LAUSR

Abstract Despite the long-held misconception surrounding pure ionic liquids (ILs), recent experimental investigations have unequivocally established that several ILs could boil at temperatures sufficiently low to avoid the risk of decomposition. Nonetheless, the experimental complexity combined with the high costs incurred for the measurement of the saturation pressure of ILs, with typical values below 0.1 (Pa), make the development of predictive models for this important thermophysical property of great importance. In the current study, utilizing 325 experimental vapor pressure data points belonging to 26 ILs with molecular weights spanning 205.265–629.763 (g·mol − 1 ), a robust multilayer perceptron (MLP) feedforward artificial neural network (ANN) has been constructed. The weights and biases of the proposed network, which is comprised of only one hidden layer with nine neurons, have been optimized using the Bayesian regularization approach. The input parameters of the ANN include the critical temperature, critical pressure, acentric factor, as well the mass connectivity index of the IL of interest, which are easily calculable using a modified Lydersen–Joback–Reid group contribution method. In addition, comprehensive comparisons have been made with two thermodynamically rigorous treatments for the prediction and correlation of the vapor pressure of pure substances, namely the zero-pressure liquid fugacity approach as well as the revised isofugacity criterion. According to the results obtained, the proposed ANN can satisfactorily be employed for the prediction of the saturation pressure of pure ILs with an overall AARD below 1%.