Experimental study on separation of isopropanol-water azeotrope by packed capillary distillation and the prediction model of artificial neural network

Capillary distillation is an azeotropic distillation technology, which combines the mechanism of surface adsorption, capillary action, as well as the distillation process. Different variations of activity coefficient and saturated vapor pressure of azeotropic components occur in the capillary pore structure, which leads to the separation. In this paper, 5A, 4A, and 3A molecular sieve capillary porous media were selected as capillary packing to study the separation effect on azeotropic mixture of isopropanol-water solution experimentally. It was found that 4A molecular sieve worked better, the best reflux ratio is 6, and the best packing height is 0.4 m. Based on the abovementioned experimental results, a back propagation (BP) neural network is developed and utilized in the separation process. The effects of parameters such as the composition of feedstock solution, reflux ratio, and packing height on the concentration of isopropyl alcohol at the top of packed capillary distillation column were predicted, and the accuracy reached 96.7%. The proposed prediction method is helpful for developing the technology of separating azeotropic solution by capillary distillation with packing suitable for industrial application.