Application an Artificial Neural Network for Prediction of Substances Solubility

Abstract

This paper presents an application of the artificial neural network methodology to prediction of solubilities of 1-1 electrolytes in nonaqueous solvents and solvent mixtures, using experimental data available in the literature. It is demonstrated that that the fundamental expressions proposed previously to describe correlations of solubility with physical-chemical properties of solvents, as well as common regression equations, exhibit large deviations and are not suitable for the description and prediction of solubility for a wide range of individual and mixed solvents. In comparison, the radial basis function artificial neural network algorithm is capable of reproducing the solubilities of such common salts as NaI, CsClO <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</inf> , NaCl and NaBr in a variety of nonaqueous solvents and solvent mixtures. Having used a training set to obtain the fitting coefficients, we are able to calculate accurately the solubilities of the 1-1 electrolytes in other mixtures of nonaqueous solvents. The reported results make it possible to predict solubilities of 1-1 electrolytes in mixed solvents without the need for additional experimental measurements.