Welcome to Scribd. Sign in or start your free trial to enjoy unlimited e-books, audiobooks & documents.Find out more
Download
Standard view
Full view
of .
Look up keyword
Like this
0Activity
0 of .
Results for:
No results containing your search query
P. 1
Active Steering of Railway Vehicles

Active Steering of Railway Vehicles

Ratings: (0)|Views: 0|Likes:
Published by Kerry Flowers

More info:

Published by: Kerry Flowers on Mar 17, 2014
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

03/17/2014

pdf

text

original

 
ACTIVE STEERING OF RAILWAY VEHICLES: AFEEDFORWARD STRATEGY
Shuiwen Shen
 
 , TX Mei
 
 , R.M.Goodall
 
 , J.Pearson
 
 , G.Himmelstein
 
 
University of Leeds LS2 9JT, Leeds UK. Tel:(40)113+3432064,email:s.shen@ee.leeds.ac.uk 
 
Department of E&E Eng, LoughboroughUniversity, LE11 3TU, LoughboroughUK.
 
Bombardier Transportation, Siegen, Germany.
Keywords:
 Active control, Vehicle dynamics, Steering, Sensi-tivity, Feedforward.
Abstract
This paper presents an active steering strategies for a railwayvehicle to improve performances on curves. By feeding for-ward a desired angle of attach and the radius of curves, a feed-forward control law is proposedand verified. In order to assessits performance issue, parameter variations such as creep coef-ficients and nonlinearitiesof the wheel conicity are considered.The simulation results show a significant improvement. Al-though the feedforward strategy depends on the knowledge of the yaw stiffness of the system, no interference with the stabil-ity make it preferable.
1 Introduction
A well-known conflict between stability and steering for rail-way vehicles with coned/profiled wheels has been a chal-lenge [1]. Though the kinematic instability of the solid-axlewheelsets canbe removedthrougha stiff yawconnection,it de-teriorates curving performances [2]. There have been many in-vestigationsonthepossiblemechanicalsolutionssuchascross-bracing,bodysteeringand primaryyaw damping,by which thetrade-off can be improved [3]. However the contradiction be-tween high-speed stability and effective curving still remains.Activecontrolcanprovidea solutionforthisproblem,althoughso far the practical applications are restricted to the secondarysuspension to improve ride quality. The incorporationof activecontrol offers a design flexibility which is not available in anymechanical suspensions. With such a flexibility, a number of publications illustrate the possibility of removingthe trade-off.In contrast to previous studies of active stability [4], active yawrelaxation[2] and activethe lateral positionand/orthe yaw mo-ment of the wheelset control [5], this paper investigates a feed-forward strategy, which offers a practical solution for improv-ing curvature performance without compromising the stability.Theremainderof this paperis organizedas follows. Inadditionto a description of a railway vehicle, a brief introduction of anactive strategy toward vehicle stability is presented in Section2. Focusing on a comparison between so-called radial and per-fect steering, the feedforward strategies are detailed in Section3 and assessed in Section 4. Finally, Section 5 gives the mainconclusions and some suggestions for future research.
2 Basic Principles
A plan view of a half-vehicle is depicted in Fig.1 (a), in whichtwo actuators are located between wheelsets and the bogieframe. Appendices B and C present the necessary parametersand their normal quantities, respectively. An introduction of three reference coordinates provides a convenience of model-ing the system, where the coordinate
 
 is attached to andmoving with the leading wheelset; the coordinate
 
 is rele-vant to the trailing wheelset; and the coordinate
 
 applies tothe bogieframe and body. The three coordinatesmovewith thesame velocity
 
 
 as the longitudinaldynamics are not modeled.Each coordinate has its own yaw velocity of 
 
 
 
 
 
 
 ,
 
 
 
 
 
 
 
 
and
 
 
 
 
 
 
 
 w.r.t.
 the track, but only oneof them is independentsince the frames are inter-related by theangles of 
 
 
 
 
 
 and
 
 
 
 
 
 
 
 . In addition, the equiva-lent yaw stiffness and damping of the primary suspensions areemployed to represent their effect on the yaw dynamics of thewheelset and bogie. The equivalent yaw stiffness and damping
w.r.t.
 the wheelsets are
 
 
 
 
 
 
 
 
 ,
 
 
 
 
 
