A comparative study between LS-SVM method and semi empirical equations for modeling the solubility of different solutes in supercritical carbon dioxide

A genetic algorithm based least square support vector machine has been used to predict the solubility of 25 different solutes in supercritical carbon dioxide. This model consists of three inputs including temperature, pressure and density of supercritical carbon dioxide and a single output which is the solubility of different solutes in supercritical carbon dioxide. The model predictions were compared with the outputs of seven well-known semi empirical correlations. Results showed that the present method has an average relative deviation of about 4.92% for 25 solutes while the best semi empirical equation resulted an average relative deviation of about 13.60% for same solutes.

Graphical abstractFigure optionsDownload full-size imageDownload high-quality image (103 K)Download as PowerPoint slideHighlights► Solubility of solutes in supercritical carbon dioxide has been studied. ► Semi empirical equations and support vector machine has been used for solubility estimation. ► Results showed support vector machine is more accurate than semi empirical equations. ► Support vector machine has an average relative deviation of about 4.92% for 25 solutes while the best semi empirical equation resulted an average relative deviation of about 13.60% for same solutes.