Vous êtes sur la page 1sur 8

Application of Online Voltage Mismatch Condition

Monitoring of Induction Motor Stator Windings in a


Mining Environment
J. Sottile F. C. Trutt J. L. Kohler
230 MMRB 687B AH National Institute for Occupational
University of Kentucky University of Kentucky Safety and Health
Lexington, KY 40506-0107 USA Lexington, KY 40506-0046 USA Pittsburgh, PA 15236 USA
jsottile@ieee.org ftrutt@ieee.org jtk4@cdc.gov

Abstract - Previous theoretical and laboratory research has external resistance in one of the lines, and the motor load can
demonstrated the utility of using voltage mismatch predictors for be varied by suitable means and the voltages and currents can
online condition monitoring of induction motor stator windings. be recorded. Next, the external resistance can be inserted in
However, the use of this monitoring system in a mining or other another line and the process can be repeated. In a field
industry application depends upon the ability to determine motor application, the training process cannot be controlled in the
impedance parameters prior to the onset of winding same fashion as in the laboratory. It may be impractical or
deterioration. Since these impedance parameters depend upon impossible to insert an external resistance in the lines to create
the motor speed, such evaluations require the collection of the necessary voltage unbalance. Motor speed will be a
voltage and current data over a range of operating speeds. In
function of the application and cannot be controlled, and in
addition, a sufficient amount of natural or created voltage
come cases, measuring motor speed may be problematic.
unbalance is required in order to evaluate these parameters. The
investigation presented in this paper uses duplicate data- The investigation presented in this paper addresses the
collection and evaluation setups in both an operating mine and in issues associated with establishing the impedance matrix in an
the laboratory in order to obtain a better understanding of the industrial setting. We begin with a brief description of the
voltage variations that are required for determination of the voltage mismatch detectors. Subsequently, a description of the
motor impedance parameters. Results are then related to the laboratory and field efforts is presented, followed by the results
feasibility of applying voltage mismatch condition-monitoring
of these studies and plans for future work.
procedures in an operating mine.

Keywords induction motor, condition monitoring II. BACKGROUND

I. INTRODUCTION A. Voltage Mismatch Detectors


A brief description of the voltage mismatch detectors is
Previous research has demonstrated that the monitoring of
presented here. Additional details can be found in [13, 14].
electrical parameters can be a useful technique for the
condition monitoring of induction motor windings [1-14]. As Consider a three-phase, wye-connected induction motor
part of this work, an approach utilizing voltage mismatch connected to a three-phase, solidly grounded wye-connected
comparison to detect stator winding deterioration or faults in power source. (If the power source or motor is in a delta
operating induction motors has been proposed and verified in configuration, it is assumed that this component is represented
the laboratory [13, 14]. Application of this voltage mismatch by its equivalent wye.)
comparison technique requires measurement of the three-phase
voltages, three-phase currents, and motor speed. In the Define V'a0, Va1, and Va2 as the phase a zero, positive, and
evaluation process, positive- and negative-sequence voltage negative sequence symmetrical components of the source
values computed from the measured line voltages are compared voltages measured with respect to ground. Similarly, we define
with positive- and negative-sequence voltages computed from Ia0, Ia1, and Ia2 as the symmetrical components of the line
measured currents and speed-dependent impedance parameters. currents and Va0, Va1, and Va2 as the symmetrical components
of the motor voltages measured with respect to the junction
Prior to implementation, a training process is necessary to point of its wye connection. The zero sequence voltages V'a0
establish the impedance parameters. This process requires that and Va0 differ by the voltage at the junction point of the motor
two independent sets of voltage and current measurements be wye connection measured with respect to the ground reference
recorded at the same motor speed. Because these terms are while the corresponding positive and negative sequence power
speed dependent, this process must be repeated for different source and motor phase voltages are equal.
motor operating speeds. In the laboratory, this process can be
conducted very conveniently. A specified supply unbalance for From a standard text, it may be shown that [15]
the motor can be established, for example, by inserting an Va0 = z00 Ia0 + z01 Ia1 + z02 Ia2, (1)

