You are on page 1of 12

Control Engineering Practice 10 (2002) 887898

Mechatronic developments for railway vehicles of the future


R.M. Goodalla,*, W. Kortum
. b
a

Department of Electronic and Electrical Engineering, Loughborough University, Loughborough, Leicestershire LE11 3TU, UK
b
DLR, Institute of Aeroelasticity, Vehicle System Dynamics, D-82234 Wessling, Germany
Received 26 January 2001

Abstract
Railway vehicles have principally been designed by mechanical engineers since railways began in the early 1800s, i.e. before
electronics and feedback control were invented. Today however they contain substantial amounts of electronic and computer
control, in particular the traction systems, which have been converted entirely. However electronic control can also be applied to the
vehicle suspension and guidance functions, which can provide large improvements in performance. More signicantly, incorporation
of sensors, controllers and actuators into the design process from the start can enable vehicle designers to take advantage of different
mechanical congurations which are not possible with a purely mechanical approachFin other words the true spirit of
mechatronics. The paper reviews the concepts, the current state-of-the-art and opportunities for mechatronic developments for
railways for the future. r 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Railways; Dynamics; Active vehicle suspension; Control

1. Introduction
1.1. Background
The steadily increasing pressure of competition is
forcing the worlds railways to reect economic criteria
in planning procedures to an ever greater extent. This
begins with maintenance for the track infrastructure and
ends with procurement costs for new vehicles. Todays
railway systems are in many respects very cost-intensive
and hence often uncompetitive in comparison with other
modes of transport.
The rail vehicles of tomorrow must therefore be more
cost-effective and energy-efcient. This means that they
need to be lighter and mechanically more straightforward. To this end it is necessary to make use of new
lightweight designs and straightforward mechanical
congurations. An important possibility for achieving
this is by making widespread use of advanced electronic
control apparatus, which needs to be embedded within
the vehicle system from the earliest stages of development.
*Corresponding author. Tel.: +44-1509-227-009; fax: +44-1509227-008.
E-mail address: r.m.goodall@lboro.ac.uk (R.M. Goodall).

The main systems on a railway vehicle for which a


mechatronic approach is appropriate are the suspension,
traction and braking systems. Of these, the suspension
system is of the greatest interest, partly because
traditionally it has been wholly mechanical, but also
because it is more fundamental to the complete vehicle
design than either of the other two. The main emphasis
will therefore be upon the use of active control in
suspensions, although some comments have also been
included about the mechatronic state-of-the-art in these
other areas.
The rst 125 years of railway suspension development
were characterised by a largely empirical approach to
suspension design. The bogie has been invented and
patented in 1809, but it took some 40 years to become
regularly used in Europe after the beginning of public
railways, although it was adopted rather more quickly in
the United States. This Empirical Design Period
extended throughout the rst half of the 20th century,
and some very successful suspension systems were in this
way designed. The use of the word empirical is not
suggesting that no calculations were carried out in the
design, rather that the parameters which determined the
suspensions fundamental performance in terms of
stability, curving, and ride quality were derived from
experience and experiment.

0967-0661/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 7 - 0 6 6 1 ( 0 2 ) 0 0 0 0 8 - 4

888

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

In the 1960s the science of vehicle dynamics came


properly of age, and what can be thought of as the
Analytical Design Period began. Although some of
the analytical ingredients had been known since the
early 1900s, it was not until the 1960s that all the bits
and pieces were put together and effective analysis
became possible (Wickens, 1998; Gilchrist, 1998). At the
same time computers became available to analyse and
predict the performance of complete railway vehicles,
which are still one of the most complex dynamic systems
in engineering and without computer analysis would
remain intractable to analyse. There had of course been
a number of fascinating, ingenious and highly intuitive
designs prior to this time, but it is notable that the new
understanding has enabled much higher operating
speeds than had previously been thought possible, and
has provided mechanical design innovations which
are based upon a fundamental re-think of suspension designFcross-bracing, steering bogies, single axle
bogies, etc.
The arrival of active suspensions has heralded the
Mechatronic Design Period, although it will be seen
that there is still a long way to go in terms of fullling its
promise. It is worth remarking that there have been a
large number of theoretical studies, but the emphasis in
this survey is upon practical implementations. The paper
will also deal with future possibilities and will include
some ideas which so far have only been explored
theoretically, but which might point the way to the
future.
1.2. Application of mechatronics
If it is assumed that there will be actuators, sensors
and controllers at the heart of future railway vehicles, it
is possible to discover ways of exploiting the synergy of
the mechanics and the electronics to achieve a superior
solution. This is of course what the discipline of
mechatronics is about, not just adding electronic control
to an existing mechanical system, but re-designing the
mechanical system to take full advantage of control. So,
what are the possibilities for mechatronics in rail
vehicles?
This can best be answered by considering the way in
which rail vehicle design has evolved historicallyFsee
Fig. 1. In the early days of railways, the vehicles were
mechanically very simple. They were, essentially, boxes
on two sets of wheels. At that time passengers thought
they were wonderful, but in fact their performance was
poor. Demands for higher speed and better ride quality
necessitated evolutionary development over a period of
150 years or more, leading to the basic structure of high
speed passenger vehicles which we have todayFfouraxled vehicles with two bogies which can run stably at
high speeds and go round the curves in a reasonable
manner, and having soft secondary suspensions to

150 years of suspension


development

Effect of Mechatronics

Fig. 1. Evolution in vehicle design.

