150 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO.

1, FEBRUARY 2007

Sensorless Control of Induction Motors by Reduced

Order Observer With MCA EXIN + Based Adaptive
Speed Estimation
Maurizio Cirrincione, Member, IEEE, Marcello Pucci, Member, IEEE, Giansalvo Cirrincione, Member, IEEE,
and Gérard-André Capolino, Fellow, IEEE

Abstract—This paper presents a sensorless technique for high- , Direct and quadrature components of
performance induction machine drives based on neural networks. the stator currents in the stator refer-
It proposes a reduced order speed observer where the speed is es- ence frame.
timated with a new generalized least-squares technique based on , Direct and quadrature components of
the minor component analysis (MCA) EXIN + neuron. With this
regard, the main original aspects of this work are the develop- the stator currents in the rotor-flux
ment of two original choices of the gain matrix of the observer, one oriented reference frame.
of which guarantees the poles of the observer to be fixed on one Space vector of the stator flux-link-
point of the negative real semi-axis in spite of rotor speed, and the ages in the stator reference frame.
adoption of a completely new speed estimation law based on the , Direct and quadrature component of
MCA EXIN + neuron. The methodology has been verified exper-
the stator flux linkage in the stator
imentally on a rotor flux oriented vector controlled drive and has
proven to work at very low operating speed at no-load and rated reference frame.
load (down to 3 rad/s corresponding to 28.6 rpm), to have good es- Space vector of the rotor flux-linkages
timation accuracy both in speed transient and in steady-state and in the stator reference frame.
to work correctly at zero-speed, at no-load, and at medium loads. A , Direct and quadrature component of
comparison with the classic full-order adaptive observer under the the rotor flux linkage in the stator ref-
same working conditions has proven that the proposed observer ex-
erence frame.
hibits a better performance in terms of lowest working speed and
zero-speed operation. Stator inductance.
Rotor inductance.
Index Terms—Field oriented control, induction machines, least- Total static magnetizing inductance.
squares (LS), neural networks, reduced order observer, sensorless
control. Resistance of a stator phase winding.
Resistance of a rotor phase winding.
Rotor time constant.
NOMENCLATURE Total leakage factor.
Number of pole pairs.
Space vector of the stator voltages in Angular rotor speed (in mechanical
the stator reference frame. angles).
, Direct and quadrature components of Angular rotor speed (in electrical an-
the stator voltages in the stator refer- gles per second).
ence frame. Sampling time of the control system.
Space vector of the stator currents in
the stator reference frame.
I. INTRODUCTION
O FAR, sensorless control of induction motors [1]–[3] has
Manuscript received June 15, 2005; revised October 28, 2005. Abstract pub-
lished on the Internet November 30, 2006. The work of G. Cirrincione has been
supported under a grant from ISSIA-CNR, Italy in the framework of the MIUR
project n. 211 entitled “Automazione della gestione intelligente della gener-
S been faced with two kinds of methods: those which employ
the dynamic model of the induction machine based on the funda-
azione distribuita di energia elettrica da fonti rinnovabili e non inquinanti e della mental spatial harmonic of the magnetomotive force (mmf) and
domanda di energia elettrica, anche con riferimento alle compatibilità interne e those based on the saliencies of the machine. Among the first,
ambientali, all’affidabilità e alla sicurezza.”
M. Cirrincione was with the ISSIA-CNR, Section of Palermo, Viale delle
the main ones are the open-loop speed estimators [4], MRAS
Scienze snc, 90128 Palermo, Italy. He is now with the Université de Technologie (model reference adaptive system) speed observers [5], even
de Belfort-Montbeliard (UTBM), 90010 Belfort Cedex, France (e-mail: m.cir- based on neural networks [6], [7], full-order Luenberger adaptive
[email protected]).
M. Pucci is with the ISSIA-CNR Section of Palermo, Institute on Intelli-
observers [8]–[11], also with neural networks [12], and reduced
gent Systems for the Automation, Viale delle Scienze snc, 90128 Palermo, Italy order speed observers [13]–[15]. Among the second, some are
(e-mail: [email protected]). based on continuous high-frequency signal injection [16]–[18]
G. Cirrincione and G.-A. Capolino are with the Department of Electrical En- and some on test vectors [19], [20]. This last kind of method-
gineering, University of Picardie-Jules Verne, 80039 Amiens, France (e-mail:
[email protected]; [email protected]). ologies, even if is very promising for position sensorless control
Digital Object Identifier 10.1109/TIE.2006.888776 thanks to the capability of tracking saliencies, either the saturation
0278-0046/$25.00 © 2007 IEEE

