary to form some appreciation of the way in which the various pa-
rameters of the system influence the eigenvalues, and in particular, the stability of
the system. Noting that the branch of the root locus which becomes unstable is that
associated with the kinematic oscillation, a perturbation procedure will be adopted
[8]. It has already been noted that the creep coefficients f
11
and f
22
in the equations
of motion are much larger than the other coefficients, and equations (2.15-16) show
that these same coefficients define the kinematic oscillation. Thus let f now repre-
sent the typical creep coefficient f
11
or f
22
and relying on the fact that f is large con-
sider the characteristic equation if all terms of order less than f
2
are neglected. Then
p
4
= p
3
= p
1
= 0 p
2
= 4f
11
f
22
l
2
/V
2
p
0
= 4f
22
f
11
l/r
0
(9)
and the characteristic equation reduces to
p
2
s
2
+ p
0
= 0 (10)
-10 0 10
0
20
60
80
100
(1/s)
(1/s)
A
0 100 200
0
20
40
60
80
100
V (m/s)
A
D
(a)
(b)
B,C
Figure 3.3 Variation of eigenvalues for elastically restrained wheelset with parameters of
Table 2.3. The root locus does not show branches B and C as their real parts are too large.
In this, and other root locus diagrams, only the positive imaginary part of the complex
pair of roots are shown.
The solution of (10) is, of course,
s = i
1
= iV( /r
0
l)
1/2
(11)
The condition that all terms of order less than f
2
be neglected may now be relaxed
by regarding terms of order f as small perturbations p
n
of the terms p
n
of order f
2
.
From (8) the corresponding perturbation in the eigenvalue s
i
will be given by
s
i
= p
n
s
n
/ n p
n
s
n-1
(12)
provided that there are no repeated roots which is the case for the system under con-
sideration. The initial values are given by equation (9) and the perturbed value of
the polynomial coefficients are
p
3
= 2( mf
11
l
2
+ If
22
)/V
p
1
= 2f
22
K
/V + 2f
11
l
2
K
y
/V - 4f
23
( f
22
+ f
11
l/r
0
)/V + 2 I
y
V( f
22
+ f
11
l/r
0
)/r
0
l
Using (12) two roots of the characteristic equation are found to be

s
1
= i
1
+
1
(13)
where

1
= - V [ f
22
K
+ f
11
l
2
K
y
- V
2
( f
11
l
2
m + f
22
I )/r
0
l
- 2f
23
( f
22
+ f
11
l/r
0
) + I
y
V
2
( f
22
+ f
11
l/r
0
)/r
0
l]/4f
22
f
11
l
2
(14)
The frequency of the oscillation corresponding to s
1
is the same as that of the kine-
matic oscillation. Insertion of typical values of the parameters from Table 2.3 shows
that the terms in f
23
(due to spin creep) and I
y
(due to gyroscopic effects) are of sec-
ondary importance and the damping at low speeds (where the V
2
terms can be ne-
glected) depends on the stiffnesses K
y
and K
. The inertia forces depend on V

2
and
are destabilising, becoming dominant at high speeds. The damping becomes nega-
tive beyond a speed V
A
given by
V
K f K f l f f f l r r l
mf l If I f f l r
A
y
y
2
22 11
2
23 22 11 0 0
11
2
22 22 11 0
2
=
+ +
+ +
{ ( / }
{ ( ) ( / )}

(15)
The explicit terms in f
23
in (15), which arise from the coupling terms in the creep
damping matrix, reduce V
A
slightly (for the example parameters, of less than 1%)
and the terms in I
y
increase V
A
slightly (for the example parameters, by 4%). These
trends can be confirmed by comparison of the eigenvalues obtained numerically. It
follows that a good approximation to V
A
is given by
V
K f K f l r l
mf l If
A
y
2
22 11
2
0
11
2
22
=
+
+
( )
( )
(16)
so that then V
A
is inversely proportional to the square root of the conicity , elastic
restraint is stabilising and inertia destabilising. Moreover, f
11
and f
22
possess similar
values and appear in both numerator and denominator of (16) with the effect that V
A
is insensitive to small variations of creep coefficient.
To the same accuracy as (13), and neglecting f
23
and I
y
, the corresponding eigen-
vector is
u
i
K f K f l V lr m f I f l i lr
y
1
22 11
2 2
0 22 11
2
0
1 2
1 0
4
=
+
+
{ / / ( / )( / / )} / ( / )
/

(17)
Using (13) to form the quadratic factor (s
2
+ 2
1
s +
1
2
) the characteristic equation
can be completely factorised in the form
(s
2
+ 2
1
s +
1
2
) (s +
2
) (s +
3
) = 0 (18)
where approximately
3
= - 2f
22
/mV (19)
4
= - 2f
11
l
2
/IV (20)
These two real eigenvalues correspond to heavily damped subsidences. Substitution
of the eigenvalues into the equations of motion yields the corresponding eigenvec-
tors, which to the same degree of approximation, indicate that
3
is associated with
a motion in which only lateral translation of the wheelset takes place, yawing dis-
placement being negligible,
[ ]
u
T
3
1 0 = (21)
Similarly,
4
is associated with a motion in which yawing of the wheelset is pre-
dominant, lateral translation being negligible,
[ ]
u
T
4
0 1 = (22)
The time constants associated with these large real roots are such that at low, or
even moderate, speeds they are comparable with the time taken for a material point
to move through the contact area as discussed in Section 2.4.5. In this case, transient
creep effects will be significant, and the equations of motion must be modified as
discussed in Section 2.4.5. It will then be found that the large roots are changed sig-
nificantly but the eigenvalues corresponding to the kinematic oscillation are virtu-
ally unchanged. Thus little error is incurred in the discussion of stability if creep
transient effects are neglected.
3.4 Instability and Feedback
The approximate solution derived above is useful in summarising the behaviour of
the system at various speeds. However, it is based on the supposition that the creep
coefficients are large compared with the other parameters. If attention is restricted
to the motion at the onset of instability, the origin of the instability in the feedback
that exists between the lateral displacement and yaw modes of the wheelset be-
comes apparent.
As discussed in Chapter 2 the equations of motion can be represented by the
block diagram shown in Figure 2.19 which indicates that the coupling terms which
arise from creep and conicity in the equations can be interpreted as providing feed-
back. It is therefore appropriate to apply the Nyquist criterion for stability of linear
feedback systems [9].
Referring to Figure 2.19, the open loop transfer function of the system is defined
by breaking the loop at some point such as A, and measuring the output at A as a
function of frequency for unit input at A; the open loop transfer function of the
system, F(i), is therefore
F i
f l r f V VI r l i f f V VI r l i
m f i V K I f l i V K
y y
y
( )
{ / ( / / ) }{ ( / / ) }
( / )( / )
=

+ + + +
2 2 2 2
2 2
11 0 23 0 22 23 0
2
22
2
11
2
Now, for sustained oscillations to occur, Nyquists criterion states that
F(i) = -1 (2)
the open loop gain of the system must be unity with a phase lag of 180. If F(i) <
-1 when the phase lag is 180 then the system will be unstable as the output will
increase with time. If F(i) is plotted in the complex plane, then for stability the
point (-1 + i0) must lie to the left of the F(i) locus as is varied from zero to in-
finity.
Another stability criterion, exactly equivalent to the Nyquist criterion, is that of
Routh [9] which is expressed in terms of the coefficients of the characteristic poly-
nomial. For the quartic (3.8), Rouths criterion for stability is that
T p p p p p p p
3 1 2 3 0 3
2
4 1
2
= > 0 (3)
Figure 3.4 shows the open loop transfer function plotted in the complex plane for
the elastically restrained wheelset, for a speed slightly above the onset of insta-
bility.
At ordinary speeds, terms in f
23
and I
y
may be neglected as discussed above. At the
speed V
B
separating instability from stability, the motion is sinusoidal and on sepa-
rating real and imaginary parts (2) expressions for V
B
and the corresponding fre-
quency
B
can be found. The speed V
B
in terms of speed V
A
found in the previous
Section and given by equation (3.16) is,
V
V
r l
mK I K
f I f ml
B
A
z y
z
2
2
0
22 11
2
2
1
2
=
( )
(4)
and the corresponding frequency
B
is given by
B
y
z
f K f l K
f I f l m
2
22 11
2
22 11
2
=
+
+
(5)
so that
B
=
A
. It is noteworthy that this frequency is independent of the speed at
which instability occurs. In addition,
B
always lies between the two natural fre-
quencies possessed by the uncoupled degrees of freedom. This fact also follows
from a consideration of phase since for sustained oscillations the open loop transfer
function must cross the real negative axis in the complex plane. As the numerator is
positive, this can only occur when the denominator is negative, and this can only
occur when the transfer functions of the uncoupled systems are of opposite signs,
i.e. at a frequency between the two natural frequencies.
-3 -2 -1 0 2 3
-3
-2.5
-2
-1
-0.5
0
0.5
1
Re(F)
Im(F)
= 0
(-1,0)
=
Figure 3.4 Open loop transfer function for elastically restrained wheelset with parameters
of Table 2.3, V = 115 m/s, creep coefficients one third of nominal giving V
B
= 110 m/s.
The motion of the system, in the mode in which the damping is zero at the speed V
B
is
y = a
1
sin(t + ) = a
2
sint (6)
where
a
2
/a
1
= ( /r
0
l )
1/2

(7)
tan = 2f
22
/V( K
y
- m
2
) = -2f
11
l
2
/V( K

- I
2
)
The amplitude ratio is therefore identical to that found in the kinematic mode of a
slowly rolling wheelset. As the critical frequency always lies between the two natu-
ral frequencies possessed by the uncoupled degrees of freedom, if K

/I < K
y
/m
tan is positive, but if K
/I > K
y
/m , tan is negative. Normally, the creep coeffi-
cient f
22
is much larger than m and K
y
, and then will be slightly less or slightly
greater than -90.
Equation (4) shows that V
A
is a close approximation to V
B
provided that either the
creep coefficients are very large in relation to the elastic stiffnesses or that the un-
coupled modal frequencies are sufficiently close. In fact, if the modal frequencies
are equal reductions in a factor on the creep coefficients have no influence on V
B
.
As mentioned above, in ordinary cases, f
11
and f
22
are large and nearly equal in
value so V
B
is relatively insensitive to their value, Figure 3.5. Calculations show
that, consistent with equation (3.15), the gyroscopic terms are then slightly stabilis-
ing and the f
23
terms slightly destabilising.
Figure 3.5 also shows the stabilising effect of large reductions in creep coefficient
in accordance with equation (4). For the parameters of the example wheelset, Table
2.3, reductions of up to 50% are only slightly stabilising, but if the creep coefficients
were significantly smaller then V
B
increases and the instability associated with the
kinematic oscillation disappears. However, another form of instability appears and
this is associated with gyroscopic effects, for if I
y
is zero there is no insta-
Figure 3.5 Effect of large reductions in magnitude of the creep coefficients on V
B
. Also shown
is the solution for the case in which I
y
= 0.
bility at low values of the creep coefficients. Examination of equation (2) for small f
shows that if K
y
/m>K
/I then there is instability above a speed given by

V
f K I m K f f f l r
I f f l r r l
y
y
2
22 23 22 11 0
22 11 0 0
2
=
+ +
+
( / ) ( / )
( / ) /

(8)
and the corresponding frequency is = (
y
/m)
1/2
. Otherwise, if K
y
/m< K
/I then
instability occurs above a speed given by
V
f l K m I K f f f l r
I f f l r r l
y
y
2
11
2
23 22 11 0
22 11 0 0
2
=
+ +
+
( / ) ( / )
( / ) /

(9)
and = (
/I)
1/2
. Using the parameters of Table 2.3 it can be seen that for ordinary
configurations this instability seems of no practical importance.
3.5 Amplitude Dependent Behaviour and Limit Cycles
In deriving the equations of motion, two major nonlinearities were identified as be-
ing those due to the wheel-rail geometry and those due to creep saturation. Unless
the displacements of the wheelset are small the behaviour of the system will depend
on the amplitude of the motion. In the linear solutions discussed so far, the solutions
are, of course, independent of amplitude. Figure 3.6(a) shows the lateral displace-
ment amplitude a as a function of speed for the linear case. There are two possibili-
ties, for the wheelset, for steady motions the steady motion along the track centre-
line A which has been shown above to be asymptotically stable for V < V
B
, and un-
stable for V > V
B
. For V = V
B
, B is a point of bifurcation and steady oscillation is
possible for any amplitude a as shown by the stability boundary D, separating stabil-
ity from instability.
If the wheelset motions cannot be considered small, nonlinearities must be taken
into account. The most direct way of investigating the stability of the system in this
case is to examine the dynamic response to a disturbance using a numerical solution
of the equations of motion, equations (2.9.8-10).
Figure 3.6(b) shows the lateral displacement amplitude a as a function of speed
for the nonlinear case. For small amplitudes the motion is consistent with the linear
model and stability exists for V < V
B
, and instability for V > V
B
. As the amplitude
increases, the motion associated with the kinematic oscillation is approximately si-
nusoidal and the creepages remain small. However, the rolling radius difference
increases and therefore the stability boundary D of Figure 3.6(a) now moves to
lower speeds as the amplitude increases. For larger amplitudes, creep saturation
increasingly occurs with two results: creep saturation is itself stabilising (compare
the stabilising effect of reducing creep coefficients in the linear case discussed
above), and as the creep lateral force due to spin is reduced, the stabilising effect of
the flange force (or gravitational stiffness) comes fully into play. Consequently, the
stability boundary E moves to higher speeds. Separating the boundary E from
boundary D is point C which occurs at speed V
C
. It is this speed that is the critical
speed of the vehicle in engineering terms, for if V < V
C
the vehicle is asymptotically
stable and any disturbance from the motion along the track centreline will decay
exponentially. When V
C
< V < V
B
there are three possibilities for steady motion. The
motion along the track centreline A is asymptotically stable. The steady oscillation
D is unstable, for if the amplitude is slightly greater than that corresponding to the
boundary D, oscillations will grow until the boundary E is reached. If the amplitude
is slightly less than that corresponding to the boundary D, oscillations will decay
until A is reached. The steady oscillation E is stable for if the amplitude is slightly
less than or slightly greater than that corresponding to the stability boundary, oscil-
lations will converge on to the boundary. The stable or sustained oscillation corre-
sponding to boundary E is termed a limit cycle, and it is this limit cycle that is prop-
erly referred to as hunting.
For V > V
B
, the motion along the track centreline is unstable and any disturbance
a
a
V
B
V
B
V
C
V
V
A A
D
B
B
D
C
E
(a)
(b)
Figure 3.6 Limit cycle diagram for (a) linear system and nonlinear system (b).
V (m/s)
50 80 110
= 0.05
= 0.10
= 0.15
Linear geometry
Linear
creep
a
(mm)
6
3
0
~
~
~
~
Figure 3.7 Effect of nonlinear geometry and creep saturation on limit cycles. The points
marked by ~ correspond to nonlinear solutions with = 0.3. Parameters for vehicle of
Table 2.3 and wheel rail geometry of Figure 2.2.
will result in the oscillation growing until the limit cycle E is reached. Similarly, if
the amplitude is greater than that corresponding to boundary E the oscillation will
also converge to the limit cycle E.
Figure 3.7 shows the limit cycle diagram for the example wheelset of Table 2.3
and wheel-rail geometry of Figure 2.2 for two cases. The first case assumes linear
creep and indicates the destabilising effect of nonlinear geometry for larger ampli-
tudes. The second case assumes linear geometry and shows the stabilising influence
of creep saturation which is enhanced for reduced values of the coefficient of fric-
tion. This diagram was derived by numerical integration of the equations of motion,
using a Runge-Kutta fourth-order method with interval adjustment in order to
achieve satisfactory accuracy. Successive trial solutions, for a given vehicle speed,
with a varying initial amplitude make it possible to locate one point on the limit cy-
cle curve. Other points on the limit cycle curve can then be most easily established
by varying the vehicle speed in steps and using as initial conditions the vector of
state variables corresponding to the limit cycle for the previous speed.
Though stability can be investigated by examining the dynamic response to a
disturbance using numerical solution of the equations of motion, methods exist for
the direct solution of the equations of motion to obtain the steady periodic response
to excitation. It is assumed that there is a periodic solution to the equations of mo-
tion at a chosen speed V. In the Describing Function method originally introduced
by Kochenburger [10] and first applied in the present context by Cooperrider et al
[11] it is assumed that the response is sinusoidal so that
x a t = sin (1)
Then the nonlinear equations of motion, equations (2.9.8-10), which can be written
as
x = F(x) (2)
are replaced by quasi-linear equations
x = G (a, )x (3)
where G the describing function matrix is a function of the amplitudes a and fre-
quency of the oscillation. This method is particularly appropriate in cases where
the nonlinearity is of the form of a single output dependent on a single input, for
example, the rolling radius difference as a function of wheelset lateral displacement.
If a typical nonlinear term of F is g(x) then it can be expanded as a Fourier series
g x A t A t ( ) sin sin ........ = + +
1 2
2 (4)
In the simplest form of the method it is assumed that the frequency response of the
rest of the system is limited, and so
g x A t ( ) sin =
1
(5)
A
1
is determined by multiplying (5) by sint and integrating over one cycle so that
A a g x td t
1
0
2
1
( ) ( ) sin ( ) =
(6)
For example, in the case of the rolling radius difference g(y) = r
r
- r
l
is replaced by
y where is the equivalent conicity and = A
1
(a)/a, as discussed in Section 2.3.
Then the feedback model of Section 4 can be extended. The frequency response is
given by equation (4.1) in which, if the small terms in f
23
and I
y
are neglected, is in
the numerator and can be considered as an amplitude dependent and frequency in-
sensitive describing function. It follows that an approximation to the stability
boundary D is given by equation (3.16) where is interpreted as the describing
function.
If the amplitude x
i
of one state is chosen, the equation
G (a, )x = s
p
x (7)
gives a set of eigenvalues s
p
. The eigenvalue with the smallest real part corresponds
to an approximation to the limit cycle, the corresponding eigenvector gives the re-
maining amplitudes and the imaginary part of the eigenvalue the frequency. By it-
eratively choosing a new value for x, evaluating the describing function, and then
repeating the process using the results of the previous step, convergence to the re-
quired solution with zero real part is obtained. The stability of the resulting limit
cycle follows from the properties of the quasi-linearised system matrix for if the
limit cycle is stable, smaller amplitudes will yield eigenvalues with positive real
parts and larger amplitudes should produce negative real parts, and conversely for
unstable limit cycles, [11].
In the case of strong nonlinearities such as creep saturation, the motion must be
approximated by further terms in the Fourier series as the motion is not approxi-
mated by a single sinusoidal component. Various methods for the determination of
limit cycles exist. For example, de Pater [12, 13] and Van Bommel [14] in their
pioneering work on nonlinear oscillations of railway vehicles used the method of
Krylov and Bogoljubov. Gasch et al [15, 16] applied Galerkins method.
It is known that apparently simple dynamical systems with strong nonlinearities
can respond to a disturbance in very complex ways. In fact, for certain ranges of
parameters no periodic solution may exist [17]. Moreover, systems with large non-
linearities may respond to a disturbance in an apparently random way. In this case,
the response is deterministic but is very sensitive to the initial conditions. Such cha-
otic motions have been studied for railway vehicles by True et al [18, 19].
3.6 Energy Balance
Insight is given to the mechanism of the instability if the energy balance at the criti-
cal speed is considered. Following Duncan [20], and returning to the linear solution
for small disturbances, as already discussed in the context of the system eigenval-
ues, there are two possibilities for instability. If there is a real eigenvalue s
p
then
q s q
p
= and q Cq s q Cq
T
p
T

= = 0 . Thus from (2.10.35)
d
dt
T U U U q Bq V
G F
T
( ) / + + + + = 0 (1)
Now q Bq
T
is positive definite for any assemblage of wheelsets so that
d
dt
T U U U
G F
( ) + + + < 0 (2)
As T, U and U
G
are positive definite, if U
F
is positive definite, motion will decay. If
U
F
is not positive definite, T + U + U
G
may increase. There may be conditions for
which there is a zero root, then the equations of motion admit a solution in which
displacements are not zero so that
( ) E E C q
g
+ + = 0 (3)
Such a condition will separate regions of static stability from regions of divergence,
in which the system has a real, positive eigenvalue. A system comprising a wheelset
which is elastically restrained in an unsymmetric way, fore and aft, can exhibit this
behaviour and is considered later.
Consider now the motion corresponding to a complex eigenvalue s = i so
that
q e a t b t
t
= +
( cos sin ) (4)

where, for the wheelset q is given by (4.6-7). Then an important term in (2.10.35) is
q Cq e b Ca
T t T

=
2
(5)
At the speed V
B
, = 0. Integrating (5) over one period , and since the kinetic and
elastic strain energies have the same values at the end of the cycle as they had at the
beginning of the cycle, (2.10.35) reduces to
V a Cb q Bqdt
T T

0
(6)
or
( f
22
a
1
2
+ f
11
a
2
2
)/V + ( f
22
+ f
11
l/r
0
)a
1
a
2
sin = 0 (7)
The first term in (7) represents the dissipation of energy by the creep damping terms
and the second term represents the energy input arising from the stiffness coupling
terms due to creep and conicity. Note that a phase difference is necessary for sus-
tained oscillation to occur.
Though the motion at the critical speed has constant amplitude, energy is be-
ing continuously dissipated by the action of creep. The rate of dissipation of energy
by the creep forces is given by equation (2.10.41) and is, of course, equal to the en-
ergy supplied by the tractive efforts which are given by equations (2.10.43-44).
Eliminating the phase angle and the yaw angle a
2
by means of equations (4.5-6)
the energy dissipated in one cycle is
2Fdt
= (2V
B
/
B
){(K
m - K
y
I)/( 2f
22
I + 2f
11
ml
2
)}
2
( f
22
+ f
11
l/r
0
)a
1
2
(8)
This depends on the separation of the modal frequencies and on the coupling terms
in the creep stiffness matrix; if the modal frequencies are equal, the dissipation per
cycle is zero, for in this case the mode of instability is the same as the kinematic
mode and there is no creepage.
3.7 Dynamic Aspects of Guidance
The steady-state behaviour of a wheelset on a uniform curve at low speeds has been
discussed above so the dynamic aspects of curve entry at speed will now be consid-
ered. The dynamic response of a complex system with nonlinear properties must
usually be obtained by step-by-step numerical integration. However, if the system is
linear, the response may also be obtained by the method of modal superposition.
This involves the calculation of the response of each of the natural modes of the
system and then summing to obtain the complete response. As this method gives
greater insight into the dynamics of the system it represents a useful starting point
for the discussion of the response of a wheelset which is, as discussed above, a non-
conservative active system. The method of modal superposition was originally ap-
plied to conservative systems, but was extended to non-conservative systems by
Halfman [21]. It was applied to the wheelset by Hobbs [22], and the theory has
thoroughly discussed by Fawzy and Bishop [23], Collar and Simpson [6] and New-
land [7, 24].
The method can be summarised as follows; reference may be made to [6] and [7]
for detailed treatments. The linear equations of motion, equations (2.10.51) are
x = A x + Y (1)
Defining normal coordinates of the system by
x = U (2)
where U is the eigenvector matrix of the system defined in equation (3.5). It is as-
sumed that the eigenvalues are not repeated which is the usual case for the wheelset.
The normal coordinates possess an orthogonality property, for premultiplying (3.5)
by U
-1
= W yields
WA U = (3)
Substituting from (2) into equation (1), premultiplying by W and using (3) an un-
coupled set of 2N equations
(s - ) = WY (4)
is obtained. For the simple case where the time dependence of Y(t) is a step func-
tion, substituting into (2) reduces (4) to a simple expansion in a series of the modal
contributions to the response. Taking Laplace Transforms, so that s becomes the
Laplacian operator, yields the modal expansion
x
u w Y
s s
p p
p
p
N
=
1
2
(5)
where u
p
is the column of U corresponding to the pth eigenvalue and w
p
is the cor-
responding row of W. Equation (5) can be used as a general basis for the discussion
of dynamic response.
Consider the dynamic response of a wheelset to entry to a uniform curve of rela-
tively large radius. The wheelset is considered to be elastically restrained to a frame
constrained to follow the centreline and take up a radial position on the track. Ref-
erence to the simplified equations of motion (2.10.47-48) shows that
Y
T
= [ 0 0 0 -2f
11
l
2
H(t)/R
0
I ] (6)
where H(t) is the unit step function. It has been shown above that, for typical values
of the parameters, the wheelset has four eigenvalues, comprising two real roots and
a complex conjugate pair. For a complex conjugate pair of eigenvalues, the corre-
sponding rows of U and columns of W will also be complex conjugate, and in
(5) pairs of terms may be grouped together to form terms of the form
1
2
2
1 1
2
1
2
+ + + s s
(7)
Interpretation of the Laplace Transforms in (5) therefore yields a corresponding
term in the modal expansion of the response of the form
e A t B t
t

1
1 1
( cos sin ) + (8)
In view of the relative size of the eigenvalues, it is apparent that in general the
response will consist of two rapidly damped transients and a damped oscillation at
kinematic frequency. This is illustrated by Figure 3.8 which shows the motion of the
wheelset immediately after entering a large radius uniform curve, as computed from
the full nonlinear equations of motion, equations (2.9.8-10). As might be expected
from the eigenvectors, the subsidence in lateral translation is not excited and the
lateral displacement is small. At t = 0 the creep couple due to the curvature of the
track is reacted by the yaw inertia of the wheelset so that the yaw acceleration is

/ = 2
11
2
0
f l R I (9)
As the yaw velocity builds up, the yaw acceleration decays and increasingly the
0 2 4
-4
-2
0
2
x 10
-3
0 2 4
-6000
-4000
-2000
0
2000
0 2 4
-2000
0
2000
4000
0 2 4
-4000
-2000
0
2000
4000
y
t
T
2r
T
2l
z
w
W
T
2r
+T
2l
T
1r
T
1l
Figure 3.9 Response of elastically restrained wheelset to entry to uniform curve, R
0
=1000 m.
V=10 m/s.
0 0.02 0.04 0.06
-6
-4
-2
0
x 10
-4
0 0.02 0.04 0.06
-2000
-1000
0
1000
2000
0 0.02 0.04 0.06
-1
-0.5
0
0.5
1
x 10
4
0 0.02 0.04 0.06
-100
-50
0
50
100
y
t
T
2l
T
2r
z
w
W
T
2r
+T
2l
T
1l
T
1l
Figure 3.8 Response of elastically restrained wheelset to entry to uniform curve for t < 0.06
s. R
0
=1000 m. V=10 m/s.
Figure 3.10 Response of elastically restrained wheelset to entry into curve, R
0
= 1000 m and
transition length L = 20 m.
0 0.1
-8
-6
-4
-2
0
x 10
-3
0 0.1
-10000
-5000
0
5000
0 0.1
-2
-1
0
1
2
x 10
4
0 0.1
-1
0
1
2
x 10
4
y (mm)
T
2r
T
2l
t (s)
T
1r
T
1l
z
y
W
T
2r
+ T
2l
Figure 3.11 Response to curve entry with no transition for t < 0.13 s. R
0
= 225 m
0 1 2 3
-10
-5
0
5
x 10
-3
0 1 2 3
-10000
-5000
0
5000
0 1 2 3
-2000
-1000
0
1000
2000
0 1 2 3
-1
-0.5
0
0.5
1
x 10
4
y (mm)
T
2r
T
2l
t (s)
T
1r
T
1l
z
y
W
T
2r
+ T
2l
Figure 3.12 Response to entry into curve, R
0
= 250 m and transition length L = 20 m.
0 1 2 3
-2
-1
0
1
x 10
-3
0 1 2 3
-4000
-2000
0
2000
0 1 2 3
-400
-200
0
200
400
0 1 2 3
-2000
-1000
0
1000
2000
y (mm)
t (s)
T
2r
T
2l
z
y
W
T
2r
+ T
2l
T
1r
T
1l
couple due to conicity is balanced by the longitudinal creep forces generated by the
yaw velocity. The yaw velocity is, approximately
( )
=
V
R
e
t
0
1
4
(10)
so that, after the initial transient, the contribution of this mode to the yaw velocity
becomes constant and is equal to -V/R
0
. The longitudinal creep at t = 0 is equal to
l/R
0
and so for radii of curvature less than about 1000 m there will be slipping.
Though transient effects modify the behaviour of the system for small t, this is not
of prime interest as far guidance is concerned. The response in the kinematic mode
is given approximately, if
1
<<
1
, for t small
y
lr
R
t V t R = = ( )( cos ) /
0
0
1
2 2
0
1 2
(11)
= = = ( )( sin )

/
lr
R V
t
y
V
Vt R
0
0
1 1 0
(12)
( )( cos ) /
= =
lr
R V
t V R
0
0
1
2
1 0
(13)
and these results are consistent with the results in Figure 3.8. The response for lar-
ger values of t, also computed from the full nonlinear equations of motion, are
shown in Figure 3.9. It can be seen that after the initial very fast transient, the re-
sponse closely matches the above approximation and the system behaves as a sim-
ple damped spring-mass system, with stiffness dependent on the conicity and speed,
and damping dependent on the suspension stiffnesses, inertia and creep coefficients.
Also shown in Figure 3.9 are the creep forces. The lateral creep forces are mainly
generated by spin and are largely cancelled out by the gravitational stiffness force.
The longitudinal creep forces decay rapidly as the wheelset is able to move out to
the equilibrium rolling line.
These results indicate one of the deficiencies of the wheelset as a guidance
mechanism in that the response time is solely a function of the conicity which is
largely out of the control of the designer and which lies within certain practical lim-
its.
Figure 3.10 shows comparable results for the wheelset which enters a similar
curve as that discussed above but in this case there is a transition of length 20 m
over which the curvature is increased linearly from zero to the steady state value.
Thus, within the transition, the integration of (2.2.3) yields for the azimuth of the
centreline of the track
= s
2
/2R
0
L (14)
and then the integration of (2.2.1-2) gives for the coordinates of the centreline of the
track
X s s R L = +
5
0
2 2
40 / ..... Y s R L s R L = +
3
0
7
0
3 3
6 336 / / . . . . . . . (15)
Consequently, this form of transition is commonly referred to as a cubic parabola. It
can be seen that the excitation of the kinematic oscillation, and the level of the
forces acting in the transient, is much reduced.
Results for a sharper curve, R
0
= 225 m with no transition curve are shown in
Figure 3.11. For small t transients similar to those discussed above occur. The
wheelset moves in a closely straight line until, after a distance (2R
0
f
)
1/2
into the
curve, flange contact takes place. Note that this distance is much less than the kine-
matic wavelength. When flange contact takes place accurate estimates of the lateral
forces require allowance for track flexibility [25]. The provision of a transition curve
allows the wheelset to respond in a more satisfactory way. Figure 3.12 shows the
dynamic response of the wheelset to a 250 m radius curve with a transition length of
20 m and shows that flange contact is avoided.
For small amplitudes of motion and over the frequency range appropriate to the
response to track irregularities, the frequency response of a wheelset may be ob-
tained from (5). Using the approximations of equations (3.13), (3.14) and (3.17) and
neglecting the contribution of the modes associated with the large real roots the fre-
quency response is found to be
H
y i
y i
i
t
( )
( )
( )

= =
+ +
1
2
2
1 1
2
2
(16)
as derived by Hobbs [22] . Illingworth [26] carried out model roller rig experiments
verifying that the theoretical description of excitation of wheelset motions by track
irregularities was substantially correct. In the case where irregularities are distrib-
uted continuously along the track the approach offered by stochastic process theory
is appropriate. Random process theory was first applied by Hobbs in 1964 [22] to
the lateral motions of a wheelset. Subsequently, extensive measurements of the
power spectra of irregularities of track have been made and have been used exten-
sively in the assessment of vehicle response.
3.8 Alternative Methods of Guidance
The early mining railways employed "peg in the slot" as a guidance system and
whilst this was superseded by the coned wheelset early in the 19th Century there is a
continuous record of invention of alternative guidance systems. In the past many
alternative systems were considered because the running gear of railway vehicles
often performed badly; rapid wear of wheels and rails, hunting and derailments oc-
curred as a result of the lack of understanding of how a wheelset should be incorpo-
rated in a vehicle. As discussed above, the smallest radius for flange free curving of
a wheelset with conventional profiles well adapted to conventional rail profiles is
about 150 m, and the longitudinal movement of a coned wheelset in taking up a ra-
dial position on sharply curved track can represent a design problem. As a target
curve radius for some innovative urban transit systems can be as low as 10 m there
is scope for alternative methods of guidance of railway vehicles.
For configurations in which the wheels are fixed on a common axle, two condi-
tions, positive centering action for small amplitudes within the flangeway clearance
and positive retention or static stability for large displacements, can be made the
basis for selection of wheel rail geometry. The possibilities are shown in Figure
3.13. Configurations (c) and (e) have positive conicity and rate of change of contact
slope with lateral displacement and hence satisfy the basic requirements. It is inter-
esting that one conforms to the conventional wheel rail combination whilst the other
resembles a plateway, a system much used in the later 18th century.
For some applications it is possible to separate out the functions of support and
guidance completely, and provide separate sets of wheels, exerting only normal
forces. There are four basic possibilities as shown in Figure 3.14 and some of these
have been used both with steel wheels on steel rails and with pneumatic tyres on
beams. However, the multiplicity of wheels and the ponderous nature of the switch-
ing arrangements make this approach relatively unattractive.
Independently rotating wheels have been frequently proposed as they eliminate
the classical hunting problem. Some of the possibilities have been surveyed by
Frederich [27]. The essential difference between a conventional wheelset and inde-
pendent wheels lies in the ability of the two wheels to rotate at different speeds and
therefore, strictly, there is an additional degree of freedom. However, the effect in
(a) (b) (c)
(d)
(e) (f)
< 0, > 0
> 0, < 0
> 0, > 0
< 0, < 0 > 0, > 0 > 0, < 0
Figure 3.13 Basic possibilities for wheel-rail geometries. (equivalent contact slope) and
(equivalent conicity) are the linearised geometric parameters defined in equations
(2.3.36) and (2.3.37).
terms of the linear wheelset model is to render f
11
= 0. The kinematic oscillation of a
conventional wheelset is therefore eliminated (the feedback loop of Figure 2.19 is
broken) but a measure of guidance is then provided by the lateral component of the
gravitational stiffness (reduced by the lateral force due to spin creep) which becomes
the flange force when the flangeway clearance is taken up, but this leads to slow self
centering action. Extensive experimental experience has shown that indeed the kine-
matic oscillation is absent but that one or other of the wheels runs in continuous flange
contact [28]. The detailed analysis of a system embodying independently rotating
wheels requires that the calculation of creepages take into account the differing wheel
rotational speeds,
r
and
l
, and this particularly affects the longitudinal creepage. If
there is only one point of contact between each wheel and the rail then, neglecting the
rotational acceleration of the wheels, each wheel will be rolling with zero longitudinal
force and zero longitudinal creepage in accordance with the simple model. However,
if a wheel contacts the rail at two distinct points, as is common with freely rotating
wheels, the analysis is more complex. Good agreement between calculation and ex-
periment is demonstrated in [29, 30, 31].
An attempt to increase the effect of the lateral resultant gravitational force but re-
duce the amount of spin is to incline substantially from the horizontal the axis of rota-
tion of the wheels, Figure 3.15(a), as put forward by Wiesinger [32]. A generic wheel-
set model including the effect of modest amounts of camber has been studied theoreti-
cally and experimentally by Jaschinski and Netter [33].
One approach in applying freely rotating wheels is to provide a pivot ahead of the
wheels. The wheelset mounted on a leading or trailing arm, a pony axle, was used on
many steam locomotives and the lack of fore and aft symmetry has a significant effect
on the dynamics of the system. In concept this is closely linked to the peg in the slot
type of system to be discussed later, in that stability depends on the lack of fore-
and-aft symmetry. A similar effect is achieved by a linkage, or the equivalent, to
the preceding vehicle in a train. An example of this is the original version of the
Figure 3.14 Possible configurations of support and guidance wheels depending on normal
forces only.
S
A
C
(a)
(b)
(c)
(d)
Figure 3.15 Some further alternative methods of guidance using independently rotating
wheels (a) cambered wheels (b) Talgo train with vehicles guided by vehicle in front (c)
Koyanagis system with guidance from a central rail ahead of the wheels (d) guidance
from a trace using a sensor S, controller C and steering actuator A.
d
2f
2fd/r
0
2f
2fd/r
0
Figure 3.16 Conventional wheelset mounted on a leading or trailing arm.

Talgo train [34], Figure 3.15(b). The same effect is achieved by guiding the pivot
point along a central rail as in the concept, Figure 3.15(c), analysed by Koyanagi
[35]. More complex but related arrangements used in light rail applications have
been considered in [36, 37]. A logical extension of these concepts is the form of
guidance, using independently rotating wheels, shown in Figure 3.15(d). Not only
does the dynamics of this system have some interesting features but a modern im-
plementation using active controls shows considerable promise, [38]. Guidance is
achieved by measuring the lateral displacement from a reference trace on the track,
and sending a steering command to an actuator through a controller.
One aspect of the dynamical behaviour of these unsymmetric fore-and-aft con-
figurations can be illustrated by considering the conventional wheelset mounted on
a leading or trailing arm, Figure 3.16. The elastic stiffness matrix becomes
E
k dk
dk d k
y y
y y
=
2
(1)
In the absence of suspension damping, assumed for simplicity, the equations of mo-
tion for the wheelset are otherwise unaltered. A stability diagram for this system is
shown in Figure 3.17 in which the stability boundaries are plotted as a function of
arm length d and speed. Mounting the spring either forward or aft is strongly stabi-
lising as far as the dynamic stability of the system is concerned for the dominant
effect is that of increasing the yaw stiffness. As the source of instability is the pres-
ence of assymmetric coupling terms in the equations of motion, between lateral
translation and yaw, it might be thought that an obvious way of stabilising the sys-
tem is to negate one or other of these terms by adding lateral stiffness offset either
in front of or behind the wheelset. This is similar to the technique known as mass bal-
ancing much used in Aeroelasticity. Unfortunately the creep stiffness coupling terms
are large in relation to the elastic stiffness couplings and impracticably large arm lengths
-3 -2 0 1 2 3
50
100
d (m)
V
(m/s)
0
D
O
Figure 3.17 Stability diagram for conventional wheelset mounted on trailing or leading
arm. Parameters from Table 2.1 except k
y
= 2 MN/m and k
=0.
or stiffnesses would be necessary. However, if the spring is moved too far forward static
instability or divergence occurs. The condition for static stability is that in the character-
istic equation p
0
> 0 [9] or
( )
/
/
=
d
f
k
l r
l r
y
2
1
0
0
(2)
The system can only be statically unstable if either l/r
0
> 1 and d > 0 or l/r
0
< 1
and d < 0. Thus, a leading arm with low conicity is prone to divergence. This form
of instability has been experienced on locomotives in the past, and is a potential
problem for articulated vehicles as discussed in Chapter 8.
As mentioned above, the case of freely rotating wheels mounted on a leading or
trailing arm may be covered by putting f
11
= 0 in the linear equations of motion. In
this case the criterion for static stability becomes simply d > 0 so that the wheels
must be behind the pivot point for static stability. In fact, more complex arrange-
ments make it possible to provide stability in both directions and various ways of
exploiting a lack of fore-and-aft symmetry are discussed in Chapter 9.
A simple model of a four wheeled vehicle with independently rotating wheels and
guided by peg-in-slot and linkage reveals the basic aspects of the dynamics of such
vehicles. These considerations equally apply to the system of Figure 3.15(d). Refer-
ring to Figure 3.18, it can be seen that guidance is obtained by steering the wheels (the
rear wheels in the opposite sense to those of the front) of the simple vehicle shown by
a linkage connected to a reference line (such as a peg in a slot) distance L
ahead of the vehicle centre of mass. Assuming a general form of linkage, then for
2a
y
t
L
Figure 3.18 Simple model of vehicle with independently rotating wheels guided by peg-in-
slot and linkage.
small displacements, the steering law can be written as
= + G y L y
t
( ) (3)
where G is the gain dependent on the form and dimensions of the linkage. The linear
equations of motion are easily formulated and take the form, writing f for f
22
,
m y + 4f( y /V - ) = 0 (4)
I fa V fa
z

