IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO.

5, OCTOBER 1994 503

An Optimal Speed Controller for

Permanent-Magnet Synchronous Motor Drives
Kuan-Teck Chang, Teck-Seng Low, and Tong-Heng Lee

Abstruct- An optimal control system synthesis method of the PM ac motor, which uses vector control and at the same
which can achieve vector and speed control simultaneously time has quick transient response, must be improved.
for permanent-magnet synchronous motor (PMSM) drives is In this paper, an optimal control system synthesis method is
proposed in this paper. A state-space multiple-input-multiple-
output (MIMO) model for PMSM is first developed and the proposed which can achieve vector control and speed control
compensation for the nonlinearities in this model is discussed. simultaneously for a PMSM drive. The PMSM has nonlinear-
A pseudo-linearized PMSM model is dynamically constructed ities, and it is necessary to perform some kind of linearization
through the state detection, and subsequently an optimal speed on the system before state-feedback control theory could be
controller is developed based on this linearized model. Integral applied. The design is further complicated as the PMSM is
control technique is incorporated to eliminate possible speed
offsets. A speed observer is further developed to eliminate the a multi-input- multi-output (MIMO) system. These problems
speed sensor from the drive. are addressed in this paper. This paper is organized as follows.
In Section 11, the nonlinear model for PMSM is given and
the possibility of using state detection to linearize the system
NOMENCLATURE is discussed. In Section 111, a speed controller is developed
Direct-axis current (A) using optimal control theory and integral control techniques. A
Quadrature-axis current (A) nonlinear speed observer is proposed in Section IV to estimate
Direct-axis voltage (V) the speed of the PMSM and thus to eliminate the speed sensor
Quadrature-axis voltage (V) in PMSM drives. In Section V, simulation studies are carried
Electrical speed (rad/s) out to verify the control and observer system, and experiments
Electrical position (rad) are carried out by using a digital signal processing system.
Load torque (N . m) Section VI concludes this paper.
Phase resistance (R)
Phase inductance (Henry) 11. STATESPACE MODELAND
Rotor inertia (kg . m2) PSEUDO-LINEARIZEDMODELOF PMSM
Damping coefficient (N . m ' shad) The dynamic model for the PMSM in the dq-transformed
Magnetic constant (N . &A)
Number of pole-pair
Observer gain vector

I. INTRODUCTION
dt w,
=[
-R/L
0
aX/J
0
-R/L
0
-X/L

-D/J
0 I,]:[ ["I
rotor reference frame is given in state-space as follows [4]:

V ECTOR control technique, incorporating fast signal pro-

cessing and fast power electronics, have made possible
the application of ac motor drives in high performance tasks
where traditionally only dc servo motor drives were applied. d
-0 - W T .
A permanent-magnet synchronous motor (PMSM) employing dt ' -
vector control is especially favorable for high performance Assuming all states are measurable, the output matrix can

"1 [':I.
servo applications because it fulfills the design criteria of a be written as:
high-performance servo drive, such as compact structure, high
air-gap flux density, high power to inertia ratio, high torque
to inertia ratio, and high torque capability [I]. As the PM ac
motor is replacing conventional dc motors for small output
y= ['
0
0
O1 0
0 1 w,
(3)

power rating variable speed control systems, the performance Several assumptions are made in the derivation of this model.
Magnetic saturation is neglected and the motor is assumed
Manuscript received January 9, 1993; revised September 19, 1993. to have a smooth-rotor and no saliency effect is considered.
The authors are with the Faculty of Engineering, National University of
Singapore, 05 1 1 Singapore. The induced emf is a sinusoidal one and the eddy current
IEEE Log Number 9403905. and hysteresis losses are negligible. As the field excitation is
02784046/94$04.00 0 1994 IEEE

