Multivariable feedforward adaptive controller

• US 5,541,833 A
• Filed: 07/28/1993
• Issued: 07/30/1996
• Est. Priority Date: 03/30/1987
• Status: Expired due to Term

• Current Assignee: Invensys Systems Incorporated (Invensys PLC)
• Original Assignee: The Foxboro Company (Schneider Electric SE)
• Inventors: Hansen, Peter D., Bristol, Edgar H.
• Primary Examiner(s): RUGGIERO, JOSEPH F
• Application Number:
  US08/097,916
• Time in Patent Office: 1,098 Days
• Field of Search: 364/148-151, 364/157-159, 364/164, 364/165, 364/153, 364/156, 364/177, 364/183, 318/561
• US Class Current: 700/45
• CPC Class Codes: G05B 11/42 for obtaining a characteristic which is both proportional and time-dependent, e.g. P. I., P. I. D.
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: G05B 13/042 in which a parameter or coefficient is automatically adjusted to optimise the performance

Abstract

Multivariable adaptive feedforward control may be accomplished by detecting the beginning and ending of a process control disturbance response, characterizing the measured inputs and process output during the disturbance by moments, which comprise time-weighted integrals performed on the process result output and inputs when the disturbance is a measured disturbance, and relating the characterized inputs and process result output in known general transfer function model equations to generate transfer function parameters which are used to calculate the coefficients of feedforward additive or multiplicative compensators.

Claims

1. A method of automatically adjusting at least one control parameter of a feedforward controller used in a control system coupled to regulate a process having inputs and a result output in a closed loop of said control system, and having a measured variable signal responsive to a measurable process disturbance response to a disturbance of said process, comprising the steps of:

a) detecting the beginning of the process disturbance response by sensing a significant change in either of the measured variable signal or the control parameter and then detecting the ending of a disturbance response;

b) determining whether the disturbance is an unmeasured disturbance response;

c) characterizing the inputs and process result output during a disturbance response by moments which comprise time-weighted integrals performed on the process result output and inputs;

d) selecting a model equation if the disturbance is a measured disturbance;

e) relating the characterized inputs and process result output in general transfer function model equations to generate transfer function parameters relating the inputs to the process result output when the disturbance is a measured disturbance; and

f) adapting transfer function model equations to relate the characterized inputs and process result output to generate transfer function parameters relating the inputs to the process result output when the disturbance is an unmeasured disturbance; and

g) adjusting said at least one control parameter.

2. The method of claim 1, wherein the step of detecting the end of a disturbance response includes peak detection.

3. The method of claim 2, wherein said peak detection includes the steps of generating an error signal corresponding to the closed-loop response of the process to an upset condition and characterizing the amplitude features of said error signal.

4. The method of claim 3, wherein said controller has a closed loop response, further including the step of determining the number of peaks previously confirmed in a significant measured response as a function of the closed loop response of the controller.

5. The method of claim 1, wherein the parameter values are rendered dimensionless by the steps of first converting the signals from engineering units to percent-of-range signals, and subsequently normalizing the percent-of-range signals by a scale factor T.

6. The method of claim 5, wherein scale factor T is equal to or greater than 20 percent of a closed loop characteristic time factor TF.

7. The method of claim 5, where scale factor T is equal to or less than 50 percent of a closed loop time factor TF.

8. The method of claim 1, where the successful adaptations of steps e) and f) are correlated to conditions at the start of the disturbance, the step of storing the adaptation information for subsequent recall.

9. The method of claim 8, further including the step of correlating the adaptations according to disturbance conditions and subsequently indexing the adaptations according to the correlated conditions.

10. The method of claim 1, wherein the step of detecting the beginning of a disturbance response comprises sensing a control parameter which exceeds a measured variable signal noise threshold.

11. The method of claim 10, further including the step of adaptively updating at least one measured variable signal noise threshold between disturbance responses.

12. The method of claim 1, where the feedforward controller includes means for receiving a set point signal, a primary measurement signal, and a secondary measurement signal, further including the step of approximating gain delay in the feedforward controller by an adaptively tuned low-pass filter.

13. The method of claim 12, wherein the low-pass filter includes a 0.7 damped quadratic.

14. The method of claim 1, where the feedforward controller includes an absolute compensator comprising means for receiving an absolute feedforward measurement load signal and for producing a dynamic feedforward compensation signal, further including the step of producing a dynamic feedforward compensation signal from the absolute feedforward measurement load signal.

15. The method of claim 1, wherein the feedforward controller includes a compensator having gain delay, further including the step of adjusting the gain delay.

16. The method of claim 1, wherein the moments are initialized at the start of a disturbance response.

17. The method of claim 1, wherein the sign index of a given moment is chosen according to a predicted direction of the manipulated-variable change needed to counteract the disturbance.

18. The method of claim 1, wherein in step e) the general transfer function model equation is:

##EQU2##

19. Multivariable adaptive feedforward control of a process by feedforward controller controlling at least one process loop in a control system regulating said process and having inputs and a process result output, in which the process is subjected to response transients due to process disturbances, wherein feedforward adaptation is accomplished by the method comprising the steps

a) detecting a beginning of the process disturbance response by sensing a significant change in either of a measured variable signal or a control parameter and then detecting the ending of the response disturbance response;

b) characterizing the inputs and process result output during a response transient by moments which comprise time-weighted integrals performed on the process result output and inputs; and

c) relating the characterized inputs and process result output in general transfer function model equations to generate model compensation parameters relating the inputs to the process result output,

d) generating a feedback control signal by

i) generating an error signal representing the difference between the set point value and the closed-loop response of the process to a process disturbance;

ii) identifying and measuring amplitude values characteristic of pattern features of said error signal; and

iii) automatically adjusting at least one process control parameter of a feedback controller to improve the difference between said measured characteristic and a target characteristic.

