Design and application of model-on-demand predictive controller to a semibatch copolymerization reactor

This article considers the design of a multivariable model-on-demand predictive controller (MoD-PC) and its application to polymer quality control in a semibatch MMA/MA copolymerization reactor. The MoD-PC is designed by combining the model-on-demand (MoD) framework with the conventional model predictive controller. For this purpose, a local autoregressive exogenous input model is constructed with a small portion of data located in the region of interest at every sample time when a model equation is needed for output prediction. This model equation is then used to calculate the optimal control input sequence. Through the open-loop test, the reaction temperature and free volume are shown to be an adequate choice for the elements of the regressor state vector in the data searching step. The validity of the identified model is corroborated by the prediction method. The results of simulation studies for the regulation and disturbance rejection problems with and without noise demonstrate that, despite the nonlinearity of the system, the MoD-PC is an effective strategy with a low computational load for the production of copolymers with desired properties.