Distributed stress wave analysis system

• US 6,351,713 B1
• Filed: 12/15/1999
• Issued: 02/26/2002
• Est. Priority Date: 12/15/1999
• Status: Expired due to Term

• Current Assignee: Curtiss-Wright Flow Control Corporation (Curtiss-Wright Corporation)
• Original Assignee: Swantech, LLC
• Inventors: Caulfield, Michael, Andrews, Sylvia, Hughes, Craig, Board, David B.
• Primary Examiner(s): ASSOUAD, PATRICK J
• Application Number:
  US09/464,018
• Time in Patent Office: 804 Days
• Field of Search: 702/42, 702/34, 702/35, 702/36, 702/182, 702/183, 702/184, 702/185, 702/81, 702/84, 702/56, 702/179, 706/912, 706/913, 706/914, 706/915
• US Class Current: 702/42
• CPC Class Codes: G01M 13/028 Acoustic or vibration analysis
• CPC Class Codes: G01N 2291/015 Attenuation, scattering
• CPC Class Codes: G01N 2291/106 one or more transducer arrays
• CPC Class Codes: G01N 29/14 using acoustic emission techniques echo of particles G01N29/046; measuring mechanical vibrations or acoustic waves in solids in general G01H1/00
• CPC Class Codes: G01N 29/4463 Signal correction, e.g. distance amplitude correction [DAC], distance gain size [DGS], noise filtering
• CPC Class Codes: G01N 29/4481 Neural networks
• CPC Class Codes: G01N 29/46 by spectral analysis, e.g. Fourier analysis or wavelet analysis spectral signal processing per se G06F17/14
• CPC Class Codes: G01N 29/52 using inversion methods other that spectral analysis, e.g. conjugated gradient inversion
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Abstract

A distributed stress wave analysis system is disclosed for detecting structure borne sounds cause by friction. The detected information is processed using feature extraction and neural network artificial intelligence software. The system consists of stress wave sensors, interconnect cables, and preferably three modules: (1) distributed processing units, (2) maintenance advisory panel, and (3) laptop computer. A derived stress wave pulse train which is independent of background levels of vibration and audible noise is used to extract signature features, which when processed by neural networks of polynomial equations, characterize the mechanical health of the monitored components. The system includes an adjustable data fusion architecture to optimize indication thresholds, maximize fault detection probability, and minimize false alarms.

Claims

1. A distributed stress wave analysis system for identifying and analyzing a fault in one or more components of a machine, comprising:

means for detecting a fault in at least one component of a machine, said means for detecting including an anomaly detection network for identifying abnormal features in acquired data followed by the performance of a first X of N confidence test, said means for detecting also including a fault detection network triggered by said anomaly detection network and the first X of N confidence test to confirm or reject a possibly discrepant condition identified by said anomaly detection network, said fault detection network performing a second X of N confidence test for confirming or rejecting the discrepant condition identified by said anomaly detection network;

means for locating the detected fault; and

means for isolating a source for the fault.

2. The distributed stress wave analysis system of claim 1 further comprising means for calculating a percent degradation of said at least one component.

3. The distributed stress wave analysis system of claim 2 further comprising means for displaying the calculated percent degradation.

4. The distributed stress wave analysis system of claim 1 further comprising means for estimating a remaining useful life for said at least one component.

5. The distributed stress wave analysis system of claim 1 wherein said first source is a gear source and said second source is a bearing source.

6. The distributed stress wave analysis system of claim 1 wherein said system is implemented with data fusion architecture to process multiple data inputs into reliable and accurate decisions.

7. The distributed stress wave analysis system of claim 6 wherein said data fusion architecture optimizes indication thresholds, maximizes fault detection probability and minimizes false alarms.

26. The distributed stress wave analysis system of claim 1 wherein said means for locating is triggered only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

27. The distributed stress wave analysis system of claim 1 wherein said means for isolating is triggered only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

39. The distributed stress wave analysis system of claim 1 wherein said means for detecting including feature extraction software to intelligently compress a large amount of data acquired from machinery mounted instrumentation into a small amount of information that accurately characterizes the raw data, said extracted features serving as inputs to said anomaly detection network.

