FUZZY CLASSIFICATION APPROACH TO FAULT PATTERN MATCHING

• US 20090037772A1
• Filed: 08/01/2008
• Published: 02/05/2009
• Est. Priority Date: 08/03/2007
• Status: Active Grant

• Current Assignee: SmartSignal Corporation (Emerson Electric Company)
• Original Assignee: SmartSignal Corporation (Emerson Electric Company)
• Inventors: Wegerich, Stephan W., Stoecker, Chad, Marcell, Richard
• Granted Patent: US 7,941,701 B2
• Time in Patent Office: Days
• US Class Current: 714/26
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0278 Qualitative, e.g. if-then rules; Fuzzy logic; Lookup tables; Symptomatic search; FMEA
• CPC Class Codes: G06F 18/24765 Rule-based classification
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

A method and apparatus are provided for diagnosing faults in a monitored system. Estimates of parameter data are generated from the system with reference parameter data characteristic of known behavior of the system. The generated estimates of the parameter data are compared with measured parameter data. A residual is determined for each of the parameters based on a relationship between the estimates of the parameter data and the measured parameter data. A fuzzy classification rule is implemented to determine a likelihood that a predetermined fault exists by using residuals for parameters used to indicate the existence of the predetermined fault.

Claims

1. A method of diagnosing faults in a monitored system, comprising:

generating estimates of parameter data for parameters monitored on the monitored system by using reference parameter data characteristic of known behavior of the system;

comparing the generated estimates of the parameter data with measured parameter data;

determining a residual for each of the parameters based on a relationship between the generated estimates of the parameter data and the measured parameter data; and

implementing at least one fuzzy classification rule to determine a likelihood that a predetermined fault exists, the fuzzy classification rules using residuals of parameters used to indicate the existence of the predetermined fault.

2. The method of claim 1 wherein the step of implementing fuzzy classification rules further comprises the step of selecting the fuzzy classification rule from a predetermined set of fuzzy classification rules that are used for all faults and all residuals.

3. The method of claim 1 wherein the fuzzy classification rule includes at least one membership function that determines the likelihood of the existence of a residual shift relative to at least one threshold for at least one parameter used to indicate the predetermined fault.

4. The method of claim 3 wherein the fuzzy classification rule includes a membership function that determines the likelihood of at least one of:

a positive residual shift,

a negative residual shift,

positive or negative residual shift,

no positive residual shift,

no negative residual shift, and

no positive or negative residual shift.

5. The method of claim 2 wherein the step of implementing fuzzy classification rules further comprises the steps of:

obtaining an output from each membership function for parameters used to indicate the predetermined fault; and

combining the outputs in a calculation to determine a single scalar fuzzy logic value.

6. The method of claim 1 further comprising the step of retrieving a fault pattern from a fault pattern library to determine which parameters are used to indicate the predetermined fault.

7. The method of claim 1, further comprising issuing a first notification in response to implementation of the fuzzy classification rules generating a fuzzy logic value exceeding a first threshold.

8. The method of claim 7 further comprising issuing at least one more notification in response to the implementation of the fuzzy classification rules generating the fuzzy logic value exceeding at least one other threshold.

9. The method of claim 1, further comprising the steps of:

implementing the fuzzy classification rule in accordance with a hierarchy of the reference parameter data and measured parameter data, including;

initially analyzing primary parameter data, and

analyzing secondary parameter data only in response to the primary parameter data corresponding to a portion of at least one fuzzy classification rule corresponding to the predetermined fault.

10. A monitoring apparatus for diagnosing faults in a system, comprising:

a reference library containing reference parameter data characteristic of known behavior of the system;

a fuzzy logic engine configured to;

generate estimates of parameter data from the system with reference parameter data characteristic of known behavior of the system;

compare the generated estimates of the parameter data with measured parameter data;

determine a residual for each of the parameters based on a relationship between the estimates of the parameter data and the measured parameter data; and

implement at least one fuzzy classification rule to determine a likelihood that a set of outputs corresponds to a predetermined fault, wherein the outputs are generated based on the residual for each of the parameters that corresponds to the predetermined fault.

11. The monitoring apparatus of claim 10, further comprising a notification engine to issue a first notification in response to implementation of the fuzzy classification rule generating a fuzzy logic value exceeding a first threshold.

12. The monitoring apparatus of claim 10, wherein the notification engine is adapted to issue at least one more notification in response to the implementation of the fuzzy classification rule generating the fuzzy logic value exceeding at least one other threshold.

13. The monitoring apparatus of claim 10, further comprising a processor to implement the fuzzy classification rule in accordance with a hierarchy of the reference parameter data and measured parameter data, wherein primary parameter data is initially analyzed, and secondary parameter data is analyzed only in response the primary parameter data corresponding to a portion of at least one fuzzy classification rule corresponding to the predetermined fault.

14. A monitoring apparatus for diagnosing faults in a system, comprising:

a real-time data processing module that provides residuals representing a relationship between generated estimates of parameter data, and measured parameter data, the generated estimates being based on reference parameter data characteristics of known behavior of the system;

a failure mode database that stores fault patterns that each indicate a predetermined fault;

a fuzzy pattern recognition module configured for;

receiving fault patterns from the failure mode database and residuals from the real-time data processing module,

implementing at least one fuzzy classification rule that uses the residuals to calculate a set of outputs,

determining if the outputs indicate a likelihood that the predetermined fault exists; and

a diagnosis and actions output module that acts depending on whether or not a predetermined fault is determined to exist and a classification of the fault.

15. The monitoring apparatus of claim 14 wherein the fuzzy pattern recognition module comprises a set of membership functions associated with the fault patterns.