NON-LINEAR MODEL PREDICTIVE CONTROL OF A SEMI-BATCH EMULSION POLYMERIZATION REACTOR
Nonlinear control, based on the well known technique of model predictive control, is a popular control technology in chemical processing. The current research effort involves the application of this technique to control the performance of a semi-batch emulsion polymerization reactor.
The first part of the research has been to interface the Non Linear Model Predictive Control (NLMPC) software with a mathematical model of an emulsion polymerization reactor. The original model of emulsion polymerization of methyl methacrylate (MMA) is first modified into a semi-batch model with heat transfer. The modified model is then interfaced with NLMPC and simulations are performed to control important output parameters such as reaction temperature, conversion profile and number of particles (which were assumed to be monodisperse), by manipulating jacket temperature, and the flowrates of the soap and initiator streams.
The results of thee simulations, along with a detailed analysis are presented. A comparison of the performance of QDMC/NLMPC with a PID shows improved control using the model predictive techniques. Important tuning issues such as the effect of weights on input changes and outputs, and the length of prediction horizon are also addressed. It has been found that the nonlinear controller is able to handle constraints and plant/model mismatches quite well, and the outputs are regulated around their respective setpoints.