Generalized predictive control using a neuro-fuzzy model

Generalized predictive control (GPC) is a well-established control technique for linear systems and can be used to control a wide range of processes. The problem is that the process must be describable by a linens model. This paper reports the authors' recent work on an adaptive fuzzy predictive control (AFPC) scheme that combines the advantages of predictive control and fuzzy control. The idea is that fuzzy set theory can be applied to describe the process in such a way as to deal with its uncertainty and the ill-defined process variables, and the controller can be designed by using a mature predictive control scheme. In particular, a radial basis function (RBF) network is adopted to accommodate the rule-based fuzzy model of the process, which, initially empty, is gradually constructed by using the unsupervised variable learning algorithm. The GPC criterion function is used to carry out the controller design. The resulting control scheme is demonstrated in the paper with application to the iron-ore sintering process.