Prediction of the lower flammability limit percent in air of pure compounds from their molecular structures

• LFL is predicted from the molecular structure of the compound alone
• ANN-SGC method predicts the LFL with a correlation coefficient of 0.9998
• Better definition of atom-type molecular groups is presented
• The method is better than others in terms of simplicity, accuracy and generality

A theoretical method is presented for predicting the lower flammability limit (LFL) volume % in air of pure compounds. Artificial neural networks were used to investigate several structural group contribution (SGC) methods available in the literature. The networks were used to probe the structural groups that have significant contribution to the LFL of pure compounds and arrive at the set of groups that can best represent LFL for about 543 substances. The 30 atomic structural groups proposed can predict the LFL of pure compounds from the knowledge of the molecular structure alone, with an average error of 0.02 volume % and a correlation coefficient of 0.9998. The results are further compared to other methods in the literature, and shown to be far more accurate.