论文部分内容阅读
基于定量构效关系(QSPR)原理,对有机过氧化物的自加速分解温度(SADT)与其分子结构间的内在联系进行了研究。应用CODESSA软件计算46种有机过氧化物的分子描述符。采用最佳多元线性回归(B-MLR)法对描述符进行筛选获得7个特征描述符,并同时建立线性回归模型。将7个特征描述符作为输入参数,采用支持向量机(SVM)法建立非线性模型。随后对模型进行验证,结果表明,B-MLR模型和SVM模型均具有良好的拟合能力、稳定性和预测能力,且SVM模型的性能(R_(train)~2=0.958,R_(test)~2=0.862)优于B-MLR模型(R_(train)~2=0.930,R_(test)~2=0.844)。通过对特征描述符的分析发现了影响SADT的主要结构因素。采用Williams图分析了模型的应用域,结果表明所有样本均在模型的应用域范围内。本文所建立的B-MLR模型和SVM模型可应用于有机过氧化物SADT的预测。
Based on the QSPR principle, the intrinsic relationship between self-accelerating decomposition temperature (SADT) of organic peroxides and its molecular structure was studied. CODESSA software was used to calculate 46 molecular descriptors of organic peroxides. The descriptors were screened by the best multivariate linear regression (B-MLR) method to get seven feature descriptors, and at the same time establish a linear regression model. Seven feature descriptors are used as input parameters, and a support vector machine (SVM) method is used to establish the nonlinear model. Then the model is validated. The results show that both the B-MLR model and the SVM model have good fitting ability, stability and predictive ability, and the performance of the SVM model (R train ~ 2 = 0.958, R test ~ 2 = 0.862) was superior to the B-MLR model (R train ~ 2 = 0.930, R test ~ 2 = 0.844). Through the analysis of feature descriptors, the main structural factors affecting SADT were found out. The Williams diagram is used to analyze the application domain of the model, and the results show that all the samples are within the application domain of the model. The B-MLR model and SVM model established in this paper can be applied to the prediction of organic peroxide SADT.