NEURAL NETS AND THE LOCAL PREDICTOR METHOD USED TO PREDICT THE TIME-SERIES OF CHEMICAL-REACTIONS

A feed-forward neural net (30-40 neurons) and the local predictor meth od are used to predict the time series of nonlinear chemical reactions . Knowledge of the underlying chemical mechanism is not necessary for performing the predictions. It will be demonstrated that the two metho ds are suitable to predict a variety of dynamic states and to distingu ish between deterministic chaos and statistical noise in a theoretical model (Brusselator) and in an experimental system (Belousov-Zhabotins ky (BZ) reaction). BZ chaos at low flow rates and statistical aperiodi city at high flow rates are predicted equally well by either method de spite the existing experimental noise.