CONTROLLING CHAOTIC CONVECTION USING NEURAL NETS - THEORY AND EXPERIMENTS

An exploratory study is conducted to assess the feasibility of using n eural networks to control flow patterns and to evaluate the performanc e of these controllers. Neural networks were used to control (suppress ) chaotic convection both in experiments and in a theoretical model of a thermal convection loop. It is demonstrated that the neural network controller can successfully cause the flow to behave in a desired way . The performance of the neural network controllers was compared with that of previously used conventional linear proportional controllers. (C) 1998 Elsevier Science Ltd. All rights reserved.