SYSTEM AND METHOD FOR PREDICTION OF GAS TURBINE TRIPS DUE TO GAS CONTROL VALVE FAILURES

• US 20090055130A1
• Filed: 08/23/2007
• Published: 02/26/2009
• Est. Priority Date: 08/23/2007
• Status: Active Grant

• Current Assignee: GE Infrastructure Technology, LLC (GE Vernova, Inc.)
• Original Assignee: General Electric Company
• Inventors: Pandey, Achalesh Kumar, Zaremba, Slawomir Marian, Jammu, Vinay Bhaskar, Babu, Ravi Yoganatha
• Granted Patent: US 8,250,017 B2
• Time in Patent Office: Days
• US Class Current: 702/183
• 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

Abstract

Systems and methods for prediction of gas turbine trips due to component failures such as electro-hydraulic valve (gas control valve) system failures. Exemplary embodiments include prediction of gas turbine trips due to component failures, the method including collecting raw gas turbine operational data and using the raw gas turbine operational data to generate a prognostic indicator for the prediction of a turbine trip due to the failed gas control valves.

Claims

1. A method for the prediction of gas turbine trips due to gas control valve failures, the method comprising:

collecting raw gas turbine operational data; and

using the raw gas turbine operational data to generate a prognostic indicator for the prediction of a turbine trip due to the failed gas control valves.

2. The method as claimed in claim 1, further comprising:

extracting gas control valve features indicative of a turbine trip from the raw gas turbine operational data;

applying a rule set to the extracted features;

determining a fused belief of gas control valve failure, based on an apriori probability and the extracted features; and

classifying the predicted gas control valve failure,

wherein extracting features from the raw gas turbine operational data includes applying domain knowledge to the extracted features.

3. The method as claimed in claim 2 further comprising calculating apriori probability and applying the domain knowledge to the calculated apriori probability.

4. The method as claimed in claim 3 wherein determining a fused belief of failure further comprises applying the calculated apriori probability to the determined fused belief of failure.

5. The method as claimed in claim 1 wherein the gas control valve failure is selected from the group consisting of:

electro hydraulic servo system faults;

sticking of spool due to poor hydraulic oil quality;

loose connection LVDT failures;

shorting LVDT failures, sticking and severed valve stem conditions;

grounding servo coil failure;

open circuit servo coil failure;

loose connection servo coil failure;

last chance filter choking;

hydraulic supply pressure loss/fluctuation;

null bias drift;

flapper nozzle choking; and

air ingress in hydraulic system.

6. The method as claimed in claim 2 wherein the extracted features includes at least one of position error;

cumulative position error, slope of cumulative position error, difference from expected based on DLN_Mode, standard deviation of the difference from expected based on DLN_Mode, and standard deviation of current.

7. A system for the prediction of gas turbine trips due to gas control valve failures, the system comprising:

a turbine;

a local computer coupled to the turbine for collecting raw turbine operational data related to a gas control valve failure from the turbine;

a remote computer coupled to the local computer;

a process, residing on the local computer, for computing a prediction of trips, the process including instructions to generate a prognostic indicator for the prediction of a turbine trip due to the failed gas control valves; and

a database coupled to the remote computer for storing information related to a trip of the turbine.

8. The system as claimed in claim 7 wherein the local computer includes a first process for extracting features from the raw component data, the raw component data and the features being indicative of gas control valve failure.

9. The system as claimed in claim 8 wherein the remote computer includes a second process for applying a rule set to the extracted features.

10. The system as claimed in claim 9 wherein the second process includes instructions to:

perform a trip reduction analysis; and

generate recommendations to prevent a trip.

11. The system as claimed in claim 9 wherein the database comprises a computer readable medium having a plurality of data nodes stored on the medium and representing a data structure, comprising:

a first data node stored in a first region of a range of memory addresses in the medium, the first data node representing a time to trip indicator;

a second data node stored in a second region of a range of memory addresses in the medium, the second data node representing trip mitigation recommendations; and

an index stored in an index region of the range of memory addresses, the index storing relationship information between the first and second data nodes, wherein the second process includes instructions to generate appropriate action to mitigate the trip.

