Application Of GA In Optimization Of Neural Fuzzy Controller And BP Neural Networks

Intelligent control develops much along with the artificial intelligence's progress in the world. Fuzzy control and neural network as two kinds of intelligent control system are applied to every field of science. As there are many deficiencies in the design, this paper uses GA to optimize the designing. In this paper, the following research results are achieved:1. In order to overcome the defects of the basic GA, this paper improves on the selecting, crossover operator and mutation operator of the basic GA. The simulation indicates that the performance of the amendatory GA is better than the basic GA's.2. In this paper the main structure and theory of fuzzy control have been researched and the shortcomings that the design of fuzzy controller are too relies on the knowledge and experience of the experts have discussed, which leads to that the design of fuzzy control is not consummate and systematization as the modern control theory. While the GA does not require that the optimized functions possess continuum differentiability, GA is proposed to optimize the parameters of fuzzy controller. And the algorithm has applied to control the inverted pendulum system. The results demonstrate that the performance of the fuzzy controller is improved.3. A BP neural network is developed for a nonlinear classification, in which a novel neural network training algorithm is proposed. It combines BP algorithm with genetic algorithm in order to overcome the shortcomings that BP algorithm is easy trapped to a local optimum, has a low speed of convergence and it is difficult to select its parameters properly. And the algorithm has applied to the identification of the shapes of ST segments successfully. The results demonstrate that the accuracy is improved and convergent rate is increased comparing with BP algorithm. It is great useful for automatic diagnosis of Cordial Disease.