Method and device for tuning and control

• US 7,937,165 B2
• Filed: 12/18/2006
• Issued: 05/03/2011
• Est. Priority Date: 12/30/2005
• Status: Active Grant

• Current Assignee: ABB Schweiz AG (ABB Limited)
• Original Assignee: ABB AB (ABB Limited)
• Inventors: Bergsten, Pontus, Modén, Per-Erik, Lindkvist, Rickard
• Primary Examiner(s): Decady; Albert
• Assistant Examiner(s): STEVENS, THOMAS H
• Application Number:
  US12/159,488
• Publication Number:
  US 20090138102A1
• Time in Patent Office: 1,597 Days
• Field of Search: 700/39, 700/305, 700/30, 700/282, 324/307, 409/147, 409/152, 409/153, 409/159, 409/193
• US Class Current: 700/39
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: Y10T 409/304648 with control means energized in response to activator stimulated by condition sensor
• CPC Class Codes: Y10T 409/304928 Adapted to hydraulically or pneumatically stimulate control
• CPC Class Codes: Y10T 409/304984 Adapted to electrically stimulate control
• CPC Class Codes: Y10T 409/30532 with means to advance work or product
• CPC Class Codes: Y10T 409/307224 with infeed control means energized in response to activator stimulated by condition sensor

Abstract

A method and a device for tuning and control of industrial processes having varying material flow rate. An adder is configured to add excitation signals to the controller output signal. A measurement system is configured to measure a property in response to the excitation signals. A model based tuning unit is adapted to estimate the value of at least one parameter with unknown value of a process model structure describing the effect of varying material flow rate, based on the measurements of the property and the output signal from the controller, and to calculate a model that describes the dynamics from controller output to controller input based on the estimated value of the parameter, and to perform model based tuning of the controller based on the model that describes the dynamics from controller output to controller input.

Claims

1. A method for tuning a controller controlling a property of an industrial process having varying material flow rate, the method comprising:

injecting excitation signals added to an output signal of the controller,

receiving measurements of said property in response to said excitation signals,

choosing a process model structure which is time continuous and comprises at least one parameter with unknown value, the parameter being independent of varying material flow rate,

estimating a value of said at least one parameter, based on said measurements of said property and an output signal from the controller,

calculating a model that describes dynamics from controller output to controller input based on the estimated value of said parameter, and on a basis thereof performing model based tuning of the controller,

reducing a parameter update size, when an unrestricted update reaches outside an area of defined allowed parameter ranges, but making a parameter update in a same direction in a parameter space, and

applying a same reduction factor to both regressors and prediction error, to be used in both parameter update and update of matrix holding information necessary in recursive estimation,

thereby keeping parameter estimates within allowed ranges.

2. The method according to claim 1, further comprising:

choosing a process model structure comprising at most one model parameter per input/output pair being affected by varying actual gain of the industrial process.

3. The method according to claim 1, further comprising:

choosing a process model structure comprising at most one model parameter per actuator being affected by varying actual response times of actuators.

4. The method according to claim 1, further comprising:

including, in the process model structure, any delays due to data processing and communication, and

choosing said process model structure such that variations of said delays do not influence more than one model parameter per model input.

5. The method according to claim 1, further comprising:

composing said process model structure of two parts, one part describing transport behavior where possible deadtimes and possible time constants are inversely proportional to the material flow rate, and another part that is independent of the material flow rate.

6. The method according to claim 1, further comprising:

deriving a first order plus deadtime model for each part of said model,

introducing a transport behavior parameter that defines a degree of pure deadtime in relation to pure first order response of the transport behavior,

having, in a case of pure deadtime transport behavior, a deadtime described by a known distance,

having, in the case of pure first order transport behavior, an associated time constant described by the known distance, and

estimating at least one of the parameters gain, flow independent time constant, flow independent deadtime and transport behavior parameter.

7. The method according to claim 1, further comprising:

estimating at least one parameter using a prediction error method.

8. The method according to claim 7, further comprising:

choosing the process model structure

9. The method according to claim 1, further comprising:

reducing, for a case of large prediction errors, an effect of measurements and regressors according to a belief factor calculated from a ratio between prediction error size and estimated standard deviation of previous prediction errors, and

thereby achieving protection against outliers.

10. The method according to claim 1, further comprising:

sampling process outputs at a basic sampling rate, applying a filter with the same basic sampling rate, and re-sampling to a slower sampling rate, this rate being used for logging of values from operation of the industrial process, and including an effect of this sampling, filtering and re-sampling in prediction calculations and parameter estimate updates.

11. The method according to claim 1, further comprising:

performing an identification experiment meaning operating the industrial process with intentional excitation influencing at least one manipulated variable, and

estimating at least one parameter on-line during said identification experiment.

12. The method according to claim 11, further comprising:

presenting results to a user in real-time, thereby allowing user decision to end the identification experiment.

13. The method according to claim 11, further comprising:

estimating fewer parameters during operation with on-line updating of the controller than during the identification experiment, and

allowing an parameter estimation result from the identification experiment to define the value of at least one parameter, and treating this parameter or these parameters as known during the operation with on-line updating of the controller.

14. The method according to claim 1, further comprising:

calculating on-line said model based tuning of the controller.

15. The method according to claim 1, further comprising:

estimating at least one parameter on-line during operation of the industrial process and while controlling it with said controller,

performing on-line said model based tuning calculations for the controller, and

applying the tuning result to said controller on-line,

thereby achieving true adaptive control.

16. The method according to claim 1, further comprising:

using lambda tuning as model based tuning method for said controller.

17. A device for tuning a controller controlling a property of an industrial process having varying material flow rate, the device comprising:

an adder for adding excitation signals to a controller output signal,

a measurement system for measuring said property in response to said excitation signals, and

a model based tuning unit adapted to estimate a value of at least one parameter with unknown value of a time continuous process model structure describing an effect of varying material flow rate, based on said measurements of said property and the controller output signal, the parameter being independent of the varying material flow rate, and to calculate a model that describes dynamics from controller output to controller input based on the estimated value of said parameter, to reduce a parameter update size, when an unrestricted update reaches outside an area of defined allowed parameter ranges, but making a parameter update in the same direction in a parameter space, to apply a same reduction factor to both regressors and prediction error, to be used in both parameter update and update of matrix holding information necessary in recursive estimation, thereby keeping parameter estimates within allowed ranges, and to perform model based tuning of the controller based on said model that describes the dynamics from controller output to controller input.

18. A non-transitory computer readable medium storing computer readable program code for causing a computer to perform the steps of:

injecting excitation signals added to an output signal of a controller,

receiving measurements of said property in response to said excitation signals,

choosing a process model structure comprising at least one parameter with unknown value, this model structure describing an effect of varying material flow rate,

estimating a value of said at least one parameter, based on said measurements of said property and an output signal from the controller,

calculating a model that describes dynamics from controller output to controller input based on the estimated value of said parameter, and on a basis thereof performing model based tuning of the controller,

reducing a parameter update size, when an unrestricted update reaches outside an area of defined allowed parameter ranges, but making a parameter update in the same direction in a parameter space, and

applying a same reduction factor to both regressors and prediction error, to be used in both parameter update and update of matrix holding information necessary in recursive estimation,

thereby keeping parameter estimates within allowed ranges.