Neural networks to estimate the water content of imidazolium-based ionic liquids using their refractive indices

Abstract

A non-linear model has been developed to estimate the water content of 1-butyl-3-methylimidazolium tetrafluoroborate, 1-butyl-3-methylimidazolium methylsulfate, and 1,3-dimethylimidazolium methylsulfate ionic liquids using their respective refractive index values. The experimental values measured to design the neural network (NN) model were registered at 298.15K. These were determined at different relative humidity values which ranged from 11.1% to 84.3%. The estimated results were compared with experimental measurements of water content obtained by the Karl Fischer technique, and the differences between the real and estimated values were less than 0.06% in the internal validation process. In addition, an external validation test was developed using bibliographical references. In this case, the mean prediction error was less than 5.4%. In light of these results, the NN model shows an acceptable goodness of fit, sufficient robustness, and a more than adequate predictive capacity to estimate the water content of the ILs through the analysis of their refractive index.

Keywords: Ionic liquid; Refractive index; Relative humidity; Water content.