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 from permanent-magnet there are no field current dynamics
and there is no cage on the rotor.
In conventional servo control designs for PMSM, a modi-
fied mathematical model is developed to simplify the nonlinear E = [0 0 - N/JIT
model given in (1). The objective of the modification is to
v(t) = ['q 'dIT
globally linearize the dynamics of the motor which contain
nonlinearities of W,id and wriq, and also to decouple the d ( t ) = TL.
electrical dynamics from the mechanical dynamics. The lin- The system matrix in (5) is time-varying and therefore it is now
earization and decoupling can be achieved by using fast power a time-varying dynamical system. The controllability analysis
electronics and the vector control techniques [l, 31. By using [5] on the system (5) on the it time-varying basis gives the
the fast electronics the currents i d and 2, are forced to follow controllability matrix as follows:
some references commanded by the vector control. The PMSM
model in (1) is then reduced to a simple dc-like, single-input-
single-output (SISO), first-order and time-invariant system to
which various control design methods can be applied: where

d
- aX. D N
r yaq- -wr - -TL.
d t ~=
M k ( t ) = -Aw(t)Mk-l(t) + zdM k - l ( t ) , = 1,2
J J
and
The speed- and position-loop controllers can then be designed
to control the resultant SISO system easily. The use of some Mo(t) = B ( t ) (6)
advanced control techniques such as Wiener-Hopf and H m
optimal control [l], adaptive control [2], variable structure which in
control [3] and others to control this linearized PMSM model
have been reported. Despite the numerous papers available,
however, the dynamics of the PMSM drive during the tran-
sients is rarely discussed.
In subsequent sections, a pseudo-linearized model of
CO = I
R/L2
-w,/L
-aX/(JL)
w,
R/L2
0

It is noted that the matrix c" is speed dependent. Closer

I MZ(t)
1 .

PMSM is derived by using the fact that all of the states of inspection of the matrix reveals that the rank of the matrix
the system, that are currents i d , 2, and speed w,, are detected is always therefore the system described by ( 5 ) is
39

and fed back to the controller. The currents are detected using controllable for inputs ' d and ' 9 .
current transducers and the speed be detected directly The representation Of the PMSM model as in (5) describes a
time-varying multiple-input-multiP1e-output system.
using the tachometer or be derived from the position feedback.
With the state detection the system described in (1) The refOrmation of the system matrix enables a pseudo-
be
reformed to: linearized system to be formed for every moment of the
motor operation. An optimal speed controller could then be
-R/L -w," developed for this pseudo-linearized MIMO system.

'
-w," -R/L

E
111. O m A L CONTROL AND INTEGRAL CONTROL

-k ]!I [El] + [ ]
-N/J
TL (4)
A full state-feedback control is now employed to develop a
speed control system for the PMSM which fulfills two control
objectives as follows:
Of the motor;
where w,"is the detected speed. Together with (2), the PMSM
model can now be written compactly as: 2) Vector control of the motor where maximum torque
performance is obtained by regulating the direct axis
d current (id) at zero value;
- X ( t ) = A , ( t ) X ( t ) + Bu(t)+ E d ( t )
dt
Y ( t )= C X ( t ) ( 5 ) A. Eigenstructure Assignment and Optimal Control
where For a MIMO system, eigenvalues and eigenvectors are to
be assigned to the closed-loop feedback system in order to
X ( t )= [iq id W,IT achieve some desired dynamical transient response character-
istics. In a multivariable system, eigenvectors determine the
-R/L -w,"

[
shape of the response modes and the eigenvalues determine
A,(t)= w," -R/L
the time domain characteristics of the response. In servo
aX/J 0 -D/J
drive control, it is always desirable to have a time-invariant
1/L 0 response, in which the convergence characteristics of the
system is clearly known to the users. In such cases, some
well-established schemes could be used for the assignment of

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 CHANG et al.: SPEED CONTROLLER FOR PERMANENT-MAGNET MOTOR DRIVES 505

be a quadratic performance index which is chosen as:

1
Gain k l

6P\ Jo =
m
+
( X T Q I X wTQzw)dt; (7)

