Fuzzy logic supervised ANN feedforward control for a batch polymerization reactor

A polymerization process is a time-varying process. In this paper, the temperature control of a highly exothermic batch polymerization reactor is studied. A fuzzy logic model supervised artificial neural network feedforward PI feedback control (FL-ANN/PI) strategy is developed, in which the FL-ANN feedforward output is applied to compensate for the effect of the major disturbance, heat of reaction. The ANN part consists of two neural networks, a fully recurrent network, which is used to estimate the effect of the heat of reaction on the reaction temperature, and an inverse backpropagation network, which calculates the feedforward signal. The fuzzy logic model supervises the final feedforward output to enhance the reliability of the ANN feedforward control.