Neural Network Estimator for Gas-Phase Ethylene Polymerization Process

Abstract

Ethylene concentration is a significant parameter in the polymerization process because it affects the molecular weight distribution of the polymer produced. However, there is a challenge to obtain a good estimation of ethylene concentration due to its highly nonlinear behavior and interaction among state variables. In this work, the neural network (NN) has been applied to estimate the concentration of ethylene monomer with generic model control (GMC) controlling the reactor temperature. The neural network has been trained with backpropagation and Levenberg-Marquart techniques. Simulation results have shown that the neural network estimator can provide good estimates of the monomer concentration under nominal condition and disturbance cases.