New clustering algorithm for identification of a nonlinear stochastic model

Many clustering algorithms have been proposed in literature to identify the premise and consequence parameters involved in the TS fuzzy model. In this paper this parameters are estimated at the same time and this from the minimization of four optimization criteria. The proposed algorithm constitutes an extension of the algorithm proposed by J.Q. Chen in 1998. However, in this paper we introduced some modification on the optimization criteria and especially the last two criteria, thus we replaced the Euclidean distance by another non-Euclidean distance when calculating the fuzzy partition matrix. The purpose of these modifications is to introduce more robustness with the algorithm especially for highly nonlinear systems and those operating in a stochastic environment. The efficiency of the algorithm is tested on an electro-hydraulic system.