Improved fuzzy clustering for identification of Takagi-Sugeno model

The paper deals with the Takagi-Sugeno fuzzy modeling approach improved with procedure for cluster initialization. Takagi-Sugeno approach decomposes the input-output data space into subspaces and then approximates the system in each subspace. A powerful method for subsystem determination is fuzzy clustering. The main drawback of the fuzzy clustering algorithms is the lack of initialization procedure. There is no reliable technique for determining the number of clusters and initial data partition in order to avoid the local minima in solving the clustering task. In this paper a Hough transform initiation procedure for finding the number of clusters and a good starting point in the clustering optimization procedure is implemented. In this way the efficiency of overall estimation procedure is increased and the accuracy of the obtained model is improved.