IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0454937
(2003-06-05)
|
등록번호 |
US-7272454
(2007-09-18)
|
발명자
/ 주소 |
- Wojsznis,Wilhelm K.
- Blevins,Terrence L.
- Mehta,Ashish
|
출원인 / 주소 |
- Fisher Rosemount Systems, Inc.
|
대리인 / 주소 |
Marshall Gerstein & Borun LLP
|
인용정보 |
피인용 횟수 :
35 인용 특허 :
14 |
초록
▼
A process controller that may be used to control a process having a set of process outputs effected by a set of process control input signals includes a multiple-input/multiple output controller that uses the process outputs to develop the set of process control input signals and a process model, wh
A process controller that may be used to control a process having a set of process outputs effected by a set of process control input signals includes a multiple-input/multiple output controller that uses the process outputs to develop the set of process control input signals and a process model, which may be a non-linear process model, that receives the set of process control input signals to produce a prediction signal for one or more of the process outputs. The multiple-input/multiple-output control element includes another process model, which may be a standard linear process model, to develop a prediction vector for each of the process outputs and includes a correction unit that modifies the prediction vector for the one or more of the process outputs using the prediction signal for the one or more of the process outputs to thereby compensate for the non-linearities of the process.
대표청구항
▼
What is claimed is: 1. A process control element for use as a portion of a process control routine implemented on a processor to control a process that includes a set of process outputs effected by a set of process control input signals, the process control element comprising: a computer readable m
What is claimed is: 1. A process control element for use as a portion of a process control routine implemented on a processor to control a process that includes a set of process outputs effected by a set of process control input signals, the process control element comprising: a computer readable memory; and a control element stored on the computer readable memory which when executed on the processor implements multiple-input/multiple output control of the process, the control element including; a first process model which receives the set of process control input signals to produce a prediction signal for one of the process outputs; and a multiple-input/multiple-output control element which receives an indication of the process outputs to develop a set of control signals, the multiple-input/multiple-output control element including a second process model that develops a prediction vector for multiple process outputs including the one of the process outputs and a correction unit coupled to an output of the first process model to receive the prediction signal for the one of the process outputs and coupled to an output of the second process model to receive the prediction vector for the one of the process outputs and that modifies the prediction vector for the one of the process outputs using the prediction signal for the one of the process outputs to produce a corrected prediction vector for the one of the process outputs. 2. The process control element of claim 1, wherein the first process model is a non-linear process model. 3. The process control element of claim 1, wherein the second process model is a linear process model. 4. The process control element of claim 1, wherein the first process model is neural network process model. 5. The process control element of claim 4, wherein the multiple-input/multiple-output control element is a model predictive control controller and wherein the set of control signals are delivered as the process control inputs signals to control the process outputs. 6. The process control element of claim 4, wherein the multiple-input/multiple-output control element is an optimizer and wherein the set of control signals includes targets for a process controller. 7. The process control element of claim 1, wherein the first process model is a non-linear process model and the second process model is a linear process model. 8. The process control element of claim 7, wherein the first process model produces a second prediction vector for the one of the process outputs as the prediction signal and the correction unit modifies the prediction vector produced by the second process model by replacing the prediction vector with the second prediction vector. 9. The process control element of claim 7, wherein the first process model produces a steady state value for the one of the process outputs as the prediction signal and the correction unit modifies the prediction vector produced by the second process model by combining the steady state value for the one of the process outputs with the prediction vector for the one of the process outputs. 10. The process control element of claim 9, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the second process model by adding to the prediction vector produced by the second process model a difference between first and second model predictions at the end of time horizon multiplied by a first or higher order exponential function. 11. The process control element of claim 9, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the second process model by adding to the prediction vector produced by the second process model a difference between first and second model predictions at the end of time horizon multiplied by a first or higher order exponential function with time constants equal to a fraction of the process output time to steady state. 12. The process control element of claim 1, wherein the first process model produces a second prediction vector for the one of the process outputs as the prediction signal and the correction unit modifies the prediction vector produced by the second process model by replacing the prediction vector with the second prediction vector. 13. The process control element of claim 1, wherein the first process model produces a steady state value for the one of the process outputs as the prediction signal and the correction unit modifies the prediction vector produced by the second process model by combining the steady state value for the one of the process outputs with the prediction vector for the one of the process outputs. 14. The process control element of claim of claim 13, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the second process model by multiplying the prediction vector produced by the second process model by a component using a ratio of the steady state value for the one of the process outputs and the steady state value at the time horizon of the prediction vector for the one of the process outputs. 15. The process control element of claim of claim 13, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the second process model by multiplying the prediction vector produced by the second process model by a component using a difference between the steady state value for the one of the process outputs and the steady state value at the time horizon of the prediction vector for the one of the process outputs. 