/ + = 4 4 0
2
(5)
Substituting from (3)
(ms
2
+ 4fs/V)y - 4f = 0 (6)
4fGay + ( I
z
s
2
+ 4fa
2
s/V + 4fGaL) = 4fGy
t
(7)
The similarity of these equations to the wheelset equations of motion should be
noted. The analysis of Section 3 can now be repeated. For the present system, the
coefficients of the characteristic polynomial are
p
4
= mI p
3
= 4f(ma
2
+ I)/V p
2
= 4fGaLm + 8f
2
a
2
IV
2
(8)
p
1
= 16f
2
GaL/V p
0
= 16f
2
Ga
A typical locus of the eigenvalues as speed is varied is shown in Figure 3.19 for
various values of the gain G. At low speeds, as in the case of the wheelset, the char-
acteristic equation can be approximately factorised in the form
(s
2
+ 2
1
s +
1
2
) (s +
2
) (s +
3
) = 0 (9)
where
2
= - 4f/mV (10)
3
= - 4fa
2
/IV (11)
1
2 2
2
4 = +
GV
a
L ma I V fa { ( ) / }
(12)
1
= V G a / (13)
For small values of G, the quadratic factor represents a steering oscillation labelled
A in Figure 3.19. This has wavelength a G / independent of speed, in which there
is zero creep. For example, if G = 1 rad/m, L = 2 m, a = 1.5 m and V = 10 rn/s then
the wavelength of the steering oscillation is 7.69m corresponding to a frequency of
1.3 Hz and the decay rate is 0.81 of critical. The decay rate of this oscillation de-
pends on L being positive; in this simple vehicle, as is intuitively obvious, the peg
must be placed forward of the centre of the vehicle. In reverse motion, V negative,
such a vehicle would be unstable. The two roots given by (10) and (11), labelled B
in Figure 3.19 correspond to subsidences in lateral translation and yaw respectively,
exactly analogous to the wheelset. At higher speeds these two real roots coalesce to
form a well damped oscillation.
For G > 4a/L
2
and low speeds the steering oscillation is replaced by two sub-
sidences. Equation (12) shows that as speed is increased, the inertia forces have the
effect of reducing the decay rate of the steering oscillation. The damping of the
steering oscillation vanishes at an approximate speed V
B
when one root of the char-
acteristic polynomial is s = i
A
. From equation (12)
V
A
2
= 4fa
2
L/(ma
2
+ I) (14)
but this approximate result is only valid for small G. A more exact speed V
B
is easily
obtained by substituting s = i in the characteristic polynomial and equating real
and imaginary parts, to give
V fa L ma I GL m a ma I
B
2 2 2 2 2 3 2 2
4 1 = + + / ( ){ / ( ) } (15)
It can be seen that increasing the gain is stabilising and the vehicle will be stable for
all values of V if
G > (ma
2
+ I)
2
/m
2
L
2
a
3
(16)
-100 -50 0 50
0
50
100
0 100 200
0
50
100
V (m/s)
A
A
B
B
Fig 3.19 Variation of eigenvalues with speed for peg-in-slot system: a = 1.25 m; f = 10 MN.;
G = 1; L = 2 m; m = 1250 kg.; I
z
= 700 kgm
2
.
Also the vehicle is stable for all values of G if
V < 2fa
2
L(ma
2
+ I) (17)
The similarity of the behaviour of the system to that of a railway wheelset is carried
over to its curving performance. In steady motion on a uniform curve the equations
of motion (4) and (5) reduce to
= 4
2
0
f mV R / (18)
4 4
2
0
fa R fa / = (19)
so that the vehicle yaws through an angle
= mV fR
2
0
4 / (20)
to react the centrifugal force, and the required steer angle is
= a R /
0
(21)
In order to generate these steer angles the vehicle must move laterally through a
distance y given by substitution into the steering law, equation (3) from (20) and
(21) given by
y L R a gR LmV fR =
2
0 0
2
0
2 2 / / / (23)
It can be seen that at low speeds if G > a/2L
2
the vehicle is displaced towards the
centre of the curve and if G < a/2L
2
it is displaced outwards. As speed increases fur-
ther outward movement takes place in order to allow the vehicle to yaw whilst gen-
erating the correct steer angle.
Another approach to the improvement of the wheelset as a guidance element is
intermediate between a conventional wheelset and one with independent wheels.
This provides a torque connection such as a damper or clutch between the two
wheels of a wheelset. Such a scheme was proposed by Benington [39] and a similar
scheme has been analysed by Choromanski and Kisilowski [40]. A 'creep controlled'
wheelset has been developed [41] which uses a controlled magnetic coupling in the
centre of the axle of the wheelset. Various control laws have been used for the con-
trol of the coupling, including feedback of creep measurements. Essentially the
wheels have a good torque connection at low frequencies so that curving ability is
maintained but at high frequencies, typical of wheelset kinematic frequencies, the
wheels are more or less uncoupled so that instability does not arise. The stability of
wheelsets with independently rotating wheels using a control engineering approach
has been considered by Goodall and Li [42].
References
1. Matsudaira, T.: Shimmy of axles with pair of wheels (in Japanese). J. of Railway
Engineering Research, (1952), pp. 16-26.
2. Matsudaira, T.: Paper awarded prize in the competition sponsored by Office of
Research and Experiment (ORE) of the International Unions of Railways (UIC).
ORE-Report RP2/SVA-C9, ORE, Utrecht, 1960.
3. Wickens, A.H.: The dynamic stability of railway vehicle wheelsets and bogies
having profiled wheels. Int. J. Solids Structures, 1 (1965), pp. 319-341.
4. Wickens, A.H.: The dynamics of railway vehicles on straight track: fundamental
considerations of lateral stability. Proc. I. Mech. Engrs. l80, part 3F (1965), pp. l-16.
5. Carter, F.W.: The electric locomotive. Proc. Inst. Civ. Engs. 221 (1916), pp. 221-
252.
6. Collar, A.R. and Simpson, A.: Matrices and engineering Dynamics. Ellis Hor-
wood, Chichester, 1987.
7. Newland, D.E.: Mechanical vibration analysis and computation. Longman, Har-
low, 1989.
8. Culick, F.E.C.: Effects of small changes in the elements A
ij
on the zeros of |A
ij
| J.
Royal Aeron. Soc. 62 (1958), pp. 898-901.
9. Porter, B.: Stability criteria for linear dynamical systems. Oliver Boyd, Edin-
burgh, 1967.
10. Kochenburger, R.J.: Frequency-response methods for analysis of a relay servo-
mechanism. Trans. AIEEE, 69 (1950), pp. 270-284.
11. Cooperrider, N.K., Hedrick, J.K., Law, E.H. and Malstrom, C.W.: The applica-
tion of quasilinearization techniques to the prediction of nonlinear railway vehicle
response. In: H.B. Pacejka (Ed.): The Dynamics of Vehicles on Roads and on
Tracks, Proceedings of the IUTAM Symposium, Delft, The Netherlands, August
1975. Swets & Zeitlinger, Lisse, 1976, pp. 314-325.
12. de Pater, A.D.: Etude du mouvement de lacet d'un vehicule de chemin de fer.
Appl. Sci. Res. A, 6 (1956), pp. 263-316.
13. de Pater, A.D.: The approximate determination of the hunting movement of a
railway vehicle by aid of the method of Krylov and Bogoljubov. Appl. Sci. Res. 10
(1961), pp. 205-228.
14. van Bommel, P.: Application de la theorie des vibrations nonlineaires sur le
problem du mouvement de lacet d'un vehicule de chemin de fer. Doctoral disserta-
tion, Technische Hogeschool Delft, 1964.
15. Moelle, D. and Gasch, R.: Nonlinear bogie hunting. In: A.H. Wickens (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 7th IAVSD Symposium, Cambridge,
September 1981. Swets and Zeitlinger Publishers, Lisse, 1982, pp. 455-467.
16. Gasch, R., Moelle, D. and Knothe, K.: The effects of non-linearities on the limit-
cycles of railway vehicles. In: J.K. Hedrick (Ed.): The Dynamics of Vehicles on
Roads and Tracks, Proc. 8th IAVSD Symposium, Cambridge, Mass., August 1983.
17. Thompson, J.M.T. and Stewart, H.B.: Nonlinear dynamics and chaos. John
Wiley, Chichester, 1986.
18. True, H.: Dynamics of a rolling wheelset. App. Mech. Reviews, 46 (1993), pp.
438-444.
19. True, H.: Railway vehicle chaos and asymmetric hunting. In: G. Sauvage (Ed.):
Linkoping, Sweden, August 1991. Swets and Zeitlinger Publishers, Lisse, 1992, pp.
625-637.
20. Duncan, W.J.: The flutter of systems with many freedoms. Aeronautical Quar-
terly 1 (1949), p. 59.
21. Halfman, R.L.: Dynamics. Addison-Wesley, Cambridge, Mass., 1959.
22. Hobbs, A.E.W.: The response of a restrained wheelset to variations in the align-
ment of an ideally straight track. British Railways Research Department Report
E542, 1964.
23. Fawzy, I. and Bishop, R.E.D.: On the dynamics of linear non-conservative
systems. Proc. R. Soc. Lond. A 352 (1976), pp. 25-40.
24. Newland, D.E. On the modal analysis of non-conservative systems. J. Sound and
Vibration 112 (1987), pp. 69-96.
25. Clark, R.A., Eickhoff, B.M. and Hunt, G.A.: Prediction of the dynamic response
of vehicles to lateral track irregularities. British Rail Research Tech. Memo. TM DA
42, 1983.
26. Illingworth, R.: Railway wheelset lateral excitation by track irregularities. In: A.
Slibar and H. Springer (Eds.): The Dynamics of Vehicles on Roads and Tracks, Proc.
5th IAVSD -2nd IUTAM Symposium, Vienna, September 1977. Swets and Zeitlinger
Publishers, Lisse, 1978, pp. 450-458.
27. Frederich, F.: Possibilities as yet unknown regarding the wheel/rail tracking
mechanism. Rail International 16, (1985), pp. 33-40.
28. Becker, P.: On the use of individual free rolling wheels on railway vehicles. Eis-
enbahn Technische Rundschau No. 11 (1970).
29. Eickhoff, B.M. Harvey, R.F.: (1989) Theoretical and experimental evaluation of
independently rotating wheels for railway vehicles. In: R. Anderson (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 11th IAVSD Symposium, Kingston,
Ont., August 1989. Swets and Zeitlinger Publishers, Lisse, 1989, pp. 190-202.
30. Eickhoff, B.M.: The application of independently rotating wheels to railway
vehicles. Proc. I. Mech. E. 205, Part3F (1991), pp. 43-54.
31. Elkins, J.A.: The performance of three-piece trucks equipped with independently
rotating wheels. In: R. Anderson (Ed.): The Dynamics of Vehicles on Roads and
Tracks, Proc. 11th IAVSD Symposium, Kingston, Ont., August 1989. Swets and
32. de Pater, A.D.: Analytisch en synthetisch ontwerpen. Technische Hogeschool
Delft, 1985, pp. 37-38.
33. Jaschinski, A. and Netter, H.: Non-linear dynamical investigations by using
simplified wheelset models. In: G. Sauvage (Ed.): The Dynamics of Vehicles on
Roads and Tracks, Proc. 12th IAVSD Symposium, Linkoping, Sweden, August 1991,
pp. 284-298. Swets and Zeitlinger Publishers, Lisse, 1992.
34. Oriol de, L.M.: El Talgo Pendular. Revista Asociacion de Investigacion del
transporte. No. 53 (1983), pp. 1-76.
35. Koyanagi, S.: A new guide system for a wheel-rail vehicle. In: A. Slibar and H.
Springer (Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc. 5th IAVSD -
2nd IUTAM Symposium, Vienna, September 1977. Swets and Zeitlinger Publishers,
Lisse, 1978, pp. 407-415.
36. Bonivert, L., Maes, P. and Samin, J.C.: Lateral dynamics of a guided light transit
vehicle. In: R. Anderson (Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc.
11th IAVSD Symposium, Kingston, Ont., August 1989. Swets and Zeitlinger
37. Chatelle, Ph., Duponcheel, J. and Samin, J.C.: Investigation on nonconventional
railway systems through a generalised multi-body approach. In: J.K. Hedrick (Ed.):
The Dynamics of Vehicles on Roads and Tracks, Proc. 8th IAVSD Symposium, Cam-
bridge, Mass., August 1983. Swets and Zeitlinger Publishers, Lisse, 1984, pp. 43-57.
38. Wickens, A.H.: Dynamics of actively guided vehicles. Vehicle System Dynamics
20 (1991), pp. 219-242.
Goodall, R.: Active railway suspensions: implementation status and technologi-
cal trends. Vehicle System Dynamics 6 (1977), pp. 87-117.
39. Bennington, C.K.: The railway wheelset and suspension unit as a closed loop
guidance control system, a method for performance improvement. Journal of Mech.
Eng. Sci. 10 (1966), pp. 91-100.
40. Choromanski, W. and Kisilowski, J.: Dynamics of railway trucks with wheelsets
with independently rotating wheels and controlled slip. In: R. Anderson (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 11th IAVSD Symposium, Kingston,
Ont., August 1989. Swets and Zeitlinger Publishers, Lisse, 1989, pp. 108-125.
41. Geuenich, W., Guenther, C. and Leo, R.: Dynamics of fiber composite bogies
with creep-controlled wheelsets. In: J.K. Hedrick (Ed.): The Dynamics of Vehicles on
Roads and Tracks, Proc. 8th IAVSD Symposium, Cambridge, Mass., August 1983.
42 Goodall, R.M. and Li,, H.: Solid axle and independently rotating wheelsets-a
control engineering assessment of stability. Vehicle System Dynamics 33 (2000), pp.
57-67.
4
Guidance of the Two-Axle
Vehicle
4.1 Introduction
As discussed in Chapter 1, guidance is the ability of a vehicle to follow the geomet-
ric layout of the track, and is mainly determined by behaviour in curves. It has been
seen in Chapter 3 that a wheelset will only be able to move outwards to the rolling
line if either the radius of curvature or the flangeway clearance is sufficiently large,
otherwise longitudinal creep forces are generated. In addition, the forces of cant
deficiency must be reacted by lateral forces acting between wheel and rail so that
lateral creep forces are generated. Moreover, for stability the wheelsets in a vehicle
must be constrained by the suspension in some way. The forces imposed by the
suspension must be reacted by further creep forces. If the vehicle has any dimen-
sional misalignments then these will also be accommodated by further straining of
the suspension and additional creep forces. It follows that the way in which the
wheelsets are connected together, through the car body or otherwise, is fundamental
to the mechanism of guidance. Consequently, the starting point for this Chapter is a
general consideration of the way in which two wheelsets may be elastically con-
nected. This is followed by a discussion of the conflict between stability and steer-
ing. Finally, the response of the complete vehicle to track curvature, cant defi-
ciency, and misalignments is considered, the attitude of the vehicle and the forces
acting being obtained by solving the equations of equilibrium.
A two-axle vehicle with a single stage suspension (or a bogie or truck unre-
strained by the car body) is considered. The symmetric two-axle vehicle is the sim-
plest form of vehicle which embodies in a practical way wheelsets and their inter-
connections. The two-axle vehicle is important in its own right, for in addition to its
widespread use in the past as both a passenger and freight vehicle, it is increasingly
seen as the light weight vehicle of the future. Moreover, its study is an essential step
to the understanding of the bogie vehicle and more complex configurations.
The pioneering work of Mackenzie [1] was mentioned in Chapter 1. His seminal
paper (which was subsequently translated and published in both France and Ger-
many) was suggested by an unintentional experiment in which the springs of the driv-
ing wheels of a six wheeled engine were tightened to increase the available adhesion.
The leading wheel mounted the rail when the locomotive approached a curve.
Mackenzie gave a qualitative but essentially correct description of the forces gener-
ated in curving based on sliding friction, and his calculations showed that the outer
wheel flange exerts against the rail a force sufficient to overcome the friction of the
wheel treads. Thus the locomotives excessive flange force was not the result of
centrifugal force, as was previously thought, but was generated by the friction of the
other wheels. He also made the comment that the vehicle seems to travel in the
direction which causes the smallest amount of sliding which foresaw a later ana-
lytical technique.
The early development of theories of curving, based on Mackenzies ideas, were
dominated, understandably, by the paramount need to avoid excessive loads on both
vehicle and track caused by steam locomotives with long rigid wheelbases travers-
ing sharp curves. Hence, in these theories the conicity of the wheelsets is ignored,
the wheelsets are considered to be rigidly held parallel in the frame and the wheels
are assumed to be in the sliding regime. The corresponding forces are then balanced
by a resultant flange force, or forces, acting parallel to, and directly below the axis
of a wheelset. If equilibrium is maintained by the outer leading flange contacting the
rail, this was referred to as free curving. If two flanges are in contact, this was re-
ferred to as constrained curving. These ideas were developed, refined and applied
over the remaining period of steam locomotive design [2], particularly by Heumann
[3], who established his minimum principle that equilibrium is attained when the
guiding force is at a minimum. Porter [4] gave the final and most complete devel-
opment of this theory. Complete solutions of the equations of motion show agree-
ment with these classical methods in the extreme conditions postulated.
In modern practice, as a wheelset incorporated in a conventional bogie is con-
strained by the longitudinal and lateral springs connecting it to the rest of the vehi-
cle, it will not be able to take up the radial attitude of perfect steering. A wheelset
will then balance a yaw couple applied to it by the suspension by moving further in
a radial direction so as to generate equal and opposite longitudinal creep forces, and
it will balance a lateral force by yawing more. On a curve of sufficiently large radius
the creepages will be small, the flangeway clearance will not be taken up, and lin-
earised conicity may be assumed. The linear theory was first developed by Boocock
[5] (who also gives experimental validation with full-scale vehicles) and Newland
[6] in 1968, and though restricted in practical application, forms a useful starting
point for the examination of the mechanics of curving.
On most curves, the analysis of curving of conventional vehicles must involve a
consideration of the progressive saturation of the creep forces and realistic wheel-
rail geometry incorporating flange contact. The first comprehensive non-linear
treatment of practical vehicles in curves was given by Elkins and Gostling in 1978
[7], in which numerical solutions were shown to give good agreement with full-
scale experiment. Their treatment covers the non-linearities caused by the movement
of the contact patch across the rail and its subsequent change in shape, creep satura-
tion and large wheel-rail contact angles.
The case of steady motion in a uniform curve is a useful, quasi-static design case
but in practice the response to much more complex track geometry must be consid-
ered. The problems of dynamic response fall into two major categories. The first
concerns the prediction of forces and motions during transients which result from
large discrete irregularities, entry into curves or the negotiation of switch and cross-
ing work. The second major category of dynamic response problem lies in the case
GUIDANCE OF THE TWO-AXLE VEHICLE 109
where irregularities are distributed continuously along the track and statistical char-
acteristics of the response are of interest.
In the severe responses to the large discrete inputs that are of interest for design
purposes, both suspension and wheel-rail contact non-linearities are important and
consequently a numerical integration of the equations of motion is necessary. Early
studies [8] involved assumptions similar to those employed by Heumann, with the
flange force being modelled by a spring. Subsequently, the model of Elkins and
Gostling was extended by Clark, Eickhoff and Hunt [9] to cover dynamic response
and their simulation was successfully validated by full-scale experiment.
4.2 Properties of the Stiffness Matrix
As the elastic connections between the wheelsets and car body have a fundamental
effect on curving, and these are described by the elastic stiffness matrix E, it is im-
portant to consider the properties of E and the various possibilities for design. Con-
sider the vehicle model of Section (2.11). Without any loss of generality, the sym-
metric motions of the car body x
b
, z
b
, and
b
can be dropped from consideration.
Moreover, no terms corresponding to the wheelset rotational freedoms appear in E.
Consequently, attention can be confined to the system described by 7 generalised
coordinates. Let y
1
, y
2
and y
b
be the lateral displacements,
1
,
2
and
b
be the yaw
angles, and
b
be the roll angle of the car body, measured from the unstrained posi-
tion of the vehicle
q = [ y
1
1
y
b
b
y
2
2
]
T
(1)
It is evident that if a two-axle vehicle is in steady motion in a uniform curve that the
attitude of the vehicle can be defined in terms of the wheelset coordinates alone as
these define the position of the car body. Therefore, consider the stiffness matrix E
*
corresponding to the reduced set of coordinates
q = [ y
1
1
y
2
2
]
T
(2)
which can be regarded as defining an equivalent inter-wheelset structure. In order to
investigate the general form of this stiffness matrix the sum and difference co-
ordinates
i
defined by, Figure 4.1,
y
1
= (

1
+

3
)/2 y
2
= (

3
-

1
)/2
(3)

1
= (

2
+

4
)/2

2
= (

4
-

2
)/2

are introduced. This may be written as
q = T (4)
where T is a transformation matrix. Application of the principle of virtual work [10]
then gives for the stiffness matrix
E* corresponding to the new set of coordinates

E* = T
T q
E*T (5)
where
q
E* corresponds to the original coordinates q. Now,
3
corresponds to a rigid
body translation so that
E*
i3
= 0, and since
E* is symmetric,
E*
3i
= 0. Also, in the
case considered here, there is a lateral plane of symmetry, so there can be no ele-
ments representing coupling between
2
and
1
. This is because the state of strain
corresponding to
2
is symmetric with respect to the plane of symmetry whereas that
corresponding to
1
and
4
is antisymmetric. Therefore,
E*
12
=
E*
24
=
E*
21
=
E*
42
= 0. Moreover, the motions described by
1
and
4
can be combined to give a
rigid body rotation, thus

= [ h 0 0 1 ]
T
As the elastic strain energy must be zero in such a motion
T
E* = 0 which yields
E*
14
= - (
E*
11
h
2
+
E*
44
)/2h
Finally, for equilibrium of the structure,
E*
44
=
E*
11
h
2
so that
E*
14
= -
E*
11
h.
The stiffness matrix, therefore, has only two independent elements,
E*
11
= k
s
the
shear stiffness and
E*
22
= k
b
the bending stiffness and is of the form
4
Figure 4.1 Sum and difference coordinates.

=
k k h
k
k h k h
s s
b
s s
0 0
0 0 0
0 0 0 0
0 0
2
(6)
From (5), if U = T
--1
q
E* = U
T
E*U (7)
and so reverting to the original coordinates,
q
E
*
=
k k h k k h
k h k k h k h k k h
k k h k k h
k h k k h k h k k h
s s s s
s b s s b s
s s s s
s b s s b s

+ +
+ +
2
2 2
(8)
This stiffness matrix is quite general and covers any form of elastic inter-wheelset
structure. Specifically, it covers the conventional arrangement of two-axle vehicle
as shown in Figure 2.22. For, referring to the stiffness matrix given by equation
(2.11.28) and considering quasi-static conditions, the coordinates representing the
displacements of the car body can be eliminated in favour of the four wheelset co-
ordinates with the result that the reduced stiffness matrix is of the form (8) where
k
s
= k
y
k
/(2k
y
h
2
+ 2k
) k
b
= k
/2 (9)
It can be seen from (9) that there is a limitation in the shear stiffness k
s
that can be
achieved as k
y
is varied as k
s
k
x
b
2
/h
2
. Of course, k
s
and k
b
can be implemented di-
rectly if some form of direct interconnection between the wheelsets is provided,
such as cross-bracing, and then there is no limit for the value of the shear stiffness
k
s
. This general approach will be applied to examine the possibilities for different
forms of suspension in the next Chapter.
4.3 Steering on Large Radius Curves
As mentioned above, the basic mechanisms of guidance can be clarified by consid-
ering the case of a vehicle with purely coned wheels on a large radius curve, for in
this case the linear equations of motion will be valid. Firstly, the steady-state behav-
iour on a uniform curve will be considered for the case of zero cant deficiency. In
this case, the speed for which the centrifugal force is exactly balanced by the lateral
component of the weight due to track cant. In the next Section, the additional input
due to cant deficiency will be discussed.
The usual definition of the generalised coordinates will be used, so that y
1
and y
2
are the lateral displacements measured from the centreline of the track, and
1
and
2
are the yaw angles measured from the radial position. Then for coned wheels and
small displacements, the gravitational stiffness and lateral force due to spin creep,
together with the other small terms, can be neglected so that the equations of equi-
librium, corresponding to the steady-state form of the simplified equations (2.10.47-
48), reduce to
[ C + E* ] q + E* q
0
= Q
c
(1)
where, assuming that f
11
= f
22
= f for simplicity,
C =
0 2 0 0
2 0 0 0
0 0 0 2
0 0 2 0
0
0
f
f l r
f
f l r
/
/
(2)
For small displacements, E* is given by (2.8) and
Q
c
= [ 0 - 2fl
2
/ R
0
0 - 2fl
2
/ R
0
] (3)
and the forces applied to the wheelsets in the reference position due to straining of
the suspension in the chosen coordinate system are given by Eq
0
where
q
0
= [ 0 h/R
0
0 - h/R
0
]
T
(4)
The solution of equations (1) giving the response to track curvature is
y*
1
, y*
2
= - h( hr
0
/l f/k
s
)/R
0
(1 + ) (5)
1
= -
2
= - h/R
0
(1 + ) (6)
where y*
1
, y*
2
are the lateral displacements measured from the rolling line so that
y*
1
, y*
2
= y
1
, y
2
+ lr
0
/R
0
(7)
and where
= f
2
l/ k
b
k
s
r
0
(8)
It can be seen from (8) that the parameter is indicative of the stiffness of the sus-
pension relative to the elements of the creep stiffness matrix. These solutions indi-
cate that when the wheelsets are restrained elastically further displacements of the
wheelsets beyond those for pure rolling are necessary for equilibrium. The leading
wheelset moves outwards beyond the rolling line defined by y = - lr
0
/R
0
. If hr
0
/l >
f/k
s
the trailing wheelset also moves outwards but if hr
0
/ll < f/k
s
the trailing wheelset
moves inwards. It can be seen from equations (5) and (6) that the wheelsets can
adopt a radial position and move out to the rolling line if, and only if, k
b
= 0. The
longitudinal creepage on each wheel is given by

1ri
= - y
i
*
/r
0
1li
= y
i
*
/r
0
(9)
and the lateral creepage is given by

2ri
=
2li
= - (10)
The longitudinal creep is largest at the leading wheelset, but the lateral creep on
both wheelsets is the same.
Dependent on the stiffness of the suspension there are two different regimes to
consider, broadly corresponding to a flexible, long wheelbase vehicle (with, typi-
cally, a low shear stiffness) and a stiff, short wheelbase vehicle, such as a bogie
(with, typically, high bending stiffness).
In the case of the flexible vehicle, >>1 . It can be seen from (6) that the wheel-
sets adopt a nearly radial attitude and so the lateral creep forces are small. In (5),
f/k
s
> hr
0
/l and so the leading wheelset moves outwards beyond the pure rolling
line, and the trailing wheelset moves inwards relative to the pure rolling line by a
similar amount. These displacements generate longitudinal creep forces which bal-
ance the couples due to the yaw stiffness and the radial position of the wheelsets,
Figure 4.2.
There are two important limitations to the validity of this linear analysis. These
are slip or creep saturation and flange contact. The lateral creep is small, so that ap-
proximately the limiting creep force is given by
W/2 = -f
1
= -k
b
h(1 + k
s
hr
0
/fl)/lR
0
(11)
so that slip occurs on all curves of radius less than R
f
where
R
f
2k
b
h/Wl (12)
when the yaw couples developed by the limiting friction forces are just insufficient
to strain the yaw suspension through angles of h/R
0
. If
f
is the flangeway clear-
ance then from (5) and (7) flange contact will occur, on the leading outer wheel, on
all curves with radius
R
0
< lr
0
/
f
(1 + k
b
h/fl
2
) (13)
approximately. Equations (12) and (13) indicate that low bending stiffness, small
wheelbase, high conicity and high axle-load promote good response to track curva-
ture.
Applying these considerations to the two-axle vehicle of Table 4.1, from (2.9) k
b
=
1.25 MN and k
s
= 50.9 kN/m. The value of = 0.1174 corresponding to the worn
wheel and rail profiles of Figure 2.2 is retained and regarded as applying to a purely
coned wheelset of the same conicity, and f = (f
11
+ f
22
)/2. The flexibility parameter
= 154.3. From (5) and (7) the curve radius that will result in the flangeway clear-
ance, assumed to be 8 mm, being taken up by the leading wheelset is R
0
= 831 m.
Then the lateral displacement of the trailing wheelset is y
2
= 0.015 mm compared
with the rolling line offset lr
0
/R
0
= - 3.45 mm. From (6) the wheelsets adopt a
nearly radial attitude
1
= -
2
= - 0.0287 mr and so the lateral creep forces T
2r1
= -
T
2l1
= -T
2r2
= T
2l2
= 204 N and are small. The longitudinal creep forces on the lead-
ing wheelset T
1r1
= -T
1l1
= - 8433 N and on the trailing wheelset T
1r2
= - T
1l2
= 6408
N. Creep saturation would occur before the flangeway clearance is taken up if <
0.226.
A comparison between theory and experiment for the curving behaviour of a
two-axle research vehicle HSFV-1 was made by Boocock [5] who found excellent
agreement over the range of track curvatures for which linear theory is applicable,
in this case for R
0
> 600 m. For sharper curves, there was considerable scatter as the
variable (and unknown) coefficient of friction influences the results, as indicated by
equation (12), as creep saturation occurs, but in any case a full nonlinear analysis is
necessary as discussed below
In the case of the stiff vehicle, <<1, (5) and (6) show that the wheelsets are
prevented from adopting a radial attitude and so outward lateral creep forces are
generated at the leading wheelset, and inward lateral creep forces are generated at
the trailing wheelset. Equilibrium is provided by yaw moments generated by longi-
tudinal creep forces, both wheelsets moving outwards. With increasing suspension
stiffness the behaviour approximates to that of a rigid vehicle. In this case, in (5)
f/k
s
<< hr
0
/l and so both wheelsets move outwards by the same amount
1
, y
2
= - ( 1 + h
2
/l
2
)lr
0
/R
0
(14)
1

= -
2

= - h/R
0
(15)
k
b
h/R
0
l
k
b
h/R
0
l
k
b
h/R
0
l
k
b
h/R
0
l
k
b
h/R
0
k
b
h/R
0
Figure 4.2 Balance of forces for flexible long wheelbase two-axle vehicle on large radius
curve.
the axles remaining parallel and the displacement from the centreline of the track
being ( 1 + h
2
/l
2
) times the displacement required for pure rolling.
Applying these considerations to the bogie of Table 4.1, from (3.5) k
b
= 20 MN
and k
s
= 7.80 MN/m. Thus, the flexibility parameter = 0.0629. From (5) and (7)
the curve radius that will result in the flangeway clearance, assumed to be 8 mm,
being taken up by the leading wheelset is R
0
= 1440 m. The lateral displacement of
the trailing wheelset y
2
= - 6.51 mm compared with the rolling line offset lr
0
/R
0
= -
1.96 mm. From (6) the wheelsets remain almost parallel
1
= -
2
= - 0.817 mr and
so the lateral creep forces T
2r1
= -T
2l1
= -T
2r2
= T
2l2
= - 5751 N. The longitudinal
creep forces on the leading wheelset T
1r1
= -T
1l1
= - 11011 N and on the trailing
wheelset T
1r2
= - T
1l2
= 8282 N. Creep saturation would occur before the flangeway
clearance is taken up if < 0.35.
Equation (5) shows that for large values of k
b
and small values of k
s
large dis-
placements can occur, the vehicle being unable to react a curvature input. In the
limit, if k
b
= and k
s
= 0 the vehicle is in neutral equilibrium on straight and
curved track. In this case, in the absence of externally applied forces, the vehicle is
able to take up an attitude on straight track defined by y
2
= -y
1
,
2
=
1
= 0 with the
bending moment between the wheelsets being equal to -2fly
1
/r
0
. Such a mode can
be termed the anti-bending mode. The vehicle is unable to react statically an exter-
nally applied yawing moment.
Actually there are two limiting cases in which static stability becomes marginal.
The other case occurs when k
s
= and k
b
= 0. In the absence of externally applied
forces the vehicle is able to take up an attitude on straight track defined by y
1
= y
2
,
Table 4.1 Example parameters for a two-axle vehicle and an isolated bogie.
r
0
= 0.45 m l = 0.7452 m
f
11
= 7.44 MN f
22
= 6.79 MN f
23
= 13.7 kNm W = 78.48 kN
Two-axle vehicle
k
y
= 0.23 MN/m k
= 2.5 MNm k
=1MNm c
y
= 50 kNs/m c
= 0
c
=50 kNms h = 3.7 m d = 0.2 m

m =1250 kg I = 700 kgm
2
m
b
=13500 kg I
xb
=16000 kgm
2
I
zb
= 170000 kgm
2
I
y
= 250 kgm
2
Isolated bogie
k
y
= 40 MN/m k
= 40 MNm k
=1MNm c
y
= 0 c
= 0
c
=10 kNms h = 1.25 m d = 0.2 m

m =1250 kg I = 700 kgm
2
m
b
=2500 kg I
xb
=1000 kgm
2
I
zb
= 3500 kgm
2
I
y
= 250 kgm
2

2
= -
1
= ly
1
/r
0
h with the shear force between the wheelsets being equal to
2fly
1
/r
0
h, the overall balance of forces being in equilibrium. Such a mode can be
termed the anti-shear mode. The vehicle is unable to react statically an externally
applied lateral force.
4.4 Response to Cant Deficiency on a Large Radius Curve
Cant deficiency
d
is defined as the angle through which the track would have to be
canted in order to negate the centrifugal force and is therefore given by
d
= V
2
/R
0
g -
0
(1)
If there is cant deficiency the equations governing the steady state motion in a curve,
equations (3.1), now become
[ C + E* ] q + Eq
0
= Q
c
+ Q
d
(2)
and the applied forces due to cant deficiency Q
d
are
Q
d
= [ -P 0 -P 0 ] (3)
where
P = mg
d
+ m
b
g
d
/2 (4)
It will be useful to first consider the response of the vehicle to cant deficiency in the
absence of the curvature input. The solution of equations (2) is then
y
1
d
, y
2
d

= - hP (- hf/k
b
1 )/2f( 1 + ) (5)
1
d
,
2
d
= - P/2f lhP/2k
b
r
0
( 1 + ) (6)
In the absence of elastic restraint, the wheelsets would yaw through an equal posi-
tive angle of P/2f in order to generate equilibriating lateral creep forces. When the
wheelsets are restrained elastically further displacements of the wheelsets are neces-
sary for equilibrium.
In the case of the flexible vehicle, in which >>1, as hf/k
b
>> 1 (6) reduces to,
approximately,
1
d
,
2
d
= - P/2f hPk
s
/2f
2
(7)
so that the wheelsets yaw through almost the same angle P/2f. Similarly, (5) reduces
to
y
1
d
, y
2
d
= h
2
P k
s
r
0
/ 2f
2
l (8)
so that the wheelsets move inwards slightly from the centreline in order to establish
equilibrium of the yaw moments.
For the two-axle vehicle of Table 4.1, consider the response to a cant deficiency
of
d
= 0.10. This corresponds to a speed of V = 29.1 m/s
2
on track with zero cant
for the same radius of R
0
= 831 m considered above. From (5) the additional lateral
displacement of the leading wheelset is y
1
d
= 0.264 mm and the additional lateral
displacement of the trailing wheelset is y
2
d
= 0.290 mm and are therefore small
compared with the displacements due to curvature. From (6) the wheelsets addition-
ally yaw through
1
d
= 0.537 mr and
2
d
= 0.566 mr and so the additional lateral
creep forces T
2r1
= T
2l1
= 3821 N and T
2r2
= T
2l2
= 4027 N. The additional longitudi-
nal creep forces on the leading wheelset T
1r1
= -T
1l1
= - 488 N and on the trailing
wheelset T
1r2
= - T
1l2
= -537 N and are small.
In the case of the stiff vehicle, in which <<1, (6) reduces to, approximately,
1
d
,
2
d
= - P/2f lhP/ 2k
b
r
0
(9)
the wheelsets also yaw through almost the same angle P/2f. The trailing wheelset
moves outwards and yaws through an angle greater than that required to react the
cant deficiency force. The leading wheelset yaws through an angle less than that
required to react the cant deficiency force. Consequently, more of the cant defi-
ciency force is reacted at the trailing wheelset than at the leading wheelset. Simi-
larly, (4) reduces to
y
1
d
, y
2
d
= - hP (- hf/ k
b
1 )/ 2f (10)
so that the leading wheelset moves outwards less than the trailing wheelset and ac-
tually moves inwards if hf < k
b
. With increasing suspension stiffness the behaviour
approximates to that of a rigid vehicle, the complete vehicle yawing through the
angle P/2f.
For the bogie of Table 4.1, the response to a cant deficiency of
d
= 0.10 corre-
sponds to a speed of V = 38 m/s
2
on track with zero cant for the same radius of R
0
=
1440m. considered above. From (5) the additional lateral displacement of the lead-
ing wheelset is y
1
d
= 0.360 mm and the additional lateral displacement of the trail-
ing wheelset is y
2
d
= - 0.937 mm and are therefore small compared with the dis-
placements due to curvature. From (6) the wheelsets additionally yaw through
1
d
= 0.506 mr and
2
d
= 0.596 mr and so the additional lateral creep forces are T
2r1
=
T
2l1
= 3606 N and T
2r2
= T
2l2
= 4242 N. The additional longitudinal creep forces on
the leading wheelset T
1r1
= -T
1l1
= - 667 N and on the trailing wheelset T
1r2
= - T
1l2
=
-1735 N.
An optimum response to cant deficiency could be considered to be when the cant
deficiency force is equally shared between the wheelsets which can only be satisfied
if k
s
= 0. It can be seen that the requirements for perfect steering and optimum re-
sponse to cant deficiency are in direct conflict.
4.5 The Conflict Between Steering and Stability
It has been shown that the curving of a vehicle in which two wheelsets are elasti-
cally interconnected depends critically on two parameters, the bending and shear
stiffnesses. The dynamic characteristics of a two-axle vehicle at low speeds will now
be considered in order to derive the influence of the bending and shear stiffness on
stability. At low speeds, the form of the equations of motion will be that of equa-
tions (2.10.49). he gravitational stiffness and lateral force due to spin creep, to-
gether with the other small terms, will be neglected. Hence the free motions of the
system, in terms of the sum and difference coordinates defined by (2.3), will be
governed by
[ B s/V + C + E
*
] = 0 (1)
where E* is given by equation (2.6), and after application of the transformation for-
mula (2.5)
B =
f
fl
f
fl
0 0 0
0 0 0
0 0 0
0 0 0
2
2
C
f
f l r
f
f l r
=
0 0 0
0 0 0
0 0 0
0 0 0
0
0
/
/
(2)
Making a trial solution in which , is proportional to e
st
the characteristic equation
of (1) becomes
fs V k f hk
f l r fl s V k
fs V f
hk f l r fl s V k h
s s
b
s s
/
/ /
/
/ /
+
+
+
=
0
0 0
0 0
0
0
0
2
0
2 2
(3)
Expanding (3) gives
p
4
s
4
+ p
3
s
3
+ p
2
s
2
+ p
1
s+ p
0
= 0 (4)
where
p
4
= f
4
l
4
/V
4
p
3
= {k
s
(l
2
+ h
2
) + k
b
}f
3
l
2
/V
3
p
2
= 2 (l/r
0
)f
4
l
2
/V
2
+ k
s
k
b
(l
2
+ h
2
)f
2
/V
2
p
1
= (l/r
0
){k
s
(l
2
+ h
2
) + k
b
}f
3
/V

0
= (/
0
)
2
4
+ (/
0
)
2

Of major interest is how the eigenvalues of equation (4) depend on the overall sus-
pension stiffnesses k
s
and k
b
. When both k
s
and k
b
are zero there are two undamped
oscillations at wheelset kinematic frequency = V(/lr
0
)
1/2
. Introduction of small
values of the stiffnesses k
s
and k
b
results in damping of these oscillations. Putting S
=
1
/
4
equations (3) can be written as two uncoupled binary equations of the form
fs V k f
f l r fl s V k
/
/ /
+
+
=
1
0
2
2
0
(5)
where, in one case, k
1
= k
s
(1 - h/S) and k
2
= k
b
and in the other case, k
1
= 0 and k
2
=
k
s
h(h - S). Equating the two quadratics resulting from equations (5) shows that for
small values of k
s
and k
b
the frequencies remain unchanged but that the real part is
given by
= - V{ k
s
( 1 - h/S ) + k
b
}/2f = Vk
s
h ( h - S )/2f (6)
so that on solving for, and eliminating, S the eigenvalues are given approximately by
s
1,2
= iV(/lr
0
)
1/2
- V(k
s
h
2
+ k
b
+ k
s
l
2
)/4fl
2
V{(k
s
h
2
+ k
b
+ k
s
l
2
)
2
- 4k
b
k
s
h
2
}
1/2
/4fl
2
(7)
Typically, one of these oscillations can be identified with a steering or bending os-
cillation whilst the other oscillation can be identified with a shear oscillation. The
steering oscillation is lightly damped, but the shear oscillation is heavily damped.
If k
b
were zero, one eigenvalue has a zero real part and the vehicle is capable of
undergoing an undamped steering oscillation at kinematic frequency with mode
shape shown in Figure 4.3(a). There is no creep and no shear deformation so that k
s
could be infinitely large without affecting the result, though as discussed previously
the system would be neutrally stable. The second root in this case is given by
s
2
= iV(/lr
0
)
1/2
- k
s
(h
2
+ l
2
)/2fl
2
(8)
and the corresponding mode shape is shown in Figure 4.3(b). The motion involves a
pure shear deformation.
Similarly, if k
s
were zero, one eigenvalue has a zero real part and the vehicle is
capable of undergoing an undamped shear oscillation at kinematic frequency with
mode shape shown in Figure 4.3(c). There is no creep and no shear deformation so
that k
b
could be infinitely large without affecting the result, though as discussed in
Section 4 the system would be neutrally stable. The second root in this case is given
by
2
= iV(/lr
0
)
1/2
- k
b
/2fl
2
(9)
(c) = - 0.5838i (d) = - 0.1480 - 0.5647i
k
s
= 0, k
b
= 1.25 MNm
(a) = 0.5838i (b) = - 0.3896 + 0.5825i
k
s
= 0.023 MN/m, k
b
= 0
Figure 4.3 Mode shapes for steering and shear oscillations showing successive positions of
the vehicle during one half period of its motion in the mode.
and the corresponding mode shape is shown in Figure 4.3(d). The motion involves a
pure bending deformation.
It may be concluded that for a symmetric two-axle vehicle both bending and shear
stiffness are required for stability at low speeds. A symmetric two-axle vehicle which
is able to steer perfectly in the sense of adopting a radial attitude within the assump-
tions discussed above would have zero bending stiffness and therefore a critical speed
of zero. Similarly, a symmetric two-axle vehicle which had optimum response to cant
deficiency by sharing the lateral force equally between the wheelsets would have zero
shear stiffness and therefore a critical speed of zero. It follows that
for the symmetric two-axle vehicle there is a compromise between steering (or re-
sponse to track curvature), cant deficiency and dynamic stability.
4.6 Motion on Sharper Curves
The linear analysis given above is useful in clarifying the basic mechanisms of the
steady state response of a vehicle to track curvature and cant deficiency and is qualita-
tively representative of motion on large radius curves where displacements of the
wheelsets are small. It can be extended by including the effects of spin and the gravi-
tational stiffness as was done in Boococks original work [5]. However, the linear
analysis is limited in applicability because on sharper curves, roughly defined as those
with curve radius R
0
< 1000 m., the displacements of all or some of the wheelsets take
up a significant proportion of the flangeway clearance. The wheel-rail geometry can-
not be represented by simple linear functions and the actual curvatures, slopes and
rolling radii must be taken into account. The contact slopes may be large, giving rise
to large amounts of spin, and resulting in the normal forces having significant com-
ponents in the lateral direction. The variation of the rotational speed of the wheelset
may influence the magnitude of the creepages. As the contact patch moves across the
rail and changes in shape, the variation of the curvatures influences the magnitude of
the creep forces. As the creepages and spin may be large, creep saturation may occur.
Hence it is necessary to consider the solutions of the full non-linear equations of mo-
tion presented in Section 2.11.
As an example, Figure 4.4 summarises the dynamic response of a two-axle vehi-
cle with the parameters of Table 4.1 to entry into a curve of radius 225 m with a
cubic parabolic transition of the type discussed in Section 3.7 with a transition
length L = 20 m. The vehicle speed V = 15 m/s and the coefficient of friction =
0.3. The track has cant
0
= V
2
/R
0
g and therefore the cant deficiency is zero. In the
initial stages of curve entry, the vehicle responds by undergoing a complex damped
response involving motions at body and kinematic frequency. The steady state mo-
tion is then established and the attitude of the vehicle is governed by the quasi-static
version of the equations of motion
F(q, V) + Eq + Eq
0
= 0 (1)
where q
0
represents the position from which the displacements are measured relative
to the unstrained suspension. Of course, (1) may be solved directly, for example by
the Newton-Raphson method. The attitude of the vehicle in plan view is shown in
Figure 4.5. The leading wheelset has moved outwards and has negative yaw relative
to the radial, so that the positive angle of attack generates large values of creep. The
outward position of the wheelset also generates a large amount of spin. Reference to
equation (2.5.13-14) shows that the variation in the rotational speed of the wheelset,
though small, has a significant effect on the magnitude of the longitudinal creepage
on both wheels. As the lateral creep is large, creep saturation results in a large pro-
portion of the available friction force being taken by the lateral creep force. Hence
only relatively small longitudinal creep forces can be generated and the yaw couple
available to strain the yaw suspension is small, and so the wheelset adopts only a
small relative angle with the car body. The trailing wheelset is displaced inwards,
and yaws positively relative to the radial, but the angle is less than on the leading
wheelset so the wheelset is able to generate longitudinal creep forces to react a lar-
ger yaw torque of the suspension. The normal force on the outer leading wheel is
increased and that on the inner reduced, with only slight variation on the trailing
wheels. Most significant is the increase in the lateral components of the normal
forces on the leading outer and trailing inner wheels. This is due to the increased
slope of the contact plane as the wheelset displacements are large. As a result, the
large lateral creep forces on the leading wheelset are reacted by the horizontal com-
0 2
-0.01
0
0.01
0 1 2
-0.03
-0.02
-0.01
0
0 1 2
-0.2
-0.1
0
0 1 2
-0.02
-0.01
0
0 1 2
-0.01
0
0.01
0.02
0.03
0 1 2
-2
0
2
0 1 2
-2
0
2
x 10
4
4
0 1 2
-2
-1
0
x 10
0 1 2
3.8
4
4.2
4
y
1
2
y
b
b
t
-V/r
0
-
1
-V/r
0
-
2
1r2
1l2
1r1
1l1
2r2
2l1
2r1
3r2
3l1
T
1l2
T
1r2
T
1l1
T
1r1
T
2l1
T
2r2
T
2l2
T
2r1
T
3r1
T
3l1
T
3r2
T
3l2
x 10
4
y
2
Figure 4.4 Dynamic response of two-axle vehicle with the parameters of Table 4.1 to entry
into a curve of radius R
0
= 225 m, V = 15 m/s, = 0.3, zero cant deficiency and cubic
parabolic transition of length 20 m.
ponent of the normal force on the outer wheel. This latter force is often identified with
the flange force. The corresponding forces applied to the track tend to cause gauge
spreading.
Next, the effect of cant deficiency is considered. The influence of cant deficiency
on the attitude of the two-axle vehicle traversing the same curve, at the same speed but
with varying angles of track cant angle is shown in Figure 4.6. As is obvious, the car
body moves in the direction of the centrifugal force on the flexible lateral suspension.
The variation of the wheelset yaw angles is small. Figure 4.7 shows the corresponding
lateral wheel forces. Cant deficiency causes additional movements of the wheelsets,
the resulting change in contact angles providing the necessary lateral components of
the normal forces on the leading outer and trailing inner wheels. These latter forces
show the strongest variation as the cant deficiency is varied.
Figure 4.7 also shows the variation in the lateral wheel forces due to cant defi-
ciency for an exactly similar case of curve entry but with a reduced coefficient of fric-
tion = 0.1. As might be expected, the magnitude of the forces is reduced, as are
the lateral displacements of the wheelsets (Figure 4.6). The yaw angles approach the
Figure 4.5 Attitude of two-axle vehicle on curve of radius R
0
= 225 m, V = 15 m/s, = 0.3.
Relative to the lateral displacements, vehicle dimensions are reduced k times, R
0
is reduced
k
2
times, and yaw rotations are multiplied by k
2
, where k = 25.
-0.2 0 0.2
-0.01
-0.005
0
0.005
0.01
d
-0.2 0 0.2
-0.02
-0.01
0
0.01
0.02
y
2
y
b
/10
y
1
1
Figure 4.6 Variation of attitude of example two-axle vehicle with cant deficiency. Solid lines
refer to = 0.3, dotted lines refer to = 0.1.
limiting values of h/R
0
.
Consider now the corresponding results for the solution of the non-linear equa-
tions of motion of an unrestrained bogie, with = 0.3, and at the same speed. Ap-
proximate allowance has been made for the cant deficiency force acting on the car
body which is applied as a lateral force and rolling moment
m
c
V
2
/2R
0
-m
c
g
0
/2
on the bogie frame. The vertical forces are similarly represented as discussed in Sec-
tion 2.11. In the steady state motion on a large radius curve, the linear theory showed
that the wheelsets of the bogie would move outwards and yaw in opposite directions
in attempting to take up radial positions. The nonlinear solution for R
0
= 1500 m con-
firms this and is shown in Figure 4.8(a). As the radius of the curve is reduced, there
is increasing yaw of the bogie and the trailing wheelset moves inwards, as shown in
Figures 4.8(b)-(e). The leading wheelset moves outwards and has increasing negative
yaw, so that the positive angle of attack generates large values of lateral creep and
only relatively small longitudinal creep forces can be generated by the leading
wheelset. The large lateral creep forces are mainly reacted by the horizontal compo-
nent of the normal force. The trailing wheelset takes up a more or less radial position,
the lateral creep forces are relatively small, so the wheelset is able to generate the lon-
gitudinal creep forces necessary for yaw equilibrium. The balance of forces in plan
view for R
0
= 225 m is shown in Figure 4.9. It can be seen that creep saturation is com-
-0.2 0 0.2
-2000
-1000
0
1000
2000
T
xl1
-0.2 0 0.2
-2
0
2
4
x 10
4
-0.2 0 0.2
-1
-0.5
0
0.5
1
x 10
4
d
-0.2 0 0.2
-2
-1
0
1
2
x 10
4
T
xr1
T
yl1
T
yr1
T
yl2
T
yr1
T
yr2
T
xr2
T
xl2
d
Figure 4.7 Variation of wheel forces of example two-axle vehicle with cant deficiency. Solid
lines refer to = 0.3, dotted lines refer to = 0.1.
plete and the flangeway clearance is taken up at the leading outer and trailing inner
wheels. This indicates partial convergence with the classical theory of Porter [4] and
the condition referred to as constrained curving. As the primary suspension is stiff,
the relative displacements between bogie frame and wheelsets are small and in Fig-
ure 4.8 it can be seen that the movement of the vehicle appears to be as a rigid body.
The effect of cant deficiency on the bogie vehicle traversing a curve with R
0
=
225 m, at the same speed but with varying angles of track cant angle is indicated in
Figures 4.10 and 4.11. Here the coefficient of friction = 0.1. Figure 4.10 shows
how the bogie yaw angle and outward displacement increase with increasing cant
deficiency. Figure 4.11 shows the corresponding lateral wheel forces.
The non-linear treatment of practical vehicles in curves was first given by Elkins
and Gostling in 1978 [7], in which numerical solutions were shown to give good
(a)
(b)
(c)
(d)
(e)
Figure 4.8 Attitude of bogie with parameters of Table 4.1 on curves of various radii with zero
cant deficiency (a) R
0
= 1500 m (b) R
0
= 800 m (c) R
0
= 500 m (d) R
0
= 225 m (e) R
0
=
100 m.
P
y2
W/2
T
3l1
sin
l1
T
2l1
cos
l1
T
2r1
cos
r1
T
1l1
T
1r1
T
2r1
cos
r1
T
1r1
T
3r2
sin
r2
P
y1
M
z1
M
z2
T
1l2
T
2l2
cos
l2
T
3l2
sin
l2
T
3r1
sin
r1
Figure 4.9 Plan-view forces acting on wheels of bogie of Table 4.1 in curve R
0
= 225 m, V =
15 m/s, = 0.3 and zero cant deficiency.
-0.2 0 0.2
-0.01
-0.005
0
0.005
0.01
d
-0.2 0 0.2
-15
-10
-5
0
5
x 10
-3
2
y
1
y
2
y
b
d
Figure 4.10 Influence of cant deficiency on attitude of bogie vehicle with parameters of Table
4.1, R
0
= 225 m, V =15 m/s., = 0.1.
agreement with full-scale experiments with both bogie and two-axle vehicles.
It should be noted that these solutions indicate that a resultant longitudinal force
exists on a wheelset in a curve. This force is greater in sharp curves, which explains
the greater drag of a train in a curve.
4.7 Response to Misalignments
The effects of misalignments may be considered in terms of the configuration of the
vehicle when the suspension is in the unstrained state. If y
10
,
10
, y
20
and
20
repre-
sent the displacements of the wheelset when the elastic elements are unstrained then
the column of initial displacements q
0
, will be given by
q
0
= [ - y
10
-
10
- y
20
-
20
]
T
(1)
and the effect of the misalignments will be given by the solution of the equations of
equilibrium for small displacements (3.1). In the case of the two-axle vehicle or bo-
gie
q
0
= [ - k
s
s
- k
b
b
+ k
s
h
s
k
s
s
k
b
b
+ k
s
h
s
]
T
(2)
-0.2 0 0.2
-1000
-500
0
500
1000
d
T
yr1
-0.2 0 0.2
-1
0
1
2
x 10
4
-0.2 0 0.2
-5000
0
5000
-0.2 0 0.2
-2
-1
0
1
x 10
4
T
1l1
T
1r1
T
yl1
d
d
T
yr2
T
yl2
T
1l2
T
1r2
Figure 4.11 Influence of cant deficiency on wheel forces for bogie vehicle of Table 4.1, R
0
= 225 m, V =15 m/s, = 0.1.
where
s
= y
10
- h
10
- y
20
-
20
(3)
b
=
10
-
20
(4)
The response to a shear misalignment is accordingly
y
1
s
, y
2
s
=
s
(- hf/ k
b
1 )/ 2( 1 + ) (5)
1
s
=-
2
s
= - lf
s
/ 2k
b
r
0
( 1 + ) } (6)
In the case of the flexible vehicle, in which >>1, as hf/k
b
>> 1 (6) reduces to, ap-
proximately,
1
s
, -
2
s
= - k
s
s
/2f (7)
Similarly, (5) reduces to
y
1
d
, y
2
d
= - hk
s
r
0
s
/2fl (8)
so that the wheelsets move laterally and yaw in opposite senses, the elastic forces
induced by a shear misalignment being directly reacted by the corresponding creep
forces.
In the case of the stiff vehicle, in which <<1, (7) reduces to, approximately,
1
d
,
2
d
= - lf
s
/ 2k
b
r
0
(9)
Similarly, (5) reduces to
y
1
d
, y
2
d
=
s
(- hf/k
b
1 )/2 (10)
so that if there is a shear misalignment
s
the wheelsets will initially move differen-
tially through
s
/2. The yaw moments thus generated are balanced by equal and
opposite yaw displacements given by (10); overall yaw equilibrium is obtained by
further lateral displacements of the wheelsets - hf
s
/2k
b.
The response to a bending misalignment is
y
1
b
, y
2
b
=
b
( hr
0
/l f/k
s
)/ 2( 1 + ) (11)
1
b
= -
2
b
=
b
/ 2( 1 + ) (12)
In the case of the flexible vehicle, >>1 . It can be seen from (12) that the yaw an-
gles are negligible and so the lateral creep forces are small. In (11) f/k
s
>> hr
0
/l
and so, approximately,
Table 4.2 Response of example vehicles of Table 4.1 to misalignments.
Two-axle vehicle
y
10
= 5 mm
y
1
= - 0.323 mm

