Methods and apparatus for model based diagnostics

• US 7,020,595 B1
• Filed: 11/26/1999
• Issued: 03/28/2006
• Est. Priority Date: 11/26/1999
• Status: Expired due to Term

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Adibhatla, Sridhar, Ashby, Malcolm J.
• Primary Examiner(s): FREJD, RUSSELL WARREN
• Application Number:
  US09/449,773
• Time in Patent Office: 2,314 Days
• Field of Search: 703/6, 703/7, 703/8, 700/28, 700/29, 700/30, 700/31, 700/32, 714/25, 714/37, 714/47, 714/745, 706/20, 706/52, 706/47, 701/3, 701/100, 731/181
• US Class Current: 703/7
• CPC Class Codes: F02C 9/00 Controlling gas-turbine plants; Controlling fuel supply in air- breathing jet-propulsion plants controlling air intakes F02C7/057; controlling turbines F01D; controlling compressors F04D27/00; controlling in general G05
• CPC Class Codes: F05D 2270/71 synthesized, i.e. parameter computed by a mathematical model
• 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
• CPC Class Codes: G05B 23/0275 Fault isolation and identification, e.g. classify fault; estimate cause or root of failure

Abstract

Systems and methods for performing module-based diagnostics are described. In an exemplary embodiment, sensor values from an actual engine plant are input to an engine component quality estimator which generates performance estimates of major rotating components. Estimated performance differences are generating by comparing the generated performance estimates to a nominal quality engine. The estimated performance differences, which are indicative of component quality, are continuously updated and input to a real-time model of the engine. The model receives operating conditional data and the quality estimates are used to adjust the nominal values in the model to more closely match the model values to the actual plant. Outputs from the engine model are virtual parameters, such as stall margins, specific fuel consumption, and fan/compressor/turbine efficiencies. The virtual parameters are combined with the sensor values from the actual engine plant in a fault detection and isolation classifier to identify abnormal conditions and/or specific fault classes, and output a diagnosis.

Claims

1. A system for monitoring engine performance, said system comprising:

an engine model;

a component quality estimator coupled to said engine model; and

a fault detection and isolation classifier coupled to said engine model and to said component quality estimator, said engine component quality estimator generates performance estimates, said model generates virtual parameters based on the performance estimates and engine operating conditions.

2. A system in accordance with claim 1 wherein said model comprises a component level model of an engine.

3. A system in accordance with claim 1 wherein said component quality estimator comprises a linear regressor.

4. A system in accordance with claim 1 wherein said fault detection and isolation classifier comprises a feed-forward neural network.

5. A system in accordance with claim 1 wherein said fault detection and isolation classifier comprises a linear regressor.

6. A system in accordance with claim 1 wherein said fault 20 detection and isolation classifier identifies fault conditions based on said virtual parameters and engine sensor values.

7. A system in accordance with claim 1 wherein said engine virtual parameters are subtracted from engine sensor values to generate sensor deltas supplied to said fault detection and isolation classifier.

8. A system in accordance with claim 7 wherein reference sensor deltas are generated and are compared to later obtained sensor deltas to generate sensor delta-deltas supplied to said fault detection and isolation classifier.

9. A system in accordance with claim 1 wherein said model 5 generates virtual sensor values based on engine operating conditions, and said component quality estimator generates quality estimates based on said virtual sensor values and engine sensor values.

10. A system in accordance with claim 9 wherein said engine component quality estimator generates an initial quality estimate and a final quality estimates, and the initial and final quality estimates are compared to generate a delta quality supplied to said fault detection and isolation classifier.

11. A method for monitoring engine performance, said method comprising the steps of:

supplying engine operating conditions to an engine model;

supplying engine sensor values to an engine component quality estimator;

generating a fault detection based at least in part on engine model outputs, and at least in part on estimator outputs; and

supplying the estimator outputs to the engine model.

12. A method in accordance with claim 11 wherein the engine model generates sensor estimates, and said method further comprises the step of generating sensor deltas by comparing the sensor estimates to engine sensor values.

13. A method in accordance with claim 11 further comprising the step of supplying the engine model outputs to the component quality estimator.

14. A method in accordance with claim 13 further comprising the steps of:

operating the engine component quality estimator to generate an initial quality estimate and a final quality estimate; and

comparing the initial quality estimate to the final quality estimate to generate a delta quality.

15. A method in accordance with claim 11 wherein the engine model generates sensor estimates, and said method further comprises the steps of:

generating a reference sensor delta by comparing the sensor estimates at a first operating point;

generating a current sensor delta at a second operating point; and

generating a sensor delta—

delta by comparing the reference sensor delta to the current sensor delta.

16. A method for monitoring engine performance, said method comprising the steps of:

generating a reference sensor delta by comparing sensor values with sensor estimates at a first operating point;

generating a current sensor delta at a second operating point;

generating a sensor delta—

delta by comparing the reference sensor delta to the current sensor delta; and

generating a fault detection based upon the sensor delta-deltas.