Neural network/expert system process control system and method

• US 5,121,467 A
• Filed: 08/03/1990
• Issued: 06/09/1992
• Est. Priority Date: 08/03/1990
• Status: Expired due to Term

• Current Assignee: Rockwell Automation Technologies Incorporated (Rockwell Automation, Inc.)
• Original Assignee: E. I. du Pont de Nemours and Company (Corteva, Inc.)
• Inventors: Skeirik, Richard D.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/562,327
• Time in Patent Office: 676 Days
• Field of Search: 364/513, 364/148, 364/149, 364/150, 364/151, 364/164, 364/165, 364/807, 364/700, 364/474.15, 364/500, 364/501, 364/502, 364/503, 395/11, 395/22, 395/23, 395/68, 395/906
• US Class Current: 706/10
• CPC Class Codes: G05B 13/029 using neural networks and expert systems
• CPC Class Codes: G06N 3/042 Knowledge-based neural networks; Logical representations of neural networks
• CPC Class Codes: Y10S 706/906 Process plant

Abstract

A neural network/expert system process control system and method combines the decision-making capabilities of expert systems with the predictive capabilities of neural networks for improved process control. Neural networks provide predictions of measurements which are difficult to make, or supervisory or regulatory control changes which are difficult to implement using classical control techniques. Expert systems make decisions automatically based on knowledge which is well-known and can be expressed in rules or other knowledge representation forms. Sensor and laboratory data is effictively used. In one approach, the output data from the neural network can be used by the controller in controlling the process, and the expert system can make a decision using sensor or lab data to control the controller(s). In another approach, the output data of the neural network can be used by the expert system in making its decision, and control of the process carried out using lab or sensor data. In another approach, the output data can be used both to control the process and to make decisions.

Claims

1. A computer-based process control method adapted to control a process using a regulatory controller, the process producing a product having at least one product property, the computer-based process control method comprising the steps of:

(1) predicting, using a neural network, output data from input data;

(2) supplying said output data to the regulatory controller for controlling the process; and

(3) using an expert system to make a decision, using data from the process, to directly or indirectly control the regulatory controller.

2. The computer-based process control method of claim 1, wherein step (3) further comprises the step of making said decision using said output data.

3. A computer-based process control method adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control method comprising the steps of:

(1) predicting, using a neural network, output data from input data;

(2) supplying said output data to an expert system; and

(3) using said expert system to make a decision, using said output data, to directly or indirectly control the controller.

4. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control system comprising:

(a) a neural network, comprising predicting means for predicting output data in accordance with input data and in accordance with weights;

(b) a controller, connected to directly or indirectly control the process, and connected to use said output data as a controller input data in place of a sensor input data; and

(c) an expert system, connected to use data from the process to make a decision and connected to directly or indirectly control the operation of said controller in accordance with said decision, comprising,

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base.

5. The computer neural network process control system of claim 4, wherein said expert system is also connected to use said output data to make said decision.

6. The computer neural network process control system of claim 4, wherein said neural network further comprises training means, responsive to a first training set, for adjusting said weights.

7. The computer neural network process control system of claim 6, further comprising lab means for producing lab data indicative of a measurement of the process, and means for entering said lab data for use in said training set.

8. The computer neural network process control system of claim 4, wherein said knowledge base includes knowledge used by said inference engine to make said decision.

9. The computer neural network process control system of claim 4, wherein said expert system is connected to directly or indirectly control the operation of the controller by changing a parameter of the controller.

10. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control system comprising:

(a) a neural network comprising

(1) first means for storing a first training input data with an associated first timestamp;

(2) second means for storing a first input data indicated by said associated first timestamp;

(3) first retrieving means, responsive to said first means for storing, for retrieving said first training input data indicated by said associated first timestamp;

(4) second retrieving means, responsive to said second means for storing, for retrieving said first input data indicated by said associated first timestamp;

(5) training means, responsive to said first retrieving means and said second retrieving means, for training said neural network in accordance with said first training input data and said first input data to produce a trained neural network; and

(6) predicting means for predicting output data in accordance with a second input data and said trained neural network;

(b) a controller connected to directly or indirectly control the process, which controller is connected to use said output data as its controller input data; and

(c) an expert system, connected to use data from the process to make a decision and connected to directly or indirectly control the operation of said controller in accordance with said decision, comprising,

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base.

11. The computer neural network process control system of claim 10, further comprising sensor means for providing said first input data and said second input data.

12. The computer neural network process control system of claim 10, further comprising monitoring means for producing action data indicative of a past action of a human operator of the process.

13. The computer neural network process control system of claim 10, further comprising output means for storing said output data with an associated second timestamp.

14. The computer neural network process control system of claim 10, further comprising lab means for producing lab data indicative of a measurement of the process, and means for entering said lab data for use in said first training input data.

15. A computer-based process control method adapted to control a process using a supervisory controller, the process producing a product having at least one product property, the computer-based process control method comprising the steps of:

(1) predicting, using a neural network, output data from input data;

(2) supplying said output data to the supervisory controller for controlling the process; and

(3) using an expert system to make a decision, using data from the process, to directly or indirectly control the supervisory controller.

16. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product, the computer-based process control system comprising:

(a) a neural network, comprising predicting means for predicting output data in accordance with input data and in accordance with weights;

(b) a controller, connected to directly or indirectly control the process, and connected to adjust a set point in accordance with said output data; and

(c) an expert system, connected to use data from the process to make a decision and connected to directly or indirectly control the operation of said controller in accordance with said decision, comprising,

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base.

17. The computer neural network process control system of claim 16, wherein said neural network further comprises training means, responsive to a first training set, for producing said weights.

18. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control system comprising:

(a) a neural network, comprising predicting means for predicting output data in accordance with input data and in accordance with weights;

(b) an expert system, connected to use said output data to make a decision and connected to directly or indirectly control the operation of the controller in accordance with said decision, comprising,

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base; and

(c) the controller, connected to directly or indirectly control the process.

19. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control system comprising:

(a) an actuator, for changing a controllable state of the process in accordance with a state of said actuator;

(b) an expert system, connected to use data from the process to make a decision as indicated by decision data, comprising;

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base; and

(c) a neural network, comprising predicting means for predicting output data in accordance with said decision data and in accordance with weights, said neural network connected to adjust said state of said actuator in accordance with said output data.

20. A computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control system comprising:

(a) a neural network, comprising predicting means for predicting output data in accordance with input data and in accordance with weights;

(b) a controller, connected to directly or indirectly control the process, and connected to adjust a set point in accordance with said output data;

(c) an expert system, connected to use said output data to make a decision and connected to directly or indirectly control the operation of said controller in accordance with said decision, comprising;

(1) a knowledge base, and

(2) an inference engine responsive to said knowledge base.

21. A modular computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the modular computer-based process control system comprising:

(a) a neural network module for predicting output data from input data;

(b) a controller, connected to directly or indirectly control the process and connected to use said output data as a controller input data in place of a sensor input data; and

(c) an expert system module, connected to use data from the process to make a decision and connected to control the operation of said controller in accordance with said decision.

22. The modular computer-based process control system of claim 21, wherein said expert system is further connected to use said output data in making said decision.

23. The modular computer-based process control system of claim 21, wherein said controller comprises a feedback control module supervising a regulatory controller.

24. The modular computer-based process control system of claim 21, wherein said controller is connected to directly or indirectly control the process by changing a first process condition affecting the product property;

and further comprising

(d) a feedforward control module means, connected to directly or indirectly control the process by changing said first process condition or a second process condition affecting the product property.

25. The modular computer-based process control system of claim 21, wherein said controller is connected to directly or indirectly control the process by changing a first process condition affecting the product property;

and further comprising

(d) a statistical test module means, connected to provide statistical data for directly or indirectly controlling the process by changing said first process condition or a second process condition affecting the product property.

26. The modular computer-based process control system of claim 21, further comprising a user interface providing a template for entering a size specification of said neural network module and/or a connectivity specification of said neural network module and/or a specification of a source of said input data, wherein said neural network module operates in accordance with said specification(s).

27. The modular computer-based process control system of claim 26, wherein said template comprises data pointers for specifying data to be used by said neural network module.

28. The modular computer-based process control system of claim 21, further comprising a user interface for configuring said neural network module using a limited set of natural language format specifications.

29. A modular computer neural network process control system adapted to control a process using a controller, the process producing a product having at least one product property, the modular computer-based process control system comprising:

(a) a neural network module for predicting output data from input data;

(b) an expert system module, connected to use said output data to make a decision and to produce decision data indicative of said decision; and

(c) a controller, connected to directly or indirectly control the process in accordance with said decision data.

30. The modular computer-based process control system of claim 29, wherein said controller comprises a feedback control module supervising a regulatory controller.

31. The modular computer-based process control system of claim 29, wherein said controller is connected to directly or indirectly control the process by changing a first process condition affecting the product property;

and further comprising

(d) a feedforward control module means, connected to directly or indirectly control the process by changing said first process condition or a second process condition affecting the product property.

32. The modular computer-based process control system of claim 29, wherein said controller is connected to directly or indirectly control the process by changing a first process condition affecting the product property;

and further comprising

(d) a statistical test module means, connected to provide statistical data for directly or indirectly controlling the process by changing said first condition or a second process condition affecting the product property.

33. The modular computer-based process control system of claim 29, further comprising a user interface providing a template for entering a size specification of said neural network module and/or a connectivity specification of said neural network module and/or a specification of a source of said input data, wherein said neural network module operates in accordance with said specification(s).

34. The modular computer-based process control system of claim 33, wherein said template comprises data pointers for specifying data to be used by said neural network module.

35. The modular computer-based process control system of claim 29, further comprising a user interface for configuring said neural network module using a limited set of natural language format specifications.

36. A computer-based process control method adapted to control a process using a controller, the process producing a product having at least one product property, the computer-based process control method comprising the steps of:

(1) predicting using a neural network output data from input data;

(2) executing in parallel the substeps of;

(a) at a first specified interval, controlling the process with the controller, using said output data; and

(b) at a second specified interval, using an expert system to make a decision, using data from the process, to change a parameter of the controller.