Predicting the explosion limits of hydrogen-oxygen-diluent mixtures using machine learning approach

Abstract

In this paper, we present a new methodology for predicting the explosion limits of hydrogen-oxygen-diluent mixtures by using machine learning approach. Results show that the explosion limits are accurately predicted with the logistic regression (LR), decision tree (DT), random forest (RF), support vector machine (SVM), and feedforward neural network (FNN) algorithms when using the optimal hyperparameters. In terms of computational cost, the LR and DT require the lower costs, the RF requires the high training and prediction costs and the training cost of the FNN is higher due to the large number of hyperparameters. In terms of prediction accuracy, the FNN predicts the explosive/non-explosive boundary more accurately with different amounts of training data. Furthermore, the receiver operating characteristic (ROC) curve and area under curve (AUC) values further prove the superiority of the five classifiers. The result of this study provides a new method for rapidly predicting explosion limits and expects to offer potential options for predicting explosion limits for more complex hydrocarbon fuels.