AI-Based Geometry Recognition for Evaluating the Feasibility of Intensified Reaction and Separation Systems

Abstract

Reactive distillation (RD) provides notable advantages over conventional processes, regarding reduced energy requirements and CO 2 emissions. However, as the benefits of RD may not be universally applicable, a comprehensive feasibility assessment is necessary. This study introduced an automated feasibility evaluation procedure for an RD column using an AI-based region recognition approach, reducing the reliance on expert knowledge and heuristics in graphical methods. Through k -means clustering-based image segmentation, topological information on the reaction and separation reachable region was extracted from ternary diagram landscapes. Subsequently, the extracted information was integrated into tray-by-tray calculations to automate the evaluation. This geometric calculation procedure was applied to assess the feasibility of RD columns with different types of reactions. The feasibility results were obtained within seconds, demonstrating the efficiency of the proposed approach. Furthermore, case studies validated the feasibility of the evaluation results for three practical examples using rigorous simulations, confirming its reliability and applicability.

Energy-Efficient Design Sequences for the Purification of Styrene Monomer from the Pyrolysis Oil of Waste Polystyrene

“…The hindrance in achieving a high-purity SM product through distillation arises not only from the closely aligned boiling points of components but also from the formation of an azeotrope among these components. , A prior study reported that an azeotrope does not form between alpha-methylstyrene (AMS) and SM under vacuum pressure conditions, but an azeotrope can be formed at atmospheric pressure . In this study, we employed the UNIQUAC thermodynamic model, which is suitable for predicting the behavior of aromatic components in Aspen Plus, to investigate the formation of azeotropes among all components across the whole range of pressures (refer to the binary interaction parameters in Table S2).…”