Experimental and Machine Learning Investigation of VLE/VLLE Behavior in the Methyl Acetate-Methanol System Using ChCl: EG Deep Eutectic Solvent

초록

The separation of azeotropic mixtures, such as methyl acetate and methanol remain a major challenge in chemical process design due to strong molecular interactions and low relative volatilities. In this study, a green and efficient deep eutectic solvent (DES) based on choline chloride (ChCl) and ethylene glycol (EG) (1:3) was introduced as a sustainable entrainer to disrupt the azeotrope at 323.15 K. Vapor-liquid equilibrium (VLE) and vapor-liquid-liquid equilibrium (VLLE) experiments revealed complete azeotrope elimination at only 0.4 wt% DES loading, confirming its superior selectivity and phase-splitting capability. Beyond the experimental findings, a Physics-Guided Artificial Neural Network-Gaussian Process Regression (PI-ANN + GPR) hybrid framework was developed to predict key thermodynamic properties. The hybrid model outperformed conventional regression methods by achieving near-perfect parity and statistically calibrated uncertainty bounds (PICP90 approximate to 0.90). This integrated experimental-computational approach not only validates ChCl: EG as a viable green entrainer but also demonstrates a scalable machine learning route for data-efficient, uncertainty-aware design of DES-assisted separations, establishing a new benchmark for reliability-driven solvent screening.

키워드