0 &Iand QZ are the weighting matrices. The application of J"

to the control system design achieves an optimal system which

/
0 50 100 speed( rad/s)
compromises the minimum state errors and minimum energy

I:[
0
Gain k2 criteria. Applying the minimum principle, the optimal control
law is obtained:
w(t) = - Q z l G T P X ( t ) (8)
0 50 100
where P is a symmetric positive definite solution of the
algebraic Ricatti equation as follows:
F T P - PGQTIGTP + P F + Q1 = 0 (9)
Or, This results in a optimally controlled system:
0. 50 100
d
Fig. 1. Gain plots for the optimal controller. -X = ( F - GQTIGTP)X
dt
eigenstructure, such as the closed-loop entire eigenstructure B. Integral Control
assignment technique proposed in [7]. Integral control is incorporated into the optimal controller
In applying the technique to a MIMO and time-invariant design for the development of the speed control system. In
system, the system matrix pair must be controllable. The state-space designs, no integral action is incorporated unless
subspaces, determined by ker[A - X I B], in which the eigen- special steps are taken in the design steps. The integral
vectors associated with each assigned eigenvalues are then action is extensively used in servo control to reduce or to
located, and the ability to select from many possible vectors eliminate steady-state errors. Integral action can be added
for a given eigenvalue provides the means for adjusting the into a state-space controller using either state augmentation
magnitudes of each mode which will appear in the output. or bias estimation technique. In the design of an optimally
Following the selection of the eigenvectors, a simple algorithm controlled PMSM, state augmentation technique is used due
is used to compute the required feedback gain matrix. The to its simplicity.
entire eigenstructure assignment technique has been used in
the design of state-feedback control laws for effective shaping C. Speed Controller Designs
of the time responses for MIMO, time-invariant systems [7].
In such systems, the designs always lead to unique feedback When using the integral control in state feedback system,
gain matrix because the system is time-invariant. the number of outputs which can be tracked cannot exceed the
However, this method is inadequate when it is used to number of control input [7]. There are two controlled inputs,
'vd and U
', which are available in the PMSM system, therefore
formulate a feedback control law for a MIMO, time-varying
PMSM system described in (5). The system matrix of PMSM only two outputs can be tracked. Since the objective of the
keeps changing with the states of the system. To maintain the controller is to have good speed tracking and maximum torque
invariant response mode, a varying eigenstructure assignment performance, speed w, and direct current i d are the natural
has to be carried out. The computation of time-varying sub- choices to be the outputs to be tracked.
spaces and the associated eigenvectors is very involved, and it With the integral control action augmented, the following
system is obtained:

.I+[
is very difficult to select the best eigenstructure of the system
at different operating points. -R/L -w: -AIL 0
An alternative to the eigenstructure assignment technique is w," -RIL 0 0
the optimal control [7]. Optimal control theory is particularly - aX/J 0 -D/J 0
attractive in this case since it can handle time-varying MIMO dt e, 0 -10 0 0

[ IF]["'
system effectively. When applied to a MIMO, time-varying -1 0
systems, the optimal control designs transfer the iterations on 0
eigenvalues and eigenvectors to the iterations on elements in a 0
cost function J". The resultant optimized designs will achieve + vd
some compromise between the use of the control effort and 0
the response speed, and at the same time guarantees a stable 0 0 0
system. Each iteration on the parameters in J" produces a
where e d and e, are integral errors in states of zd and w,
candidate design that can be evaluated in the light of the
respectively:
specifications.
In applying optimal control technique to a generalized ed 1 /(iz - id)dt
linear MIMO system of ( F ,G , H), the cost function J" can

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 506 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 5, OCTOBER 1994

Integral

Proportlond
WI , i d
Fig. 2. Optimal speed controller for PMSM.

Fig. 3. Gain plot for speed observer.

Id. iq

Obsemer

wr, Id
Fig. 4. Optimal speed controller with observer.

e, = /(wf - w,)dt
(12) vector is obtained as:
KT = [K1 K2IT = QTIBTP. (16)
2: being the reference direct current (which is usually set
Obviously the gains are also speed dependent. Computations
to zero as required by the vector control) and w,*being the
are carried out to solve the minimization problem for PMSM.
reference speed.
Typical plots for the gain variation with respect to motor speed
Writi,ng (11) as a compact, augmented version of (5):
are shown in Fig. 1, with the motor parameters as listed in
d- - Appendix and the weighting matrix QI and Q z chosen as:
-X = A,X +Bw + Fd (13)
dt rl 0 0 0 0 1
where :ris the augmented state vector. The state-feedback Q 1 = 10 0 0 0.5 0
control law to be used is:

w = -K1X - K2e 0 0 0 0 0.5

(14)

where K1 and K2 are gains obtained through the minimization Q2= 1;

