Вы находитесь на странице: 1из 6

Control Engineering Practice 10 (2002) 10231028

Mechatronic solutions for high-speed railway vehicles


T.X. Meia,*, Z. Nagyb, R.M. Goodallb, A.H. Wickensb
a

School of Electronic and Engineering, University of Leeds, Leeds, LS2 9JT, UK


b
Loughborough University, Loughborough, LE11 3TU, UK
Received 24 September 2001; accepted 26 October 2001

Abstract
This paper presents recent developments in the study of active steering for railway wheels. It demonstrates the potential benets
from this form of active control and studies what could be achieved when modern control techniques are used on the vehicles via
mechatronic components. The study is based on a mechanically simplied vehicle and active control strategies for three wheelset/
wheel-pair congurations are studied, which are solid axle wheelset, independently rotating wheelset and directly steered wheel pairs.
Various mechatronic vehicle congurations are discussed and different control schemes are presented. Two key performance
requirements, i.e. the curving performance of the wheelset and ride quality of the vehicle, are assessed for all proposed schemes.
Finally, actuation requirement is also investigated. r 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Active control; Railway vehicle; Solid axle wheelset; Independently rotating wheelset; Steered wheels

1. Introduction
The design of railway vehicles has been principally
a mechanical engineering discipline since railways
began, but this situation is starting to change.
Incorporation of electronic controls within the suspension system has become common through the use of
active tilting which enables vehicles to go faster through
curved sections of track, and these ideas are being
extended to the use of active secondary suspensions
(i.e. between the bogie and the body) which improve
the quality of ride experienced by the passengers
(Goodall, 1997). Certainly these are mechatronic
solutions because they provide solutions which are
not possible with purely mechanical suspensions. However the benets of a mechatronic approach become
more signicant if the vehicle as a whole is considered,
which leads to the idea of providing active control of
the wheels and wheelsets. Some work has been undertaken in this area (e.g. Aknin, Ayasse, & Devallez,
1991; Anon, 1997), but the important consequence
is that it enables new vehicle congurations to be
*Corresponding author. Department of Electronic and Electrical
Engineering, Loughborough University, Leicestershire, LE11 3TU,
UK. Tel.: +44-1509-228-105; fax: +44-1509-22-854.
E-mail address: t.x.mei@lboro.ac.uk (T.X. Mei).

proposed which are mechanically simple compared


with the four-axle vehicles which are normal for modern
high-speed trains, for example two-axle vehicles without
bogies, or more advanced articulated congurations.
The mechanical simplicity potentially results in large
reductions in weight, propulsion energy, and track
damage. The analysis described in this paper develops
active schemes for controlling the conventional solid
axle wheelset, wheelset with axles on which the wheels
are free to rotate independently, and wheel-pair without
axles.

2. Vehicle congurations
A two-axle vehicle is used in the study and the
parameters of the vehicle scheme are based upon the
requirements for future high-speed passenger vehicles
having a design speed of 300 km/h, the vehicle being
16 m long and 9 m between axles. Fig. 1 shows a plan
view of the vehicle with passive stabilisation. The vehicle
scheme mainly consists of a vehicle body, with two
wheelsets connected to the body via the lateral suspensions. Values of the springs and dampers in the lateral
direction are chosen to give a typical suspension
frequency of a little less than 1 Hz to give the required
ride quality.

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 1 ) 0 0 1 5 7 - 5

1024

T.X. Mei et al. / Control Engineering Practice 10 (2002) 10231028

Wheel

Wheelset
Steering lever
Frame

Passive
stabilisation

Vehicle body
Track Rod
Fig. 2. Directly steered wheel-pairs.
Fig. 1. Two-axle vehicle with passive stabilisation.

