You are on page 1of 9

14 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO.

1, JANUARY 1999

Induction Motors’ Faults Detection and Localization


Using Stator Current Advanced Signal
Processing Techniques
Mohamed El Hachemi Benbouzid, Member, IEEE, Michelle Vieira, and Céline Theys

Abstract— The reliability of power electronics systems is of broken rotor bars and mechanical unbalance [1]–[7]. All of
paramount importance in industrial, commercial, aerospace, and the presently available techniques require the user to have
military applications. The knowledge about fault mode behavior some degree of expertise in order to distinguish a normal
of an induction motor drive system is extremely important from
the standpoint of improved system design, protection, and fault- operating condition from a potential failure mode. This is
tolerant control. This paper addresses the application of motor because the monitored spectral components (either vibration
current spectral analysis for the detection and localization of ab- or current) can result from a number of sources, including
normal electrical and mechanical conditions that indicate, or may those related to normal operating conditions. This requirement
lead to, a failure of induction motors. Intensive research effort is even more acute when analyzing the current spectrum of an
has been for some time focused on the motor current signature
analysis. This technique utilizes the results of spectral analysis induction motor since a multitude of harmonics exist due to
of the stator current. Reliable interpretation of the spectra is both the design and construction of the motor and the variation
difficult since distortions of the current waveform caused by in the driven load. Many of these harmonics can be caused by
the abnormalities in the induction motor are usually minute. ovalities in the rotor, voids in the casting, slot design, etc.
This paper takes the initial step to investigate the efficiency Condition monitoring of the dynamic performance of elec-
of current monitoring for diagnostic purposes. The effects of
stator current spectrum are described and the related frequencies trical drives received considerable attention in recent years.
determined. In the present investigation, the frequency signature Many condition monitoring methods have been proposed
of some asymmetrical motor faults are well identified using for different types of rotating machine faults detection and
advanced signal processing techniques, such as high-resolution localization [8]–[14].
spectral analysis. This technique leads to a better interpretation Large electromachine systems are often equipped with me-
of the motor current spectra. In fact, experimental results clearly
illustrate that stator current high-resolution spectral analysis is chanical sensors, primarily vibration sensors based on prox-
very sensitive to induction motor faults modifying main spectral imity probes. Those, however, are delicate and expensive.
components, such as voltage unbalance and single-phasing effects. Moreover, in many situations, vibration monitoring methods
Index Terms— Broken bars, damaged bearings, fault detec- are utilized to detect the presence of incipient failure. How-
tion, induction motors, rotor eccentricity, shaft speed oscillation, ever, it has been suggested that stator current monitoring can
single-phasing effects, spectral analysis, stator current, unbal- provide the same indications without requiring access to the
anced voltage. motor [3]. Therefore, we have focused our research on the so-
called motor current signature analysis. This technique utilizes
I. INTRODUCTION results of spectral analysis of the stator current (precisely, the
supply current) of an induction motor to spot an existing or

I N GENERAL, condition monitoring schemes have concen-


trated on sensing specific failures modes in one of three
induction motor components: the stator, rotor, or bearings.
incipient failure of the motor or the drive system. It is shown
that the amount of information brought by the use of advanced
signal processing techniques, such as high-resolution spectral
Even though thermal and vibration monitoring have been uti- analysis, is lower than that deducible from the use of classical
lized for decades, most of the recent research has been directed spectral analysis property, but allows obtaining a characteristic
toward electrical monitoring of the motor with emphasis on spectral signature which can be easily distinguished from a
inspecting the stator current of the motor. In particular, a normal operating condition and then identified as a potential
large amount of research has been directed toward using the failure mode. Of particular interest are broken bars in the rotor
stator current spectrum to sense rotor faults associated with cage, rotor eccentricity, worn or damaged bearings, shaft speed
Manuscript received May 10, 1997; revised May 28, 1998. This work oscillation, and electrical-based faults (unbalanced voltage and
was supported by the French Concerted Research Program (PRC) on the single-phasing effects) [15]–[20]. Experimental investigations,
Diagnostics of Electrical Machines. Recommended by Associate Editor, for high-resolution analysis, have been carried out only for
M. Arefeen.
M. E. H. Benbouzid is with the University of Picardie “Jules Verne,” 80000 electrical-based faults detection and localization.
Amiens, France (e-mail: m.benbouzid@ieee.org).
M. Vieira and C. Theys are with the Laboratoire d’Informatique, Signaux II. STATOR CURRENT SIGNATURE ANALYSIS
et Systémes, CNRS & University of Nice “Sophia Antipolis,” 06041 Nice,
France. A current spectra contains potential fault information. Fre-
Publisher Item Identifier S 0885-8993(99)00282-3. quency components have been determined for each specified
0885–8993/99$10.00  1999 IEEE
BENBOUZID et al.: INDUCTION MOTORS’ FAULTS DETECTION AND LOCALIZATION 15

