The Mechanism Of Genetic Optimization Of Fuzzy Control System Of Knowledge Set The Final Draft

With the optimization problem of MAMDANI fuzzy system, this paper is make up a genetic fuzzy system unifying the fuzzy system and genetic algorithm. According to the FRBSs, it studies the genetic optimization for the knowledge-base of fuzzy control system which contains genetic optimization of rule-base and data-base. Specially, this paper tries to combine the Pittsburgh approach and Michigan approach.This paper proposes a fuzzy-based genetic algorithm used in fuzzy rules generation and optimization for rule-base. It solves the problem which is hard to avoid the local optimal solution or slower population diversity when using genetic algorithm to generate the fuzzy rules. The algorithm is used the rule population diversity and evolutionary speed to automatically adjust the crossover rate and mutation rate based on fuzzy logic, which realizes the automatic control rules generation of genetic fuzzy system. According to the performance index of control system, it sets the fitness function. Simulation results demonstrate that the algorithm is practical and effective. Virtually, fuzzy-based genetic algorithm is a self-adapting genetic algorithm, which is able to self-adaptive control the population diversity and evolutionary speed according to the distribution of different rules, in order to generate the most optimization rule-base.This paper introduces a genetic lateral tuning algorithm to membership function for data-base optimization. It solves the problem that using general data-base can't get the fine performance for control system when rules are fixed. It utilizes the genetic algorithm and genetic lateral tuning algorithm to obtain the fine membership function. Simulation results demonstrate that the algorithm is practical and effective. Essentially this algorithm is a modified genetic algorithm, which is presented in the genetic coding. It codes the membership function using binary model, and then decreases the search space in order to quickly optimize data-base.Finally, it creates a new fuzzy genetic network model based on the P and M and above algorithm, which integrates the global and local optimization approach. The global optimization is contained in tuning membership function before fuzzy system is fixed; Local optimization is reflected in making operation results as normalization building-out when fuzzy system is operated in each step, which feedbacks the rules involving in operation. So it combines the data-base tuning and rule-base learning as an integrated one. The data-base tuning is utilized binary model, and rule-base is simplified a normalization factor which size is decided the normalization output in fuzzy system. The feature of this model is combined with both P and M, which is contained in the processing of normalization. The network model is built in C language and simulation results demonstrate that the algorithm is practical and effective.