3. Active steering schemes


Almost all railway vehicles use the conventional
solid axle wheelset, where two coned/proled wheels
are rigidly connected to a common axle so that
they have to rotate at the same speed. The advantage
of this arrangement is that the wheelset has the ability
of natural curving/centring, but an unconstrained
solid axle wheelset is also unstable at non-zero speeds
and exhibits a sustained oscillation in the horizontal
plane, a phenomenon known as wheelset hunting
(Wickens, 1998). A typical passive vehicle therefore
also consists of springs connected from the wheelsets
to the body of the vehicle in the yaw/longitudinal
direction as shown in Fig. 1 to overcome the wheelset
hunting problem. However this added stiffness is well
known to interfere with the natural curving action of the
wheelset and deteriorate wheelset performance on
curves, resulting in largely increased creepage and
possibly severe wear of the wheels and rails. A
compromise between the wheelset stability and the
curving performance has always had to be made in the
design of railway vehicles.
To reduce the creep force between the wheels and
rail, railway engineers came up with a different wheelset
conguration. This new conguration also consists
of two wheels mounted onto a common axle, however
the two wheels are allowed to rotate independently
from each other (hence the term independently rotating
wheelset or IRW in short) (Dukkipati, Narayana
Swamy, & Osman, 1992). An even more radical
approach is to remove the axle all together (Wickens,
1993), and the two wheels are mounted onto a common
frame as shown in Fig. 2. The two wheels can then
be steered directly via a track rod and hence termed
directly steered wheel-pairs (DSW). Because the
constraint on the relative rotation of the two wheels
is removed, the longitudinal creep force and hence
the wear of the wheels is signicantly reduced. On
the other hand, the natural curving/centring ability
of the solid axle wheelset is no longer present for the
IRW or DSW and some form of guidance control
becomes necessary.

The development history of railway vehicles has


evidently shown that it is extremely difcult to solve
all the design difculties effectively with only the passive
means available. Therefore the best way (and arguably
the only feasible way) forward is to take a radical
approach by applying advanced control concepts to
steer wheelsets actively via mechatronic components.
A number of schemes are possible for the active
steering of railway wheelsets. In Fig. 3, an actuator is
placed between each wheelset and the body in the yaw
direction. This conguration can be used to replace the
passive stabilisation for the solid axle wheelset and to
stabilise the system without compromising the wheelset
curving performance, and similarly it can also be used to
steer the independently rotating wheelset for both
guidance control and stabilisation.
An alternative conguration is to apply a lateral force
between the body and each wheelset via the actuators in
parallel with lateral suspensions. It can be readily shown
that this scheme can be used to stabilise the solid axle
and independently rotating wheelsets. An adverse effect
of this conguration is that it affects the lateral mode of
the vehicle body and, as it will be shown later, worsens
the passenger ride comfort when used.
The dynamic characteristics of a DSW are very
similar to the independently rotating wheelset. The

Wheelset
Control
torque

Vehicle body

Fig. 3. Two-axle vehicle with active control.

T.X. Mei et al. / Control Engineering Practice 10 (2002) 10231028

longitudinal creepage of a non-powered wheel is almost


zero, but extra control actions are required. Because of
the special mechanical arrangement, actuators can be
readily installed to drive the track rod to control the
angle of the wheel-pairs relative to the track to achieve
both stabilisation and guidance.
Although, this paper concentrates upon active steering schemes using actuators, there are other possibilities.
E.g. a different scheme is proposed for the active control
of the independently rotating wheelset, where a differential rotation torque between the two wheels on the
same axle is used to provide the guidance control
(Gretzschel & Bose, 2000; Perez, Busturia, & Goodall,
2000).

4. Control strategies
Controls of the solid axle wheelsets and IRW/DSW
have quite distinct requirements because of their
particular characteristics/limitations. For the solid axle
wheelset, the active controller must stabilise the wheelset
to avoid kinematic oscillation, but must do this in a way
either which does not interfere with the natural curving
action or which replicates this natural action by actively
steering the wheelset to follow the pure rolling line on
curves. Excellent curving performance is an important
requirement, because for a passive two-axle vehicle this
is difcult to achieve especially when high-speed stability
is also required. On the other hand, the independently
rotating wheelset and DSW require both stabilisation
and guidance control. Achieving a pure rolling action
for IRW/DSW is not as important, but the active
controller must provide the missing curving action such
that the wheelset follows the track and the ange contact
is avoided.
In addition, railway track comprises two distinct types
(deterministic features and random irregularities) and
those two track features also inuence the development
of active control schemes. The wheelset is required to
follow the deterministic track input, but at the same time
the ride quality of the vehicle should not be compromised on the random track input. Reconciling a
controller design to accommodate both these input
types effectively is a non-trivial problem.
4.1. Solid axle wheelset
For the solid axle wheelset, stabilisation is the key
issue. It can be readily shown that the stabilisation via
active means can be achieved by either applying a yaw
torque proportional to the lateral velocity of the
wheelset or a lateral force proportional to the yaw
velocity (Mei & Goodall, 1999a). If the required
feedback can be obtained to be relative to the local
track reference, the natural curving action of the

1025

wheelset will not be affected by the stabilisation effort.


