System and method for diagnosing and validating a machine over a network using waveform data

• US 6,609,217 B1
• Filed: 05/26/2000
• Issued: 08/19/2003
• Est. Priority Date: 03/30/1998
• Status: Expired due to Term

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Johnson, John Andrew, Steen, Phillip Edward, Ramchandran, Ramesh, Bonissone, Piero Patrone, Shah, Rasiklal Punjalal, Ramani, Vipin Kewal, Chen, Yu-To
• Primary Examiner(s): Iqbal, Nadeem
• Assistant Examiner(s): Bonura, Timothy M.
• Application Number:
  US09/578,401
• Time in Patent Office: 1,180 Days
• Field of Search: 714/1-57
• US Class Current: 714/26
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS
• CPC Class Codes: G16H 40/40 for the management of medical equipment or devices, e.g. scheduling maintenance or upgrades
• CPC Class Codes: G16H 40/67 for remote operation
• CPC Class Codes: G16H 70/60 relating to pathologies

Abstract

A system and method for diagnosing and validating a machine over a network using waveform data. Historical waveform data are obtained via the network from machines having known faults along with corresponding actions for repairing the machines and are used to develop fault classification rules. The fault classification rules are stored in a diagnostic knowledge database. The database of classification rules are used to diagnose new waveform data from a machine having an unknown fault, via the network.

Claims

1. A system for diagnosing a machine from a remote facility over a network, comprising:

a diagnostic knowledge base containing a plurality of rules for diagnosing faults in a machine and a plurality of corrective actions for repairing the faults;

a diagnostic fault detector for categorizing waveform data generated from the machine as normal and faulty data, wherein the machine is connected to the remote facility over the network;

a diagnostic feature extractor for extracting a plurality of features from the waveform data categorized as faulty data; and

a diagnostic fault isolator coupled to the diagnostic feature extractor and the diagnostic knowledge base for isolating a candidate set of faults for the extracted features and identifying root causes most likely responsible for the candidate set of faults.

2. The system according to claim 1, wherein the waveform data comprises a plurality of time series plots.

3. The system according to claim 1, wherein the diagnostic fault detector comprises a gross filter and a fine filter.

4. The system according to claim 1, wherein the diagnostic feature extractor uses a time domain analysis, a frequency domain analysis, and a wavelet analysis.

5. The system according to claim 1, wherein the diagnostic fault isolator is a rule-based reasoning expert system.

6. The system according to claim 1, wherein the diagnostic fault isolator further comprises means for assigning a measure of confidence to each of the identified candidate set of faults, each measure of confidence indicating a belief that the fault is the most likely cause thereof.

7. The system according to claim 1, further comprising a diagnostic parser for removing extraneous data from the waveform data.

8. The system according to claim 1, further comprising a training unit coupled to the diagnostic knowledge base, the training unit comprising:

means for obtaining a plurality of sets of waveform data taken from a plurality of machines connected to the remote facility over the network, each of the sets of waveform data having known faults associated therewith and a corresponding corrective action for repairing the faults;

a training filter for categorizing each of the sets of waveform data as normal and faulty data;

a training feature extractor for extracting a plurality of features from each of the sets of waveform data categorized as faulty data; and

a training fault classifier for developing a plurality of rules that classify the feature extractions into a fault characterization and providing the plurality of rules to the diagnostic knowledge base.

9. The system according to claim 8, wherein the plurality of sets of waveform data comprise a plurality of time series plots.

10. The system according to claim 8, wherein the training filter comprises a gross filter and a fine filter.

11. The system according to claim 8, wherein the training feature extractor uses a time domain analysis, a frequency domain analysis, and a wavelet analysis.

12. The system according to claim 8, wherein the training fault classifier is a rule-based reasoning expert system.

13. The system according to claim 8, wherein the training unit further comprises a training parser for removing extraneous data from each of the sets of waveform data.

14. A method for diagnosing a machine having an unknown fault from a remote facility over a network, comprising the steps of:

obtaining a plurality of rules for diagnosing faults and a plurality of corrective actions for repairing the faults;

receiving new waveform data from the machine via the network;

categorizing the new waveform data as normal and faulty data;

extracting a plurality of features from the new waveform data categorized as faulty data; and

isolating a candidate set of faults for the extracted features and identifying root causes most likely responsible for the candidate set of faults.

15. The method according to claim 14, wherein the new waveform data comprises a plurality of time series plots.

16. The method according to claim 14, wherein the step of categorizing the new waveform data as normal and faulty data comprises using a gross filter and a fine filter.

17. The method according to claim 14, wherein the step of extracting a plurality of features comprises using a time domain analysis, a frequency domain analysis, and a wavelet analysis.

18. The method according to claim 14, wherein the step of isolating a candidate set of faults and identifying root causes comprises using a rule-based reasoning expert system.

19. The method according to claim 14, further comprising the step of assigning a measure of confidence to each of the identified candidate set of faults, each measure of confidence indicating a belief that the fault is the most likely cause thereof.

20. The method according to claim 14, further comprising the step of removing extraneous data from the new waveform data.

21. The method according to claim 14, wherein the step of obtaining the plurality of rules for diagnosing faults and the plurality of corrective actions for repairing the faults comprises the steps of:

obtaining a plurality of sets of waveform data taken from a plurality of machines via the network, each of the sets of waveform data having known faults associated therewith;

categorizing each of the sets of waveform data as normal and faulty data;

extracting a plurality of features from each of the sets of waveform data categorized as faulty data; and

developing a plurality of rules that classify the feature extractions into a fault characterization.

22. The method according to claim 21, wherein the plurality of sets of waveform data comprise a plurality of time series plots.

23. The method according to claim 21, wherein the step of categorizing each of the sets of waveform data as normal and faulty data comprises using a gross filter and a fine filter.

24. The method according to claim 21, wherein the step of extracting a plurality of features from each of the sets of waveform data comprises using a time domain analysis, a frequency domain analysis, and a wavelet analysis.

25. The method according to claim 21, further comprising the step of removing extraneous data from each of the sets of waveform data.

26. A system for performing a validation of waveform data generated from a machine, from a remote facility over a network, comprising:

a diagnostic knowledge base containing a plurality of rules for diagnosing faults in the machine;

a diagnostic fault detector, for categorizing the waveform data as normal and faulty data; and

a diagnostic feature extractor for extracting a plurality of features from the waveform data categorized as normal data.

27. The system according to claim 26, wherein the waveform data comprises a plurality of time series plots.

28. The system according to claim 26, wherein the diagnostic fault detector comprises a gross filter and a fine filter.

29. The system according to claim 26, wherein the diagnostic feature extractor uses a time domain analysis, a frequency domain analysis, and a wavelet analysis.

30. The system according to claim 26, further comprising a diagnostic parser for removing extraneous data from the waveform data.

31. A method for performing a validation of waveform data generated from a machine, from a remote facility over a network, comprising the steps of:

obtaining a plurality of rules for diagnosing faults;

receiving new waveform data from the machine via the network;

categorizing the new waveform data as normal and faulty data with the plurality of rules; and

extracting a plurality of features from the new waveform data categorized as normal data.

32. The method according to claim 31, wherein the new waveform data comprises a plurality of time series plots.

33. The method according to claim 31, wherein the step of categorizing the new waveform data as normal and faulty data comprises using a gross filter and a fine filter.

34. The method according to claim 31, wherein the step of extracting a plurality of features comprises using a time domain analysis, a frequency domain analysis, and a wavelet analysis.

35. The method according to claim 31, further comprising the step of removing extraneous data from the new waveform data.