Assessment of relative impacts of various geo-mining factors on methane dispersion for safety in gassy underground coal mines: an artificial neural networks approach

Abstract

Dispersing methane to a safer level is crucial for mines safety as methane has been the greatest contributor of explosion hazard in underground coal mines worldwide. Methane dispersion is affected by several geo-mining factors. This study is first of its kind, which makes an attempt to develop a model for predicting methane concentration in underground coal mines based on seven different geo-mining factors using multi-layered artificial neural networks. The main objective is to quantify the relative influences of these factors on methane dispersion in underground coal mines and identify the significant factors through sensitivity analysis. Three different architectures of neural networks were trained using the methane dispersion data generated through computational fluid dynamics simulations conducted at varied geo-mining conditions. Principal component analysis on the input set was done for dimensionality reduction, which reduced the number of variables to seven from eight while maintaining a variance of 99%. All the models performed very well, and the best model yielded mean square error of 0.0304 and R2 of 0.942. The study unveiled some new facts on the relative effects of ventilation type and surface roughness on methane dispersion. It established that air velocity is the most significant and surface roughness of mine galley is the least significant factor affecting methane dispersion in underground coal mines with relative importance of 0.25 and 0.01, respectively. The outcome of this study will be useful in design of mine ventilation system for effective coal mine methane management and enhancing mines safety.