Fault detection system and method for turbine engine fuel systems

• US 20060195248A1
• Filed: 02/14/2005
• Published: 08/31/2006
• Est. Priority Date: 02/14/2005
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Kim, Kyusung, Uluyol, Onder, Ball, Charles M.
• Granted Patent: US 7,693,643 B2
• Time in Patent Office: Days
• US Class Current: 701/100
• CPC Class Codes: F01D 21/14 responsive to other specific conditions
• CPC Class Codes: F02C 7/22 Fuel supply systems
• CPC Class Codes: F02C 7/264 Ignition
• CPC Class Codes: F02C 9/28 Regulating systems responsive to plant or ambient parameters, e.g. temperature, pressure, rotor speed F02C9/30 - F02C9/38, F02C9/44 take precedence
• CPC Class Codes: F02C 9/46 Emergency fuel control
• CPC Class Codes: F05D 2270/0831 indirectly, at the exhaust
• CPC Class Codes: F05D 2270/304 Spool rotational speed
• CPC Class Codes: F05D 2270/44 active, predictive, or anticipative
• CPC Class Codes: F05D 2270/707 fuzzy logic
• CPC Class Codes: F05D 2270/708 with comparison tables
• CPC Class Codes: F05D 2270/709 with neural networks
• CPC Class Codes: Y02T 50/60 Efficient propulsion technologies, e.g. for aircraft

Abstract

A system and method is provided that facilitates improved fault detection. The fault detection system provides the ability to detect symptoms of fault in the fuel system of a turbine engine. The fault detection system captures selected data from the turbine engines that is used to characterize the performance of the fuel system. The fault detection system includes a feature extractor that extracts salient features from the selected sensor data. The extracted salient features are passed to a classifier that analyzes the extracted salient features to determine if a fault is occurring or has occurred in the turbine engine fuel system. Detected faults can then be passed to a diagnostic system where they can be passed as appropriate to maintenance personnel.

Claims

1. A fault detection system for detecting faults in a turbine engine fuel system, the fault detection system comprising:

a sensor data processor, the sensor data processor receiving sensor data from the turbine engine and selecting a subset of sensor data, the subset of sensor data including time at lightoff, time at peak engine temperature rate of change, engine speed at lightoff time, engine speed at peak engine temperature rate of change time, engine speed at 3^rd phase time, engine temperature at peak engine temperature rate of change time, engine speed rate of change at 3^rd phase time, engine temperature rate of change at light off time and engine temperature rate of change at 3^rd phase time;

a feature extractor, the feature extractor receiving the subset of sensor data and generating principal components from the subset of sensor data; and

a classifier, the classifier receiving the generated principal components and classifying the principal components to determine if a fault occurred in the turbine engine fuel system.

2. The system of claim 1 wherein the classifier comprises a clustering system, the clustering system classifying the generated principal components into one of a plurality of clusters.

3. The system of claim 2 wherein the clustering system comprises a fuzzy clustering system, and wherein the fuzzy clustering system is trained using historical sensor data including historical sensor data obtained from turbine engines with normal fuel system components and historical sensor data obtained from turbine engines with abnormal fuel system components.

4. The system of claim 1 wherein the feature extractor performs a principal component analysis to generate the principal components, and wherein the classifier classifies using less than nine principal components.

5. The system of claim 1 wherein the classifier is trained to determine if a fault occurred in a hydro-mechanical assembly of the turbine engine fuel system.

6. A fault detection system for detecting faults in a hydro-mechanical assembly of a turbine engine fuel system, the fault detection system comprising:

a principal component analysis (PCA) feature extractor, the PCA feature extractor receiving sensor data from the turbine engine, the PCA feature extractor deriving a plurality of principal components from the sensor data; and

a fuzzy clustering classifier, the fuzzy clustering classifier receiving the plurality of principal components and classifying the plurality of principal components into one of a plurality of clusters to determine if a fault occurred in the hydro-mechanical assembly of the turbine engine fuel system.

7. The system of claim 6 wherein the fuzzy clustering classifier is trained using historical sensor data including historical sensor data obtained from turbine engines with normal hydro-mechanical assemblies and historical sensor data obtained from turbine engines with abnormal hydro-mechanical assemblies.

8. The system of claim 6 wherein the sensor data comprises a set of nine parameters, and wherein the nine parameters include time at lightoff, time at peak engine temperature rate of change, engine speed at lightoff time, engine speed at peak engine temperature rate of change time, engine speed at 3^rd phase time, engine temperature at peak engine temperature rate of change time, engine speed rate of change at 3^rd phase time, engine temperature rate of change at light off time and engine temperature rate of change at 3^rd phase time.

