Sensor fault detection, isolation and accommodation

• US 6,598,195 B1
• Filed: 08/21/2000
• Issued: 07/22/2003
• Est. Priority Date: 08/21/2000
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Wiseman, Matthew William, Adibhatla, Sridhar
• Primary Examiner(s): Ton, David
• Application Number:
  US09/643,007
• Time in Patent Office: 1,065 Days
• Field of Search: 361/184, 702/65, 702/85, 701/34, 714/745, 714/25
• US Class Current: 714/745
• CPC Class Codes: G01D 3/08 with provision for safeguarding the apparatus, e.g. against abnormal operation, against breakdown
• CPC Class Codes: G05B 23/0275 Fault isolation and identification, e.g. classify fault; estimate cause or root of failure

Abstract

A sensor in an engineering system can be tested to detect, isolate and accommodate faults. Initially, a modeled sensor value of each actual sensor value is generated as a function of a plurality of other sensors. An absolute value of a difference between the actual sensor value and the modeled sensor value is then computed and compared to a predetermined threshold. A sensor fault is detected if the difference is greater than the predetermined threshold. Once a sensor fault is detected, it is isolated using hypothesis testing and maximum wins strategies. After the fault is isolated, the fault is accommodated by substituting the modeled sensor value for the actual sensor value.

Claims

1. A method of detecting sensor faults in a system, the method comprising the steps of:

generating a modeled sensor value of each actual sensor value as a function of a plurality of other sensors;

computing an absolute value of a difference between the actual sensor value and the modeled sensor value;

comparing the difference to a predetermined threshold; and

declaring the sensor as faulted if the difference is greater than the predetermined threshold.

2. A method as claimed in claim 1 wherein the plurality of other sensors comprises all sensors except the sensor being modeled.

3. A method as claimed in claim 1 wherein the plurality of other sensors comprises a subset of all sensors.

4. A method as claimed in claim 1 wherein the step of generating a modeled sensor value comprises the step of applying a linear model matrix.

5. A method as claimed in claim 1 wherein the step of generating a modeled sensor value comprises the step of applying a nonlinear model.

6. A method of detecting and isolating sensor faults in a system, the method comprising the steps of:

generating a modeled sensor value of each actual sensor value as a function of a plurality of other sensors;

computing an absolute value of a difference between the actual sensor value and the modeled sensor value;

comparing the difference to a predetermined threshold;

declaring the sensor as faulted if the difference is greater than the predetermined threshold; and

applying fault isolation to the potentially faulted sensor.

7. A method as claimed in claim 6 wherein the plurality of other sensors comprises all sensors except the sensor being modeled.

8. A method as claimed in claim 6 wherein the plurality of other sensors comprises a subset of all sensors.

9. A method as claimed in claim 6 wherein the step of generating a modeled sensor value comprises the step of applying a linear model matrix.

10. A method as claimed in claim 6 wherein the step of generating a modeled sensor value comprises the step of applying a nonlinear model.

11. A method as claimed in claim 6 wherein the step of applying fault isolation comprises the steps of:

performing hypothesis testing on the potentially faulted sensor; and

performing maximum wins strategies on the potentially faulted sensor.

12. A method as claimed in claim 11 wherein the step of performing hypothesis testing comprises the steps of:

hypothesizing that each sensor is faulted in turn; and

concluding that the hypothesis is correct if replacing the sensor'"'"'s value by its associated modeled sensor value and then repeating fault detection causes subsequent faults detected to decrease.

13. A method as claimed in claim 11 wherein the step of performing maximum wins strategies comprises the steps of:

computing a normalized error; and

isolating fault to a sensor with a largest normalized error when the step of performing hypothesis testing indicates more than one sensor as being faulted.

14. A method of detecting, isolating and accommodating sensor faults in a system, the method comprising the steps of:

generating a modeled sensor value of each actual sensor value as a function of a plurality of other sensors;

computing an absolute value of a difference between the actual sensor value and the modeled sensor value;

comparing the difference to a predetermined threshold;

isolating the sensor as potentially faulted if the difference is greater than the predetermined threshold; and

accommodating the sensor fault by substituting the actual sensor value of the faulted sensor with its modeled sensor value.

15. A method as claimed in claim 14 wherein the plurality of other sensors comprises all sensors except the sensor being modeled.

16. A method as claimed in claim 14 wherein the plurality of other sensors comprises a subset of all sensors.

17. A method as claimed in claim 14 wherein the step of generating a modeled sensor value comprises the step of applying a linear model matrix.

18. A method as claimed in claim 14 wherein the step of generating a modeled sensor value comprises the step of applying a nonlinear model.

19. A method as claimed in claim 14 wherein the step of isolating the sensor comprises the steps of:

performing hypothesis testing on the potentially faulted sensor; and

performing maximum wins strategies on the potentially faulted sensor.

20. A method as claimed in claim 19 wherein the step of performing hypothesis testing comprises the steps of:

hypothesizing that each sensor is faulted in turn; and

concluding that the hypothesis is correct if replacing the sensor'"'"'s value by its associated modeled sensor value and then repeating fault detection causes subsequent faults detected to decrease.

21. A method as claimed in claim 19 wherein the step of performing maximum wins strategies comprises the steps of:

computing a normalized error; and

isolating fault to a sensor with a largest normalized error when the step of performing hypothesis testing indicates more than one sensor as being faulted.

22. A method as claimed in claim 14 wherein modeled sensor value used for sensor accommodation can differ from the modeled sensor value used for detection.