A design of generalized minimum variance controllers using a GMDH-type neural network for nonlinear systems

Describes the design of a generalized minimun variance controller (GMVC) using a GMDH-type neural network for nonlinear systems. The predictive value of the output required in the GMVC law is obtained by using a group method of data handling (GMDH) which is a kind of multilayered neural network. Since the predictive value of the output in GMVC law is calculated by a nonlinear model, one can expect a better control performance than that calculated by a linear model. The behavior of the proposed control scheme is evaluated on a numerical simulation example.