provide modern standards of ride quality. It is however


clear that these are heavy and mechanically complicated,
and one of the best ways of achieving lighter, simpler
vehicles is to go back to the structure of those early
vehicles.
Most people agree that this is something that cannot
be achieved effectively with passive suspension technology, but can be achieved through active control. The
lower weight and mechanical complexity reduce both
capital cost and running cost, and some of this can be
spent on the provision of the active systems and their
maintenance. The challenge is to design the active
systems so that there is a net benet.
This principle of simplifying the mechanical system by
the use of active steering and active suspensions, in other
words exchanging mechanical complexity for electronic
control complexity, is potentially an extremely important contribution to railway technology, but there are
major technical challenges to be overcome. The following sections of the paper will identify what has been
achieved so far and what challenges remain.
1.3. Research objectives
One fundamental problem with active vehicle systems
is the interaction between structural dynamics, the
forces between wheel and rail, and the control systems.
Accordingly, a proper understanding of the engineering
science is needed for an integrated application of
electronic and mechanical components to achieve the
optimum systems design of railway trains, and it is
possible to identify the following specic objectives:
1. To develop a fundamental understanding of the
dynamic response of lightweight rail vehicles with
active controls.
2. To develop methods of analysis for advanced vehicles
emphasising congurations, which take full advantage of emerging control technology (leading to
reduced weight, lower cost, lower car-body structural
vibrations, etc.).
3. To research systems architectures (sensors, actuators,
processing) which provide the level of safety,
reliability, and maintainability needed for an operational railway.

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

2. Railway vehicle design


2.1. System dynamics
It is useful to explain the essential mechanical
arrangement of a modern high-speed rail vehicle shown
in Fig. 2, which is an arrangement which has evolved
over nearly two centuries of railway operation. There is
the vehicle body for the passengers, and underneath are
two bogies each having two pairs of wheels, connected
by a solid axle to form what is called a wheelset. The
suspension arrangement which interconnects these seven
massesFone body, two bogies, four wheelsetsFis
complicated. There is the primary suspension from the
wheelsets to the bogie, mainly dealing with running
stability. Also there is the secondary suspension from
the bogies to the vehicle body, nowadays usually a very
soft set of airsprings which ensure that the passengers
experience a good ride quality. Both stages of suspension are carefully designed in the lateral and vertical
directions, and some of the rotational modes such as
yawing and rolling are important as well. Of course
what is shown here is just a passive arrangement
where the performance depends upon springs, dampers,
and so on.
Quantifying the level of the complexity is instructive.
A conventional vehicle has 7 main masses, each of which
in principle has 6 degrees-of-freedom (i.e. 3 translational, 3 rotational), resulting in an 84th order set of
differential equations. Fortunately some of these degrees-of-freedom can sometimes be neglected. However
the following are still needed: lateral and yaw for each
wheelset (assuming that vertical motion is constrained
by the track); ve for each bogie and the same for the
vehicle bodyFlateral and vertical plus the 3 rotational
modes pitch, yaw and rollFthe longitudinal motion is
essentially the vehicle speed. This reduces the model to
23 degrees-of-freedom, but a few extra states are needed
to account for exible modes and for airspring

889

dynamics, so a system of order 50 results. It might be


expected that some of the modes are going to be at high
frequency and can therefore be neglected, but analysis of
the system eigenvalues shows that all but a very few will
be below 10 Hz, so their dynamic effects are all
signicant and need to be included.
The previous paragraph highlights the dynamic
complexity, but this is compounded by substantial
non-linearities, especially associated with the properties
of the wheel-rail contact mechanics, which affect the
overall dynamics signicantly.
2.2. Control law formulation
The general scheme for an active suspension is shown
in Fig. 3. This illustrates how feedback loops are added
to enclose the dynamic system, the complexity of which
affects the formulation of the control law: classical or
intuitively based strategies can be created, but their
theoretical performance is confused by the complexity;
modern model-based strategies (e.g. optimal control, H
innity control) can cope with the complexity, but result
in very complex controllers which may be practically
difcult to implement.
Normally of course restricted models will be used for
design of specic active suspensions: an end-view model
for tilting, a side-view model for the vertical suspension,
a plan-view model for the lateral suspension and for
wheelset dynamics. These simplied models are important, but the complexity is still high and there are a
number of other higher frequency modes and non-linear
effects which need to be included, and so the design
problem remains complex.
2.3. Design requirements
Another issue relates to what has to be achieved when
designing an active suspension controller, and Fig. 4
summarises the design requirements. On the left are the
inputs, and these are separated into three types: the
deterministic track inputs describing the intended track
featuresFcurves, gradient, etc.; the random track
inputs describing the irregularities, i.e. the deviations
from the intended track position; and thirdly there are

Mechanical
system

Track inputs

Vehicle outputs
(acceleration,
displacement, etc)

Control
forces
Monitoring
system
(sensors)

Actuator
system

Drive signals

Fig. 2. Railway vehicle scheme.

Electronic
controller

Fig. 3. Scheme of an active suspension.

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

890

Controller

Vehicle system

Body acceleration
(minimise)

Sensors

Actuators

Track features
(deterministic)
Track irregularities
(stochastic)

Suspension deflection
(constrain)

Load changes

Stability
(constrain)
Curving performance
(optimise)

Fig. 4. Design requirements.

the force inputs, principally the changes in load as


people get on and off the vehicle. All these occur in the
vertical, lateral and roll direction.
On the right are the outputs and four items are shown.
The acceleration levels on the vehicle body, which
represents the quality of ride, are to be minimised. The
suspension movements must not become too large, and
must therefore be constrained. It is also necessary to
ensure that there is a minimum margin of stability,
another constraint. And nally there is curving performance, mainly about minimising any wear of the wheels
and the rails, so an optimisation is needed. There are
therefore three input types to consider, two output
measures to minimise and two design constraints to
meet, and the diagram also indicates the inuences from
the inputs to the outputs, the end result of which is a
non-trivial multi-objective design process, particularly
when combined with the dynamic complexity mentioned
earlier.