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 151

of the main flux or the rotor slotting saliencies, is usually machine in presence of DC biases. The speed observers suffering from
dependant and sometimes requires a suitable machine design this problem are those which employ open-loop flux estimators,
(open or semi-closed rotor slots for rotor slotting tracking). This e.g., open-loop speed estimators and those MRAS systems where
is not the case for the first kind of techniques, among which the the reference model is an open-loop flux estimator [5]–[7], while
full-order Luenberger observer gives very interesting perfor- speed estimators employing closed-loop flux integration, like the
mances, even if with a significant computational requirement. classic full-order adaptive observer [8], do not have this problem.
In this regard, an improvement of this observer based on total In particular, DC drifts are always present in the signal before it is
least-squares (TLSs) speed estimation has been proposed by the integrated, which causes the integrator to saturate with a resulting
authors [12], which has shown a good performance in the speed inadmissible estimation error, and also after the integration
estimation during transients, at very low-speed (down to 0.5 rad/s because of the initial conditions [22]. In general, low-pass (LP)
corresponding to 4.77 rpm) and at zero-speed. Moreover, its filters with very low cutoff frequency are used instead of pure
stability features in regenerating mode at low-speed have been integrators; however, since they fail in low-frequency ranges,
analyzed theoretically and tested experimentally. close to their cutoff frequency, some alternative solutions have
The main goal of this work is the design of an adaptive speed been devised to overcome this problem, e.g., the integrator with
observer, with a performance comparable to that obtainable with saturation feedback [22], the integrator based on cascaded LP
the full-order Luenberger observer. This is achieved by a re- filters [23], [24], the integrator based on the offset vector estima-
duced-order rotor flux observer, which results in lower com- tion and compensation of residual estimation error [4] and the
plexity and computational burden. In fact, the reduced order ob- adaptive neural integrator [25]. With regard to the reduced order
server has to solve a problem of order two, while the full-order adaptive observer, the problem of the DC drift in the integrand
observer of order four. Particularly, this paper presents a new signal exists only for those choices of the observer gain matrix
sensorless technique based on the reduced order observer, where which transform, at certain working speeds of the machine, the
the speed is estimated on the basis of a new generalized least- reduced order observer in an open-loop flux estimator, like the
squares technique, the MCA EXIN + neuron. Moreover, this current voltage model (CVM) in [26] which gives rise to a smooth
work also deals with the development of two original choices transition from the “current” to the “voltage” model according
of the gain matrix of the observer, one of which ensures that to the increase of the rotor speed (see Section III). With such a
the poles of the observer be fixed on one point of the negative choice, below a certain speed and above another one, the observer
real semi-axis, in spite of the variation of the speed of the motor, behaves like a simple open-loop estimator, and therefore suffers
with a consequent dynamic behavior of the flux estimation inde- from the mentioned problem. It is not the case of the proposed
pendent of the rotor speed. The adoption of the completely new gain matrix choice, which is described in Section III.
speed estimation law, based on the MCA EXIN + neuron, en-
sures very low operating speed at no-load and rated load (down B. Inverter Nonlinearity
to 3 rad/s corresponding to 28.6 rpm), good estimation accuracy The power devices of an inverter present a finite voltage drop
also in speed transient and correct zero-speed operation. Dif- in “on-state,” due to their forward nonlinear characteristics.
ferent from [13], which employs a combination of the reduced This voltage drop has to be taken into consideration at low-fre-
order observer, used as reference model, and the simple current quency (low-voltage amplitude) where it becomes comparable
model, used as adaptive model, to estimate the rotor speed, here with the stator voltage itself, giving rise to distortion and dis-
only the reduced order observer is employed, while the rotor continuities in the voltage waveform. Here, the compensation
speed is estimated by the MCA EXIN + algorithm, just on the method proposed by [4] has been employed. This technique is
basis of the stator voltage and current measurements and the es- based on modeling the forward characteristics of each power
timated flux. It should be remarked that the MCA EXIN + sched- device with a piecewise linear characteristics, with an average
uling is more powerful than the other existing techniques, even threshold voltage and with an average differential resistance.
least-squares based, in terms of smoother convergence transient,
shorter settling time, and better accuracy [21]. In addition, the C. Machine Parameter Mismatch
choice of MCA EXIN + neuron allows to take into consideration A further source of error in flux estimation is the mismatch
the measurement flux modeling errors, which influence the ac- of the stator and rotor resistances of the observer with their real
curacy of the speed estimation, since it is inherently robust to the values because of the heating/cooling of the machine. The load
this source of errors. This speed observer has been tested exper- dependent variations of the winding temperature may lead up to
imentally in a rotor-flux-oriented field oriented control (FOC) 50% error in the modeled resistance. Stator and rotor resistances
drive and compared with the classic full-order adaptive observer should be, therefore, estimated online and tracked during the
of [8]. Also, this paper shows a complexity analysis of the pro- operation of the drive. A great deal of online parameter estima-
posed methodology with respect to other observers, both classic tion algorithms have been devised [4], [8], requiring low com-
and based on neural networks. plexity and reduced computational burden when used in control
systems. In any case, it should be emphasized that steady-state
II. LIMITS OF MODEL-BASED SENSORLESS TECHNIQUES estimation of the rotor resistance cannot be performed in sen-
sorless drives, thus rotor resistance variations must be deduced
A. Open-Loop Integration from stator resistance estimation. In the case under study, dif-
One of the main problems of some speed observers, when ferently from [13] where the allocation of the poles of the ob-
adopted in high-performance drives, is the open-loop integration server has been chosen to minimize the sensitivity of the ob-

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 152 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Fig. 1. Block diagram of the MCA EXIN + reduced order observer.

server to the rotor resistance variations, the gain matrix choice (2e)
has been chosen (see Section III) to make the dynamic of the
flux estimation constant, by fixing the position of the poles of where all space vectors are in the stator reference frame:
the observer in a precise point of the negative real semi-axis, in stator current vector,
spite of the variation of the speed of the motor, and therefore the stator voltage vector, rotor flux vector,
sensitivity to the rotor resistance variations has not been consid-
ered as a design criterion of the observer. No stator resistance , , is the rotor speed, and
estimation algorithm [4] has been used, since the goal is to de- is the observer gain matrix. For the list of parameters, see the
velop one observer which is not too complex and computation- Nomenclature.
ally cumbersome. The proposed MCA EXIN + reduced order observer is based
on the classical reduced order flux observer structure, while a
III. MCA EXIN + REDUCED ORDER OBSERVER new speed estimation law is proposed, which is based on the
MCA EXIN + technique. Fig. 1 shows the block diagram of the
A. Reduced Order Observer Equations proposed reduced order observer, whose equations are described
The matrix equations of the reduced order flux observer, with in Section III-A. The rotor speed is estimated by a MCA EXIN
a voltage error used for corrective feedback are [13], [26], [27] + algorithm, on the basis of the estimated rotor flux linkage ,
as well as the measured stator voltage and current space vec-
tors . Moreover, since the gain matrix is time dependant, the
correction term which takes into consideration the time deriva-
tive of the gain matrix is also included in the scheme.

B. Proposed Choice of the Gain Matrix of the Observer

The choice of a suitable gain matrix of the observer has
(1) been a problem largely faced in literature [13] and [26]–[34].
It is well-known [13] that the poles of the reduced order
where observer are the couple of eigenvalues of the matrix
where
(2a)

(2b)
(2c) and .
This paper develops two new choices of the gain matrix and
proposes one of them as the most suitable for sensorless control.
(2d)
The first, called choice 1, makes the observer poles amplitude

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 153

Fig. 2. Pole locus, amplitude versus speed,  versus speed and gain locus with the proposed gain matrix choice.

constant, the second, called fixed pole position (FPP) choice, C. Other Gain Matrix Choices
fixes the position of the poles, in spite of the rotor speed. The Fig. 3 shows the observer pole locus, the amplitude of poles
FPP choice is proposed as the best for sensorless control for the versus the rotor speed, and the damping factor versus the rotor
reasons explained beneath. speed, obtained with five different gain choices of the matrix
The FPP gain matrix choice permits the position of the poles gain; the first has been developed by the authors and the other
of the observer to be fixed on the negative part of the real semi- four have been proposed in literature.
axis at distance from the origin, according to the variation 1) Choice 1: A criterion for choosing the locus of the ob-
of the rotor speed, to ensure the stability of the observer itself. server poles is to make their amplitude constant with respect for
The proposed gain choice is obtained by imposing and the rotor speed. This criterion leads either to the above proposed
and gives solution if or, if , to a semicircle pole
locus with centre in the origin, with radius and lying in the
complex semiplane with negative real part. In this last case, the
(3) position of the poles varies with the rotor speed and therefore to
Correspondingly, the time derivative of the gain matrix to be avoid instability, a maximum rotor speed must be properly
used in the observer scheme is chosen, in correspondence to which the poles of the observer lie
on the imaginary axis. The matrix gain choice which guarantees
this condition is the following:

(4)

Fig. 2 shows the observer pole locus, the amplitude of poles (5)
versus rotor speed, the damping factor versus rotor speed, and
gain locus ( versus ) as obtained with the FPP gain
matrix choice. It shows that this solution permits to keep the This matrix gain is dependant on the rotor speed, and therefore
dynamic of the flux estimation constant, because the amplitude the observer requires the correction term . With such a
of the poles is the constant and the damping factor is always matrix gain choice, the poles are complex with a constant ampli-
equal to 1. This last feature is particularly important for sensor- tude , but with a damping factor which drastically decreases,
less control in high-speed range: in fact, most of the choices of from 1 at zero-speed to about 0 at rated speed and above.
the matrix gain cause a low damping factor at high rotor speed, 2) Choice 2: Proposes the following matrix gain choice [30]:
which can easily cause instability phenomena. Actually, higher
values of the damping factor result in low sensitivity to esti-
mated speed perturbations or parameter variations. (6)

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 154 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Fig. 3. Pole locus, amplitude versus speed, and  versus speed with five matrix gain choices.

