This article was downloaded by: [McGill University Library]

On: 23 December 2012, At: 06:28

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Electric Machines & Power Systems

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uemp19

Methods of Motor Current Signature Analysis

a b
G. B. Kliman & J. Stein
a
GE Research and Development Center, P.O. Box 8, Schenectady, New york, 12301, USA
b
Electric Power Research Institute, 3412 Hiliview Avenue, Palo Alto., California, 94304, USA
Version of record first published: 06 Feb 2007.

To cite this article: G. B. Kliman & J. Stein (1992): Methods of Motor Current Signature Analysis, Electric Machines & Power
Systems, 20:5, 463-474

To link to this article: http://dx.doi.org/10.1080/07313569208909609

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 METHODS OF MOTOR CURRENT
SIGNATURE ANALYSIS

GE Research and Development Center

PO. Box 8
Schenectady. New York 1230 1, USA
J. STEIN
Electric Power Research lnstitufe
34 12 Hillview Avenue
Palo Alto, California 94304, USA
Downloaded by [McGill University Library] at 06:28 23 December 2012

ABSTRACT
Recently a technique for monitoring and diagnosing mechanical prohlems, associated with
rotating machines driven by elecuic motors, has been proposed and is now being offered hy
several commercial suppliers. This technique, known as "Motor Current Signature Analysis"
or MCSA, seeks to apply much of the long experience in vibration signatun analysis to the
analysis of motor current in effect using the rnotor as a sensor akin to an accelerometer. This
paper exploms some of the history of Be technique, presents seven1 examples. and demon-
strates a first order approach to the theory with associated cautions.
1. INTRODUCTION
A technique for diagnosing problems in mechanical loads driven by rolating electrical ma-
chines has been achieving some recognition in the marketplace recently. The basic idea.
explored in Section 4, is that variations in load torque or speed will cause corresponding
changes in the motor current, which may then he interpreted to draw conclusions as to the
health of the load.
Thus, a number of problems arise, which so far have not been mated in a thorough manner.
One such problem is the precise characterization of the motor as a sensor akin, for example.
to an accelerometer. Unlike an accelerometer the motor is convening energy and this is its
primary purpose for which the design has been optimized. We are then hced with a series of
major prohlems not present in an accelerometer. One of these problems is that very small
signals returning from the load must be detected in the presence of a large energy conversion
component. The common technique is demodulation to remove the energy conversion com-
ponent hut this must he done correctly. The motor itself introduces a wide variety of
signatures in its healthy, as well as in its faulted condition, and these must be distinguished
from load signals.
The background for this application derives from several areas m o n g which are the very
extensive experience of vibration and signature analysis in mechanical systems, modem
control theory, and motor fault analysis. This paper will deal with the background in motor
fault analysis, both for its own value and as a requirement for the separation of rnotor signa-
tures from those of the load. A first approach to a basic analysis of the motor as a sensor is
presented in order to clarify the spectra obtained using commercial demodulation methods
and spectrum analyzers.

Electric Machines and Power Systems. 20463-474,1992 463

 G. 6. KLIMAN A N D J. STEIN

2. MOTOR FAULTS
2.1. Rotor
There is, by now. an extensive academic and commercial development of theory and instru-
ments to detect problems in the rotor of an induction motor (Kliman et al. 1988, Tavner et al.
1987. Gaydon and Tucker 1981. Williamson and Smith 1982. Deleroi 1984. Mavridis 1963.
Kerzenbaum and Landy 1984, Kliman el al. 1990). All of the insuuments now on the market
depend on the detection of particular components of the current spectrum that are directly
related to rotor asymmetries. It has been shown that when a rotor asymmetry is present the
airgap flux will be perturbed and this perturbation will rotate at shaft speed. The frequencies
of the spectral components in the airgap flux are given by (Deleroi 1984):
Downloaded by [McGill University Library] at 06:28 23 December 2012

where fi = line frequency

S = per unit slip
p = number of poles
k = 1.2.3.4 ...
Because of the normal winding configuration only those for which W(pR) = I. 5.7. 11, etc..
will appear in the cunent with significant amplitude.
An example of a large motor with a small rotor asymmetry is shown in Figure 1. Also present
is a torsional oscillation, originating in the load, which could cause a false alarm in a broken
bar detector. In this case, the system measures the slip frequency quite precisely and identi-
fies (dashed venical line) the proper broken bar frequency. thus avoiding the false alann.
While this effect may be considered a nuisance in a broken bar detector it is, in facS the basis
of MCSA.

