Chaos-RBF network and its application in soft sensor of continuous catalytic reforming process

Abstract

Based on concepts of chaotic theory, a novel RBF neural network model (Chaos-RBF) is presented. For searching better weights of RBF neural network, chaotic variables are used. And Chaos-RBF is applied to intelligent soft sensor technology and optimization in the device of 600 thousand t/a UOP continuous catalytic reforming (CCR) of a refinery. Compared to several other networks, such as BP, PLS-BP, RBF and wavelet neural networks, they are used to intelligent soft sensor modeling, the results show that, Chaos-RBF have more powerful ability to obtain better neural network structure and higher precision than any other neural network model.