fault. These frequencies are derived from the physical con-


struction of the machine. It is important to note that, just as
in vibration analysis, as the fault progresses, its characteristic
spectral components continue to increase over time.

A. Eccentricity and Broken Bars


A broken bar may be distinguished from an asymmetry
by examining the harmonics sidebands. An asymmetry will
typically result in a smooth variation of air-gap flux density. It
has been shown that both rotating and nonrotating eccentricity
will give rise to current components at frequencies given by [2]

(1)

where is the electrical supply frequency, ,


the rotor bars number, the eccentricity order number,
the per unit (p.u.) slip, the number of pole pairs, and
the supply frequency harmonic rank.

B. Shaft Speed Oscillation


In the case of dynamic eccentricity that varies with rotor
position, the oscillation in the air-gap length causes variations Fig. 1. Four types of rolling-element bearing misalignment [5].
in the air-gap flux density. This, in turn, affects the induc-
tance of the machine producing stator current harmonics with
frequencies predicted by [3]

(2)

C. Rotor Asymmetry
It has been shown that when a rotor asymmetry is present,
the air-gap flux density will be perturbed and this perturbation
will rotate at shaft speed. The frequencies of the spectral
components in the air-gap flux density are given by [2]

(3)
Fig. 2. Ball-bearing dimensions.
D. Bearings Failure
Installation problems are often caused by improperly forcing with
the bearing onto the shaft or in the housing. This produces
physical damage in the form of brinelling or false brinelling of (5)
the raceways which leads to premature failure. Misalignment
of the bearing, which occurs in the four ways depicted in where is the number of balls, the mechanical rotor speed
Fig. 1, is also a common result of defective bearing installa- in hertz, the bearing pitch diameter, the ball diameter,
tion. and the contact angle of the balls on the races.
The relationship of the bearing vibration to the stator current
spectra can be determined by remembering that any air-gap III. STATOR CURRENT MONITORING SYSTEM
eccentricity produces anomalies in the air-gap flux density.
The stator current monitoring system contains the four
Since ball bearings support the rotor, any bearing defect will
processing sections illustrated in Fig. 3.
produce a radial motion between the rotor and stator of the
machine. The characteristic frequencies for ball bearings are
A. Sampler
based upon the bearing dimensions shown in Fig. 2. They are
given by [5], [21] The purpose of the sampler is to monitor a single phase
of induction motor current. This is accomplished by removing
(4) the 50-Hz excitation component through low-pass filtering and
16 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO. 1, JANUARY 1999

Fig. 3. Block diagram of single-phase stator current monitoring scheme.

sampling the resulting signal. The current flowing in single


phase of the induction motor is sensed by a current transformer
and sent to a 50-Hz notch filter, where the fundamental
component is reduced. The analog signal is then amplified and
low-pass filtered. The filtering removes the undesirable high-
frequency components that produce aliasing of the sampled
signal while the amplification maximizes the use of the analog-
to-digital (A/D) converter input range. The A/D converter
samples the filtered current signal at a predetermined sampling
rate that is an integer multiple of 50 Hz. This is continued over
a sampling period that is sufficient to achieve the required fast
Fourier transform (FFT).

