Reasoning Methods Of Expert System Based On Fuzzy Model

The development of artificial intelligence has greatly promoted the The development of expert system. Expert system can use domain accumulated experience and professional knowledge of expert, simulate thought process of human experts, solve the difficult problems which need to be settled by experts. Due to uncertainty in the practical problems, the traditional expert system is hard to set up more accurate reasoning mechanism. This paper focus on the reasoning method of expert system, and introduce fuzzy set theory and neural network method. Type-1fuzzy set include the defect in the fuzzy uncertainty, so we introduce Type-2fuzzy set to make up for the deficiencies. Besides, considering the advantages of the neural network, such as the self-learning mechanism as well as the powerful data processing ability. We combine the Type-2fuzzy set and neural network in order to improve the online self learning ability of the system. The main jobs of this thesis are as follows:1) In view of the uncertain problem of the actual problem, we propose the reasoning methods based on the GA-FNN expert system. First, establish FNN structure between fuzzy theory and neural network. Then form the expert system knowledge base with the weights and threshold value which are optimized by the genetic algorithm. At last, based on the knowledge base, we construct the reasoning model between fuzzy input and fuzzy output. This reasoning structure has parallel association function with high accuracy and fast convergence speed at the same time.2) In view of the defects of Type-1fuzzy set in the description of the uncertainty of the membership function, this paper puts forward the reasoning model based on Type-2fuzzy set. First, adopt fuzzy c-means method to extract rules automatically. We use Mamdani model in the reasoning process and with the method of Type-1centroid defuzzier to obtain the precise result. The simulation results show that, compared with the Type-1fuzzy reasoning method, the method of Type-2fuzzy reasoning method with the fuzzy interval of membership function can get more accurate results of reasoning.3) Considering the dynamic time-varying system, put forward the Type-2fuzzy neural network reasoning model. The model integrates the Type-2fuzzy theory with the advantage of handle the uncertain information and the neural network with the advantage of self-learning. In the simulation part, the nonlinear function with noise and ethylene gas phase polymerization process data are used to validate the models. The results of the study show that, the Type-2fuzzy neural network has the better tracking reasoning performance than the type-1fuzzy neural network, especially with the sudden change.