 
 
 
 whereas,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
and
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 are
 w.r.t.
the bogie. The rail/wheel creep forces generally depend bothon creepages, being relative normalised velocities between therailandwheel, andthecreepcoefficients,dependingonthever-tical loads at individual wheels. Accordingly, the lateral creepforces of the wheelsets are given by
 
 
 
 
 
 
 
 
 
 
 
 
 
(1)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(2)The torques at the wheelsets due to the longitudinal creepagesare represented as follows.
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(3)
 
  
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(4)By introducing
 
 
 
 ,
 
 
 
 
 
 ,
 
 
 
 
 
 , and
 
 
 
 
 
, a state-space representation is expressed as
 
 
 
 
 
 
 
(5)where the states
 
 , inputs
 
 , and disturbances
 
 are given by,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Wheelsethalf bodyActuator Bogie frameCo.
 FL
Co.
 FG
 Coordinate
 FT 
(a) Frames of the vehicle
F
 yl 
 F
 yt 
F
c
T
T
crpcrp
 x
 
 x
(b) General principles of active steering
Figure 1: A railway vehicle
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
and the expression of matrices
 
 ,
 
 and
 
 are given in Ap-pendix A. Generally speaking, the model developed so far,being a linear representation of the system, is suitable for thedevelopmentof the control strategies, but it is not sufficient forcontrol assessment. A more complicated version [6], includingthe model of the actuator, the sensor dynamics and the nonlin-earities of the wheel conicity, will be applied for evaluating thecontrollers.Fig.1 (b) shows a very simple representation of the railwayvehicle, being treated as a single rigid body with a mass of 
 
 
 
 
 
 
 
 
 
 
. In steady-state, only the creep forces
 
and
 
 
 are necessary to counteract the lateral unbalanceload (or cant deficiency forces
 
 
 ) for the vehicle on curves.Thus, the unnecessary creep torques
 
 
 and
 
  
 can onlyresult from an inappropriate steering. It is clear that in the per-fect steering conditions, the following must hold.
 
 
 
 
 
 
 
 
 
(6)
 
 
 
 
  
 
(7)in which
 
 
 
 
 
 and
 
 
 
 
 
 
 
 
 
 , where
 
 is thecant angle of the track on the curve. Suppose that
 
 
 and
 
 
 
, the above requirements are equivalent to
 
 
 
 
 
 
and
 
 
 
 
 
 
 
 , in which
 
 
 is the desired angle of attack for both wheelsets while
 
 
 the desired lateral displacement.Referring back to Eqn.(1), (2), (3) and (4), it yields that.
 
 
 
 
 
 
 
 
 
(8)
 
 
 
 
 
 
 
 
 
(9)Obviously, if the angle of attack and lateral deflections of thewheelsets can be readily measured, it would then be straightforward to design a control for the active steering. Unfortu-nately these signals are difficult to measure, either expensivesensors are needed or complex estimators have to be used.However, if 
 
 
 
 
 
 and
 
 
 
 
 
 , the system can meetrequirements of the perfect steering. In steady-state,
 
 
 
 
implies
 
 
 
 
 , leading to
 
 
 
 
  
 
 . Likewise,
 
 
 
 
implies
 
 
 
 
  
 . Thus
 
 
 
 
  
 
 are theonly solution. Noting that
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 is equivalentto
 
 
 
 
 while
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 to
 
 
 
 
 , it requiresonly measuring the relative movements of the wheelsets andthe bogie.Althougha numberof controllersmay be developed,this paperproposes a feedforward control law to implement the strategy.The obvious advantage is that it does not interfere with the sta-bility and therefore can be developed independently.
3 Feedforward Control Strategies
Radial steering, achieved by passively or actively operating thewheelsets in line with the radius of curvature, generally cannotresult in perfect steering except for zero cant deficiency (
i.e.
 
 
 
, the centrifugalforce is exactly balanced by the gravityforce of the vehicle on the canted track). Thus a more sophis-ticated strategy is necessary. The strategy studied in this paperfocuses on the quasi steady-state performance although how toavoid the performance becoming worse during the transition isalso under consideration.The feedforward control laws proposed for the perfect steeringare,
 
 
 
  
 
 
 
 
 
 
 
 
 
(10)
 
 
 
 
  
 
 
 
 
 
 
 
 
 
(11)in which
 
 
 is the desired angle of attack, being given by (8),and
 
  
 is the yaw stiffness and
 
 
 is the radius of curvatureused in the control law. Perfect steering reduces to radial steer-ing if 
 
 
 