B. Preprocessor
The preprocessor converts the sampled signal to the fre- Fig. 4. View of the experimental setup.
quency domain using an FFT algorithm. The spectrum gen-
erated by this transformation includes only the magnitude
information about each frequency component. Signal noise that eliminates those components that provide no useful failure
is present in the calculated spectrum is reduced by averaging information. The algorithm keeps only those components that
a predetermined number of generated spectra. This can be are of particular interest because they specify characteristic fre-
accomplished by using either spectra calculated from multiple quencies in the current spectrum that are known to be coupled
sample sets or spectra computed from multiple predetermined to particular motor faults. Since the slip is not constant during
sections (or windows) of a single large sample set. Because of normal operation, some of these components are bands in the
the frequency range of interest and the desired frequency res- spectrum where the width is determined by the maximum
olution, several thousand frequency components are generated variation in the slip of the motor.
by the processing section.
D. Postprocessor
C. Fault Detection Algorithm Since a fault is not a spurious event, but continues to
In order to reduce the large amount of spectral information degrade the motor, the postprocessor diagnoses the frequency
to a usable level, an algorithm, in fact, a frequency filter, components and then classifies them (for each specified fault).
BENBOUZID et al.: INDUCTION MOTORS’ FAULTS DETECTION AND LOCALIZATION 17

Fig. 5. Schematic view of the experimental setup.

Fig. 6. Full-loaded motor stator current power spectra. Fig. 7. Stator current power spectra around 50 Hz.

IV. EXPERIMENTAL RESULTS


To verify the generality of the presented considerations, and 8, more frequency information appears. In fact, one could
laboratory experiments were performed with an induction identify in Fig. 7 eccentricity harmonics at 26 and 74 Hz [see
motor. The investigated drive system is shown by Figs. 4 and (1)] and rotor asymmetry harmonics at 46 and 54 Hz [see
5. The stator current was sampled with a 1-kHz sampling rate (3)]. In Fig. 8, the shaft speed oscillation harmonic appears
and interfaced to a Pentium PC by a National Instruments data clearly at 246 Hz [see (2)]. When accurately analyzing the
acquisition board. All of the following stator current power power spectra, one could notice the absence of obvious bearing
spectra are based on a 4096-points FFT. failure.

A. Experiment 1 B. Experiment 2
The first experiment involved the drive system driving the In the second experiment, the stator voltages were unbal-
full load at 1444 rpm. The power spectra of Fig. 6 represents anced by adding a 0.2-p.u. resistance to one phase. The power
then, in our case, the supposed healthy motor. However, spectra of stator current is then shown by Fig. 9. One should
when zooming the power spectra, as illustrated by Figs. 7 notice the emergence of even harmonics.
18 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO. 1, JANUARY 1999

Fig. 8. Stator current power spectra for frequencies more than 200 Hz.
Fig. 11. Power spectrum of the instantaneous power for a full-loaded motor.

decreasing the distortions of the current waveform and the


spectrum noisiness as seen in Figs. 6, 9, and 10.
Therefore, in place of the stator current, some authors have
proposed the instantaneous power as a medium for the motor
signature analysis. It was shown that the amount of information
carried by the instantaneous power, which is the product of the
supply voltage and motor current, is higher than that deducible
from the current alone [22], [23]. Fig. 11 shows clearly the
differences with Fig. 7. In fact, all the fault harmonics are
translated into the frequency band 0–100 Hz. This constitutes
a great advantage because the fault harmonics domain is, in
this case, well bounded. However, as illustrated by Fig. 11, the
Fig. 9. Stator current power spectra in the case of stator voltage unbalance. power spectra is still quite noisy so as instantaneous power, at
this stage, does not bring important improvement. Therefore,
the stator current should be maintained as the main medium
for the motor signature analysis.
So, instead of using the stator current monitoring system
illustrated in Fig. 3, an alternative spectral method allows
gathering the preprocessor and the fault detection algorithm. In
fact, high-resolution spectral techniques can be used directly
from the sampled stator current waveform.

