1.1592191

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

T. X.

Mei Practical Strategies for


School of Electronic and Electrical Engineering,
University of Leeds, Leeds, LS2 9JT, UK
Phone: 44 (0)113 343 2066;
Controlling Railway Wheelsets
Fax: 44 (0)113 343 2032
e-mail: t.x.mei@ee.leeds.ac.uk Independently Rotating Wheels
R. M. Goodall This paper presents the development of an active control strategy for railway vehicles
Department of Electronic and Electrical with independently rotating wheels. The proposed control scheme is intuitively formulated
Engineering, with a simple control structure and adaptive to vehicle speed. It does not require basic
Loughborough University, Loughborough, guidance measurements (e.g., wheel-rail deflection and angle of attack) that are expensive
LEICS. LE11 3TU, UK and impractical to implement. Speed sensors are used to measure the relative rotational
Phone: 44 (0)1509 227009; speed of the two wheels on a same axle and sensors are also used to measure the relative
Fax: 44 (0)1509 227008 yaw velocity of the wheelset and the body it is connected. Both curving performance and
e-mail: R.M.Goodall@lboro.ac.uk passenger ride comfort of the actively controlled vehicle are compared with that of a
typical passive vehicle and an optimal control scheme. 关DOI: 10.1115/1.1592191兴

1 Introduction can be developed to steer the wheelset to follow the pure rolling
line on curves 关3兴. However, the actuation requirement for con-
A conventional railway wheelset is comprised of two coned/
trolling the solid axle wheelset can be quite high in general 关1兴.
profiled wheels rigidly connected to a common axle, hence the
An alternative way forward is to move away from the conven-
term solid axle wheelset. The solid axle wheelset has been suc-
tional solid axle approach and allow the two wheels on the same
cessfully used for many years, principally because it has the abil-
axle to rotate independently from each other, hence the term in-
ity of natural 共self兲 curving and centring. When travelling on a
dependently rotating wheelset or ‘‘IRW’’ in short 关1,4,5兴. By re-
curve, the wheelset moves outwards in the lateral direction so that
moving the constraint on the rotational motion of the two wheels,
the outer wheel rotates at a larger radius at the contact point with the longitudinal creep force of wheels is significantly reduced 共al-
the rail than the inner wheel to accommodate the different travel most eliminated兲 and therefore there is no longer the need for the
distance of the two wheels, resulting in a pure rolling action of pure rolling action of the wheelset. Also the much reduced creep
both wheels. However an unconstrained solid axle wheelset is force means that the actuation requirement 共the control effort兲 for
unstable at nonzero speeds and exhibits a sustained oscillation in active control schemes will be much lower. However, there are
the lateral plane which is only constrained by the wheel flanges, a disadvantages of the new wheelset configuration. One of the main
phenomenon known as ‘‘wheelset hunting.’’ This is overcome on drawbacks is that the independently rotating wheelset does not
conventional railway vehicles using springs connected from the have the natural curving ability of the conventional wheelset, and
wheelset to the bogie or the body of the vehicle. The stiffness of hence some form of guidance control must be provided.
the springs is selected such that the minimum speed where the The self-curving and centring effect can be slightly restored by
wheelset becomes unstable 共termed ‘‘critical speed’’兲 is above the gravitational force if a ‘‘worn’’ type of wheel profile with a spe-
maximum vehicle design speed. However, this added stiffness de- cially designed tread is used 关6兴. This restoring effect is, however,
grades the ability of the wheelset to curve and it may cause severe much less effective than the rolling radius effect of a conventional
wear of the wheels and rails. This is because the two wheels of the wheelset and is not large enough to affect behavior on curves 关7兴.
wheelset are forced away from the pure rolling lines and conse- In addition, instability is still observed with independently rotat-
quently contact forces are developed at the wheel-rail interface ing wheelset and additional effect is required to avoid potential
which are caused by so-called ‘‘creepages’’ between the wheel oscillations in practice 关7, 8兴. Some improved performance can be
and rail surfaces, small relative velocities which are a result of achieved by using the coupling between the two wheels 共but much
elastic deformation of the steel at the point of contact and which softer than the elastically constrained solid-axle wheelset兲, e.g., by
apply in both longitudinal and lateral directions. In some cases means of a magnet powder coupling 关9兴. An unsymmetric struc-
such as low speed sharp curves, the hard contact between the rail ture where a mixture of solid-axle and independently rotating
head and wheel flanges 共a part of wheel designed for safety, which wheelsets has also been proposed to improve the dynamic stability
ideally should not be in contact with the rail兲 will occur, which not and the ability to steer around curves 关10兴.
only causes maximum wear and noise, but also increases the risk In the design of an active steering scheme for the independently
of derailment in extreme conditions. It has always been a very rotating wheelset, it is obvious that the relative displacement be-
difficult design trade-off between the wheelset stability and the tween the wheelset and the track 共the wheel-rail deflection兲 is the
curving performance. natural choice of feedback required to provide the necessary guid-
While a compromise between different design objectives for a ance action for the wheelset to follow the track. Other measure-
railway vehicle has to be made if only passive solutions are em- ments may also be required for the stabilization and performance
ployed, studies have shown that it is possible to solve this design optimization 共e.g., angle of attack兲. However, a direct measure-
conflict by the use of active control concepts to steer the solid axle ment of the wheel-rail deflection is not readily available at present
wheelset 关1,2兴. It has been shown that active steering schemes can and will be very expensive to achieve in practice. Although state
provide necessary stabilization control without affecting the natu- estimation techniques such as Kalman filters can be used to esti-
ral curving action of the wheelset. Alternatively active controls mate the signals from other practical sensors, the control structure
can be much more complex with high orders and it can be very
Contributed by the Dynamic Systems, Measurement, and Control Division of THE difficult to make observers to work effectively in the presence of
AMERICAN SOCIETY OF MECHANICAL ENGINEERS for publication in the ASME
JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript
large parameter variations 关11兴.
received by the ASME Dynamic Systems and Control Division August 29, 2002; This paper presents the development of an intuitively formu-
final revision, March 26, 2003. Associate Editor: A. Alleyne. lated controller for the active steering of independently rotating

