Deep neural network-based modeling and control of a transesterification reactor

Abstract

Deep learning approaches have been the subject of considerable interest and challenge in various fields. In this work, we propose a deep learning methodology for automatic modeling and control applications based on an Internal Model Control (IMC) structure. Our study focuses on improving the modeling and control of a transesterification reactor using deep neural network approaches. To emulate the direct dynamics of the system, a feedforward deep multilayer perceptron (MLP) neural network was trained and then used as a neural controller within the IMC framework. Experimental results demonstrate that the proposed modeling approach can accurately capture the reactors nonlinear behavior, while the deep learning-based controller ensures robust and efficient tracking of desired reference trajectories under various operating conditions. Comparative analysis with conventional control techniques highlights the superiority of the proposed method and shows that deep neural networks, when integrated with an IMC strategy, offer a promising framework for the advanced control of complex chemical processes.