Evolving Fuzzy Models of Mechatronics Applications : Plenary Talk
Radu‐Emil Precup · 2019
Evolving fuzzy models are characterized by the continuous online rule base learning. The structure and parameters of these fuzzy models in their Takagi-Sugeno or Takagi-Sugeno-Kang (TSK) forms are obtained by several online identification algorithms. The online identification algorithms continuously evolve the parameters of the fuzzy models, which are built online by adding new or removing old local models; this process is referred to as the adding mechanism. According to the classification proposed by Dovžan, Logar and řkrjanc in 2015, three categories of online identification algorithms are considered, (1), (2) and (3), and they have specific features described as follows. (1) Adaptive algorithms start with the initial TS fuzzy model structure given by other algorithms or by user experience; the number of space partitions/clusters does not change over time, and these algorithms adapt just the parameters of the membership functions and the local models. (2) Incremental algorithms implement only adding mechanisms. (3) Evolving algorithms, which besides the adding mechanism they also implement removing and a part of them merging and splitting mechanisms as well. The presentation is organized to highlight a part of the recent results obtained by the Process Control group of the Politehnica University of Timisoara, Romania, related to the application of incremental online identification algorithms that belong to the category (2) mentioned above. A convenient implementation of these algorithms is given in two versions that depend on how the estimation of parameters in TS fuzzy model rule consequents is performed. The identification algorithms produce sets of TS fuzzy models of some representative mechatronics applications in the laboratories of the Process Control group, where the experimental validation is guaranteed in order to next enable the development of model-based fuzzy control system structures: pendulum-crane systems, twin rotor aerodynamic systems, magnetic levitation systems, and anti-lock braking systems.