simulate directly the contact of the wheel on the rail


rather than the contact of the wheelset on the track, e.g.
because of independent wheel simulations.
The design process of modern railway vehicles is
affected by a number of different disciplines inuencing
each other, such as control, exibility, aerodynamics,
hydraulics, etc. In order to apply these facilities to the
model, the packages enabling to simulate the wheel-rail
contact must be extended with new design features. For
example, some MBS packages include simple and
restricted control loops; alternatively the controller
must be added as a user-dened element. The connection of MBS or railway-dedicated packages with control
design packages is a more convenient way for the
designers. For example, the coupling of simulation
codes is described in Veitl et al. (1999). Using different
specialised packages for mechanical and control design
decreases demands on the designer. Despite using
different packages, it is possible to simulate and
optimise whole vehicle models from the earliest design
stages.

3. Mechatronic opportunities
This section provides an overview of the possibilities
that exist for mechatronic railway vehicles. It provides a
progressive overview, starting with the active control
techniques which are currently employed and moving on
through a sequence of opportunities which are increasingly further away from the current state-of-the-art in
railway technology.

2.4. Methodologies and tools


3.1. Tilting trains
Railway vehicle design is still dominated by the
mechanical engineering aspects, but taking advantage
of mechatronic opportunities requires the engineers to
take on board not only the new conceptual approach
implicit in the discipline, but also the design tools which
are required for an integrated approach to mechanical,
electronic and software system design.
Modelling and simulation of railway vehicles is
specic, not only because the model for dynamic
analysis should include models of suspension components, which can be modelled by many simulation tools,
but particularly because of the necessity of a sophisticated model for wheel/rail contact.
Because of its specicity, general modelling and
simulation do not support railway vehicle modelling.
Only some multibody system (MBS) modelling and
simulation packages have implemented the wheel/rail
module. Some dedicated packages were also developed.
The linear and the simplied non-linear theories
developed by Kalker (FASTSIM), (Kalker, 1982), are
used for modelling of wheel/rail contact, because of
many validation measurements. It is preferable to

Tilt is increasingly becoming accepted as standard


equipment for speeding up trains, and a number of
European countries have developed tilting trains: Italy
with the Pendolino concept, Sweden with its X2000
trains, Spain with the Talgo, and more recently
Germany with the VT 611 multiple units and Switzerland with a concept developed by SIG (Various authors,
1997). Further aeld Bombardier makes a tilting train in
N. America, and there are a number of Japanese systems
either in operation or at an experimental stage (Sasaki,
Kato, Konokawa, Sato, & Kakehi, 1996).
The concept of tilt. The basic idea is to lean the
vehicles inwards on curves to reduce the acceleration felt
by the passengers but this acceleration as a vehicle
passes onto a railway curve does not rise suddenly.
There is a transition from the straight to the curve,
usually lasting around 2 s, which is a deliberate design
feature so that passengers are not made uncomfortable
by too sudden an application of sideways acceleration.
Normally the track is leaned inward or canted to
reduce the lateral acceleration experienced by the

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

891

Acceleration
Perceived acceleration

Tilt
Cant
Time
Straight Transition

Curve

Fig. 5. Tilting acceleration responses.

passengers, and this also increases steadily through the


transitionFFig. 5.
At higher speeds the curving acceleration rises, and
the transition will also be more severe because the
duration of the transition will reduce, that is unless the
track is changed. This is where tilt comes in; to bring
the acceleration back to the level it was before.
However, it is not only what happens in the steady
curve that is important, but also the dynamic response
during the transition. Ideally the tilt angle of the body
should rise progressively, perfectly aligned both with the
onset of curving acceleration and the rising cant angle.
However there is a complication, because irregularities
in the track (i.e. imperfections between how the track is
meant to be and how it actually is laid) have the effect of
adding a random higher frequency element to the
acceleration perceived by the passengers. Tilting cannot
affect this because it is principally a purely lateral effect,
and the purpose of the lateral secondary suspension is to
keep these accelerations low enough. In fact, the tilt
system must avoid reacting to these irregularities,
otherwise a number of problems arise.
Mechanical schemes for tilt. A typical arrangement is
shown in Fig. 6, which is a cross-sectional view of the
X2000 train developed in Sweden by ADtranz (Anderson, Bahr, & Nilstam, 1995). The airsprings which form
the secondary suspension sit on top of a tilting bolster,
connected to the bogie by inclined swing links.
Hydraulic actuators tted between the bogie and the
bolster create the tilting action. The inclined swing links
mean that the effective tilt centre is somewhere above
the vehicle oor level, even though the action is applied
below the vehicle body. In Europe there is now a strong
trend to use electro-mechanical actuators, instead of
hydraulic actuators, which was the normal solution in
the early days of tilt. In fact it was recognised that
electro-mechanical tilting provided a better system
solution almost 20 years ago in the UK (Pennington &
Pollard, 1983), but this development was not taken any
further at the time.
Interestingly, the mechanism by which the tilt action
is provided is still changing. Fig. 7 shows the crosssections for both the rst Italian Pendolino trains

Fig. 6. Typical tilting train cross-section.

