Global state change indicator for empirical modeling in condition based monitoring

• US 20020128731A1
• Filed: 02/27/2001
• Published: 09/12/2002
• Est. Priority Date: 01/19/2001
• Status: Active Grant

• Current Assignee: SmartSignal Corporation (Emerson Electric Company)
• Original Assignee: SmartSignal Corporation (Emerson Electric Company)
• Inventors: Bell, Dave R., Xu, Xiao, Wegerich, Stephan W.
• Granted Patent: US 6,859,739 B2
• Time in Patent Office: Days
• US Class Current: 700/31
• CPC Class Codes: G05B 17/02 electric
• CPC Class Codes: G05B 23/0243 model based detection method, e.g. first-principles knowledge model

Abstract

An improved empirical model-based surveillance or control system for monitoring or controlling a process or machine provides identification of transitions between operational states. Empirical model-based estimates generated in response to receiving actual operational parameters are compared using a global similarity operator to the actual parameters to indicate whether the process or machine is operating in a stable state, or is in transition from one state to another.

Claims

1. An apparatus for detecting an operational state transition in a system instrumented with sensors, comprising:

a model for generating estimates of sensor values in response to receiving actual sensor values characterizing the operation of the system;

a global similarity engine for generating a global similarity score for a comparison of a set of estimated sensor values from said model and a related set of actual sensor values from said system; and

a state transition detector disposed to receive global similarity scores from said global similarity engine and indicate an operational state change when a global similarity score falls outside of a selected range.

2. An apparatus according to claim 1 wherein the model is an empirical model.

3. An apparatus according to claim 2 wherein the empirical model employs a similarity operator.

4. An apparatus according to claim 3 wherein the empirical model comprises a reference library of snapshots of known operational states of the system and generates the estimates from a composite of contributions from the reference library snapshots according to their similarity to the received actual sensor values.

5. An apparatus according to claim 4 wherein the similarity operator employed by the empirical model compares corresponding elements of a snapshot of received actual sensor values and a reference library snapshot to generate elemental similarities, and statistically combines the elemental similarities to generate a snapshot similarity of the received actual snapshot and the reference library snapshot.

6. An apparatus according to claim 1 wherein the global similarity engine compares corresponding elements of a snapshot of received actual sensor values and a related snapshot of estimates to generate elemental similarities, and statistically combines the elemental similarities to generate a global similarity score.

7. An apparatus according to claim 6 wherein the related snapshot of estimates is generated by the model in response to the snapshot of actual sensor values.

8. An apparatus according to claim 6 wherein the related snapshot of estimates is a prediction related to the snapshot of actual sensor values in time, and generated by the model in response to a previous snapshot of actual sensor values.

9. An apparatus according to claim 1 wherein the selected range has a lower threshold equal to a multiple of a standard deviation for a sequence of global similarity scores, subtracted from the mean of the sequence of global similarity scores.

10. An apparatus according to claim 9 wherein the multiple is in the range of one to three.

11. An apparatus according to claim 1 wherein the selected range has a lower limit selected in the range of 0.85 to 0.95.

12. An apparatus according to claim 1 wherein the state transition detector indicates an operational state change when at least a selected minimum number of global similarity scores in a shifting window of successive global similarity scores fall outside of the selected range.

14. A method according to claim 13 wherein the determining step comprises comparing corresponding elements of a snapshot of received actual parameter values and a related snapshot of estimates to generate elemental similarities, and statistically combining the elemental similarities to generate the global similarity score.

15. A method according to claim 14 wherein the related snapshot of estimates is generated in the generating step in response to the snapshot of actual parameter values.

16. A method according to claim 14 wherein the related snapshot of estimates is a prediction related to the snapshot of actual parameter values in time, and is generated in the generating step in response to a previous snapshot of actual sensor values.

17. A method according to claim 13 wherein the selected range has a lower threshold equal to a multiple of a standard deviation for a sequence of global similarity scores, subtracted from the mean of the sequence of global similarity scores.

18. A method according to claim 17 wherein the multiple is in the range of one to three.

19. A method according to claim 13 wherein the selected range has a lower limit selected in the range of 0.85 to 0.95.

20. A method according to claim 13 wherein the indication step comprises indicating an operational state change when at least a selected minimum number of global similarity scores in a shifting window of successive global similarity scores fall outside of the selected range.

13. A method for detecting an operational state change in a system having monitored parameters, comprising the steps of:

receiving actual values of the monitored parameters characterizing the operational state of the system;

generating estimates of the parameter values in response to receiving actual values;

determining a global similarity score for a comparison of a set of estimated parameter values and a related set of actual parameter values; and

indicating an operational state change when a global similarity score falls outside of a selected range.

21. A method for determining when a system monitored using a model stabilizes into an operational state, comprising the steps of:

receiving successive snapshots of actual values of monitored parameters characterizing operation of the system;

generating snapshots of estimates of the parameter values in response to receiving snapshots of actual values;

comparing each snapshot of actual values to a related snapshot of estimates to generate a global similarity score for each comparison; and

indicating stabilization of the operational state of the system based on the global similarity scores.

22. A method according to claim 21 wherein the indicating step comprises examining a moving window of successive global similarity scores and when each of at least a selected number of successive global similarity scores in the window lies within a selected range around the previous global similarity score, identifying that operational state of the system has stabilized.

23. A method according to claim 21, wherein the indicating step comprises examining a moving window of successive global similarity scores and when each of at least a selected number of successive global similarity scores in the window lies within a selected range around the mean of the previous global similarity scores in the window, identifying that operational state of the system has stabilized.

25. A program product according to claim 24 wherein said means for generating compares corresponding elements of the snapshot of received actual values and the snapshot of estimated values to generate elemental similarities, and statistically combines the elemental similarities to generate the global similarity score.

26. A program product according to claim 25 wherein the elemental similarities are generated using a BART similarity operator.

27. A program product according to claim 25 wherein the elemental similarities are generated according to the absolute value of the difference of the corresponding elements, divided by an expected range for the corresponding elements.

24. A global similarity engine program product for use in modeling a system having monitored parameters comprising:

computer readable program code means for receiving a snapshot of actual values of the monitored parameters;

computer readable program code means for receiving a snapshot of estimated values of the monitored parameters; and

computer readable program code means for generating a global similarity score for a comparison of the snapshot of actual values and the snapshot of estimated values.