1
= - 0.0178 mr
y
2
= - 0.355 mm

2
= 0.0178 mr
Two-axle vehicle

10
= 1 mr
y
1
= 0.7505 mm

1
= 0.01637 mr
y
2
= -0.1259 mm

2
= - 0.01637 mr
Bogie
y
10
= 2 mm
y
1
= 1.306 mm

1
= - 0.1623 mr
y
2
= - 3.398 mm

2
= 0.1623 mr
Bogie

10
= 0.5 mr
y
1
= 1.567 mm

1
= 0.2554 mr
y
2
= 1.727 mm

2
= - 0.2554 mr
y
1
b
, y
2
b
= k
b
r
0
b
/2fl (13)
so that in response to a bending misalignment the wheelsets move laterally and diff-
erentially, the elastic moments being directly reacted by the longitudinal creep forces.
In the case of the stiff vehicle, <<1 and (11) and (12) reduce to
1
b
,
2
b
=
b
hr
0
/2l (14)
1
b
= -
2
b
=
b
/2 (15)
so that the bending misalignment is accommodated by equal and opposite yaw dis-
placements, both wheelsets moving laterally to provide overall equilibrium of the
yaw moments.
Comparison of equations (11) and (12) with (3.5) and (3.6) shows that the response
to a bending misalignment is related to the response to curvature so that
y*
i
/(2h/R
0
) = -y
i
b
/
b
(16)
i
/(2h/R
0
) = -
i
b
/
b
(17)
which indicates, as might be expected, that the response to a misalignment in bend-
ing is the same as the response to curvature if 2h/R
0
= -
b
. It follows that the re-
sponse to a typical misalignment in bending will ordinarily be significantly less than
the response to curvature. Also comparison of equations (5) and (6) with (4.5) and
(4.6) shows that the response to a shear misalignment is related to the response to
cant deficiency so that
y
i
d
/(P/2f) = y
i
s
/(
s
/2h) (18)
(
i
d
+ P/2f )/(P/2f) =
i
s
/ (
s
/2h) (19)
which indicates, as might be expected, that the lateral displacement due to a mis-
alignment in shear is the same as that due to cant deficiency if hP/2f =
s
. The re-
sponse to misalignment in shear will ordinarily be significantly less than the re-
sponse to cant deficiency.
Though the displacements caused by typical, practical values of misalignments
are small, the creepages and creep forces can be significant and can provide a con-
tribution to creep saturation which should not be overlooked.
4.8 Flange Forces and Derailment
The conditions necessary to sustain equilibrium of the forces in flange contact were
briefly mentioned in Chapter 1. When a wheel is in flange contact with a rail, small
lateral displacements of the wheelset result in large changes in the contact angle. The
gravitational stiffness becomes large and may exceed the lateral stiffness of the track.
Realistic estimates of the forces acting between wheel and rail in this situation require
the inclusion of extra degrees of freedom representing the rail movement at each
wheel. For the frequencies under consideration here, usually quite simple spring and
damper models will be adequate to represent the dynamics of the track [9].
Derailment is the consequence of wheel climb in which large lateral forces acting
on a wheelset of a vehicle cause one wheel to climb up and over the rail. The dynam-
ics of this process was considered by Matsui [11]. Detailed solutions of the equations
of motion on irregular track have been compared with experiment [9] and good
agreement obtained. The results confirm operating experience in which many derail-
ments occur as a result of the coincidence of high lateral forces (such as those experi-
enced during hunting) with vertical unloading of a wheel, perhaps due to a track de-
fect. Sweet et al. [12] also investigated the dynamics of derailment and obtained good
agreement with model experiments.
References
1. Mackenzie, J.: Resistance on railway Curves as an element of danger. Proc. I. C.
Eng. 74 (1883), pp. 1-57.
2. Gilchrist, A.O.: The long road to solution of the railway hunting and curving
Problems. Proc. I. Mech. E. 212, Part F (1998), pp. 219-226.
3. Heumann, H.: Grundzuge der Fuhrung der Schienenfahrzeuge. Oldenbourg,
Munchen, 1953.
4. Porter, S.R.M.: The mechanics of a locomotive on curved track. The Railway Ga-
zette, London, 1935.
5. Boocock, D.: Steady-state motion of railway vehicles on curved track. Journal
Mech. Eng. Sci. 11 (1969), No. 6, pp. 556-566.
6. Newland, D.E.: Steering characteristics of bogies. Railway Gazette, 124 (1968),
No.19, pp. 745-750.
7. Elkins, J.A. and Gostling, R.J.: A general quasi-static curving theory for railway
vehicles. In: A. Slibar and H. Springer (Eds.): The Dynamics of Vehicles on Roads
and Tracks, Proc. 5th IAVSD Symposium, Vienna, September 1977. Swets and
8. Muller, C. Th.: Dynamics of railway vehicles on curved track. Proc. I. Mech. E.
180 (1965-66), Part 3F, pp. 45-57.
9. Clark, R.A., Eickhoff, B.M., Hunt, G.A.: Prediction of the dynamic response of
vehicles to lateral track irregularities. In: A.H. Wickens (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 7th IAVSD Symposium, Cambridge, September
10. Collar, A.R. and Simpson, A.: Matrices and engineering Dynamics. Ellis Hor-
wood, Chichester, 1987, p. 482.
11. Matsui, N.: On the derailment quotient Q/P. Railway Technical Research
Institute, Japanese National Railways, 1966.
12. Sweet, L.M., Karmel, A. and Fairley, S.R.: Derailment mechanics and safety
criteria for complete rail vehicle trucks. In: A.H. Wickens (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 7th IAVSD Symposium, Cambridge, August
Sweet, L.M. and Sivak, J.A.: Nonlinear wheelset forces in flange contact - Part I:
Steady state analysis and numerical results. ASME Transactions, J. of Dynamic
Systems, Measurement and Control 101, No. 3, September, 1979.
Sweet, L.M. and Sivak, J.A.: Nonlinear wheelset forces in flange contact - Part II:
Measurements using dynamically scaled models. ASME Transactions, J. of Dynamic
Systems, Measurement and Control 101, No. 3, September, 1979.
Sweet, L.M. and Karmel, A.: Evaluation of time-duration dependent wheel load
criteria for wheel climb derailment. ASME Transactions, J. of Dynamic Systems,
Measurement and Control 103, No. 3, September 1981, pp. 219-227.
13. Wickens, A.H. and Gilchrist, A.O.: Railway vehicle dynamics the Emergence
of a Practical Theory. Council of Engineering Institutions MacRobert Award Lecture,
1977.
5
Dynamic Stability of the
Two-Axle Vehicle
5.1 Introduction
In this chapter the dynamic stability of the isolated two-axle vehicle will be consid-
ered. Both long wheelbase vehicles and short wheelbase vehicles will be discussed,
the former typically representative of freight vehicles and the latter typically repre-
sentative of bogies which are unrestrained by the secondary suspension connecting
the bogie to the car body. The complete bogie vehicle will be considered in Chapter
6.
The instability of two-axle vehicles was an accepted and often unremarked oc-
currence throughout their employment on the railways. For example, the coaches
used on the Liverpool and Manchester Railway had a very short wheelbase and
were reputed to hunt violently at any speed [1]. One measure employed to control
this was to close couple the vehicles. Early bogies had very short wheelbases and
were free to swivel without restraint and tended to oscillate violently being the
probable cause of many derailments. In the case of locomotives, in the 1850s the
wheelbase of the leading bogie was increased which improved stability signifi-
cantly, [2].
The suspension of the earliest two-axle vehicles was provided by simple leaf springs
bearing directly on to shoes bolted to the vehicle frame. The lateral suspension was
only provided fortuitously as a consequence of the vertical suspension. No lateral restor-
ing force was provided until the lateral clearance of the journal axle-bearing was taken
up: thereafter some stiffness was provided by rocking of the heavily cambered spring on
its seats. After the axle-guard clearance was taken up, further motion was resisted by
bending of the axle-guard. The longitudinal suspension consisted of friction of the shoes
until the axle-guard clearance was taken up after which extremely stiff restraint was pro-
vided by the axle-guards. Needless to say, the action of such suspensions was highly
nonlinear and their modelling has been discussed in [1.8]. In the later eye-bolt suspen-
sion, more lateral flexibility, within the axle-guard clearances, was provided by swing-
links which were free to move both laterally and longitudinally. In an attempt to improve
curving, this type of suspension was later refined in the form of the double-link suspen-
sion, which became a German standard in 1890, and which provides more flexibility.
This form of suspension remains in common use in Europe and Japan. Operation of ear-
lier designs of vehicles at higher speeds in the 1950s and 1960s led to unacceptable rates
of derailment which stimulated progress in both theory and design.
Matsudaira was the first to investigate the dynamic stability of the two-axle vehicle,
using Carters theory of creep, and incorporating both the lateral and longitudinal
flexibilities between wheelsets and the car body [3]. Subsequently, it was shown in
[4] that inclusion of appropriate amounts of suspension damping, neither too small
or too large, resulted in regions of clear stability up to relatively high speeds. More-
over, a realistic theory, taking into account the nonlinear flexibility between the
wheelsets and the car body or bogie frame, and the forces acting between vehicle
and track, yielded results which are consistent with experiment [1.11]. The theory
shows that there are two ways of exploiting the suspension stiffnesses in the design
of two-axle railway vehicles for stable running at speed. The first is to use a rela-
tively flexible suspension, in which the flexibility parameter = f
2
l/k
b
k
s
r
0
defined
in Section 4.3 is large, with appropriately chosen parameters this approach is suit-
able for two-axle vehicles with relatively long wheelbase. The second approach is to
use a suspension which has large lateral and yaw stiffness in which the flexibility
parameter is small, this is appropriate to a short wheelbase vehicle such as a
bogie. These concepts were demonstrated in the laboratory on a roller rig [4] and by
track tests with a specially designed full-scale variable parameter test vehicle HSFV-
1[23]. Further experimental and theoretical experience has led to the development of
suspensions for two-axle freight and passenger vehicles capable of running at simi-
lar speeds to bogie vehicles, such as the widely used Class 143 and 144 passenger
vehicles on British Railways.
The improvement of curving performance whilst still providing the yaw restraint
necessary for stable running can be achieved by frequency sensitive yaw restraint
between the wheelsets and car body using a relaxation or yaw damper, as first pro-
posed by Hobbs [5]. The yaw damper provides little restraint at low frequencies so
that the wheelsets are able to take up an approximately radial position, but at high
frequencies the restraint is sufficient to stabilise the vehicle. The advent of reliable
dampers suitable for use in the primary suspension, in the 1970s, made it possible to
apply this concept to freight vehicles [6] and articulated passenger vehicles as dis-
cussed in Chapter 8.
As mentioned in Section 4.2 for conventional bogies in which there are primary
longitudinal and lateral springs connecting the wheelsets to a frame there is a limit
to the overall shear stiffness which can be provided in relation to the bending stiff-
ness and therefore the stability/curving trade-off in which the bending stiffness must
be minimised is constrained. This limitation is removed if the wheelsets are con-
nected directly by diagonal elastic elements or cross-bracing, or interconnections
which are structurally equivalent. Such an arrangement is termed a self-steering bo-
gie. Superficially, this arrangement is similar to systems of articulation between ax-
les by means of rigid linkages which have a long history in railway engineering. The
first application of cross-bracing appears to have been on the vehicles of the Linz-
Budweiser Pferdebahn (1827), to be seen in the Vienna Museum of Technology.
In the 1970s the self-steering bogie was successfully developed and put into ser-
vice, notably by Scheffel [7]. The essential feature of cross-bracing in modern prac-
tice is that it is elastic. Self-steering bogies are common in current practice and have
been applied to locomotives (with benefits to the maximum exploitation of adhe-
sion), passenger vehicles and freight vehicles [8]. It should be noted that inter-
DYNAMIC STABILITY OF THE TWO-AXLE VEHICLE 135
wheelset connections can be provided by means other than springs and dampers. In
[9] the equivalent of cross-bracing is provided by means of a passive hydrostatic
circuit which has a number of potential design advantages.
Equations of motion for the two-axle vehicle with a simple suspension have been
derived in Section 2.11. Whilst a simple representation of the suspension suffices to
discuss the principles of the subject, in design work it is necessary to model the de-
tails of the suspension in order to obtain accurate results. Particular examples are the
behaviour of springs which carry vertical load and the effect of trailing arm primary
suspensions [10, 11, 12] not to mention non-linearities in suspension behaviour.
However, a more fundamental aspect is concerned with the options that exist in rela-
tion to the arrangement of the suspension components and the bodies that they con-
nect. To discuss this it is necessary to consider the properties of the elastic stiffness
matrix.
As the rotation of the wheelsets cannot generate any forces in the suspension
consider the following set of seven generalised coordinates
q =[ y
1
1
y
b
b
y
2
2
]
T
(1)
(the set that would be considered in a linear analysis). As in Chapter 4, it will be
advantageous to use sum and difference coordinates,
i
, Figure 5.1,
6
Figure 5.1 Sum and difference coordinates for two-axle vehicle.
y
1
= (

1
+

4
)/2 y
2
= (

4
-

1
)/2
1
= (

2
+

5
)/2

2
= (

5
-

2
)/2 (2)
y
b
=

b
=

b
=
3
/h
Because the vehicle is symmetric and E must be symmetric, in the most general case
the stiffness matrix corresponding to the coordinates will be of the form
E
e e e
e e e e
e e e
e e e e
e e e
e e e e
e e e e
=
11 13 15
22 24 26 27
13 33 35
24 44 46 47
15 35 55
26 46 66 67
27 47 67 77
0 0 0 0
0 0 0
0 0 0 0
0 0 0
0 0 0 0
0 0 0
0 0 0
(3)
and following the procedure of Section 4.2 the condition that no elastic forces can
be generated in a rigid body lateral translation
E
l
= 0 (4)
where
l
= [ 0 0 0 2 0 1 0 ]
T
(5)
yields
e
26
= - 2e
24
e
46
= - 2e
44
e
66
= 4e
44
e
76
= - 2e
74
(6)
Applying the condition that no elastic forces can be generated by a rigid body yaw
E
y
= 0 (7)
where
y
= [ 2h 0 h 0 2 0 1 ]
T
(8)
yields
e
13
=- e
11
- e
33
/4 + e
55
/h
2
e
15
= - he
11
/2 + e
33
h/8 - e
55
/2h (9)
e
35
= e
11
h - e
33
h/4 - e
55
/h
with the result that there are a maximum of 9 independent parameters. It should be
noted that e
11
, e
33
and e
55
are not completely arbitrary as the principal minors of E
corresponding to
1
,

3
and
5
must be positive definite as the system is conserva-
tive. A similar condition applies to e
22
and e
44
.
For the two-axle vehicle of Figure 2.22 the matrix of primitive stiffnesses is
k = diag [ k
y
k
k
y
k
] (10)

and the corresponding compatibility matrix, corresponding to the set of coordinates
defined by (1), is
a
d h
d h
=

1 0 1 0 0
0 0 0 1 0 0 0
0 1 0 0 1 0 0
0 0 1 1 0
0 0 0 1 0 0 0
0 0 0 0 1 0 1
(11)
Applying (2.11.25) and (4.2.5) and comparing the result with (3), the nine inde-
pendent elements of E, in terms of the sum and difference coordinates, are
e
11
= k
y
/2 e
22
= k
/2 e
24
= e
27
= 0 e
33
= 2k
y
+ 2k
/h
2
(12)

e
44
= k
y
/2 e
47
= dk
y
e
55
= k
/2 e
77
= 2d
2
k
y
+2k

It was discussed in Section 4.2 that at low speeds, where inertia forces can be ne-
glected, the frame and suspension provides only an elastic connection between the
wheelsets. This applies not only to the conventional bogie or vehicle, in which two
wheelsets are elastically connected to a common frame by means of lateral and lon-
gitudinal springs, but applies to other forms of structure, such as direct links be-
tween the wheelsets. Whatever form of structure is chosen it can be described by the
two independent stiffness parameters; the shear stiffness k
s
and the bending stiffness
k
b
. When the attitude of the vehicle can be defined in terms of the wheelset coordi-
nates
1
,

2
,

4
and
5
alone, the result of eliminating
3
,

6
and
7
is that the deflated
stiffness matrix E* is defined in terms of the bending stiffness k
b
and shear stiffness
k
s
by (4.2.6). The discussion of Section 4.5 shows that, for stability at low speeds, k
b
and k
s
must be nonzero. For the case in which E is given by equation (3) the shear
and bending stiffnesses are given by

k
s
= ( e
11
e
33
- e
13
2
)/e
33
(13)
k
b
= e
22
- e
24
2
/ e
44
(14)
and for the simple configuration k
s
and k
b
(13) and (14) reduce to (4.2.9).
Written out in full, the equations of motion (in the simplified form of Section
2.10), in terms of the sum and difference coordinates, then become
{(m/2)s
2
+ fs/V + k
y
/2}
1
- f
2
- k
y
3
= 0 (15)
(fl/r
0
)

1
+ {(I/2)s
2
+ fl
2
s/V + k
/2}
2
= 0 (16)
- k
y
1
+ {(I
zb
/h
2
)s
2
+ 2k
y
+ 2k
/h
2
}
3
- (k
/h)
5
= 0 (17)
k
q
h
1 h
1
k
0
(c)
h
a
h
a
h h
k
k
p
k
d
k
y
/2
(b)
Figure 5.2 Alternative suspension schemes for the two-axle vehicle (a) direct shear and
bending connections between wheelsets achieved by cross-bracing (b) and (c) an ex-
ample generic suspension scheme.
k
y
/2 k
p
/2l
1
2
k
/2l
1
2
(1+h
2
/l
1
2
)k
d
/2
2l
1
(a)
{(m/2)s
2
+ fs/V + k
y
/2}
4
- f
5
- k
y
6
+ d
1
k
y
7
= 0 (18)
(fl/r
0
)
4
+ {(I/2)s
2
+ fl
2
s/V + k
/2}
5
- (k
/h)
3
= 0 (19)
- k
y
4
+ ( m
b
s
2
+ 2 k
y
)
6
- 2dk
y
7
= 0 (20)
- 2dk
y
6
- k
y
d
1
4
+ (I
xb
s
2
+ 2d
2
k
y
+ 2k
)
7
= 0 (21)
so that with this choice of coordinates, the first set of three equations is coupled to
the second set of four equations through only one term proportional to the yaw stiff-
ness k
and, moreover, the wheelbase h only occurs in conjunction with k
. These
facts can be exploited in deriving solutions to the equations of motion, and are help-
ful in understanding the dynamics of the system.
Now consider an alternative configuration employing cross-bracing. This is of-
ten applied in addition to a primary longitudinal and lateral suspension of the con-
ventional kind, Figure 5.2(a). This provides 4 parameters for optimisation. As the
overall shear stiffness can be increased beyond the limit for a conventional bogie
there are two benefits. Firstly, the reduction of overall bending stiffness improves
steering, and, secondly, as discussed later, under certain conditions dynamic stability
is improved.
In this case the matrix of primitive stiffnesses is
k = diag [ k
y
k
k
y
k
k
p
k
d
] (22)

and the compatibility matrix is, in terms of the coordinates of (1),
a
d h
d h
h h
=

1 0 1 0 0
0 0 0 1 0 0 0
0 1 0 0 1 0 0
0 0 1 1 0
0 0 0 1 0 0 0
0 0 0 0 1 0 1
0 1 0 0 0 0 1
1 0 0 0 1
(23)
where it is assumed for simplicity that the cross-bracing is in the plane of the axles.
Applying (2.11.25) and (4.2.5), the nine independent elements of E are then
e
11
= k
y
/2 + k
d
e
22
= k
/2 + k
p
e
24
= e
27
= 0 e
33
= 2k
y
+ 2k
/h
2
(24)

e
44
= k
y
/2 e
47
= dk
y
e
55
= h
2
k
d
+ k
/2 e
77
= 2d
2
k
y
+2k

Applying equations (13) and (14) the shear and bending stiffnesses are given by

k
s
= k
y
k
/(2k
y
h
2
+ 2k
) + k
d
(25)
k
b
= k
/2 + k
p
(26)
and written out in full, the equations of motion (in the simplified form of Section
2.10) then become
{(m/2)s
2
+ fs/V + k
y
/2 + k
d
}
1
- f
2
- k
y
3
hk
d
5
= 0 (27)
(fl/r
0
)

1
+ {(I/2)s
2
+ fl
2
s/V + k
/2 + k
p
}
2
= 0 (28)
- (k
/h)
5
- k
y
1
+ ( (I
zb
/h
2
)s
2
+ 2k
y
+ 2k
/h
2
)
3
= 0 (29)
{(m/2)s
2
+ fs/V + k
y
/2}
4
- f
5
- k
y
6
+ d
1
k
y
7
= 0 (30)
(fl/r
0
)
4
+ {(I/2)s
2
+ fl
2
s/V + k
/2 + h
2
k
d
}
5
- (k
/h)
3
hk
d
1
= 0 (31)
- k
y
4
+ ( m
b
s
2
+ 2 k
y
)
6
- 2dk
y
7
= 0 (32)
- 2dk
y
6
- k
y
d
1
4
+ {I
xb
s
2
+ 2d
2
k
y
+ 2k
)}
7
= 0 (33)
The discussion of Section 4.5 and equations (25) and (26) show that stability at low
speeds can be obtained with k
= 0 and with k
p
and k
d
nonzero. In this case the cou-
pling term (k
/h) between
2
and
5
is of course no longer present but a coupling
term hk
d
between
1
and
5
has now been introduced. However, a combination of
all four stiffnesses is usually exploited.
Damper elements may, of course, be disposed in similar ways to the springs so
far considered. More generally, complex suspension elements can be exploited, the
effect of which is to replace k
, for example, in the equations of motion by an im-

pedance z
(s). Figure 5.3(a) indicates the arrangement consisting of a damper in

series with a spring. The mass m
r
is small and so at the frequencies being considered
here the effect of the relaxation damper is to replace k
in the equations of motion

by k
r
c
r
s/(k
r
+ c
r
s). For oscillations at frequency the effective stiffness is k
r
c
r
2
2
/(k
r
2
+ c
r
2
2
) and the effective damping is k
r
2
c
r
2
/(k
r
2
+ c
r
2
2
). An alternative arrange-
ment, Figure 5.3(b), is to provide the equivalent of a yaw relaxation spring directly
between the wheelsets.
It will be noted that in the above schemes e
24
and e
27
are both zero, so that the
full possibilities are not being exploited. Various representations of generic two-axle
vehicles or bogies have been considered [13-19]. An example of such a scheme is
shown in Figure 5.2(b) and (c). In this case the matrix of primitive stiffnesses is
k = diag [ k
y
k
k
y
k
k
p
k
d
k
q
k
0
] (34)

and the compatibility matrix is, in terms of the coordinates of (1)

a
h d h h
d h h h
h h
h d h
h h h h
a a
a a
=
+
1 1 0 0
0 0 0 1 0 0 0
0 1 0 0 1 0 0
0 0 1 1
0 0 0 1 0 0 0
0 0 0 0 1 0 1
0 1 0 0 0 0 1
1 0 0 0 1
1 2 0 1
1 0 0 2 1
1 1 1
1 1 1
( )
( )
( )
(35)
where d
1
is the distance below the car body centre of mass of the spring k
q
. Applying
(2.11.25) and (4.2.5), the nine independent elements of E are then
e
11
= k
y
/2 + k
d
+ k
0
/2(h + h
1
)
e
22
= k
/2 + k
p
+ h
a
2
k
y
/2 +h
1
2
k
q
k
r
/2l
1
2
2l
1
c
r
/2l
1
2
m
r
/2l
1
2
(a)
(b)
Figure 5.3 The application of relaxation dampers to the two-axle vehicle (a) conventional
arrangement (b) direct yaw connection between wheelsets.
e
24
= h
a
k
y
/2 + h
1
k
q
e
27
= dh
a
k
y
+ d
1
h
1
k
q
e
33
= 2k
y
(h + h
1
)
2
/h
2
+ 2k
/h
2
+ k
0
/h
2
(36)

e
44
= k
y
/2 + k
q
e
47
= dk
y
+ d
1
k
q
e
55
= h
2
k
d
+ k
/2 + h
a
2
k
y
/2 + h
1
2
k
0
/2(h + h
1
)
e
77
= 2d
2
k
y
+2k
+d
1
2
k
q
Applying equations (13) and (14) the shear and bending stiffnesses are given by
k k
k k h h k k k k h h
h h k h h k k
s d
y y a
y a
= +
+ + +
+ + + +
2
2 2 2
1
2
0 0 1
2
1
2 2
0

( ) ( )
( ) { ( ) }
(37)
k
k
k
k k h h
k k
b p
y q a
y q
= + +
2 2
1
2
( )
( )
(38)
Note that k
0
contributes to k
s
only and this only in conjunction with k
or k
y
(pro-
vided that h
a
h
1
); k
q
contributes to k
b
only and this only in conjunction with k
y
.
Whilst a limited number of basic stiffnesses are necessary for stability, strictly a
full optimisation of the vehicle would involve selecting the optimal values of the
independent elements of the stiffness matrix E in order to meet a specified combina-
tion of stability and steering performance, and then deriving the basic stiffnesses to
achieve these. However, although it is possible to choose at will the maximum num-
ber of independent parameters, this usually comes at the price of significant com-
plexity of the mechanical arrangements in order to achieve a marginal advantage.
5.3 Stiff and Flexible Vehicles
The influence of speed on the eigenvalues for the two-axle vehicle with the wheel and
rail profiles of Figure 2.2 and with the parameters given in Table 4.1 (except that in
this Section c
y
and c
= 0) is shown in Figure 5.4. The eigenvalues at a specific speed

are given in Table 5.1. At a given speed the eigenvalues consist of (a) four relatively
large real negative eigenvalues. Two of these eigenvalues are approximately equal to -
2f
22
/mV corresponding to wheelset subsidences in lateral translation, and two eigen-
values are approximately equal to -2f
11
l
2
/IV corresponding to wheelset
At speed, V = 10 m/s Wheels Fixed Label
1 -1176 Wheelset subsidence in yaw
2 -1128 Wheelset subsidence in trans.
5,6 -3.258 10.65i -3.266 10.65i Car body upper sway
7,8 -4.711 7.638i -4.026 7.088i Car body yaw
9,10 -3.724 4.473i -3.688 4.555i Car body lower sway
11,12 -1.617 5.680i Shear oscillation
13,14 -0.6305 5.246i Steering oscillation
subsidences in yaw, in accordance with the discussion of section 3.3. At higher
speeds these combine to form heavily damped oscillations D and E. (b) two conju-
gate complex pairs at the kinematic frequency of the wheelsets labelled branches A
and B which are initially proportional to speed and which, at low speeds, correspond
to the steering and shear oscillations considered in Section 4.5. The real parts of
these roots are small and may be positive or negative depending on the speed being
considered. (c) three conjugate complex pairs C1, C2 and C3 associated with the
vehicle car body modes in which the frequencies are substantially independent of
speed and are closely equal to those of the car body oscillating on the suspension
with wheels fixed, and are therefore the solution of the following equations
{ (I
zb
/h
2
)s
2
+ 2k
y
+ 2k
/h
2
}
3
= 0
( m
b
s
2
+ 2 k
y
)
6
- 2dk
y
7
= 0 (1)
- 2d k
y
6
+ (I
xb
s
2
+ 2d
2
k
y
+ 2k
)
7
= 0
The first equation refers to yaw of the car body and the undamped natural frequency
is given by

2
= (2k
y
h
2
+ 2k
)/I
zb
(2)
The last two equations yield for the undamped frequencies
and
of the coupled
lateral translation and roll motions the characteristic equation

4
(
2
+
2
)
2
+ 4k
y
k
/m
b
I
zb
= 0 (3)
where
Table 5.1 Eigenvalues for the two-axle vehicle of Table 4.1 (seconds
-1
).
Figure 5.4 Root locus for two-axle vehicle with parameters of Table 4.1 except that c
y
= 0,
c

= 0, k
y
= 0.5 MNm.

2
= 2k
y
/m
b
(4)

2
= (2d
2
k
y
+ 2k
)}/I
xb
(5)
The height of the roll centre associated with each undamped natural mode, measured
below the plane of the lateral springs follows from the second of equations (1) and is
z
1
= d
2
/(
2
) (6)
z
2
= d
2
/(
2
) (7)
and in the usual case
>
>
and
>

so that z
I
> 0 and z
II
< 0. Mode I
is labelled lower sway and mode II upper sway. Obviously, introduction of sus-
remains valid.
Reference to Figure 5.4 shows that, at low speeds, the damping of each oscilla-
tory mode is positive. As the suspension damping is zero for this example, the
damping in the car body modes must come from the creep forces and is attributable
to the interaction between the motion of the wheelsets and the motion of the car
body. As the speed is increased, the damping of each mode becomes negative indi-
cating instability. However, the vehicle is unstable at all speeds above V
B1
the lowest
at which the damping vanishes.
As indicated in Figure 5.4, damping is lost in the modes associated with the car
body, C1, C2 and C3, at speeds V
B1
, V
B2
and V
B3
when the kinematic frequency of a
wheelset is rather close to a natural frequency of the car body on the suspension.
This mode of instability may be referred to as body instability as the modes of oscil-
lation of the car body are excited by the wheelset oscillation, and the resulting body
motions in turn sustain the wheelset oscillation, thus closing the feedback loop. This
is illustrated in Figure 5.5 where the representation of the equations of motion as a
block diagram indicates several feedback loops involving the car body freedoms, in
addition to those inherent in the wheelsets as shown in Figure 2.19.
It can be seen that at high speeds the branches W1 and W2 are similar to those
for the elastically restrained wheelset in which the car body is prevented from lateral
movement. The good agreement might be expected as the kinematic frequency is
then much greater than the frequency of the car body oscillating on the suspension.
At higher speeds, the combined effect of speed, stiffness and inertia modifies the
hk
d
hk
d
k
y
k
y
k
/h
k
/h k
y
k
y
3
wheelset
wheelset
body yaw
body l.t.
Figure 5.5 Block diagram representation of equations of motion. Dotted lines relate to variant
with inter-wheelset shear stiffness.
pension damping complicates the detail but usually the description of the modes
will vanish at speeds V
B4
and V
B5
corresponding to wheelset instabilities.
It will be noted that in Figure 5.4 the root loci corresponding to the kinematic
branches and car body suspension branches either veer or cross-over. When they
veer the eigenvectors associated with the eigenvalues on each branch of the loci be-
fore veering are interchanged during veering in a rapid but continuous way [20].
Consider the dependence of stability on the plan-view suspension stiffnesses k
y
and k
. In accordance with the discussion of Section 4.5, if either k

y
or k
are zero,
then V
B1
is zero. Similarly, if both k
y
or k
are infinite, then the solution of equations

(2.15 - 21) indicate that V
B1
is zero. For this completely rigid vehicle, the solution of
the characteristic equation consists of a pair of purely imaginary roots
s
1,2
= iV{(/lr
0
) /(1 + h
2
/l
2
)}
1/2
(8)
The frequency of the steering oscillation for a rigid vehicle given by equation (8)
was first derived by Carter [21]. For zero flexibility the damping of the steering os-
cillation is zero but unlike the superficially similar kinematic oscillation of a wheel-
set it is not a motion without creep, as can be easily be verified. The corresponding
solution of the equations including the contact stiffness
k
c
= (2
0
/ l)( 1 f
23
/
0
r
0
) (9)
given by (2.10.45) yields
V
k l h l h r
l m l h m l h I I
B
c
b b
2
2 2 2 2
0
2 2 2 2
2 2
2 2 2
=
+ +
+ + + + +
( )( )
{ ( ) ( ) }
(10)
which is ordinarily very small.
Relatively large values of V
B1
can be achieved for the case where both k

and k
y
are varied, neither stiffnesses being small. For example, if k = k
y
= k

is varied from
zero to the bifurcation speed V
B1
rises from zero (as already discussed), reaches a
maximum and then decreases to the low value given by (10). The maximum value
of V
B1
occurs when flexibility parameters similar to those which figured in the dis-
cussion of curving behaviour in Section 4.3 are of the order of unity; thus
=
f
2
l
3
/k
2
r
0
~ 1 and
y
= f
2
l
3
/k
y
2
r
0
~ 1. In practical terms, these parameters might be
thought of as characterising the behaviour of the system when the terms representing
creep forces in the equations of motion are commensurate with those representing
elastic forces. The vehicle and its behaviour can be distinguished as being stiff
when both
1 and
y
1, and is typically descriptive of an isolated short wheel-
base bogie.
If either k
y
or k
is small, then the bifurcation speeds are low. In this case of the
flexible vehicle one of the flexibility parameters is large; thus
> 1 and/or
y
>
1. In practical terms, this might be thought of as characterising the behaviour of a
long wheelbase two-axle vehicle.
damping in the kinematic oscillations, as discussed in section 3.4, and the damping
In order to gain insight into the behaviour of the flexible two-axle vehicle it is useful
to consider the case where the wheelbase is very long [22]. As already discussed,
examination of equations (2.15-21) shows that, for long wheelbase, the choice of the
sum and difference coordinates leads to two sets of equations which are uncoupled.
Furthermore, if d = 0 or the roll stiffness k
is very large, the car body roll freedom

is uncoupled or suppressed. (The influence of car body roll is considered later).
Then the two remaining sets of equations differ only in the value of A
33
; moreover,
if
zb
/h
2
= m
b
, often a good approximation, they are identical. The solutions of the
first set of equations then correspond to the steering oscillation A which is only cou-
pled with yaw of the body whilst the solutions of the second set of equations corre-
spond to the shear oscillation B which is only coupled to lateral translation of the
body. These steering and shear oscillations have already been discussed for the low
speed case in Section 4.5. In order to investigate dynamic stability one set of equa-
tions can be considered as the second set will behave similarly. Hence consider
ms f V c s k f k c s
f l r Is fl s V k
k c s m s c s k
y y y y
y y b y y
2
0
2 2
2
1
2
3
2 2 2 2
2 2 0
2 2 0 2 4 4
0
+ + +
+ +
+ +
=
( / )
/ /
(1)
In this Section, the case in which c
y
= 0 will be considered. The characteristic equa-
tion of (1) is
p
6
s
6
+ p
5
s
5
+ p
4
s
4
+ p
3
s
3
+ p
2
s
2
+ p
1
s+ p
0
= 0 (2)
where
p
6
= mm
b
I /4
p
5
= m
b
f(ml
2
+ I)/2V
p
4
= m
b
f
2
l
2
/V
2
+Ik
y
(m + m
b
/2)/2 + mm
b
k
/4
p
3
= fk
y
{(m + m
b
/2)l
2
+ I}/V + fm
b
k
/2V
p
2
= 2k
y
f
2
l
2
/V
2
+ k
y
k
(m + m
b
/2)/2 + f
2
m
b
l/r
0
p
1
= fk
y
k
/V
p
0
= 2f
2
k
y
l/r
0
An approximation to the root locus can be obtained by applying the perturbation
procedure of Section 3.3. In the characteristic polynomial, equation (2), the creep
5.4 The Flexible Vehicle with Zero Suspension Damping
easily be seen that if only the terms in f
2
are retained equation (2) reduces to
( m
b
s
2
+ 2 k
y
)( s
2
/V
2
+ l/r
0
) = 0 (3)
An approximation to the eigenvalues of equation (2) can be derived by considering
that the terms of first order in f are small perturbations of the terms retained in equa-
tion (3) so that applying equation (3.3.12) it is found that
1
= iV(/r
0
l)
1/2
- V{k
y
l
2
+ k

- ( ml
2
+ I )(/r
0
l)V
2
- 2k
y
2
l
2
/(-m
b
(/r
0
l)V
2
+ 2k
y
)}/4fl
2
(4)

2
= i (2k
y
/m
b
)
1/2
- V{k
y
2
/(-m
b
(/r
0
l)V
2
+ 2k
y
)}/f (5)
These results show that, to this approximation, the real parts of the eigenvalues are
inversely proportional to the creep coefficient; at low speeds they are proportional to
the suspension stiffness. The damping becomes negative at a speed below that at
which frequency coincidence between the kinematic frequency and the suspension
frequency occurs. This is the body instability discussed above.
In equation (4) the eigenvalue for the elastically restrained wheelset is obtained
if m
b
is made very large and equation (4) therefore predicts the wheelset instability
at higher speeds.
The exact solutions of equation (1) for the critical conditions in which purely
sinusoidal oscillations are possible, thus defining the stability boundary for small
displacements, may be obtained by substituting s = i. Expanding equation (1) and
separating real and imaginary parts, results in
m
b
(m + I/l
2
)
4
- {2k
y
( m + I/l
2
) + m
b
( k
y
+ k
/l
2
)
2
+ 2k
y
k
/l
2
= 0 (6)
V
2
= 4f
2
l
4
2
/{ 4f
2
l
3
/r
0
- ( k

2
)
2
} (7)
Equation (6) is independent of speed and creep coefficient f and always has two real
roots in
2
. The corresponding bifurcation speeds can then be found by substitution
in (7). It is clear that for k
<< 2fl(l/r
0
)
1/2
instability occurs at the kinematic fre-
quency. For sufficiently large values of k
> 2fl(l/r
0
)
1/2
the solutions of (6) and (7)
show that stability exists at all speeds; however, in this case it is no longer realistic
to assume that the wheelbase is long, and the full set of coupled equations must be
considered. If, as is usual, the body mass is much greater than the wheelset mass
relevant simple analytical solutions can be obtained by considering two limiting
conditions. If m
b
is large, equations (6) and (7) yield the solution for the elastically
restrained wheelset discussed in Section 4.2. If wheelset inertia is neglected so that
m = I = 0, then
V
2
= 2k
y
k
r
0
l/m
b
( k
y
l
2
+ k
)( 1 - k
2
r
0
/4f
2
l
3
) (8)
coefficients are numerically large compared with the other coefficients and it can
or, for k
<< 2fl(l/r
0
)
1/2
V
2
= 2k
y
k
r
0
l/m
b
( k
y
l
2
+ k
) (9)
yielding the bifurcation speed of the body instability. A similar analysis to the above
but including the contact stiffness terms k
yc
and k
c
shows that the bifurcation speed
is nonzero when k
y
and k
are zero but is very low, [22] and for all practical values
of k
y
and k
the contact stiffnesses may be neglected.

Numerical solutions of the complete system of equations of motion (2.15-21) are
displayed in Figures 5.6 and 5.7. Figure 5.6 shows the influence of k
y
on stability.
V
B1
increases from zero (or a very small value if the small terms in the equations of
motion are retained) to a maximum in accordance with the analysis given above.
Figure 5.7 shows the joint influence of k
and wheelbase h on stability. These results

indicate that the strongly stabilising effect of increasing yaw stiffness predicted by
equation (8) is only achieved for very long wheelbases as indicated at A in Figure
5.7. Two (or three if roll instability occurs) forms of body instability occur for small
values of k
in accordance with the above discussion of the long wheelbase case. As

k
is increased, the instability involving one mode of oscillation disappears, but the
instability involving the other mode occurs for all values of k
. As the wheelbase is
reduced, the stabilising effect of k
is offset by the destabilising effect of the cou-

pling terms k
/h. For larger values of k
, V
B1
is independent of k

, as shown
at 1, 2 and 3 in Figure 5.7, the coupling terms k
/h become dominant and in-

spection of the equations of motion (2.15-21) shows that an undamped kinematic
oscillation of the vehicle is possible where y
1
= y
2
,
1
=
2
=
b
and y
b
= 0. The rela-
tionship between y
1
and
1
, and y
2
and
2
is that of the kinematic oscillation, equa-
tion (3.2.12), and the frequency is given by

2
= 2h
2
k
y
/I
zb
(10)
Equating this to equation (3.2.11) yields the bifurcation speed
V
B1
2
= 2h
2
k
y
r
0
l/I
zb
(11)
and this is in close agreement with the results shown in Figure 5.7 which also shows
the associated mode shape.
For the ternary sub-system discussed above, as in the case of the elastically re-
strained wheelset, it was found that reductions in the value of the creep coefficients
are stabilising. However, for the complete vehicle, in the absence of suspension
damping, owing to the destabilising effect of the coupling terms k
/h, it is found that

the effect on V
B
of varying a factor on the creep coefficients f
11
, f
22
and f
23
is not sig-
nificant unless the wheelbase is very long.
It has been assumed in the above discussion that d = 0. An easy extension of the
above analysis would show that there are three possible modes of body instability,
two involving the upper and lower sway modes of the car body, and the other in-
volving yaw of the car body.
1 2
0
20
40
k
y
(MN/m)
V
(m/s)
Figure 5.7 (a)Variation of V
B1
with yaw suspension stiffness k
. Parameters of Table 5.1

except that c
y
= 0; c
=0; (b) mode shape of oscillation showing one half-cycle at bi-

furcation speed V
B1
for k
= 20 MNm and h = 10 m.
10 20
0
20
40
k
(MNm)
V
(m/s)
h = 4
h = 7
h = 10
A
B1
B2
B3
(a)
(b)
Figure 5.6 Variation of V
B1
with lateral suspension stiffness k
y
. Parameters of Table 5.1
except that c
y
= 0; c
=0;
The above analysis clearly indicates the mechanism of the body instabilities of the
two-axle vehicle. It is this form of instability that two-axle vehicles frequently exhib-
ited at quite low speeds [23] though detailed examination of typical designs requires
consideration of many significant suspension non-linearities [24]. The results of this
Section indicate that improvement of the stability of the flexible vehicle requires the
introduction of suspension damping and this is discussed in the next Section.
5.5 Damping and the Long Wheelbase Flexible Vehicle
Consider now the effect on the eigenvalues of varying the suspension damping c
y
indicated in Figure 5.8 for the sub-system of equations (4.1). Figure 5.8(a) corre-
sponds to c
y
= 0; the root locus has two branches A and B. Branch A starts from the
origin when V = 0 and initially has damping proportional to speed. As the speed is
increased branch A diverges so as to cross the = 0 axis at V
B1
. With further in-
creases in speed, the locus converges on point C corresponding to the natural fre-
quency of the car body on the suspension. Branch B starts from point C at low
speeds, initially has constant frequency and then converges on the root locus of the
elastically restrained wheelset. The introduction of c
y
can have the results shown in
Figure 5.8(b), (c) and (d), resulting, under certain conditions, in the elimination of
the body instability. These conditions are now investigated.

Figure 5.8 Influence of lateral suspension damping c
y
on the root locus as speed is varied
for the sub-system of equations (4.1). Parameters of Table 4.1 except c
= 0; k

= 0.5
MNm; (a) c
y
= 0 (b) c
y
= 3500 Ns/m (c) c
y
= 8000 Ns/m (d) c
y
= 17000 Ns/m
Referring to the equations of motion (4.1) in which the terms in c
y
are included,
2
can be eliminated using the second of equations (4.1) so that, making the assump-
tion that the wheelset inertia is negligible, and assuming harmonic oscillations at
frequency

2
= -
1
fl( k
/2 - fl
2
i/V)/ r
0
(f
2
l
4
2
/V
2
+ k
2
/4) (1)
Assume, consistent with the assumption of suspension flexibility, that f
2
l
4
2
/V
2
>>
k
2
/4. Then (1) reduces to

2
= - V
2
1
( k
/2 - fl
2
i/V)/ fl
3
2
r
0
(2)
3
can be eliminated using the third of equations (4.1) , so that on separating real and
imaginary parts the first of equations (4.1) becomes
V
2
k
/
2
l
3
r
0
m
b
2
{ k
y
( 2k
y
- m
b
2
) + 2c
y
2
2
}/
{(2k
y
- m
b
2
)
2
+ 4c
y
2
2
} = 0 (3)
2f/V -2Vf/
2
r
0
l+ m
b
c
y
4
/{( 2k
y
- m
b
2
)
2
+ 4c
y
2
2
} = 0 (4)
From (4) it can be seen that, for f large, if harmonic oscillations can occur they will
be at kinematic frequency, = V(/r
0
l)
1/2
. It was assumed in the above discussion
that f
2
l
4
2
/V
2
>> k
2
/4. If oscillations occur at kinematic frequency this condition
reduces to k
<< 2fl(l/r
0
)
1/2
. Equation (3) is a quadratic in and has no real roots if
k
/k
y
l
2
> 1/4(1- ) (5)
where = c
y
/(2m
b
k
y
)
1/2
. Thus if this condition is satisfied, body instability is elimi-
nated at all speeds. A necessary condition that (5) is satisfied is that k

> k
y
l
2
. It can
be seen that the criterion suggests that stability up to high speeds can be obtained by
choosing a sufficiently small lateral suspension stiffness, a sufficiently large yaw
stiffness (but not so large as to negate the assumption of long wheelbase, see below),
and an optimum value of lateral suspension damping which must be neither too
large nor too small. Thus, if the suspension parameters are chosen appropriately, so
that the body instabilities are eliminated, stability at speed is then only limited by the
wheelset instability and the critical speed is given by equation (3.3.16) and can be
made very high.
The destabilising effect of wheelset inertia can be seen by extending the analysis
leading to equation (5). Retaining the wheelset mass and yaw inertia terms in the
first and second equations of equations (4.1) and noting that instability occurs when
2
= 2k
y
/m
b
, roughly, yields
k
/k
y
l
2
> 1/4(1- ) + 4m/m
b
(6)
which suggests that stability can only be achieved if the wheelset mass is rather less
than the car body mass.
As mentioned above, in reality, the car body is able to roll, and it is found that
each of the three modes of body instability (corresponding to the three wheels-fixed
natural modes of the car body on the suspension) can be suppressed in turn by an
appropriate choice of the parameters as discussed above.
5.6 Stability of the Flexible Vehicle in General
Turning to the case of the complete vehicle in which the assumption of a long
wheelbase is no longer justified, Figure 5.9 shows the effect on stability of varying
the lateral suspension stiffness k
y
, and Figure 5.10 shows the limit cycle diagrams
for three values of k
y
for the example vehicle of Table 5.1. If the lateral suspension
damping c
y
= 0, the stability boundary would correspond to the body instability. The
linear bifurcation speed V
B
is closely given by (4.9), and is low as it roughly corre-
sponds to the speed for frequency coincidence between one of the wheels- fixed car
body modes and the wheelset kinematic mode. For certain values of k
y
and c
y
Fig-
ures 5.9 and 5.10(a) show that, for small values of k
y
, body hunting is eliminated
and the stability boundary then corresponds to wheelset hunting. Figure 5.10(a) cor-
responds to point A in Figure 5.9 and may be compared with the limit cycle dia-
gram for the elastically restrained wheelset, Figure 3.6. V
B
is then closely given by
(3.3.16). Figure 5.10(b) corresponds to point B in Figure 5.9. Body hunting occurs
for a range of low speeds, and wheelset hunting occurs beyond a higher speed. Fig-
ure 5.10(c) corresponds to point C in Figure 5.9 and shows that as k
y
is increased
beyond a certain value, body hunting occurs at a relatively low speed, and for large
values of k
y
, V
B
is given by (4.9). Figures 5.10(d)-(f) show the corresponding limit
cycle diagrams for a reduced coefficient of friction. Limit cycle amplitudes are then
restricted but hunting occurs at all speeds above V
B
in each case.
The influence of c
y
is shown in Figure 5.11. For large values of c
y
there is instab-
3 6
40
100
140
k
y
(MN/m)
V
(m/s)
0
V
B
V
C
A B C
(a)
3 6
40
100
140
k
y
(MN/m)
V
(m/s)
a = 0 mm
3
5
7
0
(b)
Figure 5.9 Stability of two-axle vehicle with parameters of Table 4.1 as a function of k
y
. (a)
critical speed V
c
and bifurcation speed V
B
. (b) stability boundaries for various assumed
amplitudes using equivalent linearisation.
ility at all speeds above the critical speed for body instability, which is close in value
to that for c
y
= 0. For certain intermediate values of k
y
, several separate regions of
instability can occur, depending on the values of the system parameters. The
stabilising effect of c
y
depends on the values of k
y
and k
, as illustrated in figure
5.11. If k
< k
y
l
2
, as in Figure 5.11(a) then no value of c
y
eliminates body instability.
As k
is progressively increased, as in Figures 5.11(b), (c) and (d) the region of

stability is increased and body instability only occurs for extreme values of c
y
. For
very large values of c
y
the mass of the body will be coupled in with the wheelset
mass and the wheelset instability will occur at low speeds. Figure 5.12 illustrates, on
a comparative basis with Figure 5.11, the destabilising effect of halving the mass
and moments of inertia of the car body.
Experimental confirmation of these results was obtained using the full-scale
variable parameter test vehicle HSFV-1 which was designed in accordance with the
damped-flexible approach described above. Roller rig experiments were carried
out with this vehicle, and the detailed agreement between theory and experiment is
presented in [25]. It was shown that, qualitatively, agreement was excellent, but
clearly refinements to the theory to take account of the influence of the rollers are
necessary. Such refinements have been considered by a number of writers [26].
(a)
140
0
8
V (m/s)
a (mm)
(b)
(c)
(d)
(e)
(f)
0
Figure 5.10 Limit cycle diagrams for two-axle vehicle with the parameters of Table 4.1 for
the three values of k
y
indicated in Figure 5.9 (a) k
y
= 0.23 MN/m and = 0.3 (b) k
y
= 2.5
MN/m and = 0.3 (c) k
y
= 5 MN/m and = 0.3 (d) k
y
= 0.23 MN/m and = 0.1 (e) k
y
=
2.5 MN/m and = 0.1 (f) k
y
= 5 MN/m and = 0.1.
Following laboratory tests, HSFV-1 was track tested where in spite of the unknown
and variable values of conicity and creep coefficients tolerable agreement was ob-
tained.
Figure 5.13 shows the joint influence of k
and wheelbase h on stability, for the

case where the suspension dampings c
and c
have their nominal values. Figure

5.13(a) shows the stability boundaries for the reduced wheelbase h = 3.3 m. In this
case, the body instability involving the bending mode is eliminated for a range of
intermediate values of k
. Figure 5.13(b) shows similar results for a smaller wheel-

base of h = 2.7 m and in this case no value of k
will eliminate the body instability.