Fig. 7. Development in tilting train schemes.

developed by Fiat, and the companys newest design


which comes into service in Switzerland in 2000 and UK
in 2001 (Ford, 1998). Originally there were large exicoil
springs for the secondary suspension, the hydraulic
tilting actuators were mounted vertically in the body of
the vehicle, and the pantograph which collects power
from the overhead line was connected by a mechanical
linkage to the bogie in order to avoid it tilting with the
vehicle body and dropping off the wire. In the new
scheme there is a big single airspring in the centre, rather
than two at the sides which is the usual arrangement. A
circular roller beam will be used rather than the inclined
link mechanism, and in line with the trend identied in
the previous paragraph these vehicles will also use
electro-mechanical actuators. Another change is that, in
order to avoid the pantograph moving with the tilting
body, the mechanical linkage used in the early systems

892

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

has been replaced with a separate controlled actuator to


counteract the tilting, i.e. much more of a mechatronic
solution.
Tilting controller strategies. The most intuitive control
approach is to put an accelerometer on the vehicle body
to measure the lateral acceleration which the tilt action
is required to reduce. The accelerometer signal is used to
drive the actuator in a direction which will bring it
towards zero, i.e. a classical application of negative
feedback. There are two problems with this approach.
Firstly, if the passengers are left without lateral
acceleration, a signicant minority experiences motion
sickness. The other problem is interaction with the
lateral suspension, and it can be shown that if the tilt
loop bandwidth is low enough not to interfere with the
lateral suspension, it is then too slow acting on the curve
transition.
The dynamic interaction problem can be avoided by
putting the accelerometer on a non-tilting part, in other
words the bogie. This provides a tilt angle command
signal for a feedback loop using a measurement of the
tilt angle, typically scaled so that the system compensates for around 6070% of the curving acceleration to
avoid motion sickness effects. However, the accelerometer on the bogie is not only measuring the curving
acceleration, but also the pure lateral accelerations due
to track irregularities. Consequently it is necessary to
add an electronic lter to reduce the effects of the
irregularities, otherwise on straight track this results in a
poor ride quality, but this creates too much delay at the
start of the curve.
The solution is to use the signal from the vehicle in
front to provide preview or precedence information,
carefully designed so that the delay introduced by the
lter compensates for the precedence time corresponding to a vehicle length. And this scheme, albeit with
detailed variations, is what most manufacturers now use.
This evolution of control strategies took place over
nearly 30 years. Initially it seemed to be a straightforward control system, and lengthy development time
arose a proper mechatronic systems approach which
recognised the complete problem domain was not
adoptedFthe complexity of the basic dynamic system,
the form of the track excitations, the tilting mechanism,
the sensing possibilities, etc. Nevertheless it can be seen
that there are now clear signs of manufacturers starting
to think mechatronically, resulting in more effective,
better-integrated designs.

bogie and the body, so the paper next looks at how the
idea of active secondary suspensions can be generalised,
principally with the objective of improving the vehicle
response to track irregularities, i.e. to improve ride quality.
Passive suspension characteristics. In a railway vehicle
the primary suspension from the wheels to the bogies is
fairly stiff, the bogies more or less follow the way in
which the track moves vertically as the vehicle travels
along. The secondary springs from the bogie to the body
are there to transmit the low frequency intended
movements so the vehicle follows the track, but at the
same time to isolate the higher frequency irregularities
to provide a good ride quality.
The amount of damping provided for the secondary
suspension is a difcult design trade-off. If it is too low
there will be a lot of activity in the resonant modes; if it
is too high the dampers transmit high frequency track
movements to the vehicle body, and modal dampings in
the region of 20% are typically chosen.
Use of active elements. Active control provides a
solution, because by replacing the dampers with
actuators, measuring the vertical velocity at each end
of the vehicle and making the actuator force proportional to the body velocity, the actuators then apply
damping to an absolute reference, and increased
damping now controls the resonance of the suspension
without making things worse at high frequencies. This is
the concept known as skyhook damping, identied
many years ago (Karnopp, 1978) but still providing the
basis for most active suspensions. Some care in the
controller design is needed because skyhook damping
can create large suspension deections when gradients
and curves are encountered (Li & Goodall, 1999), but
overall the concept is extremely benecial.
This implementation of skyhook damping gives an
important improvement in ride quality, but in fact there
are other things which can be done. For example,
normally the suspension frequency in pitching is somewhat higher than in the vertical or bouncing direction,
whereas it can readily be demonstrated that there is a
denite advantage in having a lower frequency in pitch,
but with an active suspension it is possible to bring
together the signals from the two ends of the vehicle,
separate them into bouncing and pitching and independently control the vehicle modes, in particular to make
the pitching response signicantly softerFsee Fig. 8.

3.2. Active secondary suspension


The introduction of tilting is potentially the tip of an
iceberg: once an active system has been introduced and
accepted for railway operation, the introduction of other
active solutions is technically much easier. Tilting is a
specic form of active secondary suspension between the

Sensor
Actuator Controller

Fig. 8. Active secondary suspension control scheme.