-- . . .. .., ..

FIGURE I. Example of a potential hroken har detector false alarm.

 METHODS OF MOTOR CURRENT SIGNATURE ANALYSIS 465

Rotor asymmetry, resulting from a faulty bearing, rotor ellipticity, misalignment of the shaft
with the cage, magnetic anisouopy, etc.. result in similar spectra and must be distinguished
from bar breaks and other load signals. Two approaches have been taken. One approach
(Kliman et al. 1988, Kliman et al. 1990) is to examine the relative behavior of the fundamen-
lal and higher harmonic sidebands. A rotating asymmetry will yield a line frequency
sideband very much like that for a broken rotor bar. A broken bar may be distinguished from
an asymmetry by examining the magnitudes of the harmonic sidebands. An asymmetry will
typically result in a smooth variation of airgap flux density, i.e.. the upper harmonic content
will be low. In contrast localized effects, such as a broken bar. result in large upper harmon-
ics.
The second approach to distinguish broken bars from asymmetries relies on an examination
Downloaded by [McGill University Library] at 06:28 23 December 2012

of sidebands of the bar passing frequency. It bas been shown (Cameron el al. 1987) that both
routing and nonrotating eccentricity will give rise to cumnt components at frequencies
given by:

where R = number of rotor slots

k = 1 , 2 , 3 , 4 , 5 ...
n,=l.3.5,7.9 ...
nd = 0 sutic eccenlricity
nd = 1. 2. 3 . 4 ...dynamic eccentricity

An illustration of "bar passing" spectral components of the current is shown in Figure 2 for a
4-pole motor with 51-rotor slots with and without a rotating (dynamic) eccentricity (Cameron
et al. 1987). Note that the frequencies of interest are now much higher and could go above 6
kHz on large two-pole motors. Rotating (dynamic) eccentricity, such as a rotating airgap
variation, will cause rotor currents and could be confused with a broken bar. A nonrotating
(static) asymmetry, such as an offset bearing, will not give rise to rotor currents.
Three companies now market broken bar detectors in the US.:GE. Entek and CSI (Compu-
tational Systems. Inc.). All three rely on evaluating the magnitude of the lower sideband of
the line frequency. Significant differences between them arise in the methods of finding the
proper sideband, interpreting its magnitude, evaluating asymmetry, degree of automation.
and specific hardware.
2.2. Stator Faults
The most common method of detecting faults is the "ground-fault detector." Kirchoffs
cumnt law dictates that the instantaneous sum of the currents in 3 thw-phase system.
without neutral connection, must always be zero. Hence a cumnt tnnsfomer arranged with
all three leads, passing through the window must always indicate zero unless there is an
additional return path such as ground. Ground fault detectors can be made sensitive and rapid
enough to be used for protection of personnel against shock hazards.
A technique long used for other winding faults is to trip on the presence of negative sequence
currents as an indication of a major fault. Detailed analyses of partial phase and phase-to-
phase faults have been done by an equivalent circuit method (Williamson and Mirzoian
1984). Figure 3 shows the comparison of prediction and test lor a psnial phase fault. It may
 466 G. B. KLIMAN AND J. STEIN

he observed that the negative sequence components m prominent and may become the basis
for an incipient fault detector. A complicating factor in such detectors is that a faulted (hro-
ken bar) rotor will also give rise to significant negative sequence components in the
current (Williamson and Smith 1982). Stator faults may also result in the addition of airgap
flux harmonics which, at very high resolution, will he detectable in the stator current (Pen-
man and Tavner 1987).
An approach to provide an insuument for detecting ground- and partial-phase faults was
patented (Hentschel 1974). This instrument depends on detection of the phase current unbal-
ance that must occur when there is a ground or phase fault in the motor. It is thus closely
related to the negative sequence current methods.
Rccently this approach has been further refined at Pennsylvania State University (Kohler et
Downloaded by [McGill University Library] at 06:28 23 December 2012

al. 1990). to increase the sensitivity to the point were it may he possible to evaluate insulation
dctcrioration. An illustration of the sensitivity (with simulated deterioration) is shown in
Figure 4.

FIGURE 3. Predicted and measured currents

for a phase-to-phase short circuit (Williamson
and Mirzoian 1984) (10 poles).

