Prediction of the binary surface tension of mixtures containing ionic liquids using Support Vector Machine algorithms

Abstract

The surface tension of pure ionic liquids (ILs) and their mixtures with other compounds play a key role in the design and development of many industrial processes. Therefore, its modeling is extremely important from an industrial point of view. This study examined the capability and feasibility of three intelligence algorithms for predicting the surface tension of binary systems containing ILs. To construct and test the models, 748 data points corresponding to the experimental surface tension values of binary mixtures containing ILs were extracted from the literature. The surface tension was between 0.0157 and 0.07185N·m−1. The absolute temperature (T), mole fraction and molecular weight of the IL components (xIL and MwIL) and the density of the IL components (ρIL) together with the boiling point (Tbnon-IL) and molecular weight (Mwnon-IL) of the non-IL component were considered as model input variables to differentiate between the various compounds involved in binary systems. A comparison of the experimental data and predicted values using all three methods (in terms of statistical parameters) showed good agreement; however, the CSA-LSSVM prediction was better than the other two approaches.