GENETIC ALGORITHMS AND NEURAL NETWORKS USED IN OPTIMIZATION OF A RADICAL POLYMERIZATION PROCESS

The free radical polymerization process is a typical example of a multivariable open-loop control design problem. In this work, the batch free radical polymerization of methyl methacrylate process is analyzed by the multiobjective optimization technique, which shows in a more realistic way the multivariable aspects of the polymerization manufacture. The process objectives include monomer conversion, polymerization degree, polydispersity index and reaction time. Several methods for process optimisation are applied and compared: sequential quadratic procedure, genetic algorithms and neural networks based inverse modeling.