L J
of the cost function J". e = [ e d e,IT is the state error vector. In practice the gains can be stored in the system memory
The minimization problem is solved by using the following and be scheduled and retrieved with respect to the speed of the
algebraic Ricatti equation: PMSM. The resultant optimal controller incorporating integral
control based on (13), (14), (15) and (16) is as shown in Fig. 2.
-T
A,P +
- P B Q ; ~ B ~ P PA, + Q~ = 0. (15)

Compare this equation with (9). The resultant matrix P IV. SPEEDOBSERVER
DESIGNS
obtained from the solution of (15) is a time-varying one since Observer is employed when full-state feedback control is
the matrix &, is speed dependent and time-varying. The gain used and not all of the states of the systems are accessible.

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 ~

CHANG et al.: SPEED CONTROLLER FOR PERMANENT-MAGNET MOTOR DRIVES 507

1 I t

50

0 0
loo 1
i
400 800
Speed ( r a d / s )

1200 ( m s )
01
0
I
400
I
800

Speed ( r a d / s )
1200 (ms)

q,, /
0.5
50

-0.25 I
0 4CO 800 1200
400 800 1200
iq I .

0.251 6(c) i '<d (*)

8.2 I, Oi +'+
, 'L
Ii I Gain k l
-0.25 p-,--
0 4 00 800 1200

0 400 800 1200

-1
0 400 860 1200

I
0.5 \
0 400 800 12co
I
0 400 800 1200
Fig. 5. Response for a step speed reference.

Observer is particularly useful in the PMSM drive where 0 4 GO EO0 1200

speed is not so readily available. In every PMSM drive
Fig. 6. Optimal speed control with impuise load disturbance.
a position transducer is always coupled to the machine to
provide position information for dq-frame transformation and
electronic commutations to achieve the vector control. Ad-
dition of a separate and also directly coupled tachometer to
I
measure the speed is both costly and mechanically infeasible. Digital Signal Position
One solution commonly employed is to estimate the speed Processor PM Motor Encoder
from the measurements of position transducer signals, and .p I
one feedback transducer suffices for both vector control and
speed control feedback. However, as pointed out in [14], the . cunents feedback
position % Digita1&
Interfaces
recursive algorithms used in deriving the speed from position
generally gives inaccurate estimation during the transient Fig. 7. Experimental set-up
periods; Position has to be sensed first in order to estimate
the speed, which results in an inherent lag. Quantization noise
from both the transducer and the estimation scheme is also
introduced and it affects the accuracy of the estimates.
An identity state observer is developed here where the
d
-
dt
'"1 =[
20,
- R0/ L
aA/J
- R0/ L
0
- A I i;
L-i]]

-D/J [:&I
W,

speed is estimated directly from the measurements of id

and i,, which are readily obtained using inexpensive current
transducers. A detail description of this observer is given in
[8]. However for the completeness of this paper, the essentials where and z] are the detected currents from the current
of the design are recapitulated in the following paragraphs. transducers. To simplify the assignment of the observer gains,
With the assumption that only the measurements of id and the measurement matrix y1 is the linear combination of cur-
i, are feasible, the (1) can be reformed as follows: rents:

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 508 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 5, OCTOBER 1994

Time Base : 100 mddiv Scales

Delay : 0 s Speed 1.00 V to 30 rads
Sensitivity : 1 .OO Wdiv Current 1.00 V to 0.2 A

Fig. 8. Experimental results: responses to step speed reference.

An identity state observer for (19) is constructed which

(18) assumes the form of
Lwr J
d -
Rewriting (17) and (18) to a more compact form: -dtX
= A;X + BU + A(y1 - G X )
d
-X
dt
= A;X + BU + E d where X is the estimated state vector, and A = [Zl Z2 ZgIT
y1= cx (19) is the observer proportional gain vector to be chosen to
achieve certain error characteristics. The error dynamics for
where X, B , E and d assume the forms as in ( 5 ) , and the observer system is:

Ai =

c = [l
-R/L

aA/J
0

1 01.
0
-R/L
0
-A/L-i:
ii
-D/J 1 d
-E
dt
= (Ai - AC)E

where state error vector E = X - X . The characteristic

The matrix A; varies with the measurement of current in the equation of the error is then given by
system. By current detection it is now possible to describe the
system behavior from one sample of time to the next by using det[sI - ( A ; - AC)] = 0 (20)
a linear model of (19), which is a function of the currents
measured. where I being the identity matrix. The gain vector can now be
The time-varyingobservability analysis [ 5 ] on system (19) obtained by using the conventional pole-placement technique
gives: [6].If the desired locations of the observer error roots are p l ,

det(0") = - N 2 -J
x
[(++ i d ) - i,]
p z and p3, then the desired observer characteristic equation is:

where 0" is the observability matrix of the system, and det(.)

is the matrix determinant. The system is unobservable when Equating (20) and (21), the gain vector L can be determined
det(0") = 0, that is, when by matching the coefficients of s, s2, s3. The final results show
that the observer gains are the functions of
1) selected pole locations p l , p z and p 3 ;
The observer and control system have to ensure that this 2) the motor currents, id and i,;
condition is never satisfied [ 8 ] . 3) machine parameters.

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 CHANG et al.: SPEED CONTROLLER FOR PERMANENT-MAGNET MOTOR DRIVES 509

Time Base : 100 mddiu &&

Delay : 0 s Specd 1.00 V lo 30 rads
Sensitivity : 1.00 V/div Current 1.00 V to 0.2 A

Fig. 9. Experimental results: responses to inpulsive load disturbance.

If the pole locations are to be maintained throughout the to the regulated state very quickly. The gains change as the
machine operation, the gain have to be scheduled with respect load requirement changes, as shown in Fig. 6(d)-6(f).
to the measured z d and .2, Fig. 3 illustrates the variations of Experiments are carried out to verify some key results.
gains with respect to the currents. Detail of simulation and An imbricated motor has been designed [9] and used as the
implementation of the observer can be found in [8]. experimental motor. The experimental set-up is as shown in
In essence the nonlinear observation keeps the closed- Fig. 7.
loop eigenvalues of the observer system constant. When the A digital signal processing unit using AT &T DSP 32C
nonlinearities in the machine cause the matrix Ai to vary, the is used as the central processing unit. Extensive computation
observer gains are changed in a nonlinear fashion to make has to be carried out in this DSP for various tasks, such
the closed-loop eigenvalues defined by (20) constant. The as the phase transformation for the dq frame conversion,
resultant nonlinear observation produces a consistent response the observation scheme described in Section IV, the optimal
in estimating the state. control scheme described in Section 111 and the controller
The observer developed is incorporated to the optimal feedback gains retrieval. The DSP used is capable of an
speed controller as a sojhvare speed sensor. The resultant execution time of less than 50 ns, using the pipelining effect;
controller system is as shown in Fig. 4. this makes the DSP the best choice for the controller tasks.
PWM current regulation technique is used to switch the
inverter and the inverter sources input currents to the motor.
The system is first commanded to track a speed reference of
v . SIMULATION STUDIES AND EXPERIMENTAL RESULTS
100 rads. Next the disturbance rejection ability of the control
Simulation studies are carried out to verify the validity of system is tested by injecting an impulsive load disturbance to
the proposed optimal speed controller with the speed observer. the motor. The experimental results are as shown in Fig. 8
Parameters of the experimental motor is listed in Appendix. and Fig. 9. The results are conformed with the ones obtained
The weighting matrices Q1 and Q 2 are chosen as in Section in simulations.
111. The choice of the weighting matrices does not affect the
stable steady-state characteristics. However these matrices do
affect the transient behavior of the system, and therefore have VI. CONCLUSION
to be carefully chosen. A typical simulation is as shown in Fig. An optimally controlled permanent-magnet synchronous
5, where a step speed reference is imposed to the motor. The motor drive using state-space design is proposed. In this
motor tracks the reference speed and the id is regulated to zero. system a pseudo-linearized model of the motor is first obtained
Fig. 6 shows the response of the motor when a impulsive load through the state detection. Speed control and vector control
torque is applied to the motor when it is running at a steady are simultaneously achieved by using the proposed scheme. A
speed of 100 rad/s. It can be observed that the motor returns nonlinear speed observer is also used to eliminate the speed

