System for prognosis and diagnostics of failure and wearout monitoring and for prediction of life expectancy of helicopter gearboxes and other rotating equipment

• US 5,210,704 A
• Filed: 10/02/1990
• Issued: 05/11/1993
• Est. Priority Date: 10/02/1990
• Status: Expired due to Fees

• Current Assignee: Technology International Incorporated
• Original Assignee: Technology International Incorporated
• Inventors: Husseiny, Abdo A.
• Primary Examiner(s): Lall, Parshotam S.
• Assistant Examiner(s): PIPALA, EDWARD J
• Application Number:
  US07/592,057
• Time in Patent Office: 952 Days
• Field of Search: 364/550, 364/551.01, 364/506, 364/507, 364/508, 395/911-913, 73/116, 73/117.3
• US Class Current: 702/34
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: Y10S 706/912 Manufacturing or machine, e.g. agricultural machinery, machine tool

Abstract

A wearout monitor for failure prognostics is a prognosis tool to predict incipient failure in rotating mechanical equipment. The wearout monitor provides maintenance management of a plant or process with information essential to planning preventive maintenance strategies. The monitor also assists in constructing a data base for development and implementation of policies for plant life extension, refurbishment, and modernization. The apparatus identifies systems of operation degradation of the whole system, as well as diagnosis of signs of commencing aging cycles of specific equipment, components or parts of equipment during operation. Data from the system is stored and also supplied to a central processing unit which includes an expert system, rule-based failure data bank, a predictor, a performance evaluator and a system identifier. The results of the predictions are supplied to management terminals or other indicators for subsequent use. Combination of prognostics and diagnostics of the symptoms of existing fault in mechanical equipment allows continuous on-line monitoring of systems to predict failures at early stages before leading to catastrophic breakdown and to assure safe and economic operation. By providing correlations between defect sizes and life expectancy of a rotating mechanical component, the monitor can provide the operator of the equipment with a warning time that indicates the time before loss of operation, thereby being critical to operation of transport systems wherein gearboxes can lead to loss of transmission power and subsequent loss of life particularly in helicopters.

Claims

1. A wearout monitor and control device for failure prognostics comprising:

a component susceptible to failure;

a system data acquiring means for acquiring system data relating to the operation of said component;

a data processing means for processing data received from said system data acquiring means relating to the operation of said component, said data processing means including a performance evaluation means for evaluating the performance of said component and a failure predicting means for predicting failure of said component;

a failure indicating means for indicating a failure predicted by said failure predicting means; and

means for regulating further operation of said component in response to a failure being predicted by said failure predicting means.

2. A wearout monitor as claimed in claim 1, wherein the means of acquiring system data comprise sensors for measuring data related to said component being monitored and sensors for measuring overall system parameters.

3. A wearout monitor as claimed in claim 2, wherein data received by said system data acquiring means is received by a multiband pass filter.

4. A wearout monitor as claimed in claim 3, further comprising a sample-and-hold means for receiving signals form said multiband pass filter.

5. A wearout monitor as claimed in claim 4, further comprising a multiplexer-and-switch means for receiving signals from said sample-and-hold means, wherein said multiplexer-and-switch means supplied signals to an analog-to-digital converter means, which in turn supplies signals to a data storing means for storing data representative of the state of said component.

6. A wearout monitor as claimed in claim 1, wherein data processing means includes a performance evaluation means for evaluating the performance of said component, and a failure predicting means for predicting failure of the component.

7. A wearout monitor as claimed in claim 6, wherein said performance evaluation means includes fitting said system data with an empirical time series model and establishing a model parameter of said model with a dynamic estimator means for filtering the noise in said system data.

8. A wearout monitor as claimed in claim 7, wherein said model parameter provides indication of the dynamic behavior of the component.

9. A wearout monitor as claimed in claim 1, wherein said performance evaluation means analyzes the performance of a plurality of components.

10. A wearout monitor as claimed in claim 1, wherein said failure predicting means includes an expert shell means for failure conditions and inference of failure states from a failure data bank.

11. A wearout monitor as claimed in claim 1, wherein said failure predicting means includes a rule-based failure data bank.

12. A wearout monitor as claimed in claim 1, wherein said failure indicating means includes at least one display terminal.

13. A wearout monitor as claimed in claim 1, further comprising a data storing means for storing data, wherein said data storing means receives data from said system data acquiring means and supplies the data to the data processing means.

14. A wearout monitor as claimed in claim 7, wherein said failure predicting means comprises means for matching said empirical model parameters with rule-based failure data.

15. A wearout monitor as claimed in claim 10, wherein said expert system identifies the nature and location of a failure.