With such a matrix gain choice, the poles of the observer are This matrix gain is dependant on the rotor speed, and therefore
imaginary with magnitude increasing with the rotor speed and the observer requires the correction term . With such a
the damping factor drastically reducing at increasing speed, matrix gain choice, the poles of the observer are real and lie
from 1 at zero-speed to about 0 at rated speed and above. This on the negative real semi-axis with magnitude increasing with
choice cancels the contribution of the stator current from the the speed and a damping factor constant with rotor speed and
observer (1). It has the advantage that the gain matrix is not always equal to 1.
dependent on the rotor speed, and therefore is simpler than both 5) Choice 5: Proposes the following matrix gain choice [26]:
the FPP choice and the others; for the same reason, it does not
for
even require the correction term .
for
3) Choice 3: Proposes the following matrix gain choice [13]
for
(9)
Assigned two threshold values and to the rotor speed,
the gain matrix has three different values. Below , no cor-
rection feedback is given to the observer and it behaves as the
(7) simple “current” model of the induction machine, based on its
rotor equations. Above , the correction feedback given to the
This matrix gain is dependant on the rotor speed, and therefore observer is a constant multiplied with the identity matrix, and
the observer requires the correction term . With such a it behaves as the simple “voltage” model of the induction ma-
matrix gain choice, the poles of the observer are complex with chine, based on its stator equations. Between and , the
magnitude increasing with the speed and the damping factor gain matrix linearly varies from the two limit conditions. For
reducing at increasing speed, from 1 at zero-speed to about 0.7 this reason, it has been called current voltage model (CVM),
at rated speed and above. However, in [13], it is claimed that since it gives rise to a smooth transition from the “current”
this matrix gain choice reduces the sensitivity of the observer to to the “voltage” model according to the increase of the rotor
rotor resistance variations. speed. With such a choice, the poles of the observer are com-
4) Choice 4: Proposes also the following matrix gain choice plex with magnitude first increasing and then decreasing with
[30]: the rotor speed, and a damping factor drastically reducing at
increasing speed, from 1 at zero-speed to about 0 at rated speed
and above. As mentioned above, however, this solution makes
the observer work as a simple open-loop estimator both at low
and high speeds, with the consequent dc drift integration prob-
(8) lems. This is not the case neither of the FPP choice nor the other
four ones. See [26] for the choice of and .

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 155

TABLE I
ISSUES OF ALL MATRIX GAIN CHOICES

A slightly different approach is presented in [35], which

proposes an observer where the rotor flux is estimated as the
sum of a high-pass filtered and a LP filtered flux, estimated,
respectively, by the “voltage” and the “current” models. This
leads to a correction term which depends, differently from
the other choices above, on the difference between the two
estimated fluxes, which are subsequently processed by a PI
controller. The resulting observer presents a smooth transi-
tion between “current” and “voltage” model flux estimation Fig. 4. Schematics of the LSs techniques in the monodimensional case.
which is ruled by the closed-loop eigenvalues of the observer,
determined by the parameters of the PI controller. At rotor
speeds below the bandwidth of the observer, its sensitivity to the current time sample. Note that the is applied on the stator
the parameters correspond to that of the “current” model, while voltage space vector to mean that it is computed from the DC
at high speeds its sensitivity corresponds to that of the “voltage link voltage considering the blanking time of the inverter and the
model.” In this sense, it behaves like choice 5. voltage drop on the power devices of the inverter on the basis of
Table I summarizes the features of all six choices, mainly fo- the method proposed in [4]. The same symbol on the rotor flux
cusing on the variation of the observer pole amplitude with the indicates the estimated flux.
rotor speed, the variation of the damping factor with the rotor This matrix equation, which can be written more generally
speed, the dependance on the matrix gain by the rotor speed, as , can be solved for by using LS techniques. In
and the DC drift integration problems. From the standpoint of particular, in literature there exist three LS techniques, i.e., the
the pole amplitude variation, the FPP choice and choice 1 are ordinary least-squares (OLSs), the total least-squares (TLSs),
the best, since they permit the amplitude to be constant; choices and the data least-squares (DLSs) which arise when errors are,
2, 4, and 5 permit a low variation of the pole amplitudes, while respectively, present only in or both in and in or only in
choice 3 causes a high variation. As for the damping factor vari- .
ation, the FPP choice and choice 4 are the best since they keep In classical OLSs, each element of is considered without
always equal to 1; choice 3 permits a low decrease of at in- any error: therefore, all errors are confined to . However, this
creasing rotor speeds, while choices 1, 2, and 5 cause a strong hypothesis does not always correspond to the reality: modeling
reduction of . As for the dependance of on the rotor speed, errors, measurement errors, etc., can cause errors also in .
all the choices except choice 2 suffer from this variation. As for Therefore, in real-world applications, the employment of TLSs
the DC drift integration problems, only choice 5 presents this would be very often better, as it takes also into consideration the
negative issue, especially at low and high rotor speeds. errors in the data matrix.
For the above reasons, FPP choice for the gain matrix is the In the monodimensional case ( ), which is the case
best among the six presented here for sensorless control, and has under study, the resolution of the LS problem consists in deter-
been therefore adopted in the following experimental tests. mining the angular coefficient of the straight line of equation
. The LS technique solves this problem by calculating
IV. MCA EXIN + BASED SPEED ESTIMATION the value of which minimizes the sum of squares of the dis-
The MCA EXIN + based speed estimation derives from a tances among the elements , with , and the
modification of the complete state equations of the induction line itself. Fig. 4 shows the difference among the OLS, TLS,
motor [8], [12] so that it exploits the two first scalar equations and DLS. OLS minimizes the sum of squares of the distances
to estimate the rotor speed, as shown below in discrete form for in the direction (error only in the observation vector). TLS
digital implementation, as shown in (10) at the bottom of the minimizes the sum of squares in the direction orthogonal to the
page, is the sampling time of the control algorithm and is line (for this reason, TLS is also called orthogonal regression),

(10)