The boundaries for the wheelset instability are also to be seen in Figures 5.13(a) and
(b). As k
becomes large, as discussed in Section 4, examination of the equations of

motion (2.15-21), shows that both wheelsets yaw with the car body and so the effec-
tive inertia of the combined mode is large. In addition to reducing the bifurcation
speed, the stabilising effect of lateral suspension damping is reduced.
In design applications, the emphasis is on the choice of suspension and other pa-
rameters which will provide stability under all track conditions. It is a common
practice to consider a range of values of equivalent conicity and creep coefficient
(often neglecting the stabilising influence of the contact stiffness) and apply a quasi-
linear analysis. In effect, the dependence on amplitude is replaced by dependence
on equivalent conicity. Figure 5.14 shows the effect on the stability boundaries of
varying the equivalent conicity. As expected, reduction in conicity increases the
critical speeds but also reduces the range of k
for which the body instability is elimin-

c
y
V
k
< k
y
l
2
S
S
S
S
U
U
U
U
k
= k
y
l
2
k
> k
y
l
2
k
>> k
y
l
2
S
V
V
V
c
y
c
y
c
y
Figure 5.11 Qualitative diagram showing the joint influence of k

and c
y
on stability. U =
unstable; S = stable.
ated. Thus, it can be seen that body instability is promoted by low conicity, short
wheelbase and either low or high yaw stiffness.
It can be inferred from these results that for certain parameter combinations in-
creasing conicity is actually stabilising. This is also indicated in Figure 5.15 which
shows the influence of reductions in the magnitude of the creep coefficients and
variations of conicity. In Figure 5.15(a) and (b) where the parameters relate to the
two-axle vehicle of Table 4.1 reduction in the creep coefficients is stabilising, the
c
y
V
k
< k
y
l
2
k
= k
y
l
2
k
>> k
y
l
2
k
> k
y
l
2
V
V
V
c
y
c
y
c
y
U
U
U
U
S
S
S
S
S
S
Figure 5.12 Qualitative diagram showing the effect of halving the car body mass and mo-
ments of inertia on the joint influence of k
and c
y
on stability. U = unstable; S = stable.
k
V
h = 3.3 m h = 2.7 m
k
V
U
U
U
U
U
U
S
S
S
(a) (b)
and h on stability. U =
unstable. S = stable.
k
V
= 0.4 = 0.15
= 0.1 = 0.05
U
U
U
U
U
U
U
S
S
S
S
S
U
V
V
V
k
and . U = unstable; S =
stable.
0 0.5 1
V
= 0.3 = 0.05
= 0.3 = 0.05
A
B
C
D
G
F
E
(c) (d)
(a) (b)
U
U
U
U
U
S
S
S
S
S
V
V
V

Figure 5.15 Stability charts showing the joint influence of reducing the creep coefficients f
11
,
f
22
and f
23
by a factor F and conicity (a) and (b) parameters of Table 4.1 except c
y
=
8kNs/m (c) and (d) parameters of Table 4.1 except k
y
= 2 MN/m, k
= 6 MNm and h =
3.3 m. U = unstable, S = stable.
k
p
V
= 0.4 = 0.15
= 0.1 = 0.05
U
U
U
U
S
S
S
S
V
V
V
k
p
k
p
k
p
p
and , when k
= 0. This is
drawn on a strictly comparitive basis as Figure 5.14 and shows the elimination of the
yaw body instability for most values of k
p
. U = unstable; S = stable.
k
y
V
c
y
k
r
c
r
U
U
U
U
U
U
S
S
S
S
Figure 5.17 Qualitative stability diagrams for two-axle vehicle with yaw dampers between
wheelsets and car body with k

= 0.
critical speed associated with the wheelset instability rising broadly in accordance
with the discussion of Section 3.4, as shown at A and B. In the high conicity case,
the body instability is eliminated as shown at C, but in the low conicity case the
body instability occurs over a wider range of speeds, as shown at D. In Figure
5.15(c) and (d) the parameters relate to a two-axle vehicle with the parameters of
Table 4.1 except that k
= 6 MNm and k
y
= 2 MN/m. Consistent with Figure 5.14,
body instability occurs for = 0.05 and reduction in the magnitude of the creep co-
efficients is destabilising as shown at E. However, for the high conicity case, the
body instability is eliminated for the nominal values of the creep coefficients, but
reappears when they are reduced, as shown at F. The critical speed of the wheelset
instability increases with reduction of the creep coefficients as shown at G.
5.7 The Application of Cross-Bracing and Yaw Relaxation
Reducing the wheelbase is strongly destabilising for the conventional configuration
of two-axle vehicle, and this has been traced to the coupling term (k
/h) between
2
and
5
in the equations of motion, equations (2.15-21). However, consider the appli-
cation of cross bracing to the flexible two axle vehicle so that the wheelsets are con-
nected by parallel and diagonal springs of stiffness k
p
and k
d
respectively. The equa-
tions of motion are given by (2.27-33). For clarity, the yaw stiffness k
will be
equated to zero, and then from (2.25-26) the shear stiffness k
s
= k
d
, and the bending
stiffness k
b
= k
p
.
Figure 5.16 shows the stability boundaries in the plane of V and k
p
for the case
where k
= 0 and h = 3 m. These diagrams should be compared with Figure 5.14

and, as k
p
provides a stiffness to ground without the coupling in yaw between the
wheelsets and car body there is no body instability, and the boundary shown corre-
sponds to the wheelset instability. The critical speed is quite insensitive to reduc-
tions in wheelbase, and reductions in conicity simply increase the critical speed in
accordance with (3.3.16).
The modifications to the equations of motion necessary to cover the application
of primary yaw dampers have been discussed in Section 2. A repetition of the analy-
sis in Section 4 for the long wheelbase vehicle shows that the eigenvalue corre-
sponding to the kinematic oscillation, equation (4.4), is now given by
1
= iV(/r
0
l)
1/2
- V{k
y
l
2
+ (k
r
c
r
2
V
2
)/(c
r
2
V
2
+ k
r
2
r
0
l)
- ( ml
2
+ I )(/r
0
l)V
2
- 2k
y
2
l
2
/(-m
b
(/r
0
l)V
2
+ 2k
y
)}/4fl
2
(7)
and this becomes at low speeds

1
= iV(/r
0
l)
1/2
- V
3
( c
r
2
/ k
r
- ml
2
- I -m
b
l
2
/2)/4fr
0
l
3
(8)
as at low frequencies the effective stiffness is approximately
2
c
r
2
/ k
r
. The analysis
of this simple system suggests that for small values of c
r
the effective stiffness will
be inadequate to overcome the destabilising effect of the inertia forces. Similarly,
very large values of k
r
will be destabilising, as body yaw instability is introduced.
This is illustrated by Figure 5.17 which shows the stability boundaries for a two-axle
vehicle with yaw dampers connecting the wheelsets to the car body, instead of yaw
springs, as a function of the principal parameters. With zero yaw stiffness, the stabil-
ity boundaries as k
y
is varied are broadly similar to those of the standard vehicle
shown in Figures 5.9 and 5.11. Provided k
y
is sufficiently small, stability is only
limited by the occurrence of wheelset instability at high speed. In this case the
stability boundary is given by (3.3.16) with k

replaced by k
r
.
As discussed in Section 2 an alternative arrangement is to provide the equivalent
of a yaw relaxation spring directly between the wheelsets, and as the arrangement
provides a stiffness to ground without the coupling in yaw between the wheelsets
and car body there is improved stability.
5.8 The Stiff Vehicle or Bogie
For the stiff vehicle, it is found that suspension damping has little effect on stability
and consequently, in this Section, suspension damping will be assumed to be zero.
To start with, the dynamics of the stiff vehicle are considered at low speeds. The
completely rigid vehicle has already been discussed in Section 3. The equations of
motion of a two-axle vehicle at low speeds were derived in Section 4.5. Inspection
of the characteristic polynomial (4.5.4) shows that it is a symmetric function of k
b
and k
s
(h
2
+ l
2
). In general, its eigenvalues consist of two real roots
s
1
= -k
b
V/fl
2
(1)
s
2
= - k
s
V(h
2
+ l
2
)/fl
2
(2)
corresponding to subsidences in pure bending and pure shear respectively, and a
conjugate complex pair
s
i
s
corresponding to the steering oscillation. For certain
impracticably large values of the wheelbase all the roots of the characteristic equa-
tion are real. Numerical solutions, shown in Figure 5.18, show that for small values
of k
s
and k
b
the eigenvalues are consistent with the discussion of Section 4.5. For
intermediate values of k
s
and k
b
, and ordinary values of the wheelbase h, the real part
of the eigenvalue
s
has absolute maxima for k
s
= (occurring for k
b
= fl(l/r
0
)
1/2
)
and k
b
= (occurring for k
s
(l
2
+ h
2
) = fl(l/r
0
)
1/2
) and a saddle point for k
s
(l
2
+ h
2
)
= k
b
. For either k
s
= or k
b
= , and ordinary values of the wheelbase, the maximum
value of the real part of the eigenvalue
s
increases as wheelbase is increased. Whit-
man [27] applied a perturbation procedure to establish expressions for
s
and
s
.
Similar results may be established by applying the perturbation technique of Section
3.3 to equation (4.5.4). If h = 0, a solution of equation (4.5.4) is
s = i
1
= iV( /r
0
l)
1/
(3)
The condition in which the wheelbase is non-zero may now be considered by
regarding the terms in h
2
as small perturbations p
n
of the terms p
n
that exist when
h = 0. Equation (3.3.12) then gives the corresponding perturbation in the eigenvalue
s
s
=
s
i
s
provided that there are no repeated roots, which is the case for the sys-
tem under consideration. This yields as rough, but useful, approximations

s
= -V(l/r
0
)fh
2
k
e
/ 2(l
2
+ h
2
){f
2
l
2
(l/r
0
) + k
e
2
} (4)
where k
e
is a weighted series combination of the stiffnesses
1/k
e
= 1/k
s
(l
2
+ h
2
) + 1/k
b
(5)
and

s
= V(/lr
0
)
1/2
[1 - h
2
k
e
2
/2 (l
2
+ h
2
){f
2
l
2
(l/r
0
) + k
e
2
}] (6)

Regarding the terms in parantheses as the first two terms of a binomial expansion,
(6) can be replaced by

s
= V(/lr
0
)
1/2
[1 - h
2
k
e
2
/ (l
2
+ h
2
){f
2
l
2
(l/r
0
) + k
e
2
}]
1/2
(7)
which tends to the correct limits as k
s
and k
b
are varied, so that the frequency of the
steering oscillation varies from that of the kinematic oscillation for small values of
the stiffnesses k
s
and k
b
to that of the rigid vehicle, given by equation (3.8), for large
values of the stiffnesses k
s
and k
b
.
These results give a useful qualitative impression of the behaviour of the system
though their accuracy is limited. The damping increases with wheelbase and reduces
with conicity particularly at low values of k
e
. The maximum value of
s
is
Figure 5.18 The real part of the eigenvalue corresponding to the steering oscillation plotted
in the plane of k
d
(h
2
+ l
2
) and k
p
.

smax
= - (/lr
0
)
1/2
h
2
/4l(l
2
+ h
2
) (8)
and this occurs when the flexibility parameter f
2
l/k
e
2
r
0
= 1 and is independent of
the value of the creep coefficient. Whitman [27] and de Pater [28] have given more
accurate solutions, and Scheffel [29] has given a complicated factorisation, but in all
these cases the complexity of the results is not justified by the gross assumptions
made in formulating the model.
The limiting case of small flexibility may now be considered by applying the
perturbation technique of Section 3.3 to equation (4.5.4). In this case, the stiffnesses
k
s
and k
b
in the equations of motion are much larger than the other coefficients. If k
now represents the typical stiffness k
s
and k
b
an inspection of the characteristic
equation shows that if all terms of order less than k
2
are neglected the solution for
the rigid vehicle is recovered. The condition that all terms of order less than k
2
be
neglected may now be relaxed by regarding terms of order k as small perturbations
p
n
of the terms p
n
of order k
2
. As there are no repeated roots in the case of the rigid
vehicle, equation (3.3.12) then gives the corresponding perturbation in the eigen-
value s
s
. The eigenvalue corresponding to the steering oscillation is then given by
s
s
= iV{(/lr
0
) / (1 + h
2
/l
2
)}
1/2
-Vf(l/2r
0
){h
2
/(l
2
+ h
2
)k
e
} (9)
so that the frequency is unchanged from the case of complete rigidity. Even small
flexibility introduces significant damping into the steering oscillation so that the
assumption of complete rigidity does not accord with practical reality. In contrast
with the flexible vehicle, for large values of k
e
the damping increases with f.
Table 5.2 Eigenvalues for the isolated bogie of Table 4.1 at low speed (seconds
-1
).
. V = 10 m/s Wheels Fixed Label
5,6 -16.60 189.9i -1.941 187.7i Bogie frame upper sway
7,8 -18.20 245.1i 242.0i Bogie frame yaw
9,10 -18.08 38.51i -18.06 38.60i Bogie frame lower sway
11 -44.07 Bogie shear A
12 -47.92 Bogie shear B
13,14 -0.626 3.183i Bogie steering

Now consider the influence of speed on the eigenvalues. For the typical stiff bogie,
such as one with the parameters of Table 4.1, the eigenvalues at low speed are given
in Table 5.2 and a schematic variation of the eigenvalues with speed is given in Fig-
ure 5.19. At a given speed the eigenvalues consist of (a) a relatively lightly damped
conjugate complex pair corresponding to the steering oscillation labelled branch A.
The real part of this root is negative at low speeds, and positive at high speeds, the
system losing stability at V = V
B
. (b) at low speeds, a pair of real roots, B1 and B2
corresponding to the shear and bending subsidences, equations (1) and (2). At higher
speeds these coalesce to form a heavily damped oscillation (c) three complex pairs
corresponding to the oscillations of the bogie frame on the primary suspension (not
shown in Figure 5.19) and (d) four large real roots or two complex pairs correspond-
ing to the wheelset subsidences.
Some insight into the influence of various parameters on stability is obtained by
considering the following model. As mentioned above the stiff vehicle characterises
Figure 5.19 Root locus for stiff bogie with parameters of Table 4.1.
the bogie where d, the vertical distance between the plane of the lateral springs and
centre of mass of the bogie frame, is small. If it is assumed that d = 0, the bogie
frame roll coordinate is uncoupled from the rest of the system. For a vehicle with
sufficiently stiff lateral and yaw primary suspension the natural frequencies of the
bogie frame oscillating in lateral translation or yaw on the suspension will be much
higher than the steering frequency, see for example Table 5.2. Consequently, con-
sider a model in which all the mass is concentrated at the wheelsets, and the wheel-
sets are connected by a massless suspension with shear and bending stiffness. Then,
consistent with equations (2.27-33) and (4.5.1),
{(m/2)s
2
+ fs/V + k
s
}
1
- f
2
hk
s
5
= 0 (10)
(fl/r
0
)

1
+ {(I/2)s
2
+ fl
2
s/V + k
b
}
2
= 0 (11)
{(m/2)s
2
+ fs/V }
4
- f
5
= 0 (12)
(fl/r
0
)
4
+ {(I/2)s
2
+ fl
2
s/V + h
2
k
s
}
5
hk
d
1
= 0 (13)
where, in this case, m and I are equivalent values representative of the complete bo-
gie. Making the assumption that I = ml
2
, it follows from the form of these equations
that the characteristic equation of (4.5.4) for speed V
1
, with eigenvalues
i
, will be
the same as that of (10-13) for speed V
2
, with eigenvalues s
i
, if
f
i
/V
1
= ms
i
2
/2 + fs
i
/V
2
(14)
Equation (14) can be regarded as a quadratic in s
i
. For each eigenvalue
i
of equa-
tion (4.5.4) found for low speeds there will be two roots s
i
found from equation (14)
which are roots of the characteristic equation corresponding to the equations of mo-
tion with inertia included. The combined effect of speed and inertia is therefore to
modify the damping of the steering oscillation, modify the shear and bending subsi-
dences, and to introduce the usual subsidences in lateral translation and yaw of the
wheelsets. It can be seen that in the case of a steering oscillation which is stable for
low speeds, so that the eigenvalue
i
= i where is negative, the damping will
vanish at a speed V
B
with corresponding eigenvalue s
i
= i. It can be seen that
by separating real and imaginary parts
( /V
B
) = ( /V)
0
(15)
V
B
2
= 2f(/V)
0
/m(/V)
0
2
(16)
The instability is an inertia driven instability of the steering oscillation. The mode of
instability is analogous to that of the elastically restrained wheelset in that the steer-
ing oscillation is stabilised by elastic forces generated by the primary suspension at
low speeds but at higher speeds destabilising inertia forces dominate. From (4), (6)
and (16) an estimate for the bifurcation speed (derived by Whitman [27]) may be
obtained. As ( /V)
0
<

(/lr
0
)
1/2
, a conservative estimate of V
B
is given by putting
( /V)
0
=

(/lr
0
)
1/2
resulting in
V
B
2
= 2f
2
l
2
h
2
k
e
/ m(l
2
+ h
2
){f
2
l
2
(l/r
0
) + k
e
2
} (17)
The maximum bifurcation speed obtained from (17) is
V
Bmax
2
= flh
2
k
e
/ 2m(l/r
0
)
1/2
(l
2
+ h
2
) (18)
and this occurs at k
e
= 2fl(l/r
0
)
1/2
. Figure 5.20 shows the results of a numerical ei-
genvalue analysis of equations (10)-(13) in which contours of constant bifurcation
speed V
B
are plotted in the k
s
(l
2
+ h
2
)-k
b
plane. As is evident from equations (4.5.4)
and (16), k
s
(l
2
+ h
2
) and k
b
have similar effects on stability and so the diagram has a
line of symmetry. For large values of k
s
(l
2
+ h
2
) stability is independent of k
s
and
similarly for large values of k
b
, stability is independent of k
b
. As the stiffness is in-
creased, the bifurcation speed V
B
for small oscillations increases, reaches a maxi-
mum and the decreases to zero as complete rigidity is reached. The corresponding
critical frequency is progressively reduced from that of the kinematic oscillation of a
single wheelset to that of the steering oscillation of a rigid vehicle.
The effect of small flexibility can be derived by substituting the results of (9)
into (16) yielding
V
B
2
= f
2
l
4
/m(l
2
+ h
2
)k
e
(19)
independent of the conicity . This follows from (4) as the damping in the steering
oscillation is proportional to conicity for large values of k
e
. For extreme values of
stiffness the critical speeds are low. The inclusion of the small terms in the equa-
tions, such as the contact stiffness, would be stabilising, though calculations show
that V
B
is very low.
As indicated in Figure 5.19 the steering oscillation of the example bogie loses its
stability at small amplitudes at V
B
= 63 m/s. The eigenvalues are given in Table 5.3
Table 5.3 Eigenvalues for the isolated bogie of Table 4.1 at speed V
B
(seconds
-1
).
No. V = V
B
= 63.1 m/s Label
1,2 -35.95 304.6i Wheelset yaw
3.4 -36.18 242.8i Wheelset lateral
5,6 -88.28 198.3i Bogie frame upper sway
7,8 -93.31 215.4i Bogie frame yaw
9,10 -18.57 38.19i Bogie frame lower sway
11,12 -107.0 20.94i Bogie shear
13,14 19.98i Bogie steering
and the eigenvector corresponding to the mode in which = 0 is given in Table 5.4.
This solution is a sub-critical bifurcation leading to a limit cycle with amplitude 7
mm at V=60 m/s, the details of which are shown in Figure 5.21. This shows the nu-
merical solution of the nonlinear equations of motion for the bogie with the parame-
ters of Table 4.1 and the wheel rail geometry of Figure 2.2. Figure 5.21(a) and (b)
indicate that the limit cycle is in the form of a steering oscillation very similar to that
predicted by the small amplitude solution in which y
b
lags y
1
by approximately h, y
2
lags y
b
by approximately h, and
1
,
b
and
2
roughly lag y
1
, y
b
and y
2
by 90. The
Table 5.4 Eigenvector for mode 13 of the isolated bogie of Table 4.1 at speed, V = V
B
.
5 10 15 20 25
5
10
15
20
25
k
s
(l
2
+ h
2
) (MNm/rad)
k
b
(MNm/rad)
0
55
60
65
70
75
80
85
V
B
(m/s)
Figure 5.20 Contours of constant bifurcation speed V
B
in the plane of k
s
(l
2
+ h
2
) and k
b
for a
bogie with the parameters of Table 4.1.
y
1
1

1
0.03613 + 0.2701i
y
b
0.8825 - 0.3395i

b
0.07077 - 0.07804i

b
0.08869 + 0.2718i
y
2
0.7148 - 0.6392i

2
0.05866 + 0.3142i
relative displacements between bogie frame and wheelsets are small. The frequency
of the oscillation is, however, increased as the equivalent conicity is increased. Fig-
ure 5.21(c) shows the variation of the rate of rotation of the leading wheelset which
is about 0.1% of V/r
0
. This fluctuates at twice the frequency of the oscillation as a
major component of the torque applied to the wheelset is the product of the longitu-
dinal creep force and the rolling radius both of which are varying at the frequency of
the oscillation. Figures 5.21(d) and (e) show the longitudinal and lateral creepages
for the right hand wheel of the leading wheelset. As the variation in the wheelset
rotational speed is small, reference to equation 2.5.13 shows that the longitudinal
creep of the right hand leading wheel is approximately given by r
r
/r
0
+ 1 l V
/ and
so the waveform reflects the variation of rolling radius with lateral displacement.
Similarly, reference to equation 2.5.15 shows that the lateral creep of the right hand
leading wheel is approximately given by / y V and the waveform indicates the
true departure of the response from a sinusoid. Figure 5.21(f) shows the spin of the
leading wheelset, and reference to equation 2.5.17 shows that the spin behaves like
r
/r
0
and so the waveform reflects the geometrical nonlinearity of the contact slope
variation with lateral displacement. Figures 5.21(g) and (h) shows the longitudinal
and lateral creep forces for the right hand wheel of the leading wheelset, and consid-
eration of their resultant and the normal force indicate that slipping is occurring over
a significant part of the cycle of the oscillation. Figure 5.21(i) shows the variation of
the normal force acting on the right hand wheel which is in phase with the lateral
displacement.
Figure 5.22 shows the variation of stability, as measured by the critical speed
derived by the method of describing functions, as the lateral and yaw stiffnesses
0 1
-0.01
0
0.01
t (s)
y (m)
-5
0
.005
-0.2
0
0.2
V/r
0
-0.01
0
0.01
1r1
-.002
0
.002
2r1
-1
-0.5
0
3r1
-5
0
5
T
1r1
(kN)
-5
0
5
T
2r1
(kN)
30
40
50
N
r1
(kN)
(a) (b)
(c)
(d)
(e) (f)
(g)
(h) (i)
Figure 5.21 Limit cycle of example bogie of Table 4.1, V = 60 m/s.
(assuming that k
y
= k
/l
2
) are varied. The critical speed V
C
is given by the solution of
the full linear equations of motion, for various values of the equivalent conicity and
creep coefficient. The maximum critical speed occurs at lower values of the stiff-
nesses for lower conicities, in accordance with the above discussion. For large val-
ues of the stiffnesses, the critical speed is proportional to the creep coefficient and is
largely independent of conicity.
The practical significance of these results for the conventional bogie is as fol-
lows. Because as k
y
is increased the shear stiffness is limited to a maximum value of
k
s
= k
b
/ h
2
(equation (4.2.6)) high critical speeds can more easily be achieved by
increasing yaw or bending stiffness, with consequent adverse effects on steering.
However, stability is dependent on the range of equivalent conicity and creep coeffi-
cient likely to be met in service. It can be seen from Figure 5.22 that if the typical
design specification is 0.5 > > 0.05 and f
nom
> f > f
nom
/2 where f
nom
is the value de-
rived from Kalkers analysis (Section 2.4.4.3), and reasonable margins are made for
both speed and stiffness, the maximum design speed is limited and the highest criti-
cal speeds cannot be exploited. For this reason, high speed bogies depend on re-
straint from the car body. See Chapter 6, also [30].
Equivalent levels of stability can be achieved with lower values of bending stiff-
ness by adopting cross-bracing as discussed above, which implements directly the
shear stiffness k
s
. There is then no limit in the values of shear stiffness that can be
achieved and a better trade-off between stability and curving can be obtained.

20 60 80
0
50
100
k
y
= k
/l
2
(MN/m)
V (m/s)
= 0.05
= 0.1143
= 0.05, f x 0.5 = 0.5
Figure 5.22 The joint influence of variations of conicity and creep coefficient on V
B
as the
lateral and yaw stiffnesses are varied, where k
y
= k
/l
2
.
5.9 The Three-Piece Bogie
The three-piece bogie has evolved over the last hundred years mainly for freight
vehicles in North America, but enjoys wide application throughout the world. In the
three-piece bogie, the bogie frame of a conventional bogie is replaced by two sepa-
rate side-frames which rest directly on the axle boxes through adaptors which allow
only rotational freedom. A bolster supports the car body, with a centre-plate which
allows relative yaw between bogie and body. The bolster is connected to the side-
frames by a suspension which allows some relative motion in all senses except lon-
gitudinal. Thus, the side frames have additional degrees of freedom of longitudinal
and pitching motions. This results in warping or differential rolling movements of
the wheelsets (useful on twisted track) and lozenging of the side-frames with respect
to the wheelsets in plan-view. It follows that the form of construction yields a very
low overall shear stiffness. If the warp stiffness is very large then the bogie behaves
like a conventional bogie, [31, 32]. Quite apart from the low shear stiffness of the
configuration, there are many rubbing surfaces and consequently the behaviour
tends to be highly nonlinear with the result that crabwise motions are common, with
resultant asymmetric wear of wheel profiles.
The application of cross-bracing to the three-piece bogie is particularly useful
because the anti-lozenging function of the cross-bracing allows the designer to omit
the usual constraints between the bolster and the side-frames and provide a proper
lateral suspension at that point. Moreover, if controlled flexibility is provided, by
means of a elastomeric pad mounted between the side-frames and bearing adaptors
for example, then in conjunction with cross-bracing specified suspension stiffnesses
can be provided enabling optimisation of the design. Reduced bending stiffness be-
tween the wheelsets improves curving and the improvement of the primary and sec-
ondary lateral suspension enhances ride quality and reduces loads.
Bogies incorporating these ideas were produced in the 1970's in Britain, Pollard
[33], South Africa, Scheffel [7],and in North America, List [34], though widespread
application has been retarded by the increment in first cost.
References
1. Hamilton Ellis, C.: Nineteenth Century Railway Carriages. Modern Transport
Publishing, London, 1949, p. 13.
2. White, J.H.: A History of the American Locomotive. The John Hopkins Press,
Baltimore, 1968, pp. 169-174.
3. Matsudaira, T.: in ORE Committee C9. Problems of interaction of vehicles and
track-essays awarded prizes. Report 2, Part 2, Office for Research and Experiments,
Utrecht, June 1960.
4. Wickens, A.H.: The dynamics of railway vehicles on straight track: fundamental
considerations of lateral stability. Proc. I. Mech. E. 180 (1965), Part 3F, pp. 29-44.
5. Hobbs, A.E.W.: Improvements in or relating to railway vehicles. British Patent
Specification 1261896, 1972.
6. Wolf, E.J. and Stec, F.F.: Single axle suspensions for intermodal freight cars.
Conference Proceedings: The Economics and Performance of Freight Car Trucks,
Montreal, October, 1983.
7. Scheffel, H.: A new design approach for railway vehicle suspensions. Rail Inter-
national 5 (1974), pp. 638-651.
8. Scheffel, H.: Unconventional bogie designs their practical basis and historical
background. Vehicle System Dynamics 24 (1995), No. 6-7, pp. 497-524.
9. Wickens, A.H.: British Patent 1179723, 1967.
10. Evans, J.R.: The modelling of railway passenger vehicles. In: G. Sauvage (Ed.):
The Dynamics of Vehicles on Roads and on Railway Tracks, Proc.12th IAVSD Sym-
posium, Lyon, August 1991. Swets and Zeitlinger Publishers, Lisse, 1992, pp. 144-
156.
11. Eickhoff, B.M., Evans, J.R. and Minnis. A.J.: A review of modelling methods
for railway vehicle suspension components. Vehicle System Dynamics 24, No. 6-7,
1995, pp. 469-496.
12. Sauvage, G.: Determining the characteristics of helical springs for applications
in suspensions of railway vehicles. Vehicle System Dynamics 13, 1984, pp. 13-41.
13. Wickens, A.H.: Steering and dynamic stability of railway vehicles. Vehicle Sys-
tem Dynamics 5, 1978, pp.15-46.
14. Horak, D. et al.: A comparison of the stability performance of radial and conven-
tional rail vehicle trucks. ASME J. Dynamic Systems, Measurement and Control
103, 1981, p. 181.
15. Kar, A.K., Wormley, D.N, Hedrick, J.K.: Generic rail truck characteristics. In:
H.-P. Willumeit (Ed.): The Dynamics of Vehicles on Roads and on Railway Tracks,
Proc.6th IAVSD Symposium, Berlin, September 1979. Swets and Zeitlinger Publish-
ers, Lisse, 1980, pp. 198-210.
16. Kar, A.K. and Wormley, D.N.: Generic properties and performance characteris-
tics of passenger rail vehicles. In: A.H. Wickens (Ed.): The Dynamics of Vehicles on
Roads and on Railway Tracks, Proc.7th IAVSD Symposium, Cambridge, September
17. Fujioka, T.: Generic representation of primary suspensions of rail vehicles. In
A.H. Wickens (Ed.) The Dynamics of Vehicles on Roads and on Railway Tracks,
Proc.11th IAVSD Symposium, Kingston, August 1989. Swets and Zeitlinger Pub-
lishers, Lisse, 1989, pp. 233-247.
18. Fujioka, T., Suda,Y. and Iguchi, M.: Representation of primary suspensions of
rail vehicles and performance of radial trucks. Bulletin of JSME 27 (1984), No.232,
pp. 2249-2257.
19. Hedrick, J.K. Wormley, D.N. Kim, A.K. Kar, A.K. and Baum, W.: Performance
limits of rail passenger vehicles: conventional, radial and innovative trucks. U.S.
Department of Transportation Report DOT/RSPA/DPB-50/81/28, 1982.
20. Perkins, N.C. and Mote, C.D.: Comments on curve veering in eigenvalue prob-
lems. J. Sound and Vibration 106 (1986), pp. 451-463.
21. Carter, F.W.: Railway Electric Traction. London, Edward Arnold, 1922.
22. Wickens, A.H.: The dynamic stability of a simplified four-wheeled railway ve-
hicle having profiled wheels. Int. J. Solids and Structures 1 (1965), pp. 385-406.
23. Wickens, A.H. and Gilchrist, A.O.: Railway vehicle dynamics-the emergence of a
practical theory. Council of Engineering Institutions MacRobert Award Lecture, 1977.
24. Gilchrist, A.O., Hobbs, A.E.W., King, B.L. and Washby, V.: The riding of two
particular designs of four wheeled vehicle. Proc.I.Mech.E. 180, Part 3F (1965), pp.
99-113.
25. Hobbs, A.E.W.: Lateral stability experiments with HSFV-1. British Railways
Tech. Note DYN 53, 1967.
26. Jaschinski, A. et al.: The application of roller rigs to railway vehicle dynamics.
Vehicle System Dynamics 31 (1999), pp. 345-392.
27. Whitman, A.M.: On the lateral stability of a flexible truck. ASME Journal of
Dynamic Systems, Measurement and Control 105 (1983), pp. 120-125.
28. De Pater, A.D.: The lateral behaviour of railway vehicles. CISM Courses and
lectures no.274, Springer, Vienna New York, 1982, pp. 223-279.
29. Scheffel, H.: The influence of the suspension on the hunting stability of
railway vehicles. Rail International 10 (1979), p. 662. See also: Scheffel, H.:
The dynamic stability of two railway wheelsets coupled to each other in the lateral
plane by elastic and viscous constraints. In A.H. Wickens (Ed.) The Dynamics of
Vehicles on Roads and on Railway Tracks, Proc.7th IAVSD Symposium, Cam-
bridge, September 1981. Swets and Zeitlinger Publishers, Lisse, 1982, pp. 385-400.
30. No, M., Hedrick, J.K.: High speed stability for railway vehicles considering
varying conicity and creep coefficients. Vehicle System Dynamics 13 (1984), pp.
299-313.
31. Tuten, J.M., Law, E.H. and Cooperrider, N.K.: Lateral stability of freight cars
with axles having different wheel profiles and asymmetric loading. ASME J. Engi-
neering for Industry 101 (1979), pp. 1-16.
32. Whitman, A.M. and Khaskia, A.M.: Freight car lateral dynamics-an asymptotic
sketch. ASME Trans. Journal of Dynamic Systems, Measurement and Control 106
(1984), pp. 107-113.
33. Pollard, M.G.: The development of cross-braced freight bogies. Rail Interna-
tional, September 1979, pp. 736-758.
34. List, H.A.: An evaluation of recent developments in railway truck design. ASME
Paper 71-RR-1, April 1971.
6
The Bogie Vehicle
6.1 Introduction
Long two-axle vehicles, and other vehicles with several wheelsets incorporated in a
single frame (a so-called rigid wheelbase), like locomotives, have obvious limita-
tions in curves. The first known proposal for the bogie was made in Britain by
William Chapman in 1797 [1]. It was, however, in the United States that the con-
cept was first employed to a significant extent. Dissatisfied with the performance of
the rigid wheelbase British locomotives on the lightly built and curvaceous Ameri-
can track, John B. Jervis with advice from Horatio Allen designed the first locomo-
tive with a leading swivelling bogie in 1832. In this, two wheelsets were mounted in
a frame which was free to swivel without restraint relative to the main body of the
locomotive. This radically improved curving behaviour [2]. In addition to its appli-
cation to locomotives, passenger coaches employed bogies in North America from
the 1840s. These early bogies had very short wheelbases and tended to oscillate
violently being the probable cause of many derailments. In the 1850s the wheelbase
of the leading truck of locomotives was increased which improved stability signifi-
cantly [3]. In Britain with relatively straight track there was little need for the use of
bogies until trains increased in size in the 1860s, and in any case the few occasions
when bogies had been used gave them a bad reputation with British railway engi-
neers. Fernihough pointed out the danger of bogie oscillation in his evidence before
the Gauge Commission in 1845 [4] and suggested that it might be controlled by the
frictional resistance of a bearing ring of large diameter supporting the car body on
the bogie, an idea subsequently adopted widely. The centre friction plate, or alterna-
tively, friction at the side bearers, provided yaw restraint for small relative motions
thus preventing bogie hunting on straight track at currently prevailing speeds. On
sharp curves, at low speeds, the friction was overcome and the bogie was able to
take up a more radial position on the track.
In the case of bogie passenger coaches it was appreciated that isolation of the
car body from motions of the bogie required some form of secondary suspension.
So, in addition to a secondary vertical suspension, the swing bolster which provided
lateral flexibility between bogie and car body, was invented by Davenport in 1841
[5]. In Britain, though the use of bogies in locomotives was exceptional until the
1870s, those that were built often incorporated lateral movement of the bogie pivot
restrained by some form of spring, called a centring spring [6].
In the period before adequate mathematical models became available, evolution of
the bogie vehicle had been based on rather general ideas. Much later, when calcula-
tions were attempted, the motions in the lateral plane were often assumed to be
kinematic with amplitude equal to the flange-way clearance. Hence emphasis was
given to the avoidance of coincidence between the natural frequencies of the car
body on the suspension and the kinematic frequencies. Based on these ideas, and
empirical development, surprisingly good results had been obtained before 1960,
providing that conicities were kept low by re-turning wheel treads and speeds were
moderate, for example below 160 km/h.
The discussion of Chapter 5 shows that for an isolated bogie, in which there is
no longitudinal or lateral suspension damping, for both small and large values of the
longitudinal and lateral suspension stiffnesses critical speeds are relatively low.
However, for an intermediate range of stiffnesses, the critical speed reaches a
maximum, and this maximum can be quite high. The results suggest that for many
low or moderate speed applications careful choice of the longitudinal and lateral
stiffnesses will give adequate stability without the use of secondary yaw restraint
between the bogie and the body.
With the advent of higher speeds, the consequences of bad riding became more
serious and Matsudaira [7] carried out the first mathematical modelling embracing
primary and secondary suspension stiffnesses, conicities appropriate to worn wheels
and creep. This was applied to the bogie developed for the Shin Kansen in the early
1960s which was tested to speeds of 246 km/h in 1964, and used in service
operation subsequently at speeds of 210 km/h. Generally, high speed trains still
follow a similar prescription, in which the stabilising action of the primary
suspension is augmented by the restraint offered by the secondary suspension. For
high speed trains the emphasis is on stability. For this application typical bogies are
relatively stiff with primary longitudinal stiffness in the range 30-60 MN/m reflecting
the requirement for stability at high speed. Operation is on relatively straight lines and
considerations of curving are secondary. Careful choice of bogie wheelbase, primary
suspension parameters and secondary yaw restraint provide a substantial margin of
stability, [8]. Stable operation has been demonstrated at speeds of over 500 km/h.
The models discussed here are the simplest that reveal the basic principles of
guidance and stability of the conventional bogie vehicle. In practice, for engineering
design realistic modelling must take account of much important detail [9, 10].
Many efforts have been made to improve performance in curves by making
bogie vehicles more flexible in plan view, thus permitting the axles to take up a
radial position in curves. There is no doubt that many of these configurations exhib-
ited an even wider spectrum of various hunting instabilities than more conventional
designs. Schemes for so-called body steered railway vehicles, in which provision is
made for the radial steering alignment of the wheelsets by means of linkages or levers
connecting the wheelsets with the car body, appeared early in the 19th century [11,
12]. Earlier in the 20th century significant and successful development was carried
out by Liechty [13, 14]. Schwanck [15] reported on service experience with a particu-
lar design of body-steered bogie and its advantages of reduced wheel and rail wear,
reduced energy consumption and increased safety against derailment. Many examples
of body steering are in current use.
THE BOGIE VEHICLE 175
In this Chapter, a discussion of the dynamics of the conventional bogie vehicle is
followed by a general analysis of the conflict between steering and stability of multi-
axle vehicles. This is then applied to a variety of configurations with body steered
bogies.
Consider the idealised four-axle vehicle with a body and two bogies, shown in Figure
6.1. Each bogie has a frame and two wheelsets and this sub-system has been discussed
in Chapters 4 and 5. The bogies support a car body by means of a secondary suspen-
sion, but there is no direct connection between the bogies. As in the case of the two-
axle vehicle, the wheelsets, bogie frames and car body are all assumed to be rigid and
connected by massless elastic structures and massless damper elements. The vehicle is
symmetric about a longitudinal plane of symmetry but the bogies are not assumed to
be symmetric. Therefore, for small displacements, motions which are symmetric about
this plane of symmetry are decoupled from those which are anti-symmetric and will
not be considered here. Hence in addition to the nine degrees of freedom for the
leading bogie with generalised coordinates
b
q, nine degrees of freedom for the trailing
bogie with generalised coordinates
d
q, the car body has three degrees of freedom,
lateral translation, roll and yaw with generalised coordinates
c
q.
y
1
y
b
y
c
y
d
y
2
y
3
y
4

1
4

d
b
h h
c
y
c
c
y
b
b
Figure 6.1 Generalised coordinates for bogie vehicle.
Thus the motion of the vehicle, Figure 6.1, is defined by the following set of twenty-
one generalised coordinates
q = [
b
q |
c
q |
d
q ]
T
(1)
In the discussion of stability it will usually be possible to neglect the variation of the
rotational speed of the wheelsets in which case
q = [y
1 1
y
b b b
y
2 2
y
c c c
y
3 3
y
d d d
y
4 4
]
T
(2)
so that y
b
, y
d
and y
c
refer to lateral translation of the bogie frames and vehicle body
and
b
,
d
, and
c
refer to yaw of the bogie frames and vehicle body. The other y
i
,
i
are standard wheelset coordinates. The equations of motion will be of the form of
(2.11.23) where
F = block diagonal[ F
1
O
33
F
2
O
33
F
3
O
33
F
4
] (3)
and O
33
is the 3
3 null matrix, and F

i
refers to the ith wheelset. The inertia matrix is
A = diag [m I m
b
I
xb
I
zb
m I m
c
I
xc
I
zc
m I m
b
I
xb
I
zb
m I] (4)
Figure 6.2 shows a basic form of secondary suspension for a conventional bogie
vehicle. The component stiffness and compatibility matrices are
k = diag[ k
y1
k
1
k
2
k
y2
k
2
k
2
k
yb
k
b
k
b
k
yb
k
b
k
b
k
y2
k
2
k
2
k
y1
k
1
k
2
] (5)
(6) a
h d h h
d h h h
d h d c h
=

1 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 1 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 1 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0
1 1 1
1 2 2
2 5 3 5
1 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 1 0 0
0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 1 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0
0 0 0 0 0
3 5 2 5
2 1 2
d c h d h
h d h h
0 0 0 0 0 0 0 1 1
0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1
1 1 1
+
d h h h
It should be noted that in practice the formation of E using equation (2.11.25) is
unnecessarily lengthy as there are a large number of zeros in a. Consequently, a
simpler process of assembly may be used which depends on the topology of the
connections between the bodies of the system, and which is described in any text on
the Finite Element Method.
6.3 Dynamics of the Conventional Bogie Vehicle
A set of example parameters is given in Table 6.1 for a conventional bogie vehicle in
which h
1
= h
2
= h
3
= 0 so the bogies themselves are symmetric fore-and-aft. The
example vehicle has bogies which are the same as the isolated bogie of Chapter 4 and
defined in Table 4.1. It should be noted that this vehicle has no secondary yaw re-
straint.
Figure 6.3 shows the locus of eigenvalues of the equations of motion discussed in
the previous Section as the speed is varied for the example bogie vehicle. Figure 6.3
should be compared with Figure 5.19 which gives corresponding results for the isolated
bogie. In addition to the eigenvalues corresponding to the modes already seen in the
isolated bogie there are eigenvalues approximately corresponding to motions of the car
body as if the wheelsets were fixed and which are substantially independent of speed.
h
c
h
1
h
2
h
h
5
k
y1
/2 k
y2
/2
k
yb
/2
k
2
k
1
k
b
d
1
y
b
y
c
c
d
3
d
2
k
b
k
1
k
yb
k
y1
Figure 6.2 Arrangement of bogie vehicle showing suspension stiffnesses.
primary suspension
k
y
= 40 N/m k
= 40 MNm k
=1MNm c
y
= 0 c
= 0
c

=10 kNms
secondary suspension
k
yb
= 0.45 MN/m k
b
= 0 k
b
=1 MNm c
yb
= 60 kNs c
b
= 0
c
b
= 60 kNms
creep coefficients
f
11
= 7.44 MN f
22
= 6.79 MN f
23
= 13.7 kNm
vehicle geometry
h = 1.25 m h
1
= 0 h
2
= 0 h
3
= h
4
=1.25 m h
5
= 0
c = 8.75 m d
1
= 0.2 m d
2
= 0.4 m d
3
=1 m
inertia
m =1250 kg I = 700 kgm
2
m
b
=2500 kg I
xb
=1000 kgm
2
I
zb
= 3500 kgm
2
m
c
= 22000 kg I
xc
= 30000 kgm
2
I
zc
=1 10
6
kgm
2
I
y
= 250 kgm
2
The eigenvalues of the complete system at one value of the forward speed and those
of the system with either the wheelsets fixed or the car body fixed are given in Table
6.2 where the various modes are labelled. At speeds away from that for frequency
coincidence between the car body modes and the bogie modes it is a good approxima-
tion to consider the car body fixed. It can be seen from Figure 6.3 that for the chosen
set of parameters the bifurcation speed V
B
is little changed from that for the isolated
bogie.
Table 6.1 Example parameters for conventional bogie vehicle.
Figure 6.3 Root locus as speed is varied for conventional bogie vehicle.
At speeds for which the bogie steering frequencies approach the natural frequencies
of the vehicle car body on the suspension, there is the possibility of considerable
interaction, leading to instabilities in which the amplitude of the car body is large
relative to that of the wheelsets, though for the example set of parameters such body
instabilities do not occur. Figure 6.4 shows the stability boundaries for a conventional
bogie vehicle showing V
B
as a function of the secondary yaw stiffness k
b
, secondary
lateral stiffness k
yb
, secondary yaw damping c
b
and secondary lateral damping c
yb
. A
denotes bogie instability, and B denotes body instability. For the smaller values of k
yb
V
B
is associated with the bogie instability, labelled A in Figure 6.4, discussed in
Chapter 5 and increasing k
b
is strongly stabilising. Larger values of k
b
lead to body
instability involving relatively large amplitudes of car body yaw, labelled B in Figure
6.4. This is analogous to the behaviour of the two-axle vehicle discussed in Chapter 5.
Low values of conicity and creep coefficient promote this mode of instability. Simi-
larly, k
yb
and c
yb
must not exceed certain limits if body instability is to be avoided, and
c
yb
must exceed a certain limit. It was shown in Chapter 5 that with an appropriate

No. V = 10 m/s Wheels Fixed Car Body
Fixed
Label

1 -1130 -1129 Wheelset subsidence in yaw
2 -1130 -1129 Wheelset subsidence in yaw
3 -1058 -1060 Wheelset subsidence in trans.
4 -1058 -1060 Wheelset subsidence in trans.
9,10 -25.33 190.8i -10.74 187.9i -25.22 191.1i Bogie upper sway, anti-phase
11,12 -25.36 190.5i -10.73 187.6i Bogie upper sway, in-phase
13,14 -18.20 245.1i 242.0i -18.19 245.2i Bogie yaw, anti-phase
15,16 -18.20 245.1i 242.0i Bogie yaw, in-phase
17,18 -46.55 25.42i -46.47 25.40i -46.16 27.2i Bogie lower sway, anti-phase
19,20 -46.45 24.01i -46.35 23.95i Bogie lower sway, in-phase
21 -44.09 -43.83 Bogie shear A
22 -44.09 Bogie shear A
23 -47.91 -48.02 Bogie shear B
24 -47.91 Bogie shear B
27,28 -4.304 6.96i -4.301 7.081i Car body yaw
31,32 -0.612 3.189i -0.276 3.12i Bogie steering
33,34 -0.582 3.212i Bogie steering
Table 6.2 Eigenvalues for the conventional bogie vehicle (seconds
-1
).
choice of suspension parameters and a limited range of conicity, the isolated conven-
tional bogie can possess stability up to moderately high speeds but if a wide range of
conicity and
0
100
200
1
2
100
200
1 2
100
200
1 2
100
200
0
k
yb
k
b
c
b
c
yb
0 0
V
V
V
V
A
A
A
A
B
B
B
B
Figure 6.5 Influence of secondary yaw stiffness k
b
(10
7
MNm) on critical speed V
c
(m/s) for
various values of equivalent conicity and creep coefficient for the conventional bogie ve-
hicle. S denotes the stable region.
Figure 6.4 Stability boundaries for bogie vehicle showing V
B
(m/s) as a function of the
secondary yaw stiffness k
b
(10
8
Nm), secondary lateral stiffness k
yb
(10
6
N/m), secon-
dary yaw damping c
b
(10
6
Nms) and secondary lateral damping c
yb
(10
4
Ns/m). A de-
notes bogie instability, and B denotes body instability.
4 8
0
100
200
k
b
V
c
=0.05 f/2
=0.5
=0.1143
=0.05
=0.05 f/2
S
creep coefficients are to be catered for, stability cannot be guaranteed.
Reduction in bogie mass and increases in wheelbase make a contribution to the
improvement of stability as shown in the discussion of the isolated bogie, but the
major design measure is, as discussed in the introduction, to exploit a suitably high
value of secondary yaw restraint, provided in the form of springs, dampers or relaxa-
tion dampers. Figure 6.5 shows the influence of secondary yaw stiffness k
b
on
critical speed V
c
for various values of equivalent conicity and creep coefficient for
the conventional bogie vehicle. It can be seen that the stable region of operation,
denoted by S in Figure 6.5, is much enlarged for larger values of k
b
. As noted
above, for low conicity and creep coefficients very large values of secondary yaw
stiffness are destabilising, analogous to the behaviour of the two-axle vehicle.
Figure 6.6 shows the lateral and longitudinal creep forces generated during curve
entry, for the example bogie vehicle without secondary yaw stiffness and zero cant
deficiency. The response is very similar to that of the isolated bogie shown in Figures
4.8 and 4.9 and discussed in Chapter 4. Clearly, provision of secondary yaw stiffness
in order to promote stability would degrade curving performance further, and so it is
appropriate to consider means by which the trade-off between curving and stability
might be improved.

