An SVR-based Machine Learning Model Depicting the Propagation of Gas Explosion Disaster Hazards

Abstract

Shock wave pressure, high temperature flames, and toxic gases are among the factors that cause mine ventilation system failure following a gas explosion. Herein, we propose a Support Vector Regression (SVR)-based machine learning model to quickly determine the propagation of these disaster-causing hazards throughout the whole ventilation network. FLACS was used to simulate the explosions in a straight roadway with different spatial geometric parameters and gas explosion equivalents. Four SVR machine learning models were constructed by incorporating the roadway’s cross-sectional area, length of gas filling, gas concentration, and the distance between the observation point and the explosion source as inputs, while the shock wave pressure, flame temperature, time to the maximum temperature, and pressure served as outputs. The proposed model can quickly and accurately predict the propagation of disaster-causing hazards for a given explosion position and equivalent. As such, it plays a significant role in determining the ventilation system failure mode and aids decision makers and rescuers in developing a rescue and refuge plan.