0-7803-7883-0/03/$17.00 2003 IEEE


1637
necessary that the two tests be conducted under different values
Va1 = z10 Ia0 + z11 Ia1 + z12 Ia2, and (2) of voltage supply unbalance. Using small motors in the
laboratory, these conditions are not difficult to establish. An
Va2 = z20 Ia0 + z21 Ia1 + z22 Ia2. (3) autotransformer or resistor can be used to create the necessary
unbalance and the speed can be controlled by connecting the
where the zxy parameters are functions of the motor design, motor to a generator and varying the generator load in
construction, any internal deterioration, and the operating sufficiently small increments.
speed. The zero sequence component of the line currents, Ia0, However, in an industrial setting, it may be impractical to
is approximately zero if there is no ground fault in the system. attempt to control the voltage supply and motor speed in a
Because such faults or deterioration involving ground may be similar fashion. To be accepted by the mining or other
detected by standard techniques that monitor the zero- industry, the training and monitoring processes should not
sequence component of the line currents and are relatively impact the operation (other than the installation of sensors).
easy to detect, we will concentrate on situations in which there Therefore, a more practical approach for an industrial setting is
is no ground fault present, i.e., Ia0 is assumed to be zero. to allow the motor to operate normally and to collect training
data over a period of time. During this data collection process,
With this assumption, (2) and (3) become there should be sufficient natural variations in the voltage
supply to develop the impedance parameters over the range of
Va1 = z11 Ia1 + z12 Ia2, and (4) observed operating speeds. It is worth noting here that voltage
unbalances, which are necessary for determination of the motor
impedance parameters during the training phase, are of no
Va2 = z21 Ia1 + z22 Ia2. (5)
consequence during the subsequent condition monitoring
process.
The impedance terms z11, z12, z21, and z22 can be determined
using recorded lines voltages and currents from two separate
tests on a motor running at any given speed. If these tests are III. LABORATORY INVESTIGATION
repeated at different motor speeds, a library of zxy parameters In order to experimentally investigate the sensitivity of
may be determined over a range of motor operating speeds. induction motor impedance parameter determination to the
degree of voltage unbalance and possible experimental error, a
Once these impedance parameters have been determined, it laboratory setup was constructed using instrumentation
becomes possible to monitor the line voltages and line currents identical to that which was installed in an operating mine. The
from which Va1, Va2, Ia1, and Ia2 may be calculated. Equations three line currents were measured with 200:1 current
(4) and (5) may then be used to calculate Va1 and Va2 at the transformers and burden resistors were used to convert the
measured motor operating speed and if no internal current to a voltage signal. The three line voltages were
deterioration has occurred, Va1 and Va2 calculated from the measured with three 480:2.84V potential transformers. Speed
measured currents and impedance values should match the Va1 was measured with an optical tachometer. The tachometer had
and Va2 values extracted from the measured voltages. If a 0-5 V analog output that was programmable over a user-
internal deterioration such as a turn-to-turn fault or phase-to- defined speed range. The machine used in this laboratory
phase leakage has developed, the zxy parameters will have investigation was a 208-V, 2.0-kVA, two-pole universal
changed from their normal values at that speed and there will laboratory machine (ULM) connected as an induction motor.
be a mismatch between the two sets of voltages. This This motor was coupled to a dc generator whose output was
mismatch may then be used as an indicator of internal motor connected to a bank of resistors that could be switched in
deterioration. 300W increments. A rheostat was placed in the generator field
An important advantage is that this condition monitoring circuit to control output voltage and provide finer loading
approach has been shown to be independent of initial increments. The voltage, current, and speed measurements
construction imperfections and unbalances within the motor were fed into an analog to digital converter and then into a
itself as well as changes or unbalanced conditions that may computer for analysis. Each data set consisted of seven
develop in the motor power supply system [14]. In addition, channels of input (three voltage, three current, and one speed)
performance should not be affected by mismatched sensors; consisting of 4096 points per channel sampled at a rate of
therefore, no careful sensor calibration is necessary. However, 4098.4 samples per second. This provided a resolution of
it is assumed that the sensors will not be rearranged or approximately one Hertz in the frequency domain.
replaced during both training and condition monitoring The first series of laboratory tests was conducted with the
processes. ULM running as an induction motor with a wye-connected
stator winding. In order to produce a voltage unbalance, a 3
B. Determination of the Impedance Matrix resistor (i.e., a dc resistance of 3 at room temperature) was
The calculation of the impedance matrix is a placed in line A for one series of tests, and line B for the next
straightforward process. However, because two independent series of tests. This produced a voltage unbalance factor,
sets of voltage and current measurements are required to defined as the ratio of the negative sequence voltage to the
compute the impedance parameters at a given speed, it is positive sequence voltage, of approximately 2.5%. The ULM