354 Õ Vol. 125, SEPTEMBER 2003 Copyright © 2003 by ASME Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


wheelsets. The resulting controller is simple to implement and The behavior of a railway wheelset is dominated by the contact
adaptive to the vehicle speed. Feedback is the measurement of the mechanics at the wheel-rail interface. The creep laws are now well
relative rotation speed of the two wheels on same axles, and the understood and a linearized version has been used in many studies
relative yaw velocity between the wheelset and vehicle body. The 关8,13兴. The longitudinal creep force affects the yaw motion of the
latter can be replaced with the use of passive dampers in the wheelset and in the case of independently rotating wheels the
longitudinal/yaw direction. relative rotation of the two wheels, and the lateral creep force
affects the lateral motion. Only those three motions are included
2 Vehicle Configuration in this study, as other motions are not relevant to the study of
active steering. Similarly, only the lateral and yaw motions are
A two-axle vehicle derived from a study of ‘‘mechatronic trains considered for the vehicle body. The vehicle model can be repre-
for the future’’ 关12兴 is used in the study and Fig. 1 gives a sim- sented by Eqs. 共1兲–共8兲, which is a linearized model with eight
plified plan view diagram. The parameters for this vehicle scheme degrees of freedom. The use of linearized models is justified on
are based upon the requirements for future high speed passenger the basis that an active steering scheme will improve performance
vehicles having a design speed of 300 km/h. The vehicle scheme on curves in a manner which considerably reduces the effects of
consists of a vehicle body and two wheelsets. The wheelsets are nonlinearities. The nonlinearities of a railway vehicle model are
connected to the body via springs and dampers in the lateral di- largely associated with nonlinear wheel-rail profiles and contact
rection. Note that no longitudinal springs or dampers are shown, forces, which become particularly problematic when the wheel-
although in practice some form of longitudinal connection is rail contact point approaches the wheel flanges. However, effec-
needed to transmit traction and braking forces from the wheels to tive use of active steering control will overcome this problem by
the vehicle body. However, it is common practice to omit these steering the wheelset to operate at the linear region of the wheel
when studying bogie dynamics, and in the particular case of an tread and rail surface,
active scheme it also avoids problems of introducing uncontrol-
lable degrees-of-freedom into the design model. There will also be
a vertical suspension, but this and the associated roll suspension
are not considered in the plan view design model because the
m w ÿ w1 ⫹ 冉 2 f 22
Vs 冊
⫹C s ẏ w1 ⫹K s y w1 ⫺2 f 22␺ w1 ⫺C s ẏ v ⫺K s y v

study deals with the active steering of the wheelsets for which
only the plan view dynamics of the vehicle are of prime interest.
On each wheelset, an actuator is also placed between the
⫺C s L v ␺˙ v ⫺K s L v ␺ v ⫽m w 冉 V s2
R1

