Industrial process surveillance system

• US 6,181,975 B1
• Filed: 02/24/1998
• Issued: 01/30/2001
• Est. Priority Date: 06/19/1996
• Status: Expired due to Term

• Current Assignee: ARCH Development Corporation (University of Chicago)
• Original Assignee: ARCH Development Corporation (University of Chicago)
• Inventors: Mott, Jack E., Wegerich, Stephan W, Singer, Ralph M., Gross, Kenneth C.
• Primary Examiner(s): Gordon, Paul P.
• Assistant Examiner(s): PATEL, RAMESH B
• Application Number:
  US09/028,443
• Time in Patent Office: 1,071 Days
• Field of Search: 340/506, 340/651, 700/28, 700/47, 700/48, 700/49, 700/50, 700/29, 700/80, 700/21, 700/79, 714/724, 714/725, 714/728, 714/2, 714/4, 714/15, 714/25, 714/39
• US Class Current: 700/29
• 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 23/0262 Confirmation of fault detection, e.g. extra checks to confirm that a failure has indeed occurred

Abstract

A system and method for monitoring an industrial process and/or industrial data source. The system includes generating time varying data from industrial data sources, processing the data to obtain time correlation of the data, determining the range of data, determining learned states of normal operation and using these states to generate expected values, comparing the expected values to current actual values to identify a current state of the process closest to a learned, normal state; generating a set of modeled data, and processing the modeled data to identify a data pattern and generating an alarm upon detecting a deviation from normalcy.

Claims

1. A method for monitoring a plurality of data sources in order to determine a pattern characteristic of a system, comprising the steps of:

accumulating data over time from the data sources;

processing the data to obtain optimum time correlation of the data accumulated from the plurality of data sources;

determining learned states of at least one desired pattern of the system;

using the learned states to generate expected data values of the data accumulated over time from the data sources of the system;

comparing the expected data values to current actual data values of the data from the data sources to identify a current state of the system closest to one of the learned states and generating a set of modeled data; and

determining from the modeled data a pattern for the current actual values and if the pattern deviates from a pattern characteristic of the desired pattern, an alarm notice is provided.

2. The method as defined in claim 1 further including the step of searching the time correlated data to identify maximum and minimum values for the data to determine a full range of values for the data from the data sources.

3. The method as defined in claim 1 wherein the data sources comprise at least one of industrial sensors and a financial information source.

4. The method as defined in claim 1 further including the step of modifying the system responsive to determining a deviation from a pattern characteristic of the desired pattern.

5. The method as defined in claim 1 wherein the step of determining optimum time correlation comprises comparing pairs of data from different data source, each set of data being characteristic of a separate data source and calculating a cross-correlation vector over time, applying a low pass filter to remove noise from the cross-correlation vector and determining a phase shift between the data.

6. The method as defined in claim 5 wherein the step of determining the phase shift comprises differentiating the cross-correlation vector with respect to lag time between each pair of data and performing an interpolation to compute the root of the differential of the cross-correlation vector.

7. The method as defined in claim 1 wherein the step of identifying a current state of the data closest to the learned states includes forming a combination of the learned states to identify a true state of the system.

8. The method as defined in claim 7 further including the step of substituting an expected value for faulty observations from a system data source.

9. The method as defined in claim 1 wherein the step of detecting a deviation from the desired pattern comprises applying a sequential probability ratio test to the modeled data.

10. The method as defined in claim 1 wherein the step of detecting a deviation from a desired pattern comprises performing a pattern recognition analysis using computer means.

11. The method as defined in claim 1 wherein the system is selected from the group consisting of an industrial manufacturing system, a utility operation, a business operation, an investment operator, a weather forecasting system, a database of medical information and a transportation system.

12. The method as defined in claim 1 wherein the plurality of data sources includes a plurality of data source pairs.

13. A computerized system for monitoring at least one of an industrial process and industrial sensors, comprising:

means for generating time varying data from a plurality of industrial sensors;

computer means for executing a software module for processing the time varying data to determine optimum time correlation of the data accumulated from the plurality of industrial sensors;

computer means for executing a software module for searching the time correlated data to identify maximum and minimum values for the data, thereby determining a full range of values for the data characteristic of the at least one of the industrial process and the industrial sensors;

computer means for executing a software module for determining learned states of a normal operational condition of the at least one of the industrial process and the industrial sensors and using the learned states to generate expected values characteristic of at least one of the industrial process and the industrial sensors;

computer means for executing a software module for comparing the expected values to current actual values characteristic of the at least one of the industrial process and the industrial sensors to identify a current state of the at least one of the industrial process and the industrial sensors closest to one of the learned states and generating a set of modeled data; and

computer means for executing a software module for processing the modeled data to identify a pattern for the data and upon detecting a deviation from a pattern characteristic of normal operation of the at least one of the industrial process and the industrial sensors, an alarm is generated.

