Prediction of the flash points of binary biodiesel mixtures from molecular structures

Biodiesel has already been widely used as an essential renewable and alternative energy. Great attention also has been attracted to its safety issues at the same time. In this paper, a novel mixture Quantitative Structure-Property Relationship model (M-QSPR) was developed to predict the flash points of binary biodiesel mixtures only from their molecular structures. The Simplex Representation of Molecular Structure (SiRMS) descriptors were employed to represent the molecular characteristics of biodiesel mixtures. The genetic algorithm combined with multiple linear regression (GA-MLR) was used to select the most relevant SiRMS descriptors and develop the corresponding MLR model. The resulted model was a two-parameter linear equation, with the root mean square error and average absolute error of the external test set being of 4.777 and 3.768 K, respectively. The presented model was then rigorously verified by multiple validation methods, and the results showed satisfactory fitness, robustness and predictivity of the model. This study could be reasonably considered to provide a reliable method for predicting the flash points of binary biodiesel mixtures for engineering.