Method and apparatus for detecting out-of-range conditions in power generation equipment operations

• US 7,188,050 B2
• Filed: 08/04/2005
• Issued: 03/06/2007
• Est. Priority Date: 08/25/2004
• Status: Expired due to Fees

• Current Assignee: Siemens Corp. (Siemens AG)
• Original Assignee: Siemens Corporate Research Incorporated (Siemens AG)
• Inventors: Neubauer, Claus, Yuan, Chao, Cataltepe, Zehra
• Primary Examiner(s): Bui; Bryan
• Application Number:
  US11/197,270
• Publication Number:
  US 20060048007A1
• Time in Patent Office: 579 Days
• Field of Search: 702182-183, 702/185, 702/188, 702 35- 36, 702/32, 714/26, 706/2, 706/12, 706/14, 706/15, 700 30- 32
• US Class Current: 702/183
• 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 method and apparatus for detecting out-of-range conditions representing normal operations is disclosed. A support vector machine is used to generate an improved representation of historical training data from power generation equipment that facilitates a more accurate determination of the boundary between measurements that should be considered faults and those that represent normal operating conditions. The SVM receives data collected from a plurality of independent sensors associated with the power generating equipment in order to generate a boundary substantially separating a first class of data (e.g., a fault) from a second class of data (e.g., a normal operating condition) in a support vector machine feature space. Elements of operational data are collected and compared to the boundary generated from historical training data. A determination is then made whether the element of operational data is in a particular class, such as a class associated with out-of-range conditions.

Claims

1. A method for detecting out-of-range conditions in an apparatus having a monitoring system, said monitoring system storing a support vector machine feature space having a boundary substantially separating a first class of data from a second class of data, said method comprising:

receiving an element of operational data from an independent sensor associated with said apparatus;

comparing said element of operational data to said boundary;

if said element of operational data is within said second class of data, determining that said element of operational data represents an out-of-range condition for said apparatus; and

issuing an alarm to indicate a fault occurs.

2. The method of claim 1 wherein said apparatus comprises power generation equipment.

3. The method of claim 1 wherein said support vector machine generates said boundary from a plurality of elements of training data.

4. The method of claim 1 wherein said element of operational data comprises data collected by one sensor in a plurality of sensors, said one sensor positioned at one location in a plurality of locations on said apparatus.

5. The method of claim 4 wherein said one location in said plurality of locations on said apparatus comprises a location of an independent sensor.

6. The method of claim 1 wherein said element of operational data comprises a vector.

7. The method of claim 1 wherein said second class is a class assigned to training data that is out of a normal operating range of said apparatus.

8. The method of claim 1 wherein said first class is a class assigned to data that is within a normal operating range of said apparatus.

9. A method for detecting out-of-range conditions in an apparatus comprising:

receiving a set of training data collected by one or more independent sensors, said training data comprising measurements of operational characteristics of said apparatus;

assigning a class to each element of data in said training data;

generating a boundary in a feature space, said boundary substantially separating training data assigned to one class from training data assigned to other classes, wherein said other classes are classes associated with training data that is out of a normal operating range of said apparatus; and

storing said boundary in memory.

10. The method of claim 9 wherein said apparatus comprises power generation equipment.

11. The method of claim 9 wherein said measurements comprise data elements collected by a plurality of sensors positioned at a plurality of locations on said apparatus.

12. The method of claim 9 further comprising:

collecting an element of operational data;

comparing said element of operational data to said feature space; and

determining whether said element of operational data is in said one class.

13. The method of claim 12 wherein said step of determining whether said element of operational data is in said one class comprises:

determining whether said element of operational data represents an out-of-range condition for said apparatus.

14. The method of claim 9 wherein each sensor in said plurality of sensors comprises an independent sensor.

15. The method of claim 9 wherein said training data comprises a plurality of vectors.

16. The method of claim 9 wherein said step of generating comprises estimating said boundary using a Gaussian kernel.

17. The method of claim 16 wherein said Gaussian kernel is defined as:

$\exp \left(\frac{-{<br><br>x_i-x_j<br><br>}^2}{2<br><br><br><br>{\mathrm{\sigma <br><br>}}^2}\right)$

where x_i and x_j are two vectors in the input space and σ

is the width of the Gaussian kernel.

18. The method of claim 9 wherein said one class is a class assigned to training data that is within a normal operating range of said apparatus.

19. An apparatus for detecting out-of-range conditions in equipment having a monitoring system, said monitoring system adapted to store a support vector machine feature space having a boundary substantially separating a first class of data from a second class of data, said apparatus comprising:

means for receiving an element of operational data from an independent sensor associated with said equipment;

means for comparing said element of operational data to said boundary; and

means for determining that said element represents an out-of-range condition for said equipment if said element of operational data is in said second class.

20. The apparatus of claim 19 wherein said equipment comprises power generation equipment.

21. The apparatus of claim 19 wherein said support vector machine is adapted to generate said boundary from a plurality of elements of training data.

22. The apparatus of claim 19 wherein said element of operational data comprises data collected by one sensor in a plurality of sensors, said one sensor positioned at one location in a plurality of locations on said equipment.

23. The apparatus of claim 22 wherein said one location in said plurality of locations on said equipment comprises a location of an independent sensor.

24. The apparatus of claim 19 wherein said element of operational data comprises a vector.

25. The apparatus of claim 19 wherein said second class is a class assigned to training data that is out of a normal operating range of said equipment.

26. The apparatus of claim 19 wherein said first class is a class assigned to data that is within a normal operating range of said equipment.

27. An apparatus for detecting out-of-range conditions in equipment comprising:

means for receiving a set of training data collected by one or more independent sensors, said training data comprising measurements of operational characteristics of said equipment;

means for assigning a class to each element of data in said training data; and

means for generating a boundary in a feature space, said boundary substantially separating training data assigned to one class from training data assigned to other classes, wherein said other classes are classes associated with training data that is out of a normal operation range of said equipment; and

means for storing said boundary in memory.

28. The apparatus of claim 27 wherein said equipment comprises power generation equipment.

29. The apparatus of claim 27 wherein said measurements comprise data elements collected by a plurality of sensors positioned at a plurality of locations on said equipment.

30. The apparatus of claim 29 wherein each sensor in said plurality of sensors comprises an independent sensor.

31. The apparatus of claim 27 further comprising:

means collecting an element of operational data;

means for comparing said element of operational data to said feature space; and

means for determining whether said element of operational data is in said one class.

32. The apparatus of claim 31 wherein said means for determining whether said element of operational data is in said one class comprises:

means for determining whether said element of operational data represents an out-of-range condition for said equipment.

33. The apparatus of claim 27 wherein said training data comprises a plurality of vectors.

34. The apparatus of claim 27 wherein said means for generating comprises means for estimating said boundary using a Gaussian kernel.

35. The apparatus of claim 34 wherein said Gaussian kernel is defined as:

$\exp \left(\frac{-{<br><br>x_i-x_j<br><br>}^2}{2<br><br><br><br>{\mathrm{\sigma <br><br>}}^2}\right)$

where x_i and x_j are two vectors in the input space and σ

is the width of the Gaussian kernel.

36. The apparatus of claim 27 wherein said one class is a class assigned to training data that is within a normal operating range of said equipment.