1638
was run at 3350 rpm with a balanced voltage supply for 30
minutes to allow it to reach a steady operating temperature.
Tests were conducted immediately after the warm-up period.
Motor speed was varied from 3350 rpm to 3495 rpm in 5-rpm
increments (to the extent possible). The entire series of tests
took less than one hour to conduct, and no noticeable
additional heating due to the unbalance was noticed, although
some heating probably did occur.
The graphs given in Figs. 1-4 show the four terms of the
impedance matrix for the speed range tested. Because of the
large differences in scale, they have been shown on separate
graphs to make it easier to observe changes in the values.
Three of the impedance values, z11, z21, and z22 have very
consistent behavior; z11 increases with decreasing load, as
expected; z22 is nearly constant, as expected; and z21 shows a
very gradual increase.
Unfortunately, z12 is poorly behaved. The reason for this
becomes evident if Cramers Rule is applied to the set of Figure 2. Real and imaginary components of z12 as a function of speed
equations that are used to calculate z11 and z12. When this is for the ULM.
done, the value of the numerator when determining z12 is the
difference between two large, nearly equal numbers, Ia1Va1 The second series of laboratory tests was designed to
and Ia1Va1. (Unprimed and primed values are used to investigate the effects of voltage unbalance difference between
distinguish the two data sets that are used in the determination a pair of measurements on the calculated value of the motor
of the impedance matrix.) Thus any small measurement error impedance parameters. For these tests, the motor speed was
will cause a large change in the very small value of z12 leading maintained at 3425 rpm and a voltage unbalance difference,
to instability in this calculation. (Recall that in a perfectly VUD, was defined as the magnitude of the quantity
balanced system, the off-diagonal terms would be equal to
V a 2 V a' 2
zero; in a typical system, they should be close to zero.) It VUD = ( 6)
therefore seems clear that this impedance cannot be used in a V a1 V a'1
practical system where some measurement inaccuracies must in percent. The value of the resistance inserted in line A or B
be expected. Fortunately, the dominance of z11 is such that no was varied such that the magnitude of this voltage unbalance
significant error should be introduced if z12 is assumed to be difference assumed values between 3.5% and 0.2%. The
small compared to z11 when determining the remaining motor lowest value of 0.2% was limited by the natural unbalance that
impedances and the voltage mismatches. existed in the laboratory supply system.

Figure 1. Real and imaginary components of z11 as a function of speed Figure 3. Real and imaginary components of z21 as a function of speed
for the ULM. for the ULM.

1639
Figure 4. Real and imaginary components of z22 as a function of speed Figure 6. Real and imaginary components of z21 as a function of
for the ULM. voltage unbalance difference for the ULM operating at 3425 rpm.

Ohms is essentially constant down to the lower limit of the


Results of these tests are shown in the graphs of Figs. 5 experiment. If average values for the other impedance
7. In each of these graphs, the real and imaginary parts of an parameters are calculated, z21 = 0.0985 + j0.147 Ohms and z22
impedance parameter are plotted as a function of the = 1.87 + j2.81 Ohms. If only measurements where the
magnitude of the voltage unbalance difference. The dual plots magnitude of the voltage unbalance is below 1% (where the
contained in Fig. 5 show the value of z11 calculated with z12 variations in the calculated impedance values are the greatest),
included (solid lines) and with z12 assumed to be equal to zero the average values become z21 = 0.0978 + j0.149 Ohms and z22
(dashed lines). (Data point markers were left off of this figure = 1.86 +j 2.825 Ohms. It therefore appears that average values
to show these differences more clearly.) In this latter case, the for the motor impedance parameters may give useable results,
average value of z11 for the pair of measurements was utilized. if a sufficient number of measurements are obtained, even if
some low values of unbalance difference contribute to
These results clearly show an increase in the variability of
variations in the individual computations.
the calculated values of z21 and z22 as the magnitude of the
voltage unbalance difference goes below approximately 1%.
However, when z12 is assumed to be zero, z11 = 30.6 + j14.7

Figure 5. Real and imaginary components of z11 as a function of Figure 7. Real and imaginary components of z22 as a function of
voltage unbalance difference for the ULM operating at 3425 rpm. voltage unbalance difference for the ULM operating at 3425 rpm.