⫺g ␪ c1 , (1)

wheelset and the vehicle body in the yaw direction for the imple- 2 f 11L 2g 2 f 11␭L g 2r 0 f 11L g
mentation of active control. The actuator placement is important I w ␺¨ w1 ⫹ ␺˙ w1 ⫹ y w1 ⫹ ␾˙ w1
for providing effective control of the wheelsets. A comparison of Vs r0 Vs
different configurations has shown that, whilst it is possible to 2 f 11L 2g 2 f 11␭L g
develop active steering schemes with actuators either in the yaw ⫽ ⫹ y t1 ⫹T w1 , (2)
or in the lateral directions, the latter arrangement worsens the ride R1 r0
quality 关1兴. This is because the wheelset-controlling force directly r 20 f 11 r 0 f 11L g
applied to the vehicle body in the lateral direction has an adverse ¨ ⫹
I w1 ␾ ␾˙ w1 ⫹ f 11␭•y w1 ⫹ ␺˙ w1
w1
effect on the body modes, whereas yaw actuators have a much Vs Vs
smaller effect. r 0 f 11L g
There is a variety of ways in which the torsional actuator can be ⫽ ⫹ f 11␭•y t1 , (3)
implemented in practice. One approach is to use a pair of longi- R1

冉 冊
tudinal actuators, controlled in such a way that action is only
2 f 22
applied in a differential sense. An alternative is to use a rotational m w ÿ w2 ⫹ ⫹C s ẏ w2 ⫹K s y w2 ⫺2 f 22␺ w2 ⫺C s ẏ v ⫺K s y v
actuator with a crank and linkage arrangement, but the intention in Vs

冉 冊
this paper is to describe a practical sensing and control strategy
V s2
that could be applied irrespective of the chosen actuator technol- ⫹C s L v ␺˙ v ⫹K s L v ␺ v ⫽m w ⫺g ␪ c2 , (4)
ogy, and so this level of detail has not been considered. For a R2
particular implementation it would be necessary to check and if
necessary refine the controller using a full complexity simulation 2 f 11L 2g 2 f 11␭L g 2r 0 f 11L g
I w ␺¨ w2 ⫹ ␺˙ w2 ⫹ y w2 ⫹ ␾˙ w2
model 共e.g., using an MBS package兲. Vs r0 Vs
2 f 11L 2g 2 f 11␭L g
⫽ ⫹ y t2 ⫹T w2 , (5)
R2 r0
r 20 f 11 r 0 f 11L g
¨ ⫹
I w1 ␾ ␾˙ w2 ⫹ f 11␭•y w2 ⫹ ␺˙ w2
w2
Vs Vs
r 0 f 11L g
⫽ ⫹ f 11␭•y t2 , (6)
R2

m v ÿ v ⫹2C s ẏ v ⫹2K s y v ⫺C s ẏ w1 ⫺K s y w1 ⫺C s ẏ w2 ⫺K s y w2


m v V s2
2 冉 1

1
R1 R2

m vg
2 冊
共 ␪ c1 ⫹ ␪ c2 兲 , (7)

I v ␺¨ v ⫹2L 2v C s ␺˙ v ⫹2L 2v C s ␺ v ⫺L v C s ẏ w1 ⫹L v C s ẏ w2 ⫺L v K s y w1
⫹L v K s y w2 ⫽⫺ 共 T w1 ⫹T w2 兲 . (8)
Equations 共1兲–共3兲 represent the lateral motion, the yaw motion
and the relative rotation between the two wheels of the leading
Fig. 1 Simplified plan view of a two-axle vehicle wheelset respectively. Equations 共4兲–共6兲 represent those for the