A detailed design to apply the two concepts can be
carried out by adopting a modal control technique,
where the vehicle model is decoupled into two independent subsystems. One of these subsystems may be
regarded as a wheelset attached to a mass of the body
lateral motion and the other as a wheelset attached to a
mass of the body yaw motion. Two separate controllers
may then be developed for the two subsystems
independently using classic frequency domain techniques (Mei & Goodall, 2000a). The intuitively formulated
controllers are derived based upon the fundamental
understanding of the system, which provides the
engineering insight of the problem. However the
approach can be difcult to use for complex systems
and/or for multi-objective designs.
Optimal controllers (H2) are also developed, where a
cost function is dened to minimise the tracking error of
the wheels. Depending on the approaches taken, the
optimal controllers can have just a proportional feedback gain matrix with parameters tuned to stabilise the
wheelset without interfering with the natural curving
(Mei & Goodall, 2000). Alternatively an integral action
may be added to force the wheelset to follow the pure
rolling lines (Mei, Perez, & Goodall, 2000).
4.2. Independently rotating wheelset
For the independently rotating wheelset, the wheelset
stabilisation is fairly straightforward and one option is
to use yaw damping between the wheelsets and vehicle
body (actively or passively). However the development
of an intuitively formulated controller for the missing
guidance control can be problematic because some
essential feedback measurements such as the wheelrail
deection and/or angle of attack are not readily
available. Those measurements are difcult to obtain
in practice, but studies have shown that it is possible to
use state estimation techniques to estimate the required
variables from simple practical sensors (Mei, Goodall, &
Li, 1999).
Modern controllers using optimal (H2) and robust
(HN) approaches are also developed for the vehicle with
IRW wheelsets (Mei & Goodall, 2001). Because the two
wheels are free to rotate, the design are concentrated on
issues such as guidance, ride quality, parameter uncertainties, actuator dynamics and practicality of sensing arrangement.
4.3. Directly steered wheels
The control strategies developed for the independently rotating wheelset can be readily applied to this
conguration, however this section concentrates on
control approaches using the track rod shown in Fig. 2
to steer the wheel angles.

T.X. Mei et al. / Control Engineering Practice 10 (2002) 10231028

It can be shown that a proportional control with the


wheelrail deection as feedback can be used to both
stabilise and provide the tracking guidance. But this
simple control structure may cause undesirable tracking
errors on tight curves as quasi steady state errors are
necessary for the scheme to apply the required steering
angles in opposite directions for the front and rear
wheel-pairs of a vehicle. By adding an integral action to
the controller, the tracking errors will be signicantly
reduced.
Therefore two controllers are designedFone takes
the classical structure and the other optimal control. The
common feature of the two controllers is that they both
have proportional and integral terms.

2.5

Passive (front)
2

Lateral displacement (mm)

1026

2.5

Active (solid)

2.5

Active (IRW)
0

Active (DSW)
-0.5
0

10

Time (s)

Fig. 4. Curving performance (Vs 83:3 m/s).

4.4. Actuator dynamics


All controllers presented above are implemented via
actuators, therefore it is important to study the
characteristics of the actuators and their effect on the
closed loop systems. Several approaches are available to
address the issue. The most direct approach is to include
the actuator models in the control design. Unfortunately
this often leads to very complex control structures in
particular when the actuator model is highly non-linear.
Alternatively, actuator dynamics may be disregarded at
the design stage and then included in the simulation to
evaluate the performance. Usually the simulation will
include a fast independently designed inner control loop
based upon force or actuator displacements whichever is
appropriate. This approach is very simple, but the
system stability and performance is not always guaranteed. A third approach is to design controllers with
guaranteed robustness or stability margins, which will
ensure the stability and performance without resulting in
over complicated controllers. A combination of the
second and third approaches is used in this study.

Passive
(rear)

the rear wheels behave similarly except for a time delay),


where the vehicle speed is 83.3 m/s and the curve radius
is 3500 m. With active control for the solid axle
wheelsets both wheelsets move outwards in the lateral
direction just enough to allow the natural curving on the
curve. However the wheelsets of the passive vehicle are
forced away from the ideal positions by the stabilising
yaw stiffness in the opposite directions, resulting in
increased creepage between the wheels and the track and
hence undesirable wear. Also shown in Fig. 4 are the
responses of the actively steered IRW and DSW. The
lateral displacement relative to the track is further
reduced by the active control and there is no need to
worry about the natural curving. The lateral displacement of the directly steered wheels is zero on the steady
curve, because of the extra integral action of the
controller used. When railway vehicles negotiate tighter
curves at lower speed, the advantage of the active
controls is even more obvious as indicated in Fig. 5
where the vehicle speed is 25 m/s and the curve radius is
300 m. The large wheelrail displacement exhibited by
the passive vehicle in this case would in practice cause