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 156 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

while DLS minimizes the sum of squares in the direction (er- This eigenvector can be found by minimizing the following error
rors only in the data matrix). In particular, it must be expected function:
that, in absence of noise, the results obtained with TLS are equal
to those obtained with OLS; however, in presence of increasing
(14)
noise, the performance of TLS remains higher than that of OLS,
as TLS is less sensitive to noise. For these reasons, the TLS algo-
rithm is particularly suitable for estimation processes in which which is the Rayleigh quotient of . Hence, the TLS solu-
data are affected by noise or modeling errors; this is certainly the tion is found by normalizing in order to have the last com-
case of speed estimation, where the estimated rotor flux, present ponent equal to . Resuming, TLS can be solved by applying
in , is affected both by modeling errors and noise. Therefore, MCA to the augmented matrix . [21] also proves the equiv-
a TLS technique should be used instead of the OLSs technique. alence between MCA and DLS in a very specific case. Indeed,
The TLS EXIN neuron, which is the only neural network ca- setting and (DLS) in (11) yields (14) with .
pable to solve a TLS problems recursively online, has been suc- Hence, the MCA for the matrix is equivalent to the DLS of
cessfully adopted in MRAS speed observers [6]. In this work, a the system composed of as the data matrix and of a null obser-
new generalized LS technique, the MCA EXIN + (minor com- vation vector. In particular, TLS by using MCA can be solved
ponent analysis) neuron, is used for the first time to compute the by using (12) and (13) with and with . The
rotor speed. This technique is a further improvement of the TLS advantage of this approach is the possibility of using the sched-
EXIN neuron [36], [37] and is explained below. uling. This technique is the learning law of the MCA EXIN
+ neuron [21], which is an iterative algorithm from a numer-
A. The MCA EXIN + Neuron ical point of view. It yields better results than other MCA iter-
ative techniques because of its smoother dynamics, faster con-
is the linear regression problem under hand. In [38],
vergence, and better accuracy, which are the consequence of the
all LS problems have been generalized by using a parameter-
fact that the varying parameter drives toward the solution
ized formulation (generalized TLS, GeTLS EXIN) of an error
in a smooth way. These features allow higher learning rates for
function whose minimization yields the corresponding solution.
accelerating the convergence and smaller initial conditions (in
This error is given by
[21], it is proven that very low initial conditions speed up the
algorithm).
(11)
V. IMPLEMENTATION ISSUES
where represents the transpose and is equal to 0 for OLS, 0.5 A. Control System
for TLS, and 1 for DLS. The corresponding iterative algorithm
(GeTLS EXIN learning law), which computes the minimizer by The MCA EXIN + reduced order observer has been tested
using an exact gradient technique, is given by on a “voltage” rotor flux oriented vector control scheme [6], [7]
(Fig. 5). For control purposes, the estimated speed has been fed-
back to a PI speed controller and instantaneously compared with
the measured one to compute the speed error at each instant and
(12) in each working condition. Inside the speed loop there is the
where loop. On the direct axis, the voltage is controlled
at a constant value to make the drive automatically work in the
field-weakening region. Inside the loop are, respectively, the
(13)
rotor flux-linkage loop and the loop. The voltage source in-
verter (VSI) is driven by an asynchronous space vector modula-
where is the learning rate, is the row of fed at in- tion algorithm with a switching frequency . The
stant , and is the corresponding observation. The GeTLS phase voltages have been computed on the basis of the instan-
EXIN learning law becomes the TLS EXIN learning law for taneous measurement of the DC link voltage and the switching
equal to 0.5 [38]. The TLS EXIN problem can also be solved by state of the inverter. Moreover, the method proposed in [4] for
scheduling the value of the parameter in GeTLS EXIN, e.g., it the compensation of the on-state voltage drops of the inverter de-
can vary linearly from 0 to 0.5, and then remains constant. This vices has been employed. In the case under study, the employed
scheduling improves the transient, the speed, and the accuracy IGBT modules, which are the Semikron SMK 50 GB 123, have
of the iterative technique [38]. [21] shows that a TLS problem been modeled with a threshold voltage and with an
corresponds to a MCA problem and is equivalent to a particular average differential resistance . Finally, the sam-
DLS problem. Indeed, define as the augmented ma- pling frequency of the acquired signals has been set to 10 kHz, at
trix built by appending the observation vector to the right of the which also all control loops work. For reproducibility reasons,
data matrix. In this case, the linear regression problem can be Table II shows all the parameters of the control system adopted
for the experimental implementation.
reformulated as and can be solved as a homo- As for the integration of state (1) of the reduced order ob-
geneous system ; the solution is given by the eigen- server in the discrete domain, the pure integrator in the con-
vector associated to the smallest eigenvalue of (MCA). tinuous domain has been replaced by the following discrete

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 157

Fig. 5. Implemented FOC scheme.

TABLE II
PARAMETERS OF THE CONTROL SYSTEM

filter in the -domain . This is ob-

tained by the transform of the following discrete equation:

where is the integrator input at the current time sample

and is the corresponding integrator output. This formula Fig. 6. Error surfaces with  = 0 and  = 1 and the MCA EXIN + error
is the sum of a simple Euler integrator and an additional term trajectory versus x.
taking into consideration the values of the integrand variables
in two previous time steps; it guarantees a correct integration of
TABLE III
the state equations, and thus a correct flux estimation with the PARAMETERS OF THE INDUCTION MOTOR
adopted value of differently from the simple forward Euler
integrator .
With reference to the MCA EXIN + reduced order observer,
the only two parameters set by the user have been given the
following values: and (kept constant).
With reference to the parameter , the following scheduling has
been adopted: at each speed transient commanded by the con-
trol system, a linear variation of from 0 to 1 in 0.3 s has been
given. This scheduling has been implemented in software by a
discrete integrator with the constant value 1/0.3 in input, which
permits the output to get the value 1 in 0.3 s with linear law, and
whose output is reset to zero at each change of the reference
speed of the drive. With the above scheduling, the flatness of B. Experimental Setup
the OLS error surface around its minimum, which prevents the The employed test setup consists of the following [6], [7].
algorithm from being fast, is smoothly replaced by a ravine in • A three-phase induction motor with parameters shown in
the corresponding DLS error surface, which speeds up the con- Table III.
vergence to the solution [minimum of (14)] as well as its final • A frequency converter which consists of a three-phase
accuracy. Fig. 6 shows the error surfaces obtained with diode rectifier and a 7.5 kVA, three-phase VSI.
(OLS) and (DLS) and the MCA EXIN + error trajec- • A DC machine for loading the induction machine with pa-
tory versus the two components of with regard to the DLS rameters shown in Table IV.
error surface, obtained when a speed step reference from 0 to • An electronic AC-DC converter (three-phase diode recti-
150 rad/s has been given to the drive without load. It should be fier and a full-bridge DC-DC converter) for supplying the
remarked that the proposed speed observer does not need any DC machine of rated power 4 kVA.
LP filtering of the estimated speed to be fed back to the control • A dSPACE card (DS1103) with a PowerPC 604e at
system, with consequent higher bandwidth of the speed loop. 400 MHz and a floating-point DSP TMS320F240.

