Method for developing a neural network tool for process identification

• US 5,924,086 A
• Filed: 03/13/1997
• Issued: 07/13/1999
• Est. Priority Date: 10/10/1990
• Status: Expired due to Fees

• Current Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Original Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Inventors: Samad, Tariq, Mathur, Anoop
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/816,470
• Time in Patent Office: 852 Days
• Field of Search: 395/23, 395/22, 395/902, 395/903, 395/906, 706/23, 706/25, 706/52, 364/148.03, 364/153, 364/148.02
• US Class Current: 706/25
• CPC Class Codes: G05B 17/02 electric
• CPC Class Codes: Y10S 706/902 Application using ai with detail of the ai system
• CPC Class Codes: Y10S 706/903 Control
• CPC Class Codes: Y10S 706/906 Process plant

Abstract

A tool, and the method of making the tool, for process system identification that is based on the general purpose learning capabilities of neural networks. The tool and method can be used for a wide variety of system identification problems with little or no analytic effort. A neural network is trained using a process model to approximate a function which relates process input and output data to process parameter values. Once trained, the network can be used as a system identification tool. In principle, this approach can be used for linear or nonlinear processes, for open or closed loop identification, and for identifying any or all process parameters.

Claims

1. A method for developing a tool for identifying at least one parameter of a process--wherein said parameter is characterized as a constant in an equation defining a mathematical model of said process--;

said tool capable of providing a signal representative of values of said at least one parameter which is recognized by a controller of said process, said method comprising the steps of;

determining ranges for each of said at least one parameter;

modeling said equation via a computer program;

utilizing said program to generate a set of training examples, each of said examples having (1) selected values of said at least one parameter from within said respective ranges and (2) process model output data resulting from said program when a process model input is provided to said process model for said selected values of said at least one parameter; and

training an artificial neural network on said set of training examples by;

(1) providing an input to said network comprising for each of said training examples respective values of at least said process model output data; and

(2) the output of said network comprises respectively for each of said training examples estimates of at least one of said selected values of said at least one parameter, said signal being in a form recognizable by said controller as an indicator of a present value of a process model parameter so that said controller effectuates changes to the process based on such input.

2. The method according to claim 1 wherein said equation is a linear differential equation.

3. The method according to claim 1 wherein said equation is a nonlinear differential equation.

4. The method according to claim 1 wherein said equation is an algebraic polynomial equation with said parameters being represented as coefficients of said polynomial equation.

5. The method according to claim 1 wherein the input to said network further comprises said process model input.

6. The method according to claim 5 wherein said process model input includes varying values of said process model input data.

7. The method according to claim 1 wherein said equation has the form:

##EQU6## wherein x(t) is the process model output response, t_p is a time constant parameter of the equation, K_p is a gain parameter of the equation and θ

is a delay parameter of the equation.

8. A neural network tool developed by a method for developing a tool for identifying at least one parameter of a process--wherein said parameter is characterized as a constant in an equation defining a mathematical model of said process--;

said tool capable of providing a signal representative of values of said at least one parameter which may be recognized by a controller of said process;

comprising the steps;

determining ranges for each of said at least one parameter;

modeling said equation via a computer program;

utilizing said program to generate a set of training examples, each of said examples having (1) selected values of said at least one parameter from within said respective ranges and (2) process model output data resulting from said program when a process model input is provided to said process model for said selected values of said at least one parameter; and

training an artificial neural network on said set of training examples by;

(1) providing an input to said network comprising for each of said training examples respective values of at least said process model output data; and

(2) the output of said network comprises respectively for each of said training examples estimates of at least one of said selected values of said at least one parameter, said signal being in a form recognizable by said controller as an indicator of a present value of a process model parameter so that said controller may effectuate changes to the process based on such input.

9. The neural network tool developed by the method according to claim 8 wherein said equation is a linear differential equation.

10. The neural network tool developed by the method according to claim 8 wherein said equation is a nonlinear differential equation.

11. The neural network tool developed by the method according to claim 8 wherein said equation is an algebraic polynomial equation with said parameters being represented as coefficients of said polynomial equation.

12. The neural network tool developed by the method according to claim 8 wherein the input to said network further comprises said process model input.

13. The neural network tool developed by the method according to claim 12 wherein said process model input includes varying values of said process model input data.

14. The neural network tool developed by the method according to claim 8 wherein said equation has the form:

##EQU7## wherein x(t) is the process model output response, t_p is a time constant parameter of the equation, K_p is a gain parameter of the equation and θ

is a delay parameter of the equation.