5. Performance assessment
20

Passive (front)
Lateral displacement (mm)

To obtain a fair assessment of the vehicle performance with the control schemes developed, two different
track inputs are used in the study. The deterministic
track input, including a constant curve with a cant angle
connected to straight track via transition sections of 1s
at the both ends, is used to assess the curving
performance of the wheelsets. The random track input
represents the roughness of a typical high-speed main
line, which is used to study the track following of the
wheelsets and the ride quality on the vehicle body.
For the deterministic track input, all active control
schemes developed for the vehicle with solid axle
wheelsets give much improved curving performance
when compared with the passive vehicle. Fig. 4 shows a
typical result (front wheels only for the active schemes as

15

10

Active (solid)
5

Active (IRW)
0

Active (DSW)
-5

Passive (rear)

-10
0

Time (s)

Fig. 5. Curving performance (Vs 25 m/s).

10

T.X. Mei et al. / Control Engineering Practice 10 (2002) 10231028

ange contact and noise, but this non-linearity is not


simulated in the model. With active controls the solid
axles take up a lateral wheelrail position which
corresponds to pure rolling, whereas there are very
small movements for the independently rotating wheelset and DSW.
As mentioned above, the active control schemes
should also improve the ride quality on the vehicle
body. Table 1 gives a comparison of the ride quality
(normalised by the passive vehicle) of different control
strategies, calculated at vehicle speed of 300 km/h
on straight track with random irregularities. Compared
with the passively stabilised vehicle, all active controls
except for the one with lateral actuators improve
the ride quality signicantly (26% for the solid axle
wheelset and as much as 3649% for the independently
rotating wheels). This is principally because the kinematic modes of the wheelsets are better controlled and
the body modes are not affected by the control actions.
Compared with the solid axle wheelsets, a further
improvement of 10% on ride quality is obtained for
the vehicle with IRW and 20% for the directly steered
wheels.

6. Actuation requirement
One of the key areas that must also be addressed in
the study is to identify the actuator requirements and to
study the effect of actuator dynamics on the performance of the active schemes. Results in Table 1 suggests
that the requirement of the control effort is dominated
by the requirement on random track, which is much
higher than that required for the wheelset curving
action. This is particularly true for high-speed applications as the effect of the track roughness becomes worse
when the vehicle travels faster. As indicated in Table 1,
the vehicle congurations with the yaw actuators require
about 6 times less power than the schemes with lateral
actuators, a further indication that the yaw control is a
better option. The independently rotating wheelset
apparently requires much smaller control effort than

1027

the solid axle wheelset, approximately at a ratio of 1:10.


The directly steered wheels have similar power requirement. This is because the constraint on the rotation
motion of the wheels is removed and the creep force is
consequently reduced signicantly.
A useful observation drawn from the study is that the
torque (or force) demand is relatively high, but the
average power requirement is rather low because of the
low actuator velocity. This raises an interesting question
about the gearing of actuators. Most practical actuators
available at present lack the ability of generating
relatively high torque/force at low velocity, which means
that the selection of actuators may well be decided by
the torque/force requirement rather than the power
requirement unless actuators can be developed to
accommodate both.
The response speed of actuators is another important
issue. Typical hydraulic and electro-mechanical actuators with a closed loop control can achieve a frequency
bandwidth of 1520 Hz, which should be sufcient for
implementing the active controls, but their response
speed and dynamic behaviour are seriously compromised by the actuator velocity. Even for the low velocity
applications investigated in this study, the wheelset
performance as well as the stability of the overall system
can be noticeably affected and caution should be taken
in the design to accommodate this if those actuator
technologies are to be used. Electro-magnetic actuators
on the other hand will be able to give a much faster
response to torque/force demand and the only signicant issue is how to reduce the size of the actuators. The
fact that the power requirement is low (only 10s watts
for the independently rotating wheelset) may be very
helpful in the development of compact actuators for this
particular application.

7. Conclusions
This paper has provided a comprehensive assessment
of mechatronic vehicle congurations and active control

Table 1
Performance on random track (ride quality, i.e. body acceleration, is normalised by passive vehicle)
Passive

Ride at front
Ride at centre
Ride at rear
Overall ride
Torque/force on curve (max)
Total torque/force (rms)
Actuator vel (rms)
Average power (W)

