Machine Learning Approach to Implement Mine Fire Predicting for Underground Coal Mines

Mine fire due to auto oxidation of coal leading to spontaneous combustion of coal is the leading unsolved problem worldwide. Though the exact reasons of self-heating of coal are many and unclear, the accumulated heat generated in exothermic reaction due to continuous interaction of coal with the atmospheric oxygen is the major reason of triggering mine fire. Since the chemical composition of coal is not yet fully recognized, therefore, the exact mechanism of the coal-oxygen reaction in the atmosphere in presence of water, vapour and moisture is not completely established. However, irrespective of the reason, the outcome of spontaneous fires in the coal mines is the liberation of huge quantity of poisonous and combustible gases causing heavy loss of miners as well as mining machinery in addition to loss of valuable coal reserve; most disastrous is the fact that in most of the cases of outbreak of mine-fire, it is observed that occurrence of such events leaves rarely any scope of scape causing heavy loss of miners as well as machineries. Over the last decade, researchers all over world have shown tremendous interest in Machine Learning (ML) techniques – a branch of Artificial Intelligence (AI) that have demonstrated their success in providing trustworthy solutions in a variety of real-world applications to autonomously learn the input-output characteristics. Though there are variety of methods in use to predict spontaneous heating of coal leading to catastrophic mine-fire, in this study, we propose a novel approach of integrating machine learning technique with Graham’s ratio data to predict the possibility of occurrence event of mine-fire from various environmental parameters responsible for triggering spontaneous heating of coal, so that an early warning system can be designed to take precautionary steps in advance. Our simulation study establishes validity of the proposed approach.