Representation of Adsorption Data for the Isopropanol-Water System using Neural Network Techniques

Abstract

Molecular sieves and palm stone, a newly developed bio-based adsorbent, were used to break an azeotropic isopropanol-water system via an adsorptive distillation process. Equilibrium data at different inlet water contents are presented. The data were obtained with a fixed bed adsorptive distillation process using Type 3A and Type 4A molecular sieves and palm stone. An artificial neural network (ANN) technique was used to represent the isotherm equilibrium data of this azeotropic system. The ANN prediction results were compared with the Guggenheim-Anderson-de Boer (GAB) isotherm model. It was possible to break the isopropanol-water azeotrope using this separation process with the adsorbents used in this work. Water uptake increases as the water content in the feed decreases from 16?% to 10?%. Although the GAB isotherm model was found to be applicable to the water vapor sorption data on the adsorbents examined, the ANN model fitted the equilibrium data more efficiently.