Optimizing Pressure Swing Distillation for Di-n-Propyl Ether and n-Propyl Alcohol Separation Using Aspen HYSYS and Machine Learning Algorithms

This paper presents the design and simulation of a pressure-swing distillation (PSD) process for separating and purifying di-n-propyl ether (DnPE) from n-propyl alcohol (nPA) using Aspen HYSYS software. The minimum-boiling-point azeotrope formed at atmospheric pressure makes conventional separation methods ineffective. Three critical parameters, feed stage, feed temperature, and reflux ratio, are systematically optimized to minimize energy consumption. The optimized process achieves product purities of 99.5% DnPE and 98.7% nPA while reducing energy consumption by 15% compared to conventional distillation. Additionally, an XGBoost regression model is developed to predict reboiler heat duty with 95% accuracy, further enhancing process efficiency. Particle swarm optimization is employed to identify optimal operating conditions based on the machine learning predictions. This integrated computational approach demonstrates significant improvements in separation efficiency, highlighting the industrial potential of the optimized PSD process for azeotropic mixtures.