Modified fuzzy model identification clustering algorithm for liquid level process

In this paper the problem of nonlinear system identification is investigated from a new point of view. If the nonlinear system is affected by measurement noise and if the noise cluster is arbitrarily far away, then there is no way to guarantee that any clustering algorithm will select the best cluster instead of the bad one. The proposed methodology is based to adding a noise cluster to clustering algorithm. The proposed approach allows the identification of the premise parameters and the consequence parameters together via iterative minimization using four criteria. This new technique is demonstrated by means of the identification of liquid level process.