On Data-driven Takagi-Sugeno Modeling of Heterogeneous Systems with Multidimensional Membership Functions

摘要

In case of a structural mismatch between a Takagi-Sugeno model and a nonlinear system that is to be modeled, identification algorithms tend to compensate this by a finer granular partitioning. This is in conflict with the objective of parsimonious models. A novel TS model type with multi-dimensional fuzzy sets and locally adjustable/heterogeneous fuzziness is presented with a method for the adjustment. This new approach enhances the structural flexibility of the model while the number of parameters increases negligibly. Two case studies, including compressor modeling, illustrate the performance of the proposed method.