Adaptive, on line, statistical method and apparatus for detection of broken bars in motors by passive motor current monitoring and digital torque estimation

• US 5,742,522 A
• Filed: 04/01/1996
• Issued: 04/21/1998
• Est. Priority Date: 04/01/1996
• Status: Expired due to Term

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Kliman, Gerald Burt, Yazici, Birsen
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): DAM, TUAN QUANG
• Application Number:
  US08/627,721
• Time in Patent Office: 750 Days
• Field of Search: 364/481, 364/483, 364/485, 364/550, 364/551.01, 364/554, 364/492, 324/772, 324/510, 324/511, 324/545, 324/76.11, 324/76.19, 382/219, 382/225, 382/228, 395/13, 395/20, 395/184.01, 361/3, 361/23, 361/30, 361/31
• US Class Current: 702/185
• CPC Class Codes: G01R 23/16 Spectrum analysis; Fourier analysis
• CPC Class Codes: G01R 31/343 in operation
• CPC Class Codes: G01R 31/346 Testing of armature or field windings
• CPC Class Codes: G05B 9/02 electric

Abstract

During a learning stage a motor current signal is monitored, and estimated motor torque is used to transform the current signal into a time-frequency spectra including a plurality of segments representative of good operating modes. A representative parameter and a respective boundary of each segment is estimated. The current signal is monitored during a test stage to obtain test data, and the test data is compared with the representative parameter and the respective boundary of each respective segment to detect the presence of a fault in a motor. Frequencies at which broken bar faults are likely to occur in a motor can be estimated using the estimated motor torque, and a weighting function can highlight such frequencies during estimation of the parameter. The current signal can be further subdivided into the segments by monitoring sidebands of the frequency components of current spectrum strips of each segment. Estimating the parameter and the boundary of each segment can include calculating a segment mean (the representative parameter) and variance for each frequency component in each respective segment; calculating a modified Mahalanobis distance for each strip of each respective segment; and for each respective segment, using respective modified Mahalanobis distances to calculate a respective radius about a respective segment mean to define the respective boundary.

Claims

1. An adaptive, on line, statistical method for motor fault detection of broken bars by passive motor current monitoring comprising the steps of:

monitoring a motor current signal during a learning stage;

estimating motor torque to detect load changes;

transforming the current signal into time-frequency spectra;

using the load changes to divide the time-frequency spectra into a plurality of segments representative of good operating modes;

estimating a representative parameter of each segment;

determining a respective boundary of each segment;

monitoring the motor current signal during a test stage to obtain test data, the test stage occurring after the learning stage;

comparing the test data with the representative parameter and the respective boundary of each respective segment to detect the presence of a fault in a motor.

2. The method of claim 1, further including, prior to estimating the representative parameter of each segment and determining the respective boundary of each segment, the step of estimating frequencies at which broken bar faults are likely to occur in the motor, and

wherein the steps of estimating the representative parameter of each segment and determining the respective boundary of each segment include introducing a weighting function to highlight the frequencies at which broken bar faults are likely to occur.

3. The method of claim 2, wherein the step of using the load changes to divide the time-frequency spectra includes dividing the current signal into a plurality of portions each having a specified length of time and transforming each of the plurality of portions into a respective current spectrum strip.

4. The method of claim 3, wherein the step of using the load changes to divide the time-frequency spectra further includes monitoring a sideband of a first order harmonic of each respective current spectrum strip.

5. The method of claim 4, wherein the step of using the load changes to divide the time-frequency spectra further includes monitoring additional sidebands of a higher order harmonic of each respective current spectrum strip.

6. The method of claim 3, wherein the steps of estimating the representative parameter of each segment and determining the respective boundary of each segment include:

calculating a segment mean and a segment variance for each of a plurality of frequency components in each respective segment, the representative parameter comprising the mean;

calculating a modified Mahalanobis distance for each respective current spectrum strip of each respective segment using the respective segment means and variances; and

for each respective segment, using respective modified Mahalanobis distances to calculate a respective radius about a respective segment mean to define the respective boundary.

7. The method of claim 6, wherein the step of comparing the test data with the representative parameter of each segment and the respective boundary of each segment includes calculating test modified Mahalanobis distances between the test data and each of the segments, and comparing each test modified Mahalanobis distance with a respective radius of a respective segment to determine whether a potential fault exists.

8. The method of claim 7, wherein the step of comparing each test modified Mahalanobis distance with a respective radius of a respective segment to determine whether a potential fault exists includes monitoring sidebands of harmonics of the test data prior to concluding that the potential fault exists.