FIGURE 2. Example of the effect of a IIO.ll"c-.IOY."S. CV".*"I. ..".

dynamic asymmetry in the current FIGURE 4. Change in the negative-sequence
speclrum near the bar passing predictor as a function of deterioration severity
frequencies. Current zoom spectrum (fault power) for stator-winding deterioration not
with 51-slot rotor (4 poles. I I kW) involving ground (simulated) (Kohler et al.
(Cameron et al. 1987). 1990).
 METHODS OF MOTOR CURRENT SIGNATURE ANALYSIS 467

2.3. Model Fitting

A much different approach to passive methods utilizes the tools of optimal control theory.
The published application (Cho et al. 1989). developed at MIT, was directed to rotor faults
although it could also have been applied to other aspects of the motor.
The essence of this method is the fitting of a model, e.g.. the standard equivalent circuit. A
series of measurements are made at various voltages or loads. Estimation techniques are then
used to find the best fit of the model to the measurements. In the case of rotor faults a rotor
resistance temperature compensation estimator is also utilized. Figure 5 shows the results of
one series of trials on a small motor. Three (presumably) identical rotors were tested in the
same stator. Rotors #I and #3 were left in the original condition but rotor #2 had one bar
completely broken. It was found that without the thermal compensation estimator, the results
Downloaded by [McGill University Library] at 06:28 23 December 2012

scattered over a range wider than that shown. This technique has received attention in regard
to processes and plants on a larger scale however many of the ideas are similar (Isermann
1984).
2.4. INTEGRATION O F METHODS
Effons are now k i n g made to integrate the various fault indications into a unified softwart:
structure using the techniques of artificial intelligence (AI). One of the first of these is a
system for diagnosing h e health and problems of a large steam turbine generator (Tavner et
al. 1986). The emphasis in such developments are primarily thermal and relate to the inter-
pretation of the many and different kinds of thermal sensing systems in a large synchronous
generator.
More recently work has been under way to produce an integrated motor fault detection
system aimed, primarily at the induction motor. This study seeks to address the ambiguities
of the individual techniques here described. For example, one set of the rules in the expert
system being developed compares negative sequence components resulting from stator faults
and those from rotor faults with the aid of spectrum analysis to resolve ambiguity (Gentile et
al. 1990).
3. LOAD FAULT SENSING
3.1. Bearings
In the c s e of a pump driven by an induction motor. the line current spectra when the pump
was in presumably good condition and when a deliberately damaged hearing was replaced
into the pump, were compared (Steele el al. 1982). No attempt was made at that time, to
identify specmm lines with particular physical phenomena in the motor or pump. It was
concluded that the quasi-random nature of the effeets of hall-hearing faults made spectrum
analysis unsuitable as a tool. In fact, while it was clear that the motor cumnt spectra were
different for faulted and healthy hearings characteriution of the features was elusive. Instead

FIGURE 5. Estimated rotor resistance for motors

with faulted (#2) and unfaulted (#I and #3) rotors
(Cho et aL 1989).
 468 G. 8. KLIMAN AND J. STEIN

a "real-time" approach was taken in a "pulse-over-threshold" method. Figure 6 shows the

results of a "pulse-over-threshold analysis of the filtered drive motor current signal for
various induced bearing faults in the driven pump (Steele et al. 1982). The correlations are
rather sviking hut there was appmntly no further testing performed either in the laboratory
or in the field to exploit the method.
3.2. Valves
More recently another approach to passive current monitoring has been developed at Oak
Ridge National Laboratory (Haynes 1989) using time domain analysis, which has also been
offered by Bechtel C o p (Bechtel 1989), in addition to the more traditional frequency do-
main analysis. This approach was developed specifically for the diagnosis of motor-operated
valves (MOV).
Downloaded by [McGill University Library] at 06:28 23 December 2012

Figure 7 is an example of the identification of specific events in the operation of 3 MOV by a

closeexamination of the average motor current as a function of time. On the other hand.
frequency domain analysis displays quile different characteristics for the MOV driven by an
induction motor, as illustrated in Figure 8. Notice that in this case, care has been taken to
associate spectnrm lines with particular physical phenomena in the geared valve. which is the
motor load.
While the major feamres, such as motor speed and worm gear tooth meshing (WGTM)
frequency wcre present and identifiable. the origin and nature of the WGTM side hands were
unclear. It was known, however from experiments at ORNL, for example, that these side
bands would change as the packing friction was changed. The motor speed, sometimes
referred to as the "MCSA motor speed" has a specific origin which will ht: made clear in
Section 4.2.

