Statistical signatures used with multivariate statistical analysis for fault detection and isolation and abnormal condition prevention in a process

• US 7,526,405 B2
• Filed: 10/13/2006
• Issued: 04/28/2009
• Est. Priority Date: 10/14/2005
• Status: Expired due to Fees

• Current Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Original Assignee: Fisher-Rosemount Systems Incorporated (Emerson Electric Company)
• Inventors: Miller, John Philip
• Primary Examiner(s): TSAI, CAROL S W
• Application Number:
  US11/549,447
• Publication Number:
  US 20070088528A1
• Time in Patent Office: 928 Days
• Field of Search: 702/108, 702/109, 702/110, 702/121, 702/179, 702/183, 702/184, 702/185, 702/22, 702/23, 702/188, 700 18- 23, 700/17, 700 26- 28, 700 48- 53, 700 86- 88, 700/83, 700/108, 700/121, 700/110, 714/37, 714/38, 714/46, 714/47, 714/25, 438/14, 709/220, 709/221, 709/222, 709/228, 709/223, 382/144, 382/145, 382/147
• US Class Current: 702/179
• 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

Abstract

A system and method for monitoring a process in a process plant and detecting an abnormal condition includes collecting data representative of the operation of the process, performing a multivariate statistical analysis to represent the operation of the process in a known state based on a set of collected reference data, where the reference data includes a statistical measure of the operation of the process in the known state. The system and method may further include representing the operation of the process in an unknown state based on a set of monitored data, where the monitored data includes a statistical measure of the operation of the process in an unknown state, and using the output of the multivariate statistical analysis as an input, and comparing the process in the unknown state to the multivariate statistical representation of the operation of the process in the known state to determine the operational state of the process.

Claims

1. A multivariate statistical analysis system comprising:

a processor;

a memory coupled to the processor; and

a first set of machine-readable instructions for execution by the processor, stored in the memory, and comprising a first analysis tool that, when executed;

receives a first set of statistical measures indicative of the operation of a process in a process plant, each statistical measure comprising the result of a statistical calculation performed on two or more measured values of a particular parameter of the process;

performs a multivariate statistical analysis on the first set of statistical measures; and

outputs a representation of the operation of the process based on the first set of statistical measures.

2. The system of claim 1, wherein the received first set of statistical measures is representative of the operation of the process in a known state.

3. The system of claim 1, wherein the first statistical measures are calculated within a device being monitored within the plant.

4. The system of claim 1, wherein the first analysis tool is adapted to generate the first statistical measures indicative of the operation of the process from first data collected by a data collection module, wherein the data are related to a plurality of devices within the process plant.

5. The system of claim 1, wherein the multivariate statistical analysis comprises a principal component analysis.

6. The system of claim 1, further comprising a second analysis tool that, when executed:

receives a second set of statistical measures indicative of the operation of a process;

receives the representation of the operation of the process based on the first set of statistical measures; and

represents the operation of the process in an unknown state based on the second set of statistical measures,

wherein the second set of statistical measures is representative of the operation of the process in an unknown state.

7. The system of claim 6, further comprising an execution engine adapted to implement the first and second analysis tools resulting in a multivariate statistical representation of the operation of the process in a known state and from which the operation of the process in an unknown state may be determined.

8. The system of claim 6, wherein the second analysis tool is adapted to determine the presence of an abnormal process condition.

9. The system of claim 8, further wherein the second analysis tool generates an alarm based on the presence of the abnormal process condition.

10. The system of claim 8, wherein the second analysis tool is adapted to identify the abnormal process condition.

11. The system of claim 10, further wherein the second analysis tool is adapted to generate an alarm corresponding to the identified abnormal process condition.

12. The system of claim 6, wherein the second analysis tool is adapted to predict an occurrence of an abnormal process condition.

13. The system of claim 12, further wherein the second analysis tool is adapted to generate an alarm based on the predicted occurrence of the abnormal process condition.

14. The system of claim 12, wherein the second analysis tool is adapted to identify the abnormal process condition of the predicted occurrence of the abnormal process condition.

15. The system of claim 14, wherein the second analysis tool adapted to generate an alarm corresponding to the identified abnormal process condition.

16. The system of claim 1, wherein each statistical measure of the first set of statistical measures comprises one of the group consisting of:

a mean, a mean change, a median, a median change, a standard deviation, a standard deviation change, a variance, a skewness, a kurtosis, a root-mean-square (RMS), a rate of change, a range, a minimum, and a maximum.

17. A method for monitoring a process in a process plant, the method comprising:

collecting data representative of an operation of the process;

calculating a statistical measure on the collected data representative of the operation of the process in a known state, wherein the statistical measure is the result of a statistical calculation on two or more measured values of a particular parameter of the process; and

performing a multivariate statistical analysis to represent the operation of the process in a known state based on a set of collected reference data comprising the statistical measure of the operation of the process in the known state.

