Multiple-Model Identification Using ANFIS for Nonlinear Systems

This study presents a new technique for identifying nonlinear systems using multiple models. In this technique the identification structure used is ANFIS, consequent parts are performed by multiple models and the interpolation of local models is performed by the membership functions of the Takagi Sugeno fuzzy system. The identification technique uses a number of multiple model concepts to initiate the fuzzy parameters, and the learning algorithm of the ANFIS structure adjusts the membership functions so that the fuzzy multiple models properly reproduce the nonlinear system. One experimental tank of water was hand built to exemplify the technique and to compare the classic multiple model and the intelligent one.