Alert generation for trend performance analysis

• US 6,408,259 B1
• Filed: 02/01/2000
• Issued: 06/18/2002
• Est. Priority Date: 02/01/2000
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Goebel, Kai Frank, Doel, David Lacey
• Primary Examiner(s): Shah, Kamini
• Assistant Examiner(s): VO, HIEN XUAN
• Application Number:
  US09/492,685
• Time in Patent Office: 868 Days
• Field of Search: 702/183, 702/182, 702/185, 702/179, 702/104, 701/34, 701/35, 701/40, 701/29, 701/99, 701/100, 701/106, 706/900, 706/905, 706/906
• US Class Current: 702/183
• CPC Class Codes: B64D 43/00 Arrangements or adaptations of instruments arrangements of cameras B64D47/08; aeronautical measuring instruments per se G01C
• CPC Class Codes: G05B 23/0232 based on qualitative trend analysis, e.g. system evolution
• CPC Class Codes: G05B 23/0297 Reconfiguration of monitoring system, e.g. use of virtual sensors; change monitoring method as a response to monitoring results
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles

Abstract

This invention provides an improved alert generation system and method for trend performance analysis. A multi-variate abnormal condition detector normalizes data obtained from a process and classifies the data in a multi-dimensional space defined for the variables in the process. The normalized data are classified into either a normal or abnormal class. An alert evaluator evaluates the data classified in the abnormal class and determines whether the data are a true alert or a false alert. In another embodiment, the multi-variate abnormal condition detector and alert evaluator are used in parallel with a trend performance analysis tool to validate alerts generated therefrom.

Claims

1. A system for generating an alert from data obtained from a process, comprising:

a normalizer for normalizing the data;

a classifier for classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the process and the abnormal class is indicative of potential alert conditions in the process;

an alert level evaluator for evaluating a vigilance level of the classified data for related variables, said alert level evaluator increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

an alert score generator for generating an alert score for each of the related variables according to the vigilance level.

2. The system according to claim 1, wherein said normalizer comprises means for centering the data about a selected center point.

3. The system according to claim 1, wherein said classifier uses a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

4. The system according to claim 1, further comprising a statistical property estimator for estimating a plurality of statistical properties of the obtained data.

5. The system according to claim 4, wherein said statistical property estimator estimates a mean and standard deviation of the classified data.

6. The system according to claim 4, further comprising an adaptor for smoothing the statistical properties estimated by said statistical property estimator.

7. The system according to claim 1, wherein the alert condition is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

8. The system according to claim 1, wherein the alert score generator aggregates each of the alert scores for the related variables into a composite score.

9. A system for generating an alert from data obtained from an aircraft engine process, comprising:

a normalizer for normalizing the data obtained from said aircraft engine process;

a classifier for classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the aircraft engine process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the aircraft engine process and the abnormal class is indicative of potential alert conditions in the aircraft engine process;

an alert level evaluator for evaluating a vigilance level of the classified data for related variables, said alert level evaluator increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

an alert score generator for generating an alert score for each of the related variables according to the vigilance level.

10. The system according to claim 9, wherein said normalizer comprises means for centering the data about a selected center point.

11. The system according to claim 9, wherein said classifier uses a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

12. The system according to claim 9, further comprising a statistical property estimator for estimating a plurality of statistical properties of the obtained data.

13. The system according to claim 12, wherein said statistical property estimator estimates a mean and standard deviation of the classified data.

14. The system according to claim 12, further comprising an adaptor for smoothing the statistical properties estimated by said statistical property estimator.

15. The system according to claim 9, wherein the alert condition is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

16. The system according to claim 9, wherein the alert score generator aggregates each of the alert scores for the related variables into a composite score.

17. The system according to claim 9, wherein the multi-dimensional variable space comprises engine variables comprising exhaust gas temperature, fuel flow, core speed, exhaust gas temperature divergence, fuel flow divergence and core speed divergence.

