A new algorithm for T-S fuzzy modeling based on hierarchical fuzzy-clustering

Aimed at the disadvantage of fuzzy c-means clustering algorithm in T-S modeling, a new modeling approach based on hierarchical fuzzy-clustering is proposed. Firstly, the input variables were determined by grey relational analysis algorithm. The input space was partitioned into some local regions by entropic clustering algorithm, then the centers of the clusters were further proceed using fuzzy c-means clustering algorithm and premise structure and parameters were determined. Finally, a least square algorithm was provided to determine the consequent part of each rule. The proposed method was applied to the well-known Box-Jenkins gas-furnace data and ball mill system in a power plant. The modeling results demonstrate that the algorithm is simple, useful and the precision of model is high.