Robust process model identification in model based control techniques
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G05B-013/02
G05B-011/01
G06F-019/00
G06F-011/30
G06F-007/60
G06F-017/10
G21C-017/00
H03F-001/26
H04B-015/00
출원번호
UP-0403361
(2006-04-13)
등록번호
US-7840287
(2011-01-22)
발명자
/ 주소
Wojsznis, Wilhelm K.
Mehta, Ashish
Thiele, Dirk
출원인 / 주소
Fisher-Rosemount Systems, Inc.
대리인 / 주소
Marshall, Gerstein & Borun LLP
인용정보
피인용 횟수 :
18인용 특허 :
31
초록▼
A robust method of creating process models for use in controller generation, such as in MPC controller generation, adds noise to the process data collected and used in the model generation process. In particular, a robust method of creating a parametric process model first collects process outputs b
A robust method of creating process models for use in controller generation, such as in MPC controller generation, adds noise to the process data collected and used in the model generation process. In particular, a robust method of creating a parametric process model first collects process outputs based on known test input signals or sequences, adds random noise to the collected process data and then uses a standard or known technique to determine a process model from the collected process data. Unlike existing techniques for noise removal that focus on clean up of non-random noise prior to generating a process model, the addition of random, zero-mean noise to the process data enables, in many cases, the generation of an acceptable parametric process model in situations where no process model parameter convergence was otherwise obtained. Additionally, process models created using this technique generally have wider confidence intervals, therefore providing a model that works adequately in many process situations without needing to manually or graphically change the model.
대표청구항▼
What is claimed is: 1. A method of generating a process model for a process implemented on one or more physical devices, comprising: collecting process data indicative of process operation from the process implemented on the one or more physical devices; after collecting the process data and prior
What is claimed is: 1. A method of generating a process model for a process implemented on one or more physical devices, comprising: collecting process data indicative of process operation from the process implemented on the one or more physical devices; after collecting the process data and prior to determining the process model for the process implemented on the one or more physical devices, adding noise to the process data to produce conditioned process data; determining the process model for the process implemented on the one or more physical devices from the conditioned process data including the added noise; and generating, based on the determined process model, parameters used to control the process. 2. The method of claim 1, wherein collecting process data includes upsetting the process using a known process upset signal and collecting process data indicative of a process response to the process upset signal. 3. The method of claim 2, wherein adding noise to the process data comprises adding noise to a process output signal which is indicative of the process response to the process upset signal. 4. The method of claim 2, further comprising adding noise to the process upset signal prior to upsetting the process using the process upset signal to thereby add noise to data being collected as the collected process data. 5. The method of claim 1, wherein determining the process model includes determining one or more parameters of a parametric process model. 6. The method of claim 5, wherein determining the process model includes determining one or more parameters of an auto-regressive with external inputs process model. 7. The method of claim 1, wherein determining the process model includes determining a non-parametric process model. 8. The method of claim 7, wherein determining the process model includes determining a finite impulse response process model. 9. The method of claim 1, wherein adding noise to the process data to produce the conditioned process data includes adding random noise to the process data. 10. The method of claim 9, wherein adding random noise to the process data includes adding zero-mean noise to the process data. 11. The method of claim 10, wherein adding random noise to the process data includes adding evenly distributed random noise with a maximum amplitude between 0.2 and 0.5 percent of the range of the collected process data. 12. The method of claim 10, wherein adding random noise to the process data includes adding evenly distributed random noise with a maximum amplitude of approximately 0.4 percent of the range of the collected process data. 13. The method of claim 1, wherein determining the process model from the conditioned process data includes estimating a process dead time for the process from the collected process data, and using the dead time and the conditioned process data to determine the process model. 14. The method of claim 13, wherein estimating the process dead time includes generating a further process model from the collected process data and determining an estimate of the process dead time from the further process model. 15. The method of claim 14, wherein generating the further process model includes generating a finite impulse response model. 16. The method of claim 1, further including screening the collected process data or the conditioned process data prior to generating the process model from the conditioned process data. 17. The method of claim 1, wherein adding noise to the process data includes determining an amplitude of the noise as a function of the collected process data. 18. The method of claim 17, wherein determining an amplitude of the noise includes determining an amplitude of the noise as a function of the range of the collected process data, a mean of the collected process data or a standard deviation of the collected process data. 19. The method of claim 1, wherein adding noise to the process data includes determining an amplitude of the noise as a function of a process input signal used to generate the process data indicative of process operation. 20. A model generation system for generating a process model from a process implemented on one or more physical devices in a process control environment including one or more processors and a computer readable memory, the model generation system comprising: a first routine stored on the computer readable memory and executable on at least one of the one or more of the processors to collect from the process implemented on the one or more physical devices in the process control environment, process data indicative of process operation for at least a portion of the process; a second routine stored in the computer readable memory and executable on at least one of the one or more of the processors to, after collecting the process data and prior to generating the process model, add noise to the process data to produce conditioned process data; and a model generation routine stored in the computer readable memory and executable on at least one of the one or more of the processors to determine the process model for the process implemented on the one or more physical devices from the conditioned process data including the added noise. 21. The model generation system of claim 20, wherein the model generation routine is a parametric model generation routine that is executable to generate a parametric model by determining one or more parametric model parameters from the conditioned process data. 22. The model generation system of claim 21, wherein the model generation routine is an auto-regressive with external inputs process model generation routine. 23. The model generation system of claim 21, wherein the model generation routine includes a process parameter routine that is executable to estimate a dead time of the process and a model parameter estimation routine that is executable to determine the one or more parametric model parameters from the conditioned process data and the estimate of the process dead time. 24. The model generation system of claim 23, wherein the process parameter routine is executable to produce anon-parametric model for the process and determines the process dead time from the non-parametric model. 