三元互溶混合液体闪点预测研究

Flash point is a major property to evaluate the flammability hazard of liquid mixtures. The inherent quantitative relationship between the flash points and the structures of ternary miscible flammable liquid mixtures datasets (M-QSPR) was studied based on quantitative structure-property relationship (QSPR) principles in this article.The Genetic Algorithm method combined with Multiple Linear Regression (GA-MLR) was used to screen out a set of structural parameters which makes significant contribution to the flash points of the relevant liquid mixtures. The new flash point forecasting models of the corresponding liquid mixtures by Multiple Linear Regression (MLR) and Support Vector Machine (SVM) from the perspective of structural information based on Kay's mixing rule were built. The researches indicate that the prediction model which has better prediction ability, can be used to study the impact of major structure of ternary miscible mixed liquid on flash point.