16. A method of controlling a process having a set of process outputs effected by a set of process control inputs, the method comprising: using a first process model to develop a prediction vector for one or more of the process outputs; using a second process model to produce a prediction signal for one of the process outputs using the set of process control signals; providing the prediction vector and the prediction signal as inputs to a correction unit; correcting the prediction vector for the one of the process outputs in the correction unit using the prediction signal for the one of the process outputs to produce a corrected prediction vector; and using the corrected prediction vector to produce a set of control signals for use in controlling the process. 17. The method of claim 16, wherein using the first process model includes using a first linear process model as part of a multiple-input/multiple-output control routine that receives an indication of the process outputs and wherein using the corrected prediction vector includes using the multiple-input/multiple-output control routine to produce the control signals to be delivered to the process control inputs from the corrected prediction vector. 18. The method of claim 17, wherein using the first linear process model as part of a multiple-input/multiple-output control routine includes using a model predictive control routine as the multiple-input/multiple-output control routine. 19. The method of claim 17, wherein using the second process model includes using a non-linear process model. 20. The method of claim 19, wherein using the second process model includes using a neural network process model. 21. The method of claim 20, wherein using the second process model includes producing a steady state prediction value of the process output as the prediction signal. 22. The method of claim 20, wherein using the second process model includes producing a high-fidelity prediction vector for the process output as the prediction signal. 23. The method of claim 16, wherein using the second process model includes producing a steady state value for the one of the process outputs as the prediction signal and wherein correcting the prediction vector includes modifying the prediction vector produced by the first process model by combining the steady state value for the one of the process outputs with the prediction vector for the one of the process outputs. 24. The method of claim 23, wherein using the first process model includes producing the prediction vector for the one of the process outputs to include a steady state value at a time horizon and wherein modifying the prediction vector includes adding to the prediction vector a difference between the steady state value at the time horizon of the prediction vector and the prediction signal at the time horizon multiplied by a first or higher order exponential function. 25. The method of claim 23, wherein using the first process model includes producing the prediction vector for the one of the process outputs to include a steady state value at a time horizon and wherein modifying the prediction vector includes adding to the prediction vector a difference between the steady state value at the time horizon of the prediction vector and the prediction signal at the time horizon multiplied by a first or higher order exponential function with time constants equal to a fraction of the process output time to steady state. 26. The method of claim 16, wherein using the first process model includes using a first linear process model as part of a control optimizer that receives an indication of the process outputs and wherein using the corrected prediction vector includes using the control optimizer to produce a set of target signals for a controller using the corrected prediction vector. 27. A process controller for use in controlling a process having a set of process outputs effected by a set of process control input signals, the process controller comprising: a multiple-input/multiple output controller that receives an indication of the process outputs and develops the set of process control input signals, the multiple-input/multiple-output controller including a first process model that develops a prediction vector for one of the process outputs; a second process model that receives the set of process control input signals to produce a prediction signal for the one of the process outputs; and a correction unit coupled to an output of the first process model to receive the prediction vector for the one of the process outputs and coupled to an output of the second process model to receive the prediction signal for the one of the process outputs and that modifies the prediction vector for the one of the process outputs using the prediction signal for the one of the process outputs to produce a corrected prediction vector for the one of the process outputs. 28. The process controller of claim 27, wherein the second process model is a non-linear process model. 29. The process controller of claim 27, wherein the first process model is a linear process model. 30. The process controller of claim 27, wherein the second process model is neural network process model. 31. The process controller of claim 27, wherein the multiple-input/multiple-output controller is a model predictive control controller. 32. The process controller of claim 31, wherein the second process model is a non-linear process model. 33. The process controller of claim 32, wherein the second process model is a neural network process model. 34. The process controller of claim 27, wherein the first process model is a linear process model and the second process model is a non-linear process model. 35. The process controller of claim 34, wherein the second process model produces a second prediction vector for the one of the process outputs and the correction unit modifies the prediction vector produced by the first process model by replacing the prediction vector with the second prediction vector. 36. The process controller of claim 34, wherein the second process model produces a steady state value for the one of the process outputs as the prediction signal and the correction unit modifies the prediction vector produced by the first process model by combining the steady state value for the one of the process outputs with the prediction vector for the one of the process outputs. 37. The process controller of claim 36, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the first process model by adding to the prediction vector produced by the first process model a difference between first and second model predictions at the end of time horizon multiplied by a first or higher order exponential function. 38. The process controller of claim 36, wherein the prediction vector for the one of the process outputs includes a steady state value at a time horizon and wherein the correction unit modifies the prediction vector produced by the first process model by adding to the prediction vector produced by the first process model a difference between first and second model predictions at the end of time horizon multiplied by a first or higher order exponential function with time constants equal to a fraction of the process output time to steady state.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.