Research On The Design Of Fuzzy Controller Based On Data Mining

In the process of modern industrial control, fuzzy control is widely used in complex process and object control for its virtues such as easy realization, robust. But how to obtain fuzzy rule is a difficult problem to the designer of controller when previous knowledge is not exist or hypothesis condition is not sufficient. But sometimes people can collect plentiful input-output datum in the process of industrial production, and data mining technique can mine useful knowledge or rules from plentiful datum. In such a case, designing elementary framework of fuzzy controller using fuzzy data mining technique, mining fuzzy associated rules and forming fuzzy control query table from data base, optimizing the parameters of membership functions and improving the performance of fuzzy controller utilizing genetic algorithm, and so on. These have become important tasks in the field of the application of fuzzy control.Firstly, the fundamental principles of fuzzy control and data mining are analyzed in this paper. Through studying associated rules algorithm and analyzing the operational principle of mining associated rules from data base, a new method of mining and forming fuzzy control query table is put forward. Then fuzzy model is mined from data base utilizing fuzzy data mining technique and a digital elementary framework of fuzzy controller is designed. By studying the merit and the demerit of simple genetic algorithm, a real-coded genetic algorithm (RGA) is put forward, and is used to optimizing the parameters of membership functions, and combined with least-squares methods (LSM) to determine rules consequent parameters. Moreover, an index examination of giving attention to much performance is adopted to determine the proper research range, the number of generations and the total number of partitioned point. It disposed preferably the tradeoff between complexity and approximation accuracy of the system.