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 158 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

TABLE IV achievable in infinite precision (see [36] for more details). This
PARAMETERS OF THE DC MACHINE kind of error accumulates in time without bound, leading in the
long run (ten of million of iterations) to an eventual overflow
(the so called numerical divergence). The source of this di-
vergence is both the analog-to-digital (A/D) conversion and the
finite word length used to store all internal algorithmic quanti-
ties. The degradation of the solution is proportional to the condi-
tioning of the input, i.e., to the eigenvalue spectrum of the input
autocorrelation matrix [36]. Decreasing the learning rate in
C. Hardware Choice the infinite precision algorithm leads to improved performance.
With regard to the hardware implementation of the pro- Neverthless, this decrease increases the deviation from infinite
posed sensorless control technique, a floating point DSP precision performance, while its increase magnifies numerical
(TMS320F240) has been employed, since its programming errors, so a tradeoff is required [41]. A technique for avoiding
needs a lower development time than a fixed point DSP. In overflows caused by this kind of divergence, at the expense of
general, the choice between floating or fixed point DSPs is some increase of complexity and small degradation of the solu-
highly dependent on the application at hand, its complexity, and tion, is the technique called leakage (see [36] for its details).
its cost reduction needs. For industry products, the production Finally, the MCA EXIN + algorithm does not need partic-
volumes play a relevant role in this choice: for high volumes, ularly critical operations (the most complex operation is a dot
the lower cost of a fixed point DSP justifies the amount of product between vectors of dimension two) and explicitly em-
any nonrecurring engineering (NRE) cost [39], like control ploys neither the covariance matrix [40] nor its inverse compu-
software development, while for low volumes, the NRE cost tation, as needed in some versions of the recursive least-squares
dominates and thereby floating point choice seems to be the (RLSs). In any case, it should be remarked that, independently
best, as it reduces the development costs. In the industrial of the DSP choice, the elements of data matrix and the obser-
drives market, where high production volumes dominate, the vation vector (both bidimensional vectors) processed by the
cost issue is important, and thus the potential implementation MCA EXIN + algorithm must be properly scaled before pro-
of each technique on fixed point DSP is a real advantage. cessing, so that the estimated will result in a quantity ranging
The big difference between floating and fixed point DSPs is between and 1, otherwise, the effect of the parameter in
the coding of the numbers. Fixed point chips, differently from (11) is null. This intrinsic algorithm demand further facilitates
floating point ones, need a scaling factor to be managed by the the potential conversion of the software into fixed point format.
user. If all variables of the program are normalized between
and 1, not even the scaling factor is needed. Moreover, fixed D. Computational Complexity
point DSPs are also capable of a finer resolution than fixed
From the computational point of view, the MCA EXIN + re-
point ones (of the same word length) because of the extra bits
in the mantissa. In drive control, the development of the en- duced order observer has been compared here with some neural
tire control system, including the observer, can be performed network-based observers, in particular, the TLS EXIN full-order
by working in per-unit (p.u.), i.e., by dividing each values by observer [12] and the TLS EXIN MRAS observer with adap-
the corresponding base value: this assumption highly simpli- tive neural integrator [6], and with the classic full-order ob-
fies the transport of the software from floating point to fixed server [8]. This comparison has been done on the basis of the
point format, exploiting also its better resolution. In addition number of floating operations (flops) needed by each algorithm
to this, many development tools have been recently proposed for every iteration. The comparison is shown in Table V. If the
to convert control software from floating point to fixed point correction term is not adopted in the proposed observer,
representation [39]. the most demanding observer is the TLS EXIN full-order ob-
Some specific suggestions for the implementation on fixed server which requires 126 flops + 3 IF-THEN instructions and
point DSPs of an MRAC speed controller for a FOC induction then there is the MCA EXIN + reduced order speed observer
motor drive with a LSs-based parameter estimator are given in with 120 flops + 4 IF-THEN instructions. If the correction term
[40]. [40] explicitly addresses the difficulties associated with the is adopted, the proposed observer is the most complex,
large dynamic range of the covariance matrix (used by the LSs with 147 flops + 4 IF-THEN instructions. However, a series of
algorithm) with respect to the finite length of the DSP word. The experimental tests performed with and without this term have
proposed solution is the adoption of a double-word fixed point shown that the MCA EXIN + reduced order observer can be im-
representation, which increases the dynamic range of the pro- plemented either without the without a significant wors-
cessed data, and consequently reduces the rounding/truncation ening of the performance of the flux and speed estimation. Then,
errors, at the expense of higher execution time. there is the TLS EXIN MRAS observer with the adaptive neural
With regard to the proposed technique, the state equations of integrator which requires 87 flops + 3 IF-THEN instructions,
the reduced order observer can be implemented in software by and finally the classic full-order observer requiring 76 flops.
simple operations (sums and multiplications/divisions) between However, it should be remarked that both the TLS EXIN
quantities in p.u. With specific regard to the MCA EXIN + algo- full-order observer in [12] and the classic full-order observer in
rithm, any quantization error, which is larger in fixed point than [8] are implemented with a matrix gain correction term which
in floating point DSPs, can degradate its solution (typical of all is null, i.e., without any feedback term. The adoption of such a
of gradient-based algorithms), with respect to the performance term would require the product of a (2 4) matrix for a (4 1)

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 159

TABLE V
COMPLEXITY OF THE PROPOSED OBSERVER COMPARED WITH OTHERS IN LITERATURE

Fig. 7. Reference, estimated, and measured speed during a 50–0–50 rad/s test at no-load (experimental).

vector, where the gain matrix could be either constant or variable appendix). Moreover the results obtained experimentally have
with the speed of the machine in dependence on the desired ob- been compared with those obtained with the full-order classic
server dynamics, and thus would highly increase the complexity adaptive observer proposed in [8]. The parameters of the full-
of both observers. It can be concluded that with in all order classic observer are exactly the same as those suggested
cases the reduced order observer requires fewer flops than the in [8]. Note also that in the full-order classic observer no com-
full-order ones. In any case, the total flops of the different ob- pensation of the inverter nonlinearity has been considered. On
servers is of the same order. the other hand, the parameter estimation method of [8] has been
adopted in the full-order classic observer only. Simulations have
been performed in Matlab®–Simulink®. With regard to the ex-
VI. EXPERIMENTAL RESULTS
perimental tests the speed observer as well as the whole control
The proposed MCA EXIN + reduced order observer has been algorithm have been implemented by software on the DSP of
verified in simulation and experimentally on a test setup (see the dSPACE 1103.

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 160 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Fig. 8. Reference, estimated, and measured speed during a set of speed step references (experimental).

A. Dynamic Performances TABLE VI

BANDWIDTH OF THE SPEED LOOP VERSUS THE REFERENCE
SPEED (EXPERIMENTAL)
As a first test, the drive has been operated at the constant
speed of 50 rad/s at no-load, then a zero step reference has been
given and the drive has been operated at zero-speed for almost
2 s, and then a step speed reference of 50 rad/s at no-load has
been given. Fig. 7 shows the waveforms of the reference, esti- estimated, and measured speed during a set of speed reversal,
mated (used in the feedback loop), and measured speed during respectively, from 3 to , from 4 to , from 5
this test. It shows that the measured speed and the estimated one to , and from 6 to . These last figures show
both follow correctly the reference, even at zero-speed. that the drive is able to perform a speed reversal also at very low
Subsequently, the transient behavior of the observer at very speeds, i.e., in very challenging conditions. However, it should
low speeds has been tested. First, the drive has been given a set be noted that the lower the speed reference, the higher the time
of speed step references at very low speed, ranging from 3 rad/s needed for the speed reversal, as expected, because of the reduc-
(28.65 rpm) to 6 rad/s (57.29 rpm). Fig. 8 shows the wave- tion of the speed bandwidth of the observer at decreasing speed
forms of the reference, estimated and measured speed during references, which is typical of all observers.
this test, and Table VI shows the 3 dB bandwidth of the speed
loop versus the reference speed of the drive. Both Fig. 8 and B. Low-Speed Limits
Table VI show a very good dynamic behavior of the drive with In this test, the drive has been operated at a constant very
a bandwidth which, however, decreases from 69.3 rad/s at the low-speed (3 rad/s corresponding to 28.65 rpm), at no-load and
reference speed of 6 rad/s to 12.3 rad/s at 3 rad/s. This con- rated load. Fig. 10 shows the reference, estimated, and measured
sideration is confirmed by Fig. 9 which shows the reference, speed during these tests. It shows that the steady-state speed

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 161

Fig. 9. Reference, estimated, and measured speed during a set of speed reversal (experimental).

