Prediction of Surface Tension of Various Aqueous Amine Solutions Using the UNIFAC Model and Artificial Neural Networks

In the first stage of this work, a thermodynamic approach based on the equality of chemical potentials together with the original UNIFAC activity coefficient model was used to predict the surface tension of 11 different aqueous amine solutions over the entire concentration range at the temperature range of 293.15-348.15 K. The considered aqueous amine solutions include one of the following amines: monoethanolamine, diethanolamine, di-isopropanolamine, methyl diethanolamine, 1-amino-2-propanol, triethanolamine, 3-dimethylamino-l-propylamine, 3-amino-1-propanol, dimethylethanolamine, 2-methylamino ethanol, and 2-ethylamino ethanol. The effect of temperature, amine concentration in the aqueous solution, and hydrophobic groups such as -CH2 and -CH(3 )on the surface tension of primary, secondary, and tertiary amines was investigated. In the second stage, the multilayer perceptron neural network model is designed to predict the surface tension of the aqueous solutions from two different classes of independent variables. The UNIFAC and two developed multilayer perceptron models were used to predict 650 surface tension values, which were in turn used to compare to experimental datapoints with the average absolute relative deviations of 6.79, 1.17, and 1.12%, respectively.