Systems and methods for diagnosing the cause of trend shifts in performance data

• US 7,031,878 B2
• Filed: 12/16/2002
• Issued: 04/18/2006
• Est. Priority Date: 12/16/2002
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Yu, LiJie, Cuddihy, Paul E., Cleary, Daniel J.
• Primary Examiner(s): Barlow, John
• Assistant Examiner(s): VO, HIEN XUAN
• Application Number:
  US10/320,173
• Publication Number:
  US 20040117150A1
• Time in Patent Office: 1,219 Days
• Field of Search: 702/182, 702/183, 702/184, 702/185, 701/100, 701/29, 701/30, 703/8
• US Class Current: 702/182
• CPC Class Codes: H04L 43/00 Arrangements for monitoring or testing data switching networks

Abstract

Automated systems and methods for diagnosing the cause of a trend shift in performance data associated with a system including selecting a fuzzy model describing how the mean associated with a parameter shifts when a predetermined problem occurs and receiving performance data associated with the system. The diagnostic systems and methods also including applying a preferred set of piecewise regressions. The diagnostic systems and methods further including selecting a start date and selecting data sets near the start date and an end date that have a relatively large size without violating normal scatter. The diagnostic systems and methods still further including measuring the mean shift between a plurality of samples using one or more statistical tests and combining the results with the fuzzy model to achieve a diagnosis.

Claims

1. An automated method for diagnosing the cause of a trend shift in performance data associated with a system, the method comprising:

selecting a fuzzy model describing how the mean associated with a parameter shifts when a predetermined problem occurs;

receiving performance data associated with the system;

applying a preferred set of piecewise regressions;

selecting a start date;

selecting data sets near the start date and an end date that have a predetermined size without violating normal scatter;

measuring the mean shift between a plurality of samples using one or more statistical tests; and

combining the results with the fuzzy model to achieve a diagnosis.

2. The diagnostic method of claim 1, further comprising removing outlying data points from the performance data by calculating a local standard deviation minus a data point in question and removing the data point if the data point falls outside of a predetermined range.

3. The diagnostic method of claim 1, wherein the start date provides a relatively large shift without leveling or changing direction between the start date and an end date.

4. The diagnostic method of claim 1, wherein the one or more statistical tests comprise a two-sample t-test.

5. The diagnostic method of claim 1, further comprising providing a plurality of validation cases associated with the system.

6. The diagnostic method of claim 5, further comprising validating the fuzzy model using the plurality of validation cases.

7. The diagnostic method of claim 5, further comprising providing an evaluation function.

8. The diagnostic method of claim 7, further comprising using the evaluation function to determine how well the fuzzy model differentiates the diagnosis from a plurality of incorrect diagnoses for the plurality of validation cases.

9. The diagnostic method of claim 8, wherein using the evaluation function to determine how well the fuzzy model differentiates the diagnosis from the plurality of incorrect diagnoses for the plurality of validation cases comprises outputting a plurality of confidence values for the diagnosis and the plurality of incorrect diagnoses.

10. The diagnostic method of claim 9, further comprising determining the degree of separation between the diagnosis and the plurality of incorrect diagnoses.

11. The diagnostic method of claim 10, further comprising optimizing the degree of separation between the diagnosis and the plurality of incorrect diagnoses by randomly varying the fuzzy model within predetermined guidelines and recalculating the evaluation function.

12. The diagnostic method of claim 11, further comprising combining and evaluating deviations that exhibit favorable diagnostic performance using a genetic algorithm.

13. The diagnostic method of claim 12, further comprising validating the fuzzy model to ensure that no over-fitting of the plurality of validation cases has occurred.

14. The diagnostic method of claim 1, wherein the system comprises an aircraft engine system.

15. An automated method for diagnosing the cause of a trend shift in performance data associated with an aircraft engine system, the method comprising:

selecting a fuzzy model describing how the mean associated with a parameter shifts when a predetermined problem occurs;

achieving a diagnosis using the fuzzy model;

providing a plurality of validation cases associated with the aircraft engine system;

validating the fuzzy model using the plurality of validation cases;

providing an evaluation function; and

using the evaluation function to determine how well the fuzzy model differentiates the diagnosis from a plurality of incorrect diagnoses for the plurality of validation cases.

16. The diagnostic method of claim 15, wherein using the evaluation function to determine how well the fuzzy model differentiates the diagnosis from the plurality of incorrect diagnoses for the plurality of validation cases comprises outputting a plurality of confidence values for the diagnosis and the plurality of incorrect diagnoses.

