Self-tuning neuro-fuzzy generalized minimum variance controller

The development of a Self-Tuning Neuro-Fuzzy Generalized Minimum Variance (GMV) controller is described. It uses fuzzy expert knowledge of the dynamic weightings to meet desired closed-loop stability and performance requirements. The controller is formulated in a polynomial system approach mixed with a Neuro-Fuzzy model and Fuzzy Self-Tuning mechanism. The proposed method is applied to a model of the Continuous Stirred Tank Reactor with Cooling Jacket and is compared with a PI controller, GMV controller with the correct model and a Fuzzy-PI controller. Simulation results are presented to demonstrate the performance of the proposed method.