Nonlinear model predictive control of a batch polymerization process

Abstract

A Newton-type nonlinear model predictive control (NMPC) algorithm is applied through a simulation study to a batch reactor for polymerization of phenol-formaldehyde resole resins. The NMPC algorithm is combined with a recursive state and parameter estimation algorithm and a nonlinear state-space model, which is used both for the estimation and the prediction. The state-space process model is developed from first principles, and uncertain parameters are estimated off-line from logged process measurements. Some of the parameters can not be assumed to be constant during the batch, and should be estimated online during operation. NMPC strategies are compared with the conventional reactor control. The basic strategy is to maximize the reaction rate, with constraints on the reactor temperature and the heat of reaction. The latter constraint is determined by available cooling capacity. The simulation study indicates a potential reduction in batch cycle time with good safety margins.