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 2003, Vol. 125 Õ 355

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


trailing wheelset. Equations 共7兲 and 共8兲 are for the lateral and yaw interactions between lateral, yaw, and rotation motions within
motions of the vehicle body. The longitudinal dynamics 共and the each wheelset. Also there are interactions between the two
absolute forward rotation of the wheels兲 are not modeled. Those wheelsets via the body motions. Second, the deterministic and
motions are not particularly relevant in the study of railway ve- random track inputs affect the wheelset behavior differently and it
hicle dynamics because of the large time constant of the forward is important to accommodate the two track features effectively in
motion. The vehicle speed is therefore assumed as constant, and the control design. Third, the study is aimed at avoiding the use of
for the study of different speeds it is considered adequate to just
impractical/expensive measurement or the requirement of a com-
change the value of V s in the model. The reference directions 共in
the plan view diagram as shown in Fig. 1兲 are from left to right for plex state estimator, and therefore minimizing the cost and com-
the lateral movements and clockwise for the yaw movements. The plexity of the active control scheme. Finally the controller should
positive reference for the relative rotation between the two wheels be adaptive to the vehicle travelling speed, as the wheelset dy-
of each wheelset is in the direction where it makes the wheelset to namics are speed dependent.
yaw in the positive direction. There are two different types of To reduce the order of the system model and complexity of the
railway track input. The curve radius (R 1 , R 2 ) and cant angles control design, a modal control approach can be employed to
( ␪ c1 , ␪ c2 ) are the deterministic features corresponding to the in- de-couple the model into two independent subsystems. This tech-
tended track alignment which are designed to satisfy passenger nique has been proved very effective for a vehicle with solid-axle
comfort requirements, whereas the track random inputs (y t1 , y t2 ) wheelsets 关14兴, but it can be readily adapted for vehicles with
are the unintended track feature which is the deviation from the IRW’s. By adding and subtracting corresponding motion equa-
intended alignment. tions for the two wheelsets, i.e., yaw, lateral, and rotational motion
equations of one wheelset are added to or subtracted from corre-
3 Control Designs sponding ones of the other wheelset, it can be shown that one of
the resulting subsystems may be regarded as a IRW wheelset at-
A detailed analysis of the independently rotating wheelset re-
tached to a mass of the body lateral motion as shown in Fig. 2
veals that the kinematic instability of the solid axle wheelset is not
removed by allowing the two wheels to rotate freely 关8兴. There- 共termed lateral mode兲 and the other as a IRW attached to a mass of
fore some form of stabilization effort should be provided in addi- the body yaw motion as given in Fig. 3 共termed yaw mode兲. The
tion to the guidance action necessary to control the wheelsets to output measurements from the two wheelsets are decomposed to
follow the track. give feedback signals required by the lateral and yaw controllers,
A number of factors affect the controller design. First, the sys- respectively, and the output signals from the two controllers are
tem model is highly interactive with high order 共14th order:- 5th then recombined to control the two wheelsets accordingly. Vari-
order for each wheelset and 4th order for the body兲. There are ables used in the figures are defined as

Fig. 2 Lateral mode „mode A …

Fig. 3 Yaw mode „mode B …

356 Õ Vol. 125, SEPTEMBER 2003 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


Table 1 Eigenvalues of lateral and yaw modes back 共for which passive yaw or longitudinal stiffness may be
„ Vs Ä83.3 mÕs… used兲. Stabilization with yaw velocity or lateral acceleration feed-
back can provide better damping, although there is a constraint on
Lateral mode Yaw mode
damping freq. 共Hz兲 damping freq. 共Hz兲 its bandwidth.
This paper studies the use of yaw velocity feedback, as it offers
⫺0.11 6.13 ⫺0.11 6.13 the flexibility for passive or active implementation. For the pas-
⫺0.11 6.13 ⫺0.11 6.13
0 0 sive stabilization, a damper is placed between each wheelset and
1.0 37.3 1.0 37.3 vehicle body which will provide an extra safety measure in the
1.0 64.7 1.0 64.7 case of actuator failure. However, the active stabilization is pre-
0.21 0.93 0.22 0.97 ferred because actuators will be required to implement the guid-
0.21 0.93 0.22 0.97
ance control anyway and also an adaptive scheme may then be
developed to give adequate damping at different vehicle speeds.
The stabilization effort for the leading and trailing wheelsets can
be expressed in Eqs. 共9兲 and 共10兲,
y wa ⫽y w1 ⫹y w2 , y wb ⫽y w1 ⫺y w2 ,
␺ wa ⫽ ␺ w1 ⫹ ␺ w2 , ␺ wb ⫽ ␺ w1 ⫺ ␺ w2 , T w1 s ⫽C w • 共 ␺˙ w1 ⫺ ␺˙ v 兲 , (9)

