Parameter optimization of fuzzy-neural-network decoupling controller for adjusting temperatures of regenerative reheating furnace

This paper addresses the problem of controlling the temperature of the combustion process in a billet regenerative reheating furnace, and presents a method of optimizing the parameters of a fuzzy-neural-network decoupling controller (FNNDC) through the combination of an immune clone and a self-adaptive mutation rate. First, a recurrent neural network is built to model the process based on data from actual runs. Then, the number and parameters of hidden neurons in the FNNDC are determined by means of a fuzzy c-means clustering strategy. Finally, an algorithm for immune-clone evolution (ICE) is combined with a self-adaptive mutation rate to optimize the connecting weights of the FNNDC. This method features global optimization and high precision in local searches. Simulation results demonstrate the validity of this method and its superiority over genetic algorithms.