17. The diagnostic method of claim 16, further comprising determining the degree of separation between the diagnosis and the plurality of incorrect diagnoses.

18. The diagnostic method of claim 17, further comprising optimizing the degree of separation between the diagnosis and the plurality of incorrect diagnoses by randomly varying the fuzzy model within predetermined guidelines and recalculating the evaluation function.

19. The diagnostic method of claim 18, further comprising combining and evaluating deviations that exhibit favorable diagnostic performance using a genetic algorithm.

20. The diagnostic method of claim 19, further comprising validating the fuzzy model to ensure that no over-fitting of the plurality of validation cases has occurred.

21. An automated system for diagnosing the cause of a trend shift in performance data associated with a system, the diagnostic system comprising:

a fuzzy model describing how the mean associated with a parameter shifts when a predetermined problem occurs;

means for receiving performance data associated with the system;

a first algorithm operable for applying a preferred set of piecewise regressions;

a second algorithm operable for selecting a start date;

a third algorithm operable for selecting data sets near the start date and an end date that have a predetermined size without violating normal scatter;

a fourth algorithm operable for measuring the mean shift between a plurality of samples using one or more statistical tests; and

means for combining the results with the fuzzy model to achieve a diagnosis.

22. The diagnostic system of claim 21, further comprising a fifth algorithm operable for removing outlying data points from the performance data by calculating a local standard deviation minus a data point in question and removing the data point if the data point falls outside of a predetermined range.

23. The diagnostic system of claim 21, wherein the start date provides a relatively large shift without leveling or changing direction between the start date and an end date.

24. The diagnostic system of claim 21, wherein the one or more statistical tests comprise a two-sample t-test.

25. The diagnostic system of claim 21, further comprising means for receiving a plurality of validation cases associated with the system.

26. The diagnostic system of claim 25, further comprising means for validating the fuzzy model using the plurality of validation cases.

27. The diagnostic system of claim 25, further comprising an evaluation function.

28. The diagnostic system of claim 27, wherein the evaluation function is used to determine how well the fuzzy model differentiates the diagnosis from a plurality of incorrect diagnoses for the plurality of validation cases.

29. The diagnostic system of claim 28, wherein the evaluation function is used to determine how well the fuzzy model differentiates the diagnosis from the plurality of incorrect diagnoses for the plurality of validation cases by outputting a plurality of confidence values for the diagnosis and the plurality of incorrect diagnoses.

30. The diagnostic system of claim 29, further comprising means for determining the degree of separation between the diagnosis and the plurality of incorrect diagnoses.

31. The diagnostic system of claim 30, further comprising a sixth algorithm operable for optimizing the degree of separation between the diagnosis and the plurality of incorrect diagnoses by randomly varying the fuzzy model within predetermined guidelines and recalculating the evaluation function.

32. The diagnostic system of claim 31, further comprising a genetic algorithm operable for combining and evaluating deviations that exhibit favorable diagnostic performance.

33. The diagnostic system of claim 32, further comprising means for validating the fuzzy model to ensure that no over-fitting of the plurality of validation cases has occurred.

34. The diagnostic system of claim 21, wherein the system comprises an aircraft engine system.

35. An automated system for diagnosing the cause of a trend shift in performance data associated with an aircraft engine system, the diagnostic system comprising:

a fuzzy model describing how the mean associated with a parameter shifts when a predetermined problem occurs;

means for achieving a diagnosis using the fuzzy model;

means for receiving a plurality of validation cases associated with the aircraft engine system;

means for validating the fuzzy model using the plurality of validation cases; and

an evaluation function, wherein the evaluation function is used to determine how well the fuzzy model differentiates the diagnosis from a plurality of incorrect diagnoses for the plurality of validation cases.

36. The diagnostic system of claim 35, wherein the evaluation function is used to determine how well the fuzzy model differentiates the diagnosis from the plurality of incorrect diagnoses for the plurality of validation cases by outputting a plurality of confidence values for the diagnosis and the plurality of incorrect diagnoses.

37. The diagnostic system of claim 36, further comprising means for determining the degree of separation between the diagnosis and the plurality of incorrect diagnoses.

38. The diagnostic system of claim 37, further comprising an algorithm operable for optimizing the degree of separation between the diagnosis and the plurality of incorrect diagnoses by randomly varying the fuzzy model within predetermined guidelines and recalculating the evaluation function.

39. The diagnostic system of claim 38, further comprising a genetic algorithm operable for combining and evaluating deviations that exhibit favorable diagnostic performance.

40. The diagnostic system of claim 39, further comprising means for validating the fuzzy model to ensure that no over-fitting of the plurality of validation cases has occurred.