14. The computerized system as defined in claim 13 wherein the industrial sensors include data sources which are selected from the group consisting of physical sensors, chemical sensors, biological sensors and financial data over time.

15. The computerized system as defined in claim 13 wherein said computer means for determining optimum time correlation comprises means for comparing pairs of data from different data sources, each characteristic of a separate data source and calculating a cross-correlation vector over time, applying a low pass filter to remove noise from the cross-correlation vector and determining a phase shift between the data.

16. The computerized system as defined in claim 15 wherein said computer means for determining the phase shift comprises means for differentiating the cross-correlation vector with respect to lag time between each pair of data and performing an interpolation to compute the root of the differential of the cross-correlation vector.

17. The computerized system as defined in claim 13 wherein said computer means for identifying a current state of the system closest to the learned state includes means for forming a combination of the learned states to identify a true state of the system.

18. The computerized system as defined in claim 17 further including computer means for executing a software module for substituting an expected value for faulty observations from at least one of the industrial sensors.

19. The computerized system as defined in claim 13 further including means for modifying the at least one of an industrial process and industrial sensors being monitored responsive to detecting a deviation from normal operation.

20. The computerized system as defined in claim 13 wherein said computer means for detecting a deviation pattern comprises means for performing a pattern recognition analysis.

21. A system for monitoring a data source characteristic of a process, comprising:

means for generating time varying data from a data source;

computer means for executing a software module for processing the time varying data to determine optimum time correlation of the data accumulated from the data source;

computer means for executing a software module for searching the time correlated data to identify maximum and minimum values for the data, thereby determining a full range of values for the data from the data source;

computer means for executing a software module for determining learned states of a desired operational condition of the data source and using the learned states to generate expected values of the data source;

computer means for executing a software module for comparing the expected values to current actual values of the data source to identify a current state of the data source closest to one of the learned states and generating a set of modeled data; and

computer means for executing a software module for processing the modeled data to identify a pattern for the data and upon detecting a deviation from a pattern characteristic of normal operation of the process, an alarm is generated.

22. The system as defined in claim 21 further including means for modifying the process being monitored responsive to the alarm.

23. The system as defined in claim 21 wherein the data source is selected form the group consisting of an industrial data source, a utility operational data source, a business data source, a financial data source, a weather data source, a medical data source, a transportation data source and a chemical data source.

24. The system as defined in claim 21 wherein said computer means for determining optimum time correlation comprises means for comparing pairs of data from different data sources, each characteristic of a separate data source and calculating a cross-correlation vector over time, applying a low pass filter to remove noise from the cross-correlation vector and determining a phase shift between the data.

25. The system as defined in claim 21 wherein said computer means for identifying a current state of the process closest to the learned state includes means for forming a combination of the learned states to identify a true state of the process.

26. The system as defined in claim 21 further including means for identifying a current state of the process closed to the learned state comprises means for forming a combination of the learned states to identify a true state of the process.

27. The system as defined in claim 21 further including computer means for executing a software module for substituting an expected value for faulty observations from the data source.

28. The system as defined in claim 21 wherein said means for processing the modeled data to identify a pattern for the data comprises means for applying a sequential probability ratio test.

29. The system as defined in claim 21 wherein said means for processing the modeled data to identify a pattern for the data comprises means for performing a pattern recognition analysis.

30. A system for monitoring a plurality of data sources in order to determine a pattern characteristic of a process, comprising:

means for accumulating data over time from the data sources;

means for processing the data to obtain optimum time correlation of the data accumulated from the plurality of the data sources;

means for determining learned states of at least one desired pattern of the process;

means for using the learned states to generate expected data values of the data accumulated over time from the data sources of the process;

means for comparing the expected data values to current actual data values of the data from the data sources to identify a current state of the process closest to one of the learned states and generating a set of modeled data; and

means for determining from the modeled data pattern for the current actual values and if the pattern deviates from a pattern characteristic of the desired pattern, an alarm notice is provided.

31. The system as defined in claim 30 further including means for searching the time correlated data to identify maximum and minimum values for the data to determine a full range of values for the data from the data sources.

32. The system as defined in claim 30 wherein said data sources are selected from the group consisting of an industrial data source, a utility operational data source, a business data source, a financial data source, a weather data source, a medical data source, a transportation data source and a chemical data source.

33. The system as defined in claim 30 further including means for substituting an expected value for faulty observation from one of said data sources.

34. The system as defined in claim 30 wherein said means for determining from the modeled data a pattern for the current actual values comprises a sequential probability ratio test.

35. The system as defined in claim 30 further including means for modifying the process being monitored responsive to the alarm notice.

36. A system for monitoring a process, comprising:

input means for acquiring sensor data descriptive of an operational state of said process;

memory means for storing a plurality of data values characteristic of at least one normal state of operation of said process;

means for computing a measure of similarity of the operational state of the process with each of the plurality of data values characteristic of the at least one normal state of operation of the process; and

alarm means for generating a signal indicative of a difference between the operational state and the at least one normal state of operation of the process, based on a sequence of such measures of similarity over successively acquired ones of said sensor data.

