Nonlinear-approximator-based automatic tuner

• US 5,847,952 A
• Filed: 06/28/1996
• Issued: 12/08/1998
• Est. Priority Date: 06/28/1996
• Status: Expired due to Term

• Current Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Original Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Inventors: Samad, Tariq
• Primary Examiner(s): Lee, Thomas C.
• Assistant Examiner(s): Marc, McDieunel, Marc, McDieunel
• Application Number:
  US08/671,996
• Time in Patent Office: 893 Days
• Field of Search: 364/148, 364/157, 364/158, 364/165, 348/4
• US Class Current: 700/48
• CPC Class Codes: G05B 13/027 using neural networks only

Abstract

An automatic tuner for control systems that produces, as output values, parameters of an arbitrary controller. The controller is in a control loop so that its output effects changes in actuators and regulates a physical process. The controller has either linear or nonlinear controller components, or a combination of both. The tuner has a nonlinear approximator that has been optimized off-line. The off-line optimization is done without supervised learning so that desired outputs of the nonlinear approximator do not need to be available, and separate optimization to generate the desired outputs is not necessary. The off-line optimization can also rely on arbitrary criteria. Such optimization ensures robustness of generated controller parameters so that the input process characteristics do not need to be highly accurate. The inputs to the nonlinear approximator consist of two sets of input parameters, either of which may be empty. A first set of input parameters can relate to process characteristics. A second set of input parameters can relate to desired closed-loop system behavior. The output values may be proportional and/or integral and/or derivative gains for PID-like controllers, or otherwise be parameters for delay-compensation controllers, controllers that consist of lead-lag terms in combination with PID controllers, higher-order linear controllers, or nonlinear controllers of predetermined structure. The nonlinear approximator may be implemented as a compositional sigmoidal mapping, a multilayer perception structure, a fuzzy logic model, a radial basis function network, a polynomial expansion, or other parametrized nonlinear structure.

Claims

1. A tuner for producing internal gain parameters of a controller, comprising:

preprocessing means, having an input and an output, for processing a set of parameters at the input into a set of transformed parameters at the output;

nonlinear approximator means, having an input connected to the output of said preprocessing means and having an output, for operating on the set of transformed parameters to produce a set of normalized controller tuning parameters at the output wherein said nonlinear approximator means is offline optimized in model based simulation without advance supervised learning; and

postprocessing means, having a first input connected to the output of said nonlinear approximator means, and having a second input and an output, for scaling the set of normalized controller tuning parameters into a set of scaled controller tuning parameters at the output, wherein the scaled controller tuning parameters are fed to the controller for optimally controlling a process.

2. The tuner of claim 1 wherein said nonlinear approximator means is a neural network.

3. The tuner of claim 1 wherein said nonlinear approximator means is a fuzzy approximator.

4. The tuner of claim 1 wherein said nonlinear approximator means is a radial basis function approximator.

5. The tuner of claim 1 wherein said nonlinear approximator means is an approximator that utilizes nontranscendental mathematical functions.

6. The tuner of claim 1 wherein said nonlinear approximator means is an approximator that utilizes computationally simple compositional sigmoidal mappings.

7. A tuner for producing internal gain parameters of a controller, comprising:

preprocessing means, having an input and an output, for transforming a set of input parameters at the input into a set of normalized parameters at the output,

nonlinear approximator means, having an input connected to the output of said preprocessing means and having an output, for algorithmically operating on the set of normalized parameters in that said nonlinear approximator means is setup offline using a system that integrates closed-loop simulations with an optimization algorithm that optimizes parameters of said nonlinear approximator means to minimize a cost function that indicates performance of the closed-loop simulations to result in a set of normalized tuning parameters at the output; and

postprocessing means, having a first input connected to the output of said nonlinear approximator means, having a second input connected to the input of the preprocessing means and having an output, for scaling the set of normalized tuning parameters into controller tuning parameters at the output.

