Thermal conductivity estimation of nitrogen-containing liquid organic compounds using QSPR methods from molecular structures

Abstract Thermal conductivity is an essential thermodynamic property. Therefore, the development of method for predicting thermal conductivity is utmost importance. In this study, 142 thermal conductivity data were collected which was divided into training set, validation set and prediction set. The structures of compounds were optimized in Gaussian 09W and the molecular descriptors were extracted by the Dragon software. And then we developed a new quantitative structure-property relationship (QSPR) linear model with 6 parameters for predicting the thermal conductivity of nitrogen-containing liquid organic compounds using multiple linear regression method (MLR). The squared correlation coefficient of training set, validation set and prediction set were 0.9685, 0.9808 and 0.9713, respectively. And the average absolute relative deviations (AARD) of each subset were 5.52%, 8.78% and 8.56%, respectively. In addition, the applicability domain of the developed model was analysed with Williams plot. This work can provide guidance for calculating the thermal conductivity of nitrogen-containing liquid organic compounds.