Application of ANN in predicting VLE data of CO2-Aqueous-Alkanolamine System

Abstract

The removal of acid gases from gas streams by using suitable solvent like alkanolamine, commonly referred to as gas sweetening, is a technology that has been in use industrially for over half a century. For the rational design of gas treating processes the equilibrium solubility of acid gases like CO2 and H2S over alkanolamines (vapour-liquid equilibrium of the acid gases over alkanolamines) are essential besides the mass transfer and chemical kinetics. Representation of the experimental data with model is required, so that one can systematically correlate and predict the vapour-liquid equilibria (VLE) of these systems. In this work artificial neural network (ANN) has been used to predict the equilibrium solubility of CO2 over the alkanolamine solventslike mono-ethanolamine(MEA), di-ethanolamine (DEA) and Piperazine (PZ) instead of using any rigorous thermodynamic model. A feed forward network with back propagation and a radial basis network have been used here in an effort to predict the VLE data of CO2-MEA-water, CO2- DEA-water and CO2-Piperazine-water systems with a comparable accuracy to those predictions based on rigorous thermodynamic model. It has been found that the predictions are withinaccuracy of ±5%.