A new method for predicting the upper flammability limits of fuel mixtures

The upper flammability limit (UFL) of fuel mixtures is of great importance for safety considerations in engineering. This study proposed a quantitative structure-property relationship (QSPR) model to predict the UFL of fuel mixtures from only their molecular structures. Mixtures descriptors were calculated by mole weighted average of the molecular descriptors of pure chemicals. A set of optimal mixture descriptors that significantly contribute to the UFL of fuel mixtures were determined by performing the genetic algorithm along with multiple linear regression (GA-MLR). The resulted QSPR model is a multilinear equation with four variables. After performing the rigorous model validations using multiple strategies, the results demonstrated the satisfactory robustness, goodness of fit, and predictive capability of the developed model. Also, the applicability domain (AD) of the model was defined. The proposed model can be reasonably expected to reliably predict the UFL of fuel mixtures and provide some guidance for prioritizing design of safer fuel mixtures with desired properties.