Set-membership identification of T-S fuzzy models using support vector regression
He Liqing, Xianfang Sun · 2009
In this paper, the problem of identifying nonlinear systems with unknown-but-bounded (UBB) noise is investigated. The fuzzy inference theory and support vector regression (SVR) learning mechanism are used to construct a T-S model for the nonlinear system based on input and output data with UBB measurement noise. After the structure of a T-S model is determined using SVR, all the feasible parameters in its consequent part are found by the optimal bounding ellipsoid (OBE) algorithm and then a class of feasible nonlinear models are found which are consistent with the given noise bound series and input-output data set. The simulation results illustrate that the proposed method is effective.