40. The distributed stress wave analysis system of claim 1 wherein said fault detection network incorporating features extracted from multiple sensors.

42. The distributed stress wave analysis system of claim 1 wherein said anomaly detection network having one or more classification test each having an output determined by a variable passing through a limit threshold.

43. The distributed stress wave analysis system of claim 42 wherein successive classification test outputs are subjected to the first X of N confidence test requiring that the classification test output remains the same for X out of N measurement and test cycles.

44. The distributed stress wave analysis system of claim 42 wherein a “

X”

parameter of the first X of N confidence test, a “

N”

parameter of the first X of N confidence test, and the limit threshold for each limit threshold are adjustable.

8. A method for identifying and analyzing a fault in one or more components of a machine, comprising the following steps:

(a) detecting a fault in at least one component of a machine, said step of detecting a fault including the steps of identifying abnormal features in acquired data by a anomaly detection network followed by the performance of a first X of N confidence test, triggering a fault detection network to confirm or reject a possibly discrepant condition identified by said anomaly detection network and the first X of N confidence test, said fault detection network performing a second X of N confidence test for confirming or rejecting the discrepant condition identified by said anomaly detection network;

(b) locating the detected fault; and

(c) isolating a source for the fault.

9. The method of identifying and analyzing a fault of claim 8 further comprising the step of calculating a percent degradation of said at least one component.

10. The method of identifying and analyzing a fault of claim 9 further comprising the step of displaying the calculated percent degradation.

11. The method of identifying and analyzing a fault of claim 9 further comprising the step of estimating a remaining useful life for said at least one component.

12. The method of identifying and analyzing a fault of claim 8 wherein said method employs diagnostic techniques Utilization Interval, Decision Threshold, and an “

X of N”

Confidence Test, to achieve specified system accuracy in terms of probability of detection and probability of false alarm.

13. The method of identifying and analyzing a fault of claim 8 wherein said method employs an “

X of N”

Confidence Test, to reduce system sensitivity to transient conditions and reduce probability of false alarm.

28. The method for identifying and analyzing a fault of claim 8 wherein said detected fault it located only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

29. The method for identifying and analyzing a fault of claim 8 wherein the source of the fault is isolated only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

14. A distributed stress wave analysis system for identifying and analyzing a fault in one or more components of a machine, comprising:

at least one sensor;

at least one distributed processing unit in communication with said at least sensor, said at least one distributed processing unit extracting shock and friction information received from said at least one sensor, wherein said at least one distributed processing unit performs extraction of feature information relating to time domain and frequency domain characteristics of the shock and friction raw signals received from said at least one sensor and compresses the raw data into extracted feature information;

a maintenance advisory panel in communication with said at least one distributed processing unit, and

a computer.

15. The distributed stress wave analysis system of claim 14 wherein said at least one sensor is a plurality of sensor groups and said at least one distributed processing unit is a plurality of distributed processing units, each of said sensor groups in communication with a corresponding one of said plurality of distributed processing units, wherein the plurality of distributed processing units are interfaced to each other by said maintenance advisory panel.

16. The distributed stress wave analysis system of claim 15 wherein said maintenance advisory panel having a visual indicator for each of said plurality of distributed processing units to provide a visual indication of a detected fault by one or more of the distributed processing units.

17. The distributed stress wave analysis system of claim 14 wherein said computer is a laptop computer.

18. The distributed stress wave analysis system of claim 14 wherein said system is implemented with data fusion architecture to process multiple data inputs into reliable and accurate decisions.

19. The distributed stress wave analysis system of claim 18 wherein said data fusion architecture optimizes indication thresholds, maximizes fault detection probability and minimizes false alarms.

32. The distributed stress wave analysis system of claim 14 wherein said at least one distributed processing unit is located remote of said computer.

20. A method for identifying and analyzing a fault in one or more components of a machine, comprising the following steps:

(a) extracting friction and shock signals from broadband noise received from at least one sensor associated with at least one monitored component of a machine;

(b) detecting abnormal friction/shock signatures from the extracted signal, said step of detecting abnormal friction/shock signatures including the steps of identifying a possibly discrepant condition by an anomaly detection network and performance of a first X of N confidence test, which when passed triggers a fault detection network to confirm or reject the possibly discrepant condition identified by said anomaly detection network and the first X of N confidence test, and confirming or rejecting the discrepant condition identified by said fault detection network by performing a second X of N confident test;

(c) locating a fault causing the detected abnormal friction/shock signature;

(d) isolating a cause of the fault to a gear source or a bearing source.

