Intelligent Monitoring System and Method for Building Predictive Models and Detecting Anomalies

• US 20070265713A1
• Filed: 02/05/2007
• Published: 11/15/2007
• Est. Priority Date: 02/03/2006
• Status: Active Grant

• Current Assignee: Recherche 2000 Inc.
• Original Assignee: Recherche 2000 Inc.
• Inventors: Veillette, Michel, Berriah, Said, Tremblay, Gilles
• Granted Patent: US 7,818,276 B2
• Time in Patent Office: Days
• US Class Current: 700/30
• CPC Class Codes: G05B 17/02 electric
• CPC Class Codes: G05B 23/0245 based on a qualitative model, e.g. rule based; if-then decisions

Abstract

There is described a method and system for detecting an anomaly in data originating from a component used in an industrial process. The anomaly detection steps comprise: providing predictive models in which at least one of the predictive models corresponds to at least one operating mode from a plurality of operating modes of the component; obtaining operating data from the component, the operating data being indicative of a given operating mode from the operating modes; selecting at least one of the predictive models based on the given operating mode; generating estimated data using the selected at least one predictive model; comparing the operating data with the estimated data; and generating an alarm trigger when the comparison meets a given anomaly criteria thereby constituting the detection of the anomaly.

Claims

1. A method for detecting an anomaly in data originating from a component used in an industrial process, the method comprising:

providing predictive models in which at least one of the predictive models corresponds to at least one operating mode from a plurality of operating modes of the component;

obtaining operating data from the component, the operating data being indicative of a given operating mode from the operating modes;

selecting at least one of the predictive models based on the given operating mode;

generating estimated data using the selected at least one predictive model;

comparing the operating data with the estimated data; and

generating an alarm trigger when the comparison meets a given anomaly criteria thereby constituting the detection of the anomaly.

2. The method of claim 1, further comprising determining anomaly signatures representative of a set of deviations of operating data from reference data;

and wherein the meeting of the given anomaly criteria comprises identifying at least one of the anomaly signatures.

3. The method of claim 2, further comprising associating the identified anomaly signature to a specific problem associated to the component.

4. The method of claim 1, further comprising building the predictive models.

5. The method of claim 4, wherein the building further comprises generating estimated data from reference operating data;

and producing residuals by calculating differences between the reference operating data and the generated estimated data.

6. The method of claim 5, wherein the building further comprises determining, from the reference data, the plurality of operating modes for the component.

7. The method of claim 6, wherein the building further comprises identifying the training operating mode from the plurality of operating modes and using the predictive model which corresponds to the identified training operating mode in the generating of estimated data.

8. The method of claim 7, wherein the building further comprises using the reference data to specify the range of validity of the plurality of predictive models.

9. The method of claim 8, wherein the range specification further comprises providing at least one of a data tag with high and low limits and a time interval for the limits;

a data mean;

a data distribution; and

a data aging.

10. The method of claim 9, wherein the building further comprises preprocessing the reference data by performing at least one of filtering irrelevant readings, generating numerical variables and extracting specific features.

11. The method of claim 7, wherein the building further comprises using the residuals and the calculated data to assess the precision limits of the model.

12. The method of claim 1, further comprising identifying, from the operating data, the given operating mode from the plurality of operating modes.

13. The method of claim 1, further comprising using the alarm trigger to perform at least one of shutting down an industrial process, and switching a value in a status tag that triggers an event on a User Interface, an event database, or a relay trigger.

14. The method of claim 1, wherein the component comprises plurality of components.

26. The system of claim 14, wherein the calculation engine is further for identifying, from the operating data, the given operating mode from the plurality of operating modes.

27. The system of claim 14, wherein the anomaly recognition engine is further for using the alarm trigger to perform at least one of shutting down an industrial process, and switching a value in a status tag that triggers an event on a User Interface, an event database, or a relay trigger.

15. A system for detecting an anomaly in data originating from a component used in an industrial process, the system comprising:

a predictive model builder for providing predictive models in which at least one of the predictive models corresponds to at least one operating mode from a plurality of operating modes of the component;

an input for receiving operating data from the component, the operating data being indicative of a given operating mode from the operating modes;

a calculation engine for selecting at least one of the predictive models based on the given operating mode;

the calculation engine further for generating estimated data using the selected at least one predictive model; and

an anomaly recognition engine for comparing the operating data with the estimated data and for generating an alarm trigger when the comparison meets a given anomaly criteria thereby constituting the detection of the anomaly.

16. The system of claim 15, wherein the calculation engine is further for determining anomaly signatures representative of a set of deviations of operating data from reference data;

and wherein the anomaly recognition engine is further for identifying at least one of the anomaly signatures in determining the meeting of the given anomaly criteria.

17. The system of claim 16, further comprising a diagnosis generator for associating the identified anomaly signature to a specific problem associated to the component.

18. The system of claim 15, further comprising a predictive model builder for building the predictive models.

19. The system of claim 18, wherein the predictive model builder comprises a prediction builder for generating estimated data from reference operating data;

and producing residuals by calculating differences between the reference operating data and the generated estimated data.

20. The system of claim 19, wherein the predictive model builder further comprises an activation condition builder for determining, from the reference data, the plurality of operating modes for the component.

21. The system of claim 20, wherein the activation condition builder is further for identifying the training operating mode from the plurality of operating modes and using the predictive model which corresponds to the identified training operating mode in the generating of estimated data.

22. The system of claim 21, wherein the predictive model builder further comprises a characterisation builder for using the reference data to specify the range of validity of the plurality of predictive models.

23. The system of claim 22, wherein the characterisation builder is further for providing at least one of a data tag with high and low limits and a time interval for the limits;

a data mean;

a data distribution; and

a data aging.

24. The method of claim 22, wherein the characterization builder further uses the residuals and the calculated data to assess the precision limits of the model.

25. The system of claim 22, wherein predictive model builder further comprises a preprocessing builder for preprocessing the reference data by performing at least one of filtering irrelevant readings, generating numerical variables and extracting specific features.