1640
IV. FIELD INVESTIGATION

A. Training Procedure
During the field effort, data files were collected at an
underground limestone mine using a system that was identical
to the one used in the laboratory. This allowed for a
comparison of the quality of data between laboratory and the
operating mine and also provided an opportunity to debug the
data collection system in the laboratory if problems were
encountered in the mine. The motor selected for the study was
a 150 hp conveyor belt drive motor that transported rock from
the mines primary crusher. The potential transformers and
current transformers were located at the secondary of a
12,470:480V transformer that supplied the motor and crusher.
The data acquisition system was set up approximately six feet
from the transformer enclosure. The conveyor motor was
located approximately 200 feet from the transformer, and
required 230 feet of signal wire to reach the data collection
system. Figure 8 shows the motor and the optical speed
(a) Laboratory speed signal.
sensor.
A total of 120 files were collected over a three week
period, using the same sampling parameters as those used in
the laboratory. The observed voltage unbalance factor was
less than 0.5% for all of the files recorded. During the data
collection process, it was observed that the motor operating
speed was relatively narrow, ranging from 1787 to 1798 rpm,
indicating that this motor was probably oversized for this
application. We also observed that the tachometer signal had
significantly more noise than the signal recorded in the
laboratory, as illustrated in Figure 9. This was problematic
because the z11 impedance parameter is very sensitive to
changes in speed. During these tests, the magnitude of the
zero-sequence current was monitored and was less than 0.3 A,
justifying the assumption of (4) and (5).
An algorithm was developed in which pairs of files were
matched to compute the impedance parameters. Because of
the uncertainty in the speed measurements, files were matched
primarily on the basis of positive sequence impedance, and the (b) Mine-site speed signal.
speed was computed as the average of the two measurements.
Figure 9. Sample of tachometer speed signals.

The second criterion was that the voltage unbalance be


different to insure independent equations were used in the
calculations. Of the 120 data files collected during the
training process, 27 combinations were initially selected as
meeting the criteria. Of the 27 files initially selected, 23
provided useable information for determining the impedance
parameters. While the algorithms for selecting useable
combinations of data files were applied by hand in this case,
programming techniques may be utilized to automatically
incorporate these decisions into an automated system.
Figures 10 through 12 show the results of the impedance
matrix training process for the data collected at the mine site. It
is noted here that these values were determined from an
average voltage unbalance factor of less than 0.5%. As a
Figure 8. Conveyor motor and optical speed sensor. result, the values are more widely scattered than those collected

1641
Real Real

Imaginary Imaginary

Figure 10. Real and imaginary components of z11 as a function of Figure 12. Real and imaginary components of z22 as a function of
speed for the 150hp conveyor motor. speed for the 150hp conveyor motor.

in the laboratory at higher levels of unbalance; however, clear Figure 13 illustrates the results of this process. The
trends were still established. Another difference is that z21 of mismatch values are the magnitudes of the vector difference
the conveyor was approximately constant while z21 for the between the calculated and measured voltage values. Figure 13
ULM exhibited a slight increase with increasing speed. (a) shows that the positive sequence voltage mismatch is less
than 12 V in all cases, with a mean value of 6.7 V. These
B. Condition Monitoring values represent 4.33% and 2.42%, respectively, of the rated
Once the impedance parameters were determined, a line-to-neutral voltage of this motor. The plot also shows that
simulated health-monitoring procedure was conducted. This the mismatches are fairly uniform over the measured speed
procedure consisted of randomly selecting 15 files (that were range. This result is encouraging because of the sensitivity of
not used in the training process) over the measured speed range z11 to speed, especially near synchronous speed. Figure 13 (b)
and computing the positive and negative sequence voltage shows the negative sequence voltage mismatch values are all
mismatches. Because these files were collected during the less than 0.5 V with a mean mismatch value of 0.27 V.
same time period as those used for training, the mismatch Although these mismatches are much smaller than their
values should be small and representative of what would be positive sequence counterparts, they represent a larger fraction
expected for condition monitoring of a healthy motor. of the measured negative sequence voltage compared with the
positive sequence values. For example, the average measured
negative sequence voltage was approximately 1.2 V, so the
average mismatch is approximately 23% of the average
observed negative sequence voltage. This is expected because
Real the negative sequence values are much smaller than the
positive sequence values and are more difficult to measure
accurately.
Because it is not possible to simulate motor deterioration in
this setting and there was no apparent deterioration or failure
that occurred during the monitoring period, it was not possible
to predict the magnitude of the change in the mismatch
predictors with deterioration as compared to random variations
Imaginary for the healthy motor. However, some insight into expected
performance can be obtained by comparing these results with
results of simulated deterioration tests previously conducted on
the ULM. Fig. 14 is a repeat of Fig. 2 in [14] showing the
change in the voltage mismatch detectors with increasing
deterioration current for healthy operation and for increasing
levels of simulated deterioration. In this figure, we observe a
non-deteriorated Va1 mismatch of approximately 2.0 V or
approximately 1.75% of the rated line-to-neutral voltage
Figure 11. Real and imaginary components of z21 as a function of speed compared with the mine site mean of 6.7 V (2.42% of rated)
for the 150hp conveyor motor.
and maximum value of 12 V (4.33% of rated). The ULM non-
deteriorated Va2 mismatch is approximately 0.2 V compared