37. A system according to claim 36, wherein said means for computing further generates a set of estimated sensor data computed from the measure of similarity.

38. A system according to claim 37, wherein said alarm means is responsive to a difference between the sensor data and the estimated sensor data.

39. A system according to claim 38, wherein said alarm means applies a sequential probability ratio test to the difference between the sensor data and the estimated sensor data.

40. A system according to claim 39, wherein said means for computing determines the measure of similarity at least by comparing the sensor data to at least one value in the set of reference sensor data to compute a measure of association for each such comparison.

41. A system according to claim 40, wherein said means for computing determines the measure of similarity as a combination of a plurality of the association measures.

42. A system according to claim 41, wherein said means for computing determines the set of virtual sensor data as a linear combination of contributions from the set of reference sensor data wherein the contributions are determined based on the measure of similarity.

43. A system according to claim 37, further comprising presentation means disposed to receive the estimated sensor data and the sensor data, and responsive to the signal from said alarm means, presents a set of composite sensor data comprising at least a portion of the sensor data and at least a portion of the estimated sensor data.

44. A system according to claim 36, further comprising correlation means for receiving successive sensor data from said input means and presenting the sensor data to said processing means in a modified succession reflective of time delays of correlated effects between sensors in the sensor data.

45. A system according to claim 44, wherein said correlation means compares pairs of successive sensor data from different ones of the input means to compute a cross-correlation vector over time, removes noise from the cross-correlation vector by applying a low pass filter and determines a phase shift to apply to the sensor data for presentation to the processing means.

46. A system according to claim 36, further comprising memory update means for adding further reference sensor data to said memory means descriptive of additional normal states of operation of the process.

47. A method of determining an operational state of a process, comprising the steps of:

collecting reference sensor data descriptive of normal states of operation of the process;

acquiring recent sensor data from at least one sensor descriptive of the operational state of the process;

computing a measure of association of the recent sensor data to reference sensor data of each of the normal states of operation of the process; and

providing a composite of association measures as a determination of the operational state of the process.

48. A method according to claim 47, wherein said computing step further comprises for each of the at least one normal state, the steps of:

comparing each element of the recent sensor data corresponding to a particular sensor to each element of reference sensor data corresponding to the particular sensor for one of the data values of the at least one normal state to provide a similarity value for each such step of comparing; and

statistically combining all the similarity values for the one of the data values of the at least one normal state to compute a measure of association of the recent sensor data to the reference sensor data for the at least one normal state.

49. A method according to claim 48, wherein said step of statistically combining comprises averaging all such similarity values to provide the measure of association.

50. A method according to claim 48, comprising the further steps of:

creating estimated sensor data from combinations of the reference sensor data based on the measure of association of the recent sensor data with the reference sensor data for each of the data values of the at least one normal state; and

evaluating the estimated sensor data as a further determination of the operational state of the process.

51. A method according to claim 50, comprising the further steps of:

obtaining a residual measure from the difference between the recent sensor data and the estimated sensor data;

applying a statistical significance test to the residual measure; and

generating a signal representative of a statistically significant difference between the operational state of the process and the at least one normal state of operation of the process.

52. A method according to claim 51 wherein said step of applying a statistical significance test comprises applying a sequential probability ratio test.

53. A method according to claim 47, wherein said step of collecting reference sensor data comprises the steps of:

gathering sensor data from the process when the process is in one of the normal states;

identifying in the gathered sensor data at least one set of data from the at least one sensor at moments when the sensor attains a highest value and a lowest value;

combining in a collection each set of data identified in the previous step; and

removing from the collection a redundant set of data.

54. A method according to claim 47, comprising the further steps of:

determining if the operational state of the process is an additional normal state of operation, and if it is, then adding the recent sensor data to the collected reference sensor data.

55. A system for monitoring a process, comprising:

input means for acquiring sensor data descriptive of an operational state of said process;

memory means for storing a plurality of reference sensor data sets descriptive of normal states of operation of said process;

means for computing a measure of similarity of the sensor data descriptive of the operational state of the process with each of the reference sensor data sets descriptive of the normal states of operation of the process; and

alarm means for generating a signal indicative of a difference between the operational state and the normal states of operation of the process, based on the measure of similarity over successively acquired ones of said sensor data.

56. A system according to claim 55, wherein said means for computing further generates a set of estimated sensor data computed from the measure of similarity.

57. A system according to claim 55, wherein said alarm means is responsive to a difference between the sensor data and the estimated sensor data.

58. A system according to claim 55, wherein said alarm means applies a sequential probability ratio test to the difference between the sensor data and the estimated sensor data.

59. A system according to claim 55, further comprising presentation means disposed to receive the estimated sensor data and the sensor data, and responsive to the signal from said alarm means, presents a set of composite sensor data comprising at least a portion of the sensor data and at least a portion of the estimated sensor data.