Industry-University Cooperative Research Grant: Model Predictive Control of Nonliner, Stochastic Processes

点击翻译按钮获取中文摘要

Abstract - Rawlings - 9311420 Of the advanced process control methods, model-predictive control (MPC) is probably the most popular method in industrial use today. Linear MPC refers to the class of control algorithms that compute a manipulated variable profile by utilizing a linear process model to optimize a linear or quadratic open-loop performance objective over a future horizon subject to linear inequality constraints on the state and manipulated variables. The first move of this open-loop optimal manipulated variable profile is implemented. This procedure is repeated at each time interval and the process measurements are used to update the optimization problem. While the process model is linear, the controller is nonlinear because of the inequality constraints. The objective of this research is to extend this theory to develop a more flexible and industrially relevant framework for process control applications. The issue of how to extend existing theory to the case of incomplete state measurements from nonlinear, stochastic systems will be investigated. This extension is important for industrial applications because the number and types of available sensors are limited and the sensors are subject to noise and occasional failure. This combined constrained state estimation and state feedback control, subject to the process constraints, would allow greater flexibility in designing the controller for challenging process control applications. The application of the method to nonlinear processes will also be explored, as most chemical processes are inherently nonlinear. This is an industry/university co operative research project; DuPont is cosponsoring the research. ! L !This program cannot be run in DOS mode. $ NE 4 =E V H @ P | V B P a ^ 0 ` 0 moire WEP KERNEL GDI USER ? ? ? c ? ? ? ? x ? moire1.exe SAVERDLGPROC SAVERINIT LAbstract - Rawlings - 9311420 Of the advanced process control methods, model-predictive control (MPC) is probably the most 0 0 ! ! D 0 0 ( Times New Roman Symbol & Arial " h k{ { { ~ a / Maria K. Burka Maria K. Burka