Adaptive Neurofuzzy Identification and Control

A powerful algorithm is introduced to build an adaptive Takagi-Sugeno neurofuzzy model online from zero fuzzy rules for unknown nonlinear SISO systems. The proposed technique creates the fuzzy rules and adapts the membership functions in the IF statement as well as the linear model in the THEN statements. In addition, the algorithm searches for redundant rules to be eliminated to get as less number of rules as possible. The proposed technique has been applied to nonlinear plant models commonly encountered in chemical reactors to elaborate its efficiency. An adaptive controller based on the induced neurofuzzy model is developed and applied to SISO nonlinear systems showing the efficient behavior of the proposed controller. A stability study is also included. The proposed algorithm is then extended to build neurofuzzy models for nonlinear MIMO systems in an automated manner online, based on which a control scheme is built to drive nonlinear MIMO systems to follow linear reference models. The power of the proposed technique is demonstrated though its application on a MIMO nonlinear system thus elaborating its efficiency.