Artificial neural network feedforward/feedback control of a batch polymerization reactor

Control of polymerization reactors is a challenging problem due to nonlinear behavior of most polymer reactions. When the reaction is carried out in a batch reactor, the problem becomes even more difficult. In this work, the temperature control of batch polymerization of methylmetacrylate (MMA) is considered. The mathematical model developed by Ross and Laurence (1976) for suspension polymerization of MMA is used for computer simulation and control. The heat generation term is considered as a load and estimated via a trained feedforward artificial neural network. A feedforward/feedback control algorithm is used for controlling the reactor, and the performance of proposed scheme is compared with a well tuned PI controller. Simulation studies show that the neural network is able to estimate the heat generation term very well and considerable improvement in the closed loop performance has been observed.