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

The combination of this modal approach and skyhook


damper gives substantial improvements in ride quality,
and both theoretical and experimental investigations
have shown up to 60% reduction in rms acceleration
levels.
Applications. The only active secondary suspension
which has run regularly in service in Europe (apart from
tilt) is a low grade active lateral suspension using
pneumatic actuators on the Fiat Pendolino trains, and
this same facility is provided on some of the newer Fiat
tilting bogies seen in Fig. 7. However, there are also a
number of successful examples of full scale experimental
implementations (Goodall, 1997) which help to understand the mechatronic opportunities.
One of the earliest major experimental study was
undertaken by British Rail in the early 1980s. This
covered both lateral and vertical secondary suspensions
using a variety of actuator technologies, and was the
rst practical demonstration that signicant improvements in ride quality could be achieved, of around 35
50% during mainline testing (Pollard & Simons, 1983).
In the early 1990s there was a development by
ADtranz in the UK, what they called their active
hold-off device. This was designed to improve ride
quality simply by holding the suspension away from the
lateral bump stop on curves, which avoided the high
transient accelerations due to contact with the bump
stops. The system has been tested comprehensively,
including evaluation on a service vehicle (Allen, 1994).
ADtranz in Sweden has looked at a different
conceptFa semi-active electronically controlled lateral damper for use on their X2000 tilting train. Other
people have studied the use of semi-active devices, but
they have usually modied a normal hydraulic damper,
whereas the ABB approach is very novel and patented
(Roth & Lizell, 1995). Essentially it is a displacement
pump driven by the movements of the actuator, and
braked by a miniature electronically controlled brake,
an arrangement which seems to achieve a wide range of
linear operation, and is a good example of a mechatronic approach at a component level rather for the
whole system. Alstom has also looked at semi-active
systems in laboratory tests (ONeil & Wale, 1992),
including the assessment of devices using electrorheological uids, the characteristics of which are
directly affected by the application of an electrical
voltage.
The Japanese are also studying active suspensions.
Hydraulic and pneumatic actuators have been tested on
their WIN350 train which is being used for research and
development aimed at achieving 350 km/h operation for
the Shin Kansen trains (Sakurai, Usonomiya, Yagi, &
Yoshimoto, 1996). A semi-active lateral damper is used
on leading vehicles of the Series 700 Shinkansen trains
(Norinao, 1997) to provide a satisfactory ride quality for
300 km/h operation. There seems to be a lot of Japanese

893

active suspension work, much of which appears to be


closely related. They have a 3 car Try-Z test train for
evaluating 3 different active suspension concepts for
conventional trains, i.e. not only for the Shin Kansen
(Sasaki et al., 1996).
A very signicant development has been conducted by
Siemens SGP in Austria on an experimental vehicle.
They have semi-active secondary vertical dampers, fully
active pneumatic actuators to keep the lateral suspension centred on curves, and electro-mechanical tilt
actuators (mentioned before in connection with trends
in tilting trains). What is particularly interesting about
this development is the way the control systems have
been integrated, including a full set of inertial sensors,
and this almost certainly represents the most advanced
mechatronic implementation which is currently available (Stribersky, Muller, & Rath, 1998).
Most of the serious experimental work at present is
concerned with the lateral direction, because generally
speaking that is the harder design problem. It is
probable that once these have been proved in operation
the deciencies of the vertical suspensions will be
highlighted, and this may prompt a spate of work on
active vertical suspension. For example a laboratory rig
has been used to assess the use of actively controlled
damping for a secondary air suspension (Tang, 1996).
3.3. Controlling wheels and wheelsets
Control of the primary suspension to give active
steering/guidance is a much more substantial step in
technology, and this sub-section reviews the principles
and explains the kind of practical implementations
which have been investigated.
Wheelset dynamics. It is necessary to understand how
the conventional railway wheelset works, which has
been a vital element of railway vehicles since railways
began, but the way it works is not always obvious to
people. A wheelset consists of two coned or otherwise
proled wheels rigidly connected by an axleFFig. 9. On
straight track the wheelset runs in a centralised position,
but when a curve is encountered the wheelset naturally
moves outwards; this causes the outer wheel to run on a
larger radius and the inner on a smaller radius. Being

Wheel flange

Rolling radius
Lateral movement

Solid axle

Track

Fig. 9. Features of a railway wheelset.

894

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

connected by the axle the wheels must still rotate at the


same rotational speed, so the outer wheel moves faster
along the track, and the effect is to make the wheelset go
around the curve. A popular misconception is that it is
the ange which makes the vehicle follow the curve, but
in fact this is not the caseFit is entirely a consequence
of the proling of the wheels and the designers job is to
avoid contact between the ange and the rail otherwise
large amounts of wear results.
Unfortunately, while this inherent curving action is
obviously just what is needed, there is a problem, and
this arises when you start to look at the dynamics of the
wheelset. Its motion occurs both laterally and in the
rotational yaw direction. The forces on the wheelset
arise from so-called creepages between the wheel and
rail, small relative velocities which arise because of
elastic deformation of the steel at the point of contact
and which apply in both the longitudinal and the lateral
directions. The overall effect is an instability, principally
a kinematic oscillation (Wickens, 1969). Adding mechanical dampers does not stabilise the wheelset, and it
is necessary to add springs, and the normal solution is to
have two wheelsets connected via longitudinal and
lateral springs within a bogie as seen in Fig. 10. The
lateral springs are necessary to transmit the curving
forces, but the longitudinal springs are mainly there to
stabilise the wheelsets, essentially by providing a yaw
stiffness.
However on a curve these stabilising longitudinal
springs produce forces which interfere with the natural
curving action of the wheelset. The result is that on the
tighter curves the wheel ange will be in contact with the
side of the rail, causing wear of the wheels, wear of the
rails and often signicant amounts of noise. There is
therefore a difcult design trade-off: stiff springs give
stable high-speed running, but poor curving; soft springs
mean that the curving performance is better, but stable
running is only possible at low speeds. Great ingenuity
has been applied in nding mechanical solutions to this
design trade-off (Illingworth & Pollard, 1982): crossbraced bogies with carefully designed linkages between
the wheelsets; steering bogies with linkages to the
vehicle body which try and get the wheelsets into more
or less the right position on steady curves (Von