Fig. 10. Reference, estimated, and measured speed during a constant speed of 3 rad/s at no-load and rated load with the MCA EXIN + reduced order observer
(experimental).

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 162 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Fig. 11. Reference, estimated and measured speed during a constant speed of 4 rad/s at no-load and rated load with the classic full-order observer (experimental).

Fig. 12. Reference, measured , estimated speed, and load torque at the constant speed reference of 30 rad/s with two consecutive load torque steps of 65 Nm
(experimental).

estimation error is very low, equal to 2.45% at no-load and to when giving a constant reference speed of 4 rad/s (38.19 rpm),
7.67% with rated load. For comparison reasons, the test has been at no-load and at rated load for the classic full-order observer. It,
also performed with the full-order classic observer [8]: Fig. 11 therefore, shows that the mean estimation error is about 30% at
shows the reference, estimated, and measured speed, obtained no-load and 30.5% at rated load. The comparison shows a better

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 163

Fig. 13. (a) Reference, estimated, measured speed, and position at zero-speed at no-load (experimental). (b) Reference, estimated, measured speed, and position
at zero-speed with 5 Nm load torque (experimental).

accuracy in the speed estimation of the MCA EXIN + reduced has been operated at the constant speed of 30 rad/s and two
order observer than the one the classic full-order adaptive ob- subsequent load torque square waveforms of amplitude
server, even at a higher reference speed (4 rad/s against 3 rad/s). have been applied. Fig. 12 shows the reference, measured, and
Below 2 rad/s, however, the speed accuracy estimation of the estimated speed during this test, as well as the applied load
MCA EXIN + reduced order observer drastically reduces. torque, created by the torque of a controlled DC machine. These
figures show that the drive response occurs immediately when
C. Rejection to Load Perturbations the torque steps are given. Moreover, even during the speed tran-
In this test, to verify the robustness of the speed response of sient caused by the torque step, the estimated speed follows the
the proposed observer to a sudden torque perturbation, the drive real one very well.

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 164 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Fig. 14. Reference, estimated, measured speed, and speed estimation error during zero-speed operation at no-load with the classic full-order observer (experiment).

D. Zero-Speed Operation This work deals with those sensorless techniques of induction
Finally, to test the zero-speed operation capability of the ob- machine drives based on the fundamental harmonic of the mmf.
server, the drive has been operated for 60 s fully magnetized at In particular, it is in the framework of previous LSs based sen-
zero-speed with no-load. Fig. 13(a), which shows the reference, sorless techniques developed by the authors. However, the target
estimated, measured speed and position during this test shows of this work is the design of an observer with performances com-
the zero-speed capability of this observer, and highlights a not parable to those obtainable with the full-order Luenberger ob-
perceptible movement of the rotor during this test, which is con- server, but with lower computational burden. The main original
firmed from the graph of the measured position. The same kind aspects of this work are the following: 1) the development of two
of test has been performed at the constant load torque of 5 Nm. original choices of a gain matrix of the observer, one of which
Fig. 13(b) shows the reference, estimated, measured speed and (the FPP choice) ensures the poles of the observer to be fixed on
position during this test, and highlights that the measured speed one point of the real axis, in spite of the variation of the speed
is in average close to zero and the rotor has an undesired angular of the motor, with a resulting dynamic behavior of the flux es-
movement of 2 rad, achieved in 60 s with a constant applied timation of the observer independent of the rotor speed and 2)
load torque of 5 Nm. This is the ultimate working condition at the adoption of a completely new speed estimation law based
zero-speed. Above 5 Nm load torque, the rotor begins to move on the MCA EXIN + neuron, which guarantees lower operating
and instability occurs. For comparison reasons, Fig. 14 shows speed at no-load and rated load, good estimation accuracy also
the reference, estimated, measured speed, and the instantaneous in speed transient and correct zero-speed operation. The choice
speed estimation error obtained with the classic full-order ob- of the MCA EXIN + neuron allows the observer to take into
server [8] at zero-speed with no-load. The classic observer at consideration the measurement flux modeling errors, which in-
almost 15 s after the magnetization of the machine, has an un- fluence the accuracy of the speed estimation.
stable behavior and the machine eventually runs at 45 rad/s A suitable test setup has been developed for the experimental
with a mean speed estimation error of 13.74 rad/s. The com- assessment of the methodology. An experimental compar-
parison shows a better accuracy in the speed estimation of the ison with the classic full-order observer has shown that the
MCA EXIN + reduced order observer than the classic full-order MCA EXIN + reduced order observer can work correctly
adaptive observer, which has an unstable behavior after a few down to 3 rad/s (28.65 rpm), while the classic full-order ob-
seconds. server presents a worse speed estimation accuracy at 4 rad/s
(38.19 rpm). Moreover, the MCA EXIN + reduced order
VII. CONCLUSION observer works properly at zero-speed without load and with
This paper presents a new sensorless technique which is based medium/low loads, whereas the classic full-order observer has
on a reduced order observer where the speed is estimated by not the same performance, at least with the observer tuning
a new neural LSs-based technique, the MCA EXIN + neuron. proposed in [8].

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
CIRRINCIONE et al.: SENSORLESS CONTROL OF INDUCTION MOTORS BY REDUCED ORDER OBSERVER 165

