On designing an optimal fuzzy neural network controller using genetic algorithms

This paper presents an optimal scheme for the design of a fuzzy neural network as a controller through simulating the process of the controlled system. The structure and the parameters of the fuzzy neural network are first optimized by a three-stage strategy applying genetic algorithms off-line. The defuzzification part of the fuzzy neural network is then reconstructed and is optimized to refine the control rules online. The simulation results demonstrate that the responses are more favorable than that of conventional fuzzy controller and conventional fuzzy neural network controller trained by expert data.