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.
 510 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 5, OCTOBER 1994

sensor from the motor drive. Simulation and experimental Kuan-Teck Chang obtained the B.Eng. Degree
results show the feasibility and effectiveness of the proposed with First Class Honors from the University of
Malaya, Maylasia, in 1989.
controller. He joined the University of Singapore as a re-
search engineer in 1989 and is presently the Labora-
tory Manager of the Magnetics Technology Centre,
Singapore. He is also pursuing the Ph.D. degree
part-time at the Department of Electrical Engineer-
APPENDIX
ing of the National University of Singapore.
Motor parameters used in the simulations and experiment: Mr. Chang’s research interests include permanent
magnet machine control and servo control in data
N = 4
storage devices.
L 0.0183 H
=
R 1.8 R
=
J 1.04 x lop3 kgm2
=
D : 0.005 Nmskad Teck-Seng Low received the B S c . and Ph.D.
x = 0.125 “/A degrees from the University of Southampton,
Southampton, U.K., in 1978 and 1982, respectively.
Input Voltage = 110 V He joined the National University of Singapore
Rated Current = 5 A as a Lecturer in 1983, and is presently an
Associate Professor in the Department of Electrical
Engineering. He was seconded by the University
REFERENCES to be the Director of the Magnetics Technology
Centre.
[I] T. H. Liu and C. H. Liu, “A multiprocessor-based fully digital control Dr. Low’s research interests include the design
architecture for permanent-magnet synchronous motor drives,” IEEE and control of permanent-magnet motors, robotics,
Trans. Power Electron., vol. 5 , no. 4, pp. 413-423, Oct. 1990. and the application of neural networks in machine design and control. He
[2] Raymond B. Sepe and J. H. Lang, “Real-time adaptive control of the has been active in the IEEE internationally. He is a member of the IEEE
permanent-magnet synchronous motor,” IEEE Trans. Indus. Appl., vol. Transnational Committee and also of the Administration Committee of the
27, no. 4, pp. 7 W 7 1 6 , July/Aug. 1991. IEEE Industrial Electronics Society.
13) K. W. Lim, T. S. Low, M. F. Rahman, and L. B. Wee, “A discrete-
time variable-structure controller for a brushless DC motor drive,” IEEE
Trans. Indus. Electron., vol. 38, no. 2, pp. 102-107, Apr. 1991.
[4] P. C. Krause, Analysis of Electrical Machinery New York: McGraw
Hill, 1987. Tong-Heng Lee received the B.A. Degree with
151 C. T. Chen, Linear System Theory and Design. New York: Holt, Rinehart First Class Honors in the Engineering Tripos from
and Winston, 1984. Cambridge University, England, in 1980, and the
[6] G. F. Franklin, J. D. Powell, and A. Emami-Naeini, Feedback Con- M.Sc. and Ph.D. degrees from Yale University, New
trol of Dynamic Systems. Reading, Mass: Addison-Wesley Publishing Haven, Connecticut, in 1984 and 1987, respectively.
Company, 1988. His doctoral work at Yale was in the area of adaptive
[7] J. D’Azzo and C. H. Houpis, Linear Control System Analysis and De- system theory.
sign: Conventional and ModernNew York McGraw-Hill International, Currently he is the Head of the Control Division
1988. and Deputy Director of the Centre of Intelligent
[8] T. S .Low, T. H. Lee, and K. T. Chang, “A nonlinear speed observer for Control at the Department of Electrical Engineering
permanent-magnet synchronous motors,” IEEE Trans. Indus. Electron., __
of the National University of Singapore.
vol. 40, no. 3, June 1993. Dr. Lee’s research interests are in the areas of stability of motion, neural
I91 T. S. Low, K. J. Binns, M. F. Rahman, and L. B. Wee, “A NdFeB networks and expert systems for control, and issues in the design of controllers
excited permanent-magnet m o t o r d e s i g n and performance,” 3rd IEE for fast servo systems. He was a recipient of the Cambridge University Charles
Int. Con$ on Elec. Drives, London, 1987. Baker Prize in Engineering.

Authorized licensed use limited to: Indian Institute of Technology (ISM) Dhanbad. Downloaded on January 30,2024 at 04:08:33 UTC from IEEE Xplore. Restrictions apply.