Madeyski, 1999), but none of these fundamentally


overcomes the problem which has been outlined.
Active control of wheels and wheelsets. It is now
possible to discuss where control ts in, because
appropriate use of active control can provide new levels
of performance not achievable with a conventional
passive bogie.
One possibility is where the longitudinal springs seen
in Fig. 10 are replaced by actuators, and this enables
characteristics to be obtained which either are not
possible with a purely mechanical solution, or at least
extremely difcult in practice. In the more general case a
wheelset can be actively connected in both directions,
including directly to the vehicle body (i.e. without a
bogie), which provides the basis for identifying a range
of possible control approaches. These are outlined in the
sections which follow, although it is important to
emphasise that these are very much concept diagrams
to describe the basic idea, and a lot of extra engineering
detail would be needed in practice.
The rst option in Fig. 11 shows the longitudinal
springs in series with the actuatorsFthe idea here is that
the higher frequency oscillations of the wheelset are
stabilised by the springs, with low bandwidth active
control provided by the actuators, essentially to relax
the force they produce when the vehicle goes round
curves so as to allow the wheelset to take up its natural
curving position. This approach has been studied
applied to a conventional bogie, and it is been shown
possible to get extremely good performance with really
quite a simple control law which minimises the wheelset
yaw torque applied at low frequencies (Shen & Goodall,
1997). Actuator power levels are extremely low, a few
tens of watts only, which means that practical implementation is not a problem. The same idea has also been
applied to single-axle running gear in Germany, and
forms the basic scheme in some integral railcars,
manufactured in Austria and now running in service,
(Sommerer, 1999).
Fig. 12 shows another approach, which is the other
way round to the previous paragraphFpassively

Frame/vehicle body

Actuator

Fig. 10. Mechanical scheme of a conventional bogie.

Fig. 11. Actively steered wheelset.

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

Rotary
actuator

895

Rotary
actuator

Bearings

Fig. 12. Actively stabilised wheelset.

steered, actively stabilised. As mentioned earlier, conventional dampers do not stabilise the wheelset; however, applying a yaw torque to the wheelset which is
proportional to the lateral velocity of the wheelset
produces a form of active damping which is stabilising,
and a detailed analysis shows that it does not interfere
the natural curving either. In fact a number of control
laws are possible for this scheme, not only the active
damping principle but other ideas as well (Mei &
Goodall, 1999).
Wheelsets with independently rotating wheels. A
modication of the basic wheelset which railway
suspension designers have looked at many times over
the years is wheelsets in which the wheels are free to
rotate independently on the axles (Dukkipati, Narayanaswamy, & Osman, 1993). It is commonly stated that
this removes the instability because the two wheels are
no longer connected mechanically. In fact if you analyse
this carefully this is not the case (Goodall & Li, 2000),
but stabilisation is much more straightforward, and in
this case dampers are effective. However, the natural
curving action disappears, again because the connection
between the wheels has disappeared, and so active
control in this case is there primarily to steer the
wheelset through the curves, as Fig. 13 shows.
Fig. 14 presents an important alternative idea which
can be used with independently rotating wheelsFrather
than use actuators, it is possible to control the torque
being applied to the two wheels separately, the sum of
the torques providing propulsion and braking as
normal, with the difference being used to steer the
wheelset. This idea has been investigated in Germany
and demonstrated on a laboratory rig (Gretzschel &
Bose, 1999). Also in Germany there is an experimental
vehicle which also controls the relative motion of
adjacent wheels (Frederich, 1999), and Fig. 15 shows
the wheelset assembly.
Directly steered wheels. So far the paper has looked at
applying control to conventional wheelsets, but the next
scheme, seen in Fig. 16, is totally different because there
is no axle. The wheels are pivotted about a vertical axis,
and there is a track rod connecting the wheels to keep

Fig. 13. Wheelset with independently rotating wheels.

TL

TR

Fig. 14. Torque-controlled independently rotating wheels.

Fig. 15. Torque-controlled running gear.

Fig. 16. Directly steered wheels.

896

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

them parallel. There are a number of ways in which


steering can be achievedFhere the steering angle is
directly adjusted by an actuator, but differential torque
control is another choice (Powell & Wickens, 1995) i.e.
the idea described in the previous paragraph but applied
via a different mechanism. The wheels are of course now
completely independent and must therefore be steered.
Here we can see signicant mechatronic inuences
starting to appear, because the basic mechanical scheme
has been fundamentally changed to accommodate the
benets provided by active control.
Range of options for active wheel/wheelset control. The
preceding sub-sections have indicated a great variety of
possibilities for achieving active steering. Firstly, there is
the control approachFis the control being used to steer,
to stabilise or both? Secondly, what method is providing
the controlFactuator control or torque control? Also
to be decided is the wheel congurationFnormal solidaxle, independently rotating wheels on an axle, or
steered wheels? Superimposed on these choices is also
the type of control hardwareFsensors, controllers
actuators, etc.Fand so the decision as to which is the
best solution for a particular application is complex and
still open, and it is clear that signicant research is still
needed to help resolve some of these issues.
3.4. Traction and braking
Electric traction systems have been common in
railways for many years, and the addition of electronics
and control has mainly been associated with exploitation of power electronic drives. Probably the main
consequence of this has been the ability to replace d.c.
motors with smaller and more efcient a.c. motors. In
general this has resulted in an easing of the mechanical
design, rather than the kind of fundamental change
which follows from mechatronic integration, although
there have been cases of vehicles for low speed systems
in which hub motors have been directly connected to
the wheels, thereby removing the need for a gearbox
and/or mechanical drive and creating substantial simplications in bogie design. The other signicant impact
is that the traction system has been increasingly able to
provide braking effort through electrical regenerative
action, although the net effect upon the main mechanical braking system has so far been relatively small.
Advanced control can provide a much more effective
use of the adhesion which is available at the wheel/rail
interface, and there are examples of theoretical and
experimental studies which demonstrate the ability
through the sophisticated power control to operate at
the peak of the adhesion/slip characteristic, (Beck &
Engel, 1996; Schwartz & Pfeiffer, 2000). Wheelslide
protection of railway braking systems has been tted for
many years, but the pneumatic brake actuators which
are predominant in the industry are insufciently fast to

do other than provide fairly simple on-off control when


wheelslide is detected. Recent examples of wheelslide
protection are making much more extensive use of the
sophistication which is possible through electronic
control (Anon, 1999). In addition, a number of braking
system manufacturers are investigating more advanced
braking actuators (e.g. electro-mechanical systems), at
least in part because of the enhanced control capability
which these will provide, and in many senses these
developments reect the changes away from traditional
mechanical engineering solutions which have also been
seen with the suspension systems and towards a
mechatronic approach.