␾ wa ⫽ ␾ w1 ⫹ ␾ w2 , ␾ wb ⫽ ␾ w1 ⫺ ␾ w2 , T w2 s ⫽C w • 共 ␺˙ w2 ⫺ ␺˙ v 兲 . (10)

y ta ⫽y t1 ⫹y t2 , y tb ⫽y t1 ⫺y t2 ,
The control gain is selected using the frequency design method
1/R ta ⫽1/R 1 ⫹1/R 2 , 1/R tb ⫽1/R 1 ⫺1/R 2 . with the assistance of Figs. 2 and 3. In this case, the decoupled
lateral and yaw modes are very similar to each other as indicated
All variables are the sums or differences between the correspond-
in Table 1 and therefore the same control gains are selected for the
ing variables of the two wheelsets or the track geometries at the
two modes resulting in the local control for each wheelset as
two wheelsets, in particular ÿ ra ÿ rb represent the lateral centrifu-
shown in Eqs. 共9兲 and 共10兲. Also the body yaw rate, which has
gal forces due to the track curvature and cant angle, and ␪¨ ra ␪¨ rb little effect on the stability, is added in the controllers so that the
are angular accelerations due to curve transitions of the track. relative yaw velocity between the body and each wheelset can be
Table 1 shows eigenvalues of the two modes for the passive used which can be provided by measuring the actuator movement
共uncontrolled兲 vehicle. The eigenvalues in the first two lines rep- directly.
resent the unstable kinematic modes of the two wheelsets and the
eigenvalue at the origin (s⫽0 in third line兲 indicates the missing 3.2 Guidance Control. An independently rotating wheelset
guidance action caused by the two wheels being able to rotate does not have the ability of self-curving/centring, i.e., it will not
freely. The fourth and fifth lines give the eigenvalues of the be able to follow the track by itself without making flange contact.
wheelset high frequency mode, whereas the lower frequency This problem is caused by the two wheels being able to rotate
modes for the lateral and yaw motions of the vehicle body are freely which, from the control point of view, is indicated by the
shown in the last two lines. Clearly it is necessary both to stabilize zero eigenvalue for each wheelset as given in Table 1.
the wheelset kinematic mode and to provide the track following An obvious way to restore the guidance action of the wheelset
control. is to control the wheel-rail deflections directly. This will be of
course very expensive to achieve in practice 共because the feed-
3.1 Wheelset Stabilization. It can be readily demonstrated back is not available兲 and an alternative must be found. This paper
by applying standard frequency analysis methods to Figs. 2 and 3 proposes a control method derived from the fundamental feature
that the wheelset stabilization can be achieved by using one or a of the solid axle wheelset that enables its self-curving action, i.e.,
combination of three wheelset variables as feedback: the yaw two wheels rotating at the same speed. Although there is no
angle, the yaw velocity and the lateral acceleration. Figure 4 ‘‘hard’’ connection between the two wheels on an IRW axle, a
shows the root locus diagram of the lateral mode with the different control action can be formulated such that the actuator will steer
feedback measurements, and similar results can be obtained for the axle to achieve the zero-speed difference. This approach adds
the yaw mode. As shown in the diagram, the achievable damping a damping effect between the two wheels via the active means,
ratio of the kinematic mode is limited with the yaw angle feed- however, it does not result in the stiff connection of the solid axle
wheelset which forces the two wheels to be at the same angular
position 共rather than velocity兲 at all times. Equations 共11兲 and 共12兲
show the guidance actions for the front and rear wheelsets on the
vehicle, respectively. The mathematical explanation is that this
guidance action effectively moves the eigenvalue at the origin to
somewhere in the negative half plane,

T w1 g ⫽K w • ␾˙ w1 , (11)

T w2 g ⫽K w • ␾˙ w2 . (12)

The important practical implication is that the relative wheel


speed will already be available because wheel speeds are mea-
sured for controlling wheel slide during braking.
3.3 Speed Adaptation. Wheelset dynamics are speed de-
pendent and frequencies of both the kinematic and high frequency
modes of a wheelset vary with the wheelset travelling speed, but
the variations are nonlinear functions of the speed. In general a
vehicle stabilized for high-speed operations will remain stable at
lower speed, but the guidance on both curved and random tracks
Fig. 4 Root locus diagram of the lateral mode tends to be poor resulting in unacceptably large wheel-rail lateral

