Prediction of auto-ignition temperatures for binary liquid mixtures based on electro-topological state indices

The auto-ignition temperature (AIT) values of 168 sets of binary flammable liquid mixtures composed of different components and volume ratios were measured by AITTA 551 auto-ignition temperature tester. The mixed electro-topological state indices (ETSI) values of different atom types were calculated. The modified particle swarm optimization (MPSO) algorithm with exponential decreasing inertia weight (EDIW) was applied to optimize the support vector machine (SVM) hyper-parameters and MPSO-SVM prediction model was established. The model was employed in research for predicting the AIT of mixtures according to the mixed ETSI values of different atom types. The results showed that it could effectively predict the AIT of binary liquid mixtures based on electro-topological state indices. The squared correlation coefficient (R²) and average absolute error (AAE) of MPSO-SVM model were 0.991 and 3.962 K, respectively. In terms of model generalization performance and prediction accuracy, the result of MPSO-SVM model was obviously superior to the results of multiple linear regression (MLR), grid search method (GSM-SVM), genetic algorithm (GA-SVM) and particle swarm optimization (PSO-SVM). This study provided an effective method to predict the AIT of binary liquid mixtures for engineering.