21. The method of identifying and analyzing a fault of claim 20 further comprising the step of calculating a percent degradation of the at least one component detected to contain the fault.

22. The method of identifying and analyzing a fault of claim 21 further comprising the step of displaying the calculated percent degradation.

23. The method of identifying and analyzing a fault of claim 20 further comprising the step of estimating a remaining useful life for the at least one component detected to contain the fault.

24. The method of identifying and analyzing a fault of claim 20 wherein said method employs diagnostic techniques Utilization Interval, Decision Threshold, and an “

X of N”

Confidence Test, to achieve specified system accuracy in terms of probability of detection and probability of false alarm.

25. The method of identifying and analyzing a fault of claim 20 wherein said method employs an “

X of N”

Confidence Test, to reduce system sensitivity to transient conditions and reduce probability of false alarm.

30. The method for identifying and analyzing a fault of claim 20 wherein said detected fault it located only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

31. The method for identifying and analyzing a fault of claim 20 wherein a source of the fault is isolated only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

33. A distributed stress wave analysis system for identifying and analyzing a fault in one or more components of a machine, comprising:

means for detecting a fault in at least one component of a machine, said means for detecting including feature extraction software to intelligently compress a large amount of data acquired from machinery mounted instrumentation into a small amount of information that accurately characterizes the large amount of data, said extracted features serving as inputs to an anomaly detection network for identifying abnormal (anomalous) features in the data acquired from machinery mounted instrumentation, followed by the performance of a first X of N confidence test, said means for detecting also including a fault detection network triggered by said anomaly detection network and its associated X of N confidence test to confirm or reject a possible discrepant condition identified by said anomaly detection network, said fault detection network performing a second X of N confidence test for confirming or rejecting the discrepant condition identified by said anomaly detection network;

means for locating the detected fault; and

means for isolating a source for the fault.

35. The distributed stress wave analysis system of claim 33 wherein said means for locating is triggered only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

36. The distributed stress wave analysis system of claim 33 wherein said means for isolating is triggered only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

41. The distributed stress wave analysis system of claim 33 wherein said fault detection network incorporating features extracted from multiple sensors.

45. The distributed stress wave analysis system of claim 33 wherein said anomaly detection network having one or more classification test each having an output determined by a variable passing through a limit threshold.

46. The distributed stress wave analysis system of claim 45 wherein successive classification test outputs are subjected to the first X of N confidence test requiring that the classification test output remains the same for X out of N measurement and test cycles.

47. The distributed stress wave analysis system of claim 45 wherein a “

X”

parameter of the first X of N confidence test, a “

N”

parameter of the first X of N confidence test, and the limit threshold for each limit threshold are adjustable.

34. A method for identifying and analyzing a fault in one or more components of a machine, comprising the following steps:

(a) detecting a fault in at least one component of a machine, said step of detecting a fault including the steps of running feature extraction software to intelligently compress a large amount of data acquired from machinery mounted instrumentation into a small amount of information that accurately characterizes the large amount of data, said extracted features serving as inputs to an anomaly detection network to identify abnormal (anomalous) features in the data acquired from the machinery mounted instrumentation, followed by the performance of a first X of N confidence test, triggering a fault detection network to confirm or reject a possible discrepant condition identified by said anomaly detection network and its associated X of N confidence test, said fault detection network performing a second X of N confidence test for confirming or rejecting the discrepant condition identified by said anomaly detection network;

(b) locating the detected fault; and

(c) isolating a source for the fault.

37. The method for identifying and analyzing a fault of claim 34 wherein said detected fault it located only after said fault detection network and the second X of N confidence test confirms the discrepant condition.

38. The method for identifying and analyzing a fault of claim 34 wherein the source of the fault is isolated only after said fault detection network and the second X of N confidence test confirms the discrepant condition.