Design of a Process Controller Based on an Upgraded Elman Neural Network

An upgraded Elman neural network with extra learning capability is utilized to construct process controllers for industrial processes. The basic structure of this upgraded Elman network is introduced. A comprehensive learning algorithm is developed which optimizes not only the feedforward but also the self-feedback connections of such partially recurrent neural networks. The identification system used in the controller design is arranged in a parallel pattern. The control system is devised in a feedforward plus feedback format based on the inverse model identified of the process under control. Numerical results for the control of a pH neutralization process are also presented.