Fire safety assessment models based on machine learning methods for the coal industry

Abstract

One of the major coal industry problems is the auto-oxidation of coal, which leads to fires and human-induced disasters. This paper aims to use machine learning techniques and the most effective algorithm for preventing fires in mines. Two machine learning techniques, the naive Bayes classifier and support vector machines (SVMs), were employed to achieve the objective. The algorithm was developed based on the dependency of the indicating gas amount on the coal temperature. The accuracy of the techniques was assessed using the non-conformity matrix and related parameters. The data collected was processed using Microsoft Office. It was discovered that before the appearance of an open fire, there were two other temperature intervals without combustion and smoldering, demonstrating a clear dependence of smoke and carbon monoxide concentrations on the coal temperature. This technique is excellent for forecasting the first step of combustion, which is confirmed by the corresponding parameter values. The results demonstrated that the methodology was excellent in laboratory studies. The algorithm proposed in this paper can be used as a new method of predicting fires in coal mines and elaborating a branch of machine learning to establish more effective ways for detecting combustion in its early stages.