VI. HIGH-RESOLUTION SPECTRAL ANALYSIS


The classical spectral estimation techniques which have
been used are among the most robust ones, allowing com-
Fig. 10. Stator current power spectra in the case of stator open phase.
putationally efficient algorithms like FFT. However, a main
disadvantage of the classical spectral estimation is the impact
C. Experiment 3 of side-lobe leakage due to the inherent windowing of finite
The last experiment dealt with the analysis of a single-phase data sets. Window weighting allows mitigating the effect of
effect corresponding to a stator open phase. Fig. 10 illustrates side lobes at the expense of decreasing the spectral resolution
in this case the power spectra of the stator current. As with the which can be no better than the inverse of acquisition time
previous remark, we notice the emergence of even harmonics, (i.e., s Hz).
but with less amplitude. In order to improve the statistical stability of the spectral
These two cases of electrical-based faults (unbalanced volt- estimate, i.e., to minimize the estimate variance, pseudoensem-
age and single-phasing effects) have been quite easily detected, ble averaging by segmenting the data was introduced at the
compared to the healthy power spectra of Fig. 6, and classified price of further decreasing the resolution. So, tradeoffs among
by the postprocessor, thanks only to the even harmonics stability, resolution, and leakage suppression are necessary.
amplitude analysis. A class of spectral techniques based on an eigenanalysis of
the autocorrelation matrix has been promoted in the digital
V. DISCUSSIONS signal processing research literature. They may improve or
Generally, not denying the diagnostic value of classical maintain high resolution without sacrificing as much stability,
spectral analysis techniques, induction motors faults detection which allows keeping only the principal spectral components
via stator current signature analysis could be improved by of the signal and decreasing noise influence.
BENBOUZID et al.: INDUCTION MOTORS’ FAULTS DETECTION AND LOCALIZATION 19

A. Eigenanalysis-Based Frequency Estimators of eigenvalues exceeding a given threshold [24]. More gener-
To give the details of these estimators is not the purpose ally, a model order selection criterion can be used—the Akaike
of this paper (see [24] and [25] for details). The principle is information criterion for example.
just briefly reviewed. With regard to autocorrelation matrix order selection, we
The data are assumed to consist of -pure sinusoids in white would like the dimension of to be large so that the
majority of the eigenvalues are equal to However, doing
noise, a suitable model for the stator current
so runs the risk of statistically unstable eigenvalue estimates
and high-computational load [25].
(6)
B. Experimental Results
where is a data, is an additive complex white noise
To verify the generality of the presented estimators, the
sample of variance and and are, respectively,
above algorithms have been implemented to improve induction
amplitude, frequency, and phase of the th sinusoid.
machines faults detection and localization in the two following
These frequency estimation techniques are based on an
experiments.
eigenanalysis which divides the information in the autocor-
1) Experiment 1—Distinction of Three Close Frequencies
relation matrix , where is the matrix order, into two
and : Usually, a machine fault modifies the stator
vector subspaces—one a signal subspace and the other a noise
current spectrum of the healthy machine by changing the
subspace. The eigenvalues are
amplitude of some components already present in the spectra.
Some faults affect spectral components whose frequencies
Two well-known eigenanalysis-based frequency estimators
depend on the p.u. slip related itself to the load. However,
have been used: MUSIC and ROOT-MUSIC. It should be for low-slip values, the frequency resolution of the classical
noticed that ROOT-MUSIC has been well known for detecting spectral estimation technique could not be sufficient for distin-
sinusoids in Gaussian white noise. guishing close frequencies, for example, the electrical supply
1) MUltiple SIgnal Classification (MUSIC) Estimator: This frequency with
estimator is based on the property that the noise subspace In fact, the classical spectral estimation technique used
eigenvectors of a Toeplitz autocorrelation matrix are or- in Figs. 6 and 10 is the Welch’s periodogram with a Han-
thogonal to the signal vectors , where ning window calculated on the data length (4096 samples).
(7) The minimal frequency resolution is s Hz,
evaluated by the mainlobe frequency bandwidth of a Fourier
For , the sinusoidal frequencies are estimated as the
transform of the Hanning window. However, it is important
peaks of the pseudospectrum given by
to notice that this value is available for close frequencies
of the same amplitude. In our case, the resolution will be
degraded since the considered frequencies are not of the same
amplitudes. In practice, for , this classical technique
is unable to distinguish the frequencies and
Use of eigenanalysis-based frequency estimators will be
(8)
useful even if a prefiltering is necessary to select the frequen-
cies under interest. In our case, to select the frequencies to
Theoretically, when , the th sinusoidal frequency
analyze, a solution is to use a pass-band filter centered on
In practice, the obtained spectrum have largest
For the load at 1444 rpm, stator current power spectrum
peaks at the principal frequency locations. obtained after filtering is given in Fig. 12. It can be noticed
2) ROOT-MUSIC Estimator: Setting to in that the three principal spectral components are clear and
(8), such as the polynomial is correspond to the previously defined frequencies (i.e., 46, 50,
and 54 Hz) in Fig. 6. In this case, the high-resolution property
(9) of the eigenanalysis-based frequency estimators cannot be
underscored.
The polynomial coefficient is the sum of the matrix 2) Experiment 2—Stator Voltage Unbalance Underscoring:
elements on the th diagonal. In the case of stator voltage unbalance, one of the principal
The ROOT-MUSIC algorithm determines roots of and spectral components modified by the electric fault is the third
then selects the roots closest to the unit circle, corresponding harmonic of supply frequency (i.e., 150 Hz) whose amplitude
with increases in a significant way whatever the load.
These estimators require the determination of the number The two principal spectral components of the stator current
of frequencies spectrum are the first and the fifth harmonics (50 and 250 Hz)
3) Choice of and : These estimators require the a pri- for the healthy motor (Fig. 6) and the first and third harmonics
ori knowledge of the number of frequencies and the auto- (50 and 150 Hz) for the stator voltage unbalance (Fig. 9). The
correlation matrix order MUSIC algorithm has been applied for each case with
If is unknown in the case of eigenanalysis-based tech- and , and results are given, respectively, in Figs. 13
niques, the number of sinusoids will be chosen as the number and 14. It should be noticed, contrary to Figs. 6 and 9, that
20 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO. 1, JANUARY 1999