25. The model generation system of claim 24, wherein the process parameter routine is executable to produce a finite impulse response model as the non-parametric model. 26. The model generation system of claim 20, further including a third routine stored on a computer readable memory and executable on at least one of the one or more of the processors to produce a process controller using the process model. 27. The model generation system of claim 26, wherein the process controller is a model predictive control based controller. 28. The model generation system of claim 20, wherein the first routine includes a signal generator routine that is executable to produce a known process upset signal to upset the process and a collection routine that is executable to collect process data indicative of a process response to the process upset signal. 29. The model generation system of claim 28, wherein the first routine is further executable to add noise to the process upset signal so that the collected process data indicative of the process response is at least a portion of the conditioned process data. 30. The model generation system of claim 29, wherein the second routine is executable to add zero-mean, random noise to the process upset signal. 31. The model generation system of claim 30, wherein the second routine is executable to enable a user to select a magnitude associated with the zero-mean, random noise to be added to the process upset signal. 32. The model generation system of claim 20, wherein the second routine is executable to add zero-mean, random noise to the process data. 33. The model generation system of claim 32, wherein the second routine is executable to enable a user to select a magnitude associated with the zero-mean, random noise to be added to the process data. 34. The model generation system of claim 32, wherein the second routine is executable to add evenly distributed random noise with a maximum amplitude between 0.2 and 0.5 percent of the range of the process data. 35. The model generation system of claim 32, wherein the second routine is executable to add evenly distributed random noise with a maximum amplitude of approximately 0.4 percent of the range of the process data. 36. The model generation system of claim 20, wherein the second routine is executable to determine an amplitude of the noise added to the process data as a function of the process data. 37. The model generation system of claim 36, wherein the second routine is executable to determine the amplitude of the noise as a function of a range or a mean or a standard deviation of the process data. 38. The model generation system of claim 20, wherein the second routine is executable to determine an amplitude of the noise added to the process data as a function of a process input signal used to generate the process data. 39. A method of generating a control or simulation block for controlling or simulating at least a portion of a process, comprising: delivering a known process upset signal to the process to cause the process to undergo a change; collecting from the process, process data indicative of a response to the process upset signal; after collecting the process data and prior to determining any process model corresponding to the process data, adding noise to the process data to produce conditioned process data; determining a process model from the conditioned process data including the added noise; and generating the control or simulation block using parameters generated based on the determined process model. 40. The method of claim 39, further comprising adding noise to the process upset signal prior to delivering the process upset signal to the process to thereby add noise to collected process data. 41. The method of claim 39, wherein determining the process model includes determining one or more parameters of a parametric process model. 42. The method of claim 41, wherein determining the process model includes determining one or more parameters of an auto-regressive with external inputs process model. 43. The method of claim 39, wherein determining the process model includes determining a non-parametric process model. 44. The method of claim 39, wherein adding noise to the process data to produce conditioned process data includes adding random noise to the process data. 45. The method of claim 44, wherein adding random noise to the process data includes adding zero-mean noise to the process data. 46. The method of claim 45, wherein adding noise to the process data includes adding evenly distributed random noise with a maximum amplitude between 0.2 percent and 0.5 percent of the range of the process data. 47. The method of claim 45, wherein adding noise to the process data includes adding evenly distributed random noise with a maximum amplitude of approximately 0.4 percent of the range of the process data. 48. The method of claim 39, wherein determining the process model from the conditioned process data includes estimating a process dead time for the process from the process data, and using the estimated dead time and the conditioned process data to determine the process model. 49. The method of claim 48, wherein estimating the process dead time includes generating a further process model from the process data and determining an estimate of the process dead time from the further process model. 50. The method of claim 39, further including screening the process data or the conditioned process data prior to generating the process model from the conditioned process data. 51. The method of claim 39, wherein generating the control or simulation block using the determined process model includes generating a model predictive control block. 52. The method of claim 51, wherein generating the model predictive control block includes generating a multiple-input-multiple-output controller. 53. The method of claim 39, wherein generating the control or simulation block using the determined process model includes generating single-input-single-output control block. 54. The method of claim 39, wherein determining a process model from the conditioned process data includes determining a single-input-single-output process model. 55. The method of claim 39, wherein adding noise to the process data includes determining an amplitude of the noise as a function of the process data. 56. The method of claim 55, wherein determining an amplitude of the noise includes determining an amplitude of the noise as a function of the range of the process data, a mean of the process data or a standard deviation of the process data. 57. The method of claim 39, wherein adding noise to the process data includes determining an amplitude of the noise as a function of a process upset signal.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (31)
Ganesamoorthi, Sai; Blevins, Terrence L.; Tzovla, Vasiliki; Wojsznis, Wilhelm K.; Mehta, Ashish, Adaptive predictive model in a process control system.
Wojsznis, Wilhelm K.; Blevins, Terrence L.; Seeman, Richard C.; Nixon, Mark J., Integrated optimal model predictive control in a process control system.
Melvin, III,Lawrence S.; Shiely,James P., Method and apparatus for identifying a manufacturing problem area in a layout using a process-sensitivity model.
Hess, Todd M.; Kroll, Andreas; Lindberg, Carl-Fredrik S. M.; Modén, Per Erik; Petersson, Mikael; Praprost, Kenneth L., System and methodology and adaptive, linear model predictive control based on rigorous, nonlinear process model.
Demoro, Esdras Piraguacy; De Oliveira, Autur Toledo Martins; Axelrud, Celso, System for on line inference of physical and chemical properties and system for on line.
Beveridge, Robert Allen; Whalen, Jr., Richard J., Dynamic matrix control of steam temperature with prevention of saturated steam entry into superheater.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.