1642
25
Voltage Unbalance = 5.2%
n = 3303 rpm

20 Ia2 Measured x 10

Voltage (V) or Cuttent (A)


Ia2 Calculated x 10

15
Va2 Mismatch Calculated x 10

Va2 Mismatch Measured x 10


10

5
Va1 Mismatch Measured

Va1 Mismatch Calculated


0
0 2 4 6 8 10 12 14 16
Deterioration Current (Percent of simulated short-circuit value)

Figure 14. Experimental and theoretical voltage mismatches with a


large supply voltage unbalance for the ULM [14].
(a) Positive sequence voltage mismatch.

C. Implementation Issues
The objectives of this study were to obtain a better
understanding of the voltage variations that are required to
develop the motor impedance parameters, and to establish
whether it is feasible to conduct this process in an industrial
setting.
The results of the field study are encouraging, especially
when considering the conditions under which they were
conducted. For example, the data collection process was
conducted manually, in which an operator sat at the data
collection system and triggered the acquisition process. This
was appropriate for this study because it allowed the operator
to monitor the data collection process and deal with any
unforeseen problems. However, compared with an automated
system, it limited the number of opportunities for collecting
files in which the voltage unbalance difference is large enough
to establish the impedance parameters with a high degree of
accuracy. When fully implemented, an automated process
(b) Negative sequence voltage mismatch. could be conducted almost continuously over several weeks
Figure 13. Positive and negative sequence voltage mismatch values for
thereby providing better estimates of the impedance
the 150 hp conveyor motor. parameters.
Another difficulty that is more problematic in an industrial
with the mine site mean and maximum mismatch values of application is the determination of motor speed. For this study,
0.27 V and 0.5 V, respectively. an optical tachometer was used because it was accurate,
These comparisons indicate that the field site results for the convenient to use, and it worked well in the laboratory.
healthy motor are similar to the laboratory study previously However, in the mine environment, the tachometer signal was
conducted, although the non-deteriorated mismatches are noisy, making it difficult to accurately measure motor speed.
somewhat larger. It is anticipated that other industrial sites This hampered the determination of the impedance parameters
would exhibit similar results. These results also indicate the because pairs of files had to be matched primarily on the basis
need for using trends in the mismatch values rather than simply of positive sequence impedance and the speed was determined
establishing a threshold value. A major point of interest from by averaging the values of two files. It is expected that similar
Figure 14 is that in percentage terms, the maximum positive noise problems would be encountered at other locations, or that
and negative sequence voltage mismatches observed for the it may simply be impractical to mount a tachometer.
healthy motor in the mine tests are exceeded in the ULM when There are alternatives to using a tachometer for measuring
the deterioration current is limited to approximately 5% of the speed. One technique is to obtain the motor running speed
short circuit value. This is a very low level of deterioration, from analysis of the amplitude spectrum of the line current.
and at this, as well as higher simulated levels of deterioration, However, this approach requires a relatively long record length
no abnormal operation of the ULM was observed. (approximately 30-50 seconds) to obtain sufficient resolution in