16. A method for predicting failure of and regulating components in a system during a continuous operation of the components, comprising the steps of:

providing at least one component to be monitored;

providing a means for receiving data relating to the operation of said component;

providing a means for processing the data relating to the operation of said component, said means for processing the data including an expert system, a means for predicting failure, and a performance evaluating means;

providing a means for indicting predicted failure of said component;

receiving data indicative of the operation of said component with said means for receiving data;

processing data received from said means for receiving data by said means for processing data including evaluating the performance of said component by said means for evaluating performance, and analyzing the data to predict failure of said component by said means for prediction failure;

indicating predicted failure of said component by said means for indicating predicted failure; and

regulating further operation of said component in response to a predicted failure.

17. A method as claimed in claim 16, further comprising providing a means for component identification.

18. A method as claimed in claim 17, further comprising providing a rule-based failure data bank.

19. A method as claimed in claim 16, further providing a template means and applying said template means to the processed data to determine a match for prediction of expected local failures, expected global failures, and for indicating whether no match has occurred.

20. A method as claimed in claim 16, wherein said means for data processing conducts a trend analysis of predictions, said trend analysis of predictions includes comparing extrapolated predictions with prior extrapolated predictions, checking the trend of the comparison, and supplying information for matching.

21. A method as claimed in claim 20, further comprising providing a rule-based failure data bank;

and applying information from said rule-based failure data bank during checking of a trend relating to the prior extrapolated predictions.

22. A method as claimed in claim 21, further comprising providing a means for updating rules and updating rules in said rule-based failure data bank.

23. A method as claimed in claim 22, wherein said step of updating rules includes:

providing a means for supplying signals to said means for updating rules; and

,

supplying signals to said means for updating rules to cause updating of said rule-based failure data bank.

24. A method of predicting incipient failure of and regulating the operation of rotating mechanical components, comprising the steps of:

providing at least one component to be monitored;

developing knowledge based models of a system employing said component;

examining the process design and analyzing failure modes, causes of failure, and the effects of failure on the system;

examining operation procedures constructing rule-based operation rules indicative of failure modes, causes of failure and effects of failure;

monitoring said component and said system and predicting failure of said component and said system; and

regulating further operation of said system in response to a failure being predicted.

25. A method as claimed in claim 24, wherein said analyzing failure modes, causes of failure, and the effects of failure are performed on the overall system and said component including the effects of component failure on the overall system.

26. A method as claimed in claim 24, wherein said constructing rule-based operation rules relates changes in instrumentation signals to the failure modes, failure causes, and failure effects.

27. A method as claimed in claim 24, further comprising:

synthesizing an empirical time series model from measured signals;

estimating the parameters of the empirical model through time series analysis of measured data using a dynamic estimator;

predicting the optimum dynamic behavior of the system and the component by evaluating the noise covariance matrix and dynamic estimator gain;

employing a feature matching algorithm to identify the empirical model parameters from a knowledge-based model; and

using a rule-based expert system to identify the nature and location of failure.

28. A method as claimed in claim 24, further comprising performing process simulation and analysis scenarios of anticipated failures.

29. A method as claimed in claim 28, constructing matching templates of failure from simulation results.

30. A method as claimed in claim 24, constructing matching templates of failure from said rule-based operation rules.

31. An apparatus for on-line diagnosis of defects and failure, prognosis of incipient failures and controlling of mechanical systems having rotating equipment, comprising:

at least one component of a system susceptible to incipient failure;

two independent channel means for acquisition of signals from global sensors and local sensors;

a data acquisition means for processing the signals from the said channel means for use by a central processing unit;

said central processing unit including means for analyzing the data input from said data acquisition means relating to the operation of said component;

analyzing means for identifying abnormalities in said data input in comparison with reference normal data compiled in a knowledge base means stored in said central processing unit;

identifying means for identifying the type and extent of the abnormality relating to the operation of said component by the aid of said knowledge base means;

display means for displaying the abnormalities identified by said identifying means and relating to the operation of said component; and

control means connected to the system for controlling the operation of the system and shutting down the operation of the system in response to an abnormality detrimental to the operation of the system.

32. Apparatus as set forth in claim 31, wherein said data acquisition means comprises:

a multiband pass filter means for processing signals according to predefined frequency bands;

a means for sampling data according to predetermined sampling intervals;

a hold means, a multiplexer means and switch means for managing the flow of data to the central processing unit means in batches appropriate for processing according to predetermined data acquisition strategy; and

,

a converter means for converting analog signals to digital signals.

33. Apparatus as set forth in claim 31, wherein analysis means comprises a fast Fourier transform analyzer and a time series analysis module.

34. Apparatus as set forth in claim 33, wherein said analysis means provides indices of normal and abnormal behavior in the operation of said component and generates empirical models from said data of said component to be monitored.

