Nonlinear dynamic system modeling based on T-S fuzzy model with structural risk minimization

A number of techniques based on Takagi-Sugeno (T-S) fuzzy models from measured data have been introduced to construct nonlinear dynamic system, due to their capability to approximate any nonlinear behavior. However, most attention has been focused on antecedent structure identification, there is a small methods that provide the investigation or improvement for the consequent parameters identification of T-S fuzzy model. Consequently, this paper proposes a novel method to identify nonlinear dynamic system based only on measured data, in which concentrates on the identification of consequent parameters with structural risk minimization for T-S fuzzy model. The proposed method combines the advantages of fuzzy system theory and some ideas from Least Squares Support vector Machine (LS-SVM). Gustafson-Kessel clustering algorithm(GKCA) is first applied to split training data into R clustering subsets and structural risk based on LS-SVM is decomposed into R terms likewise. Following that, the decomposed structural risk is to be identified consequent parameters of T-S fuzzy model. Finally, the viability and superiority of the method are verified by nonlinear dynamic system simulation.