Fig. 12. Power spectrum after pass-band filter at 1444 rpm.

Fig. 13. MUSIC frequency estimate for the healthy machine.

Fig. 14. MUSIC frequency estimate for the stator voltage unbalance.
BENBOUZID et al.: INDUCTION MOTORS’ FAULTS DETECTION AND LOCALIZATION 21

spectral signatures illustrated by Figs. 13 and 14 can be easily ACKNOWLEDGMENT


used for the detection and identification of a potential voltage The authors thank A. Berrehilli and N. Farid for setting up
unbalance. the experimental system. The valuable comments made by the
The frequency values obtained with the ROOT-MUSIC reviewers are also acknowledged.
estimator for and are 50.11 and 250.43 Hz
for the healthy motor and 50.08 and 150.22 Hz for the stator
voltage unbalance. REFERENCES

C. Discussions [1] J. R. Cameron, W. T. Thomson, and A. B. Dow, “Vibration and current


monitoring for detecting airgap eccentricity in large induction motors,”
With regard to these results, MUSIC and ROOT-MUSIC Proc. Inst. Elect. Eng., vol. 133, pt. B, pp. 155–163, May 1986.
[2] G. B. Kliman and J. Stein, “Induction motor fault detection via passive
methods allow keeping only the main frequencies without current monitoring,” in Proc. 1990 Int. Conf. Electrical Machines, vol.
other spectral information. This is a first step to an automated 1, Cambridge, MA, pp. 13–17.
diagnostic system. [3] , “Methods of motor current signature analysis,” Elec. Mach.
Power Syst., vol. 20, pp. 463–474, Sept. 1992.
The main advantages of the high-resolution spectral analysis [4] R. R. Schoen, B. K. Lin, T. G. Habetler, J. H. Schlag, and S. Farag,
can be underscored in the last experiments: “An unsupervised, on-line system for induction motor fault detection
using stator current monitoring,” IEEE Trans. Ind. Applicat., vol. 31,
1) taking the stator current model into consideration; pp. 1280–1286, Nov./Dec. 1995.
2) noise influence reduction; [5] R. R. Schoen, T. G. Habetler, F. Kamran, and R. G. Bartheld, “Motor
3) extraction of the useful information (prefiltering and bearing damage detection using stator current monitoring,” IEEE Trans.
Ind. Applicat., vol. 31, pp. 1274–1279, Nov./Dec. 1995.
choice of the number of sinusoids). [6] T. W. S. Chow and G. Fei, “Three phase induction machines asym-
Even if the primary motivation for eigenanalysis-based metrical faults identification using bispectrum,” IEEE Trans. Energy
Conversion, vol. 10, pp. 688–693, Dec. 1995.
frequency estimators, i.e., the high-resolution property, has not [7] F. Filippetti, G. Franceschini, C. Tassoni, and P. Vas, “A fuzzy logic
been underscored in these experiments, this property can be approach to on-line induction motor diagnostics based on stator current
shown for lower slip values. In comparison with the classical monitoring,” in Proc. 1995 IEEE Int. Power Tech. Conf., vol. EMD,
Stockholm, Sweden, pp. 156–161.
spectral analysis techniques, the eigenanalysis-based spectral [8] J. Penman, M. N. Dey, A. J. Tait, and W. E. Bryan, “Condition
methods are computationally intensive. The approximate rela- monitoring of electrical drives,” Proc. Inst. Elect. Eng., vol. 133, pt.
B, pp. 142–148, May 1986.
tive computational complexity of MUSIC and ROOT-MUSIC [9] D. J. T. Siyambalapitiya, P. G. McLaren, and P. P. Acarnley, “A rotor
algorithms is , where is the number of samples, whereas condition monitor for squirrel-cage induction machines,” IEEE Trans.
classical spectral technique one is [24]. Ind. Applicat., vol. 23, pp. 334–340, Mar./Apr. 1987.
[10] P. J. Tavner, K. K. Amin, and C. Hargis, “An electrical technique
for monitoring induction motor cages,” in Proc. 1987 IEE Int. Conf.