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 2003, Vol. 125 Õ 357

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


Table 2 Control gains of the full state feedback optimal controller

For wheelset 1 For wheelset 2


Lateral velocity of wheelset 1 ⫺2.5282⫻10 4
1.2395⫻101
Lateral displacement of wheelset 1 ⫺4.7478⫻106 4.6510⫻102
Yaw velocity of wheelset 1 ⫺8.2903⫻104 6.6830
Yaw angle of wheelset 1 ⫺1.2671⫻107 1.8065⫻103
Relative rotational velocity of two wheels of 2.4497⫻104 ⫺1.4791
wheelset 1
Lateral velocity of wheelset 2 1.2395⫻10 ⫺2.5282⫻104
Lateral displacement of wheelset 2 4.6510⫻102 ⫺4.7478⫻106
Yaw velocity of wheelset 2 6.6830 ⫺8.2903⫻104
Yaw angle of wheelset 2 1.8065⫻103 ⫺1.2671⫻107
Relative rotational velocity of two wheels of ⫺1.4791 2.4497⫻104
wheelset 2
Lateral velocity of body frame ⫺2.8207⫻104 ⫺2.8207⫻104
Lateral displacement of body frame ⫺3.0760⫻105 ⫺3.0760⫻105
Yaw velocity of body frame ⫺1.2614⫻105 1.2614⫻105
Yaw angle of body frame ⫺1.3789⫻106 1.3789⫻106

movements. Some form of gain scheduling or adaptation will be solid axle wheelset, and the primary objective here is to produce a
necessary if the vehicle with active controls is to be used across a good guidance control such that the wheelsets follow the track and
wide range of speed. The adaptation of speed proposed in this the flange contact is avoided.
study is obtained based on engineering insight of the problem. The The advantage of the active controls is better illustrated at
stabilization action is set to be proportional to the square of the lower speed. Figure 6 compares the curving performances of the
vehicle speed, a measurement that is readily available on all different vehicles at the speed of 25 m/s. The large wheel-rail
trains, and a constant value is used to compensate the low damp- displacement 共20 mm兲 of the passive system in this case would in
ing for very low speed operations. The steering control is made to practice cause flange contact 共the wheel design is such that the
be proportional to the vehicle speed. Equations 共13兲 and 共14兲 give maximum available movement is typically around 10 mm兲, but
the final control laws for the front and rear wheelsets, respectively. this effect is not represented by the linearized model used in this
These now include both stabilization and guidance functions, study. With the proposed active control scheme, the wheelset lat-
which offers a very simple solution for the difficult problems of a eral displacement is only about 5 mm, therefore avoiding flange
quite complex dynamic system, contact. On transitions, the maximum movement observed with
the proposed controller is 6 –7 mm which is starting to enter the
T w1 ⫽ 共 C w0 ⫹C w •V s2 兲 • 共 ␺˙ w1 ⫺ ␺˙ v 兲 ⫹K w •V s • ␾
˙ ,
w1 (13) nonlinear region of the wheel-rail contact, but brief enough not to
be a major concern. It is possible to eliminate the problem com-
T w2 ⫽ 共 C w0 ⫹C w •V s2 兲 • 共 ␺˙ w2 ⫺ ␺˙ v 兲 ⫹K w •V s • ␾
˙ .
w2 (14) pletely by modifying wheel rail profiles to effectively increase the
conicity for low speed and tight curves.
4 Simulation Results On straight track with random input, the active control also
To study the responses of the actively controlled vehicle on demonstrates excellent tracking performance, even though the
different tracks, both deterministic and random track inputs are wheel-rail deflections are not part of the feedback. Figure 7 shows
used in the simulation. The deterministic track inputs for the speed the lateral displacement of the front wheelset relative to the track
of 83.3 and 25 m/s are typical high- and low-speed railway curves at the maximum vehicle speed of 83.3 m/s and similar results are
with constant radius of 3500 and 300 m, respectively. The curves obtained for other speeds. The movements are such that the maxi-
are connected to the straight track via transitions, which last mum deflection is only very occasionally reached—refinement of
around 1 s 共80 and 24 m in the two cases兲 with a constant rate of the control law could reduce this if necessary.
change of the radius. Note also that the track is canted during the One of the other design parameters of a railway vehicle is the
curve to reduce the lateral acceleration experienced by the passen- passenger ride comfort on the vehicle body and it is very impor-
gers. The random track input is derived from a filtered white noise tant that the active control scheme also improves the ride quality
generator to give an appropriate power spectrum for the lateral 共at least it should not be deteriorated兲. Table 3 compares the ride
deviations.
For comparison purposes, results from two other vehicles have
been used. One is a passive vehicle with solid axle wheelsets
stabilised by mechanical yaw stiffness, and the other is a vehicle
with independently rotating wheelsets actively controlled by a full
state feedback optimal control. Table 2 gives the control gains.
Figure 5 compares the wheelset lateral displacements of the
three schemes on a curved track, where the vehicle speed is 83.3
m/s. The wheelsets of the passive vehicle are forced away from
their pure rolling positions by the stabilizing yaw stiffness in the
opposite directions. This will result in increased creepage between
the wheels and the track and hence undesirable wear. When the
independently rotating wheelsets are used, both the proposed ac-
tive controller and the full state optimal control achieve a good
control of the wheelsets. A pure rolling action is achieved with the
proposed control, while a more orderly control especially on curve
transitions is obtained with the optimal control, which is expected.
It should be noted that achieving close to pure rolling action for Fig. 5 Wheelset lateral displacement „ Vs Ä83.3 mÕs, radius
the independently rotating wheelset is not as crucial as for the Ä3500 m…