Figure 6.6 Dynamic response of bogie vehicle to curve entry. V = 15 m/s, R
0
= 225 m,
length of transition L = 20 m.
0 2 4
-0.01
0
0.01
0.02
0 2 4
-15
-10
-5
0
x 10
-3
0 2 4
-2
-1
0
1
2
x 10
4
0 2 4
-15000
-10000
-5000
0
5000
1
b
2
4
3
d
c/10
y
c
4
d
b
1
3
2
1r
2l
4r 3l 1r
3l
4r
2l
T
1
T
2
t
6.4 Steering and Stability of Multi-Axle Vehicles in General
It has been shown in Chapter 3 that for two-axle vehicles there is a conflict between
steering and stability. The approach taken there is now generalised to vehicles with
any number of axles and it will be shown that it is possible to achieve a better
resolution of this conflict if there are more than two axles [16, 17].
Consider the elastic stiffness matrix E. For stability and steering only motions
which are anti-symmetric with respect to the longitudinal plane of symmetry are
considered, and the vehicle is assumed to move forward at constant speed. As in the
case of the two-axle vehicle, two rigid body motions are possible, lateral translation
and yaw, described by the column matrices q
l
and q
y
. No elastic forces can be
generated in a rigid body motion so that
Eq
l
= 0 (1)
Eq
y
= 0 (2)
For a system of rigid bodies with R degrees of freedom connected by M massless
elastic elements, imposition of the conditions (1) and (2) means that E will have
degeneracy P = 2 and with a suitable choice of coordinates will be of the partitioned
form
E
O O
O E
R
R
=
+
22 2 2
2 2
,
,
(3)
where E
+
is of order (R - 2). It follows that there are (R
2
- 3R + 2)/2 disposable
elements. As there can only be (R - 2) independent rows in a, the M
R compatibility
matrix defined by equation (2.11.5), the number of corresponding disposable stiffness
elements will be equal to (R - 2).The remaining (R - 3)(R - 2)/2 disposable elements
are geometrical and are contained in a. The most general choice of elastic stiffness
matrix is therefore possible if the number of physical parameters, element stiffnesses
and their positions, equals or exceeds this number. There is no limit to the number of
element stiffnesses and hence to the number of rows in the compatibility matrix a or
the order of k, the primitive stiffness matrix of the elastic elements.
If the vehicle has fore-and-aft symmetry, choice of coordinates which describe
either motions which are symmetric or anti-symmetric with respect to the plane of
symmetry results in the following partitioned form for E
where E
1
is of order S
S where S = (R-2)/2 if R is even and S = (R-3)/2 if R is odd; E

2
is of order U
U where U = (R-2)/2 if R is even and U = (R-1)/2 if R is odd. Therefore

the number of independent elements in E for a symmetric vehicle is (R
2
-2R)/4 if R is
even and is (R
2
-2R+1)/4 if R is odd.
E
O O O
O E O
O O E
S U
S SU
U US
=
22 2 2
2 1
2 2
(4)
If (2.11.25) is substituted into (1) and (2) then
kaq
l
= 0 (5)
kaq
y
= 0 (6)
An additional condition on E is provided by consideration of steering on curves.
Following the discussion on the motion of an unrestrained wheelset on a curve in
Chapter 2, assume that the lateral creep force due to spin is exactly negated by the
gravitational stiffness. Then the equilibrium of the vehicle in a curve being traversed
at equilibrium speed and hence zero cant deficiency will be determined by the
simplified form of equations of equilibrium
[ C + E ]q
c
= Q
c
(7)
where q
c
represents a column of the values of the generalised coordinates consistent
with steady state motion on a uniform curve, and Q
c
are the corresponding forces
applied by the track. Zero lateral and longitudinal creep, or perfect steering, will occur
when
Cq
c
= Q
c
(8)
and so
Eq
c
= 0 (9)
Substitution of equation (2.11.25) into equation (9) shows that for perfect steering
kaq
c
= 0 (10)
Thus, equations (1), (2) and (10) imply that the elastic stiffness matrix for a vehicle
with perfect steering will have degeneracy P > 3. Repeating the arguments used above
it is found that the number of independent elements in E is equal to (R
2
- 5R + 6)/2 for
an unsymmetric vehicle and to (R
2
- 4R + 4)/4 (R even) or to (R
2
- 4R + 3)/4 (R odd)
for a symmetric vehicle.
Dynamic stability requires that all the real parts of the eigenvalues of
| As
2
+ (B/V + D)s + C + E | = 0 (11)
are negative. As discussed in Chapters 3 and 5, in general, as speed is increased,
stability is lost at a bifurcation speed V
B
at which at least one eigenvalue becomes
purely imaginary. A necessary condition for the existence of a nonzero critical speed
is that the eigenvalues of the system at low speeds
|(B/V + D)s + C + E | = 0 (12)
all have negative real parts, as discussed in Chapter 3. If all N wheelsets are identical,
and if there were no structural connections between the various rigid bodies of the
system, E and D would be null, and equations (12) would reduce to N uncoupled sets
of binary equations where N is the number of wheelsets. There would be N identical
eigenvalues s =

iV(
/lr
0
)
1/2
corresponding to a kinematic oscillation of each of the
identical wheelsets. The elements of the corresponding eigenvectors q
k
will be
y
i
=
i i
= i
i
(
/lr
0
)
1/2
all other q
i
= 0 (13)
where the
i
are arbitrary in magnitude so that N quantities remain undetermined. Now
if
Eq
k
=0 (14)
equation (12) will be satisfied, and there will be at least one undamped mode at low
speeds involving a kinematic oscillation of the wheelsets. The condition that equation
(14) is satisfied is that P > N where P is the degeneracy of the E matrix. If P
N
then no solution corresponding to a kinematic oscillation can exist, and thus P
N is a
necessary but not sufficient criterion for dynamic stability.
A necessary condition for a vehicle with wheelsets of equal conicity to be
dynamically stable and capable of perfect steering in the case of zero cant deficiency
is therefore
3
N (15)
The case where the wheelsets have unequal conicities is discussed in Chapter 9 where
it is shown that a small margin of stability can be obtained but is generally of little
practical importance.
Now consider further the behaviour of the system at low speeds. Denoting the
generalised coordinates associated with the wheelsets by q
s
and the remaining
coordinates by q
v
so that for the bogie vehicle
q
v
= [ y
b b b
y
c c c
y
d d d
]
T
(16)
q
s
= [ y
1 1
y
2 2
y
3 3
y
4 4
]
T
(17)
the equations of motion can be written in the partitioned form
+ + +
+ + + + +
0
0
) / (
2
2
v
s
uv uv uv us us
rv rv rs rs rs rs rs
q
q
E s D s A E s D
E s D E C s D V B s A
(18)
Making the substitution s = VD and letting V tend to 0 yields the two sets of uncoupled
equations governing the motion at low speeds

[B
rs
D + C
rs
+ E
rs
- E
rv
E
uv
-1
E
us
] q
s
= 0 (19)
or
[B
rs
D + C
rs
+ E
rs
*
] q
s
= 0 (20)
where E
rs
*
= E
rs
- E
rv
E
uv
-1
E
us
is the deflated stiffness matrix involving only the
wheelset coordinates and
[ A
uv
s
2
+ D
uv
s + E
uv
] q
v
= 0 (21)
indicating that at low speeds the eigenvalues of the system fall into two distinct sets.
Firstly, a set associated with the motions of the wheelsets, modified by the influence
of the quasi-static elastic interaction between the wheelsets through the elastic
connections to the bogie frames and car body. Secondly, a set associated with the
natural modes of the vehicle body oscillating on the suspension as if the wheelsets
were fixed. This generalises the results already found for the two-axle and bogie
vehicles.
The eigenvalues of (20) are the same as those of (12) and so the condition for
stability at low speeds is dependent on E
*
. Similarly, the equations governing steady
motion in curves, equations (7), can be replaced by
[ C
rs
+ E
rs
*
] q
s
= Q
r
c
(22)
The conditions for stability and perfect curving discussed above therefore also apply
to the reduced system involving only the wheelset freedoms, and it follows that the
number of independent elements in E
*
is correspondingly reduced.
A similar approach has been followed by de Pater [18, 19] with some significantly
different details.
6.5 Steering and Stability of a Generic Bogie Vehicle
This general approach is now applied to bogie vehicles in which rather general forms
of suspension are incorporated so as to meet the dual requirements of steering and
stability. As the complete vehicle is assumed to have a transverse plane of symmetry
the structure of the stiffness matrices is simplified. Clearly, reversal of the direction of
motion should result in identical equations of motion if account is taken of the sign
convention chosen for the generalised coordinates. The complete stiffness matrix
takes the form
bb
E
bc
E
bd
E
E =
cb
E
cc
E
cd
E
(1)
db
E
dc
E
dd
E
It will be assumed that there is no direct connection between the leading and trailing
bogies, so that
bd
E and
db
E are null. The relationship between the coordinates of the
leading bogie and the trailing bogie is therefore defined by
q
b
= Tq
d
(2)
where T = 0 except that T
16
= T
33
= T
44
= T
61
= -1 and T
27
= T
55
= T
72
= 1. Hence
e
66
-e
67
e
63
e
64
-e
65
e
61
-e
62
e
77
-e
73
-e
74
e
75
-e
71
e
72
e
33
e
34
-e
35
e
31
-e
32
e
44
-e
45
e
41
-e
42
(sym)
e
55
-e
51
e
52
e
11
-e
12
e
22
e
86
-e
87
e
83
e
84
-e
85
e
81
-e
82
e
96
-e
97
e
93
e
94
-e
95
e
91
-e92
-e
10,6
e
10,7
-e
10,3
-e
10,4
e
10,5
-e
10,1
e
10,2
From the above discussion, if there were direct connections between the leading and
trailing bogies and their wheelsets the total number of parameters would be 64 assuming
that the rigid body conditions were satisfied and the vehicle is symmetric. This number is
much reduced if there are no direct connections between the leading and trailing bogies.
Further reductions in the number of disposable parameters will occur if it is assumed that
the bogies are themselves symmetric and also if the curving condition is imposed.
Fully generic schemes, such as those discussed in Chapter 4, could be the starting
point for an analysis of the bogie vehicle. However, a more direct approach is to derive
the simplest arrangements which are capable of both perfect steering and stability as fully
generic schemes are unnecessarily complex. The spirit of the discussion so far,
concerning the resolution of the conflict between steering and stability, is that the vehicle
should be made as flexible as possible in plan view, consistent with the need for stability.
Whilst the range of configurations with the maximum number of parameters are of
(3)
dd
E =
cd
E =
(4)
interest for two and three-axle vehicles, the situation for a four-axle vehicle is quite
different. As discussed above, in the case of the vehicle with four axles under
consideration here, a fully generic configuration meeting the criterion expressed by
equation (3.15) would have a very large number of parameters, far more than is needed
for practical configurations which should be as simple and as flexible as possible.
If, however, attention is confined to low speeds or quasi-static conditions then the
equations of motion can be reduced to the form of (4.20) and (4.22), as discussed in
Section 3, involving only 8 wheelset coordinates. If the curving condition, equation
(4.10), is applied in addition to the rigid body conditions, equations (4.5) and (4.6), the
stiffness matrix E
*
can be expressed in the form of equation (4.4). There will therefore
be a maximum of 9 and a minimum of 5 independent parameters in E
*
. It will be
shown below how this suggests an approach to the generation of generic
configurations for the steered bogie vehicle.
Noting that fully generic arrangements may be useful in generating design
variants, the further development of the argument takes as its starting point the
simpler form of arrangement shown in Figure 6.2. The arrangement is similar to
conventional practice, except that provision is made for asymmetry in both geometry
and in the magnitude of the stiffnesses.
The solution of equations (4.10) will initially be discussed for this simplified
generic arrangement. Assuming zero cant deficiency, equations (4.10) become

y
1
{ y
1
+ h
11
- y
b
+ d
1b
- (h + h
1
)
b
} = 0 (5)
k
1
( 1
-
b
) = 0 (6)
k
y2
{- y
b
+ d
1b
+ (h + h
2
)
b
+ y
2
- h
22
} = 0 (7)
k
2
( b
-
1
) = 0 (8)
k
yb
{ y
b
+ d
2b
+ h
5b
- y
c
+ d
3c
- (c + h
5
)
c
} = 0 (9)
k
b
(
b
-
c
) = 0 (10)
k
1 b
= 0 (11)
k
b
(
b
-
c
) = 0 (12)
k
2 b
= 0 (13)
k
y1
{ y
4
- h
14
- y
d
+ d
1d
+ (h + h
1
)
d
} = 0 (14)
k
1
(
4
-
d
) = 0 (15)
k
y2
{- y
d
+ d
1d
- (h + h
2
)
d
+ y
3
+ h
23
} = 0 (16)
k
2
(
d
-
3
) = 0 (17)
k
yb
{ y
d
+ d
2d
- h
5d
- y
c
+ d
3c
+ (c + h
5
)
c
} = 0 (18)
k
b
(
d
-
c
) = 0 (19)
k
1 d
= 0 (20)
k
b
(
d
-
c
) = 0 (21)
k
2 d
= 0 (22)
From equations (11) to (13) and (20) to (22)
b = c = d = 0.
From equations (6),
(8), (15) and (17) the wheelsets can only take up a radial position if k
1
= k
2
=0.
Similarly, from equations (10) and (19) k
b
=k
d
= 0. Therefore for zero creep all
inter-body yaw stiffnesses must be zero.
If h
1
and h
2
are fixed then

equations (5) and (7) define y
b
and
b
, and equations
(14) and (16) define y
d
and
d
. If h
5
is fixed then

equations (9) and (18) define y
c
and
c
. If h
1
, h
2
and h
5
are zero then
y
b
= ( y
1
+ y
2
)/2
b
= ( y
1
- y
2
)/2h (23)
y
d
= ( y
3
+ y
4
)/2
d
= ( y
3
- y
4
)/2h (24)
y
c
= ( y
b
+ y
d
)/2
c
= ( y
b
- y
d
)/2c (25)
so that in this case the attitude of the vehicle is completely defined by the lateral
positions of the wheelsets y
i
.
In the case of motion on a uniform curve, the analysis of Chapter 4 shows that in
order to achieve zero lateral and longitudinal creep each wheelset moves outwards
and aligns itself radially. Because of symmetry y
d
= y
b
,
d = b, d
= -
b
, y
4
= y
1
, y
3
=
y
2
,
4
= -
1
,
3
=-
2
and
c
= 0. For example, in the case of linear conicity and the
outer and inner wheelsets are identical, the lateral displacements are
y
1
= y
4
= - lr
0
/
R + (c + h)
2
/2R y
2
= y
3
= - lr
0
/
R + (c - h)
2
/2R (26)
and the yaw displacements are
1
= -
4
= (c + h)/R
2
= -
3
= (c - h)/R (27)
and the displacements of the car body and bogie frames are given from (23-25) as
y
b =
y
c
= y
d
= -lr
0
/
R + (c
2
+ h
2
)/2R
b = d
= c/R
c = 0
(28)
In the general case, where h
1
, h
2
and h
5
are not zero, Figure 6.7 indicates the
geometry in which the wheelsets adopt a radial position with radial displacements
appropriate to their conicity, and the frames and car body take up the positions
connecting the wheelsets without strain of the suspension elements. Thus, under the
conditions considered here, perfect steering is obtained because of the combined
action of creep and conicity which is the primary mechanism of guidance.
Additional elastic restraint is needed in order to stabilise the kinematic oscillations
of the wheelsets. For the system with the elastic stiffness matrix E as derived above it
may be easily verified that there are 4 pairs of imaginary eigenvalues at low speeds
corresponding to undamped kinematic wheelset oscillations. This is consistent with the
necessary criterion for stability, equation (4.15), derived above. As the degeneracy of E is
8, additional stiffnesses are required to reduce the degeneracy to 4 so that the criterion may
be satisfied. Four additional stiffnesses must be provided, and the corresponding rows in
the compatibility matrix a must satisfy the rigid body conditions expressed by equations
(4.5) and (4.6) and the steering condition expressed by equation (4.10). In addition they
must be linearly independent from the existing rows of a. Since the corresponding rows of
the compatibility matrix will involve the wheelset yaw angles they can be considered as
defining "steering laws", a term used in the literature. These sub-matrices of k and a
relating to the additional elastic elements will be denoted by k
s
and a
s
for convenience.
Various possibilities are considered in the next section.
These results are consistent with the above discussion concerning the number of
independent parameters in E. The basic system, without the additional stiffnesses
corresponding to the steering laws, has 3 independent stiffnesses k
y1
, k
y2
and k
yb
, and a
minimum of 5 are required. The two additional stiffness parameters will each appear twice
in the 4 additional rows of a
s
, two rows for the leading bogie and two rows for the trailing
bogie. (The roll stiffnesses do not affect the degeneracy of E).
Since no elastic forces can be generated in perfect steering, as expressed by equations
(4.10), it is clear that the above solutions hold even if all the elastic stiffnesses are very
large. For the configuration under discussion, in addition to equations (4.1), (4.2) and (4.9)
being satisfied, symmetry implies that there is a fourth solution Eq
d
= 0 where q
d
is anti-
symmetric so that y
d
= -y
b
,
d =b, d
=
b
, y
4
= -y
1
, y
3
= -y
2
,
4
=
1
,
3
=
2
and

y
c
=
c
=
0. It can be seen that when the elastic stiffnesses are all large the system acting as a free
body would have four degrees of freedom, rigid body lateral translation, rigid body yaw
and symmetric and antisymmetric bending as a mechanism, corresponding to degeneracy
of four in the elastic stiffness matrix, Figure 6.8.
Figure 6.7 Geometry of basic bogie vehicle on curve: a bogie frame, b unstrained spring con-
necting bogie frame to car body (corresponding to k
yb
), c and d unstrained springs corre-
sponding to primary lateral suspensions k
y1
and k
y2
.
Though the application of direct structural connections between the bogies is not
considered here, it is worth noting that the antisymmetric bending mode can be sup-
pressed, without degrading the ability of the vehicle to steer properly, by providing a
direct shear connection between the bogies. Such inter-couplers are used in practice
with the objective of reducing forces in curves, [20].
6.6 Application to Specific Configurations
Analytical studies of body-steered bogie vehicles were initiated by Bell and Hedrick [21]
and Gilmore [22] who identified various instabilities which were promoted by low
conicities and reduced creep coefficients. A considerable body of work by Anderson and
Smith and colleagues is reported in [23] to [28], covering the analysis of a vehicle with
bogies having separately steered wheelsets. Weeks [29] has described dynamic
modelling and track testing of vehicles with steered bogies, noting the enhanced
sensitivity of this type of configuration to constructional misalignments.
The theoretical considerations of the previous Section are illustrated by the scheme
shown in Figure 6.9. Two additional stiffnesses k
a
and k
b
, representing linear springs
Figure 6.8 Mechanism modes for basic bogie vehicle: a rigid body lateral translation, b rigid
body yaw, c symmetric bending, d antisymmetric bending.
between the wheelsets and the car body, are added to the basic scheme
discussed above. Then the matrices k
s
and a
s
for the leading bogie are also shown in
Figure 6.9, and there will be similar matrices for the trailing bogie. The first two rows
relate to the stiffnesses k
a
and k
b
, and satisfy equations (4.5), (4.6) and (4.10) as do the
corresponding rows for the trailing bogie. In accordance with the above discussion only
four additional stiffnesses are required to decrease the degeneracy of the stiffness matrix
to the maximum of four that is required.
The addition of interwheelset shear stiffness k
c
such as might be provided by
cross-bracing is covered by the third row in a
s
which relates to k
c
and is also shown in
Figure 6.9 for the leading bogie. There is a corresponding row for the trailing bogie and
both rows satisfy equations (4.5), (4.6) and (4.10). This element could replace either k
a
and k
b
and the criteria for stability and perfect steering would still be satisfied. Thus a
h h
k
y
b
y
b
b
d
c7
d
b7
k
a
k
c
h
7
h
9
k
b
h
3
c
k
a

b
k
s
= [ k
a
k
b
k
c
]
1 - h
7
0 0 0 0 0 -1 d
c7
-(c + h - h
7
) 0 0 0 0 0 0 0
a
s
= 0 0 0 0 0 1 h
9
-1 d
c9
-(c - h +h
9
) 0 0 0 0 0 0 0
1 - h
3
0 0 0 - 1 -h
4
0 0 0 0 0 0 0 0 0 0
Figure 6.9 Simplified generic bogie with additional stiffnesses: top, plan view, bottom, end
view looking forward at the leading bogie and outer wheelset. (h
7
= - ch/(c + h);
h
9
= - ch/(c - h); h
4
= 2h - h
3
).
d
b c
a
f
g
Figure 6.10 Wiesinger bogie: a and d pivots connecting wheelsets to bogie frame g (k
y1
and
k
y2
very large), b pivot connecting bogie frame to car body (k
yb
very large), c pivot
connecting steering arm e to car body (k
a
very large), f linkage providing shear
connection between the wheelsets (k
c
very large).
c
5
e
3
c
1
c
2
e
1
e
2
c
3
c
6
c
4
e
4
k
s
= [ k
a
k
b
k
c
]
0 - 1 0 0 1 + h/c 0 0 0 0 - h/c 0 0 0 0 0 0 0
a
s
= 0 0 0 0 1 - h/c 0 -1 0 0 h/c 0 0 0 0 0 0 0
1 - h
3
0 0 0 -1 -h
4
0 0 0 0 0 0 0 0 0 0 0
k
a
= e
1
e
3
c
2
2
(c
5
- c
1
)
2
/{e
1
(c
2
- c
1
)
2
+ e
3
(c
5
- c
1
)
2
}
k
b
= e
2
e
4
c
3
2
(c
6
- c
4
)
2
/{e
2
(c
4
- c
3
)
2
+ e
4
(c
6
- c
4
)
2
}
c
2
(c
5
- c
1
)/c
1
(c
5
- c
2
) = c/(c + h)
c
3
(c
6
- c
4
)/c
4
(c
6
- c
3
) = c/(c - h)
Figure 6.11 Linkage steered bogie with separately steered wheelsets.
practical choice for design is to steer the outer wheelset from the car body and steer the
inner wheelset by means of a shear connection from the outer wheelset. This is
exemplified by designs of steered bogie due to Robinson (1881) (cited by Smith [12])
and Wiesinger (1932) [11], Figure 6.10. The need for elastic elements in order to
achieve a margin of stability and satisfactory response was not then recognised; so in
these schemes k
a
and k
c
are very large, leading to the use of the pivots a, b, c and d.
However, the analysis above suggests that it is possible to dispense with pivots and
accommodate the necessary relative displacements by the use of suitably disposed elastic
elements. A serious drawback of this configuration is that the direct coupling of the
wheelset motions to the car body degrades stability and ride quality as shown in [30].
In the configuration shown in Figure 6.11 each wheelset is separately steered
through a linkage connected to the car body, and is the basis of the steered bogie
developed by UTDC. Such arrangements have been analysed by Smith and Anderson
[23-27] and Shen [31]. It can be seen that the effective stiffnesses k
a
and k
b
are
functions of the linkage stiffnesses e
1
, e
2
and e
3
. Expressions for k
a
and k
b
are derived
by considering the equilibrium of the system in terms of the generalised coordinates
and the angular position of the link, and then eliminating the latter as inertia effects in
the links are negligible.
As in the previous example, alternative configurations can be derived by providing
inter-wheelset shear stiffness k
c
. Then any one of k
a
, k
b
and k
c
, or the corresponding
element stiffnesses, could be set to zero and the condition for stability satisfied.
Alternative configurations could therefore have either the outer or the inner wheelsets
steered together with an interwheelset shear connection. This could represent a useful
simplification, which does not appear to have been studied in modern times, though it
is the equivalent of the schemes by Robinson and Weissinger mentioned above.
Careful choice of the dimensions of the linkages is necessary in order to keep
within practical sizes. For this reason schemes have been developed in which the
linkage is in the vertical plane.
A more common variant of body steered bogie in which the wheelsets are jointly
steered by one lever is shown in Figure 6.12. To represent this configuration 3 addi-
tional rows of the compatibility matrix are required for the leading bogie, corre-
sponding to stiffnesses k
a
, k
b
and k
c
which are functions of the element stiffnesses e
1
,
e
2
and e
3
as shown in Figure 6.12. The matrices k
s
and a
s
for the leading bogie are
also shown in Figure 6.12, and there will be similar matrices for the trailing bogie.
Elements of this design go back as far as 1841, and there are many modern exam-
ples. The third row of a is equal to the sum of the first and second rows and thus
only two of the three rows in the compatibility matrix for the linkage are independ-
ent. k
c
is equivalent to an inter-wheelset shear stiffness, and thus the arrangement
not only provides the steering action defining the yaw angles of the wheelsets rela-
tive to the body, but also imposes inter-wheelset shear restraint.
Figure 6.13 shows a scheme which received intensive development by Liechty [11].
A further example is provided by the configuration shown in Figure 6.14 in which the
wheelsets are steered by a linkage pivoted on a member joining the wheelsets. This is
clearly derived from the scheme of Figure 6.11 and provides a useful model for the
discussion of stability below.
Figure 6.15 shows the dynamic response of a vehicle with the configuration of Figure
6.14 and parameters given in Table 6.1 to entry into a curve consisting of a linearly
increasing curvature over a distance of 20 m and the constant curvature of 225 m. The
track is assumed to be canted so cant deficiency is zero. Figure 6.15 should be
compared with Figure 6.6 which gives the corresponding results for the conventional
bogie vehicle. When the vehicle is completely on the uniform section of the curve, the
wheelsets move outwards and adopt a radial position with negligible longitudinal
creep and creep forces. The spin on the outer wheels is relatively large and results in
corresponding lateral creep forces but the lateral forces on the inner wheels are small.
These results are consistent with equations (3.2.8-9). The bogie frames and car body
also adopt radial positions consistent with the wheelset positions so that in the steady
state the suspensions are not strained.
However, during the transition quite large longitudinal and lateral creep forces are
exerted, as on the transition the suspension is strained and the resulting forces must
be reacted by inertia and creep forces. These forces reduce to zero as the vehicle
k
s
= diag[ k
a
k
b
k
c
]
0 - 1 0 0 1 + h/c 0 0 0 0 - h/c 0 0 0 0 0 0 0
a
s
= 0 0 0 0 1 - h/c 0 -1 0 0 h/c 0 0 0 0 0 0 0
0 -1 0 0 2 0 -1 0 0 0 0 0 0 0 0 0 0
k
a
= e
1
e
3
c
2
2
(c
5
- c
1
)
2
/ k
b
= e
2
e
3
c
1
2
(c
5
- c
3
)
2
/ k
c
= e
1
e
2
c
2
2
(c
3
- c
1
)
2
/
= {e
1
(c
2
- c
1
)
2
+ e
2
(c
3
- c
1
)
2
+ e
3
(c
5
- c
1
)
2
}
c
2
(c
5
- c
1
)/c
1
(c
5
- c
2
) = c/(c + h)
c
3
(c
5
- c
1
)/c
1
(c
5
- c
3
) = c/(c - h)
Figure 6.12 Steered bogie with shared wheelset linkage.
c
1
k
s
= diag[ k
a
k
b
k
c
]
a
h c c
c h c
h h
s
=

1 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 1 0 0 0 0 0 0 0 0 0
1 0 0 0 1 0 0 0 0 0 0 0 0 0 0
2 2
2 2
k
a
= e
1
e
3
c
1
2
/
k
b
= e
2
e
3
c
1
2
/

k
c
= e
1
e
2
c
2
2
/

= e
1
c
2
2
+ e
2
c
2
2
+ e
3
c
1
2
Figure 6.13 Liechtys bogie. c
2
= h
2
/c.
k
s
= diag[ k
a
k
b
k
c
]
a
h c h h c h h c
h c h h c h h c
h h
s
=
+ +

( / ) / ( / ) / /
( / ) / ( / ) / /
1 2 1 0 0 0 1 2 0 0 0 0 0 0 0 0 0 0
1 2 0 0 0 0 1 2 1 0 0 0 0 0 0 0 0 0
1 0 0 0 1 0 1 0 0 0 0 0 0 0 0
Figure 6.14 Bogie with wheelsets separately steered from an inter-wheelset member.
Expressions for k
a
and k
b
and the relationship between the c
i
are the same as the scheme
of Figure 6.11.
e
1
e
2
e
3 e
4
c
2
c
1
c
3
c
4
c
6
a
k
c
c
5
e
2
e
3
e
1
h
c
2
c
1
leaves the transition. This is consistent with the results given in Figure 6.15. A
compensating steering arrangement that would account for transition geometry has
been proposed by Smith [32].
The concept of the body-steered bogie has been applied to freight vehicles with three
piece bogies by Scales [33], List [34] and Shen [31] shows that dynamic stability as
well as curving performance is significantly improved in comparison with a
conventional design.
So far it has been shown that a general formulation of the equations of motion of a
symmetric railway vehicle with two unsymmetric bogies, capable of perfect steering,
makes it possible to consider a wide variety of possible configurations on a common
basis. Though there is a large number of disposable parameters theoretically
available, only a limited number are needed to derive practical configurations which
are simple and yet are capable of perfect steering. The configurations discussed satisfy
the necessary conditions for dynamic stability at low speeds, and so the actual range of
values of parameters necessary to achieve both static and dynamic stability must be
discussed next.
Figure 6.15 Dynamic response of steered bogie vehicle of Figure 6.14 to curve entry. V = 15
m/s, R
0
= 225 m, length of transition L = 20 m.
6.7 Stability of Bogie Vehicles with Steered Wheelsets
In this section the dynamic behaviour, as exemplified by the root locus as speed and
suspension stiffness is varied, of the body-steered four-axle bogie vehicles is con-
sidered. The configuration of steered bogie vehicle chosen for discussion is that of
Figure 6.14.
Consider firstly the behaviour at low speeds, and how the eigenvalues of the
system vary with suspension stiffness. Figure 6.16 shows the variation of the eigen-
values with k, a factor applied to k
y
, k
a
, k
b
and k
c
. When k = 0 there are four un-
damped oscillations at wheelset kinematic frequency
=V(
/lr
0
)
1/2
. Introduction of
stiffness results in damping of these oscillations, the real part of the eigenvalue be-
ing initially proportional to k. Two of these oscillations can be identified with the
steering or bending oscillation A and D of the leading and the trailing bogies respec-
tively. The other two oscillations can be identified with the shear oscillation B and C
of the leading and the trailing bogies respectively. In both cases the steering oscilla-
tion is lightly damped. The shear oscillations are heavily damped, splitting into four
subsidences B1, B2, C1 and C2 as k increases. It can be seen that if these results are
compared with those for the isolated bogie discussed in Section 5.8, the eigenvalues
corresponding to the steering and shear oscillations are significantly changed in
character. In fact, both oscillatory instability and divergence occur for certain values
of k, and the conditions for this are discussed below in detail.
Not shown in Figure 6.16 are eigenvalues which are largely independent of k.
These eigenvalues, of which there are 9 complex conjugate pairs, refer to modes
Figure 6.16 Variation of the eigenvalues for the body-steered vehicle at low speeds as a
function of a factor on the stiffnesses k
a
and k
c
(eigenvalues of large modulus not
shown).
Figure 6.17 Root locus as speed is varied for body-steered bogie vehicle of Figure 6.11. The
left hand plot shows the root locus in the complex plane and the right hand plot is the
variation of the imaginary parts with speed. Parameters of Table 6.1 with k
a
= 40 MNm,
k
b
= 40 MNm, k
c
= 0, k
y
= 0.

which are substantially oscillations of the vehicle body and the bogie frames on the
primary and secondary suspension as if the wheels were fixed. The influence of
speed on the eigenvalues for the steered bogie vehicle is shown in Figure 6.17, for
values of k
y
, k
a
, k
b
and k
c
which give stability at low speeds. Comparison with Fig-
ure 6.3 shows that the results are very similar to those for a conventional bogie vehi-
cle, with the notable exception that the branches A1 and A2, initially proportional to
speed, corresponding to the bogie steering modes, differ. Inspection of the eigenvec-
tors shows that one branch refers to motions mainly involving the leading bogie and
the other branch refers to the trailing bogie. The eigenvalues associated with the
vehicle body and bogie frame modes, substantially independent of speed, are
broadly similar to those of the conventional bogie vehicle, as are the 8 relatively
large real negative eigenvalues corresponding to wheelset subsidences in lateral
translation and yaw.
As in the case of the conventional bogie vehicle, at speeds for which the
steering frequencies approach the natural frequencies of the vehicle body on the
suspension, there is the possibility of body instability. Such body instability can be
eliminated by a suitable choice of secondary suspension parameters as discussed for
the conventional bogie vehicle. As speed is increased, the combined effect of speed
and inertia modifies the damping in the steering oscillations. For each steering oscil-
lation which is stable for low speeds, the damping will vanish at a bifurcation speed
V
B
. The instability corresponds to a form of bogie hunting, and unlike the conven-
tional bogie vehicle where the two bifurcation speeds are approximately the same, as
Figure 6.17 shows those for the body steered vehicle can be quite different.
6.8 Simple Bogie Model
Examination of the reduced order stiffness matrix E
*
defined in equations (4.19) and
(4.20) shows that, for practical vehicles, E
*
is of the form E
*
= E
*
0
+
E where
E is
small compared with E
0
*
, and
e
11
e
12
-e
11
e
14
0 0 0 0
e
12
e
22
-e
12
e
24
0 0 0 0
-e
11
-e
12
e
11
-e
14
0 0 0 0
e
14
e
24
-e
14
e
44
0 0 0 0
0 0 0 0 e
11
e
14
-e
11
e
12
0 0 0 0 e
14
e
44
-e
14
e
24
0 0 0 0 -e
11
-e
14
e
11
-e
12
0 0 0 0 e
12
e
24
-e
12
e
22
where e
11
= k
a
2
/4h
2
+ k
b
2
/4h
2
+ k
c
e
12
= - k
a/
2h - k
c
h
e
14
= - k
b
/2h - k
c
h
e
22
= k
a
+ k
c
h
2
e
24
= k
c
h
2
e
44
= k
b
+ k
c
h
2
where
= 1 + h/c,
= 1 - h/c. E
0
*
is the stiffness matrix derived by assuming that the
car body is restrained from deviating from a uniform forward motion.
As in the case of the conventional bogie vehicle, for the purposes of trend studies a
good approximation to the eigenvalues associated with bogie motions can be obtained
by treating the vehicle body as an inertial mass.
Accordingly, referring to the equations of motion put y
c
,
c
,
c
= 0. Furthermore, if
it assumed that k
y
is large and remembering that k

= 0
, the bogie frame co-ordinates
are given by y
b
= ( y
1
+ y
2
)/2 and
b
= ( y
1
- y
2
)/2h. If d
1
= 0 then the bogie frame roll
coordinate is laterally uncoupled from the rest of the vehicle. The system is thus re-
duced to two uncoupled sets of four degrees of freedom each, involving y
1
,
1
, y
2
,
2
E
0
*=
(1)
and y
3
,
3
, y
4
,
4
. Finally the secondary lateral suspension stiffness k
yb
has only a
small influence on the stability of the bogies and will be neglected.
Introducing sum and difference coordinates
i
defined by equation (4.2.3)
and neglecting all suspension damping terms, the equations of motion will be
{(m/2 + m
b
/4)s
2
+ fs/V + e
11
}
1
- f
2

e
12
2
+ e
144 = 0
(3)
(f
l/r
e
12
)

1
+ {(I/2)s
2
+ fl
2
s/V + e
22}
2
e
244
= 0 (4)
{(m/2 + m
b
/4)s
2
+ fs/V }
3
- f

4
= 0
(5)

e
141
e
242
+ (f
l/r)
3
+ {
( I/2)s
2
+ fl
2
s/V + e
44}
4
= 0 (6)
where now
e
11
= k
a
2
/4h
2
+ k
b
2
/4h
2
+ k
c
e
12
= - k
a
/4h + k
b
/4h
e
14
= - k
a
/4h - k
b
/4h - k
c
h
e
22
= k
a
/4 + k
b
/4
e
24
= k
a
/4 - k
b
/4
e
44
= k
a
/4 + k
b
/4 + k
c
h
2
The upper sign refers to the leading bogie and the lower sign refers to the trailing
bogie. A particularly simple case arises when the bogie frame mass m
b
= 0, corre-
sponding to the configuration shown in Figure 6.14 with a light inter-wheelset
member, for then if it is assumed that I = ml
2
it is possible to write
D = ms
2
/2f + s/V (7)
then a trial solution q = ae
D t
leads to the characteristic equation
p
4
D
4
+ p
3
D
3
+ p
2
D
2
+ p
1
D + p
0
= 0 (8)
where
p
4
= f
4
l
4
p
3
= f
3
l
2
{ k
a
(1/2 +
2
l
2
/4h
2
) + k
b
(1/2 +
2
l
2
/4h
2
) + k
c
( l
2
+ h
2
)}
p
2
= 2f
4
l
2
+ f
2
k
a
k
c
{l
2
+ h
2
+ l
2
(1-
)
2
}/4 + f
2
k
a
k
b
{1+ l
2
/h
2
+ l
2
(
)
2
/4h
2
}/4
+ f
2
k
b
k
c
{l
2
+ h
2
+ l
2
(1-
)
2
}/4 f
3
l
2
(k
b
- k
a
)(1 -
)/4h
p
1
= f
3
{ k
a
(1/2 +
2
l
2
/4h
2
) + k
b
(1/2 +
2
l
2
/4h
2
) + k
c
(l
2
+ h
2
)}
f
2
(1 -
){-k
a
k
c
(

-1)h + k
b
k
c
(

- 1)h + k
a
k
b
(

-
)/2h}/4
p
0
= f
4
2
+ f
2
(k
a
k
b
/h
2
+ k
a
k
c
+ k
b
k
c
)/4 f
3
(k
b
- k
a
)(1 -
)/4h
and
l/r
0
.
6.9 Stability of Simple Bogie Model
First consider stability at low speeds so that D becomes equal to s/V. In this case the
results are applicable to the configurations shown in Figures 6.11 and 6.14. Rouths
criterion for oscillatory stability, equation (3.4.3), will be applied. It can be shown
that if any two of the stiffnesses k
i
are zero, T
3
= 0 and an undamped oscillation at
wheelset kinematic frequency occurs in accordance with the discussion of Section 4.
Consequently, it is of interest to discuss the three cases, each for the leading and
trailing bogie, firstly where the outer wheelset is steered and there is shear stiffness
between the wheelsets, secondly where the inner wheelsets are steered and there is
shear stiffness between the wheelsets and thirdly where both outer and inner wheel-
sets are steered.
Firstly consider the possibility of oscillatory instability at very low speeds. This
can occur only for very large values of the stiffnesses k
i
and then the sign of T
3
is
determined by the sign of p
1
so that on substitution for r and s the condition for os-
cillatory stability is
(1 -
){k
a
k
c
+ k
b
k
c
+ k
a
k
b
/h
2
} < 0 (1)
where the upper sign refers to the leading bogie and the lower to the trailing bogie.
Thus the stability depends on (1-
).For the leading bogie, oscillatory instability

occurs if
<1, but not if

>1. For the trailing bogie, oscillatory instability occurs
if
>1, but not if

<1.
The mechanics of the instability is illustrated by examination of the case where
k
b
= 0 and both k
a
and k
c
are made very large, which implies that
-(
/
h)
1
+
2
+

4
= 0 (2)
1
- h
4
= 0 (3)
Therefore defining new generalised coordinates
1
and
2

so that
or
= (5)
and performing the transformation of equations (3) to (6) using (4.2.5), the equa-
tions of motion at low speeds become
fs
1
/V - f

2
= 0 (6
)
(f
l/r)
1
+ [f{(l
2
+ h
2
+ l
2
(
- 1)
2
}s/V + fh(1 -

)(1 -
)
]
2
= 0 (7)
The characteristic equation then reduces to the quadratic
{l
2
+ h
2
+ l
2
(

- 1)
2
}D
2
+ h(1 -
)(1 -
)D + (
l/r)

= 0 (8)
Equations (4) and (5) indicate that if

=1 the bogie behaves as a rigid two-axle ve-
hicle. Note the similarity of the frequency of the steering oscillation
= (
l/r)V/{l
2
+ h
2
+ l
2
(
- 1)
2
}
1/2
(9)
to that of a rigid bogie with the same wheelbase,
= (
l/r)V/(l
2
+ h
2
)
1/2
. Thus the
mechanism of the oscillatory instability lies in the modification of this oscillation by
the action of the steering linkage. The oscillation is illustrated in Figure 6.18.
Figure 6.18 Mode shape of low speed oscillatory instability.
Figure 6.19 Stability diagrams showing the bifurcation speed V
B
(m/s) for the body-steered
vehicle with separately steered wheelsets of Figure 6.11 as a function of k
a
(10
7
MNm)
and k
b
(10
7
MNm). Complete vehicle (left hand), trailing bogie with car body fixed (cen-
tre) and leading bogie with car body fixed (right hand). Coned wheels,

= 0.05.
Figure 6.19 gives the stability boundaries as k
a
and k
b
are varied for the case where k
c
= 0 as given by the solution of the equations of motion for the complete vehicle with
coned wheels and

= 0.05, the trailing bogie with car body fixed and the leading bo-
gie with car body fixed. It can be seen that the stability diagram for the complete vehi-
cle is closely approximated by the superposition of the diagrams for the bogies with
car body fixed in accordance with the discussion in Section 7. The low speed oscilla-
tory instability just discussed occurs only for very large values of k
a
and k
b
for ordi-
nary values of the system parameters, but static instabilities do occur within the practi-
cal range of k
a
and k
b
. A close approximation to these results follows from equations
(8.8) because for static stability p
0
> 0 or
4f
2
+ k
a
k
b
/h
2
+ k
a
k
c
+ k
b
k
c
f(k
b
- k
a
)(1 -
)/h > 0 (10)

Assuming the usual case where
<1 , examination of equation (10) shows that

for the case where the outer wheelset is steered (k
b
= 0) the trailing bogie is
statically stable for all values of k
a
and k
c
and the leading bogie is statically
unstable for sufficiently large values of k
a
and sufficiently small values of k
c
.
Instability occurs for k
c
= 0 if k
a
> 4fh
/

( 1 -
) and for k
a
very large if
k
c
< f

( 1 -
)/h.

if the inner wheelset is steered the leading bogie is statically stable for all
values of k
b
and k
c
and the trailing bogie is statically unstable for sufficiently
large values of k
b
and sufficiently small values of k
c
. Instability occurs
for k
c
= 0 if k
b
> 4fh
/

( 1 -
) and for k
b
very large if k
c
< f

( 1 -
)/h.

if both wheelsets are steered, and k
c
= 0, the leading bogie is statically unsta-
ble for sufficiently large values of k
a
and for sufficiently small values of k
b
.
Instability will occur for k
b
= 0 if k
a
> 4fh
/

( 1 -
) and for k
a
very
large if k
b
< f
h( 1 -
). The trailing bogie is statically unstable for suffi-

ciently large values of k
b
and for sufficiently small values of k
a
. Instability
occurs for k
a
= 0 if k
b
> 4fh
/

( 1 -
) and for k
b
very large if k
a
< f

h( 1 -
).
In all cases low values of conicity promotes divergence and increasing conicity in-
Figure 6.20 Stability boundaries showing contours of the bifurcation speed V
B
for body
steered bogies with wheelsets separately steered from an interwheelset member (Figure
6.14) with car body fixed: outer wheelset steered through stiffness k
a
with inter-wheelset
shear connection k
c
, trailing bogie (a) and leading bogie (b); inner wheelset steered k
b
with inter-wheelset shear connection k
c
, trailing bogie (c) and leading bogie (d); both in-
ner and outer wheelsets steered with no inter-wheelset shear connection, trailing bogie
(e) and leading bogie (f). Coned wheels,

= 0.05. D refers to divergence, O refers to
low speed oscillatory instability.
creases static stability. The divergence boundaries shown in Figure 6.19 occur for
values of the stiffnesses closely in accordance with the simple criteria discussed
above. Note that these static instabilities all occur if

=1. They are the result of
mounted on a pivoted arm as discussed in Chapter 3.
Inspection of equation (10) shows that static instability can be eliminated com-
pletely if the linkage stiffnesses k
a
and k
b
are chosen to be different such that k
b
/k
a
=
/.
The form of these results is very similar to those obtained by Bell and Hed-
rick [21] for a bogie in which the relative angular displacement between the wheel-
sets was proportional to the relative motion between the bogie frame and car body, a
configuration similar in some respects to that of Figure 6.10. The results given here
may also be compared with those given by Smith and Anderson [25] for a bogie in
which there is a fixed relationship between k
a
and k
b
.
Figure 6.19 also shows contours of the bifurcation speed V
B
. The instability cor-
responds to a form of bogie hunting which is an inertia driven instability of the
steering oscillation. It is similar to the instability of a conventional bogie, but as the
linkage restraining the bogie can be made stiff without degrading curving perform-
ance, quite high critical speeds can be achieved when parameters are chosen ap-
propriately. Figure 6.20 indicates the stability as a function of the stiffnesses k
i
for
the leading and trailing bogies for each of the cases where one of the three k
i
are
zero. For a complete vehicle values of the k
i
must be chosen to give a margin of sta-
bility for both leading and trailing bogies.
Moreover, as a range of conicities and creep coefficients must be considered in
practice, values of the stiffnesses k
i
must be chosen accordingly. Figure 6.21 shows
an example of the effect of varying equivalent conicity and creep coefficient on sta-
bility. In this case, the usual bogie instability is promoted by large conicities and the
instability of the leading bogie discussed above is promoted by low conicities and
creep coefficients. The result is a restricted range of speeds for stable operation un-
less the range of conicities is restricted.
Figure 6.21 Stability of the body-steered vehicle of Figure 6.11 showing the effect of varying
conicity and creep coefficient on the bifurcation speed V
B
(m/s) as k
a
(10
7
MNm) and k
b
(10
7
MNm) are varied. f/2 indicates that f
11
, f
22
and f
23
are halved. k
y
= 6 MN/m.
assymmetry rather than steering as such. For if

=1 then the bogie behaves as an
unsymmetric vehicle. The stability of unsymmetric vehicles is discussed in Chapter 9
but the salient feature of the static instability is revealed by considering a wheelset
cient it is possible that schemes in which all three stiffnesses are implemented could
provide better performance.
More generally, the approach followed in this Chapter makes it possible to con-
sider many possible configurations on a common basis. Though there are many dis-
posable parameters theoretically available, only a limited number are needed to de-
rive practical configurations which are relatively simple and which improve the
resolution of the conflict between steering and stability. Full assessment of a given
design must depend on design detail and performance in sharp curves, a situation in
which these schemes are likely to be applied, obviously requires a nonlinear analy-
sis. Furthermore, it has been shown that the introduction of more refined suspension
arrangements results in more forms of potential static and dynamic stability.
References
1. Lewis, M.J.T.:Early Wooden Railways. Routledge and Kegan Paul, London, 1974, p.
291.
2. Stover, J.F.: American Railroads. University of Chicago Press, Chicago, 1961, p. 25.
3. White, J.H.: A History of the American Locomotive. The John Hopkins Press, Balti-
more, 1968, pp. 169-174.
4. Ahrons, E.L.: The British Steam Railway Locomotive 1825-1925. The Locomotive
Publishing Co., London, 1927, p. 62.
5. White, J.H.: The American Railroad Passenger Car. The John Hopkins Press, Balti-
more, 1978, p. 497.
6. Ahrons, E.L. p. 157.
7. Matsudaira, T.: Hunting problem of high-speed railway vehicles with special reference
to bogie design for the New Tokaido Line. Inst. Mech. Eng. Proc. 180, Part 3F (1965),
pp. 58-66.
8. Sauvage, G.: Running quality at high speeds. In: O. Nordstrom (Ed.): The Dynamics
of Vehicles on Roads and Tracks, Proc. 9th IAVSD Symposium, Linkoping, Sweden,
June 1985. Swets and Zeitlinger Publishers, Lisse, 1986, pp. 496-508.
9. Evans, J.R.: The modelling of railway passenger vehicles. In: G. Sauvage (Ed.): The
Dynamics of Vehicles on Roads and Tracks, Proc. 12th IAVSD Symposium, Linkoping,
Sweden, August 1991. Swets and Zeitlinger Publishers, Lisse, 1992, pp. 144-156.
In general, detailed optimisation of the parameters is necessary considering not only
the linkage stiffnesses but the lateral primary suspension stiffness. Whilst it has been
shown that arrangements providing only two of the stiffnesses k
a
, k
b
and k
c
are suffi-
11. Liechty, R.: Das Bogenlaufige Eisenbahn-Fahrzeug. Schulthess, Zurich, 1934.
12. Smith, R.E.: Steering rail vehicle axles
a historical review. Proc. CSME Canadian