4. Future mechatronic vehicles


As mechatronic trains develop more and more
electronics will appear within railway vehicles, and more
profound integration of the various systems is possible.
The propulsion system already has a lot of electronics,
and this can be linked in with active steering. As
mentioned earlier, the possibility has already been
recognised of avoiding the need for steering actuators
if the traction motors driving the wheels are differentially controlled to achieve the steering action. Obviously this needs a high level of integration between the
suspension and drive control systems, particularly to
accommodate the requirements of safety criticality. An
exciting new concept is the idea of a wheel motor in
which there is mechanical integration of the wheel and
the traction motor (Shelly, 1999)Fsee Fig. 17.
It must be remembered that the suspension, guidance,
propulsion and braking forces of a railway vehicle all
pass through the contact point between the wheels and
the rail, a patch of compressed metal perhaps 2 cm in
diameter. There is a tendency to design these systems

Fig. 17. Wheel motor.

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898
Vehicle motions
Suspension actions
Drive actions

Sensors

Drive
controller

Motor/
brakes

Suspn.
controller

Suspn.
actuators

Steering
controller

Steering
actuators

Vehicle
dynamics

Control system

Vehicle

897

A third generation of systems offering a further


increase in functionality, probably highly integrated
with other vehicle dynamic systems such as traction
and braking, and using information from facilities
such as track prole databases, satellite positioning
system etc.

5. Conclusions

Steering actions
Track information

Track database

Vehicle position

Fig. 18. Integrated control scheme.

independently, but once they are all electronically


controlled, it will be possible to optimise the use of this
contact patch through an integrated control system, and
Fig. 18 shows an overall scheme. There will be a set of
sensors measuring what is happening, a controller which
carries out drive, suspension and steering control, and
the forces produced all act upon the vehicle dynamic
system. The diagram also shows a track database system
which could contain the intended alignment of the track,
although important questions arise regarding the
accuracy of the track information. Nevertheless, this
kind of conceptual approach raises some extremely
interesting possibilities.
So it is possible to look forward to a full implementation of mechatronic principles for the design of running
gear, using single wheels in which electronics is applied
to an integrated control system, controlling traction,
braking and suspension forces in an optimal way. The
concept, in particular the use of active steering, is
potentially liberating from the operational viewpoint,
because in principle it is possible to dictate the direction
through switches from the vehicle rather than the track.
Conventional track switches could disappear such that
the track is continuous in both directions through the
switch, with the sensors causing the vehicle to follow
one route or the other, and if this is combined with
electronically rather than mechanically coupled
vehicles all sorts of possibilities arise which could
signicantly enhance the operational exibility of railway operation.
Most of the developments which exist at present are
what might be called rst generation systems, having
limited functionality with restricted control laws, usually
applied locally to the actuators. It is, however, possible
to envisage subsequent generations for which the
mechatronic opportunities (and challenges) will be much
greater:
A second generation of systems with a higher degree of
functionality (e.g. lateral and vertical), integrated
measurement systems for higher reliability and faulttolerance, and control laws which take a more complete
view of the vehicle or train system.

Everywhere one looks at the technology of transport


vehicles, the increasing importance of mechatronic
solutions to vehicle design is apparent. Aircraft have
essentially already made the transition, with full
integration of controls and computation and the
mechanical/aerodynamic structure through y-by-wire.
The automotive industry has also made substantial
changes with electronic engine controls and anti-lock
braking systems, and other changes are imminent:
brake-by-wire, active stability control, and steer-bywire.
The railway industry is following, but given the
signicantly longer product life cycle it is probably
inevitable that it is lagging behind the aircraft and
automotive industries. Nevertheless active tilting is now
well established, and a number of practical developments have been reported recently which involve a high
degree of control applied to the running gear. It is
interesting to see that, although the concepts of active
secondary suspensions have been understood for many
years, the actual take-up on real, operational vehicles is
relatively low. They only really affect ride quality, for
which the commercial return is relatively low, whereas
the burgeoning interest in various forms of active
steering is motivated by substantial changes in vehicle
conguration. The engineering challenges are greater,
but the operational benets are likely to be substantial if
safe, reliable, cost-effective solutions can be developed.
It seems inevitable that rail vehicles will become
increasingly mechatronic in some form or another,
and comparison with the other industries mentioned
makes it clear that such an evolution (or revolution?) is
strategically important for railways. An EC research
project, in which both authors were involved, has made
an important contribution to the evolution of this new
mechatronic technology for trains, the aim being to
identify the opportunities, to provide the scientic basis
upon which practical applications can be developed, and
to disseminate the research results and potential benets
(Ellis & Goodall, 1999).

References
Allen, D. H. (1994). Active bumpstop hold-off device. Proceedings of
the railtech 94. IMechE Paper C478/5/013, Birmingham, UK.

