Determination of boiling points of azeotropic mixtures using quantitative structure-property relationship (QSPR) strategy

Azeotropes, which are solutions that contain two or more chemicals, are very important in industry. Experimental techniques as well as theoretical approaches such as ab initio have been developed for estimating mixture properties and phase equilibrium data. Both approaches are accurate and effective, but are costly and time-consuming. The quantitative structure-property relationship (QSPR) method, which is efficient and extremely fast, could be a viable alternative approach. In this work, we developed QSPR models for prediction of boiling points (Tb) of binary azeotropes. The Tb values of azeotropic mixtures were investigated by means of multiple linear regressions (MLRs). Two different data matrixes were calculated for characterizing azeotropic mixtures based upon the centroid approximation and the weighted-contribution-factor approximation. The ant colony optimization algorithm (ACO) was employed to select relevant descriptors. For both approximations, significant QSPR models were obtained by using the ACO-MLR algorithm. The descriptors that appeared in the best MLR models are related to those properties, including mass, ability to form H-binding, numbers of heteroatom, solvation entropy, and solvation energy, that control the boiling point.