A grey wolf optimizer-based support vector machine for the solubility of aromatic compounds in supercritical carbon dioxide

Abstract

The prediction of solute solubility in supercritical carbon dioxide (SCCO2) is crucial for the development of supercritical applications. Many models have been developed to calculate the solubility of aromatic compounds. In this work, a grey wolf optimizer-based support vector machine (GWO-SVM) was proposed for correlating solute solubility in SCCO2. The proposed GWO-SVM model utilized the temperature, pressure and the density of SCCO2 as input parameters and the solubility of different solutes in SCCO2 as target parameter on the basis of gray correlation analysis. The new model successfully correlated solute solubility of 18 compounds (1148 data points including 814 training data points and 334 testing data points) in SCCO2, which were collected from the published literature. A comparison of the 27 commonly used empirical models and the proposed GWO-SVM model showed that the overall average absolute relative deviation of the proposed model is the lowest (3.20%). It was also found that the overall average absolute relative deviation is less dependent on material type for the proposed GWO-SVM model.