18. The method of claim 17, wherein calculating a statistical measure on the collected data representative of the operation of the process comprises performing a statistical calculation on the collected data representative of the operation of the process to generate the statistical measure of the operation of the process in the known state.

19. The method of claim 17, further comprising calculating a statistical measure on the collected data representative of the operation of the process in an unknown state.

20. The method of claim 17, wherein the multivariate statistical analysis comprises a principal component analysis.

21. The method of claim 17, wherein the collected data comprises raw data associated with a plurality of devices within the process plant, and wherein calculating a statistical measure on the collected data representative of the operation of the process comprises calculating a statistical measure on the raw data.

22. The method of claim 17, wherein the collected data is representative of an operation of a device within the process, and wherein calculating a statistical measure on the collected data representative of the operation of the process comprises calculating a statistical measure within the device.

23. The method of claim 17, wherein the representation of the operation of the process in a known state comprises a representation of the operation of the process in one or more known abnormal conditions.

24. The method of claim 17 further comprising representing the operation of the process in an unknown state based a set of monitored data comprising a statistical measure of the operation of the process in the unknown state and using the output of the multivariate statistical analysis as an input.

25. The method of claim 24 further comprising comparing the process in the unknown state to the multivariate statistical representation of the operation of the process in the known state to determine the operational state of the process.

26. The method of claim 24, further comprising:

comparing the representation of the process in an unknown state to the representation of the process in a known state; and

detecting the presence of an abnormal process condition from the comparison.

27. The method of claim 26, further comprising generating an alarm based on the detection of an abnormal process condition.

28. The method of claim 26, further comprising identifying the abnormal process condition.

29. The method of claim 28, further comprising generating an alarm corresponding to an identified abnormal process condition.

30. The method of claim 26, further comprising predicting the occurrence of an abnormal process condition.

31. The method of claim 30, further comprising generating an alarm based on the predicted occurrence of an abnormal process condition.

32. The method of claim 30, further comprising identifying the abnormal process condition of the predicted occurrence of an abnormal process condition.

33. The method of claim 32, further comprising generating an alarm corresponding to an identified abnormal process condition.

34. The method of claim 17, wherein the statistical measure comprises one of the group consisting of:

a mean, a mean change, a median, a median change, a standard deviation, a standard deviation change, a variance, a skewness, a kurtosis, a root-mean-square (RMS), a rate of change, a range, a minimum, and a maximum.

35. A system for monitoring a process in a process plant, the system comprising:

a first analysis tool for;

receiving a first set of statistical measures, each of the first set of statistical measures being the result of a statistical calculation on two or more measured values of a particular parameter of the process, each of the first set of the statistical measures indicative of the operation of the process while the process is in one or more known conditions; and

performing a multivariate statistical analysis on the first set of statistical measures to generate a multivariate statistical representation of the operation of the process in the one or more known conditions; and

a second analysis tool for;

receiving a second set of statistical measures, each of the second set of statistical measures being the result of a statistical calculation on two or more measured values of the particular parameter of the process, each of the second set of statistical measures indicative of the operation of the process while the process is in an unknown condition;

receiving the multivariate statistical representation of the operation of the process;

projecting the received second set of statistical measures onto the multivariate statistical representation of the operation of the process to represent the operation of the process in the unknown condition; and

determining the unknown condition based on the observation of the second set of statistical measures projected onto the multivariate statistical representation as compared with the observation of the first set of statistical measures projected onto the multivariate statistical representation.

36. The system of claim 35, wherein the first analysis tool is adapted to execute a principal component analysis to generate the multivariate statistical representation.

37. The system of claim 35, wherein the first analysis tool is adapted to assign an observation of the first set of statistical measures projected onto the multivariate statistical representation to an abnormal condition.

38. The system of claim 37, wherein the first analysis tool is adapted to assign an observation of the first set of statistical measures projected onto the multivariate statistical representation to an abnormal condition according to a score discriminant.

43. The system of claim 35, wherein the second analysis tool is adapted to detect an abnormal condition based on the determination of the unknown condition.

44. The system of claim 43, wherein the second analysis tool is adapted to identify the abnormal condition based on the observation of the second set of statistical measures projected onto the multivariate statistical representation as compared with the observation of the first set of statistical measures projected onto the multivariate statistical representation, wherein the known condition comprises a known abnormal condition.

45. The system of claim 43, wherein the second analysis tool is adapted to generate an alarm based on the detected abnormal condition.

46. The system of claim 43, wherein the detected abnormal condition comprises a predicted occurrence of the abnormal condition.

47. The system of claim 43, wherein the detected abnormal condition comprises an actual occurrence of the abnormal condition.

