Use of the DIPPR database for development of QSPR correlations: Normal boiling point

Tabulation of evaluated physical property constants and of their estimated uncertainties makes the DIPPR database a valuable tool for developing correlations for physical properties of pure fluids. In this study, we have used the DIPPR database to develop a group-contribution method for the normal boiling point (NBP) of pure compounds. The resultant correlation uses the molecular descriptors of molecular weight and van der Waals volume in addition to Domalski-Hearing (DH) second-order group definitions. A training set of 1141 evaluated normal boiling points of >95% accuracy was selected from the database and correlated. The average absolute deviation (AAD) was 7.8 K (1.9%) with zero bias. Estimations of NBP for a test set of 384 compounds not used in the regression gave an AAD of 13.0 K (2.7%). The results suggest that the method is comparable in accuracy to the best methods available for estimating NBP, but it has the convenient feature of DH group designations that are immediately compatible with currently available DH algorithms and software.