VII. CONCLUSIONS Electrical Machines and Drives, London, U.K., pp. 43–46.
[11] D. Leith, N. D. Deans, and I. D. Stewart, “Condition monitoring of
This paper has taken the initial step to investigate the electrical machines using real-time expert system,” in Proc. 1988 Int.
efficiency of current monitoring for diagnostic purposes. The Conf. Electrical Machines, vol. 3, Pisa, Italy, pp. 297–302.
[12] P. J. Tavner, “Condition monitoring—The way ahead for large electrical
effects of stator current spectrum have been described and machines,” in Proc. 1989 IEE Int. Conf. Electrical Machines and Drives,
the related frequencies determined. In the present investiga- London, U.K., pp. 159–162.
tion, the frequency signature of some asymmetrical motor [13] D. J. T. Siyambalapitiya and P. G. McLaren, “Reliability improvement
and economic benefits of on-line monitoring system for large induc-
faults have been well identified using advanced signal pro- tion machines,” IEEE Trans. Ind. Applicat., vol. 26, pp. 1018–1025,
cessing techniques, such as high-resolution spectral analysis. July/Aug. 1990.
Experimental results have demonstrated that the stator current [14] S. Chen, E. Zhong, and T. A. Lipo, “A new approach to motor condition
monitoring in induction motor drives,” IEEE Trans. Ind. Applicat., vol.
high-resolution spectral analysis, proposed as a medium for 30, pp. 905–911, July/Aug. 1994.
induction motors faults detection, has definite advantages [15] W. Deleroi, “Broken bars in squirrel cage rotor of an induction motor,
over the traditionally used FFT spectral analysis, and, more Part 1: Description by superimposed fault currents,” (in German), Arch.
für Elektrotechnik, vol. 67, pp. 91–99, 1984.
generally, this technique will be useful in all faults modifying [16] R. Belmans, A. Vandenput, and W. Geysen, “Influence of torsional
main spectral components. vibrations on lateral oscillations of induction motors rotors,” IEEE
Extensive experimental studies are necessary to full assess Trans. Power Apparatus Syst., vol. 104, no. 7, pp. 1832–1837, 1985.
[17] IAS Motor Reliability Working Group, “Report of large motor reliability
usefulness of the proposed technique for the preventive main- survey of industrial and commercial installations, Part I,” IEEE Trans.
tenance diagnostics and failure prevention in drive systems Ind. Applicat., vol. 21, pp. 853–864, July/Aug. 1985.
[18] , “Report of large motor reliability survey of industrial and
with induction motors. commercial installations, Part II,” IEEE Trans. Ind. Applicat., vol. 21,
pp. 865–872, July/Aug. 1985.
APPENDIX [19] , “Report of large motor reliability survey of industrial and
commercial installations, Part III,” IEEE Trans. Ind. Applicat., vol. 23,
RATED PARAMETERS OF THE MACHINE UNDER TEST pp. 153–158, Jan./Feb. 1987.
[20] A. H. Bonnett and G. C. Soukup, “Cause and analysis of stator and rotor
failures in three-phase squirrel-cage induction motors,” IEEE Trans. Ind.
Power 4 kW Applicat., vol. 28, pp. 921–937, July/Aug. 1992.
Frequency 50 Hz [21] R. L. Schiltz, “Forcing frequency identification of rolling element
Voltage 220/380 V bearings,” Int. J. Sound Vibration, pp. 16–19, May 1990.
[22] R. Maier, “Protection of squirrel-cage induction motor utilizing instan-
Current 15/8.6 A taneous power and phase information,” IEEE Trans. Ind. Applicat., vol.
Speed 1440 rpm 28, pp. 376–380, Mar./Apr. 1992.
Pole pair 2 [23] S. F. Legowski, A. H. M. S. Ula, and A. M. Trzynadlowski, “Instan-
taneous power as a medium for the signature analysis of induction
22 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO. 1, JANUARY 1999