Engineering Centennial Convention, 1987.
13. Liechty, R.: Studie uber die Spurfuhrung von Eisenbahnfahrzeugen. Schweizer
Archiv f. Angewandte Wissenschaft und Technik, 3 (1937), pp. 81-100.
14. Liechty, R.: Die Bewegungen der Eisenbahnfahrzeuge auf den schienen und die da-
bei auftretenden Krfte. Elektrische Bahnen, 16 (1940), pp. 17-27.
15. Schwanck, U.: Wheelset steering for bogies of railway vehicles. Rail Engineering
International 4 (1974), pp. 352-359.
16. Wickens, A.H.: Steering and dynamic stability of railway vehicles. Vehicle System
Dynamics, 5 (1975), pp. 15-46.
17. Wickens, A.H.: Stability criteria for articulated railway vehicles possessing perfect
steering. Vehicle System Dynamics, 7 (1979), pp. 33-48.
18. de Pater, A.D.: Optimal design of a railway vehicle with regard to cant deficiency
forces and stability behaviour. Delft University of Technology Lab. for Eng. Mech. Re-
port 751, 1984.
19. de Pater, A.D.: Optimal design of railway vehicles. Ingenieur-Archiv 57 (1987), pp.
25-38.
20. Topham, W.L.: Methods of reducing flange wear on diesel and electric locomotives.
J. Inst. Loco. Engineers 49 (1959), pp. 771-825.
21. Bell, C.E. and Hedrick, J.K.: Forced steering of rail vehicles: stability and curving
mechanics. Vehicle System Dynamics 10 (1981), pp. 357-385.
22. Gilmore, D.C.: The application of linear modelling to the development of a light
steerable transit truck. In: A.H. Wickens (Ed.): The Dynamics of Vehicles on Roads and
Tracks, Proc. 7th IAVSD Symposium, Cambridge, September 1981. Swets and Zeitlinger
23. Fortin, J.A. and Anderson, R.J.: Steady-state and dynamic predictions of the curving
performance of forced-steering rail vehicles. In: J.K. Hedrick (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 8th IAVSD Symposium, Cambridge, Mass., August
10. Eickhoff, B.M., Evans, J.R. and Minnis, A.J.: A review of modelling methods for
railway vehicle suspension components. Vehicle System Dynamics 24 (1995), pp. 469-
496.
25. Smith, R.E and Anderson, R.J.: Characteristics of guided-steering railway trucks.
Vehicle System Dynamics, 17 (1988), pp. 1-36.
26. Anderson, R.J., Fortin, C.: Low conicity instabilities in forced-steering railway
vehicles. In: M. Apetaur (Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc.
10th IAVSD Symposium, Prague, August 1987. Swets and Zeitlinger Publishers, Lisse,
1988, pp. 17-28.
27. Smith, R.E.: Forced-steered truck and vehicle dynamic modes-resonance effects due
to car geometry. In: A. Apetaur (Ed.): The Dynamics of Vehicles on Roads and Tracks,
Proc. 10th IAVSD Symposium, Prague, August 1987. Swets and Zeitlinger Publishers,
Lisse, 1988, pp. 423-424.
28. Smith, R.E.: Dynamic characteristics of steered railway vehicles and implications for
design. Vehicle System Dynamics, 18 (1989), pp. 45-69.
29. Weeks, R.: The design and testing of a bogie with a mechanical steering linkage. In:
M. Apetaur (Ed.): The Dynamics of Vehicles on Roads and Tracks, Proc. 10th IAVSD
Symposium, Prague, August 1987. Swets and Zeitlinger Publishers, Lisse, 1988, pp. 497-
508.
30. Li, W.: The dynamics of perfect steering bogie vehicles and its improvement with a
reconfigurable mechanism. Doctoral Dissertation, Loughborough University, 1995.
31. Shen, Z.Y., Yan, J.M., Zen, J., and Liu, J.X.: Dynamical behaviour of a
forced-steering three-piece freight car truck. In: M. Apetaur (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 10th IAVSD Symposium, Prague, August 1987.
32. Smith, R.E.: Multiaxle steered articulated railway vehicle with compensation for
transitional spirals. U. S. Patent Application 157565, 1989.
33. Scales, B.T.: Behaviour of bogies on curves, Railway Engineering, December, 1972.
34. List, H.A.: Means for improving the steering behaviour of railway vehicles. Annual
Meeting of the Transportation Research Board, Washington, D. C. January, 1976.
24. Fortin, J.A.C., Anderson, R.J. and Gilmore, D.C.: Validation of a computer
simulation forced-steering rail vehicles. In: O. Nordstrom (Ed.): The Dynamics of
Vehicles on Roads and Tracks, Proc. 9th IAVSD Symposium, Linkoping, June 1985.
7
The Three-Axle Vehicle
7.1 Introduction
Various forms of three-axle vehicle have been used widely in the past. In most of
these designs the wheelsets were connected to the car body by a conventional sus-
pension similar to that used in two-axle vehicles. Negotiation of curved track was
catered for by allowing greater flexibility or clearances for the central wheelset.
However, there is also a long history of inventions which attempt to ensure that
wheelsets are steered so that they adopt a more or less radial position on curves. It
was argued that three axles, connected by suitable linkages, would assume a radial
position on curves and then re-align themselves correctly on straight track. As will
be shown below, a wide range of new potential instabilities are introduced. In fact,
the three-axle configuration is also important because it gives considerable insight
into the dynamic behaviour of articulated vehicles discussed in Chapter 8.
Three-axle vehicles were in use from an early date. According to Liechty [1] a
three-axle vehicle in which the lateral displacement of the central axle steered the
outer axles through a linkage was tried out on the Linz-Budweis railway in 1826.
Germain patented a design in 1837 in which radial steering was provided [2], and in
1844 Themor built a similar vehicle which was operated for some time [1]. Fidler
also patented a similar arrangement in 1868 [3], Figure 7.1(a). In these last three
schemes the outer wheelsets were pivoted to the car body. In 1889, Robinsons ar-
rangement [4] introduced the refinements of guides for the central wheelset, the
body pivots was placed slightly inboard of the outer wheelsets, and the central
wheelset had a much smaller radius than the outer wheelsets. Fayes 1898 patent [5]
removed the guides for the central wheelset in order to avoid reported difficulties
with Robinsons design on reverse curves. There were, of course, many different
ways of providing inter-wheelset steering, such as complex linkages, and this is ex-
emplified by the variety of designs produced since.
Fidler introduced direct shear connection between the outer wheelsets in 1868,
[3]. The central wheelset was mounted without lateral freedom in the car body, Fig-
ure 7.1(b). A similar arrangement was invented by Grover in 1880 [6].
All these developments were based on very simple ideas about the mechanics of
vehicles in curves. In this Chapter, the basic instabilities of three-axle vehicles with a
single car body will be considered, and how they are related to the natural steering
properties of the three-axle vehicle. Particular emphasis will be given to the various
possibilities for the connections between the wheelsets and the car body in order to
meet the conflicting requirements of stability and curving.
The three-axle vehicle was first examined in this context in a series of papers
[7]-[14]. A similar approach, though with slightly different assumptions, has been
followed by de Pater [15, 16] and Keizer [17].
7.2 Steering and Stability of Three Axle Vehicles
Consider the idealised three-axle vehicle shown in Figure 7.2. As the rotation of the
wheels cannot generate any forces in the suspension consider the set of nine general-
ised coordinates
q = [ y
1
1
y
2
2
y
b
b
y
3
3
]
T
(1)
as indicated in Figure 7.2.
The conditions governing steering and stability of multi-axle vehicles have been
discussed in Section 6.4 and it can be seen that the criterion expressed by equation
(6.4.15) can be satisfied by a three-axle vehicle. As in the case of the bogie vehicle,
the total number of disposable parameters is very large. Fully generic schemes could
Figure 7.1 Three-axle vehicles with steered wheelsets due to Fidler (a) outer wheelsets piv-
oted on car body (b) central wheelset pivoted on car body, direct shear connection be-
tween outer wheelsets.
THE THREE-AXLE VEHICLE 211
be the starting point for an analysis of the three-axle vehicle. However, it is more
important to derive the simplest arrangements which are capable of both perfect
steering and stability. The main thrust will be that the vehicle should be made as
flexible as possible in plan view, consistent with the need for stability.
Noting that fully generic arrangements may be useful in generating design vari-
ants, the further development of the argument is based on a simpler form of ar-
rangement shown in Figure 7.3. Firstly, consider the case where all wheelsets are
similar and each wheelset is connected to the car body in the conventional manner.
Then, the element stiffness and compatibility matrices are
k = diag [ k
y1
k
1
k
1
k
y2
k
2
k
2
k
y1
k
1
k
1
] (2)
a
d h
d
d h
=

1 0 0 0 1 2 0 0
0 0 0 0 0 1 0 0 0
0 1 0 0 0 0 1 0 0
0 0 1 0 1 0 0 0
0 0 0 0 0 1 0 0 0
0 0 0 1 0 0 1 0 0
0 0 0 0 1 2 1 0
0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 1 0 1
1
4
1
*
*
(3)
(a)
(b)
y
1
3 y
2
y
3
b
y
b
2h*
b y
b
Figure 7.2 Generalised coordinates for three-axle vehicle.
The solution of equations (6.4.10) will initially be discussed for this simplified ge-
neric arrangement. Equations (6.4.10) become
k
y1
(y
1
- y
b
+ d
1
b
- 2h*
b
) = 0 (4)
k
1
b
= 0 (5)
k
1
(
1
-
b
) = 0 (6)
k
y2
(- y
b
+ d
4
c
+ y
2
) = 0 (7)
k
2
b
= 0 (8)
k
2
(
b
-
2
) = 0 (9)
k
y1
( y
3
- y
b
+ d
1
b
+ 2h*
b
) = 0 (10)
k
1
b
= 0 (11)
k
1
(
3
-
b
) = 0 (12)
From equations (5), (8) and (11)
b
=0. From equations (6), (9), (12),
either
1
=
b
or k
1
= 0 (13)
either
2
=
b
or k
2
= 0 (14)
either
3
=
b
or k
1
= 0 (15)
k
y2
/2
k
y2
/2
k
y1
/2
k
y1
/2
k
2
/2
k
2
/2
k
1
k
1
k
y1
/2
k
y1
/2
2h*
h
1
Figure 7.3 Schematic showing suspension stiffnesses for three-axle vehicle.
k
yd
k
yd
k
t
For steering without lateral or longitudinal creep on a uniform curve, within the as-
sumptions of Section 6.4, and relative to the unstrained vehicle
y
1
= y
3
= - lr
0
/R
0
+ 2h*
2
/R
0
(16)
y
2
= - lr
0
/R
0
(17)

1
= -
3
= 2h*/R
0
(18)

2
= (19)
so that from (19) and (14)
b
= 0 and the value of k
2
can be arbitrarily chosen.
From (4), (7) and (10), either k
y2
= 0 and y
b
= y
1
= y
3
or k
y1
= 0 and y
b
= y
2
.
Figure 7.4 indicates the geometry for each of these configurations in which the
wheelsets adopt a radial position with radial displacements appropriate to their
conicity, and the car body takes up the positions connecting the wheelsets without
strain of the suspension elements.
These two configurations satisfy the conditions for steering so it is now neces-
sary to consider stability. For the first configuration, in which k
y2
= 0 and k
y1
, k
1
,
k
2
, and k
2
are non-zero, and the degeneracy of E is 5, and the stability criterion,
equation (6.4.15), is not satisfied and there are two pairs of imaginary eigenvalues
at low speeds corresponding to undamped kinematic wheelset oscillations. At least
two additional stiffnesses are required to reduce the degeneracy to three as required
by the stability criterion.
For the second configuration, in which k
y1
= 0 and k
y2
, k
1
, k
2
, and k
2
are non-
zero, the degeneracy of E is 6, and again the stability criterion, equation (6.4.15), is
not satisfied and there are three pairs of imaginary eigenvalues at low speeds corre-
sponding to three undamped kinematic wheelset oscillations. In this case at least
Figure 7.4 Geometry of two alternative configurations on uniform curve: top, k
y1
= 0 and y
b
=
y
2
; bottom, k
y2
= 0 and y
b
= y
1
.
three additional stiffnesses are required to reduce the degeneracy to three as re-
quired by the stability criterion.
As in the case of the bogie vehicle, for these additional stiffnesses, the corre-
sponding rows in the compatibility matrix a must satisfy the rigid body conditions
expressed by equations (6.4.5) and (6.4.6), and the steering condition expressed by
equation (6.4.10). In addition they must be linearly independent from the existing
rows of a. (Of course, the k
i
can be repeated as long as the corresponding row in a is
linearly independent, thus increasing the rank of a ). The sub-matrices of k and a
relating to the additional elastic elements will be denoted by k
s
and a
s
for conven-
ience.
Figure 7.3 shows that further elastic connections corresponding to direct shear
connections between adjacent wheelsets (of stiffness k
yd
) and the outer wheelsets (of
stiffness k
t
) have been introduced. As in the case of the body-steered bogie, in
schemes of the early period, pivots were used to provide lateral restraint and yaw
freedom. For example, in Robinsons scheme k
y1
and k
2
were very large though the
principles discussed above remain valid. As mentioned above, such a vehicle would
be unstable at low speeds. The sub-matrices of k
s
and a
s
relating to these additional
elastic elements are
k
s
= diag [ k
yd
k
yd
k
t
] (20)
a
h h
h h
h h
s
=

1 1 0 0 0 0 0
0 0 1 0 0 0 1
1 2 0 0 0 0 0 1 2
1 2
2 1
* *
(21)
where h
2
= 2h
*
- h
1
. The condition for stability at low speeds has been shown, in
Section 6.4, to depend on the stiffness matrix E* involving only the wheelset coor-
dinates. In the present case the number of disposable parameters will be (R
2
- 4R +
3)/4 = 4 and so E
*
may be expressed in terms of the given elements e
13
*, e
15
*, e
16
*,
and e
22
* and then
e
11
*= - e
13
*- e
15
*
e
12
*= e
13
*h*+ e
16
*
e
14
*= 4e
15
*h* - 2e
16
* + e
13
*h*
e
23
*= - e
13
*h*
e
24
*= - 2e
22
* - 3e
13
*h*
2
- 4e
16
*h*
e
25
*= - e
16
*
e
26
*= e
22
* + e
13
*h*
2
e
33
*= - 2e
13
*
e
34
*= 0
e
35
*= e
13
* (22)
e
36
*= e
13
*h*
e
44
*= 4e
22
* + 16e
16
*h* - 16e
15
*h*
2
+ 2e
13
*h*
2
e
45
*= - e
14
*
e
46
*= e
24
*
e
55
*= e
11
*
e
56
*= - e
12
*
e
66
*= e
22
*
For the enhanced configuration, E* has the following four independent elements
e
13
* = - k
yd
e
15
* = - k
t
- k
y1
k
2
/2(8k
y1
h*
2
+ k
2
) (23)
e
16
* = - 2k
t
h*
e
22
* = h
1
2
k
yd
+ 4k
t
h*
2
From equation (6.4.15), for stability, the degeneracy of E* must be not more than
three and this is satisfied if either (a) k
y1
= 0 and k
yd
, k
t
, k
y2
and k
2
are non-zero, (b)
k
y2
= k
t
= 0 and k
y1
, k
2
and k
t
are non-zero, or (c) k
y2
= k
2
= 0 and k
y1
, k
yd
and k
t
are
non-zero.
7.3 Steering with Unequal Conicities
The above discussion of curving requires modification to account for unequal conic-
ities of the wheelsets. There is a geometric requirement that a bending displacement
enabling all three wheelsets to take up radial alignment is compatible with the lateral
displacements required, by their various equivalent conicities, for pure rolling.
Figure 7.5 shows the geometry when such a vehicle negotiates a curve. For zero
longitudinal and lateral creep, the wheelsets must adopt a radial position, and, as-
suming for simplicity purely coned wheels, move outwards lr
0
/R
0
(for the outer
wheelsets) and lr
1
/
1
R
0
(for the central wheelset). If the radial displacements are
small compared with the radius of the curve, the geometry of Figure 7.5 then yields
the position of an effective pivot, measured from the central points between the
wheelsets, necessary for perfect steering
h*- h
1
= l( r
0
/ - r
1
/
1
)/2h* (1)
The possible range of configurations which satisfy this equation is constrained by
two practical considerations. Firstly, large values of conicity are not achievable in
practice. Secondly, assuming that the difference in conicity between outer and cen-
tral wheelsets is achieved by a difference in wheel diameters, it is unlikely that in a
good mechanical design the ratio of diameters could exceed about two, or excep-
tionally three. Figure 7.6 then summarises the possible range of configurations
which steer perfectly on curves for the case where h* = 1.25 m. Lines A and B give
the relationship between
1
and for the virtual pivot position being at the central
and outer wheelsets respectively. For longer wheelbases the position of the effective
pivot lies nearer the centre of the wheelbase. The lines C and D represent the limits
of a 2:1 ratio in conicity between central and outer wheelsets. A smaller ratio in
conicity leads to a pivot position between the wheelsets as shown at E, and large
ratios in conicity could require pivot positions outside the wheelbase.
lr
1
/R
0
1
lr
0
/R
0
h
1
Figure 7.5 Geometry of three-axle vehicle for zero lateral and longitudinal creep on uniform
curve when conicities are not uniform.
0 0.1 0.2 0.3
0
0.1
0.2
0.3
1
l/r
0

1
l/r
0
A
C
E
D
B
Figure 7.6 Relationship between conicities and virtual pivot position for negotiation of uniform
curve with zero longitudinal and lateral creep. h* = 1.25 m.
7.4 Stability of Vehicle with Uniform Conicity
In order to discuss stability attention will be concentrated on option (c) of Section 2
in which k
y2
= k
2
= 0 and k
y1
and the inter-wheelset stiffnesses k
yd
and k
t
are non-
zero. Moreover in this Section, the case where h
1
= h
2
= h* and
1
= will be con-
sidered, deferring discussion of non-uniform conicity to the following Section.
It is convenient to consider, firstly, the behaviour at low speeds. As already dis-
cussed, at low speeds, the car body degrees of freedom can be eliminated and the
equations of motion will be of the form of (6.4.20). Neglecting the small terms asso-
ciated with the contact stiffness, and assuming for simplicity that all the creep coef-
ficients are equal, the non-zero terms of B and C are
B
11
= B
33
= B
55
= 2f B
22
= B
44
= B
66
= 2fl
2
C
12
= C
34
= C
56
= -2f C
21
= C
65
= 2fl/r
0
C
43
= 2fl
1
/r
1
and the characteristic equation becomes
p
6
D
6
+ p
5
D
5
+ p
4
D
4
+ p
3
D
3
+ p
2
D
2
+ p
1
D + p
0
= 0 (1)
where D = d/Vdt, = l/r
0
and
1
=
1
l/r
0
and
p
6
= 1
p
5
= (l
2
+ 4h*
2
)k
t
/fl
2
+ (2l
2
+ h
1
2
+ h
2
2
)k
yd
/fl
2
p
4
= (2 +
1
)/ l
2
+ {2(l
2
+ 4h*
2
)(2l
2
+ h
1
2
+ h
2
2
) - (l
2
+ 2h*h
1
)
2
}k
yd
k
t
/2f
2
l
4
+ {(2l
2
+ h
1
2
+ h
2
2
)
2
- (l
2
- h
2
2
)
2
}k
yd
2
/4f
2
l
4
p
3
= (l
2
+ 4h*
2
)( +
1
)k
t
/f l
4
+ (
1
l
2
+ 3l
2
+ h
1
2
1
+ h
1
2
+ 2h
2
2
)k
yd
/f l
4
+ (3l
2
+ 8h*
2
)(3l
2
+ h
1
2
)k
yd
2
k
t
h
2
2
/4f
3
l
6
p
2
= ( + 2
1
)/l
4
+ {
1
( l
2
+ h
1
2
)
2
+ 8h
2
2
l
2
+ 2( l
2
+ h
1
2
)( l
2
+ h
2
2
)
- h
1
2
l
2
( - 1)
2
- 2h
1
h
2
l
2
( - 1)(1 -
1
)}k
yd
2
/4f
2
l
6
+ [(2h* - h
1
)(2h* - h
1
)l
2
+ 2(l
2
+ 4h*
2
){
1
(l
2
+ h
1
2
) + (l
2
+ h
2
2
)}
-
1
(l
2
+ 2h*h
1
)
2
]k
yd
k
t
/2f
2
l
6
p
1
= ( l
2
+
1
l
2
+ h
1
2
1
+ h
2
2
)k
yd
/f l
6
+
1
(l
2
+ 4h*
2
)k
t
/f l
6
+ {24h*
2
+ 4h*h
1
(
1
- ) + (
1
+ 2)(3l
2
+ h
1
2
)}k
yd
2
k
t
h
2
2
/4f
3
l
6
p
0
=
2
1
/ l
6
+ {4
2
h
2
2
-
1
h
1
2
( - 1)
2
- 2h
1
h
2
( - 1)(1 -
1
)}k
yd
2
/4f
2
l
6
+
1
(2h* - h
1
)(2h* - h
1
)k
yd
k
t
/2f
2
l
6
Figure 7.7 shows the variation of the eigenvalues with k, a factor applied to k
yd
and k
t
. When k = 0 there are three undamped oscillations A, B and C at wheelset
kinematic frequency = V(/lr
0
)
1/2
as discussed above. Introduction of stiffness
results in damping of these oscillations, the real part of the eigenvalue being ini-
tially proportional to k. One of these oscillations can be identified with a steering or
bending oscillation A of the complete vehicle, in which there is very little inter-
wheelset shear. As k is varied from zero to very large values the steering oscillation
remains lightly damped. As k increases, the frequency of the steering oscillation
decreases, tending towards a constant value for large values of k. The other two os-
cillations B and C involve much more shear of the inter-wheelset structure and are
heavily damped, splitting into four subsidences B1, B2, C1 and C2 as k increases.
The damping of one of these subsidences, C2, eventually decreases as k is increased
so that at large values of k it vanishes.
An important feature of these results is the small damping in the steering oscill-
Figure 7.7 Root locus for three-axle vehicle at low speeds as a factor k on the stiffnesses k
t
and k
yd
is varied, for the parameters of Table 7.1.
Table 7.1 Example parameters for three-axle vehicle.
suspension
k
y1
= 0.23 MN/m k
1
= k
2
= 0 k
1
= k
2
= 1MNm k
y2
= 0
c
y1
= 100 kNs/m c
1
= c
2
= 0 c
1
= c
2
= 50 kNms d
y2
= 0
k
yd
= 0.70 MN/m k
t
= 0.20 MN/m c
t
= c
yd
= 0
creep coefficients
f
11
= 8.83 MN f
22
= 8.06 MN f
23
=17.7 kNm W = 100.6 kN
vehicle geometry
h* = 4.125 m d
1
= d
4
= 0.2 m r
0
= 0.45 m l = 0.7452 m
inertia
m =1250 kg I = 700 kgm
2
m
c
=27000 kg I
xc
=32400 kgm
2
I
zb
= 269000 kgm
2
I
y
= 250 kgm
2

ation A and in the subsidence C2 at large values of k. In fact, both oscillatory insta-
bilities and divergence occur for certain values of the parameters, and the conditions
for stability will now be discussed in detail.
If k is very large, it is appropriate to neglect terms in k of order less than two in
(1) which then reduces to
p
4
D
4
+ p
3
D
3
+ p
2
D
2
+ p
1
D + p
0
= 0 (2)
where
p
4
= {2(l
2
+ 4h*
2
)(2l
2
+ h
1
2
+ h
2
2
) - (l
2
+ 2h*h
1
)
2
}k
yd
k
t
/2f
2
l
4
+ {(2l
2
+ h
1
2
+ h
2
2
)
2
- (l
2
- h
2
2
)
2
}k
yd
2
/4f
2
l
4
p
3
= (3l
2
+ 8h*
2
)(3l
2
+ h
1
2
)k
yd
2
k
t
h
2
2
/4f
3
l
6
p
2
= {
1
( l
2
+ h
1
2
)
2
+ 8h
2
2
l
2
+ 2( l
2
+ h
1
2
)( l
2
+ h
2
2
)
- h
1
2
l
2
( - 1)
2
- 2h
1
h
2
l
2
( - 1)(1 -
1
)}k
yd
2
/4f
2
l
6
+ [(2h* - h
1
)(2h* - h
1
)l
2
+ 2(l
2
+ 4h*
2
){
1
(l
2
+ h
1
2
) + (l
2
+ h
2
2
)}
-
1
(l
2
+ 2h*h
1
)
2
]k
yd
k
t
/2f
2
l
6
p
1
= {24h*
2
+ 4h*h
1
(
1
- ) + (
1
+ 2)(3l
2
+ h
1
2
)}k
yd
2
k
t
h
2
2
/4f
3
l
6
p
0
= {4
2
h
2
2
-
1
h
1
2
( - 1)
2
- 2h
1
h
2
( - 1)(1 -
1
)}k
yd
2
/4f
2
l
6
+
1
(2h* - h
1
)(2h* - h
1
)k
yd
k
t
/2f
2
l
6
If =
1
and h
1
=h
2
and k is taken to the limit then this reduces to
p
3
D
3
+ p
1
D = 0 (3)
the solutions of which consist of a purely imaginary root corresponding to an un-
damped oscillation frequency given by
I
2
= 9V
2
(l
2
+ 3h*
2
)/(8h*
2
+ 3l
2
)( h*
2
+ 3l
2
) (4)
and a zero root corresponding to a mode of neutral stability. This zero root implies
that the vehicle is capable of quasi-static misalignment as discussed in connection
with two-axle vehicles, Section 4.3, and the corresponding mode shape is shown in
Figure 7.8(a). The oscillation at frequency
I
is one in which the creep forces are in
overall equilibrium even though the motion of individual wheelsets is not one of
pure rolling. This is analogous to the behaviour of a rigid two-axle vehicle, Section
5.3. Figure 7.8(b) shows that the mode shape of this oscillation resembles the steer-
ing oscillation mentioned above.
Instability at low speeds can occur either by the real part of a complex conjugate
pair of eigenvalues becoming positive (oscillatory instability) or by a real eigenvalue
becoming positive (divergence). As the instabilities occur for large values of the
shear stiffnesses it is most instructive to use the perturbation method introduced in
Section 3.3. The terms in the expressions for the coefficients which are of order k
2
can be considered as small perturbations p
i
of the initial values p
i
* which are of
order k
3
. The corresponding perturbation in the ith eigenvalue then follows from
equation (3.3.12). Thus, from (2), the initial values p
i
* are
(a)
(b)
Figure 7.8 Mode shapes for three-axle vehicle, k
yd
= k
t
= (a) corresponding to zero root (b)
one cycle corresponding to s
i
= i
I
where
I
is given by equation (4).
p
4
* = 0
p
3
* = (3l
2
+ 8h*
2
)(3l
2
+ h*
2
)k
yd
2
k
t
h*
2
/4f
3
l
6
p
2
* = 0 (5)
p
1
* = {24h*
2
+ 3(3l
2
+ h*
2
)}k
yd
2
k
t
h*
2
/4f
3
l
6
p
0
* = 0
and
p
4
= (3l
4
+ 16h*
2
l
2
+ 12h*
4
)k
yd
k
t
/2f
2
l
4
+ (3l
2
+ h*
2
)(l
2
+ 3h*
2
)k
yd
2
/4f
2
l
4

p
3
= 0
p
2
= {(3l
2
+ h*
2
)(l
2
+ 3h*
2
) + h*
2
l
2
(1 + )
2
}k
yd
2
/4f
2
l
6
+ {(2 - )(2 - 1)h*
2
l
2
+ (3l
4
+ 16h*
2
l
2
+ 12h*
4
)}k
yd
k
t
/2f
2
l
6
(6)

p
1
= 0
p
0
= (1 + )
2
h*
2
k
yd
2
/4f
2
l
6
+ (2 - )(2 - 1)h*
2
k
yd
k
t
/2f
2
l
6
The approximations for the eigenvalues found using equation (3.3.12) are s
I
=

I
where
I
is given by equation (3) and
= - V(p
4
I
4
- p
2
I
2
+p
0
)/(- 3p
3
I
2
+ p
1
) (7)
and s
II
=

= - Vp
0
/p
1
. Equation (7) can be expressed in the form
= Vp
4
(
I
2
-
1
2
)(
I
2
-
2
2
)/2p
1
(8)
where
1
and
2
are the roots of
p
4
s
4
+ p
2
s
2
+ p
0
= 0 (9)
which are given by
1
2
= V
2
{(1 + )
2
h*
2
k
yd
+ 2(2 - )(2 - 1)h*
2
k
t
}/
{2(3l
4
+ 16h*
2
l
2
+ 12h*
4
)k
t
+ (3l
2
+ h*
2
)(l
2
+ 3h*
2
)k
yd
}
(10)

2
2
= V
2

Equation (7) shows that for stability either
1
<
I
<
2
or
1
>
I
>
2
(11)
This is identical to Leonhards criterion [18] and is, of course fully equivalent to
Rouths criterion; it provides a simple way of discussing the stability of the system.
From (4) and (10) it can be seen that
I
<
2
for all values of h* and l, independent
of . Therefore for stability
1
<
I
which for given k
yd
and k
t
is a quadratic in .
For practical values of the parameters this has solutions for small values of , which
can occur in practice, and very large values which are not of interest. It follows from
(4) and (10) that for very small values of stability cannot be achieved if k
yd
> 4k
t
.
For large values of k
yd
and h* >> l the condition that
1
<
I
implies that oscillatory
instability will only occur for low values of conicity, such as < 1/10. The mode
shape of the unstable mode is similar to that shown in Figure 7.8(a) and thus insta-
bility of the steering oscillation which occurs for large values of k
yd
and k
t
is pro-
moted by low conicity, inter-wheelset flexibility, and short wheelbase. Even where
it occurs, it should be noted that the real part of the corresponding pair of eigenval-
ues, though positive, remains small.
The criterion for static stability is p
0
> 0 or
3
+ (1+ )
2
k
yd
2
h*
2
/4f
2
+ (2 -)(2- 1)k
t
k
yd
h*
2
/2f
2
> 0 (12)
Divergence can only occur if <1/2 (or >2 which is beyond the practical range of
values), and for k
t
> k
yd
/4. As divergence only occurs when is small (12) reduces to
k
yd
/ k
t
> 2(2 - )(1 - 2)/ (1 + )
2
(13)
independent of wheelbase and creep coefficient. The mode of neutral static stability
which exists for very large values of k
yd
and k
t
is destabilised by the introduction of
flexibility between the wheelsets.
It is now possible to consider the influence of speed on the eigenvalues and this is
illustrated by the root locus shown in Figure 7.9. As speed is increased the eigenvalues
corresponding to branches A, B and C1and C2 are initially proportional to speed and
are closely equal to the solutions of (1). For the chosen set of parameters there is sta-
bility at low speeds. Similar to the case of the two-axle vehicle, the eigenvalues asso-
ciated with the vehicle body modes, D1, D2 and D3 are substantially independent of
speed and are closely equal to the wheels-fixed eigenvalues as illustrated in Table 7.2.
For the three-axle vehicle, like other configurations, at speeds for which the kinematic
or steering frequencies approach the natural frequencies of the vehicle body on the
suspension, there is the possibility of body instabilities. As in the case of the two-axle
vehicle, stability can be obtained by choosing suitable values for k
y1
, k
y2
, d
y1
, and d
y2
.
As a result, for the chosen set of parameters there is little interaction between the
wheelset modes and the modes involving the car body oscillating on the suspension so
body instabilities do not arise. In addition there are the usual 6 relatively large eigen-
values corresponding to the wheelset subsidences in lateral translation and yaw,
Figure 7.9 Root locus as speed is varied for three-axle vehicle with parameters of Table 7.1
except that k
t
= k
yd
= 0.1 MN/m.
At Speed, V = 10 m/s Wheels - Fixed Label
1 -1290 Wheelset subsidence
3,4 -1377 6.066i Wheelset subsidence
9 -45.30 -48.19 Car body yaw
10 -2.415 -2.415 Car body yaw
13 -14.18 Shear subsidence C1
15,16 -1.796 5.722i Shear oscillation B
17,18 -0.7354 5.770i Steering oscillation A
Table 7.2 Eigenvalues for three-axle vehicle with parameters of Table 7.1 except that k
t
= k
yd
= 0.1 MN/m.
two of which appear at E and X in Figure 7.9. As the speed is increased, the steering
oscillation A loses stability at a bifurcation speed V
B
= 48.7 m/s.
Figure 7.10 summarises stability as a function of k
yd
and k
t
, for the extreme cases
of = 0.5 and = 0.05 with the creep coefficients halved. Analogous to the case of
the two-axle vehicle, the behaviour is different depending on the value of k
yd
/2fl
1/2
and k
t
/2fl
1/2
. The stiffness to ground provided by the yaw stiffness for the two-
axle vehicle is provided by k
yd
and k
t
for the three axle vehicle. As already dis-
cussed, stability depends on non-zero values of k
yd
and k
t
; if k
yd
= 0, there will be
two undamped eigenvalues at wheelset kinematic frequency at low speed, one mode
involving motion of the central wheelset, and the other mode involving motion of
the outer wheelsets such that there is zero shear between the outer wheelsets. If k
t
=
0 there will be one undamped eigenvalue at wheelset kinematic frequency at low
speed, the motion involving zero shear between adjacent wheelsets.
As mentioned above, a suitable choice of parameters has been made so that the
body instabilities do not occur. If both k
yd
and k
t
are non-zero, examination of the
root locus has shown that it is the steering oscillation, branch A, that becomes un-
stable above the bifurcation speed V
B
. V
B
is increased if k
yd
and k
t
are increased. This
region is labelled O1 in Figure 7.10. This instability is analogous to wheelset insta-
bility in which the inertia forces induced by the steering oscillation drive the oscilla-
0.2 0.8 1
50
100
150
k
yd
(MN/m)
V (m/s)
= 0.5
0.8 1
50
100
150
k
yd
(MN/m)
V (m/s)
= 0.05 f x 0.5
0.2
0.8 10
50
100
150
k
t
(MN/m)
V (m/s)
0.2 0.8 1
50
100
150
k
t
(MN/m)
V (m/s)
0.2
0
0
S
O1
O1
S
S
D
O2
0
0
S
S
O1
O2
D
O1
Figure 7.10 Joint effect of conicity and creep coefficient on stability as a function of k
t
and
k
yd
for three-axle vehicle with the parameters of Table 7.1. D = divergence, O = oscilla-
tory instability and S = stable.
tion, there being little involvement of the car body. Low conicity and creep coeffi-
cients, stabilise the instability O1, but a separate instability O2 is introduced where
the critical speed reduces with increasing k
yd
and k
t
. The mode shape of this oscilla-
tion is similar to that of the low speed steering oscillation shown in Figure 7.8(b)
and involves significant yaw of the car body. In addition, the combination of low
conicity and creep coefficient introduces divergence over a range of small values of
k
yd
and above a limiting value of k
t
in accordance with equation (13). For very large
values of k
yd
and k
t
, beyond the range of values in Figure 7.10, there is a further re-
gion of oscillatory instability which occurs at low speeds in accordance with equa-
tion (11).
In the light of these results, it might be thought that the selection of values for
k
yd
and k
t
to obtain the maximum critical speed for a range of conicity and creep
coefficient is a difficult compromise. However, it should be borne in mind that a
number of the instabilities, both dynamic and static, discussed above are associated
with eigenvalues which small in magnitude. This is particularly true when the sys-
tem has low values of equivalent conicity. In these circumstances, judgement of
parameter values should be informed by nonlinear response studies.
7.5 Stability with Unequal Conicities
The preceding results apply to a configuration in which the wheelset conicities are
uniform and the inter-wheelset shear structure is symmetric so that the shear spring
k
yd
is located midway between the wheelsets. As discussed above, it is necessary to
vary the spring position h
1
(measured from the leading wheelset) if the conicities are
not uniform in order to achieve perfect steering. However, it is convenient to con-
sider the influence of h
1
and conicity on stability separately.
Figure 7.11 shows the effects on stability of variations of h
1
, and
1
for the
vehicle with parameters of Table 7.1, based on the linearised equations of motion.
Both oscillatory and static instabilities occur. From equation (4.1), for static stability
p
0
> 0 so in the case where =
1
then
2
+ {4h
2
2
+ h
1
( - 1)
2
(2h
2
- h
1
)}k
yd
2
/4f
2
+ (2h* - h
1
)(2h* - h
1
)k
yd
k
t
/2f
2
> 0 (1)
Without discussing (1) in detail, it can be seen in general terms that movement of
the pivot towards the central wheelset and increasing k
t
all promote divergence. This
is illustrated by the results in Figure 7.11(c).
Two regions of oscillatory instability occur as h
1
is varied and are shown in Fig-
ure 7.11(c). Movement of the position of the spring k
yd
towards either the outer or
inner wheelsets is destabilising and oscillatory instability occurs at all speeds when
either h
1
= 0 or h
1
= 2h*. This is as expected as s = iV(/r
0
l)
1/2
is then a solution of
the characteristic equation for all values of ,
1
, k
yd
and k
t
, an undamped kinematic
oscillation occurring because of the lack of elastic restraint.
As indicated in Figure 7.11(c) the steering oscillation of the example three-axle ve-
hicle, with h
3
= 3 m., loses its stability at small amplitudes at V =168.4 m/s. The ei-
genvalues are given in Table 7.3 and the eigenvector corresponding to the mode in
which = 0 is given in Table 7.4. This solution is a sub-critical bifurcation leading
to a limit cycle in which the lateral displacement of all three wheelsets have a simi-
lar amplitude of 5 mm at V= 178 m/s, the details of which are shown in Figure 7.12.
This shows the numerical solution of the nonlinear equations of motion for the vehi-
cle with the parameters of Table 7.1 and the wheel rail geometry of Figure 2.2. Fig-
ures 7.12(a) and (b) indicate that the limit cycle is in the form of a steering oscilla-
tion very similar to that predicted by the small amplitude solution and there are
many points of similarity with the corresponding results for the unrestrained bogie
shown in Figure 5.21. The displacements of the car body are small. The frequency
of the oscillation is 7.46 Hz and is closely equal to that given by the small amplitude
eigenvalues. Figure 7.12(c) shows the variation of the rate of rotation of the leading
wheelset which is about 0.1% of V/r
0
. This fluctuates at twice the frequency of the
oscillation as a major component of the torque applied to the wheelset is the product
of the longitudinal creep force and the rolling radius both of which are varying at the
frequency of the oscillation. Figures 7.12(d) and (e) show the longitudinal and lateral
creepages for the right hand wheel of the leading wheelset. As the variation in the
wheelset rotational speed is small, reference to equation 2.5.13 shows that the largest
0.1 0.3
0
0.5
100
200
300

V (m/s)
D S
O
0.1 0.3 0 0.5
100
200
300

1
V (m/s)
S
O
O
0 4 8
100
200
300
h
1
(m)
V (m/s)
D S
O O
(c)
(a)
(b)
Figure 7.11 The influence of unequal conicities and effective pivot point on stability for the
three-axle vehicle with parameters of Table 7.1. D = divergence; O = oscillatory instabil-
ity; S = stable.
Table 7.3 Eigenvalues for the three-axle vehicle of Table 7.1, h
1
= 3 m, V = V
B
= 168.4 m/s.
Table 7.4 Eigenvector for mode 10 of the isolated bogie of Table 7.1 , V = V
B
= 168.4 m/s.
contribution to the longitudinal creep is made by the variation in rolling radius with
lateral displacement and this is reflected in the waveform. Similarly, reference to
equation 2.5.15 shows that the largest contribution to the lateral creep is made by the
angle of yaw, as indicated by the waveform. Figure 7.12(f) shows the spin of the
leading wheelset, and reference to equation 2.5.17 shows that the spin behaves like
r
/r
0
and so the waveform reflects the geometrical nonlinearity of the contact slope
variation with lateral displacement. Figures 7.12(g) and (h) shows the longitudinal
and lateral creep forces for the right hand wheel of the leading wheelset, and consid-
eration of their resultant and the normal force indicate that slipping is occurring over
part of the cycle of the oscillation. Figure 7.12(i) shows the variation of the normal
force acting on the right hand wheel which is in phase with the lateral displacement
and is relatively large in magnitude.
1,2 -41.04 269.1i Large wheelset roots
3.4 -38.29 159.7i Large wheelset roots
5 -176.7 Large wheelset root
6 -147.8 Large wheelset root
7,8 -36.03 83.43i Shear oscillation
9 -79.52 Shear subsidence
10,11 47.50i Steering
12 -46.98 Car body yaw
13,14 -2.44 9.32i Car body upper sway
15,16 -4.30 3.68i Car body lower sway
17 -4.64 Shear subsidence
18 -2.42 Car body yaw
No. Eigenvalue(s
-1
) Label
y
1
1

1
-0.2454 + 0.3839i
y
2
-0.2901 - 0.8140i

2
0.1594 + 0.09210i
y
b
-0.01508 + 0.001898i

b
0.002628 - 0.00006271i

b
0.07224 - 0.05877i
y
2
-1.0462 - 0.19032i

2
0.4814 - 0.4148i
Figures 7.11(a) and (b) show the effect of variation of and
1
on stability. For
static stability p
0
> 0 so if h = h
1
then from equation (4.1)
1
+ (1 + ){(1 +
1
) + -
1
}k
yd
2
h
2
/4f
2
+
1
(2 - )(2 - 1)k
yd
k
t
h
2
/2f
2
> 0 (2)
Reduction of promotes divergence as shown in Figure 7.11(a)
In order to get a feel for the behaviour of the system, it is useful to consider the
simple case of a vehicle in which k
t
= 0 and k
yd
is very large. Figure 7.13 shows the
root locus as h
1
and
1
are varied. Solution of (4.1) for this case shows that there are
two purely imaginary eigenvalues when =
1
and h
1
= h
2
. One of these, A in Fig-
ure 7.13, corresponds to kinematic oscillations of the wheelsets in which the phase
between the wheelset motions is determined by the shear connection between the
wheelsets so that the shear spring is unstressed. The other, B in Figure 7.13, refers to
a steering oscillation in which creep occurs. The frequency of this steering oscilla-
tion, when =
1
and h
1
= h
2
, is given by

1
2
= (1 + )
2
V
2
h*
2
/(3l
2
+ h*
2
)(l
2
+ 3h*
2
) (3)
It can be seen that even if the conicity is zero, a steering oscillation occurs in which
the wavelength is solely determined by the geometry of the vehicle.
Consider the variation of the conicities, but maintaining h
1
= h
2
. As
1
is reduced in
t (s) 0 0.1 0.2
-0.01
0
0.01
0 0.1 0.2
-.005
0
.005
0 0.1 0.2
-1
0
1
0 0.1 0.2
-5
0
.005
0 0.1 0.2
-.005
0
.005
0 0.1 0.2
-0.5
0
0.5
0 0.1 0.2
-20
0
20
0 0.1 0.2
-20
0
20
0 0.1 0.2
0
50
100
y
1 y
2
y
3
y
b
b
-V/r
0
-
1
1r1
2r1
3r1
3l1
T
1r1
(kN) T
2r1
-T
3r1
(a)
(b) (c)
(d)
(e)
(f)
(g)
(h) (i)
Figure 7.12 Displacements, creepages and forces in the limit cycle of three-axle vehicle of
Table 7.1 (except that h
3
= 3 m) at V = 178 m/s.
value, the two eigenvalues approach each other and, at a critical value, frequency
coalescence takes place, further reductions in
1
leading to eigenvalues of the form
i. This form of instability may be termed flutter as it is closely related to
instabilities experienced with other non-conservative systems such as those arising
in Aeroelasticity [20]. It can be shown [11] that the equations of motion of the sys-
tem with k
t
= 0 and k
yd
very large are of the same form as those of undamped aeroe-
lastic systems which have been extensively studied [20].
Increasing
1
beyond the value
1
= results in the natural frequencies diverg-
ing in value. The frequency of the steering oscillation, branch B, reduces in value
until

1
= 2(1 + )/(1 - 2
2
) (4)

consistent with (3), when it becomes zero. Further reductions in the value of
1
re-
sult in eigenvalues of the form and divergence occurs.
Now, in this simple case of a vehicle in which k
t
= 0 and k
yd
is very large, con-
sider the variation of the position of the inter-wheelset shear spring k
yd
, but main-
taining =
1
. Solution of (4.1) for this case shows that there are two purely imagi-
nary eigenvalues. Irrespective of the position of the pivot, one of these corresponds
to kinematic oscillations of the wheelsets, and the other refers to a steering oscilla-
tion. The frequency of this steering oscillation is given by
Figure 7.13 Root locus as h
1
and
1
are varied for three-axle vehicle of Table 7.1 with =
0.025 and k
t
=0.

2
2
2
2 2 2
2 1
2
2
1
2
2
2 2 2
2
2 2
1 1
2
=
+
+ +
V h h h
l h h l h
{ ( ) ( ) ( ) }
{( ) ( ) }
(5)
When the effective pivots of the shear springs move towards the central axle, the
frequency of the steering oscillation reduces in value until
h
1
/2h = 2/(3 - ) (6)

when it becomes zero. Further increases in h
1
result in roots of the form and di-
vergence occurs. As the pivots move toward the outer axles, the frequency of the
steering oscillation reaches a maximum and then decreases in value until it vanishes
at a point outside the wheelbase when
h
1
/2h = -2 /(1 - 3 ) (7)

