Prediction of flash points of pure organic compounds: Evaluation of the DIPPR database

The flash point is one of the most important flammability properties of compounds for the design of inherently safe processes. Many models have been developed to predict the flash point using the DIPPR database. However, for only 740 of the 1628 organic compounds available in the DIPPR database, the data for both flash point and normal boiling point were experimentally determined. For the other compounds, at least one of these properties was predicted and therefore is not appropriate for model development. The present study introduces a model to predict the flash points of pure organic compounds using their molecular structures and normal boiling points. The new model exploits the equality of the relative errors observed for the normal boiling point and flash point values predicted using the Joback method. Consequently, the relative error of the predicted normal boiling points can be used as a scaling factor to modify the predicted flash points. The ability of the model to evaluate the accuracy of a database was investigated. The ratio of the relative error of the predicted flash point to the relative error of the predicted normal boiling point obtained using the Joback method was proposed as a measure to evaluate the accuracy of flash point data.