35. Apparatus as set forth in claim 34, wherein said identifying means comprises:

means for identifying component abnormalities from said indices and said models;

pattern recognition means for comparison of abnormalities with said knowledge base;

prognostic means for predicting faults before their occurrence;

diagnostic means for detecting failures and monitoring symptoms of potential faults; and

,

means for localizing failure by attributing the failure to one of a plurality of components,

wherein said prognostic and diagnostic means operate independently and in parallel as a redundant detection system.

36. Apparatus as set forth in claim 35, further comprising a pattern recognition means for recognizing patterns including templates of defect patterns.

37. Apparatus as set forth in claim 35, wherein said comparison involves identifying minimum distance between failure indices and said templates of defect patterns.

38. Apparatus as set forth in claim 31, wherein said knowledge base means is continuously updated with new failure data and includes a self learning means for upgrading the knowledge base.

39. Apparatus as set forth in claim 38, wherein the knowledge base includes a knowledge base for diagnosis, a knowledge base for prognosis and maintenance management means.

40. Apparatus as set forth in claim 35, wherein said identifying means includes a diagnostic logic means and prognostic logic means.

41. Apparatus as set forth in claim 40, wherein said diagnostic logic means and said prognostic logic means provide diagnosis and prognosis messages respectively indicating trends, symptoms and signals of incipient failures of said component.

42. Apparatus as set forth in claim 41, wherein said messages and said maintenance management means are used by a decision means to determine the type and location of actual faults or incipient failure of said component.

43. Apparatus as set forth in claim 31, wherein said display means comprises a printing means, a video display means and graphics display means.

44. A method for on-line diagnosis and prognosis of failure in rotating mechanical equipment and controlling the operation of the equipment, comprising:

providing mechanical components constituting mechanical equipment which is susceptible to failure;

sensing local conditions of specific components and global conditions of a system containing said components;

monitoring said components by performing analysis of signals representing said local and global conditions;

performing time series analysis and signal analysis on said signals to provide empirical model identifying equipment operating conditions relative to preselected indices appropriate for representation of normal and abnormal conditions of the operation of said equipment;

performing diagnosis and prognosis independently using separate knowledge bases and using pattern recognition to recognize symptoms of failure or incipient failure in any one of said components;

producing messages indicative of changes in trends or of identified symptoms of the operation of said equipment; and

regulating operation of said equipment if the changes in trends or in identified symptoms relate to a potentially failing component.

45. The method of claim 44, further comprising, providing a learning knowledge base for continuous updating of stored data and for constructing new templates for pattern recognition.

46. The method of claim 44, wherein the step of monitoring includes:

global monitoring of said components using a global database comprising data relating to operating conditions to provide diagnosis and prognosis fault messages as they relate to said components;

fault searching using a rule base formed by fault tree analysis associating predetermined faults to certain equipment, assigning probabilities to each fault, and tracing the root causes of major faults identified by said global monitoring; and

,

local monitoring of said components within said equipment using a local database comprising data relating to conditions of said components and said fault searching to provide diagnosis and prognosis fault messages as they relate to said components of said equipment.

47. The method of claim 44, further comprising the steps of:

acquiring signal analysis data from the sensors of the equipment;

setting up the diagnostic and prognostic procedure;

implementing an on-line monitoring method iteratively by searching for abnormal conditions;

evaluating the pattern of each index, and evaluating different matched filters for different indices;

performing prognosis with a first prognostic knowledge base;

performing diagnosis with the same logic of the prognostic step with a second knowledge base; and

shutting down the equipment and replacing a potentially defective component if the potential failures are determined to be serious.

48. The method of claim 47, wherein setting up the diagnostic and prognostic procedure comprises:

determining an adequate sample size sampling frequency, and filtering strategy of the signals data;

initiating the model parameters by fitting the signal data in normal conditions;

selecting an index vector which is sensitive to malfunctions;

calculating an index vector in normal conditions;

setting the threshold vectors for prognostics and diagnostics.

49. The method of claim 47, wherein the step of implementing the on-line monitoring method comprises the steps of:

measuring input data vectors from various sensors relating to selected components;

estimating model parameters in the monitored conditions;

updating the model parameters for the monitored conditions;

storing the model parameters in model storage files;

computing an index vector in the monitored conditions;

calculating the values of prognostic vectors and diagnostics vectors as composite vectors of the index vector of said normal conditions and index vector of said monitored conditions;

comparing said prognostic vector with a prognostic failure threshold index vector as criterion for incipient failure of any one of said components;

comparing said diagnostic vector with a diagnostic failure threshold index vector as criterion for the presence of fault;

waiting for the next sample if no abnormality evaluated by prognostics/diagnostics exists;

indicating abnormality if any one of the criteria for prognostics and diagnostics is violated; and

,

implementing said prognostic and diagnostic monitoring sequentially and in parallel.

50. The method of claim 49, wherein evaluating the pattern of each index, and evaluating different matched filters for different indices;

includes performing pattern recognition through a series of pattern evaluations in the matched filter.