898

R.M. Goodall, W. Kortum


. / Control Engineering Practice 10 (2002) 887898

Anderson, E., Bahr, H. V., & Nilstam, N. G. (1995). Allowing higher


speeds on existing tracksFdesign considerations of the X2000
train for Swedish State Railways. Proceedings of the IMechE Pt F,
209(2), 93104.
Anon. (1999). Westinghouse Brakes wheelslide protection system.
Modern Railways, 336.
Beck, H.-P., & Engel, B. (1996). Traction drive control with PI
state controller and Kalman lterFrst experimental results.
Proceedings of the 13th IFAC congress, San Francisco, Vol. P
(pp. 343348).
Dukkipati, R. V., Narayanaswamy, S., & Osman, M. O. M. (1993).
Independently rotating wheel systems for railwaysFa state-of-theart review. Vehicle System Dynamics, 21, 297330.
Ellis, B., & Goodall, R. M. (1999). The mechatronic train: Requirements and concepts. Proceedings of world congress on railway
research, Tokyo, Japan, October, 1999.
Ford, R. (1998). Tilting trains hold the key to Virgins ambitious
franchise. Railway Gazette International, October, 1998 (pp.
707710).
Frederich, F. (1999). Nullebenen-Konzept der Spurfuhrung.
.
Glasers
Annalen, 123(7/8), 269278.
Gilchrist, A. O. (1998). The long road to the solution of the railway
hunting and curving problems. Proceedings of the IMechE Part F,
212, 219226.
Goodall, R. M. (1997). Active railway suspensions: Implementation
status and technological trends. Vehicle System Dynamics, 28,
87117.
Goodall, R. M., & Li, H. (2000). Solid axle and independently rotating
railway wheelsetsFa control engineering assessment of stability.
Vehicle System Dynamics, 33, 5767.
Gretzschel, M., & Bose, L. (1999). A mechatronic approach for active
inuence on railway vehicle running behaviour. Proceedings of the
16th IAVSD symposium dynamics of vehicles on roads and tracks,
Pretoria.
Illingworth, R., & Pollard, M. G. (1982). The use of steering
suspension to reduce wheel and rail wear in curves. Proceedings
of the Institution of Mechanical Engineers, 196, 379385.
Kalker, J. J. (1982). A fast algorithm for the simplied theory of rolling
contact. Vehicle System Dynamics, 11, 1113.
Karnopp, D. (1978). Are active suspensions really necessary? ASME,
78-WA/DE/2.
Li, H., & Goodall, R. M. (1999). Linear and non-linear skyhook
damping control laws for active railway suspensions. Control
Engineering Practice, 7(7), 843850.
Mei, T. X., & Goodall, R. M. (1999). Wheelset control strategies for a
2-axle railway vehicle. Proceedings of the 16th IAVSD symposium,
Pretoria, August, 1999.
Norinao, H. (1997). Active and semi-active suspensions smooth
300 km/h ride. Railway Gazette International, April 1997 (pp.
241242).
ONeil, H. R., & Wale, D. (1992). Improvements in the ride quality
using a semi-active lateral suspension system. Proceedings of the

IMechE seminar S043, railway dynamicsFfor today and tomorrow,


IMechE, London.
Pennington, K. W., & Pollard, M. G. (1983). The development of an
electro-mechanical tilt system for the advanced passenger train.
IMechE Conference Paper C299/83 (pp. 2128).
Pollard, M. G., & Simons, N. J. A. (1983). Passenger comfortFthe
role of active suspensions. Proceedings of the IMechE, 198D(35),
115.
Powell, A. J., & Wickens, A. H. (1995). Active guidance of railway
vehicles using traction motor torque control. Proceedings of the
14th IAVSD symposium, Ann Arbor, MI, USA.
Roth, P.-A., & Lizell, M. (1995). Lateral semi-active damping system
for trains. Proceedings of the 14th IAVSD symposium, Ann Arbor,
MI, USA.
Sakurai, K., Usonomiya, M., Yagi, Y., & Yoshimoto, Y. (1996).
Development of a new Shinkansen vehicle series 500 for realization
of 300 km/h commercial operation. Proceedings of the STech 96,
Paper C514/031/96, IMechE.
Sasaki, K., Kato, H., Konokawa, T., Sato, Y., & Kakehi, Y. (1996).
Active tilting control of series E991 e.m.u. experimental train
(development of third generation active tilting control). Proceedings of the STech96, Paper C514/055/96, IMechE.
Schwartz, H. J., & Pfeiffer, R. (2000). Wheel creep control for light rail
vehicles. Railway Technical Review, No. 2 (pp. 412).
Shelly, T. (1999). Motors make the wheel go round. Eureka, 19(5),
3839.
Shen, G., & Goodall, R. M. (1997). Active yaw relaxation for
improved bogie performance. Vehicle System Dynamics, 28(45),
273289.
Sommerer, R. (1999). IntegralFein Gliederzugsystem fur
. den
Vollbahnbetrieb. Glasers Annalen, 123(2), 7384.
Stribersky, A., Muller, H., & Rath, B. (1998). The development of an
integrated suspension control technology for passenger trains.
Proceedings of the IMechE Pt F, 212, 3342.
Tang, J. S. (1996). Passive and semi-active airspring suspensions for
rail passenger vehiclesFtheory and practice. Proceedings of the
IMechE, 210, 103117.
Various authors (1997). Proceedings of the IMechE seminar S479 on
tilting trains for the UK?
Veitl, A., Gordon, T., Van de Sand, A., Howell, M., Valasek, M.,
Vaculin, O., & Steinbauer, P. (1999). Methodologies for coupling
models and codes in mechatronic system analysis and design.
Proceedings of the 16th IAVSD symposium, Pretoria, August, 1999.
Von Madeyski, T. (1999). Running-gear technologyFinteractions with
the track and the vehicle body. Railway Technical Review, Nos. 34
(pp. 1017).
Wickens, A. H. (1969). General aspects of the lateral dynamics of
railway vehicles. Journal of Engineering for Industry and Transportation, ASME Series B, 9(3), 869878.
Wickens, A. H. (1998). The dynamics of railway vehiclesFfrom
Stephenson to Carter. Proceedings of the IMechE Part F, 212,
209217.

You might also like