Ftsm — Fast Takagi-sugeno Fuzzy Modeling

Takagi-Sugeno type fuzzy models are widely used for model-based control and model-based fault diagnosis. They provide high accuracy with relatively small and easy to interpret models. The problem that we address in this paper is that data driven identification of such fuzzy models is computationally costly. Whereas most identification algorithms for Takagi-Sugeno models restrict the model’s generality in order to simplify the identification, a different approach is taken here: we apply resilient propagation (RPROP), an efficient nonlinear optimization technique, for parameter identification in order to achieve a fast TakagiSugeno modeling (FTSM) that is suited to model high-dimensional data sets containing a large number of data.