Fuzzy model identification by FCM clustering algorithm and fuzzy controller design

• Main Authors: Wei-En Chang, 張維恩
• Other Authors: Chung-Chun Kung
• Format: Others
• Language: en_US
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/37147018727277599666

碩士 === 大同工學院 === 電機工程研究所 === 87 === ABSTRACT Fuzzy controllers have been successfully applied in many cases to which conventional control algorithms are difficult to be applied. Recently, it was proven that fuzzy systems are capable of approximating any real continuous function to arbitrary accuracy. This result motivates us to use the fuzzy systems as identifiers for nonlinear dynamic systems, and then design the fuzzy controllers based on the fuzzy system. In the process of the identification, most of the techniques used in the literatures for fuzzy c-mean (FCM) clustering are based on off-line operation. Namely, these techniques require all the objects or the distance matrix to be available before the start of any FCM clustering routine and it seems impractical in some cases. In this thesis, we propose a modified FCM clustering algorithm named the recursive fuzzy c-mean (RFCM) clustering algorithm. Comparing with the traditional FCM, the proposed algorithm has the following advantages: (1) Lower memory size, and lower computational complexity. (2) It is more applicable than traditional FCM in system identification and fuzzy control. There are three main objectives in this thesis: (1) We derive the clustering algorithm to partition the universal data set. (2) The Takagi and Sugeno’s fuzzy models are used as identifier for nonlinear dynamic systems. (3) The fuzzy model-based controller design method for tracking control is proposed based on this fuzzy system. Simulation results show that the identification algorithm exhibits good performances and the fuzzy model-based controllers could perform successful tracking ability.