Fuzzy Rbf Neural Network Controller Based On Genetic Algorithm Design

The perfect union of Fuzzy Control technology, Neural Network and Genetic Algorithm has shown great potential for process control. Based on improvements to the above union, this paper introduces a novel promising method for design of a robust Fuzzy RBF Neural Network Controller, independent of the object model and priori knowledge. Application to the typical nonlinear system CSTR (Continuous Stirred Tank Reactor) shows the method has good perfomiance and promise. The two key points to the success of the method are: 1. Improvement to Genetic Algoritlun: the way that employs fuzzy coding to represent the structure of the Fuzzy RBF Neural Network greatly lessens the parameter optimization task thus speeding the Genetic Algorithm; also the introduction of Excellent Schema Self-learning Fuzzy Genetic Algorithm improves the performance; 2. Freely use of two controller structures: Based on the equivalence of two fuzzy neural networks, alternate of the two structures makes full use of their advantages.