1
1
1
1
N/A
N/A
N/A
N/A

Solid axle
Yaw control

Lateral control

0.787
0.712
0.710
0.743
120 Nm
12.5 kN m
35 mrad/s
438 W

2.60
2.29
2.73
2.58
800 N
18.8 kN
140 mm/s
2632 W

IRW

DSW

0.589
0.685
0.651
0.636
40 Nm
1.61 kN m
24 mrad/s
38 W

0.485
0.551
0.529
0.516
21 N
16 kN
4.0 mm/s
64 W

1028

T.X. Mei et al. / Control Engineering Practice 10 (2002) 10231028

strategies for the active steering of railway wheelsets


with and without axles. The study has shown that the
active control schemes proposed for railway vehicles
offer a solution to the difcult design trade-off between
the stability and curving performance of a conventional
passive vehicle. The vehicle curving performance is
signicantly improved when the control is applied in
either yaw or lateral direction. With active controls, it is
possible to achieve the pure rolling action for the solid
axle wheelset on curves, or to provide the missing
guidance action (track following) for the independently
rotating wheels. When the control is applied in the yaw
direction, the ride quality on the vehicle body is also
signicantly improved. However if the control is applied
in the lateral direction, the ride quality is seriously
worsened. The results have also shown that the best
performance is achieved by using independently rotating
wheels (either with an axle or in the DSW conguration), rather than solid axles.
Actuation issues have also been studied. It has been
shown that the actuator power level is mainly decided by
the requirement of the active steering on random track
and the vehicle speed. A key feature of the requirement
is that, although the average power requirement is low,
the torque/force demand is relatively high compared
with the actuator velocity. This is an important factor
that should be taken into account in the development of
the actuators. The response speed and dynamic behaviour of the actuators is also a crucial issue in the
application of the active steering schemes and special
attention should be given to the effect of the actuator
velocity on the overall system performance and stability.
The important overall conclusion is that mechatronic
solutions enable mechanically simple vehicle congurations to achieve the kind of performance which would
not be possible with a purely mechanic approach.

Acknowledgements
Authors wish to acknowledge the support of the
European Community for funding the project BE97
4387, which made this study possible.

References
Aknin, P., Ayasse, J. B., & Devallez, A. (1991). Active steering of
railway wheelsets, 12th IAVSD conference. France: Lyon.
Anon (1997). Dosaged torque, BahnTech (report of Deutsche Bahn
Ag), 3/97 (pp. 1011).
Dukkipati, R. V., Narayana Swamy, S., & Osman, M. O. M.
(1992). Independently rotating wheel systems for railway vehiclesF
a state of the art review. Journal of Vehicle System Dynamics, 21,
297330.
Goodall, R. M. (1997). Active railway suspensions: Implementation
status and technological trends. Vehicle System Dynamics, 28(23),
87117.
Gretzschel, M., & Bose, L. (2000). A new concept for integrated
guidance and drive of railway running gears. 1st IFAC conference
on mechatronic systems, Darmstadt, Germany, September 2000.
Mei, T. X., & Goodall, R. M. (1999). Wheelset Control strategies for a
2-axle railway vehicle, Proceedings of 16th IAVSD symposium:
Dynamics of vehicles on roads and tracks, Pretoria, South Africa,
AugustSeptember 1999.
Mei, T. X., & Goodall, R. M. (2000a). Modal control for active
steering of railway vehicles with solid axle wheelsets. Vehicle
System Dynamics, 34/1, 2541.
Mei, T. X., & Goodall, R. M. (2000b). LQG solution for active
steering of solid axle railway vehicles. Journal of IEE Proceedings
Control Theory and Applications, 147, 111117.
Mei, T. X., & Goodall, R. M. (2001). Robust control for
independently rotating wheelsets on a railway vehicle using
practical sensors. IEEE Transactions on Control Systems Technology, 9(4), 599607.
Mei, T. X., Goodall, R. M., & Li, H. (1999). Kalman lter for the state
estimation of a 2-axle railway vehicle. Proceedings of 5th European
control conference 99, Karlsruhe, Germany, CA-10-F812.
Mei, T. X., Perez, J., & Goodall, R. M. (2000). Design of optimal PI
controls for the perfect curving of railway vehicle with solid axle
wheelsets. UK: UKACC Control 2000.
Perez, J., Busturia, J. M., & Goodall, R. M. (2000). Control strategies
for active steering of bogie-based railway vehicles, 1st IFAC
conference on mechatronic systems, Darmstadt, Germany.
Wickens, A. H. (1993). Dynamic stability of articulated and steered
railway vehicles guided by lateral displacement feedback. 13th
IAVSD symposium, Chengdu, China.
Wickens, A. H. (1998). The dynamics of railway vehiclesFfrom
Stephenson to Carter. Proceedings of the Institution of Mechanical
Engineers, Part-F, 212, 209226.

Вам также может понравиться