Deep learning-driven green solvent design and process intensification towards isopropyl alcohol-water azeotrope system

Abstract

Extraction distillation and azeotropic distillation are two important methods for separating isopropyl alcohol (IPA) and water azeotrope. However, azeotropic distillation is generally more energy-intensive than extractive distillation separation process. Therefore, solvent design and process intensification for the extractive distillation process are the keys to addressing the problems of azeotropic separation and reducing energy consumption. In this contribution, a deep learning-based solvent high throughput screening framework was proposed to design the green solvent for separating the IPA/water mixtures. All properties, such as thermodynamic properties and EH&S properties, used for screening were predicted by the deep learning-based predictive models. From more than 10⁸ individual molecules, five green solvent candidates were screened for the separation of IPA/water azeotrope. The energy consumption analysis of 5 solvents shows that ethylene glycol as solvent has the lowest separating energy consumption. Finally, the heat integration and heat pump distillation of the extractive distillation separation process was carried out, and the energy-saving potential reached 45.86%.