Block-structured models composed of nonlinear fuzzy dynamic and static parts : a case study

The paper addresses issues of the dynamic fuzzy Takagi- Sugeno models identification for multi-step ahead prediction. In the case of highly nonlinear models, standard Takagi-Sugeno models may be hard to identify if they should be designed for recurrent prediction generation. In such a case, alternative fuzzy block-structured models composed of fuzzy dynamic and fuzzy static parts may be useful. Two main benefits of the proposed models are: (1) possibility to speed-up model tuning procedure, (2) potential to fine-tune an already available, standard Takagi-Sugeno model. The benefits offered by the proposed models are illustrated using the example of identification of a nonlinear process – a system consisting of two tanks of different shapes (cylindrical and conical ones).