18. A method for generating an alert from data obtained from a process, comprising:

normalizing the data;

classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the process and the abnormal class is indicative of potential alert conditions in the process;

evaluating a vigilance level of the classified data for related variables, said evaluating comprising increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

generating an alert score for each of the related variables according to the vigilance level.

19. The method according to claim 18, wherein said normalizing comprises centering the data about a selected center point.

20. The method according to claim 18, wherein said classifying comprises using a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

21. The method according to claim 18, further comprising estimating a plurality of statistical properties of the obtained data.

22. The method according to claim 21, wherein said estimating estimates a mean and standard deviation of the classified data.

23. The method according to claim 20, further comprising smoothing the estimated statistical properties.

24. The method according to claim 18, wherein the alert is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

25. The method according to claim 18, wherein said generating an alert score comprises aggregating each of the alert scores for the related variables into a composite score.

26. A method for generating an alert from data obtained from an aircraft engine process, comprising:

normalizing the data obtained from said aircraft engine process;

classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the aircraft engine process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the aircraft engine process and the abnormal class is indicative of potential alert conditions in the aircraft engine process;

evaluating a vigilance level of the classified data for related variables, said evaluating comprising increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

generating an alert score for each of the related variables according to the vigilance level.

27. The method according to claim 26, wherein said normalizing comprises centering the data about a selected center point.

28. The method according to claim 26, wherein said classifying comprises using a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

29. The method according to claim 26, further comprising estimating a plurality of statistical properties of the data.

30. The method according to claim 29, wherein said estimating estimates a mean and standard deviation of the classified data.

31. The method according to claim 26, further comprising smoothing the estimated statistical properties.

32. The method according to claim 26, wherein the alert is representative of at-least one of a failure, a fault condition, or symptoms of an incipient failure condition.

33. The method according to claim 26, wherein said generating an alert score comprises aggregating each of the alert scores for the related variables into a composite score.

34. The method according to claim 26, wherein the multi-dimensional variable space comprises engine variables comprising exhaust gas temperature, fuel flow, core speed, exhaust gas temperature divergence, fuel flow divergence and core speed divergence.

35. A system for validating an alert generated from a trend performance analysis tool used to monitor data obtained from a process, comprising:

a normalizer for normalizing the data monitored by the trend performance analysis tool;

a classifier for classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the process and the abnormal class is indicative of alert conditions in the process;

an alert level evaluator for evaluating a vigilance level of the classified data for related variables, said alert level evaluator increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

an alert score generator for generating an alert score for each of the related variables according to the vigilance level, wherein the alert generated from the trend performance analysis tool is valid if the alert score satisfies a predetermined score.

36. The system according to claim 35, wherein said normalizer comprises means for centering the data about a selected center point.

37. The system according to claim 35, wherein said classifier uses a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

38. The system according to claim 35, further comprising a statistical property estimator for estimating a plurality of statistical properties of the data.

39. The system according to claim 38, wherein said statistical property estimator estimates a mean and standard deviation of the classified data.

40. The system according to claim 38, further comprising an adaptor for smoothing the statistical properties estimated by said statistical property estimator.

41. The system according to claim 35, wherein the alert condition is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

42. The system according to claim 35, wherein the alert score generator aggregates each of the alert scores for the related variables into a composite score.

43. The system according to claim 35, further comprising an alert validator that compares the alert from the trend performance analysis tool with the alert score.

44. A system for validating an alert generated from a trend performance analysis tool used to monitor data obtained from an aircraft engine process, comprising:

a normalizer for normalizing data obtained from said aircraft engine process and monitored by the trend performance analysis tool;

a classifier for classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the aircraft engine process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the aircraft engine process and the abnormal class is indicative of potential alert conditions in the aircraft engine process;

an alert level evaluator for evaluating a vigilance level of the classified data for related variables, said alert level evaluator increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

an alert score generator for generating an alert score for each of the related variables according to the vigilance level, wherein the alert generated from the trend performance analysis tool is valid if the alert score satisfies a predetermined score.

