Prediction of lower flammability limits of hydrocarbons based on quantitative structure-property relationship

The quantitative relationships between the lower flammability limits (LFL) and the molecular structures of hydrocarbon compounds were investigated based on the quantitative structure-property relationship (QSPR) studies. Various structure parameters were calculated to describe the structure characteristics of the molecules based on their structures. A set of structure parameters having significant contribution to the LFL were chosen as the molecular descriptors by employing the variable selection method of genetic algorithm (GA). Both the multiple linear regression (MLR) and support vector machine (SVM) were employed to model the possible quantitative relationship existed between these selected descriptors and LFL, respectively, and the corresponding prediction models for the LFL of hydrocarbons were constructed based on the molecular structures. The models were tested by internal and external validations. The results show that, for both models, the predicted LFL values agree well with the experimental ones, and the predicted errors are within the range of the experimental error of LFL measurements. The mean absolute error and the root mean square error for the test set of the SVM model are 0.036% and 0.046%, respectively, which are better than those of the MLR model and previous models. This paper provides a new method for predicting LFL of hydrocarbons for engineering.