Hierarchical Fuzzy Identification for Complex Systems
Ping Zhang · Control theory & applications · 2002
Drawn ideas from GMDH algorithm, a new approach to hierarchical fuzzy identification for complex systems is presented to overcome the serious limitation of Takagi and Sugeno's model (T_S model), that is the problem of curse of dimensionality. Firstly, a new hierarchical fuzzy model, which consists of a number of hierarchically connected special T_S type models with two input variables, is described in detail. Then, a concrete method for identifying the hierarchical fuzzy model is also proposed. The main features of the presented method are: a) Fuzzy C_means (FCM) is used to evaluate the significance of each input variable for rationally constructing the hierarchical model in the stage of structure identification; b) Since parameters to be identified are properly determined in advance, they can be obtained rapidly by using extended Kalman filter algorithms. Finally, an example is given to demonstrate the validity of the approach.