(403e) New Liquid Flammability Rating with the Assist of Machine Learning

Currently, the classification and rating of liquid's flammability is based on its flashpoint. However, flashpoint becomes insignificant when the liquid is present in the form of aerosol. This study uses DIPPR 801 database to find a new liquid flammability rating with the consideration of the possibility of atomization. Over 800 organic chemicals have been studied and principle component analysis is applied. Also, several unsupervised learning algorithms have been utilized to conduct the clustering on the database. The new liquid's flammability rating standard will extend the existing liquid flammability standards, such as NFPA 704 and GHS of Classification and Labelling of Chemicals.