Further outward movement of the pivots result in roots of the form and diver-
gence occurs.
These results show that for the three-axle vehicle there are many potential modes
of instability. However, the lateral stiffness and damping between the car body and
wheelsets may be chosen appropriately to eliminate the body instabilities, in a simi-
lar way to that prescribed for the two-axle vehicle. Then stability may be obtained
up to high speeds providing that extreme values of conicity and inter-wheelset stiff-
nesses are avoided. The historical difficulties with the stability of three-axle vehicles
with inter-wheelset linkages can be ascribed to the linkages either being deficient in
that the criteria of equation (6.4.15) are not satisfied or that linkages were too stiff.
7.6 Dynamic Response
In addition to problems with stability, experience with three-axle vehicles has shown
that whilst steady-state curving may be improved, the form of the inter-wheelset con-
nections may give rise to large steering errors on non-uniform curves [5]. Accord-
ingly, it is important to consider the dynamic response of three-axle vehicles to track
geometry and the following examples illustrate the principles involved.
Figures 7.14 and 7.15 show the dynamic response of the three-axle vehicle of
Table 7.1 (except that k
yd
= k
t
= 0.1 MN) to entry into a curve of radius R
0
= 225 m,
with a cubic parabolic transition of length L
0
= 20 m as computed from the full
nonlinear equations of motion. The vehicle speed V = 15 m/s and the coefficient of
friction = 0.3. The track is canted so that the cant deficiency is zero throughout.
The results may be compared with the results for the elastically restrained wheelset
in Section 3.7 and the results for the two-axle vehicle in Section 4.6. Figure 7.14
shows the motion of the wheelsets which after an initial transient on the transition
take up a radial attitude with an outwards lateral displacement closely equal to that
necessary to negate longitudinal creep as given by equation (3.2.23). In the transient,
the wheelsets attempt to follow the curve, and the steering errors are quite small.
However, on very sharp curves flange contact readily takes place, and steering errors
can be large. This effect can be pronounced on reverse curves as has been found in
the past. Figure 7.15 shows the time history of the creep forces acting on the leading
wheelset. As in the case of a single wheelset, once the steady-state motion is estab-
lished, the lateral creep forces are mainly generated by spin and are largely cancelled
out by the gravitational stiffness force, and the longitudinal creep forces are small.
However, during the transition the deviations of the wheelsets from a radial attitude
induce elastic forces generated by the inter-wheelset suspension which have to be
0 2 4
-0.005
0
0.005
0 2 4 0 2 4
y
1
y
2
y
3
3
t (s)
Figure 7.14 Dynamic response of the wheelsets of the three-axle vehicle with the parameters
of Table 7.1 (except that k
yd
= k
t
= 0.1 MN) to curve entry with a cubic parabolic transi-
tion of length 20 m, V = 15 m/s and R
0
= 225 m.
0 2 4
-6
-4
-2
0
2
0 2 4
-10
-5
0
5
0 2 4
-10
0
10
0 2 4
-10
0
10
y
1
(mm)
1
(mr)
t (s)
T
2r1
(kN)
T
2l1
(kN)
T
1r1
(kN)
T
1l1
(kN)
T
2r1
+ T
2l1
z
y
W
Figure 7.15 Dynamic response of the leading axle of the three-axle vehicle with the parame-
ters of Table 7.1 (except that k
yd
= k
t
= 0.1 MN) to curve entry with a cubic parabolic
transition of length 20 m, V = 15 m/s and R
0
= 225 m.
reacted by longitudinal creep forces, which therefore reach a peak and then subside
to their steady-state value. Figure 7.16 presents the corresponding results for the
same vehicle but with the nominal values of the stiffnesses k
yd
and k
t
( k
yd
= 0.7
MN, k
t
= 0.2 MN). The increased stiffness of the inter-wheelset structure results in
larger longitudinal creep forces both in the steady-state and the transition.
It is clear from these indicative results that the three-axle vehicle offers the pos-
sibility of improved steady-state curving in which the wheelsets take up a closely
radial position, but the dynamic response in transitions and reverse curves requires
careful consideration. Thus, the trade-off between stability and curving is not com-
pletely eliminated.
References
1. Liechty, R.: Das Bogenlaufige Eisenbahn-Fahrzeug. Schulthess, Zurich, 1934, p.
23.
2. White, J.H.: The American Railroad Freight Car. The John Hopkins Press, Bal-
timore, 1993, p. 168.
3. Fidler, C.: British Patents 2399, 3825, 1868.
0 2 4
-10
-5
0
0 2 4
-15
-10
-5
0
5
0 2 4
-20
-10
0
10
20
0 2 4
-2
0
2
y
1
(mm)
1
(mr)
t (s)
T
2r1
(kN)
T
2l1
(kN)
T
1l1
(kN)
T
1r1
(kN)
z
w
W (kN)
T
2r1
+ T
2l1
Figure 7.16 Dynamic response of the leading axle of the three-axle vehicle with the parame-
ters of Table 7.1 to curve entry with a cubic parabolic transition of length 20 m, V = 15
m/s and R
0
= 225 m.
4. Elsner, H.: Three-Axle Streetcars. N.J. International, Hicksville, 1994, Vol. 1,
Chapter 1.
5. Elsner, H.: Three-Axle Streetcars. N.J. International, Hicksville, 1994, Vol. 1, p.
60.
6. Liechty, R.: Das Bogenlaeufige Eisenbahn-Fahrzeug. Schulthess, Zurich, 1934,
p. 31.
7 Wickens, A.H.: Steering and dynamic stability of railway vehicles. Vehicle System
Dynamics 5, No. 1-2 (1975), pp. 15-46.
8. Wickens, A.H.: Stability criteria for articulated railway vehicles possessing per-
fect steering. Vehicle System Dynamics 7, No.1 (1979), pp. 33-48.
9. Wickens, A.H.: Static and dynamic stability of a class of three-axle railway vehi-
cles possessing perfect steering. Vehicle System Dynamics 6, No.1 (1977), pp. 1-19.
10. Wickens, A.H.: Flutter and divergence instabilities in systems of railway vehi-
cles with semi-rigid articulation. Vehicle System Dynamics 8, No.1 (1979), pp. 33-
48.
11. Wickens, A.H.: The stability of a class of multi-axle railway vehicles possessing
perfect steering. In: K. Magnus (Ed.): Proc. IUTAM Symposium on Dynamics of
Multibody Systems, Munich, August-September 1977, pp. 345-356. Springer-
Verlag, Berlin, 1978.
12. Wickens, A.H.: Static and dynamic stability of unsymmetric two-axle railway
vehicles possessing perfect steering. Vehicle System Dynamics 11, No. 2 (1982), pp.
89-106.
13. Wickens, A.H.: Stability optimisation of multi-axle railway vehicles possessing
perfect steering. ASME Journal of Dynamic Systems Measurement and Control 110,
No.1 (1988), pp. 1-7.
14. Wickens, A.H.: Static and dynamic stability of a generalised symmetric three-
axle railway vehicle possessing perfect steering. Archives of Transport Quarterly 1,
No.2 (1989), pp. 139-160.
15. de Pater, A.D.: Optimal design of a railway vehicle with regard to cant defi-
ciency forces and stability behaviour. Delft University of Technology, Laboratory
for Engineering Mechanics, Report 751, 1984.
16. de Pater, A.D.: Optimal design of railway vehicles. Ingenieur-Archiv 57, No.1
(1987), pp. 25-38.
17. Keizer, C.P.: A theory on multi-wheelset systems applied to three wheelsets. In
O. Nordstrom (Ed.): The Dynamics of Vehicles on Roads and Tracks. Proc. 9th
IAVSD Symposium, Linkoping, June 1985. Swets and Zeitlinger Publishers, Lisse,
1986, pp. 233-249.
18. Porter, B.: Stability criteria for linear dynamical systems. Oliver and Boyd, Ed-
inburgh, 1967, p. 37.
19. Bolotin, V.V.: Non-conservative problems of the theory of elastic stability. Per-
gamon, Oxford, 1963.
20. Done, G.T.S.: The flutter and stability of undamped systems. British Aeronauti-
cal Research Council Reports and Memorandum No.3553, 1966.
8
Articulated Vehicles
8.1 Introduction
The economics of railways in which expensive infrastructure is justified by large
traffic flows requires the operation of either many vehicles with short headways or
very large vehicles to provide the required capacity. Hence the concept of the train.
In many cases the interaction between the vehicles in a train is minimised by the
form of coupling between the vehicles, so that longitudinal forces can be transmitted
between car bodies, but the coupler is capable of transmitting little or no lateral
force or yaw couple. In this case it is a good approximation to treat each vehicle as
if it were isolated and the lateral dynamics of each vehicle can be considered to be
largely independent of that of the rest of the train. However, the need to improve
curving performance, maximise the use of the clearance gauge, minimise axle loads,
reduce mass, aerodynamic drag and cost has led to many designs in which there is
articulation of the car bodies of a vehicle or train, so that the connections between
vehicles form an essential part of the running gear. In this Chapter articulated vehi-
cles, in which the relative motion between the car bodies is used to influence the
stability and guidance of the vehicle, are considered.
The first articulated locomotive was designed by Horatio Allen in 1832 [1].
Though this had a short career, it probably stimulated several of the articulated de-
signs for the Semmering Contest in 1851. Thereafter, there was a succession of ar-
ticulated locomotives the development of which is described by Weiner [2], and
which sought to resolve the conflict between the long wheelbase made necessary by
high power and the large curvature of many railway lines. Most of these had un-
symmetric fore-and-aft configurations which are considered in Chapter 9.
In the early days of the railways, it had become customary to link together two
and three axle vehicles not only by couplings but also by side chains to provide yaw
restraint between adjacent car bodies in order to stabilise lateral motions [3]. The
need to lengthen vehicles stimulated measures which allowed bending in plan view.
As an alternative to the use of the bogie, in 1837 W.B. Adams proposed an articu-
lated two-axle carriage [4]. The first three-axle vehicle with articulated car body was
proposed by Fidler in 1868 [5]. In Fidlers patent the wheelset is mounted on the car
body at the point where the car body is tangential to the curve, and the central
wheelset is mounted on a steering beam, Figure 8.1(a). Machlachan [6] proposed a
similar configuration in 1878 but made the significant addition of a shear connection
236
(a)
(b)
(c)
(d)
(e)
Figure 8.1 Historical articulated railway vehicle configurations. (a) Fidler, 1868. (b) Machla-
chan, 1878. (c) Barber, 1907. (d) Liechty, 1931. (e) Configuration using steering linkage.
or pivot between the car bodies, Figure 8.1(b). In Barbers 1907 patent [7] the outer
bodies are pivoted together but instead of the wheelset being mounted on the body
at the point of tangency, the outer wheelsets are freely pivoted and are connected to
the central wheelset by cross-bracing, Figure 8.1(c). The steering beam has disap-
peared. A similar objective is achieved by Liechtys 1931 patent [8] in which the
outer wheelset is mounted on an arm pivoted on the car body and actuated by the
steering beam, Figure 8.1(d). A similar scheme has been used recently in the Boa de-
sign [9]. An alternative approach is to use a linkage, Figure 8.1(e), similar to the ar-
rangement used in body-steered bogies, but driven by the angle between the car bodies.
In trains of bogie vehicles, the use of the Jacob bogie which is shared by adja-
cent car bodies reduced mass (and cost and drag) of an articulated rake. However,
the resulting fixed consist was disliked by some operators, and individual body
lengths had to be shorter to maintain limits on the throwover (lateral displacements
of the car bodies) on curves. Nevertheless, articulation was successfully exploited
by Gresley [10] in a number of carriage designs between 1900-1930. In 1939, Sta-
ARTICULATED VEHICLES 237
(a)
(b)
(c)
(d)
(e)
Figure 8.2 Examples of some of the many configurations of vehicles with articulated car bod-
ies and single-axles. Equivalent bogie versions are common.
nier employed a double point articulation, one at each end of the bogie, which al-
lowed longer car bodies [11]. Application to a new generation of passenger trains in
the U.S. in 1930-50 was less successful, and generally articulated trains acquired a
reputation for bad riding. More recently, articulation of bogie vehicles has been
used, most successfully, in high speed trains such as the TGV [12].
Another important modern development is the use of articulation on vehicles
with single axle running gear, as discussed in Chapters 4 and 5. The dynamic be-
haviour of the Talgo train is influenced by its unsymmetric configuration and the
problems arising will be considered in Chapter 9. Current examples of trains which
have single-axle running gear such as the Copenhagen S-Tog [13] embody forced
steering of the wheelsets through mechanical linkages or hydraulic actuators driven
by the angle between adjacent car bodies. Extensive design calculations were car-
ried on this train [13] and its lateral stability is also discussed in [14]. Another recent
example is provided by the Wien trams [15].
A wide range of articulated configurations have been used in practice, particu-
larly for trams, and some of the variations are shown in Figure 8.2. This figure is
238
based in part on [16]. The simplest form of articulated vehicle is one with two car
bodies supported on three or four axles. Not only has this been a common configu-
ration, particularly for trams, but its study reveals many of the dynamic characteris-
tics of articulated trains with many axles. For a vehicle with three or more axles it
was shown in Chapter 7 that stability could be achieved without compromising
steering on uniform curves, the stiffness necessary for stability being provided in a
way that does not impede radial steering. This provides a theoretical basis for the
discussion of articulated vehicles exploiting single-axles in this Chapter.
As in the case of the bogie vehicle considered in Chapter 6, two broad ap-
proaches to configurations can be distinguished. The first may be termed forced
steering (steering derived from relative motions between car bodies) and the second
self steering (steering derived from wheelset motions only). Various possible future
stages in the development of single-axle suspensions may be envisaged, in which
mechanical linkages are progressively displaced by systems with actuators and sen-
sors and either passive or active control.
8.2 Steering and Stability
As a wide variety of vehicles with articulated car bodies exist, it will be necessary to
consider a limited number of variants, so the behaviour of a number of representa-
tive configurations with three and four axles will be discussed. Figure 8.2(a) shows
two two-axle vehicles coupled together; Figure 8.2(b) and 8.2(c) show similar con-
3
y
3
y
d
d
h
y
c
c
y
2
2
y
b
b
y
1
1
h
1
c
y
2
2
y
1
1
y
b
b
y
3
3
y
4
4
y
d
d
c
h h
Figure 8.3 Generalised coordinates for articulated vehicles.
239
k
1
k
y1
/2
k
y2
/2
k
yb k
yb
k
2
k
b
k
1
k
y1
/2
k
2
k
2
k
1
k
b
k
y2
/2 k
y2
/2 k
y1
/2
k
yb
k
1
(a)
(b)
Figure 8.4 Simplified arrangement of suspension showing basic stiffnesses for articulated
vehicles.
figurations but with one axle removed; Figure 8.2(d) and (e) show symmetric three-
axle vehicles with either two or three car bodies, including vehicles with two car
bodies carrying a steering beam on which the central wheelset is mounted. Compari-
son of the behaviour of these various configurations will give insight into the dy-
namics of articulated vehicles in general.
Consider the equations of motion of these configurations. It is clear that the con-
figurations of Figure 8.2(b) and (c) are derivable from that of Figure 8.2(a) by the
deletion of a wheelset, and such unsymmetric configurations will be discussed in
Chapter 9. Moreover, the configurations of Figure 8.2(d) and (e) are equivalent in
terms of their degrees of freedom. Consequently, making the same assumptions that
have been made in the case of the two-axle vehicle and other configurations, the
motions of these vehicles may be defined by the generalised coordinates shown in
Figure 8.3.
The form of the elastic stiffness matrix will be determined by applying the crite-
ria for stability of a vehicle capable of perfect curving given in Chapter 6. The solu-
tion of equations (6.4.10) will initially be discussed for the three-axle vehicle with
three car bodies of Figure 8.2(d) so that neglecting the variation of the rotational
speed of the wheelsets
q = | y
1
1
y
b
b
y
2

2
y
c

c

c
y
3

3
y
d

d

d
|
T
(1)
so that y
b
, y
d
and y
c
refer to lateral translation of the vehicle bodies and
b
,
d
, and
c
refer to yaw of vehicle bodies. The other y
i
,
i
are standard wheelset coordinates. The
equations of motion will be of the form of (2.11.23) where
ARTICULATED VEHICLES
240
F = block diagonal[ F
1
O
33
F
2
O
33
F
3
O
33
] (2)
and O
33
is the 3 3 null matrix, and F
i
refers to the ith wheelset. The inertia matrix is
A = diag [m I m
b
I
xb
I
zb
m I m
c
I
xc
I
zc
m I m
b
I
xb
I
zb
] (3)
The component stiffness and compatibility matrices for the basic form of secondary
suspension shown in Figure 8.4(a) are
k = diag[ k
y1
k
|1
k
1
k
yb
k
|b
k
b
k
y2
k
|2
k
2
k
yb
k
|b
k
b
k
y1
k
|1
k
1
] (4)
a
d h
d h d h c
d
=

1 0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0
0 1 0 0 1 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 1 0 0 0 0 0
0 0 0 1 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 1 0 0 0
1
2 1 3 1
4
0 0
0 0 0 0 0 0 0 1 0 0 1
0 0 0 0 0 0 0 0 1 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 1 0 0 0 0 1
0 0 0 0 0 0 0 0 0 0 1 0 1
0 0 0 0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 1
3 1 2
1

(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
d c h d h
d h
1
(5)
Then equations (6.4.10) become
k
y1
( y
1
- y
b
+ d
1
b
- h
b
) = 0 (6)
k
1
b
= 0 (7)
k
1
(
1
-
b
) = 0 (8)
k
yb
{ y
b
- d
2
b
- h
1
b
- y
c
+ d
3
c
- (c - h
1
)
c
} = 0 (9)
k
|b
(
b
-
c
) = 0 (10)
k
b
(
b
-
c
) = 0 (11)
241
k
y2
(- y
c
+ d
4
c
+ y
2
) = 0 (12)
k
2
c
= 0 (13)
k
2
(
c
-
2
) = 0 (14)
k
yb
{ y
d
- d
2
d
+ h
1
d
- y
c
+ d
3
c
+ (c - h
1
)
c
} = 0 (15)
k
|b
(
d
-
c
) = 0 (16)
k
b
(
d
-
c
) = 0 (17)
k
y1
(y
3
- y
d
+d
1
d
+ h
d
) = 0 (18)
k
1
d
= 0 (19)
k
1
(
3
-
d
) = 0 (20)
From equations (7), (13) and (19)
b
=
c
=
d
= 0. From equations (8), (11), (14),
(17), and (20)
either
1
=
b
or k
1
= 0 (21)
either
c
=
b
or k
b
= 0 (22)
either
2
=
c
or k
2
= 0 (23)
either
d
=
c
or k
b
= 0 (24)
either
3
=
d
or k
1
= 0 (25)
For the case where the outer wheelsets have radius r
0
and conicity and the inner
wheelset has radius r
1
and conicity
1
, for steering on a uniform curve without lateral
and longitudinal creep, the displacements measured from the unstrained position are
y
1
= y
3
= - lr
0
/R
0
+ (c + h )
2
/2R
0
(26)
y
2
= - lr
1
/
1
R
0
(27)
1
= -
3
= (c + h )/R
0
(28)
2
= 0 (29)
Because of symmetry y
d
= y
b
,
d
= -
b
, and
c
= 0 so that from (23) the value of
242
k
2
can be arbitrarily chosen. From (12), either k
y2
= 0 or y
c
= y
2
. Choosing the latter
case, from (6), (9), (26) and (27)
y
b
= {- h
1
lr
0
/- hlr
1
/
1
+ h
1
(c + h )
2
/2}/R
0
( h + h
1
) (30)
b
= {- lr
0
/+ lr
1
/
1
+ (c + h )
2
/2}/R
0
( h + h
1
) (31)
If the condition (21) is satisfied by putting
1
=
b
then, from (28) and (31),
- lr
0
/ + lr
1
/
1
+ (c + h )( c - h - 2h
1
)/2 = 0 (32)
or, if the inner and outer wheelsets have the same radius and conicity
h
1
= (c - h)/2 (33)
If this condition is satisfied the conditions for perfect steering can be satisfied with a
nonzero yaw stiffness k
1
.
The remaining conditions (22) and (24) requires that k
b
= 0. This implies that
there are two main possibilities for symmetric articulated three-axle vehicles. The
first has geometry which satisfies the geometric requirement (32) or (33) and the
second does not satisfy this requirement so that k
1
= 0. In addition, there is, of
course, the further possibility of the vehicle with a single car body which has been
discussed in Chapter 7.
An analogous analysis can be carried out for the configurations of Figure 8.2(a),
where the basic stiffnesses are shown in Figure 8.4(b). As this vehicle consists of
two-axle vehicles coupled together with simple suspension elements, it is not neces-
sary to reiterate the equations of motion. The conditions for steering give the simple
result
y
1
= y
4
= - lr
0
/R
0
+ (c + h )
2
/2R
0
(34)
y
2
= y
3
= - lr
1
/
1
R
0
+ (c - h )
2
/2R
0
(35)
1
= -
4
= (c + h )/R
0
(36)
2
= -
3
= (c - h )/R
0
(37)
y
b
= y
d
= - lr
0
/2R
0
- lr
1
/2
1
R
0
+ (c
2
+ h
2
)/2R
0
(38)
b
= -
d
= - lr
0
/2hR
0
+ lr
1
/2
1
hR
0
+ c/R
0
(39)
Similar results, of course, apply to the configurations of Figures 8.2(b) and 8.2(c).
Figure 8.5 indicates the geometry for each of these configurations in which the
wheelsets adopt a radial position with radial displacements appropriate to their
conicity, and the car bodies take up the positions connecting the wheelsets without
243
strain of the suspension elements. Thus, for small displacements and large radius
curves, steering without lateral and longitudinal creep is obtained because of the
combined action of creep and conicity which is the primary mechanism of guid-
ance. It is evident that on curves with moderate and large curvature, a similar analysis
may be carried out based on the full nonlinear equations of motion, cf. Section 3.2.
These configurations satisfy the conditions for steering so it is now necessary to
consider stability. For simplicity it will be assumed that all wheelsets have the same
radius and conicity. The necessary condition for the vehicle to be able to steer with-
out lateral and longitudinal creep and be dynamically stable at low speeds is given
by (6.4.15).
For the three-axle configuration of Figure 8.4(a) if the geometrical condition
(33) is not satisfied and the yaw stiffness k
1
is zero, the degeneracy of the basic
stiffness matrix E is 7, so that in this case three additional stiffnesses are required to
(c + h)
2
/2R
0
h
1
y
1
l/R
0
r
0
y
b
b
h
c
(a)
(b)
(c + h)
2
/2R
0
c
h
l/R
0
r
0
y
1
y
b
b
(c)
Figure 8.5 Attitude on curve of various articulated configurations.
244
reduce the degeneracy to 3 so that the criterion may be satisfied. For configurations
where the geometrical condition (33) is satisfied and there are nonzero yaw stiff-
nesses k
1
, the degeneracy of E is 4, so that in this case only one additional stiffness
is required to satisfy the criterion.
The basic stiffness matrix E for the four-axle configuration of Figure 8.4(b) has
degeneracy 7 and therefore requires 4 additional stiffnesses in order to satisfy the
criterion.
Thus additional stiffnesses must be provided, and this is considered next.
8.3 Application to Specific Configurations
Based on the configurations discussed in Section 1, Figure 8.6 shows an extension of
Figure 8.4 in which further elastic connections have been introduced. Figure 8.6(a) and
(b) show the body-steered and self-steered versions of the three-axle vehicle and Fig-
ure 8.6(c) and (d) show the corresponding versions of the four-axle vehicle.
Firstly, it is useful to consider the early schemes shown in Figure 8.1 in the light
of the above criteria. As mentioned in Chapter 7, pivots were originally used to pro-
vide lateral restraint and yaw freedom, instead of elastic elements. For example, in
Fidlers patent for the three-axle vehicle, Figure 8.1(a), the geometric condition
(2.33) is satisfied and k
y1
, k
y2
, k
yb
, k
1
, and k
2
, were very large though the principles
discussed above remain valid. As mentioned above, E has degeneracy of 4 and the
vehicle would be unstable at low speeds. The steering beam is equivalent to the cen-
tral car body of Figure 8.2(d). In Machlachans scheme, Figure 8.1(b), there is a
shear connection or pivot between the outer car bodies, which may be regarded as
an additional stiffness k
yc
. From the above discussion, and assuming that the geo-
metrical condition (2.33) is satisfied, the addition of an additional row and corre-
sponding stiffness k
yc
would have the effect of stabilising the vehicle so that all ei-
genvalues had negative real parts at low speeds. In Barbers scheme, Figure 8.1(c),
the outer bodies are connected by a shear stiffness k
yc
but the geometrical condition
(2.33) is not satisfied. Instead the outer wheelsets are freely pivoted and are con-
nected to the central wheelset by cross-bracing, with shear stiffness k
yd
. In this case
the three stiffnesses k
yc
and k
yd
(leading and trailing bay) would ensure that the de-
generacy of E would be reduced to 3 as required by the stability criterion. A similar
objective is achieved by Liechtys scheme, Figure 8.1(d) in which the outer wheel-
set is controlled by a steering beam. Again, the steering beam is the central car body
of Figure 8.2(d) steered through springs of stiffness k
d
and k
e
, but if the steering
beam is small its independent coordinates can be eliminated and its effect repre-
sented as shown in Figure 8.7. It is therefore an equivalent arrangement to a steering
linkage connecting the yaw angle of the leading wheelset with the angle between
adjacent car bodies.
Thus, these additional stiffnesses are provided in two distinct ways, consisting of
the basic configuration as defined in Section 3 with, either body steering or self steer-
ing. In the former case the outer wheelsets are steered by a linkage, of stiffness k
st
,
from the angle between the outer and central car body, the third additional
245
k
s
=diag [ k
st
k
st
k
yc
]
a
c h h h c h h
c h h c h h h
d c d c
s
=
+ + + +
+ + +

(
(
(
0 0 0 2 0 0 0 0 2 0 0 0 0 0
0 0 0 0 0 0 0 0 0 2 0 0 0 2 2
0 0 1 0 0 0 0 0 0 0 1
1 1
1 1
5 5
( )
k
st
= e
1
e
2
c
2
2
(c
3
- c
1
)
2
/{e
1
(c
2
- c
1
)
2
+ e
2
(c
3
- c
1
)
2
} c
2
(c
3
- c
1
)/c
3
(c
2
- c
1
) = (c + h)/(c - h -2h
1
)
e
1
k
yc
c
1 c
2
c
3
e
2
(a)
k
yd
k
yd
k
t
(b)
k
s
= diag[ k
yd
k
yd
k
t
]
a
h c h h
c h h h
c h c h
s
=
+
+

(
(
(
0 0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 1 0 0 0
1 0 0 0 0 0 0 0 0 1 0 0 0
3 3
3 3
k
s
= [ k
st1
k
st2
k
st2
k
st1
]
k
st1
k
st1
k
st2
k
st2
(c)
a
h c h c
h c h c
h c h c
h c h c
s
=

+
+

(
(
(
(
0 1 0 0 1 2 0 0 0 0 0 0 2 0 0
0 0 0 0 1 2 0 1 0 0 0 0 2 0 0
0 0 0 0 2 0 0 0 1 0 0 1 2 0 0
0 0 0 0 2 0 0 0 0 0 0 1 2 0 0
/ /
/ /
/ /
/ /
k
yd
k
yd
k
ye
k
t
k
yf
k
yf
(d)
k
s
= [ k
yd
k
yd
k
t
k
ye
k
yf
k
yf
]
a
h h
h h
c h c h
c h c h
c c
c c
s
=

+ +

(
(
(
(
(
(
(
(
1 0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0 1
1 0 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 1 1 0 0 0 0 0
1 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 1
Figure 8.6 Additional stiffnesses for articulated vehicles.
246
c
1 c
1
c-h
c
2
b

2
O
P
R
S
Q
h
c
c c c h
c 2 2
1 2
2
=
+ +
a
s
= [ 0 0 0 0 -1+ h/2c 0 1 0 0 0 0 -h/2c 0 0 ]
Figure 8.7 Steering beam applied to rear wheelset of leading vehicle showing the equivalence
to a steering linkage.
stiffness required for stability is provided by k
c
, as shown in Figure 8.6(a). In the
latter case the adjacent wheelsets connected by a shear spring k
yd
and the outer
wheelsets connected by a shear spring k
t
, as shown in Figure 8.6(b).
Similarly, there are equivalent schemes for the four-axle vehicles with two car
bodies as shown in Figures 8.6(c) and (d).
For long articulated vehicles direct connection of the wheelsets may seem diffi-
cult to implement, but it is possible to utilise passive sensors and actuators replicat-
ing the stiffness and damping characteristics of mechanical linkages and arranged so
as to be de-coupled from the motions of the car bodies.
8.4 Stability of an Articulated Three-Axle Vehicle
As might be expected, there is considerable similarity between the behaviour of the
articulated three-axle vehicle and the three-axle vehicle with single car body consid-
ered in Chapter 7. Firstly, the scheme in which the adjacent wheelsets connected by
a shear spring k
yd
and the outer wheelsets connected by a shear spring k
t
, as shown
in Figure 8.6(b) will be considered in detail.
Consider the influence of speed on the eigenvalues of the configuration of Fig-
ure 8.6(b) with the nominal set of parameters given in Table 8.1. The root locus as
speed is varied is shown in Figure 8.8, and values for the eigenvalues at V = 10 m/s
are given in Table 8.2. It was shown in Section 6.4 that, at low speeds, the equations
of motion can be separated into uncoupled sets involving either the wheelset co-
ordinates or the car body coordinates. The resulting system of equations involving
the wheelset coordinates is of exactly the same form, at low speeds, as those derived
in Section 7.2 for the three-axle vehicle with single car body and, in particular, the
form of the stiffness matrix E* is as given in (7.2.22-23). It follows that the system
has six eigenvalues proportional to speed consisting of a steering oscillation A, a
247
Table 8.1 Example parameters for three-axle articulated vehicle.
suspension of basic system
k
y1
= k
y2
= 0.23 MN/m k
1
= 0 k
|1
= k
|2
= 1MNm
c
y1
= 200 kNs/m c
1
= c
2
= 0 c
|1
= c
|2
= 50 kNms c
y2
= 50 kNs/m
k
yb
= 1 MN/m k
b
= 0 k
|b
= 0
c
yb
= 0 c
b
= 0 c
|b
= 0
additional suspension for body-steered system
k
st
= 30 kNm k
yc
= 1 MN/m k
2
= 2 MNm c
b
= 0.1 MNms c
st
= c
yc
= c
2
= 0
additional suspension for self-steered system
k
yd
= 0.1 MN/m k
t
= 0.1 MN/m k
2
= 5 MNm c
t
= c
yd
= c
2
= c
b
= 0
vehicle geometry
r
0
= 0.45 m l = 0.7452 m
h = 2 m h
1
= 3.75 m c = 6.25 m d = 0.2 m
creep coefficients
f
11
= 8.83 MN f
22
= 8.06 MN f
23
=17.7 kNm
inertia
m =1250 kg I = 700 kgm
2
m
b
=10000 kg I
xc
=12000 kgm
2
I
zb
= 130000 kgm
2
m
c
=7000 kg I
xc
=8400 kgm
2
I
zb
= 9000 kgm
2
I
y
= 250 kgm
2
shear oscillation B and two shear subsidences C1 and C2 which are closely equal to
the solutions of equation (7.4.1). For the chosen set of parameters there is stability at
low speeds. A further set of eigenvalues is associated with wheelset motions and con-
sist of six large subsidences (three roots are equal to -2f/mV and three roots are equal
to -2fl
2
/IV) and these are not shown in Figure 8.8. The eigenvalues associated
Figure 8.8 Root locus of the three-axle configuration of Figure 8.6(b) with the nomi-
nal set of parameters given in Table 8.3.
248
Table 8.2 Eigenvalues for articulated three-axle vehicle with parameters of Table 8.1.
At Speed, V = 10 m/s Wheels - Fixed Label
7,8 -0.6423 45.74i -0.02164 45.83i D9
9,10 -2.368 22.60i -2.445 22.60i D8
12,13 -2.909 10.40i -2.909 10.40i D7
14,15 -2.259 8.833i -2.181 8.927i D6
16,17 -2.162 8.801i -2.165 8.797i D5
18,19 -3.820 7.333i -0.2211 7.246i D4
20,21 0.2710 5.210i Steering oscillation A
22,23 -0.8677 5.601i Shear oscillation B
25,26 -3.199 4.351i -3.266 4.448i D2
27,28 -1.303 3.329i -1.317 3.305i D1
29,30 -3.162 4.389i -3.012 4.471i D3
Figure 8.9 Natural modes of articulated vehicle with parameters of Table 8.3 with wheels fixed
and zero suspension damping, D = 0.
249
V =
8.5
e =
4.829
73.0 37.36
78.9 42.26
Figure 8.10 Mode shapes of the critical modes at the bifurcation speeds of articulated vehicle
with parameters of Table 8.3, showing one cycle of the oscillation.
with the vehicle body modes, D1-D9, are substantially independent of speed except
at speeds for which the kinematic or steering frequencies approach the natu-
ral frequencies of the vehicle body on the suspension. They are closely equal to the
wheels-fixed eigenvalues, as illustrated in Table 8.2. The mode shapes of these
wheels-fixed modes are shown in Figure 8.9 for the case where the suspension
damping is zero. It is noteworthy that many of the wheels-fixed modes are com-
pletely uncoupled from the wheelset motions but in other cases there is considerable
interaction, leading to body instability. For the given parameters, body instability
occurs at a bifurcation speed V
B1
= 8.5 m/s and the associated mode shape is shown
in Figure 8.10(a). As in the case of the two-axle vehicle, stability can be obtained by
choosing suitable values for k
y1
, k
y2
, d
y1
, and d
y2
, but in addition inter-car body
dampers may be effective. At higher speeds, the shear oscillation B loses stability at
a bifurcation speed V
B2
= 73.0 m/s. As shown in Figure 8.10(b), this is analogous to
the wheelset instability, there is little motion of the car bodies and is predicted accu-
rately if the car bodies are assumed fixed. A further instability at a bifurcation speed
V
B3
= 78.9 m/s involves yaw of the central car body and the mode shape is shown in
Figure 8.10(c).
250
Figure 8.11 Variation of the imaginary part of the eigenvalues for the three-axle configura-
tion of Figure 8.6(b) with the nominal set of parameters given in Table 8.1.
1
0
80
0.5
80
80
1
80
80
80
0
0
0 0.5
0.5 0
0
0.5
0.5
V (m/s)
V (m/s)
V (m/s)
k
yd
(MN/m) k
yd
(MN/m)
k
t
(MN/m)
k
yd
(MN/m)
k
t
(MN/m)
k
2
(MNm) k
2
(MNm)
= 0.5 = 0.05, f x 0.5
O2
O1
S
S
D
O1
O1
S
S
S S
O1
O2
D
O2
Figure 8.12 Joint influence of conicity, creep coefficient and stiffnesses k
yd
, k
t
and k
2
on sta-
bility for articulated three-axle vehicle of Fig. 8.6(b) with parameters of Table 8.1 except
that c
y1
= c
y2
= c
yb
= 0.1 MNs/m, k
yd
= 0.7 MN/m, k
t
= 0.2 MN/m and k
2
= 5 MNm. S de-
notes stability, D divergence and O1 and O2 oscillatory instabilities.
251
0.4
0
80
= 0.5
V (m/s)
1 0
80
k
yc
(MN/m)
V (m/s)
S
1 0
80
k
y2
(MNm)
V (m/s)
S
S
0.4
0
80
= 0.5 f/2
k
st
(MNm)
V (m/s)
S
1 0
80
k
yc
(MN/m)
V (m/s)
S
1 0
80
V (m/s)
k
st
(MNm)
S
A
B B
C
k
y2
(MNm)
O
O
O
O
O
O
Figure 8.13 Joint influence of conicity, creep coefficient and stiffnesses k
st
, k
yc
and k
2
on
stability for articulated three-axle vehicle of Figure 8.6(a) with forced steering with pa-
rameters of Table 8.3. S denotes stability and O oscillatory instability.
Figure 8.11 shows the variation of the imaginary parts of the eigenvalues with speed
for a range of low speeds. Frequency coincidence occurs between the steering oscil-
lation A and shear oscillation B and the car body on suspension modes, but in only
one case in the chosen speed range does instability occur. Clearly, coupling of the
wheelset modes and the car body modes depends not only on a degree of frequency
coincidence but also on compatibility of mode shape for the energy transfer that wo-
uld sustain the amplitude of the oscillation. In this connection, the zig-zag mode
shape of the car body mode D4 is particularly likely to couple with the steering os-
cillation, and this illustrates the tendency of articulated vehicles to suffer this form
of body instability.
The joint influence of the stiffnesses k
yd
, k
2
and k
t
, conicity and creep coefficient is
shown in Figure 8.12. This may be compared with Figure 7.10 which shows a similar
plot for the three-axle vehicle with a single car body. If any of the stiffnesses k
yd
, k
t
and
k
2
are zero then the critical speed is zero in accordance with the discussion in the pre-
ceding Section. For large values of the conicity, instability is confined to smaller val-
ues of k
yd
, k
2
and k
t
. The instability, labelled O1 in Figure 8.12, is the body instability
discussed above. The instability O2 is analogous to the wheelset instability. For low
values of the conicity and creep coefficients, divergence occurs for a range of values of
k
yd
and for larger values of k
t
. Moreover, the instability O2 occurs at lower speeds.
252
Turning to the scheme of Figure 8.6(a) in which the wheelsets are connected by
a steering linkage to the car bodies, it is found that the root locus is qualitatively
similar to the root locus for the variant with direct connections between the wheel-
sets, but there is a very lightly damped steering oscillation at low speeds. Three ad-
ditional points arise. Firstly, as might be expected, in this case there is more interac-
tion between the car body modes and the wheelset modes. Secondly, the addition of
yaw damping, c
b
, between the car bodies is effective in eliminating the body insta-
bility. Thirdly, very large values of c
b
are effective in stabilising the system at low
speeds, and increasing the damping in the steering mode, even when k
yc
= 0 and the
stability criteria discussed in Section 3 are not satisfied. However, the bifurcation
speeds associated with the wheelset modes are little affected.
Hence Figure 8.13 shows the influence of the stiffnesses k
st
, k
yc
and k
2
on sta-
bility for extreme values of the conicity and creep coefficient for the scheme of Fig-
ure 8.6(a) and the parameters of Table 8.1. In accordance with the discussion above
if the contact stiffnesses were neglected then for stability at low speeds either k
yc
or
k
2
is needed in addition to k
st
. If both k
yc
and k
2
are zero, the critical speed is de-
termined by the contact stiffnesses and would be low. Initially, increases in k
st
in-
crease the critical speed, A in Figure 8.13(a). Beyond a certain value of k
st
the criti-
cal speed is not influenced by k
st
, B in Figure 8.13(a), because the mode of instabil-
ity involves mainly kinematic motion of the central wheelset. For large values of k
st
,
an instability involving interaction with the lateral bending mode of the car bodies
occurs. This instability occurs for smaller values of k
st
in the case of small values of
conicity and reduced creep coefficient as shown in Figure 8.13(b). Thus, in this case
as in others that have already been discussed, there is a conflict between the require-
ment that k
st
must be sufficiently large in order to achieve a large margin of stability at
high conicity, but this will tend to encourage instability for low values of conicity and
creep coefficient. Figures 8.13(c)-(f) show that both k
2
and k
yc
are effective in stabilis-
ing the system for low conicities and creep coefficients, but not at high conicities.
8.5 Stability and Response of an Articulated Four-Axle Vehicle
For the configuration of Figure 8.6(c) and the nominal set of parameters given in
Table 8.3, the root locus as speed is varied is shown in Figure 8.14. As might be
expected, in this case there are eight eigenvalues proportional to speed representing
four oscillations at kinematic frequency O. These oscillations are stable at low
speeds provided that elastic stiffness has been provided in accordance with the pre-
scription of the Section 3. In addition there are the usual set of eigenvalues also as-
sociated with wheelset motions and consisting of eight large subsidences (at low
speeds, four roots are equal to -2f/mV and four roots are equal to -2fl
2
/IV) which
become heavily damped oscillations at higher speeds as shown at S in Figure 8.14.
As usual, the eigenvalues associated with the vehicle body modes, D, are more or
less independent of speed except at speeds for which the kinematic or steering fre-
quencies approach the natural frequencies of the vehicle body on the suspension.
They are closely equal to the wheels-fixed eigenvalues. As in the case of the three-
axle vehicle,
253
Table 8.3 Example parameters for four-axle articulated vehicle
suspension of basic system
k
y1
= k
y2
= 0.23 MN/m k
1
= k
2
= 0 k
|1
= k
|2
= 1MNm k
yb
= 40 MN/m
c
y1
= 50 kNs/m c
1
= c
2
= 0 c
|1
= c
|2
= 50 kNms c
y2
= c
yb
= 0
additional suspension for body-steered system
k
st1
= 2.5MN/m k
st2
= 2.5MN/m d
st1
= d
st2
= 0
additional suspension for self-steered system
k
yd
= 1MN/m k
t
= 0.1MN/m k
yf
= 0.1MN/m k
ye
= 1MN/m
c
t
= c
yd
= c
yf
= 0
vehicle geometry
r
0
= 0.45 m l = 0.7452 m
h = 3.5 m c = 6 m d = 0.2 m
creep coefficients
f
11
= 7.44 MN f
22
= 6.79 MN f
23
=13.7 kNm
inertia
m =1250 kg I = 700 kgm
2
m
b
=10000 kg I
xc
=12000 kgm
2
I
zb
= 130000 kgm
2
m
c
=7000 kg I
xc
=8400 kgm
2
I
zb
= 9000 kgm
2
I
y
= 250 kgm
2
Figure 8.14 Root locus of the four-axle configuration of Figure 8.6(c) with the nominal set
of parameters given in Table 8.3.
254
it is noteworthy that many of the wheels-fixed modes are completely uncoupled
from the wheelset motions but in other cases there is major interaction, potentially
leading to body instability. One of the oscillations at kinematic frequency, shown at
O1 in Figure 8.14, has very low damping and is associated with near coincidence
between one of the wheels-fixed modes and a steering mode. At higher speeds, all
four steering modes lose stability at bifurcation speeds between 85.5 and 89.5 m/s.
This is analogous to the wheelset instability, there is little motion of the car bodies
and is predicted accurately if the car bodies are assumed fixed. Then application of
equation (3.3.15) shows that in this case it is the stiffness of the steering linkage k
st1
which determines the bifurcation speed.
A full survey of the stability boundaries will not be given here, but as an exam-
ple Figure 8.15 shows stability as a function of speed and the stiffness in the steer-
ing linkage k
st
(= k
st1
=k
st2
), for three values of the equivalent conicity and a case
of reduced creep coefficient. If k
st
is zero then the system is unstable at low speed
as already discussed. For small values of k
st
instability O1 occurs above a certain
speed. For large values of k
st
another form of instability O2, which involves rela-
tively large amounts of yaw of the car bodies, occurs for which low conicity and
creep coefficient is destabilising. This again illustrates the difficult trade-off for
many configurations of railway vehicle, for if a design has to cater for a wide range
20
0
100
20
0
100
20
0
100
20
0
100
V (m/s)
V (m/s)
k
st
(MNm)
k
st
(MNm)
O1
S
O2
S
k
st
(MNm)
k
st
(MNm)
S
S
V (m/s)
V (m/s)
= 0.5 = 0.2
= 0.05
= 0.05 f x 0.5
Figure 8.15 The effect of variations of conicity and creep coefficient on stability as a function
of the stiffness of the steering linkage k
st
and speed for the four-axle vehicle with the pa-
rameters of Table 8.5. S = stability, O1 and O2 instability.
255
of conicities, comparison of the boundaries for high and low conicity in Figure 8.15
shows that the speed range for stability is severely reduced.
These results can be compared with those for the configuration of Figure 8.6(d),
for a set of nominal parameters, given in Table 8.3, comparable with those used in
the discussion of the configuration of Figure 8.6(c), in which the wheelsets are con-
nected directly and not through the car bodies. Figure 8.16 shows the root locus as
speed is varied, and can be compared with Figure 8.14. In this case the eight eigen-
values which are proportional to speed, as suggested by the analysis of Section 6.4,
consist of two conjugate complex pairs corresponding to steering oscillations,
shown as O1 and O2 in Figure 8.16 and four real roots, shown at R. One of the lat-
ter is very small indicating a condition of marginal static stability, though for the
chosen set of parameters there is stability at low speeds in accordance with the
discussion of Section 3. The lightly damped kinematic oscillation of the body
steered configuration is absent. The usual set of eigenvalues associated with
wheelset motions consists of six large subsidences (three roots are equal to -2f/mV
and three roots are equal to -2fl
2
/IV) and these are not shown in Figure 8.16. The six
pairs of eigenvalues associated with the vehicle body modes are substantially
independent of speed except at speeds for which the kinematic or steering
frequencies approach the natural frequencies of the vehicle body on the sus-
pension and they are closely equal to the wheels-fixed eigenvalues. These are shown
as D1-D5 in Figure 8.16. (D6 is off the scale). It is noteworthy that in the present
case the wheels-fixed modes are completely uncoupled from the wheelset motions.
motions.re 8.17 shows stability as a function of speed and the stiffnesses k
yd
(= k
ye
),
k
yf
and k
t
for two extreme values of the equivalent conicity and creep coefficient.
Note that in accordance with the discussion of Section 3, stability at low speeds is
obtained if either k
t
or k
yf
are zero. However, an appropriate choice of stiffnesses
Figure 8.16 Root locus of the four-axle configuration of Figure 8.6(d) with the nominal set
of parameters given in Table 8.3.
256
additional to the minimum required for stability at low speeds enhances stability.
Compared with the body steered vehicle, the range of stability is much increased for
larger values of conicity and there is considerable scope for optimisation. However,
divergence occurs for smaller values of k
yd
and larger values of k
yf
and k
t
when the
conicity is low and the creep coefficient is reduced. It was seen above that this be-
haviour is characteristic of some articulated vehicles having three axles. The condi-
tions for static stability depend in a complicated way on the inter-wheelset stiff-
nesses, as indicated for three-axle vehicles.
In order to assess the practical significance of these results it is necessary to con-
sider a typical solution of the complete nonlinear equations of motion for the di-
rectly steered system of Table 8.3. Figures 8.18(a) and (b) show, for the set of pa-
rameters indicated, the dynamic response to an initial condition y
1
(0) = 4 mm ap-
plied to the lateral displacement of the leading wheelset. The response is dominated
by a lightly damped oscillation, following this transient, by the vehicle taking up a
steady-state attitude in which overall balance of the creep forces is achieved. Figure
8.18(d) shows the stability boundaries, consistent with those shown in Figure 8.17,
in the k
yd
(= k
ye
) - plane. As is a function of the wheelset lateral displace-
ment, it can be
Figure 8.17 Qualitative diagram showing the joint influence of conicity, creep coefficient
and stiffnesses k
yd
(= k
ye
), k
yf
and k
t
on stability for a directly steered four-axle articu-
lated vehicle. S denotes stability, D divergence and O oscillatory instability.
= = 0.05 f /2
V
k
yd
k
yd
V
k
yf
k
yf
V
k
t
k
t
S
D
S
O
S
S
O
D
O
D
S
S
257
0 2 4 6
-6
-4
-2
0
2
4
t (s)
y (mm)
0 2 4 6
-1.5
-1
-0.5
0
0.5
1
t (s)
(mr)
8 8
1
4
3
b d
2
2
d
3 1
4
b
(a)
(b)
(c)
1 2
0
0.1
0.2
k
yd
= k
ye
(MN)
A
B
A1
C
D
C1
S
O1
O2
D
(d)
Figure 8.18 Dynamic response and stability of vehicle with low conicity subjected to
initial condition of 4 mm applied at leading wheelset. (a) and (b) lateral and yaw
displacements (c) steady-state attitude of vehicle (d) stability chart. Parameters of
Table 8.3 except that k
yd
= k
ye
= 1 kN/m, k
t
= 2 MN/m, k
f
= 1 MN/m, creep coeffi-
cients halved and wheel-rail geometry consistent with = 0.065.
258
Figure 8.19 Dynamic response of linkage steered vehicle of Figure 8.6(d) to curve entry.
Speed 15 m/s, length of transition 20 m., curve radius 225 m cant deficiency zero. y
1
lat-
eral displacement,
1
yaw angle, T
1r
and T
1l
longitudinal creep forces on right and left, T
2r
and T
2l
lateral creep forces, for leading wheelset.
Figure 8.20 Dynamic response of self-steered vehicle of Figure 8.6(c) to curve entry. Speed
15 m/s, length of transition 20 m., curve radius 225 m cant deficiency zero.
2
0
-10
t (s)
0
3
y
1
(mm)
1
(mr)
-5
-10
10
T
1l
T
1r
T
2r
T
2l
T
1l
5
-15
(kN)
(kN)
y
1
(mm)
2
-10
0
t (s)
0 3
1
(mr)
0.5
-1.5
0
1.5
-1.5
0
(kN)
(kN)
2
-10
0
T
1l
T
1r
T
2r
T
2l
259
seen that if A corresponded to the starting point of the transient, the motion would
decay towards the stability boundary at B, and in this case this point corresponds to
a zero real root. Reference to equation (3.7.5) shows that for the equivalent linear
system with a zero real root the response to an impulse tends to a motion in the
mode corresponding to the zero real root. In fact, Figure 8.18(c) shows that the atti-
tude of the vehicle is closely similar to that indicated by the corresponding linear
system. If C were the starting point, the motion would decay to point D and a limit
cycle. If A1 and C1 were the starting points then the motions would grow until the
steady state is established at B and D respectively. It is clear that because, in this
case, the instabilities O2 and D are associated with low conicity, displacements and
creep forces will be small. This is not the case with the instability O1.
The self-steered configuration therefore allows considerable freedom in the se-
lection of the parameters. While both configurations can be expected to perform
well on curves of constant radius, the forces generated in a transition or reverse
curve will be influenced by the suspension parameters and there is conflict between
the requirements for stability and dynamic response. This can be illustrated by com-
paring the dynamic response in curve entry for examples of the two configurations.
Figure 8.19 shows a typical dynamic response of the linkage-steered vehicle of
Figure 8.6(d) on entry to a curve in which the transition has linearly increasing cur-
vature. The track is canted and is being traversed at the speed for zero cant defi-
ciency. On the transition, the varying angles between the wheelsets and the car bod-
ies induces forces in the steering linkage so that the leading wheelset fails to steer
and moves out beyond the rolling line, large lateral and longitudinal creep forces
being induced.
These forces are significantly reduced when running on the uniform part of the
curve as the wheelsets take up a radial position and attempt to move out to the roll-
ing line. For a much smaller radius curve, the wheelsets are still able to take up a
radial position but because of the restricted flange-way clearance, cannot move out
to the rolling line and hence significant creep forces are generated. If the cant defi-
ciency is not zero then further yaw movements of the wheelsets are necessary and
the creep forces are increased.
In the case of the self-steered configuration of Figure 8.6(c) it is possible to se-
lect suitably low values of the stiffnesses which, whilst maintaining stability for a
range of conicities, also give low forces in the transition to a uniform curve. This is
illustrated in Figure 8.20.
References
1. White, J.H.: American Locomotives, Revised Ed., The John Hopkins Press, Bal-
timore, 1997, p. 509.
2. Weiner, L.: Articulated Locomotives, Constable, London, 1930.
3. Fryer, C.E.J.: A History of Slipping and Slip Carriages, Oakwood, Oxford, 1997
p. 7.
260
4. White, J.H.: The American Railroad Passenger Car, The John Hopkins Press,
Baltimore, 1978, p. 627.
5. Fidler, C.: British Patents 2399, 3825. 1868.
6. Liechty, R.: Das Bogenlaeufige Eisenbahn-Fahrzeug, Schulthess, Zurich, 1934,
p. 31.
7. Barber, T.W.: British Patent 24632, 1907.
8. Liechty, R.: British Patent 390036, 1933.
9. Anon.: International Railway J. September, 1988. p. 26.
10. Jenkinson, D.: British Railway Carriages of the 20
th
Century, Volume 1: The
End of an Era 1901-22. Patrick Stephens, Wellingborough, 1988, p. 155.
11. Jenkinson, D.: British Railway Carriages of the 20
th
Century, Volume 2: The
Years of Consolidation 1923-53. Patrick Stephens, Wellingborough, 1990, p. 197.
12. Tachet, P. and Boutonnet, J.-C.: The Structure and Fitting Out of the TGV vehi-
cle Bodies. French Railway Techniques 21, No. 1 (1978), pp. 91-99.
13. Rose, R.D.: Lenkung und Selbstlenkung von Einzelradsatzfahrwerken am
Beispiel des KERF im S-Tog Kopenhagen. Proc. 4
th
International Conference on
Railway Bogies and Running Gears, Budapest, 1998, pp. 123-132.
14. Slivsgaard, E. and Jensen, J.C.: On the dynamics of a railway vehicle with a
single-axle bogie, Proc. 4
th
Mini Conference on Vehicle System Dynamics, Identifi-
cation and Anomalies, Budapest, 1994, pp. 197-207.
15. Anon.: Wien Light Rail Vehicles by Duwag-Bombardier. Railway Gazette In-
ternational, September, 1992, p. 586.
16. Charlton, E.H.: Articulated Cars of North America. Light Railway Transport
League, London, 1966, Figure 34.
9
Unsymmetric Vehicles
1 Introduction
Though early railway passenger and freight vehicles were generally symmetric, lo-
comotives early adopted an unsymmetric configuration in order to maximise the
axle-load on driving wheels and make use of the available adhesion. Additional
smaller wheelsets were soon provided to improve the steadiness of running. It has
already been mentioned, in Chapter 6, how the introduction of the leading swivel-
ling bogie improved both stability and curving behaviour, and it was a matter of
experience that the running behaviour of these unsymmetric configurations was
strongly dependent on the direction of motion [1].
A later example of an unsymmetric vehicle design, used in trams, is provided by
maximum traction trucks devised in the 1890s and in which the driving wheels
were followed or preceded by pony wheels which had a diameter about two-thirds
of the driving wheels [2]. In this case each bogie was unsymmetric but the complete
vehicle was usually symmetric.
The classic work on the stability of unsymmetric railway vehicles is by Carter
[3] which was directed toward the configurations then current in railway practice.
Carter applied Routh's stability theory, not only to electric bogie locomotives, then
exhibiting many problems of instability, but also to a variety of steam locomotives.
In his mathematical models, a bogie consists of two wheelsets rigidly mounted in a
frame, and locomotives comprise wheelsets rigidly mounted in one or more frames.
Following Carters first paper of 1916 the theory was elaborated in a chapter of his
1922 book [4]. Carters next paper [5] gave a comprehensive analysis of stability
within the assumptions mentioned above. As he was concerned with locomotives the
emphasis of his analyses was on the lack of fore-and-aft symmetry characteristic of
the configurations he was dealing with, and he derived both specific results and de-
sign criteria. Carter's work expressed, in scientific terms, what railway engineers had
learnt by hard experience, that stability at speed required rigid-framed locomotives
be unsymmetric and uni-directional.
His analysis of the 0-6-0 locomotive found that such locomotives were unstable
at all speeds if completely symmetric and he comments that this class of locomotive
is much used in working freight trains; but is not employed for high speed running
on account of the proclivities indicated in the previous discussion.
Carter analysed the 4-6-0 locomotive both in forward and reverse motion and
found that in forward motion beyond a sufficiently high speed or sufficiently stiff
bogie centring spring (laterally connecting the bogie to the locomotive body)
oscillatory instability occurs, but as the mass of the bogie is small compared with the
main mass of the locomotive, the resulting oscillation was unlikely to be dangerous
at ordinary speeds. Carters stability diagrams, the first of their kind in the railway
field, are shown in Figure 9.1(a) and 9.1(b).
In reverse motion, Figure 9.1(b), Carter found that beyond a certain value of
the centring spring stiffness buckling of the wheelbase occurred which would tend
to cause derailment at the leading wheelset. As this wheelset is incorporated in the
main frame of the locomotive, the lateral force acting between wheel and rail would
be proportional to the mass of the main frame and would be correspondingly large,
and potentially dangerous. This was the explanation of a number of derailments at
speed of tank engines such as the derailment of the Lincoln to Tamworth mail train
at Swinderby on 6 June 1928, as discussed in his final paper [6].
Carters analysis of the 2-8-0 with a leading Bissel, or single-axle bogie, similarly
explained the need for a very strong aligning couple for stability at high speed, whilst
noting that in reverse motion a trailing Bissel has a stabilising effect for a large and
useful range of values of aligning couple. Rocard [7,8] also considered unsymmetric
configurations and proposed the use of different conicities fore-and-aft.
A general theory for the stability of unsymmetric vehicles and the derivation of
theorems relating the stability characteristics in forward motion with those in reverse
motion was given in [9,10]. In the case of articulated two-axle vehicles at low
speeds it was shown that a suitable choice of elastic restraint in the inter-wheelset
connections results in static and dynamic stability in forward and reverse motion and
which will steer perfectly, without any modification dependent on the direction of
motion. However, the margin of stability is small.
A further practical result of Carters work was a series of design measures, the
subject of various patents [11], for the stabilisation of symmetric electric bogie lo-
comotives, because the introduction of the symmetric electric locomotive had been
accompanied by a more or less common experience of lateral instability at high
speed. In particular, Carter suggested an arrangement of the running gear in which
1 2 3 4
20
40
60
k
y
(MN/m)
V (m/s)
S
O
1 2 3 4
20
40
60
k
y
(MN/m)
V (m/s)
S
D
O
0 0
(a) (b)
Figure 9.1 Carters stability diagram for the 4-6-0 locomotive in (a) forward motion and (b)
reverse motion. k
y
is the centring stiffness. (Recalculated in modern units from [5]). S =
stable; O = oscillatory instability; D = divergence.
UNSYMMETRIC VEHICLES 263
the configuration was altered depending on the direction of motion. This makes it
easier to resolve the conflict of stability and steering with an adequate margin of
stability. Such re-configurable systems have been studied by Li [12].
Chapter 3 discusses the use of the pony axle (a wheelset mounted on a leading or
trailing arm) in conjunction with freely rotating wheels, and the example of the
Talgo train was cited, in which the wheelsets are mounted on the car bodies in an
unsymmetric way.
A further source of lack of symmetry in railway vehicles can occur when a con-
figuration which is intended to be symmetric suffers from badly distributed loading
or unequal wheel wear, thus giving rise to asymmetry about a transverse plane.
Asymmetry about a longitudinal plane can, of course, also occur for the same rea-
sons; however, this tends to merely alter the equilibrium position of the vehicle, and,
while it should be taken into account in detailed numerical calculations, no new
phenomena are introduced. Moreover, there is evidence to suggest that asymmetry
about a transverse plane due to unequal wheel wear is more important for vehicles
which have inherently poor curving ability, such as the three-piece freight truck. In
this case, calculations have been described by Tuten, Law, and Cooperrider [13].
Illingworth [14] suggested the use of unsymmetric stiffness in steering bogies.
Elkins [15] showed both by calculation and experiment that a configuration of bo-
gie, with the trailing axle having independently rotating wheels and the leading axle
conventional, significantly improved stability and curving performance and reduced
rolling resistance.
Suda et al [16-17] have studied bogies with unsymmetric stiffnesses and sym-
metric conicity, and their development work has led to application in service [18].
The concept has been extended to include lack of symmetry of the wheelsets by
equipping the trailing axle with freely rotating wheels. This provides a practical ex-
ample of a re-configurable design as the wheelsets are provided with a lock which is
released on the trailing wheelset (allowing free rotation of the wheels) and locked on
the leading wheelset (providing a solid axle). The lock is switched depending on the
direction of motion.
In modern vehicles the provision of secondary suspension between bogies and
car body has the result that lack of symmetry in the car body rarely introduces new
phenenoma. However, lack of symmetry in the bogie itself can be exploited to im-
prove performance and this is discussed below.
9.2 Stability Theorems for Rigid and Semi-Rigid Vehicles
Important insights are obtained by considering, first of all, the static and dynamic
stability of vehicles in which the wheelsets are incorporated rigidly in one or more
frames, which are themselves connected by joints imposing non-elastic constraints.
Thus, in these cases the elastic stiffness matrix E is null. A rigid vehicle is one in
which all the interwheelset elastic stiffnesses are infinite, so that the number of de-
grees of freedom M = 2, corresponding to lateral translation and yaw. A semi-rigid
vehicle is defined as one in which E is null but 2N > M > 2, where N is the number
of wheelsets, so that in addition to possessing rigid body freedoms, the vehicle is
able to articulate like a mechanism. Now referring to the simple form of the equa-
tions of motion of a single wheelset, because the wheelset is symmetrical about a
transverse plane through its centre, for a motion in the reverse direction, while re-
taining the same definition of generalised coordinates, the equations of motion
(2.10.47-48) become
m
y + 2f
22
y /V + k
y
y + 2f
22
= Q
y
(1)
-2f
11
ly/r
0
+ I
z