48. The system of claim 35, wherein the second set of statistical measures comprises one of the group consisting of:

a mean, a mean change, a median, a median change, a standard deviation, a standard deviation change, a variance, a skewness, a kurtosis, a root-mean-square (RMS), a rate of change, a range, a minimum, and a maximum.

49. The system of claim 35, wherein the first set of statistical measures comprises one of the group consisting of:

a mean, a mean change, a median, a median change, a standard deviation, a standard deviation change, a variance, a skewness, a kurtosis, a root-mean-square (RMS), a rate of change, a range, a minimum, and a maximum.

39. A system for monitoring a process in a process plant, the system comprising:

a first analysis tool for;

receiving a first set of statistical measures, each of the first set of statistical measures being the result of a statistical calculation on two or more measured values of a particular parameter of the process, each of the first set of the statistical measures indicative of the operation of the process while the process is in one or more known conditions; and

performing a multivariate statistical analysis on the first set of statistical measures to generate a multivariate statistical representation of the operation of the process in the one or more known conditions; and

a second analysis tool for;

receiving a second set of statistical measures, each of the second set of statistical measures being the result of a statistical calculation on two or more measured values of the particular parameter of the process, each of the second set of statistical measures indicative of the operation of the process while the process is in an unknown condition;

receiving the multivariate statistical representation of the operation of the process;

projecting the received second set of statistical measures onto the multivariate statistical representation of the operation of the process to represent the operation of the process in the unknown condition; and

determining the unknown condition based on the observation of the second set of statistical measures projected onto the multivariate statistical representation as compared with the observation of the first set of statistical measures projected onto the multivariate statistical representation;

wherein the first analysis tool is adapted to assign an observation of the first set of statistical measures projected onto the multivariate statistical representation to an abnormal condition according to a score discriminant; and

further wherein the score discriminant comprises a maximum score discriminant calculated as;

$g_i<br><br>\left(x\right)=<br><br>-\frac{1}{2}<br><br>{\left(x-{\stackrel{\_}{x}}_i\right)}^T<br><br>{P<br><br>\left(P^T<br><br>S_i<br><br>P\right)}^{-1}<br><br>P^T<br><br>\left(x-{\stackrel{\_}{x}}_i\right)+<br><br>\ln <br><br>\left(p_i\right)-\frac{1}{2}<br><br>\ln <br><br>\left[\det <br><br>\left(P^T<br><br>S_i<br><br>P\right)\right]$

x=vector of original process variable measurements

g_i=likelihood that x belongs to abnormal condition class i

x_i=a mean vector of all observations belonging to class i

P =a principal component analysis loading matrix

S_i=a covariance matrix of all observations belonging to class i

p_i=a priori probability of an observation belonging to class i.

40. The system of claim 39, wherein the second analysis tool is adapted to associate the observation of the first set of statistical measures to an abnormal condition based upon a maximum value of g_i.

41. A system for monitoring a process in a process plant, the system comprising:

a first analysis tool for;

receiving a first set of statistical measures, each of the first set of statistical measures being the result of a statistical calculation on two or more measured values of a particular parameter of the process, each of the first set of the statistical measures indicative of the operation of the process while the process is in one or more known conditions; and

performing a multivariate statistical analysis on the first set of statistical measures to generate a multivariate statistical representation of the operation of the process in the one or more known conditions; and

a second analysis tool for;

receiving a second set of statistical measures, each of the second set of statistical measures being the result of a statistical calculation on two or more measured values of the particular parameter of the process, each of the second set of statistical measures indicative of the operation of the process while the process is in an unknown condition;

receiving the multivariate statistical representation of the operation of the process;

projecting the received second set of statistical measures onto the multivariate statistical representation of the operation of the process to represent the operation of the process in the unknown condition; and

determining the unknown condition based on the observation of the second set of statistical measures projected onto the multivariate statistical representation as compared with the observation of the first set of statistical measures projected onto the multivariate statistical representation;

wherein the first analysis tool is adapted to assign an observation of the first set of statistical measures projected onto the multivariate statistical representation to an abnormal condition according to a score discriminant; and

further wherein the score discriminant is calculated as;

g_i(x)=(x−

x_i)^TP(P^TS_iP)<sup>−

1</sup>P^T(x−

x_i)+ln[det(P^TS_iP)]

wherein;

x=vector of original process variable measurements likelihood that x belongs to abnormal condition class i

g_i=a mean vector of all observations belonging to class i

x_i=a mean vector of all observations belonging to class i

P=a principal component analysis loading matrix

S_i=a covariance matrix of all observations belonging to class i.

42. The system of claim 41, wherein the second analysis tool is adapted to associate the observation of the first set of statistical measures to an abnormal condition based upon a minimum value of g_i.