Designing a Fuzzy Logic Controller with a NonParametric Similarity-Based Clustering Algorithm

Abstract

Clustering algorithms are used to produce powerful extension to design a three-term (proportional plus integral plus derivative—PID) fuzzy logic controller(FLC). They can be used to eliminate the presumption of the existence of expert information and extract rules that can satisfactorily represent the systems. In this paper, a non-parametric clustering algorithm based on data similarity, which is free of user-defined parameters is proposed. This algorithm is simple and fast. For comparison purposes, two methods of extracting the rules for a three-term FLC from the generated clusters were presented. These two methods entail on the use of the linguistictype model and the Takagi-Sugeno-Kang (TSK)-type model. Two applications representing second-order systems and thirdorder systems were used to analyze the performance of the proposed design methods and compared with other design methods. The analysis shows that the proposed design methods are efficient and superior to other design methods with respect to transient response, accuracy, and robustness to variation of defuzzification methods.