Development of a quantitative structure–liquid thermal conductivity relationship for pure chemical compounds

In this study, a quantitative structure-liquid thermal conductivity relationship was developed for pure mostly organic chemical compounds. An extensive data set including more than 19,000 evaluated liquid thermal conductivity data for 1635 pure mostly organic chemical compounds at various temperatures and at atmospheric pressure below the normal boiling point and at saturated pressure above the normal boiling point temperature was provided from DIPPR 801 database to develop, validate and assess the predictive capabilities of the obtained correlation. Parameters of the model comprise 20 molecular descriptors including temperature. The obtained model shows a good agreement with DIPPR 801 data. The model shows the same values of the average absolute relative deviation and root mean square error of 7.4% and 0.01 for the training set (80% of the data), the validation set (10% of the data), and the test set (10% of the data). In terms of wide range of applicability, the model shows reasonable accuracy.