Fuzzy neural network for nonlinear-systems model identification

This paper presents a model identification approach of nonlinear systems where only the input-output data of the identified system are available. To automatically acquire the fuzzy rule-base and the initial parameters of the fuzzy model, an unsupervised clustering method is used in structure identification. Based on the cluster result, a fuzzy neural network (FNN) is constructed to match with it. The FNN is trained by its learning algorithm to obtain a precise fuzzy model and realize parameter identification. The network has universal approximation capability, a property very useful in, e.g. modeling and control application. Finally, the effectiveness of the proposed technique is confirmed by the simulation results of two nonlinear systems.