T-S model identification based on silhouette index and improved gravitational search algorithm

In this paper, an approach based on silhouette index (SI) and improved gravitational search algorithm (IGSA) is presented to deal with T-S model identification problem. Clustering algorithm employing SI and IGSA is introduced for structure identification. The SI considers both the intra-cluster cohesion and the inter-cluster separation which can highly assess the accuracy of clustering. One cluster represents a fuzzy rule. Cluster center is regarded as the gauss membership function center parameter, which is identified by IGSA. The improved algorithm IGSA is also used for parameter identification of T-S model. It introduces the mutation of genes to standard GSA and considers the best solution, which enhances the search space and improves the ability of sharing global information. The simulation results produced by a two order nonlinear system and Box-Jenkins gas stove illustrate the effectiveness of the proposed method.