1643
the frequency domain for an accurate speed determination. In [5] M.A. Cash and T.G. Habetler, Insulation failure detection in the stator
this application, and many mining applications, the motor load windings of ASD-driven induction machines using standard deviation of
line currents, Record of the 33rd IEEE Industry Applications Society
and speed may vary during this length of time such that it Conference, pp. 299-303, Oct. 1998.
would be necessary to implement training and monitoring [6] A.J.M. Cardoso, S.M.A. Cruz, and D.S.D. Fonseca, Inter-turn stator
algorithms based on average values observed during the winding fault diagnosis in three-phase induction motors by Parks
sampling interval. Vector approach, IEEE Transactions on Energy Conversion, vol. 14,
no. 3, pp. 595-598, 1999.
Plans for future work include the investigation of more [7] R.M. Tallam, T.G. Habetler, R.G. Harley, D.J. Gritter, and B. Burton,
reliable and unobtrusive methods for accurately measuring Neural network based on-line stator winding turn fault detection for
motor speed; the use of longer sampling intervals and the induction motors, Record of 35th IEEE Industry Applications Society
application of averaging techniques; the development and Conference, vol. 1, pp. 375-380, Oct. 2000.
testing of an automated system for the training and health [8] S. Williamson and K. Mirzoian, Analysis of cage induction motors with
monitoring processes; and installation of several prototype stator winding faults, IEEE Transactions on Power Apparatus and
Systems, vol. 104, pp. 1838-1842, 1985.
systems at industrial sites. These systems would then be used
[9] M.A. Cash, T.G. Habetler, and G.B. Kliman, Insulation failure
to carry out the training process and conduct condition prediction in induction motors using line-neutral voltages, Record of
monitoring for an extended period of time. the 32nd IEEE Industry Applications Society Conference, pp. 208-212,
Oct. 1997.
V. CONCLUSION [10] R.M. Tallam, T.G. Habetler, and R.G. Harley, Stator winding turn-fault
detection for closed loop induction motor drives, IEEE Transactions on
The results of the training and monitoring processes for the Industry Applications, vol. 39, no. 3, May/June 2003, pp. 720-725.
voltage mismatch detectors are very encouraging. Laboratory [11] G.B. Kliman, W. J. Premerlani, R.A. Koegl, and D. Hoeweler, A new
tests have indicated that training can be conducted at fairly low approach to on-line turn fault detection in AC motors, Record of 31st
levels of voltage unbalance, provided enough samples are used IEEE Industry Applications Society Conference, pp. 686-693, Oct. 1996.
to help reduce random error. Based on these results, a field [12] J.L. Kohler, J. Sottile, and F.C. Trutt, Condition monitoring of stator
study was conducted at an underground limestone mine in windings in induction motors: Part I - Experimental investigation of the
effective negative-sequence impedance, IEEE Transactions on
which the training process was conducted without the benefit Industry Applications, vol. 38, no. 5, Sep/Oct. 2002, pp.1447-1453.
of an artificially induced supply unbalance or control of the [13] J. Sottile, F.C. Trutt, and J.L. Kohler, Condition monitoring of stator
motor speed. This process was successful despite its being windings in induction motors: Part II - Experimental investigation of
hampered by noisy signals from the optical tachometer used to voltage mismatch detectors, IEEE Transactions on Industry
obtain the motor speed and relatively small natural supply Applications, vol. 38, no. 5, Sep/Oct. 2002, pp.1454-1459.
unbalances. Subsequently, a simulated health monitoring [14] F.C. Trutt, J. Sottile, and J.L. Kohler, Online Condition Monitoring of
process using randomly selected sets of mine data showed that Induction Motors, IEEE Transactions on Industry Applications, vol.
38, no. 6, Nov/Dec 2002, pp.1627-1632.
reasonably small voltage mismatch values were observed for
[15] S.A. Nasar and F.C. Trutt, Electric Power Systems, CRC Press, 1999.
the healthy motor, i.e., less than 4.33% of rated voltage for the
positive sequence mismatch and less than 0.5 V for the
negative sequence mismatch. These mismatches indicate that
the computed impedance parameters are fairly representative of
the motor. Further comparison of this mine data with
laboratory observations showed that normal mismatches in the
mine data were exceeded at low levels of simulated stator
winding deterioration. As a result, it appears that voltage
mismatch condition monitoring could provide useful early
warning information in this industrial setting.

REFERENCES
[1] J. Sottile and J.L. Kohler, An on-line method to detect incipient failure
of turn insulation in random-wound motors, IEEE Transactions on
Energy Conversion, vol. 8, pp. 762-768, Dec. 1993.
[2] J. Penmen, H.G. Sedding, B.A. Lloyd, and W.T. Fink, Detection and
location of interturn short circuits in the stator windings of operating
motors, IEEE Transactions on Energy Conversion, vol. 9, pp. 652-658,
Dec. 1994.
[3] M. Arkan, and P.J. Unsworth, Stator fault diagnosis in induction motors
using power decomposition, Record of the 34th IEEE Industry
Applications Society Conference, pp. 1908-1912, Oct. 1999.
[4] R.R. Schoen and T.G. Habetler, A new method of current-based
condition monitoring in induction machines operating under arbitrary
load conditions Electric Machines and Power Systems, vol. 25, pp. 141-
152, 1997.

1644

Vous aimerez peut-être aussi