51. The method of claim 50, wherein said performing prognosis and performing diagnosis comprise the steps of:

invoking prognostic functions and diagnostic functions;

retrieving the model parameters from the model storage file for predicting future faults;

predicting a future data vector;

calculating a specific prognostic index vector for failure pattern recognition;

storing and providing pattern classification of each index;

identifying and indicating all possible types of defect;

checking possible failure alternatives if one defect has been prognosed;

managing a maintenance scheme if prognostic failures are identified;

displaying the information of prognostic failures and correct compensation;

sounding an alarm when the prognostic mechanism and its knowledge base are unable to manage a potential fault; and

,

supplementing said prognostic knowledge base by the new potential fault for future usage.

52. A wearout monitoring method for predicting life expectancy of rotating equipment and controlling the operation of the equipment, comprising:

providing at least one component of the rotating equipment to be monitored;

detecting defect sizes;

analyzing anticipated failures of said components, identifying failure modes, failure causes, and failure effects;

analyzing and classifying historical data of operation and failure of the equipment and said components according to a taxonomy of failure modes, failure causes, failure effects, types of defects, and size of defects;

constructing a knowledge base from said analyzed historical data and said analyzed anticipated failures as they relate to different loading conditions;

specifying defect threshold size that results in wearout failure of said component;

constructing a life expectancy prediction model based on predetermined defect growth rates and said defect threshold size;

computing the remaining life of said component corresponding to each defect size; and

controlling the operation of the equipment in response to the remaining life of said component.

53. The method of claim 52, wherein said knowledge base contains data from dynamic simulation of the equipment and said component under different loading conditions and data derived from said dynamic simulation on life expectancy and defect growth rates.

54. The method of claim 53, wherein said knowledge base contains data derived from structural analysis of said component as it relates to local stresses and loading stresses and the associated strain leading to development of defects.

55. The method of claim 52, wherein said computing of the remaining lief is provided as a warning time that indicates the time before a catastrophic operation failure which is less than the remaining life of said component.

56. The method of claim 55, further comprising the incorporation of said warning time and prognostics of incipient failure into a warning system.

57. The method of claim 56, wherein said prognostics of incipient failure uses additional indicators from a group including decline of torque and or temperature increase in a lubricant.

58. The method of claim 54, wherein each of said structural analysis and said dynamic simulation use a finite element method.

59. The method of claim 58, wherein said finite element method provides stress and strain distribution in the component.

60. The method of claim 52, wherein said life expectancy prediction model comprises:

analyzing drift and diffusion of defects in the component;

calculating drift coefficients and diffusion coefficients from a predetermined rate of defect growth and defect size; and

computing mean times to first failure for each type of defect according to a predetermined statistical distribution of defects.

61. The method of claim 60, wherein drift coefficient is the mean of a defect size statistical distribution and said diffusion coefficient is the variance of said distribution.

62. The method of claim 60, wherein a warning time is determined to be the minimum mean time to first failure.

63. The method of claim 62, wherein the said minimum mean time to first failure is compared to a time drop of torque and to other time characteristics of operation.

64. The method of claim 63, wherein the shortest characteristic time is a time at which to abort the operation of said component.

65. An apparatus for determining on-line operation life of a component of rotating equipment before the rotating equipment can no longer transmit the required operating power and for aborting the operation of the equipment, comprising:

at least one component of the equipment having an operational life expectancy;

a chip detecting means for detecting the size of defects by the size of chips in said component and generating an analog signal indicative thereof;

a debris indicating means for indicating debris from said component of the rotating equipment and generating an analog signal indicative thereof;

means for converting analog signals from said chip detecting means and said debris indicating means;

data processing means for processing data and implementing a life expectancy prediction;

means for displaying a warning time; and

control means for aborting the operation of the equipment in response to a predetermined warning time.

66. The apparatus of claim 65, wherein said data processing means comprises test data files and means for computing drift coefficients and diffusion coefficients from statistical analysis of defect size distribution and subsequently computing a mean time to first failure associated with each defect.

67. Apparatus as set forth in claim 66, wherein said data processing means processes data from said indicator and said detector and sample relevant test data from said test data files.

68. Apparatus as set forth in claim 66, wherein said mean time to first failure of each defect are used to predict life expectancy and compute said warning time.

69. Apparatus as set forth in claim 65, wherein said means for displaying the warning time comprises a digital display means for displaying said warning time in hours, minutes and seconds and a warning light for indicating when said warning time is short.

70. Apparatus as set forth in claim 66, wherein said data processing means computes the change in parameters associated with operation of the equipment other than defect growth including torque and temperature rise.

71. Apparatus as set forth in claim 66, wherein said test data files are generated from off-line testing, engineering investigation and simulated computation of warning time as it relates to defect size and type.