, -mGWme*
Ma- 0 0 . 8.0 I., 10 I, 0 IS

FIGURE 6. Results of a pulse-over-threshold FIGURE 7. Typical transient features seen

analysis of the filtered current of a motor at the beginning of the close-to-open stroke
driving a pump having various defective of an MOV (Haynes 1989) (4 hp. 4 pole).
bearings (Steele et al. 1982) (2.6 kW. 4 pole).
 METHODS OF MOTOR CURRENT SIGNATURE ANALYSIS 469

Note that Figure 8 is not the raw-current spectrum bul has first been demodulated, assuming
the 60 Hz line frequency as the carrier, by special circuitry developed at ORNL (Smith et al.
1990). The features of the demodulated signal spectrum arise both from the motor itself and
from the load and may be identified by the use of Equations (1) and (6) in Section 4.1.
3.3. Compressors
A related technique of motor current signature analysis directed specifically lo study the
behavior of hermetic rotary compressors has been developed independently at GE. In such
compressors the high- and low-pressure regions are separated by a vane in sliding contact
with a metal ring being driven off axis by a 2-pole induction motor. Figure 9 is an example of
how the free rotation of the ring is manifested in the current. The other prominent local
features are the results of asymmetry in the rotor and the cyclic variation of the motor torque
due to the compressor.
Downloaded by [McGill University Library] at 06:28 23 December 2012

FIGURE 8. MOVZOO motor current spectrum

analysis (close-to-open suoke. 4 hp, 4 pole)
(Haynes 1989).

FIGURE 9. Typical current spectrum for a properly operating rotary com-

pressor and drive motor in the vicinity of line frequency (0.25 hp, 2 pole).
 G. 0. KLIMAN AND J. STEIN

4. SHAFT SPEED OSCILLATION

4.1. Spectra
To predict the current spectra as a result of speed oscillation we use the steady-slate equiva-
lent circuit for simplicity realizing that it is a gross approximation for the kind of effects
under examination here. It is assumed that the motor is operating in the sinusoidal steady
state at l i e frequency hut with a small, low-frequency torque and speed variation applied to
the shaft by the load. In order to display the major effects, the equivalent circuit is simplified
to include only the slip dependent rotor resistance. This assumption was checked by imple-
menting a simulation using a full d-q axis representation of the motor.
The shaft speed of the motor is assumed to consist of an average and an oscillating compo-
nent given by Eq. (3), and the line voltage is assumed to be a pure sinusoid given by Eq. (4):
Downloaded by [McGill University Library] at 06:28 23 December 2012

the line current, i s . is then given by Eq. (5):

i,(t)=(VI R,).(N, - N o ) / N;sin(2@1)+

(VI R,).(N, I No).sin(21rf,t)~Mksin(2~kjct)
k

where No = mechanical synchronous speed (rpm)

R, = rotor resistance
N,,, = unperturbed shaft speed (rpm)
Mk = fourier components of the speed fluctuation
jc = fundamenwl frequency of the speed fluctuation
V = voltage amplitude
Notice that Eq. (5) consists of a line frequency component that is just what it would be
present if there were no speed variation plus<.asymmetrical series of sidebands reflecting the
shape of the shaft-speed variation -as if th+otor were acting as an amplitude modulator
with line frequency as the carrier, and the shaft speed as the modulating signal.
In the special case of the fundamental frequency,jc, of the modulation being the shaft fre-
quency ilself, i.e., thc load torque varies in synchronism with shaft position, the frequencies
of the harmonics may he expressed more simply in terms of the line frequency and slip alone.

4.2. Interpretation of Spectra

Table 1 compares the frequencies (Eq. 6) for 2- and Cpole motors with the "hroken bar"
frequencies of Eq. 1. First, note that in a 4-pole motor there are two components very near to
shaft frequency. Second, note that there will always he a component at slip frequency but it
 METHODS OF MOTOR CURRENT SIGNATURE ANALYSIS

TABLE I. Comparison of broken bar current spectrum and shaft

oscillation current spectra

will he due to higher harmonics as the number of poles in the motor is increased. Third, for
2-pole motors. the shaft oscillation harmonics will be identical lo those for broken ban with
Downloaded by [McGill University Library] at 06:28 23 December 2012

the exception of an interchange of line and rotor frequency components.

