有机过氧化物自加速分解温度的神经网络研究  

Research on the self-accelerating decomposition temperature of organic peroxides by neural network

作　　者：秦正龙 Qin Zhenglong(School of Chemistry&Materials Science,Jiangsu Normal University,Xuzhou 221116,Jiangsu,China)

机构地区：[1]江苏师范大学化学与材料科学学院,江苏徐州221116

出　　处：《江苏师范大学学报（自然科学版）》2024年第4期47-50,共4页Journal of Jiangsu Normal University：Natural Science Edition

基　　金：江苏省高校品牌专业建设工程项目(PPZY201992)。

摘　　要：为了研究有机过氧化物自加速分解温度(SADT)与其结构之间的定量构效关系,计算了46种有机过氧化物的分子形状指数(nK)和电性距离矢量指数(M i).通过筛选优化,得到分子形状指数的2K、3K、4K及电性距离矢量指数的M 9、M 21、M 26、M 27、M 33,共8个结构参数,并将它们作为BP神经网络的输入层变量,有机过氧化物SADT作为输出层变量,采用8∶4∶1的网络结构,获得了令人满意的定量结构与性质关系(QSPR)神经网络预测模型,其总相关系数为0.997,自加速分解温度的预测值与实验值的一致性令人满意,平均误差仅为0.9℃,优于文献方法.结果表明,有机过氧化物自加速分解温度与8种结构参数之间呈现良好的非线性关系.In order to study the quantitative structure-activity relationship between the self-accelerating decomposition temperature(SADT)and the structure of organic peroxides,molecular shape index(nK)and electrical distance vector index(M i)of 46 organic peroxides were calculated and 8 structural parameters,namely 2K,3K and 4 K of the molecular shape indices,and M 9,M 21,M 26,M 27 and M 33 of the electrical distance vector indices were obtained by screening and optimization.With the eight parameters as input variables of neural network and the SADT as output variable,the 8∶4∶1 network structure was adopted and BP neural network method was used to establish a satisfying quantitative structure-property relationship(QSPR)prediction model.The total correlation coefficient was 0.997.The predicted values of SADT by the model were in satisfactory agreement with the experiment values,with an average error of only 0.9℃,which was better than literature methods.The results showed that there was a good nonlinear relationship between the self-accelerating decomposition temperature and the eight molecular structure parameters.

关 键 词：有机过氧化物 自加速分解温度 分子形状指数 电性距离矢量指数 定量结构性质相关性 神经网络 

分 类 号：X937[环境科学与工程—安全科学] O6-39[理学—化学]