Distal supervised learning control and its application to CSTR systems

Abstract

In this paper, distal supervised learning control is considered for the nonlinear continuous-stirred tank reactor (CSTR) systems. Multilayer neural networks (BP) are introduced to construct the distal supervised learning control system. The proposed controller consists of an expert coordinator and two BP networks. Extreme control mode or distal supervised learning control mode is activated by expert coordinator based on control errors. The effectiveness of the proposed controller is illustrated through an application to control acetic anhydride hydrolysis reaction in a CSTR system. Results show that the proposed distal supervised learning control is strong in self-learning and easy to realize, and helpful for improving nonlinear control performance.