Bayesian belief networks for industrial processes

• US 6,415,276 B1
• Filed: 08/13/1999
• Issued: 07/02/2002
• Est. Priority Date: 08/14/1998
• Status: Expired due to Term

• Current Assignee: Science & Technology Corporation
• Original Assignee: The University of New Mexico
• Inventors: Aradhye, Hrishikesh B., Heger, Arlen S.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/374,438
• Time in Patent Office: 1,054 Days
• Field of Search: 706/24, 706/52, 702/84, 702/183
• US Class Current: 706/52
• 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 9/02 electric
• CPC Class Codes: G06F 18/29 Graphical models, e.g. Bayesian networks

Abstract

A method and an apparatus for diagnosis of sensors and/or processes through use of Bayesian belief networks. More specifically, the method and apparatus of the present invention achieve sensor and/or process fault detection, isolation, and accommodation.

Claims

1. A method for diagnosis of sensors comprising:

a) providing at least one sensor-status node wherein each sensor-status node comprises a known probability table;

b) providing at least one process-variable node wherein each process-variable node comprises a known probability table;

c) providing at least one sensor-reading node wherein each sensor-reading node comprises a probability table conditional on at least two known probability tables;

d) providing the at least one sensor-reading node with at least one sensor reading;

e) inferring a status of at least one sensor.

2. The method of claim 1 wherein steps a) through e) comprise at least one Bayesian belief network.

3. The method of claim 1 further comprising estimating at least one value of at least one process variable.

4. The method of claim 1 further comprising updating at least one known probability table after providing the at least one sensor-reading node with at least one sensor reading.

5. The method of claim 4 further comprising monitoring of the at least one updated probability table for assessing at least one member selected from the group consisting of sensor fault detection, sensor fault classification, process fault detection, and process fault classification.

6. The method of claim 1 wherein inferring further comprises process modeling.

7. The method of claim 6 wherein process modeling comprises at least one member selected from the group consisting of continuous-value model, discrete-value model, linearized model, neural network model, fuzzy logic model, steady-state model, unsteady-state model, static model, and dynamic model.

8. The method of claim 1 further comprising inputting data from an external observer.

9. The method of claim 1 wherein providing at least one sensor reading comprises providing a reading from at least one member selected from the group consisting of temperature sensors, concentration sensors, pH sensors, level sensors, flowrate sensors, and volume sensors.

10. A sensor diagnostic apparatus comprising:

calculation means for calculating probabilities of a Bayesian belief network wherein said network comprises at least one sensor-status node, at least one process-variable node, and at least one sensor-reading node;

input means for inputting information to said calculation means;

storage means for storing information; and

output means for outputting information from said calculation means.

11. The apparatus of claim 10 wherein said calculation means further comprises solution means for solving at least one process model.

12. The apparatus of claim 10 wherein said calculating means further comprises estimating means for estimating at least one value of at least one process variable.

13. The apparatus of claim 10 wherein said calculating means further comprises updating means for updating at least one known probability table.

14. The apparatus of claim 10 wherein said calculating means further comprises monitoring means for monitoring of at least one updated probability table for assessing at least one member selected from the group consisting of sensor fault detection, sensor fault classification, process fault detection, and process fault classification.

15. The apparatus of claim 10 wherein said calculating means further comprises process modeling.

16. The apparatus of claim 15 wherein said process modeling comprises at least one member selected from the group consisting of continuous-value model, discrete-value model, linearized model, neural network model, fuzzy logic model, steady-state model, unsteady-state model, static model, and dynamic model.

17. The apparatus of claim 10 wherein said information comprises data from an external observer.

18. The apparatus of claim 10 wherein said information comprises information from a reading from at least one member selected from the group consisting of temperature sensors, concentration sensors, pH sensors, level sensors, flowrate sensors, and volume sensors.