Fault detection in synchronous machines using pattern recognition

Patent metadata

Field	Value
Publication number	US-2026072084-A1
Application number	US-202319107833-A
Country	US
Kind code	A1
Filing date	Aug 31, 2023
Priority date	Sep 2, 2022
Publication date	Mar 12, 2026
Grant date	—

Abstract

A method of fault detection in a synchronous machine is provided. The method comprises: using at least one sensor to determine a measure of a stray magnetic field generated by the synchronous machine; processing the determined measure in order to obtain a frequency spectrum for the stray magnetic field; analysing the frequency spectrum for the stray magnetic field in order to identify a predetermined pattern, characteristic of one or more fault conditions, within the frequency spectrum; and determining one or more faults within the synchronous machine based on identifying the pattern.

First claim

1 . A method of detecting one or more faults in a synchronous machine, the method comprising:
using at least one sensor to determine a measure of a stray magnetic field generated by a synchronous machine; processing the determined measure in order to obtain a frequency spectrum for the stray magnetic field; analysing the frequency spectrum for the stray magnetic field in order to identify a predetermined pattern, characteristic of one or more fault conditions, within the frequency spectrum; and determining one or more faults within the synchronous machine based on identifying the pattern.

2 . The method of claim 1 , wherein the one or more faults includes an inter-turn short circuit fault, or a dynamic eccentricity fault, or a short-circuit fault, or a static eccentricity fault, or a misalignment, or a demagnetization fault, or a combination of two or more such faults.

3 . The method of claim 1 , wherein the synchronous machine is a wound field synchronous machine or is a permanent-magnet synchronous machine or is a synchronous generator.

4 . (canceled)

5 . (canceled)

6 . The method of claim 1 , wherein the at least one sensor comprises a search coil located on or adjacent a stator core back side or a frame of the synchronous machine.

7 . (canceled)

8 . The method of claim 1 , wherein the at least one sensor comprises at least one non-invasive sensor.

9 . (canceled)

10 . (canceled)

11 . The method of claim 1 , wherein the method is carried out during no-load or full-load operation of the synchronous machine.

12 . (canceled)

13 . The method of claim 1 , wherein the predetermined pattern is independent of a power rating or of a topology of the synchronous machine.

14 . The method of claim 1 , comprising using a machine learning algorithm to identify the predetermined pattern.

15 . The method of claim 1 , wherein analysing the frequency spectrum comprises determining whether the frequency spectrum matches one of two or more different predetermined patterns, each predetermined pattern being characteristic of a different respective fault condition.

16 . The method of claim 1 , comprising signalling a detection of the one or more faults to an output device or over a network connection.

17 . The method of claim 1 , wherein the stray magnetic field is a combination of radial and axial magnetic fields.

18 . (canceled)

19 . The method of claim 1 , wherein the predetermined pattern is a pattern of amplitudes of respective frequency components within the frequency spectrum.

20 . (canceled)

21 . The method of claim 19 , wherein the predetermined pattern comprises an increasing sequence of amplitudes between a stator terminal frequency, fs, of the synchronous machine and twice fs, and a decreasing sequence of amplitudes between twice fs and three times fs, and wherein the predetermined pattern is characteristic of an inter-turn short circuit fault.

22 . The method of claim 19 , wherein the predetermined pattern comprises a decreasing sequence of amplitudes between a stator terminal frequency, fs, of the synchronous machine and twice fs, and an increasing sequence of amplitudes between twice fs and three times fs, and wherein the predetermined pattern is characteristic of a dynamic eccentricity fault.

23 . The method of claim 19 , wherein the predetermined pattern comprises the amplitudes of frequency components between a stator terminal frequency, fs, of the synchronous machine and twice fs all being within a predetermined maximum range, and wherein the predetermined pattern is characteristic of a combined inter-turn short circuit fault and dynamic eccentricity fault.

24 . (canceled)

25 . A fault detection system comprising a data processing apparatus configured to receive a measure of a stray magnetic field generated by a synchronous machine;
wherein the data processing apparatus is configured to:
process the received measure in order to obtain a frequency spectrum for the stray magnetic field;
analyse the frequency spectrum for the stray magnetic field in order to identify a predetermined pattern, characteristic of one or more fault conditions, within the frequency spectrum; and
determine one or more faults within the synchronous machine based on identifying the pattern.

26 . (canceled)

27 . The fault detection system of claim 25 , wherein the data processing apparatus is or comprises a microcontroller or a field programmable gate array.

28 . (canceled)

29 . The fault detection system of claim 25 , further comprising at least one sensor configured to determine the measure of the stray magnetic field.

30 . A synchronous machine comprising the fault detection system of claim 25 .

31 . A non-transitory computer-readable storage medium comprising instructions that, when executed on a data processing apparatus, will cause the data processing apparatus to:
process a received measure of a stray magnetic field generated by a synchronous machine in order to obtain a frequency spectrum for the stray magnetic field; analyse the frequency spectrum for the stray magnetic field in order to identify a predetermined pattern, characteristic of one or more fault conditions, within the frequency spectrum; and determine one or more faults within the synchronous machine based on identifying the pattern.

32 . (canceled)

Assignees

• Norwegian Univ Sci & Tech Ntnu

Inventors

• Ehya Hossein

Classifications

• G01R33/06

  using galvano-magnetic devices · CPC title

• G01R31/346

  Testing of armature or field windings · CPC title

• G01R23/16

  Spectrum analysis; Fourier analysis · CPC title

• H02K11/20

  for measuring, monitoring, testing, protecting or switching (rectifiers H02K11/04; power electronics H02K11/33) · CPC title

• G06N20/00

  Machine learning · CPC title