20. The method of claim 19, further including the steps of sensing a conflict between the feedforward control and the feedback control signal, and detuning the feedforward controller such that rejection of unmeasured-load disturbances by the feedback controller is not compromised.

21. The method of claim 19, wherein in further including step e, computing new compensation parameters from past model compensation parameters.

22. The method of claim 19, wherein said feedforward control is related to a signal value and said feedback control signal and said feedforward control signal are used to control the same process loop, introducing incremental correction first.

23. The method of claim 19, wherein the step of detecting the beginning of a disturbance response comprises sensing a control parameter which exceeds a measured variable signal noise threshold.

24. The method of claim 23, further including the step of adaptively updating at least one measured variable signal noise threshold between disturbance responses.

25. The method of claim 19, wherein the feedforward controller includes means for receiving a set point signal, a primary measurement signal, and a secondary measurement signal, further including the step of approximating gain delay in the feedforward controller by an adaptively tuned low-pass filter.

26. The method of claim 25, wherein the low-pass filter includes a 0.7 damped quadratic.

27. The method of claim 19, wherein the feedforward controller includes an absolute compensator comprising means for receiving an absolute feedforward measurement load signal and for producing a dynamic feedforward compensation signal, further including the step of producing a dynamic feedforward compensation signal from the absolute feedforward measurement load signal.

28. The method of claim 19, wherein the feedforward controller includes a compensator having gain delay, further including the step of adjusting the gain delay.

29. The method of claim 19, wherein the moments are initialized at the start of a disturbance response.

30. The method of claim 19, wherein a given moment sign index is chosen according to a predicted direction of a manipulated-variable change needed to counteract the disturbance.

31. The method of claim 21, wherein the past model compensation parameters are zeroed before the new compensation parameters are computed.

32. The method of claim 19 wherein in step c) the general transfer function model equation is:

##EQU3##

33. An adaptive feedback controller for use with a process having multiple variables, comprising:

a) means for manipulating one of said process variables;

b) a primary controller comprising means for (i) receiving a set point signal, a primary measurement signal, a differentiated feedforward signal, and an integral feedback signal and for (ii) producing a primary controller output signal;

c) a secondary controller comprising means for (i) receiving a secondary control set point signal and a secondary measurement signal, and (ii) outputting a secondary control signal for controlling said manipulating means; and

d) a function block comprising means for (i) receiving an absolute load signal, a dynamic absolute compensation signal, the primary controller output signal, the secondary measurement signal, and (ii) for producing the secondary control set point signal and the integral feedback signal.

34. The controller of claim 33, further including an absolute compensator comprising means for receiving an absolute feedforward measurement load signal and for producing a dynamic feedforward compensation signal.

35. The controller of claim 33, wherein the compensator further includes a gain delay.

36. The controller of claim 35, further including means for receiving a set point signal, a primary measurement signal, and a secondary measurement signal, and wherein the gain delay is approximated by an adaptively tuned low-pass filter.

37. The controller of claim 36, wherein the filter includes a 0.7 damped quadratic.

38. The controller of claim 34, having moment and model coefficient calculation time scales, and a closed-loop characteristic time TF used to determine an expected end of an isolated response, to set filter parameters for each of one or more identifier inputs, and to normalize the time scale for moment and model coefficient calculations, and TF is the coefficient of the first order term in the closed-loop characteristic equation:

##EQU4## where D is the derivative time, I is the integral time, P is the proportional band of the feedback controller, and N is the index for the primary measurement signal.

39. The controller of claim 33, further including an incremental adapter comprising means for receiving a set point signal, a primary measurement signal, an integral feedback signal, and incremental loads, and for producing an incremental feedforward compensation signal.

40. The controller of claim 33, further including an incremental adapter comprising means for receiving a set point signal, a primary measurement signal, a controller output signal, and incremental loads, and for producing an incremental feedforward compensation signal.

41. The controller of claim 39, further including means associated with each load for load signal filtering, compensating, adding the result to the sum, then integrating the sum in a feedback controller to effectively add the accumulated total incremental feedforward compensation to the feedback controller output.

42. The controller of claim 40, further including means associated with each load for load signal filtering, compensating, adding the result to the sum, then integrating the sum in a feedback controller to effectively add the accumulated total incremental feedforward compensation to the feedback controller output.

43. The controller of claim 39, wherein the compensator further includes a gain delay.

44. The controller of claim 41, further including means for updating the gain delay terms while the process is in steady-state, without bumping the process.

45. The controller of claim 44, wherein the gain delay is approximated by an adaptively tuned low-pass filter.

46. The controller of claim 45, wherein the filter includes a 0.7 damped quadratic.

47. The controller of claim 41, having moment and coefficient calculation time scales, and a closed-loop characteristic time TF used to determine an expected end of an isolated response, to set filter parameters for each of one or more identifier inputs, and to normalize the time scale for moment and model coefficient calculations, and TF is the coefficient of the first order term in the closed-loop characteristic equation:

##EQU5## where D is the derivative time, I is the integral time, P is the proportional band, and N is the index for the primary measurement signal.

48. The controller of claim 35, further including means for adding the absolute load signal and the dynamic absolute compensation signal.

49. The controller of claim 48, wherein the primary controller and the absolute compensator have an effective gain which is substantially proportional to an absolute feedforward signal produced from addition of an absolute feedforward measured load signal and the feedforward compensation signal.

50. The controller of claim 39, further including means for zeroing the incremental feedforward compensation subsequent to each accumulation before new compensation parameters are computed.

51. The method of claim 40, further including means for zeroing the incremental feedforward compensation subsequent to each accumulation before new compensation parameters are computed.