Illuminating Flash Point: Comprehensive Prediction Models

Tu Le, Mathew Ballard, Phillip Casey, Ming Liu, David Alan Winkler

    Research output: Contribution to journalArticlepeer-review

    7 Citations (Scopus)


    Flash point is an important property of chemical compounds that is widely used to evaluate flammability hazard. However, there is often a significant gap between the demand for experimental flash point data and their availability. Furthermore, the determination of flash point is difficult and costly, particularly for some toxic, explosive, or radioactive compounds. The development of a reliable and widely applicable method to predict flash point is therefore essential. In this paper, the construction of a quantitative structure - property relationship model with excellent performance and domain of applicability is reported. It uses the largest data set to date of 9399 chemically diverse compounds, with flash point spanning from less than -130°C to over 900°C. The model employs only computed parameters, eliminating the need for experimental data that some earlier computational models required. The model allows accurate prediction of flash point for a broad range of compounds that are unavailable or not yet synthesized. This single model with a very broad range of chemical and flash point applicability will allow accurate predictions of this important property to be made for a broad range of new materials.

    Original languageEnglish
    Pages (from-to)18-27
    Number of pages10
    JournalMolecular Informatics
    Issue number1
    Publication statusPublished - Jan 2015


    • Domain of applicability
    • Flash point
    • Neural network
    • Quantitative structure-property relationship
    • Robust model prediction


    Dive into the research topics of 'Illuminating Flash Point: Comprehensive Prediction Models'. Together they form a unique fingerprint.

    Cite this