T–S Fuzzy Model Identification Based on a Novel Hyperplane-Shaped Membership Function

Abstract

The fuzzy membership function is a crucial factor that may affect the model structure and modeling accuracy. Although hyperplane-shaped clustering (HPSC) has been widely used in Takagi–Sugeno (T–S) fuzzy model identification, there is no well-designed membership function in these approaches. The commonly used bell-shaped Gaussian function is actually inappropriate to employ together with HPSC in fuzzy modeling. In this paper, a hyperplane-shaped fuzzy membership function is designed to match HPSC for T–S fuzzy model identification for the first time. Experimental results on several benchmark problems indicate that modeling accuracies have been greatly improved from the original modeling approach by replacing the proposed membership function.