A Mutual Information Based Multifeature Pattern Recognition Criterion for Identification and Optimization of Thermally Safe Operating Conditions for Single Diffusion-Controlled Reactions Occurring in Isoperibolic Liquid–Liquid Semibatch Reactors

Abstract

Prevention of thermal runaway and establishment of safe-economic operating conditions are crucial for the practical operation of single diffusion-controlled liquid–liquid reactions occurring in isoperibolic semibatch reactors. Hence, in this study, a mutual information-based multiple feature recognition criterion (MI-MFR) was innovatively proposed, for which the maximum relevance minimum redundancy approach was introduced in feature screening; in combination with the support vector machine algorithm, various thermal behaviors can be effectively recognized. On this basis, a series of boundary diagrams were established accordingly, the impact of the variation in model parameters on which was systematically explored. Subsequently, an improved generalized thermal safety region was developed by introducing the upper thresholds in normal and cooling failure scenarios. Meanwhile, the robustness and progressiveness of the established MI-MFR criterion were confirmed by comparison with other conventional criteria. Finally, MI-MFR-based procedures were constructed to identify thermal behaviors and optimize safe-economic operating conditions in practices, based on which the optimum operating conditions of toluene mononitration were obtained.