Integration of Multilayer Perceptron Networks and Linear Dynamic Models

Abstract

Recently neural network dynamic modeling has drawn much attention not only from academia but also from industry. It has been shown that neural networks can model complex chemical processes. However a neural network dynamic model might be difficult to develop when training data does not contain sufficient nonlinear information. Owing to the capacity of modem control systems.a great deal of steady state data is stored. This paper presents a Neural Network Hammerstein (NNH) modeling approach in order to fully utilize the abundant steady state information. To a complex polymerization process is used to demonstrate the proposed modelingapproach