In this paper

a new inlereiitial control strategy based on RBF network for a non adiabalie tubular fixed -bed reactor is presented. An on-line self learning inferential estimator based on RHF network is developed

and then a predictive control system with output of the estim nor as feedback variable is designed. It is shown via simulation

where a phthalic anhydride reactor is taken as a research plant

thai the estimator has the performance of on line self - learning ability and good accuracy. Moreover

the performance of the control system is satisfactory.