REFERENCES [24] J. O. P. Pinto, B. K. Bose, and L. E. Borges de Silva, “A stator-flux-ori-

ented vector-controlled induction motor drive with space-vector PWM
[1] K. Rajashekara, A. Kawamura, and K. Matsuse, Sensorless Control of and flux-vector synthesis by neural network,” IEEE Trans. Ind. Appl.,
AC Motor Drives. Piscataway, NJ: IEEE Press, 1996. vol. 37, no. 5, pp. 1308–1318, Sep./Oct. 2001.
[2] P. Vas, Sensorless Vector and Direct Torque Control. Oxford, U.K.: [25] M. Cirrincione, M. Pucci, G. Cirrincione, and G. A. Capolino, “A new
Oxford Science, 1998. adaptive integration methodology for estimating flux in induction ma-
[3] Holtz, “Sensorless control of induction motor drives,” Proc. IEEE, vol. chine drives,” IEEE Trans. Power Electron., vol. 19, pp. 25–34, Jan.
90, pp. 1359–1394, Aug. 2002. 2004.
[4] J. Holtz and Q. Juntao, “Drift- and parameter-compensated flux [26] L. Harnefors, “Design and analysis of general rotor-flux-oriented
estimator for persistent zero-stator-frequency operation of sensor- vector control systems,” IEEE Trans. Ind. Electron., vol. 48, pp.
less-controlled induction motors,” IEEE Trans. Ind. Appl., vol. 39, pp. 383–390, Apr. 2001.
1052–1060, Jul.-Aug. 2003. [27] G. C. Verghese and S. R. Sanders, “Observers for flux estimation in
[5] C. Shauder, “Adaptive speed identification for vector control of induc- induction machines,” IEEE Trans. Ind. Electron., vol. 35, pp. 85–94,
tion motors without rotational transducers,” IEEE Trans. Ind. Appl., Feb. 1988.
vol. 28, no. 5, pp. 1054–1061, Sep./Oct. 1992. [28] R. Nilsen and M. P. Kazmierkowski, “Reduced-order observer with pa-
[6] M. Cirrincione, M. Pucci, G. Cirrincione, and G. A. Capolino, “A new rameter adaption for fast rotor flux estimation in induction machines,”
TLS based MRAS speed estimation with adaptive integration for high Proc. Inst. Elect. Eng. D, Control Theory Appl., vol. 136, no. 1, pp.
performance induction motor drives,” IEEE Trans. Ind. Appl., pp. 1–22, 35–43, Jan. 1989.
Jul./Aug. 2004. [29] ——, “New reduced-order observer with parameter adaptation for flux
[7] M. Cirrincione and M. Pucci, “An MRAS speed sensorless high perfor- estimation in induction motors,” in Proc. IEEE Power Electron. Spe-
mance induction motor drive with a predictive adaptive model,” IEEE cialists Conf., Jun.-Jul. 29–3, 1992, vol. 1, pp. 245–252.
Trans. Ind. Electron., vol. 52, pp. 532–551, Apr. 2005. [30] G. Franceschini, M. Pastorelli, F. Profumo, C. Tassoni, and A. Vagati,
[8] H. Kubota, K. Matsuse, and T. Nakano, “DSP-based speed adaptive “About the gain choice of flux observer in induction servo-motors,”
flux observer of induction motor,” IEEE Trans. Ind. Appl., vol. 29, pp. in Proc. Conf. Record IEEE Industry Appl. Soc. Annu. Meeting,, Oct.
344–348, Mar./Apr. 1993. 7–12, 1990, vol. 1, pp. 601–606.
[9] H. Kubota, I. Sato, Y. Tamura, K. Matsuse, H. Ohta, and Y. Hori, [31] A. Damiano, G. Gatto, I. Marongiu, and A. Pisano, “Synthesis and dig-
“Stable operation of adaptive observer based sensorless induction ital implementation of a reduced order rotor flux observer for IM drive,”
motor drives in regenerating mode at low speeds,” in Proc. IAS Annu. in Proc. IEEE ISIE, Bled, Slovenia, pp. 729–734.
Meeting, Oct. 2001, pp. 469–474. [32] C. Zell and A. Medvedev, “Reduced-order flux observers with arbitrary
[10] H. Kubota, K. Matsuse, and Y. Hori, “Behavior of sensorless induction convergence rate,” in Proc. 4th IEEE Conf. Control Appl., Sep. 28–29,
motor drives in regenerating mode,” in Proc. PCC , Japan, 1997, pp. 1995, pp. 793–798.
549–552. [33] S.-U. Kim, I.-W. Yang, E.-J. Lee, Y.-B. Kim, J.-T. Lee, and Y.-S. Kim,
[11] C. Lascu, I. Boldea, and F. Blaabjerg, “A modified direct torque control “Robust speed estimation for speed sensorless vector control of induc-
for induction motor sensorless drive,” IEEE Trans. Ind. Appl., vol. 36, tion motors,” in Proc. Record IEEE Ind. Appl. Conf., Oct. 3–7, 1999,
pp. 122–130, Jan.-Feb. 2000. vol. 2, pp. 1267–1277.
[12] M. Cirrincione, M. Pucci, G. Cirrincione, and G. Capolino, “An adap- [34] C.-M. Lee and C.-L. Chen, “Observer-based speed estimation method
tive speed observer based on a new total least-squares neuron for in- for sensorless vector control of induction motors,” Proc. Inst. Elect.
duction machine drives,” IEEE Trans. Ind. Appl., vol. 42, no. 1, pp. Eng. D, Control Theory Appl., vol. 145, no. 3, pp. 359–363, May 199.
89–104, Jan./Feb. 2006. [35] P. J. Jansen and R. D. Lorenz, “Observer-based direct field orienta-
[13] H. Tajima and Y. Hori, “Speed sensorless field-orientation of the in- tion: Analysis and comparison of alternative methods,” IEEE Trans.
duction machine,” IEEE Trans. Ind. Appl., vol. 29, no. 1, pp. 175–180, Ind. Appl., vol. 30, pp. 945–953, Jul./Aug. 1994.
Jan./Feb. 1993. [36] G. Cirrincione, M. Cirrincione, J. Hérault, and S. Van Huffel, “The
[14] Y.-N. Lin and C.-L. Chen, “Adaptive pseudoreduced-order flux ob- MCA EXIN neuron for the minor component analysis: Fundamen-
server for speed sensorless field-oriented control of IM,” IEEE Trans. tals and comparisons,” IEEE Trans. Neural Netw., vol. 13, no. 1, pp.
Ind. Electron., vol. 46, pp. 1042–1045, Oct. 1999. 160–187, Jan. 2002.
[15] J. Song, K.-B. Lee, J.-H. Song, I. Choy, and K.-B. Kim, “Sensorless [37] G. Cirrincione and M. Cirrincione, “Linear system identification by
vector control of induction motor using a novel reduced-order extended using the TLS EXIN neuron,” Neurocomputing, vol. 28, no. 1-3, pp.
Luenberger observer,” in Proc. Record IEEE Conf. Industry Appl., Oct. 53–74, Oct. 1999.
2000, vol. 3, pp. 1828–1834. [38] G. Cirrincione, M. Cirrincione, and S. Van Huffel, “The GeTLS EXIN
[16] P. L. Jansen and R. D. Lorenz, “Transducerless position and velocity neuron for linear regression,” in Proc. IJCNN, Como, Italy, Jul. 2000,
estimation in induction and salient AC machines,” IEEE Trans. Ind. pp. 285–289.
Appl., vol. 31, pp. 240–247, Mar./Apr. 1995. [39] C. Inacio and D. Ombres, “The DSP decision: Fixed point or floating?,”
[17] F. Briz, M. W. Degner, A. Diez, and R. D. Lorenz, “Static and dy- IEEE SPECTRUM, vol. 33, no. 9, pp. 72–74, Sep. 1996.
namic behavior of saturation-induced saliencies and their effect on car- [40] L. Chen, J. C. Balda, and K. J. Olejniczak, “Model reference adap-
rier–signal–based sensorless AC drives,” IEEE Trans. Ind. Appl., vol. tive control-implementation considerations on an integer-based DSP,”
38, pp. 670–678, May/Jun. 2002. in Proc. 30th IEEE Ind. Appl. Conf. IAS Annu. Meeting, Oct. 8–12,
[18] N. Teske, G. M. Asher, M. Sumner, and K. J. Bradley, “Analysis and 1995, vol. 2, pp. 1612–1618.
suppression of high-frequency inverter modulation in sensorless posi- [41] R. D. Gitlin, J. E. Mazo, and M. G. Taylor, “On the design of gradient
tion-controlled induction machine drives,” IEEE Trans. Ind. Appl., vol. algorithms for digitally implemented adaptive filters,” IEEE Trans. Cir-
39, pp. 10–18, Jan./Feb. 2003. cuits Syst., vol. 20, no. 2, pp. 125–136, Mar. 1973.
[19] C. S. Staines, G. M. Asher, and M. Sumner, “Rotor position estima-
tion for induction machines at zero and low frequency utilising zero se-
quence currents,” in Proc. Record IEEE 39th IAS Annu. Meeting Conf. Maurizio Cirrincione (M’03) received the Laurea
Industry Appl. , Oct. 3–7, 2004, vol. 2, pp. 1313–1320. degree in electrical engineering from the Politecnico
[20] J. Holtz and H. Pan, “Acquisition of rotor anisotropy signals in sensor- of Turin, Turin, Italy, in 1991 and the Ph.D. degree
less position control systems,” IEEE Trans. Ind. Appl., vol. 40, no. 5, in electrical engineering from the University of
pp. 1379–1387, Sep./Oct. 2004. Palermo, Palermo, Italy.
[21] G. Cirrincione, “A Neural Approach to the Structure from Motion From 1996 to 2005, he has been a Researcher at
Problem,” Ph.D. dissertation, INPG (Institut National Polytechnique the Section of Palermo, ISSIA-CNR (Institute on
de Grenoble), Grenoble, France, 1998. Intelligent Systems for the Automation), Palermo.
[22] J. Hu and B. Wu, “New integration algorithms for estimating motor Since 2005, he has been a Full Professor of Con-
flux over a wide speed range,” IEEE Trans. Power Electron., vol. 13, trol Systems at the Technological University of
pp. 969–977, Sep. 1998. Belfort-Montbéliard, France His current research
[23] L. E. Borges de Silva, B. K. Bose, and J. O. P. Pinto, “Recurrent- interests are neural networks for modeling and control, system identification,
neural-network-based implementation of a programmable cascaded LP intelligent control, electrical machines, and drives.
filter used in stator flux synthesis of vector-controlled induction motor Dr. Cirrincione was awarded the prize “E.R.Caianiello” for the Best Italian
drive,” IEEE Trans. Ind. Electron., vol. 46, pp. 662–665, Jun. 1999. Ph.D. Thesis on neural networks.