358 Õ Vol. 125, SEPTEMBER 2003 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


Table 3 Vehicle performance „rms values, Vs Ä83.3 mÕs… use only the basic and simple measurements and it provides a
practical and inexpensive solution for the difficult problems of a
Passive Optimal 共IRW兲 Proposed 共IRW兲
complex dynamic system.
Front 5.30% g 3.12% g 4.16% g It has been shown that the proposed controller is adaptive to the
Center 4.10% g 2.81% g 3.48% g vehicle speed and works well on both curved and random tracks.
Rear 5.06% g 3.32% g 4.41% g Excellent curving performance is achieved and there will be no
Overall 4.86% g 3.09% g 4.04% g
Act. torque N/A 1.6 kNm 1.1 kNm flange contact even on very low speed curves. The control scheme
Act. velocity N/A 24 m rad/s 33 m rad/s also improves the passenger ride comfort when compared with a
typical passive vehicle on the random track input.
Although an optimal control might outperform the proposed
scheme, the simplicity of the control structure and the simple
requirement of practical feedback measurement will be the posi-
quality 共the rms lateral acceleration兲 of the three vehicles and the tive arguments when considered for real applications.
control effort requirement of the active schemes. Clearly the pro-
posed control method improves the ride quality by 13–20% with
an overall improvement of 17% when compared with the passive Acknowledgment
vehicle, but not as much as the optimal control scheme. It seems The authors wish to acknowledge the support of the European
that the optimal control performs better on both curved and ran- Community for funding the Project No. BE97-4387, which made
dom tracks, but it requires a full state feedback with a state ob- this study possible.
server and a much more complex control structure.
The control torque required for the proposed control scheme is
lower than for optimal control, but the power requirements are
Appendix: Vehicle variables and parameters
fairly similar as the corresponding actuator velocity is higher. As
shown in the table, the average power requirement of each actua- Variables Definitions and values
tor is small, i.e., in the region of 30 W for both control schemes. Cs Lateral damping per wheelset 共37 kN s/m兲
C w0 , C w Stabilizing damping per wheelset 共1000, 5.49兲
f 11 , f 22 Longitudinal and lateral creep coefficients
5 Conclusions 共nominal 10 MN兲
g Gravity (9.8 m/s2 )
This paper has presented the development of a novel control I v, I w Vehicle yaw inertia (558,800 kg m2 ) and
scheme for the active steering of the independently rotating rail-
wheelset yaw inertia (700 kg m2 )
way wheelset. The controller has been intuitively formulated to I w1 Wheel inertia 100 (kg m2 )
Ks Lateral stiffness per wheelset 共511 kN/m兲
Kw Coefficient of active guidance control 共24兲
Lg Half gauge of wheelset 共0.7 m兲
Lv Half wheelset spacing of vehicle 共4.5 m兲
m v, m w Vehicle mass 共30,000 kg兲 and wheelset mass
共1250 kg兲
r0 Wheel radius 共0.45 m兲
R1 R2 Radius of the curved track at leading and trailing
wheelsets 共3500 m兲
T w1 , T w2 Control torque for leading and trailing
wheelsets, respectively
Vs Vehicle travel speed 共83.3 m/s兲
y t1 , y t2 Track lateral irregularity at leading and trailing
wheelsets
y w1 , y w2 , y v Lateral displacement of leading, trailing wheelsets
and body
␪ c1 , ␪ c2 Cant angle of the curved track at the leading
and trailing wheelsets (6 0 )
␭ Wheel conicity 共nominal value 0.2兲
Fig. 6 Wheelset lateral displacement „ Vs Ä25 mÕs, radius ␾ w1 , ␾ w2 Relative rotation angle of two wheels at leading
Ä300 m… and trailing wheelset
␺ w1 , ␺ w2 , ␺ v Yaw angles of leading, trailing wheelset
and vehicle body