8. The tuner of claim 7 wherein:

the input parameters are either or both of two types of parameters, wherein the two types of parameters are;

parameters characterizing a system to be controlled; and

parameters characterizing a desired closed-loop behavior; and

said nonlinear approximator means is optimized offline without supervised learning and wherein the offline optimization does not require generation of optimized tuning parameters for optimizing said nonlinear approximator means.

9. The tuner of claim 8 wherein said nonlinear approximator means is a neural network.

10. The tuner of claim 8 wherein said nonlinear approximator means is a fuzzy approximator.

11. The tuner of claim 8 wherein said nonlinear approximator means is a radial basis function approximator.

12. The tuner of claim 8 wherein said nonlinear approximator means is an approximator that utilizes nontranscendental mathematical functions.

13. The tuner of claim 8 wherein said nonlinear approximator means is an approximator that utilizes computationally simple compositional sigmoidal mappings.

14. A system for setting up a tuner comprising:

a tuner having inputs;

approximator parameters;

controller parameters;

a closed-loop control simulator;

a cost function; and

an optimization algorithm; and

wherein;

the tuner includes a nonlinear approximator, the behavior of which can be modified by adjusting the approximator parameters;

the tuner outputs controller parameters which are used in simulations with the closed loop control simulator;

a value of the cost function is computed from closed loop control simulations;

the value of the cost function is input to the optimization algorithm which optimizes approximator parameters;

optimized approximator parameters minimize the value of the cost function;

a minimized value of the cost function leads to appropriate controller parameters; and

appropriate controller parameters result in optimized controller performance.

15. The system of claim 14 wherein the nonlinear approximator is a neural network.

16. The system of claim 14 wherein the nonlinear approximator is a fuzzy approximator.

17. The system of claim 14 wherein the nonlinear approximator is a radial basis function approximator.

18. The system of claim 14 wherein the nonlinear approximator is an approximator that utilizes nontranscendental mathematical functions.

19. The tuner of claim 14 wherein the nonlinear approximator is an approximator that utilizes computationally simple compositional sigmoidal mappings.

20. A tuner for producing internal gain parameters of a controller, comprising:

preprocessing means, having an input and an output, for processing a set of input tuning parameters that indicate desired or appropriate performance characteristics at the input into a set of transformed parameters at the output;

nonlinear approximator means, having an input connected to the output of said preprocessing means and having an output, for operating on the set of transformed parameters to produce a set of normalized tuning parameters at the output; and

postprocessing means, having a first input connected to the output of said nonlinear approximator means, and having a second input and an output, for scaling the set of normalized tuning parameters into a set of scaled tuning parameters at the output, wherein the scaled tuning parameters are fed to the controller for optimally controlling a process.

21. The tuner of claim 20 wherein said nonlinear approximator means is a neural network.

22. The tuner of claim 20 wherein said nonlinear approximator means is a fuzzy approximator.

23. The tuner of claim 20 wherein said nonlinear approximator means is a radial basis function approximator.

24. The tuner of claim 20 wherein said nonlinear approximator means is an approximator that utilizes nontranscendental mathematical functions.

25. The tuner of claim 20 wherein said nonlinear approximator means is an approximator that utilizes computationally simple compositional sigmoidal mappings.

26. The tuner of claim 20 further comprising a set of input tuning parameters that indicate characteristics of open loop dynamics of a system to be controlled.

27. The tuner for producing actual values of internal gain parameters of a controller that is to be used for controlling a process, comprising:

preprocessing means, having an input and an output, for processing a set of parameters at the input into a set of transformed parameters at the output;

nonlinear approximator means, having an input connected to the output of said preprocessing means and having an output, for operating on said set of transformed parameters to produce normalized actual values of internal gain parameters of said controller; and

postprocessing means having a first input connected to the output of said nonlinear approximator means, and having a second input and an output, for scaling said normalized actual values of internal gain parameters into said actual values of internal gain parameters of said controller at the output, wherein said actual values of internal gain parameters are input to the controller for optimally controlling said process.