Machine learning modeling of XGBoost and SVR for predicting rice husk dust ignition energy based on crushed brown rice content

Abstract

There exists the risk of dust explosion during the rice husk dust production and transportation. It is of great significance to clarify the parameter of the minimum ignition energy (MIE) of the dust cloud for reducing the dust explosion. This study explored the influence of the inclusion proportion of crushed brown rice (abbreviated as crushed brown rice) on the MIE of rice husk dust cloud. It was found that with the incorporation of crushed brown rice, the MIE of rice husk dust showed an increasing trend. However, a small amount of crushed brown rice can improve the dispersion of dust clouds, thereby reducing their ignition energy and increasing the risk of dust explosions. Moreover, two machine learning modeling methods, namely the Extreme Gradient Boosting algorithm (XGBoost) and the Support Vector Regression algorithm (SVR), were utilized to model the MIE of the rice husk. The grid search was employed to automatically optimize the hyperparameters. In the optimal state of the model, the R² value of the SVR model, which is 0.913, is higher than that of the XGBoost model, which is 0.883, and the prediction error is lower at the same time, thus demonstrating that the SVR model is more suitable for the study of the MIE of the rice husk. Grain dust processing and transportation should therefore take these findings into consideration.