Reduction of Fuzzy Rules by Using Hybridization with Clustering

The fuzzy controllers have found numerous applications to control systems mainly due to its robustness to external disturbances. The design of any fuzzy controller is mainly dependent on knowledge base which consists of rule base and membership functions. In conventional fuzzy logic controller (FLC) the performance of the process control can be improved if the number of rules are increased which is achieved by either increasing the number of inputs or by increasing the number of linguistic variables. In either case, as the number of rules increases, the computation time increases there by taking longer time for processing. In this paper a three-input single-output FLC with seven membership functions is considered for process control. With three input and seven linguistic variables the total number of rules will be 343 and with these many number of rules the process becomes very slow. In this paper a novel method using hybrid fuzzy controller to reduce the rules present in the rule base is proposed. In the proposed method the rules are reduced from 343 to 98 rules and for further reduction in number of rules an approach using clustering is also proposed which reduces the rules to 61 and at the same time the performance of the process control remains similar to that of the conventional fuzzy controller. Due to the reduction in number of rules the computational memory utilized is also reduced considerably. The proposed method is applied to second order and third order system and the simulation results are presented and analyzed to show the effectiveness of the proposed method