Multi-input multi-output generic non-interacting controller

• US 5,920,478 A
• Filed: 06/27/1997
• Issued: 07/06/1999
• Est. Priority Date: 06/27/1997
• Status: Expired due to Term

• Current Assignee: Granite Creek Systems, Inc., Okaleaf Engineering, Inc.
• Original Assignee: Oakleaf Engineering, Inc.
• Inventors: Elia, Curtis F., Ekblad, Mark K.
• Primary Examiner(s): Grant, William
• Assistant Examiner(s): Baumgardner, Carolyn T.
• Application Number:
  US08/884,254
• Time in Patent Office: 739 Days
• Field of Search: 364/148, 364/149, 364/724.011, 364/724.06, 364/151, 318/139, 455/179
• US Class Current: 700/29
• CPC Class Codes: G05B 11/32 with inputs from more than one sensing element; with outputs to more than one correcting element
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

A multi-input multi-output (MIMO) controller is configured as a generic, non-interacting, multivariable controller with adaptive capabilities. The controller transforms the coupled multivariable control problem to multiple decoupled single-input single-output loops. The decoupling occurs in the input/output map of the system being controlled. The transformed system can then be controlled using any control synthesis technique to meet the system performance objectives, such as Proportional plus Integral plus Derivative (PID) control. An empirical dynamic model is continually identified from on-line data. From this model, static decoupling transformations based on a singular value decomposition are determined and updated and the controller gains are calculated.

Claims

1. A controller for controlling a plurality of controllable inputs to a dynamic process where the plurality of controllable inputs affect a plurality of process outputs of the dynamic process and control of the plurality of process outputs is achieved by manipulating one or more of the plurality of controllable inputs to the dynamic process, the controller comprising:

a parameter calculator which calculates a plurality of decoupling parameters and at least one recoupling parameter based on at least one process output of the plurality of process outputs or at least one controllable input of the plurality of controllable inputs;

an output decoupler, adapted to receive at least two process outputs of the plurality of process outputs and adapted to generate decoupled output signals based on the at least two process outputs and the plurality of decoupling parameters;

a tunable controller, adapted to receive the decoupled output signals, the tunable controller comprising a plurality of tuners, each of which tunes a control system formed between one of the decoupled output signals and a decoupled controllable input signal output by the tunable controller which is controlled according to a set of one or more control targets; and

an input recoupler, adapted to receive the decoupled controllable input signals from the tunable controller and the plurality of recoupling parameters from the parameter calculator, wherein the input recoupler is further adapted to transform the decoupled controllable input signals according to the at least one recoupling parameter into the one or more of the plurality of controllable inputs being manipulated.

2. The controller of claim 1, wherein the decoupling parameters are an indication of a dynamic process model.

3. The controller of claim 1, further comprising a control design engine which adapts to process variations through successive identification tests or which schedules the decoupling parameters and controller gains as a function of at least one of the plurality of process outputs.

4. The controller of claim 1, further comprising a control design engine, wherein the decoupling parameters include transformations which transform a coupled multi-variable control loop into a plurality of decoupled single-input single-output loops.

5. The controller of claim 1, wherein the plurality of decoupling parameters are decoupling parameters obtained through a Singular Value Decomposition on DC gain values determined from an identified dynamic process model.

6. The controller of claim 1, wherein said tunable controller further comprises means for individual channel tuning of each of the plurality of tuners.

7. The controller of claim 1, wherein said tunable controller further comprises means for ganged channel tuning of each of the plurality of tuners.

8. The controller of claim 1, wherein said tunable controller further comprises means for tuning based on a single scalar entity.

9. A controller as set forth in claim 8, wherein the tuning based on a single scalar entity uses Singular Value Decomposition on DC gain values as well as a plurality of process time constants obtained from an identified dynamic process model.

10. A controller as set forth in claim 1, wherein the controller is a controller which does not constrain its control synthesis to a particular control synthesis technique.

11. A controller as set forth in claim 1, wherein the tunable controller controls the at least two process outputs as decoupled output signals and controls at least two others of the plurality of process outputs as coupled output signals.

12. A controller as set forth in claim 1, wherein the tunable controller controls each of the tuners is a Proportional plus Integral (PI) controller, a Proportional plus Derivative (PD) controller or a Proportional plus Integral plus Derivative (PID) controller.

13. A controller for controlling a plurality of controllable inputs to a dynamic process where the plurality of controllable inputs affect a plurality of process outputs of the dynamic process and control of the plurality of process outputs is achieved by manipulating one or more of the plurality of controllable inputs to the dynamic process, the controller comprising:

a parameter calculator which calculates a plurality of recoupling parameters and is adapted to receive at least one process output of the plurality of process outputs or at least one controllable input of the plurality of controllable inputs;

a tunable controller, adapted to receive at least two process outputs of the plurality of process outputs, the tunable controller comprising a plurality of tuners, each of which tunes a control system formed between one process output from one of the plurality of process outputs and a decoupled controllable input signal output by the tunable controller which is controlled according to a set of one or more control targets; and

an input recoupler, adapted to receive the decoupled controllable input signals from the tunable controller and the plurality of recoupling parameters from the parameter calculator, wherein the input recoupler is adapted to transform the decoupled controllable input signals according to the plurality of recoupling parameters into the one or more of the plurality of controllable inputs being manipulated.

14. A method of decoupling a plurality of process outputs of a dynamic process from a plurality of process inputs to the dynamic process, comprising the steps of:

measuring responses of the dynamic process to test signals to determine a dynamic process model for the dynamic process;

generating an output transformation matrix, a tuning matrix and an input transformation matrix from the dynamic process model;

transforming signals from the plurality of process outputs through the output transformation matrix to form a plurality of decoupled process output signals;

transforming the plurality of decoupled process output signals according to the tuning matrix and independent controls of at least one of the plurality of decoupled process output signals to form a plurality of decoupled process input signals;

transforming the plurality of decoupled process input signals through the input transformation matrix to form a plurality of recoupled process input signals; and

applying the plurality of recoupled process input signals to the process inputs of the dynamic process;

wherein the output transformation matrix, tuning matrix and input transformation matrix are such that the input/output map of the plurality of decoupled process input signals to the plurality of decoupled process output signals is decoupled relative to the dynamic process, thereby converting the dynamic process from a multi-variable control system into a plurality of single-input single-output control systems.

15. The method of claim 14, wherein the output transformation matrix is an identity matrix.

16. The method of claim 14, wherein the step of measuring comprises the steps of:

applying an input step signal to each of the process inputs;

detecting an effect of the input step signals;

applying an output step signal for each of the process outputs, prior to the step of transforming according to the tuning matrix;

detecting an effect of the output step signals; and

generating a dynamic process model based on the effects detected in the steps of detecting.

17. The method of claim 14, wherein the dynamic process model is a function of at least one of a gain, a time constant, a delay, a variable gain, a variable time constant, and a variable delay.

18. The method of claim 14, wherein the dynamic process model is a model of a nonlinear dynamic process.