motors,” IEEE Trans. Ind. Applicat., vol. 32, pp. 904–909, July/Aug. Michelle Vieira was born in Belfort, France, in
1996. 1972. She received the B.Sc. degree in telecommu-
[24] S. L. Marple, Digital Spectral Analysis with Applications. Englewood nications in 1994 from the University of Nancy-
Cliffs, NJ: Prentice-Hall, 1987. Metz, France, and the M.Sc. degree in signal pro-
[25] S. M. Kay, Modern Spectral Estimation. Englewood Cliffs, NJ: cessing in 1995 from the University of Nice, Sophia
Prentice-Hall, 1988. Antipolis, France. She is currently working toward
the Ph.D. degree in monitoring and diagnosis of
induction motors via signal processing at the Uni-
versity of Nice.
Mohamed El Hachemi Benbouzid (S’92–M’94)
was born in Batna, Algeria, in 1968. He received
the B.Sc. degree in electrical engineering in 1990
from the University of Batna, Batna, and the M.Sc.
and Ph.D. degrees in electrical and computer en-
gineering in 1991 and 1994, respectively, from the Céline Theys was born in Paris, France, in 1967.
National Polytechnic Institute, Grenoble, France. She received the B.Sc., M.Sc., and Ph.D. degrees
After graduation, he joined the University of in electrical engineering and digital signal process-
Picardie “Jules Verne,” Amiens, France, where he is ing from the University of Nice, Sophia Antipolis,
an Associate Professor of Electrical and Computer France, in 1989, 1990, and 1993, respectively.
Engineering. His current research interests include She is currently an Associate Professor at the
electric machines and drives, electromagnetics computational, and electro- University of Nice. Her research interests include
mechanical actuation as well as techniques for energy savings. He is leading digital signal processing for the detection of abrupt
a research program on induction machine drives monitoring and diagnostics changes and fault diagnosis.
for the French Picardie Region. He has published more than 40 technical
papers including 15 refereed publications in journals.
Dr. Benbouzid is a Member of the French Society of Electrical Engineers
SEE and is active in PES and IAS-IEEE.

You might also like