9. The system of claim 8 wherein the plurality of principal components comprises five principal components generated from the nine parameters, and wherein the five principal components are selected based on statistical distributions of principal components generated from historical sensor data in the presence and absence of hydro-mechanical assembly anomalies.

10. The system of claim 6 wherein the sensor data comprises a selected set of sensor data, the selected set of sensor selected based on statistical distributions of historical sensor data in the presence and absence of hydro-mechanical assembly anomalies.

11. The system of claim 10 wherein the historical sensor data is standardized for ambient conditions.

12. The system of claim 10 wherein the historical sensor data is filtered based on power level angle profiles of the turbine engines.

13. The system of claim 10 wherein the historical sensor data includes only historical data from normal auto mode startups of the turbine engines.

14. A method of detecting faults in a turbine engine fuel system, the method comprising the steps:

selecting a set of sensor data from the turbine engine;

deriving a plurality of principal components from the selected set of sensor data using a principal component analysis; and

classifying the plurality of principal components into one of a plurality of clusters to determine if a fault occurred in the turbine engine fuel system.

15. The method of claim 14 wherein the plurality of clusters are obtained by training using historical sensor data, the historical sensor data including historical sensor data obtained from turbine engines with normal fuel systems and historical sensor data obtained from turbine engines with abnormal fuel systems.

16. The method of claim 14 wherein the selected set of sensor data comprises nine parameters, and wherein the nine parameters include time at lightoff, time at peak engine temperature rate of change, engine speed at lightoff time, engine speed at peak engine temperature rate of change time, engine speed at 3^rd phase time, engine temperature at peak engine temperature rate of change time, engine speed rate of change at 3^rd phase time, engine temperature rate of change at light off time and engine temperature rate of change at 3^rd phase time.

17. The method of claim 16 wherein the plurality of principal components comprises five principal components generated from the nine parameters, and wherein the five principal components are selected based on statistical distributions of principal components generated from historical sensor data in the presence and absence of hydro-mechanical assembly anomalies.

18. The method of claim 14 wherein the step of selecting a set of sensor data from the turbine engine comprises selecting based on statistical distributions of historical sensor data in the presence and absence of fuel system anomalies.

19. The method of claim 18 wherein the historical sensor data is standardized for ambient conditions.

20. The method of claim 18 wherein the historical sensor data is filtered based on power level angle profiles of the turbine engines.

21. The method of claim 18 wherein the historical sensor data includes only historical data from normal auto mode startups of the turbine engines.

22. The method of claim 14 wherein the plurality of clusters are obtained by training using historical sensor data, the historical sensor data including historical sensor data obtained from turbine engines with normal hydro-mechanical assemblies in the turbine engine fuel system and historical sensor data obtained from turbine engines with abnormal hydro-mechanical assemblies in the turbine engine fuel system, and wherein the historical data is standardized for ambient conditions and filtered based on power level angle profiles of the turbine engines.

23. The method of claim 14 wherein the step of selecting a set of sensor data from the turbine engine comprises selecting based on a previously performed correlation analysis of historical sensor data to identify indicative parameters for a selected failure mode in the fuel system.

24. An apparatus comprising:

a) a processor;

b) a memory coupled to the processor;

c) a fuel system fault detection program for detecting faults in a hydro-mechanical assembly of a turbine engine fuel system, the fuel system fault detection program residing in the memory and being executed by the processor, the fuel system fault detection program including;

i) a principal component analysis (PCA) feature extractor, the PCA feature extractor receiving sensor data from the turbine engine, the PCA feature extractor deriving a plurality of principal components from the sensor data; and

ii) a fuzzy clustering classifier, the fuzzy clustering classifier receiving the plurality of principal components and classifying the plurality of principal components into one of a plurality of clusters to determine if a fault occurred in the hydro-mechanical assembly of the turbine engine fuel system.

25. A program product comprising:

a) a fuel system fault detection program for detecting faults in a hydro-mechanical assembly of a turbine engine fuel system, the fault detection program including;

i) a principal component analysis (PCA) feature extractor, the PCA feature extractor receiving sensor data from the turbine engine, the PCA feature extractor deriving a plurality of principal components from the sensor data; and

ii) a fuzzy clustering classifier, the fuzzy clustering classifier receiving the plurality of principal components and classifying the plurality of principal components into one of a plurality of clusters to determine if a fault occurred in the hydro-mechanical assembly of the turbine engine fuel system; and

b) computer-readable signal bearing media bearing said program.