 . Clearly, the controllaws dependon the accuracyof the yaw stiffness
 
 
 and the curvature
 
 . Perfect steeringis only attained if the parameters are exact.To prove that above control laws could result in the perfectsteering, two new variables are introduced, which are
 
 
 
 
 
 
 
and
 
 
 
 
 
 
 
 . Then, the steady-state version of equations of the system becomes
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(12)
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
(13)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
(14)in which,
 
 
 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
By introducing
 
 
 
 
 
 
 
 
 
 
 and
 
 
 
 
 
 
 
 
 ,then
 
 
 
 
 
 
 
where
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
  
It is true in this case that the singular values of 
 
 
 all exist andpositive, thus
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
The perfect feedforward control law (10) and (11) with preciseknowledge of yaw stiffness and curvature results in
 
 
 
and
 
 
 
 (since
 
 
 
 
 
 
 
 
 ), therefore
 
 
 
 ,
 
 
 
and
 
 
 
 are unique solution of the system, im-plying the perfect steering conditions of 
 
 
 
 
 
 
 
 and
 
 
 
 
 
 
. Excluding
 
 
 from the control law leads to
 
 
 
 
 
 
 
 
 
, which in turn results in
 
 
 ,
 
 
 and
 
 
 be-ing not equal to zero. This confirms that radial steering is notperfect, and the distance from the ideal depends on the size of 
 
 
since
 
 
 
 
 
 
 
 
 
 
 (finally on the magnitude of 
 
 
 ).Onthe otherhand,the performanceofthe proposedcontrollawis affected by parameter variations such as the inaccuracies of the curvature
 
 
 and the yaw stiffness
 
 
 
 . The
 
 
 (
i.e.
 
 
 
 
 
 
) leads to
 
 
 
 
 
 
 
 
 . Thus,
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
it is clear that the system performance depends both on the
0 5 10 15 20−7−6−5−4−3−2−10
time (sec)
   L  a   t  e  r  a   l  c  r  e  e  p   f  o  r  c  e   (   k   N   )
perfect steering radial steering stability only
(a) Lateral forces
05101520
1.6
1.4
1.2
1
0.8
0.6
0.4
0.200.20.4
time (sec)
  y  a  w  c  r  e  e  p   t  o  r  q  u  e   (   k   N  m   )
stability only radial steering perfect steering
(b) yaw torques
Figure 2: Creep forcescurvaturevariation and the curvatureitself. Moreover,the vari-ations of yaw stiffness
 
 
 
 (
i.e.
 
  
 
 
 
 
 
 
 )yields
 
 
 
 
 
 
 
 
 and
 
 
 
 
 
 
 
 . Then, the in-equation given below reveals how
 
 
 
 affects the steering.
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Equation (12) implies that only two of 
 
 
 ,
 
 
 , and
 
 
 areindependent, indicating the idea of 
 
 
 
 
 and
 
 
 
 
 is suf-ficient to achieve the perfect steering. The equal lateral forces(
i.e.
 
 
 
 
 ) alone however does not necessarily lead to zerolongitudinal creep as
 
 
 
 can also give
 
 
 
 
 
 
 
 .
4 Simulation and analysis
15
 
10
 
505101500.10.20.30.40.50.60.70.80.9
Lateral displacement [mm]
   C  o  n   i  c   i   t  y   [
   −
   ]
Flank contact Conicity for straight forward case
λ
=0.05or so
(a) Nonlinearity of the conicity
05101520
0.500.511.522.533.544.5
time [sec]
   d   i  s  p   [  m  m   ]
Nonlinear conicity Normal case
(b) Lateral displacements
05101520
0.8
0.6
0.4
0.200.20.40.6
time [sec]
   Y  a  w  c  r  e  e  p   t  o  r  q  u  e   [   k   N  m   ]
Non trailingNon leadingNor leadingNor trailing
(c) Yaw creep torques
Figure 3: Results for the nonlinearityComputer simulations are used to compare the performance of the different strategies, which are the stability control only, theradial and the perfect steering.Fig.2 shows that the perfect feedforward strategy yields equallateral and zero longitudinal creep forces in quasi steady-statewhereas the radial steering and the stability control only strat-egy do not. If the lateral creepages of the leading (solid-line)and trailing (dash-line) wheelset are unequal, the yaw creeptorquesatleadingandtrailingwheelsetsareessentialtobalancethe torque they generated. The stability control only strategy isthe worst. Clearly, none of them can reach perfect steering dur-

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->