Figure 10 is an example of the expected current spectrum resulting from a shaft speed oscil-
lation in a typical four-pole motor. It has been assumed that the spectrum of the shaft speed
oscillation is some decreasing function of frequency and is synchronous to the shaft so that
Eq. 6 applies. It must he pointed out that in addition to the usual nature of shaft toque penur-
hation the inertia of the motor will result in an approximate first order low pass filter with a
critical frequency of only a few hertz, funher attenuating high-frequency components.
It is apparent, from Table 1, that there is an opponunity for confusion between rotor asym-
metry signals and shaft oscillation signals. All small motors will exhibit rotor asymmetry
signals resulting from casting flaws such as porosity, blow holes, or variable lamination
contact. A large bar wound motor will usually but not always have smaller asymmetries in
their healthy swte. Not all of the airgap frequencies of Table I will appear in the current with
significant amplitude as only those harmonics closely matching the pole pitch will result in
significant swtor voltage. This characteristic effectively attenuates all hut those near the usual
Is', 5". 7". etc., harmonics of the motor current. However if the dynamic range of detection
is sufficiently large they may be found.
Figure 10 illusvates the effect where it has been assumed that most of the rotor asymmetry
harmonics have been attenuated except, of course, for the one near 60 Hz. Shaft oscillation
harmonics are identified by the symbol SO-k where k is the order of the original torque speed
harmonic. Similarly the rotor asymmetry harmonics are denoted hy BB-k. A relatively strong
shaft speed oscillation has also been assumed. Note, however, that if very small shaft speed
or torque signals are involved the rotor asymmetry signals may distoR the pattern. Even here
the sideband of 60 Hz is severely affected. The shaft oscillation harmonic SO-p12 always
appears alone.
Figure I I shows this same spectrum as (ideally) inverted by demodulation. The normally
seen asymmetry sideband now appears at twice slip frequency (not slip frequency as swted in
some commercial brochures). The "MCSA speed" is now identified as the fundamental of a
synchronous shaft oscillation (or it may result from a severe rotor asymmetry such as broken
cage bars). It is worth noting that if one is dealing with a large. bar-wound machine of good
quality and it is driving a constant toque, high inenia load, such as a large fan, the "MCSA
speed" may not appear at all at an attainable measurement dynamic range.
 G. B. KLIMAN AND J. STEIN

FIGURE 10. Typical current spectrum: 4- FIGURE I I. Typical demodulated current

pole motor. specvum for the 4-pole motor of Figure 10.
Downloaded by [McGill University Library] at 06:28 23 December 2012

5. CONCLUSION
Several examples of passive current monitoring techniques have been described. The number
of papers being published and companies getting into business indicate that this field is
beginning lo take off, and that such techniques will be routinely applied on a large scale.
In this paper not all motor diagnostic techniques were discussed nor were all the possible
signatures that both healthy and faulted moton may introduce. Also not discussed were the
issues of AID conversion, demodulation and other digital signal processing techniques to
deal with the very large dynamic range of these motor current signals. There is much to do
before MCSA can be regarded as a high confidence sensor system in all areas. Nonetheless
MCSA is destined to hecome a very useful, economic method for monitoring and diagnosing
motor loads and these issues will have to he addressed.
6. ACKNOWLEDGMENT
The authors wish to thank GE Corporate Research and Development, and the Elecuic Power
Research Institute for the use of their facilities in the preparation of this paper.
7. REFERENCES
1. "Bechtel-KWU Alliance Motor Operated Valve (MOV) Services Presentation."
Bechtel. Gaithersburg, MD. August 1989.
2. Cameron JR,Thomson WT,and Dow AB: "On-Line Current Monitoring of Induction
Motors - A Method for Calculating the Level of Airgap Eccentricity." Conf on Elec-
IricMachines nndDrives, IEE Publication No. 282, pp. 173-178, 1987.
3. Cho KR. Lang IH,and Umans SD: "Detection of Broken Rotor Bars in Induction
Motors Using State and Parameter Estimation." IEEE-IAS, Conf Proc. IAS Annual
Meeting. San Diego. OcL 1989.
4. Deleroi W: "Broken Bar in Squirrel Cage Rotor of an Induction Motor, Pan I: De-
scription by Superimposed Fault Currents," Arch. fur Elektrotechnik. Val. 67.
pp. 91-99. 1984.
5. Gaydon BG and Tucker CFI: "Detection of Rotor Defects in Induction Motors Using
Minicomputer," CEGB Report SUSSD/RN/80/074. June 198 1.
 METHODS OF MOTOR CURRENT SIGNATURE ANALYSIS 473

