Experimental demonstration of nonlinear model-based control techniques on a lab-scale distillation column

Distillation is an important and energy intensive operation in the chemical industry, and the economic incentives for simultaneously controlling the product compositions are obvious. Recent changes in competition, the market, and process instrumentation and computers have made process control a primary focus in the chemical industry. The National Research Council Committee identified Computer Assisted Process and Control Engineering as one of the priority areas in its 1987 report and urged universities to integrate process control in their curriculum.; This dissertation reports on a Nonlinear Process Model-Based Control (NPMBC) technique developed for the control of a lab-scale distillation column. NPMBC uses a nonlinear approximate model of a process directly for control purposes. The model is not a rigorous simulator but incorporates the major characteristics of the process and for effective control must adapt to the process as the process changes. A few adjustable parameters are used to match the model to the measured process data. This study uses Generic Model Control (GMC) technique, one type of NPMBC.; Collaboration with industry provided guidance, instrumentation, and funding to automate a unit operations laboratory distillation column with industrial data acquisition and control equipment. Experimental setup, development of the steady state controller model, and experimental demonstration of the unconstrained and constrained GMC techniques are the main features of this study. The column is operated with a binary methanol-water system at atmospheric pressure. Top and bottom compositions are controlled using reflux and vapor boilup as the manipulated variables. Process characteristics include nonlinear multivariable interactions, non-ideal thermodynamic behavior, significant unintentional disturbances, and disparate dynamics. Adjustable model parameters are tray efficiency and vapor boilup bias.; Two constraint control strategies designed to solve problems involving constraints on manipulated variables have been developed. This constrained controller works in conjunction with the unconstrained GMC algorithm. The controller is tested in servo (setpoint changes), regulatory (feed flow and feed composition disturbances), and constraint (constraints on vapor boilup at large feed flow and feed composition disturbances) modes and is shown to work satisfactorily.