45. The system according to claim 44, wherein said normalizer comprises means for centering the data about a selected center point.

46. The system according to claim 44, wherein said classifier uses a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

47. The system according to claim 44, further comprising a statistical property estimator for estimating a plurality of statistical properties of the data.

48. The system according to claim 47, wherein said statistical property estimator estimates a mean and standard deviation of the classified data.

49. The system according to claim 47, further comprising an adaptor for smoothing the statistical properties estimated by said statistical property estimator.

50. The system according to claim 44, wherein the alert condition is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

51. The system according to claim 44, wherein the alert score generator aggregates each of the alert scores for the related variables into a composite score.

52. The system according to claim 44, wherein the multi-dimensional variable space comprises engine variables comprising exhaust gas temperature, fuel flow, core speed, exhaust gas temperature divergence, fuel flow divergence and core speed divergence.

53. The system according to claim 44, further comprising an alert validator that compares the alert from the trend performance analysis tool with the alert score.

54. A method for validating an alert generated from a trend performance analysis tool used to monitor data obtained from a process, comprising:

normalizing the data monitored by the trend performance analysis tool;

classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the process and the abnormal class is indicative of potential alert conditions in the process;

evaluating a vigilance level of the classified data for related variables, said evaluating comprising increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

generating an alert score for each of the related variables according to the vigilance level, wherein the alert generated from the trend performance analysis tool is valid if the alert score satisfies a predetermined score.

55. The method according to claim 54, wherein said normalizing comprises centering the data about a selected center point.

56. The method according to claim 54, wherein said classifying comprises using a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

57. The method according to claim 54, further comprising estimating a plurality of statistical properties of the data.

58. The method according to claim 57, wherein said estimating estimates a mean and standard deviation of the classified data.

59. The method according to claim 57, further comprising smoothing the estimated statistical properties.

60. The method according to claim 54 wherein the alert is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

61. The method according to claim 54, wherein said generating an alert score comprises aggregating each of the alert scores for the related variables into a composite score.

62. The method according to claim 54, further comprising comparing the alert from the trend performance analysis tool with the alert score.

63. A method for validating an alert generated from a trend performance analysis tool used to monitor data obtained from an aircraft engine process, comprising:

normalizing the data obtained from said aircraft engine process and monitored by the trend performance analysis tool;

classifying the normalized data in a multi-dimensional space defined for a plurality of variables in the aircraft engine process, the normalized data classified into a normal class or an abnormal class defined in the multi-dimensional variable space, wherein the normal class is indicative of normal operating conditions for the aircraft engine process and the abnormal class is indicative of potential alert conditions in the aircraft engine process;

evaluating a vigilance level of the classified data for related variables, said evaluating comprising increasing the vigilance level following a suspicious data reading data classified in the abnormal class and decreasing the vigilance level following a data reading classified in the normal class; and

generating an alert score for each of the related variables according to the vigilance level, wherein the alert generated from the trend performance analysis tool is valid if the alert score satisfies a predetermined score.

64. The method according to claim 63, wherein said normalizing comprises centering the data about a selected center point.

65. The method according to claim 63, wherein said classifying comprises using a fuzzy knn algorithm to classify data in the normal class and the abnormal class.

66. The method according to claim 63, further comprising estimating a plurality of statistical properties of the data.

67. The method according to claim 66, wherein said estimating estimates a mean and standard deviation of the classified data.

68. The method according to claim 66, further comprising smoothing the estimated statistical properties.

69. The method according to claim 63, wherein the alert is representative of at least one of a failure, a fault condition, or symptoms of an incipient failure condition.

70. The method according to claim 63, wherein said generating an alert score comprises aggregating each of the alert scores for the related variables into a composite score.

71. The method according to claim 63, wherein the multi-dimensional variable space comprises engine variables comprising exhaust gas temperature, fuel flow, core speed, exhaust gas temperature divergence, fuel flow divergence and core speed divergence.

72. The method according to claim 63, further comprising comparing the alert from the trend performance analysis tool with the alert score.