Study On Intelligent Control Based On Improved Fuzzy Neural Network

Fuzzy neural network focuses both on logic reasoning technology of fuzzycontrol and self-learning capability of neural network control, and tries to makeappropriate integration of those advantages, which is one of the emergentlyimportant researches in the intelligent control nearly several years. All theseresearches are of profound theory significance and practical application value.There are many kinds of fuzzy neural networks presently. But, they manifestmainly on the combination of frame and arithmetic. Based on the study offuzzy neural network's development, this paper proposed the idea that appliedsome theories and methods of fuzzy control to neural network, and confirmedthe feasibility by simulation, which opened new ways for fuzzy neuralnetwork's development.Firstly, after further research on the temporary fuzzy neural network'sdevelopment, a new classification method was proposed. It analyzed the fuzzyneural network's frame from the point of classification of neural network andfuzzy logic reasoning. And, there introduced a new feasible optimizationarithmetic in the arithmetic research, which was called particle swarmoptimization arithmetic.Secondly, a typically new network which was called normal fuzzy neuralnetwork was proposed in detail. It gave concrete arithmetic and cleardemonstration about the network's convergence by theorem-Stone Weierstrass.And its performance was confirmed by concrete identification.Thirdly, there introduced an improved fuzzy neural network control basedon identification of RBF. The scaling factor for fuzzification and the scalinggain were imbedded in the network, which composed the network's input layerand output layer. Instead of generally approximate method, the networkemployed the radial basis function (RBF) neural network to offer preciseinformation of Jacobian for the system. Finally, it provided an ideal schemethat transformed the basic range into fuzzy range effectively.At last, enlighten by the idea of self-adjusted factor, the paper proposed afuzzy radial basis function (RBF) neural network with self-adjusted factorwhich was optimized by the particle swarm optimization algorithm.