9. The method of claim 7, wherein the segments comprise good segments and further including:

when a potential fault exists, forming a test segment using consecutive test data and determining a test mean and a test variance of each of a plurality of frequency components in the test segment;

calculating good Bhattacharya distances between each good segment and each of the other good segments;

calculating test Bhattacharya distances between the test segment and each of the good segments;

making a determination whether any test Bhattacharya distance between the test segment and a respective good segment is less than any good Bhattacharya distance between the respective good segment and each of the other good segments; and

when the determination is negative, indicating an alarm.

10. An adaptive, on line, statistical method for induction motor broken bar fault detection by passive stator current monitoring comprising the steps of:

monitoring a stator current signal and estimating motor torque during a learning stage;

transforming the stator current signal into a time-frequency spectra including a plurality of segments representative of good operating modes by

dividing the current signal into a plurality of portions each having a specified length of time,

transforming each of the plurality of portions into a respective current spectrum strip,

monitoring changes in the estimated motor torque, and

monitoring sidebands of harmonics of each respective current spectrum strip;

estimating frequencies at which broken bars faults are likely to occur in an induction motor using the estimated motor torque;

estimating a representative parameter and a respective boundary of each segment using a weighting function to highlight the frequencies at which broken bar faults are likely to occur;

monitoring the stator current signal during a test stage to obtain test data, the test stage occurring after the learning stage; and

comparing the test data with the representative parameter and the respective boundary of each respective segment and monitoring sidebands of harmonics of the test data to detect the presence of a fault in the induction motor.

11. The method of claim 10,

wherein the step of estimating the representative parameter and the respective boundary of each segment includes

calculating a segment mean and a segment variance for each of a plurality of frequency components in each respective segment, the representative parameter comprising the mean,

calculating a modified Mahalanobis distance for each respective current spectrum strip of each respective segment using the respective segment means and variances, and

for each respective segment, using respective modified Mahalanobis distances to calculate a respective radius about a respective segment mean to define the respective boundary;

and

wherein the step of comparing the test data with the representative parameter and the respective boundary of each segment includes

calculating test modified Mahalanobis distances between the test data and each of the segments, and

comparing each test modified Mahalanobis distance with a respective radius of a respective segment to determine whether a potential fault exists.

12. A passive motor broken bar fault detection apparatus comprising:

a current sensor for monitoring a motor current signal during a learning stage, and for monitoring the motor current signal during a test stage to obtain test data, the test stage occurring after the learning stage;

a computer for, during the learning stage, estimating motor torque, transforming the current signal into a time-frequency spectra including a plurality of segments representative of good operating modes using the estimated motor torque, and estimating a representative parameter and a respective boundary of each segment, and, during the test stage, for comparing the test data with the representative parameter and the respective boundary of each respective segment to detect the presence of a fault in a motor.

13. The apparatus of claim 12, wherein means for the computer transforming the current signal includes means for dividing the current signal into a plurality of portions each having a specified length of time and transforming each of the plurality of portions into a respective current spectrum strip.

14. The apparatus of claim 13, wherein the means for the computer transforming the current signal further includes means for monitoring sidebands of harmonics of each current spectrum strip.

15. The apparatus of claim 13, wherein means for the computer estimating the representative parameter and the respective boundary of each segment includes

means for calculating a segment mean and a segment variance for each of a plurality of frequency components in each respective segment, the representative parameter comprising the mean;

means for calculating a modified Mahalanobis distance for each respective current spectrum strip of each respective segment using the respective segment means and variances; and

means for using, for each respective segment, respective modified Mahalanobis distances to calculate a respective radius about a respective segment mean to define the respective boundary.

16. The apparatus of claim 15, wherein the segments comprise good segments, and

wherein means for the computer comparing the test data with the representative parameter and the respective boundary of each segment includes

means for calculating test modified Mahalanobis distances between the test data and each of the good segments, and comparing each test modified Mahalanobis distance with a respective radius of a respective good segment to determine whether a potential fault exists,

means for, when a potential fault exists, forming a test segment using consecutive test data and determining a test mean and a test variance of each of a plurality of frequency components in the test segment,

means for calculating good Bhattacharya distances between each good segment and each of the other good segments and test Bhattacharya distances between the test segment and each of the good segments,

means for making a determination whether any test Bhattacharya distance between the test segment and a respective good segment is less than any good Bhattacharya distance between the respective good segment and each of the other good segments, and

means for, when the determination is negative, indicating an alarm.