Semi-continuous copolymer composition distribution predictive control using a double ANN model structure

Copolymer composition distribution (CCD) is essential for product quality in copolymer manufacturing. In this work, we implement a double ANN structure for the on-line one shot control of MA-VAc semi-continuous latex copolymerization system. The control strategy is firstly assuming that the system is operating in a semi-starved condition. The feeding rate of MA can only be adjusted once in a single batch. Based on an intermediate measurement, a hybrid ANN model, that combines the information provided by the experimental data and theoretical model simultaneously, is implemented to predict the product quality at the end of the batch. However, it also has been found that because of the effects of measuring error, implementing a double ANN structure is better than implementing a single ANN. A critical parameter is identified by the first ANN. The parameter, in turn, is used as an input of the second ANN, that is a hybrid model. Both the experimental and simulation studies show that the proposed double ANN is superior to a single ANN structure. Besides, the experimental studies also show that the ANN model predictive control is promising for the CCD control of a semi-continuous latex system.