Prediction of spontaneous coal combustion tendency using multinomial logistic regression

Abstract

Spontaneous combustion of coal is a complex underground mining disaster, which mainly threats mine safety and efficiency. Several factors usually cause spontaneous combustion of coal, such as gas concentration, ventilation and coal properties. In this study, spontaneous combustion tendencies of coal mines were predicted considering the effective parameters for an underground coal mine in Turkey. Multinomial logistic regression, a multivariate statistical technique, was applied. Gas concentrations (CH4, CO, O2) and air velocity were defined as factors affecting spontaneous coal combustion. Fire hazard levels of the coal mines were determined as ‘normal situation’ and ‘potential combustion’. It was observed that CH4 and CO variables and CH4 × CO interaction were effective in the formation of clusters. The results indicate that Mine I is more liable to spontaneous combustion than Mine II and Mine III. At the same time, the effects of variations in factors are examined in the study.