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.
 166 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 54, NO. 1, FEBRUARY 2007

Marcello Pucci (M’03) received the Laurea degree of the Department of Electrical Engineering. From 1981 to 1994, he has
and Ph.D. degree in electrical engineering from the been Associate Professor at the University of Dijon, Dijon, France, and the
University of Palermo, Palermo, Italy, in 1997 and Mediterranean Institute of Technology, Marseille, where he was founder and
2002, respectively. Director of the Modeling and Control Systems Laboratory. From 1983 to
In 2000, he was a host student at the Institut of 1985, he was Visiting Professor at the University of Tunis, Tunisia. From 1987
Automatic Control, Technical University of Braun- to 1989, he was the Scientific Advisor of the Technicatome SA Company,
schweig, Germany, working in the field of control of Aix-en-Provence, France. In 1994, he joined the University of Picardie “Jules
AC machines. Since 2001, he has been a Researcher Verne,” Amiens, France, as a Full Professor, Head of the Department of
at the Section of Palermo, ISSIA-CNR (Institute on Electrical Engineering (1995–1998), and Director of the Energy Conversion
Intelligent Systems for the Automation), Palermo. and Intelligent Systems Laboratory (1996–2000). He is now Director of the
His current research interests are electrical machines, Graduate School in Electrical Engineering, University of Picardie “Jules
control, diagnosis and identification techniques of electrical drives, intelligent Verne.” In 1995, he was a Fellow European Union Distinguished Professor
control, and power converters. of Electrical Engineering at Polytechnic University of Catalunya, Barcelona,
Spain. Since 1999, he has been the Director of the Open European Labora-
tory on Electrical Machines (OELEM), a network of excellence between 50
partners from the European Union. He has published more than 250 papers
Giansalvo Cirrincione (M’03) received the Laurea in scientific journals and conference proceedings since 1975. He has been the
degree in electrical engineering from the Politec- Advisor of 13 Ph.D. and numerous M.Sc. students. In 1990, he has founded
nico of Turin, Turin, Italy, in 1991 and the Ph.D the European Community Group for teaching electromagnetic transients and
degree from the Laboratoire d’Informatique et he has coauthored the book Simulation & CAD for Electrical Machines, Power
Signaux (LIS) de l’Institut National Polytechnique Electronics and Drives (ERASMUS Program Edition). His research interests
de Grenoble (INPG), Grenoble, France, in 1998. are electrical machines, electrical drives power electronics, and control systems
He was a Postdoctoral at the Department of Sig- related to power electrical engineering.
nals, identification, system theory and automation Prof. Capolino is the Chairman of the France Chapter of the IEEE Power
(SISTA), Leuven University, Leuven, Belgium, Electronics, Industrial Electronics and Industry Applications Societies and the
in 1999 and since 2000, he has been an Assistant Vice-Chairman of the IEEE France Section. He is also member of the AdCom
Professor at the University of Picardie-Jules Verne, of the IEEE Industrial Electronics Society. He is the co-founder of the IEEE
Amiens, France. Since 2005, he has been Visiting Professor at the Section of International Symposium for Diagnostics of Electrical Machines Power Elec-
Palermo, ISSIA-CNR (Institute on Intelligent Systems for the Automation), tronics and Drives (IEEE-SDEMPED) that was held for the first time in 1997.
Palermo, Italy. His current research interests are neural networks, data analysis, He is a member of steering committees for several high reputation international
computer vision, brain models, and system identification. conferences. Since November 1999, he has been Associate Editor of the IEEE
TRANSACTIONS ON INDUSTRIAL ELECTRONICS.

Gérard-André Capolino (A’77–M’82–SM’89–

F’02) received the B.Sc. degree in electrical en-
gineering from Ecole Supérieure d’Ingénieurs de
Marseille, Marseille, France, in 1974, the M.Sc.
degree from Ecole Supérieure d’Electricité, Paris,
France, in 1975, the Ph.D. degree from the University
Aix-Marseille I, Marseille, in 1978, and the D.Sc.
degree from the Institut National Polytechnique de
Grenoble, Grenoble, France, in 1987.
In 1978, he joined the University of Yaoundé
(Cameroon) as an Associate Professor and Head

Authorized licensed use limited to: Reva Institute of Tehnology and Management. Downloaded on December 14, 2008 at 00:04 from IEEE Xplore. Restrictions apply.