Thermocouple failure detection in power generation turbines

• US 6,363,330 B1
• Filed: 10/09/1998
• Issued: 03/26/2002
• Est. Priority Date: 04/10/1998
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Morjaria, Mahesh Amritlal, Alag, Satnam Singh Sampuran
• Primary Examiner(s): SHAH, KAMINI S
• Application Number:
  US09/169,334
• Time in Patent Office: 1,264 Days
• Field of Search: 702/130, 702/131, 702/132, 236/158.B, 374/179, 290/40.R, 290/4.R, 290/40.B, 700/287, 60/392.81
• US Class Current: 702/132
• CPC Class Codes: G01K 7/026 Arrangements for signalling failure or disconnection of thermocouples

Abstract

Claims

1. A method of thermocouple failure detection in power generation turbines comprising:

creating redundancy estimates from temperature readings generated by said thermocouples;

predicting an expected value from each temperature reading;

comparing said temperature readings and said redundancy estimates with said expected value;

fusing said redundancy estimates and said expected values to generate a fused thermocouple value;

and generating a thermocouple confidence by comparing said fused thermocouple value and said temperature readings;

wherein said predicting of expected values is completed by creating a history of the process.

2. A method in accordance with claim 1, wherein said redundancy estimates are created by physical redundancy using multiple thermocouples to monitor the same temperature.

3. A method in accordance with claim 1, wherein said redundancy estimates are created by functional redundancy using measurements from non-redundant thermocouples to estimate temperature.

4. A method in accordance with claim 3, wherein said functional redundancy is represented as x_i=ƒ

(y₁, - - - y<sub>i−

1</sub>, y_i+1, - - - y_n) where y₁, y₂ - - - y_n are measurements from n thermocouples measuring temperatures at the point each respective thermocouple is located, x_i.

5. A method in accordance with claim 4, wherein said function is solved using nonlinear regression.

6. A method in accordance with claim 4, wherein said function is solved using a neural network.

7. A method in accordance with claim 4, wherein said function is solved using an analytical first principal model.

8. A method in accordance with claim 3, wherein said functional redundancy estimates are created by monitoring temperature signals at two adjoining thermocouples, one on either side of an at-issue thermocouple.

9. A method in accordance with claim 1, wherein said history is created by building an adaptive time-series model incorporated in a state space form for predicting variables at a next sampling time.

10. A method in accordance with claim 9, wherein the state of the variable of said next sampling period is:

x(k+1)=x(k)+u(k)+w(k) where u(k) is an unknown input to be estimated, w(k) is noise, and x(k) is the temperature detected by a respective thermocouple.

11. A method in accordance with claim 10 wherein x(k) and w(k) are independent, random and Gaussian distributed.

12. A method in accordance with claim 10, wherein u(k) is estimated using a polynomial technique.

13. A method in accordance with claim 10, wherein u(k) is estimated using a neural network.

14. A method in accordance with claim 10, wherein u(k) is estimated using a fuzzy logic technique.

15. A method in accordance with claim 1, wherein comparing said temperature readings and said redundancy estimates with said expected value is completed using a validation gate.

16. A method in accordance with claim 15, wherein said validation gate is defined by a 3 sigma area.

17. A method in accordance with claim 1, wherein said thermocouple confidence is calculated for each thermocouple within each set of thermocouples that a respective thermocouple belongs.

18. A method in accordance with claim 17, wherein a failed thermocouple is detected if the thermocouple confidence for a respective thermocouple is zero in each set of thermocouples that a respective thermocouple belongs.

19. A thermocouple failure detection system (10) for a power generation turbine assembly (12) comprising:

a plurality of thermocouples (14) positioned to monitor temperature within said turbine assembly (12); and

a computer (16) coupled to said plurality of thermocouples (14);

wherein temperature signals generated by respective thermocouples (14) are provided to said computer (16) and utilized within an algorithm that is embedded within said computer to generate a thermocouple confidence value, which value provides an indication of thermocouple operation or alternatively thermocouple failure;

wherein said algorithm is performed in a language selected from the group consisting of C, C++, JAVA, basic, Visual Basic, MATLAB, and Fortran.

20. A system in accordance with claim 19, wherein said computer is selected from the group consisting of a mainframe, a microcomputer, a minicomputer and a supercomputer.

21. A system in accordance with claim 19 wherein said algorithm comprises a method of thermocouple failure detection in power generation turbines comprising:

creating redundancy estimates from temperature readings generated by said thermocouples;

predicting an expected value from each temperature reading;

comparing said temperature readings and said redundancy estimates with said expected value;

fusing said redundancy estimates and said expected values to generate a fused thermocouple value;

and generating a thermocouple confidence by comparing said fused thermocouple value and said temperature readings.

22. A system in accordance with claim 21, wherein said redundancy estimates are created by physical redundancy using multiple thermocouples to monitor the same temperature.

23. A system in accordance with claim 21, wherein said redundancy estimates are created by functional redundancy using measurements from non-redundant thermocouples to estimate temperature.

24. A system in accordance with claim 23, wherein said functional redundancy is represented as x_i=ƒ

(y₁, - - - y<sub>i−

1</sub>, y_i+1, - - - y_n) where y₁, y₂ - - - y_n are measurements from n thermocouples measuring temperatures at the point each respective thermocouple is located, x_i.

25. A system in accordance with claim 24, wherein said function is solved using nonlinear regression.

26. A system in accordance with claim 24, wherein said function is solved using a neural network.

27. A system in accordance with claim 24, wherein said function is solved using an analytical first principal model.

28. A system in accordance with claim 23, wherein said functional redundancy estimates are created by monitoring temperature signals at two adjoining thermocouples, one on either side of an at-issue thermocouple.

29. A system in accordance with claim 21, wherein said predicting of expected values in completed by creating a history of the process.

30. A system in accordance with claim 29, wherein said history is created by building an adaptive time-series model incorporated in a state space form for predicting variables at next sampling time.

31. A system in accordance with claim 30, wherein the state of the variable of said next sampling period is:

x(k+1)=x(k)+u(k)+w(k) where u(k) is an unknown input to be estimated, w(k) is noise, and x(k) is the temperature detected by a respective thermocouple.

32. A system in accordance with claim 31 wherein x(k) and w(k) are independent, random and Gaussian distributed.

33. A system in accordance with claim 31, wherein u(k) is estimated using a polynomial technique.

34. A system in accordance with claim 31, wherein u(k) is estimated using a neural network.

35. A system in accordance with claim 31, wherein u(k) is estimated using a fuzzy logic technique.

36. A system in accordance with claim 21, wherein comparing said temperature readings and said redundancy estimates with said expected value is completed using a validation gate.

37. A system in accordance with claim 21, wherein said validation gate is defined by a 3 sigma area.

38. A system in accordance with claim 21, wherein said thermocouple confidence is calculated for each thermocouple within each set of thermocouples that a respective thermocouple belongs.

39. A system in accordance with claim 38, wherein a failed thermocouple is detected if the thermocouple confidence for a respective thermocouple is zero in each set of thermocouples that a respective thermocouple belongs.