三元互溶混合液体闪点预测研究

Prediction study for flash points of ternary miscible liquid mixtures

闪点是评价易燃液体火灾危险性重要指标。文中基于定量结构-性质相关性(QSPR)原理对三元互溶混合液体的闪点与其结构信息间的内在定量关系(M-QSPR)开展了理论研究。从结构信息角度计算混合物描述符,并采用"Kay's mixing rule"混合规则对混合液体的结构特征进行表征,应用遗传-多元线性回归(GA-MLR)算法,优化筛选出与混合体系闪点最密切的结构参数作为输入参数,分别采用多元线性回归(MLR)和支持向量机(SVM)算法建立三元互溶混合液体闪点的理论预测模型,并将其与文献已有模型比较。研究结果表明:预测模型不仅有较好的预测能力,还可揭示影响三元互溶混合液体闪点的主要结构因素及其影响规律。

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.