CONTROL OF A CHAOTIC POLYMERIZATION REACTOR - A NEURAL-NETWORK-BASED MODEL-PREDICTIVE APPROACH

Continuous polymerization processes may be very sensitive to small cha nges of the operation conditions. Continuous VA (vinyl acetate) soluti on homopolymerization reactors may present multiple steady-states and oscillatory behavior. A predictive control scheme that uses an interna l model of the process is employed to stabilize such reactors and make them less sensitive to disturbances while subject to ''hard'' control action constraints. An ANN (artificial neural network) is used as the internal model, leading to fairly good predictions of the reactor beh avior, including its multiplicities. The performance of the resulting control algorithm is compared to that of a ''well-tuned'' conventional proportional-integral-derivative (PID) controller.