Gentile G et al.: "Analysis Approach of Induction Motor Stator Faults to On-Line

Diagnostics." ICEM90, Cambridge, MA, 1990, pp. 491-497.
Gonzalez Al and Lowenfeld S: "On-Line Diagnosis of Turbine-Generators Using
Artificial Intelligence." IEEE Trans. on Energy Conversion, Vol. E C - I . No. 2, pp. 68-
74. June 1986.
Haynes HD: "Aging and Service Wear of Electric Motor Operated Valves Used in
Engineered Safety-Feature Systems of Nuclear Power Plants." ORNL ReporL
NUREGICR-4234,ORNL-6170N2. Vol. 2. Aug. 1989.
Hentschel M: "Circuit Arrangement for Recognition of a Defect Caused by Phase
Breakdown or Ground ShoR in the Current Supply of a Three-Phase Motor." U. S.
Downloaded by [McGill University Library] at 06:28 23 December 2012

Patent 3.824.430. July 1974.

Isermann R: "Process Fault Detection Based on Modeling and Estimation Methods - A
Survey." Auromatica, Vol. 20, No. 4, pp.387-404, 1984.
Kerzenbaum I and Landy CF: '"The Existence of Large lnterbar Cumnts in Three-
Phase Squirrel Cage Motors With Rotor-Bar andlor End Ring Faults," IEEE. PAS-103,
No. 7, July 1984.
Kliman GB et al.: "Noninvasive Detection of Broken Rotor Bars in Operating Induc-
tion Motors," IEEE-PES, Trans. EC, Vol. 3, No. 4, pp. 873-879, Dec. 1988.
Kliman GB et al.: '"The Detection Broken Ban in Motors," EPRI GS-6589-L Project
2331-1, January 1990.
Kohler JL, Tmtt FC. and Santa Cmz C: "Incipient Failure Prediction for Electrical
Machines.'' Proc. 44th Meeting of the Motor Failure Prevention Group (MFPG).
Virginia Beach, VA, pp. 97-109. April 1990.
Mavridis D: "On the Performance of Three-Phase Squirrel Cage Induction Motors
During the Presence of Bar or Ring Breaks," Ph.D. thesis, Univ. of Hanover (Ger-
many). 1963.
Penman J and Tavner PJ:ConditionMonitoring of Electrical Machines, Wiley & Sons.
New York, 1987.
Smith SF, Castlebemy KN, and Nowlin CH: "Machine Monitoring Via Motor-Current
Demodulation Techniques," Proc. 44th Meeting of the MFPG, Virginia Beach, VA.
pp. 87-96, April 1990.
Steele ME, Ashen RA, and Knight LC:"Applications of an Electrical Method for
Condition Monitoring of Motors," MOTORCON Proc., pp. 321-335, Sept. 1982.
Tavner MA, Gaydon BG, and Ward BA: "Monitoring Large Generators and Motors."
IEE Proc., Val. 133, PI. B, No. 3, pp. 169-179, May 1986.
Tavner PJ, Armin KK, and Hargis C: "An Electricd Technique for Monitoring Induc-
tion Motor Cages," IEE Con$ on Electrical Machines and Drives. IEE Publicarion No.
282, pp. 43-46, 1987.
 474 G. 6 . KLIMAN AND J. STEIN

21. Williamson S and Mirzoian K: "Analysis of Cage Induction Motors With Stator
Winding Faults.'' IEEE-PES. Summer Power Meeling. July 1984.
22. Williamson S and Smith AC: "Steady State Analysis of 3-Phase Cage Motors With
Rotor Bar and End Ring Faults," IEEProc., Vol. 129, PI. B, No. 3, pp. 93-100, May
1982.

Manuscript rece~vedin final form April 7, 7992

Request reprints from G. 8.Kliman
Downloaded by [McGill University Library] at 06:28 23 December 2012