References
关1兴 Mei, T. X., and Goodall, R. M., 1999, ‘‘Wheelset Control Strategies for a
2-Axle Railway Vehicle,’’ 16th IAVSD Symposium: Dynamics of Vehicles on
Roads and Tracks, Pretoria, South Africa.
关2兴 Aknin, P., Ayasse, J. B., and Devallez, A., 1991, ‘‘Active steering of railway
wheelsets,’’ 12th IAVSD Conference, Lyon, France.
关3兴 Mei, T. X., Perez, J., and Goodall, R. M., 2000, ‘‘Design of optimal controls
for perfect curving of solid axle wheelset,’’ UKACC International Conference:
Control 2000, Cambridge, UK.
关4兴 Gretzschel, M., and Bose, L., 1999, ‘‘A Mechatronic approach for active in-
fluence on railway vehicle running behavior,’’ 16th IAVSD Symposium Dynam-
ics of Vehicles on Roads and Tracks, Pretoria, S. Africa.
关5兴 Powell, A., Goodall, R. M., Walker, C., Nuet, D., and Noel, Y., 1998, ‘‘Com-
parison of the mechanical steering system used on the MF88 trains of the Paris
Metro with an active guidance system,’’ International Congress Railtech ’98
‘‘Technology for Business Needs,’’ Birmingham, UK, pp. 85–95.
关6兴 Satou, E., and Miyamoto, M., 1991, ‘‘Dynamics of a bogie with independently
Fig. 7 Wheel-rail deflection on random input „ Vs Ä83.3 mÕs… rotating wheels,’’ 12th IAVSD Symposium, France, 1991, pp. 519–534.

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 2003, Vol. 125 Õ 359

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab


关7兴 Eickhoff, B. M., 1991, ‘‘The application of independently rotating wheels to 关11兴 Mei, T. X., Goodall, R. M., and Li, H., 1999, ‘‘Kalman Filter for the State
railway vehicles,’’ IMechE Proc. (Part F), 205, pp. 43–54. Estimation of a 2-Axle Railway Vehicle,’’ Proceedings of 5th European Con-
关8兴 Goodall, R. M., and Li, H., 2000, ‘‘Solid axle and independently-rotating trol Conference 99, Karlsruhe, Germany, CA-10-F812.
railway wheelsets—A control engineering assessment,’’ Veh. Syst. Dyn., 33, 关12兴 Ellis, B., and Goodall, R. M., 1999, ‘‘The Mechatronic Train: Requirements
pp. 57– 67. and Concepts,’’ Proceedings of World Congress on Railway Research, Tokyo,
关9兴 Geuenich, W., Guenther, C., and Leo, R., 1983, ‘‘Dynamics of fiber composite
Japan, pp. 1–7.
bogies with creep controlled wheelsets,’’ Proc. 8th IAVSD Symposium of Dy-
关13兴 Wickens, A. H., 1998, ‘‘The dynamics of railway vehicles—From Stephenson
namics of Vehicles on Roads and Tracks, Cambridge, MA, USA, pp. 225–238.
关10兴 Suda, Y., 1990, ‘‘Improvement of high speed stability and curving performance to Carter,’’ IMechE Proc. (Part F), 212, pp. 209–217.
by parameter control of trucks for rail vehicles considering independently ro- 关14兴 Mei, T. X., and Goodall, R. M., 2000, ‘‘Modal control for active steering
tating wheelsets and unsymmetric structure,’’ JSME Int. J., Ser. III, 33, pp. of railway vehicles with solid axle wheelsets,’’ Veh. Syst. Dyn., 34Õ1,
176 –182. pp. 25– 41.

360 Õ Vol. 125, SEPTEMBER 2003 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jdsmaa/26320/ on 03/28/2017 Terms of Use: http://www.asme.org/ab

You might also like