用分子结构有限元分析方法估算多类芳烃的液体导热率  

Estimation of Thermal Conductivity of Liquid Aromatic Hydrocarbons Using Finite Element Analysis Method in Molecular Structure

作　　者：仇明华[1,2] 刘万强[1,2] 刘凤萍[1] 陈冠凡[1] 岳明[1] 

机构地区：[1]湖南科技大学化学化工学院,湖南湘潭411201 [2]理论有机化学与功能分子教育部重点实验室,湖南湘潭411201

出　　处：《过程工程学报》2018年第1期196-201,共6页The Chinese Journal of Process Engineering

基　　金：国家自然科学基金项目(编号:21472040);湖南省教育厅科学研究项目(编号:13C302)

摘　　要：通过对芳烃分子3D结构的有限元分析,建立了分子结构的刚度矩阵和振动方程,用MATLAB软件计算得到相应分子结构的固有频率和矩阵特征值,用多元回归分析方法建立了液体芳烃分子导热率的QSPR模型,用线性规划法对模型指数进行优化、均值化和差值化处理.结果表明,301个液体芳烃导热率实验数据的训练集实验值与计算值的相关系数高达0.979,平均绝对误差小于0.0024 W/（m？K）,平均相对绝对误差小于2.02%;22个液体芳烃导热率实验数据的测试集实验值和预测值的相关系数为0.978,平均绝对偏差小于0.0028 W/（m？K）,平均相对绝对偏差小于2.99%.The 3D molecular structure of aromatic hydrocarbon was analyzed by finite element analysis to establish the stiffness matrix and vibration equation, from which nature frequency and stiffness matrix eigenvalues were calculated with MATLAB software, then QSPR model for the conductivity of liquid aromatic hydrocarbons was developed by multivariate regression method, and further performing optimization, mean value and D-value treatment on the model. The results showed that for training set（301 liquid aromatic hydrocarbons）, the correlation coefficient between the results calculated by QSPR model and the experimented data was 0.979, the mean absolute error was less than 0.002 4 W/（m×K）, the relative error was less than 2.02%, for test set（22 liquid aromatic hydrocarbons）, the correlation coefficients was 0.978, the mean absolute error was less than 0.002 8 W/（m×K）, the relative error was less than 2.99%.

关 键 词：热导率 有机物 芳烃 分子结构 有限元分析 

分 类 号：O642.1[理学—物理化学] TQ013.1[理学—化学]