| Advanced process control strategies are a means by which chemical plants can operate effectively and economically to meet challenges like changing market demands, tighter process efficiency, better product quality, and stricter environmental standards. Of the advanced process control strategies developed in the last two decades, Model Predictive Control (MPC) algorithms are most widely used. However, MPC design and implementation is time consuming and complex. A key to the continued success of the MPC algorithms is the development of systematic design algorithms that can be understood and used by plant operators, and can make control system design decisions easier and intuitive.; This thesis presents design tools to improve the current practice of controller tuning and process identification, the two important steps in MPC design. Dynamic Matrix Control (DMC), one of the widely used MPC algorithm, is selected for study. Robust design of DMC controllers that explicitly presents compromises between stability, performance, and computing efficiency, using move suppression parameters, blocking intervals, and condensing intervals as the tuning parameters, is developed. For the selection of blocking and condensing intervals, a parametric design algorithm (that can be completely automated) using wavelet packet transform, perturbation analysis, and information theory is explained and illustrated. In the area of process identification, input test signal design and prefilter design are addressed. The influence of process knowledge and control system specifications on the Pseudo Random Binary Signal input signal is quantified. A systematic prefilter design procedure that provides explicit information on the compromises in design is presented. The Tennessee-Eastman Challenge Problem and the Shell Process Control Problem are used as the test-beds to evaluate the tools developed and to identify new research problems in the area of process control. |