+ 2f
11
l
2
/V + k
= Q

(2)
and in general the equations of motion for reversed motion of a complete vehicle are
[ As
2
+ Bs/V - C + E] q = Q (3)
Thus, the equations of reversed motion are obtained by reversing the sign of the
creep stiffness matrix. In the case of the semi-rigid or rigid vehicle the equations of
motion reduce to
[As
2
+ (B/V)s C ] q = Q (4)
where + refers to forward motion and - refers to reversed motion. The trial solution
q e
st
leads to the determinantal equation
As
2
+ (B/V)s C = 0 (5)
On making the substitution s = VD this becomes
AV
2
D
2
+ BD C = 0 (6)
and for low speeds, this reduces to
BD C = 0 (7)
which expands to, for forwards motion,
p
M
D
M
+ p
M-1
D
M-1
+..........+ p
1
D + p
0
= 0 (8)
For reversed motion, since all the elements of C change sign, some of the p
i
change
sign. In particular, p
0
and p
1
will have opposite signs. The condition for static stabil-
ity is that p
0
> 0 and a necessary (but not sufficient) condition for dynamic stability
is that all the p
i
have the same sign. Hence, if an unsymmetric rigid or semi-rigid
vehicle is statically and dynamically stable in forward motion, it must be dynami-
cally unstable in reverse motion.
If the number of degrees of freedom M is even, p
0
does not change sign in re-
versed motion and if the vehicle is statically stable in both directions, it will be dy-
namically unstable in one direction. If M is odd, p
0
does change sign in reversed
motion and the vehicle will be statically unstable in one direction of motion.
Hence, an unsymmetric semi-rigid or rigid vehicle can possess a margin of sta-
bility in one direction only. This margin of stability is derived from the lack of
symmetry, for if the vehicle were symmetric since (8) must be invariant with the
direction of motion it follows that p
M-1,
p
M-3
,...... must all be zero. Then, in many
cases (8) can be factorised in the form (for M even)
p
M
(s
2
+
1
2
) (s
2
+
2
2
)........ (s
2
+
P
2
) = 0 (9)
where
1
,
2
,.........
P
are a set of P = M/2 steering frequencies, or (for M odd)
p
M
(s
2
+
1
2
) (s
2
+
2
2
)........ (s
2
+
P
2
)s = 0 (10)
involving a set of P = (M-1)/2 steering frequencies, with the addition of a zero root.
This zero root indicates that the vehicle is capable of quasi-static misalignment simi-
lar to that discussed for two-axle vehicles in Section (4.3). In other cases, roots coa-
lesce to form a pair of roots i indicating oscillatory instability or alternatively
roots of the form occur indicating divergence.
The assumption of low speeds has made it possible to neglect the inertia terms in
the equations of motion. For semi-rigid vehicles in the more general case, (6) leads
to a characteristic equation of order 2M. However, the coefficients p
1
and p
0
will
remain unchanged, and it follows that irrespective of speed, an unsymmetric semi-
rigid or rigid vehicle will be unstable in at least one direction of motion. This sug-
gests that the examination of stability at low speeds is likely to yield useful informa-
tion on the behaviour of various configurations, as is done in the following.
9.3 Unsymmetric Rigid Vehicle
The general theory is exemplified by the unsymmetric rigid vehicle as considered by
Carter [5]. Whereas Carter considered rigid assemblies of an arbitrary number of
wheelsets, it will be sufficient for present purposes to consider only the case of a
two-axle vehicle or bogie. If y and represent the lateral displacement and yaw of
the vehicle then the standard form of the simplified equations of motion of Section
2.10 applies where
A
m m c
m c I
=
0 0
0 0
(1)
B
f
f h l
=
+
4 0
0 4
2 2
( )
(2)
C
f
f fh
=
0 4
2 2
1 2 1 2
( ) ( )
(3)
where m
0
= 2m + m
b
and I
0
= I
zb
+ 2mh
2
and c is the distance of the centre of mass
ahead of the mid point, and
1
=
1
l/r
0
,
2
=
2
l/r
0
.
The characteristic equation of this system is
p
4
s
2
+ p
3
s
3
+ p
2
s
2
+ p
1
s + p
0
= 0 (4)
At low speeds, this reduces to
p
2
s
2
+ p
1
s + p
0
= 0 (5)
and then
p
2
= 2(h
2
+ l
2
)/V
2
p
1
= h(
1

2
)/V
p
0
= (
1
+
2
)
In accordance with the general theory discussed in the previous section, since the
number of degrees of freedom is even, p
0
is invariant with change in direction of
motion and is essentially positive. On the other hand, the sign of p
1
changes when
the direction of motion changes. Thus, the distribution of conicity can be arranged to
give stability in one direction of motion, but not in both. This was suggested as a
means of stabilising a vehicle by Rocard [7] who states that a successful experiment
was made by French National Railways in 1936. Equation (5) yields an eigenvalue
i where, if p
1
2
<< 4p
0
p
2
,
+
h V
h l
( )
( )
1 2
2 2
4
(6)

2 1 2
2
2 2
2
=
+
+
( )
( )
V
h l
(7)
This steering oscillation was discussed in Section 5.3 for a symmetric vehicle and is
the analogue for a rigid vehicle of the kinematic oscillation of a single wheelset. If
the vehicle is symmetric, the oscillation is undamped, though it may be readily veri-
fied that creep occurs throughout the motion. If the vehicle is unsymmetric, the
damping in the steering oscillation will be positive if the conicity of the leading
wheelset is larger than that of the trailing wheelset.
Consider now the stability of the vehicle at speed, firstly assuming that c = 0.
Making the reasonable assumption that I
zb
= m
0
(h
2
+ l
2
), it follows from the form of
the equations of motion that the characteristic equation (5) for speed V
1
, with eigen-
values
i
, will be the same as (4) for speed V
2
, with eigenvalues s
i
, if
4f
/V
1
= m
0
s
i
2
/2 + 4fs
i
/V
2
(8)
Equation (8) can be regarded as a quadratic in s. Typically, the solutions of (4) will
consist of a complex eigenvalue together with two large real roots
s
1
= -4f/m
0
V (9)
s
2
= -4f(h
2
+ l
2
)/I
0
V (10)
The combined effect of speed and inertia is therefore to modify the damping of the
steering oscillation, and to introduce subsidences in lateral translation and yaw of
the vehicle. It can be seen that in the case of a steering oscillation which is stable
for low speeds, so that the corresponding real part of the eigenvalue = i is
negative, the damping will vanish at a speed V
B
and the corresponding eigenvalue
will be s
i
= i. By separating real and imaginary parts
( /V
B
) = ( /V)
0
(11)
V
B
2
= 4f(/V)
0
/m
0
(/V)
0
2
(12)
The instability is an inertia driven instability of the steering oscillation. Substituting
from (6) and (7), (12) yields
V
fh
m
B
2 1 2
0 1 2
2
=
+
( )
( )

(13)
For a bogie with the parameters of Table 4.1 with
1
= 0.15 and
2
= 0.075 equation
(13) gives a bifurcation speed V
B
= 34.4 m/s indicating that lack of symmetry offers
scope for useful enhancement of stability.
Numerical solutions of (4) show that variation of the longitudinal position c of
the centre of mass of the vehicle have only small influence on the bifurcation speed.
9.4 Steering of a Vehicle with Unsymmetric Inter-Wheelset Structure
It was shown in Chapter 5 that the behaviour of a symmetric vehicle is dominated
by the elastic connections between the wheelsets, and it can therefore be expected
that lack of symmetry in that respect may be important. In considering the properties
of the stiffness matrix for a general unsymmetric inter-wheelset structure, the argu-
ment of Section 4.2 may be followed but in the present case fore-and-aft symmetry
is not assumed. In terms of the sum-and-difference coordinates of Section 4.2 the
stiffness matrix then takes the form [10]
E
k k k h
k k k h
k h k h k h
s sb s
sb b sb
s sb s
=
0
0
0 0 0 0
0
2
(1)
and there are only three independent elements. In addition to the bending and shear
stiffnesses k
b
and k
s
already defined, there is a coupling term k
sb
indicative of lack of
fore-and-aft symmetry. For a symmetric vehicle k
sb
= 0 and E reduces to the form
given in Section 4.2. Reverting to the usual definition of the generalised co-
ordinates, using equations (4.2.3), so that y
1
and y
2
are the lateral displacements
measured from the centreline of the track, and
1
and
2
are the yaw angles, E be-
comes
E
k k k h k k k h
k k h k k h k h k k h k k h
k k k h k k k h
k k h k k h k k h k k h k h
s sb s s sb s
sb s b sb s sb s b s
s sb s s sb s
sb s b s sb s b sb s
=

+ + +
+ +
+ + + +
2
2
2 2
2 2
(2)
Consider the steady-state behaviour on a uniform curve for the case of zero cant
deficiency. Then for coned wheels and small displacements, the gravitational stiff-
ness and lateral force due to spin creep, together with the other small terms, can be
neglected so that the equations of equilibrium, corresponding to the steady-state
form of equations (2.10.47-48), reduce to
[ C + E ] q + Eq
0
= Q
c
(3)
where, assuming that f
11
= f
22
= f for simplicity,
C =
0 2 0 0
2 0 0 0
0 0 0 2
0 0 2 0
1 0
2 0
f
f l r
f
f l r
/
/
(4)
E is given by (2) and
Q
c
= [ 0 - 2fl
2
/R
0
0 - 2fl
2
/R
0
]
T
(5)
and the forces applied to the wheelset due to straining of the suspension to achieve
the reference position in the chosen coordinate system are given by Eq
0
where
q
0
= [ 0 h/R
0
0 -h/R
0
]
T
(6)
For motion on a curve with zero lateral and longitudinal creep
[ ]
q lr R lr R =
0 1 0 0 2 0
0 0 / /
T
(7)
The solution of equations (3) is satisfied by (7) if both
k
k
lr
h
sb
s
=
0
1 2
2
1 1

(8)
and
k
s
k
b
- k
sb
2
= 0 (9)
These are the conditions for steering on a uniform curve with zero creep and are the
same for forward and reverse motion. It can be seen that (9) is the generalisation,
applicable to the present case of unsymmetric elasticity, of the requirement that for
perfect steering in a uniform curve the bending stiffness must be zero. Similarly, (8)
is equivalent to a geometric requirement that a "bending displacement" about the
point of articulation is compatible with lateral displacements of the wheelsets neces-
sary, by their respective conicities, to ensure pure rolling.
Figure 9.2 indicates the geometry when a vehicle with a structural arrangement
of the kind meeting the criteria (8) and (9) negotiates a curve. For zero creep, both
wheelsets must adopt a radial position with outward displacements lr
0
/R
0
1
for the
leading wheelset and lr
0
/R
0
2
for the trailing wheelset. If the radial displacements
are small compared with the radius of the curve, the geometry of Figure 9.2 then
yields (8).
lr
0
/R
0
1
lr
0
/R
0
2
h - k
sb
/k
s
h + k
sb
/k
s
Figure 9.2 Attitude of unsymmetric vehicle negotiating a uniform curve with zero longitudi-
nal and lateral creep
An alternative approach due to Suda [16,17] imposes the steering condition that
1
and
2
are zero, so that the lateral creep is zero in a curve. It is possible to meet this
criterion using only the unsymmetric inter-wheelset structure, with the wheelset
conicities being equal. However, the wheelsets do not move out to the rolling line
and longitudinal creep forces are not zero. For the wheelsets to take up a radial posi-
tion and move outwards to the rolling line on a curve both conditions (8) and (9)
must be satisfied and this requires lack of symmetry in both the distribution of
conicity and elasticity.
These considerations may be illustrated with reference to the example unsymmet-
ric inter-wheelset structures shown in Figure 9.3 which obey (9). Figure 9.3(a) shows
an arrangement based on inclined tension members which provide a virtual pivot, and
Figure 9.3(b) shows an arrangement embodying a single shear spring at the articula-
tion point. In both cases a pivot position (virtual or real) can be provided at any point
within (or, theoretically, beyond) the wheelbase. Another way in which (8) or (9) can
be implemented is by offsetting the lateral stiffnesses k
y1
and k
y2
as shown in Figure
9.4 (a). Suda [16,17] has studied and experimented with the scheme shown in Figure
9.4(b) (applying the condition for zero lateral creep described above) and this is the
basis for a design of bogie actually put into commercial service.
2a
(a)
2b
k
s
= k(b - a)sin
2
k
sb
= k(b - a)sincos/2
k
b
= k(h
2
sin
2
abcos)
k
h
1
h
2
h h
k
(b)
k
s
= k
k
sb
= k(h
2
- h
1
)/2
k
b
= k(h
2
- h
1
)
2
/4
Figure 9.3 Examples of unsymmetric inter-wheelset structures which satisfy the condition
given by equation (9).
The possible range of configurations of vehicle which satisfy (8) and (9) is con-
strained by the practical considerations mentioned in connection with the three-axle
vehicle. It is assumed that that values of the non-dimensional equivalent conicities
1
and
2
are always less than unity, and the ratio of conicity between leading and
trailing wheelsets is will not exceed about two, or exceptionally, three. Then Figure
9.5 summarises the possible range of configurations which can steer without lateral
and longitudinal creep on curves. It can be seen that the pivot position falls within
the wheelbase for those configurations which meet these practical constraints. This
is indicated by the lines A and B giving the relationship between
1
and
2
for the
pivot position at the front and rear axles respectively. The lines C and D represents
k
y
/2 k
y
/2
k
2
k
1
h
a
h
b
h h
k
s
= 2k
y1
k
y2
k
/
k
sb
= k
y1
k
y2
k
(h
a
- h
b
)/
k
b
= [(k
y1
+ k
y2
)k
2
+ k
y1
k
y2
k
{(h + h
a
)
2
+ (h + h
b
)
2
}/
= 2(k
y1
+ k
y2
)k
+ k
y1
k
y2
(2h + h
a
+ h
b
)
2
}
k
y1
/2 k
y2
/2
k
(a)
(b)
k
s
= (k
1
+ k
2
)k
y
/
k
sb
= hk
y
(k
2
- k
1
)/
k
b
= {2k
1
k
2
+ h
2
k
y
(k
2
+ k
1
)}/
= 2(k
1
+ k
2
) + 4h
2
k
y
Figure 9.4 Two examples of unsymmetric structures connecting wheelsets to bogie frame.
the limits of a 2:1 ratio in conicity between the axles. A smaller ratio in conicity
leads to a pivot position between the wheelsets as shown at E and F. Conversely,
configurations which do not meet these practical constraints require pivot positions
well outside the wheelbase as shown at G and H.
9.5 Stability of a Two-Axle Articulated Vehicle
Consider the stability at low speeds of an unsymmetric two-axle vehicle. For sim-
plicity, it is assumed that the creep coefficients for each wheelset are equal. Equa-
tions of motion of the form (2.7) are obtained where
B
f
fl
f
fl
=
2 0 0 0
0 2 0 0
0 0 2 0
0 0 0 2
2
2
(1)
C
f
f
f
f
=
0 2 0 0
2 0 0 0
0 0 0 2
0 0 2 0
1
2
(2)
0 0.1 0.2 0.3 0.4 0.5
0
0.1
0.2
0.3
0.4
0.5
2

1
A
C
D
B
G
E
F
H
Figure 9.5 Configurations of unsymmetric vehicles capable of steering with zero longitudinal
and lateral creep.
and E is given by (4.2).
The trial solution q e
st
leads to a characteristic quartic polynominal with coef-
ficients given by
p
4
= 1
p
3
= (k
s
l
2
+ k
s
h
2
+ k
b
)V/fl
2
p
2
= {2(k
s
k
b
- k
sb
2
)(l
2
+ h
2
) + 2f
2
l
2
(
1
+
2
) k
s
fhl
2
(
1
-
2
)
k
sb
fl
2
(2-
1
-
2
)}V
2
/2f
2
l
4
(3)
p
1
= { (k
s
k
b
- k
sb
2
)(
1
-
2
)h + 2k
sb
fh(
1
-
2
)
+ f(k
s
l
2
+ k
s
h
2
+ k
b
) (
1
+
2
)}V
3
/2f
2
l
4
p
0
= {(k
s
k
b
- k
sb
2
)(
1
+
2
) + 2f
2
2
k
s
fh(
1
-
2
)
k
sb
f(
1
+
2
- 2
1
2
)}V
4
/2f
2
l
4
where refers to forward (+) or reversed (-) motion, respectively, and
i
=
i
l/r
i
.
The semi-rigid case will be considered first so that for the vehicle shown in Fig-
ure 9.3(b) k = . Then (3) reduces to the cubic polynominal with coefficients given
by
p
3
= (2l
2
+h
1
2
+h
2
2
)l
2
p h Vl h Vl
2 1 1
2
2 2
2
1 1 = ( ) ( )
(4)
p l h V l h V
1 2
2
1
2 2
1
2
2
2 2
= + + + ( ) ( )
p h V h V
0 2 1 1
3
1 2 2
3
1 1 = ( ) ( )
where the signs refer to forward and reverse motion respectively. Figure 9.6
shows how the eigenvalues vary as the equivalent conicity of the trailing wheelset
varies for the bogie with the parameters of Table 4.1 except that
1
= 0.22 and h
1
=
0.75 m. It can be seen that the eigenvalues when
2
=
1
comprise a zero root corre-
sponding to the quasi-static misalignment discussed in Section 4.3 and an oscillatory
root i corresponding to the kinematic oscillation of a single wheelset. How-
ever, for the unsymmetric vehicle, when
2
<
1
there is dynamic instability and
static stability, and for
2
>
1
there is dynamic stability. When
2
is sufficiently
greater than
1
there is divergence. In reverse motion, the frequency of the steering
oscillation remains unaltered but the real parts of the eigenvalues change sign. This
is, of course, in accordance with the discussion in Section 2.
If
1
is small, the characteristic polynomial can be factorised approximately, yield-
ing a real root
1

1
0
1
2 1 1 1 2 2
2
2
1
2
1
2
2
2
1 1
= =
+
+ + +
p
p
h h V
l h l h
{ ( ) ( )}
( ) ( )
(5)
and a complex pair = i where

2 1
3
2
2
1
2
1
2
2
2 2
2
1
2
2
2 2
2
= =
+ + +
+ +
p
p
l h l h V
l h h l
{ ( ) ( )}
( )
(6)
=
( ) p p p p
p p
2 1 3 0
3 1
2
=
+ + +
+ + + + +
( ){ ( )( ) ( )( )}
( ){ ( ) ( )}

2 1 1
2
2
2
1 2
2
1
2
2
2 2
1
2
2
2
2
2
2
2
1
2
1
2
1 1
2 2
h l h h l h V
l l h h l h l h
(7)
0 0.2 0.4 0.6
0
0.5
1
2
0 0.2 0.4 0.6
-0.4
-0.2
0
0.2
0.4
2
Figure 9.6 Behaviour of eigenvalues of semi-rigid unsymmetric vehicle at low speeds as
conicity of rear wheelset is varied. Parameters of Table 4.1 except that
1
= 0.22, h
1
=
0.75.
Though these expressions are based on an approximate solution of the characteristic
equation, they do describe accurately enough the behaviour of the system as the
conicities and pivot position vary. The real root represents a subsidence or diver-
gence depending on the relative magnitudes of h
1
and h
2
and
1
and
2
. In fact, the
exact condition for static stability is p
0
> 0 or

2 1 1 1 2 2
1 1 0 h h ( ) ( ) + > (8)
or

2 1 1 1 2
1 1 1 < + / { ( ) / } h h (9)
The oscillatory root represents an oscillation with frequency roughly equal to the
mean frequency of the kinematic frequencies of the two wheelsets, and is influenced
1
0
1
1
S
O
1
0
1
1
O
D
1
0
1
1
S
O
D
1
0
1
1
O
D
D,O
1
0
1
1
D
D,O
h
1
= 0
h
1
= h/3, h
2
= 2h/3
h
1
= 2h/3, h
2
= h/3
h
2
= 0
h
1
= h
2
= h
Figure 9.7 Regions of stability in the
1
,
2
plane as a function of h
1
and h
2
for the semi-rigid
vehicle at low speeds; S = stability, D = divergence and O = oscillatory instability.
by joint position. The damping in this oscillation is strongly dependent on the differ-
ence in equivalent conicities and the pivot position exercises a less important influ-
ence. The condition for stability is simply

2 1
> (10)
The regions of stability, as given by (9) and (10) are plotted in the
1
,
2
plane for
various spring positions in Figure 9.7. It is possible to select parameters which will
ensure both static and oscillatory stability for one direction of motion, but these pa-
rameters would lead to both static and oscillatory instability in reverse motion.
Since, in most practical cases,
1
<1 and
2
< 1 a necessary condition for static
stability is a
1
h
2
> a
2
h
1
. This criterion can be combined with (10) to give the neces-
sary but not sufficient condition
h
2
/h
1
>
2
/
1
>1 (11)
Thus, for stability in forward motion, the pivot must be nearer the front wheelset
than the back wheelset and the rear wheelset must have larger conicity (e.g., smaller
radius) than the front wheelset. These results contrast with those obtained for the
rigid vehicle.
9.6 The Influence of Elastic Stiffness on Stability
The influence of the stiffness on the stability of the two-axle vehicle shown in Fig-
ure 9.3(b) is now considered. The analysis will show that the configuration studied
is not only stable for a range of parameters in both forward and reverse motion, but
it is also capable of perfect steering.
Consider the behaviour of the eigenvalues as a factor k on the stiffnesses k
s
, k
b
and k
sb
is varied, and the steering conditions (4.8) and (4.9) are applied. There are
two regimes in which the behaviour is quite distinct. The first of these is for small
values of kl/2f and the second is for all other values of kl/2f. As will be explained,
the former regime is one of small interaction between the wheelsets and the latter
regime is one of intensive interaction.
When kl/2f = 0 the eigenvalues are purely imaginary and distinct, corresponding
to undamped kinematic oscillations. As the eigenvalues are distinct it is possible to
employ the perturbation analysis of Section (3.3), equation (3.3.12). Therefore, for
small values of kl/2f , the eigenvalues are
1
i
1
and
2
i
2
where

1
2
2
0
2
1 0
2
1
2
=
V
lr
V k k h l r
fl
sb s
( )( / )
(1)
1
2 2 2
2 4 = + + V k l k h k k h fl
s s b sb
( ) /

2
2
2
0
2
2 0
2
1
2
=
+ V
lr
V k k h l r
fl
sb s
( )( / )
(2)
2
2 2 2
2 4 = + + + V k l k h k k h fl
s s b sb
( ) /
the signs referring to forward and reverse motion respectively. Both
1
and
2
are
negative for all values of k
s
and k
b
and the system is stable.
These results are consistent with a wheelset oscillating under asymmetric elastic
restraint, the other wheelset being restrained from lateral motion. In fact, (1) can be
derived by the solution of the binary set of equations of motion involving y
1
and
1
,
whilst (2) similarly can be derived from the sub-system involving y
2
and
2
. The
analysis is then a simple extension of that for a single symmetric wheelset given in
Chapter 3, to which the present analysis is reducible. These two wheelset sub-
systems (leading and trailing wheelsets) are, for the small values of kl/2f under con-
sideration here, weakly-coupled in the sense of Milne [19], and his analysis could be
applied to give more complete rigour. The validity of the perturbation method ap-
plied here depends on the eigenvalues of the sub-systems being sufficiently sepa-
rated. This requirement is, of course, not satisfied for the symmetric vehicle with
equal conicity on both wheelsets, and a different approach is needed as employed in
Chapter 4.
As kl/2f is increased the interaction between the wheelsets dominates the beha-
viour of the vehicle. Some insight into the pattern of behaviour is given by the re-
sults for the symmetric vehicle discussed in Chapter 4, where it was shown that two
modes of oscillation exist; a lightly damped steering oscillation which takes place at
the kinematic frequency and a "shear" oscillation which for smaller values of kl/2f
takes place at the kinematic frequency. As the real part of the eigenvalue increases
rapidly as kl/2f increases this oscillation is replaced by two subsidences. Similar
results may be expected for the unsymmetric vehicle, but there is the possibility of
static and oscillatory instability.
The condition for static stability is that p
0
> 0 and if the steering conditions (4.8)
and (4.9) are applied, this becomes
2f
1
2
k
s
h(
1
-
2
) k
sb
(
1
+
2
- 2
1
2
) > 0 (3)
which reduces to (5.8) as kl/2f becomes large. Rouths condition for oscillatory sta-
bility is that the test function of equation (3.4.3) should be positive or, on substitu-
tion
(
1
-
2
)[-A
2
{ f(
1
-
2
) - k
s
h(2
1
-
2
) - k
sb
(
1
-
2
)} + k
sb
h{2k
s
h(
1
-
2
)
- 2k
sb
(
1
+
2
) + 4k
sb
} 4k
sb
2
h
2
f(
2
-
1
)] > 0 (4)
where A k l k h k
s s b
= + +
2 2
.
Equation (4) reduces to (3.10) as kl/2f becomes large.
For small values of kl/2f, for the vehicle shown in Figure 9.3(b), (4) is satisfied
for all values of the parameters consistent with (1) and (2), and the vehicle is oscilla-
tory stable for both forward and reverse motion, except when
1
=
2
when it is of
course marginally stable, irrespective of the position of the spring.
Figure 9.8 summarises these results in the form of a stability diagram in the k
s
, h
1
plane, for both forward and reverse motion at low speeds, where the position of the
spring and the conicities are chosen so that (4.8) and (4.9) are satisfied. As the posi-
tion of the spring moves aft from the central position, there is the possibility of both
oscillatory and static instability. If the spring were placed at the trailing axle, then
from (5) for oscillatory stability
k
f l h h
h l h
<
+ +
+
{( ) }( )
( )( )
2 2 2 4
1 2
3 2 2
2
4 4
4 2 1

(5)
and from (3) for static stability
k
f
h
<
1
1
1 ( )
(6)
As the position of the spring moves forward from the central position neither oscil-
latory or static instability occurs. Thus, if the criteria (5), and (6) are satisfied, the
vehicle possesses static and oscillatory stability, at low speeds, for either direction of
motion and the ability to steer perfectly on curves. This result may be contrasted with
the conflict between steering and stability exhibited by the symmetric two-axle vehi-
cle discussed in Chapter 4.
0 1.25
2.5
0.5
1
1.5
2
k
s
(MN/m)
h
1
(m)
S O
O,D
0 1.25 2.5
0.5
1
1.5
2
k
s
(MN/m)
h
1
(m)
S
O
D
S
S
0.1 0.08 0.06 0.05 0.04

2
0.1 0.08 0.06 0.05 0.04

2
(a)
(b)
Figure 9.8 Stability diagram for unsymmetric vehicle at low speeds for the case where equa-
tions (4.8) and (4.9) are satisfied.
1
= 0.1. (a) forward motion (b) reverse motion.
Figure 9.9 shows the stability boundaries for the vehicle in both forward and
reverse motion as a function of the effective pivot position and vehicle speed. It can
be seen from these results that the choice of parameters in order to achieve static and
dynamic stability and steering without creep, in both directions of motion, is limited,
as only moderate values of the shear stiffness are destabilising. For physically real-
isable masses the critical speed associated with instability in the steering mode
(driven by the inertia forces) is quite low. Even if stability at low speeds is achieved,
the margin of stability is very small resulting in low damping in the steering oscilla-
tion at low speeds.
The significance of this simple example lies not in its direct practical use but in
the fact that it demonstrates that two-axle configurations exist that resolve the con-
flict between steering and stability.
9.7 Applications of Unsymmetry
Considering the model analysed in the previous Section, suppose that, in addition to
the shear stiffness k
1
at the pivot point, yaw restraint k
2
is also provided. This could
be provided by a relaxation damper so that steering at low speeds is not compro-
mised, the conditions for steering without creep, equations (4.8-9) will be satisfied,
and stability at speed will be enhanced. Then Figure 9.10 shows the effect on stabil-
ity of separately introducing different conicities on the leading and trailing wheelsets
and elastic unsymmetry in the interwheelset connections. The bifurcation speed is
shown as a function of the shear and bending stiffnesses and these results can be
compared with those for the symmetric bogie shown in Figure 5.20. If the conicity is
greater on the trailing wheelset than on the leading wheelset, and if the effective
pivot point is towards the rear of the vehicle, then stability is higher than the corre-
0
1.25
2.5
2
6
10
h
1
(m)
V (m/s)
S S
O
0 1.25
2.5
2
6
10
S S
O
h
1
(m)
V (m/s)
(a)
(b)
Figure 9.9 Stability diagram for unsymmetric vehicle for the case where equations (4.6) and
(4.7) are satisfied. (a) forward motion (b) reverse motion. k
s
= 10 kN/m.
sponding symmetric vehicle for a range of stiffnesses. But, in each case, operation in
reverse results in lower bifurcation speeds.
As the margin of stability is adequate for one direction of motion if appropriate
parameters are chosen this suggests the application of a variable configuration which
changes with the direction of motion. For the system discussed above, in order to
preserve the conditions for steering without creep at low speeds it would be neces-
sary to vary both the position of the effective pivot point and the ratio of the conic-
ities. Obviously, this would be very difficult to achieve in practice.
A practical application of similar ideas has been described by Suda [16,17] in
which depending on the direction of travel, elastic stiffness parameters are switched
fore-and-aft. Three configurations are considered (a) conventional or radial bogies
with unsymmetric interwheelset connections (b) bogies with symmetric interwheel-
set structures but using independently rotating wheels on the trailing axle (c) bogies
with a combination of unsymmetric inter-wheelset connections and independently
25
0
25
k
1
(N/m)
k
2
(MNm)
25
0
25
25
0
25
0
25
2.5
k
1
(N/m)
k
2
(MNm)
k
1
(N/m)
k
1
(N/m)
k
2
(MNm)
k
2
(MNm)
V
BMAX
= 2.08V
0
V
BMAX
= 0.86V
0
V
BMAX
= 0.91V
0
V
BMAX
= 1.77V
0
1
= 0.075,
2
= 0.15 h
1
= 0.625, h
2
= 1.875
forward
reverse
Figure 9.10 Stability diagrams for unsymmetric vehicle with both shear stiffness k
1
and yaw
stiffness k
2
at joint showing the effects of differing conicities (left hand) and joint posi-
tion (right hand) in forward motion (top row) and reverse motion (bottom row). V
0
is the
maximum bifurcation speed when
1
=
2
= 0.1 and h
1
= h
2
= 1.25.
rotating wheels on the trailing axle. Suda proposes that the change from independent
rotation of the wheels to a solid axle when the direction of motion is changed is ef-
fected using an electromagnetic clutch. In (a) it is found that even though the steer-
ing condition is relaxed stability is low. A practical method of increasing stability in
this case is to switch the parameters when the direction of motion is changed is by
using switched primary longitudinal dampers in the conventional configuration of
bogie. Yaw restraint is large for the trailing wheelset and low for the leading wheel-
set. In both (b) and (c) steering performance must be relaxed in order to achieve
adequate stability.
In actual application to the bogie vehicles of the JR-Central Series 383 train [18],
the longitudinal stiffness of the wheelsets in each vehicle was originally configured
in a soft-hard-soft-hard arrangement, with the first soft wheelset leading and the
configuration reversed when the train changed direction. However, experiments
showed that there was no need to have soft restraint on the third wheelset, and the
configuration was modified to soft-hard-hard-soft removing the need to switch the
configuration when the train reverses direction. The dynamic behaviour of symmet-
ric bogie vehicles with two unsymmetric bogies has been considered in Chapter 6.
References
1. Ahrons, E.L.: The British Steam Railway Locomotive 1825-1925. Bracken Books,
London, 1987.
2. Goodwin, A.M.: The evolution of the British electric tramcar truck. Tramway and
Light Railway Society, London, 1977.
3. Carter, F.W.: The electric locomotive. Proc. Inst. Civil. Engrs. 221 (1916), pp.
221-252.
4. Carter, F.W.: Railway Electric Traction. Edward Arnold, London, 1922.
5. Carter, F.W.: On the stability of running of locomotives. Proc. Roy. Soc. 121,
Series A (1928), pp. 585-611.
6. Carter, F.W.: The running of locomotives, with reference to their tendency to de-
rail. Institution of Civil Engineers, Selected Engineering Paper, No.81, 1930.
7. Rocard, Y.: La stabilite de route des locomotives. Actual. Sci. Ind. 234 (1935),
Part 1.
8. Rocard, Y.: General Dynamics of Vibrations. Crosby Lockwood, London, 1960.
(trans. from French, 1st pub. 1943)
9. Wickens, A.H.: Steering and stability of unsymmetric articulated railway vehi-
cles. Trans. A.S.M.E. Journal of Dynamic Systems, Measurement and Control. 101
(1979) pp. 256-262.
10. Wickens, A.H.: Static and dynamic stability of unsymmetric two-axle railway
possessing perfect steering. Vehicle System Dynamics 11 (1982) pp. 89-106.
11. Carter, F.W.: Improvements In and Relating to High Speed Electric or Other
Locomotives, British Patent Specifications 128,106 (1918), 155,038 (1919), and
163,185 (1920).
12. Li, W.: The Dynamics of Perfect Steering Bogie Vehicles and its Improvement
with a Re-Configurable Mechanism. Doctoral Dissertation, Loughborough Univer-
sity, 1995.
13. Tuten, J.M., Law, E.H., and Cooperrider, N.K.: Lateral stability of freight cars
with axles having different wheel profiles and asymmetric loading, ASME Paper
No. 78-RT-3, 1978.
14. Illingworth, R.: The use of unsymmetric plan view suspension in rapid transit
steering bogies. Hedrick, J. K. (Ed.): The Dynamics of Vehicles on Roads and
Tracks, Proc. 8th IAVSD Symposium, Cambridge, Mass. August 1983, pp. 252-265.
Swets and Zeitlinger Publishers, Lisse, 1984.
15. Elkins, J.A.: The performance of three-piece trucks equpped with independently
rotating wheels. Anderson, R. (Ed.): The Dynamics of Vehicles on Roads and Tracks,
Proc. 11th IAVSD Symposium, Kingston, Ontario, August 1989, pp. 203-216. Swets
and Zeitlinger Publishers, Lisse, 1984.
16. Suda, Y.: Improvement of high speed stability and curving performance by pa-
rameter control of trucks for rail vehicles considering independently rotating wheel-
sets and unsymmetric Structure. JSME International Journal, Series III, 33-2
(1990), pp. 176-182.
17. Suda, Y.: High speed stability and curving performance of longitudinally un-
symmetric trucks with semi-active control, Vehicle System Dynamics, 23 (1994), pp.
29-52.
18. Yoshikawa, K.: Faster through the curves on JR-Central, Railway Gazette Inter-
national, August 1997, pp. 531-532.
19. Milne, R.D.: The analysis of weakly coupled dynamical systems. Int. J. Control,
2 (1st series) (1965), pp. 171-199.
Index
actuators, 246 optimum response to, 117
alternative guidance systems, 94-102 response to, on large radius curve, 116
Amontons law of friction, 32 centre friction plate, 173
anti-bending mode, 115 centrifugal forces, 47
anti-shear mode, 116 centring spring, 173, 262
antisymmetric degrees of freedom, 65 chaotic motions, 87
articulation, 1, 16, 99, 235 characteristic polynomial, 77
asymptotic stability, 76, 84 Class 143/144 passenger vehicles, 134
axle-guard clearance, 133 close coupling of vehicles, 133
axle load, 64,71 conicity, 21-31
coning, 3-8, 11, 28
bending oscillation, 119 constrained curving, 108
bending stiffness, 110, 137 constrained motion, 48
bifurcation, 84 constraints, 19
Bissel, 262 contact forces, 42
block diagram, 59, 81, 145 contact mechanics, 32
body instability, 145-154, 179, 254 contact patch, 6, 7, 32-40
body steering, 174, 190 contact slope difference, 21-31
bogie: contact stiffness, 13, 57-59
conventional bogie: Coulombs laws of friction, 32
dynamic response, 181 creep, 6, 33-36, 40-42
eigenvalues, 177-179 creep coefficients, 36, 58, 59
equations of motion, 175-176 'creep controlled' wheelset, 102
generalised coordinates, 175 creep damping matrix, 55
parameters, 178 creep forces, 7-8, 36-37, 47, 50, 54-60
stability, 179 creep saturation, 84, 113
steered bogie: creep stiffness matrix, 55
dynamic response, 194 critical speed, 85
eigenvalues, 197-198 cross-bracing, 111, 134, 138, 139, 159
equations of motion, 185-186 cross-level, 19
stability, 197-205 cylindrical treads, 3
unrestrained bogie:
curving, 124-127 degrees of freedom, 1, 19
eigenvalues, 160-163 derailment, 13-14
equations of motion, 61-66 describing functions, 30, 86
generalised coordinates, 61 divergence, 99, 222, 229, 265
limit cycles, 166-167 double-link suspension, 133
parameters, 115 dynamic response:
stability, 160-160, 168 conventional bogie, 181
buckling, 262 four-axle artic. vehicle, 256- 257
braking or tractive torque, 50 steered bogie, 194
Brunel, 5 three-axle vehicle, 230-232
two-axle vehicle, 121
cant, 15, 19
cant deficiency, 76 effective conicity, 28
elastic stiffness matrix, 55
eigenvalue problem, 76-77
eigenvalues:
conventional bogie vehicle, 177-179
four-axle articulated vehicle, 252-255
steered bogie, 197-198
three-axle vehicle, 217-223
three-axle artic. vehicle, 246-250
two-axle vehicle, 143-145
unrestrained bogie, 160-162
wheelset, 76-81
energy balance, 50-52, 56-58, 88
equation of compatibility, 62
equation of energy balance, 52, 56-57
equations of motion, 46-66
conventional bogie vehicle, 175
four-axle articulated vehicle, 242
steered bogie, 185-187
three-axle vehicle, 210-212
three-axle artic. vehicle, 239-240
two-axle vehicle, 61-66
unrestrained bogie, 61-66
wheelset, 46-60
equivalent conicity 11, 13, 30-31
effective conicity, 28
eye-bolt suspension, 133
FASTSIM, 37
feedback loop for wheelset, 8, 51, 81
feedback loops for two-axle vehicle, 145
flange, 2-5, 9, 11, 13-15, 94, 123, 130
flange contact, 2-4, 13-15, 113
flange wear, 15
flange-way clearance, 2, 3, 75
flexibility parameter, 112, 134, 146
flexible, long wheelbase vehicle, 113,
146
flutter, 229
forced steering, 16, 174, 238
four-axle articulated vehicle:
dynamic response, 259
eigenvalues, 252-255
equations of motion, 242
generalised coordinates, 238
parameters, 253
stability, 252
free curving, 108
freedoms, 19
Galerkins method, 87
gauge spreading, 123
generalised applied forces, 48-50
generalised coordinates:
steered bogie, 175
three-axle vehicle, 210
three-axle articulated vehicle, 239
unrestrained bogie, 61
wheelset, 49-65
gravitational forces, 47
gravitational stiffness, 13, 21, 50
gravitational stiffness matrix, 55
guidance, 10, 107
gyroscopic forces, 47
gyroscopic matrix, 55
Hertz stiffness, 48
Heumann minimum principle, 108
high speed trains, 174, 237
HSFV-1 vehicle, 114, 134, 155
hunting, 9, 10, 85
impedance of yaw spring, 140
independently rotating wheels, 95, 263
inertia matrix, 55, 65
instability, 8, 9, 10, 16, 145
inter-couplers, 190
Jacob bogie, 236
kinematic oscillation, 5-6, 9, 71-74
Klingels formula, 5-6
Krylov and Bogoljubov method, 87
lateral creepage, 6-7, 34, 40
lateral creep force, 6-7, 36-39, 42
laws of friction, 32
Liechtys bogie,195
limit cycles, 84-86, 153
linkage-steered bogie, 192-195
locomotives, 261-262
longitudinal creepage, 6-7, 34, 40
longitudinal creep force, 6-7, 36-39, 42
lower sway mode, 144
misalignment, response of two-axle
vehicle to, 127
modal expansion, 90
modal superposition, 89
INDEX 286
model roller rig experiments, 94
normal coordinates, 89
Nyquist criterion, 81
open loop transfer function, 81
orthogonality, 90
parameters:
unrestrained bogie, 115
wheelset, 59
path coordinates for two-axle vehicle, 61
peg-in-slot, 94, 99-102
perturbation procedure, 79
plane of symmetry, 61, 65
plateway, 95
pony axle, 96-98, 263
potential energy, 52, 56
quasi-linear equations, 86, 155
radial steering, 16, 182
Rayleighs dissipation function, 57
re-configurable systems, 263, 280-281
Redtenbachers formula, 3-4, 74-75
relaxation damper, 134, 140, 159
reversed motion, equations of motion, 264
ride, 15
rigid two-axle vehicle:
curving, 114
eigenvalues, 146
rotation matrix, 45
roller rig, 71, 154
Rouths criterion for stability, 81
self-steering, 16, 134, 238
semi-rigid vehicle, 263
sensors, 246
shear oscillation, 119, 143
shear stiffness, 110, 137
Shin Kansen, 174
side bearers, 173
simplified equations of motion, 60
spin, 6-8, 13, 18, 34-40
static instability, 10, 99, 115, 220, 256
steering, 117, 118, 182, 210, 238-242
steering oscillation, 119
steering beam, 244
steering laws, 189
Stephenson, 5
stiff, short wheelbase vehicle, 113
stiffness matrix:
deflated, 137
degeneracy, 183
properties, 109, 135, 182, 185
steered bogie vehicle, 185-186, 191-
195
two-axle vehicle, 62, 65
unrestrained bogie, 62, 65
unsymmetric structure, 268
wheelset, 55
S-Tog, 237
subsidences, 80
sum and difference coordinates, 109
suspension for simple two-axle vehicle,
62
swing bolster, 173
switching, 3, 95
system matrix, 60
Talgo, 98, 237
tangential tractions, 32-35
three-dimensional geometry, 31
three-piece bogie, 169, 263
throwover, 236
track curvature, response to, on large
radius curves, 111
track flexibility, 94
track irregularities, 1, 15
transient effects, 40, 81
transition, 15, 93-94
two-point contact, 31, 39, 49
upper sway mode, 144
veering of eigenvalues, 146
Wiesinger bogie, 192
worn profiles, 13, 28, 39
yaw relaxation, 134, 140, 159
yaw damper, 134, 140, 159

Fundamentals of Rail Vehicle Dynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fundamentals of Rail Vehicle Dynamics

Uploaded by

Copyright:

Available Formats

FUNDAMENTALS OF RAIL VEHICLE DYNAMICS

and using (8) and (9)

= . For conical wheels, R

cos cos sin sin sin

RAIL VEHICLE DYNAMICS 50

and V can be specified and

. The various steps in the solution of the equations

, and multiplying (8.17) by V and adding, giving

from equations (12)-(14), (6.1)-(6.12) and (5.13)-

will be represented by y and

Figure 2.18 Energy balance for wheelset.

are zero, and in terms of the operator D, these equations

. The inertia forces depend on V

/I then there is instability above a speed given by

( cos sin ) (4)

Figure 3.16 Conventional wheelset mounted on a leading or trailing arm.

E* corresponding to the new set of coordinates

=50 kNms h = 3.7 m d = 0.2 m

=10 kNms h = 1.25 m d = 0.2 m

and, moreover, the wheelbase h only occurs in conjunction with k

, for example, in the equations of motion by an im-

(s). Figure 5.3(a) indicates the arrangement consisting of a damper in

in the equations of motion

= 0) is shown in Figure 5.4. The eigenvalues at a specific speed

. In accordance with the discussion of Section 4.5, if either k

are infinite, then the solution of equations

is very large, the car body roll freedom

the contact stiffnesses may be neglected.

and wheelbase h on stability. These results

in accordance with the above discussion of the long wheelbase case. As

is offset by the destabilising effect of the cou-

/h. For larger values of k

/h become dominant and in-

/h, it is found that

. Parameters of Table 5.1

=0; (b) mode shape of oscillation showing one half-cycle at bi-

is progressively increased, as in Figures 5.11(b), (c) and (d) the region of

and wheelbase h on stability, for the

have their nominal values. Figure

. Figure 5.13(b) shows similar results for a smaller wheel-

will eliminate the body instability.

becomes large, as discussed in Section 4, examination of the equations of

for which the body instability is elimin-

Figure 5.14 Qualitative diagram showing the joint influence of k

= 0 and h = 3 m. These diagrams should be compared with Figure 5.14

3 null matrix, and F

S where S = (R-2)/2 if R is even and S = (R-3)/2 if R is odd; E

U where U = (R-2)/2 if R is even and U = (R-1)/2 if R is odd. Therefore

).For the leading bogie, oscillatory instability

<1, but not if

>1, but not if

)/h > 0 (10)

<1 , examination of equation (10) shows that

). The trailing bogie is statically unstable for suffi-

a historical review. Proc. CSME Canadian

RAIL VEHICLE DYNAMICS 222

You might also like