12. The system as claimed in claim 11 wherein the appropriate action to mitigate the trip is at least one of a graceful shutdown of the turbine and a repair of the failed gas control valve.

13. The method as claimed in claim 8 wherein the gas control valve failure is selected from the group consisting of:

electro hydraulic servo system faults;

sticking of spool due to poor hydraulic oil quality;

loose connection LVDT failures;

shorting LVDT failures;

sticking and severed valve stem conditions;

grounding servo coil failure;

open circuit servo coil failure;

loose connection servo coil failure;

last chance filter choking;

hydraulic supply pressure loss/fluctuation;

null bias drift;

flapper nozzle choking; and

air ingress in hydraulic system.

14. The method as claimed in claim 8 wherein the extracted features includes at least one of position error, cumulative position error, slope of cumulative position error, difference from expected based on DLN_Mode, standard deviation of the difference from expected based on DLN_Mode, and standard deviation of current.

15. In a power plant having a turbine, a method for the prediction of gas turbine trips due to gas control valve failures, the method comprising:

collecting gas control valve failure data having failure signatures from the turbine for collection in a remote computer;

processing the gas control valve failure data to extract features based on the failure signatures and based on domain knowledge of the turbine;

applying a rule set on the extracted features based on apriori probability and the domain knowledge; and

determining a fused belief of failure calculated from the rule set applied to the extracted features.

16. The method as claimed in claim 15 further comprising:

classifying the component failure from the fused belief of failure; and

generating a prognostic indicator, wherein the prognostic indicator includes a band of at least one of no trip, low probability of trip, and high probability of trip.

17. The method as claimed in claim 15 wherein the apriori probability is calculated in the remote computer based on sensor fusion methods including at least one of Neamen Pearson method, Demster Shafer method and Bayesian Fusion as follows:

$P<br><br>\left(T/E_i\right)=\frac{P<br><br>\left(T\right)<br><br>\underset{i=1<br><br>\mathrm{\dots <br><br>}<br><br><br><br>n}{\prod <br><br>}<br><br>P<br><br>\left(E_i/T\right)}{P<br><br>\left(T\right)<br><br>\underset{i=1<br><br>\mathrm{\dots <br><br>}<br><br><br><br>n}{\prod <br><br>}<br><br>P<br><br>\left(E_i/T\right)+P-\left(T\right)<br><br>\underset{i=1<br><br>\mathrm{\dots <br><br>}<br><br><br><br>n}{\prod <br><br>}<br><br>P<br><br>\left(E_i/-T\right)}$

wherein Ei is a conditional probability of obtaining features of the component failure data, P(T) is an apriori belief value for a trip, P(−

T) is an apriori belief for no trip, P(Ei/F) is a conditional probability for obtaining features for a trip event, and P(Ei/−

T) is a conditional probability for obtaining features for a non-trip event.

18. The method as claimed in claim 15 wherein the gas control valve failure is selected from the group consisting of:

electro hydraulic servo system faults;

sticking of spool due to poor hydraulic oil quality;

loose connection LVDT failures;

shorting LVDT failures;

sticking and severed valve stem conditions;

grounding servo coil failure;

open circuit servo coil failure;

loose connection servo coil failure;

last chance filter choking;

hydraulic supply pressure loss/fluctuation;

null bias drift;

flapper nozzle choking; and

air ingress in hydraulic system.

19. The method as claimed in claim 15 wherein the extracted features includes at least one of position error, cumulative position error, slope of cumulative position error, difference from expected based on DLN_Mode, standard deviation of the difference from expected based on DLN_Mode, and standard deviation of current.

20. The method as claimed in claim 15 further comprising generating an appropriate action to mitigate a trip of the gas turbine, wherein the appropriate action to mitigate the trip is at least one of a graceful shutdown of the turbine and a repair of the failed component.