Mining Data in a Longwall Coal Mine: An Integration of Rock Mechanics & Data Mining to Predict Intersection Stability

"Studies have shown that the roof in coal mines is nearly twice as likely to fail at intersections as at entries due to enlarged roof spans, stress redistribution and other factors. However, the relative stability of intersections in a mine varies, and an improved understanding of what factors impact roof instability can lead to efficient, proactive mining practices that both enhance stability and mining efficiency.A study was undertaken at an underground longwall coal mine to develop a model to predict the stability at the intersections of entries and cross-cuts. The statistical analyses relied upon observational and measured data regarding the development of tension cracks, seepage, intersection geometry, mining practices, geological attributes and geomechanical factors at 783 intersections. Each intersection was given a NIOSH roof. Common data mining techniques, such as multivariate linear regression, multinomial logistic regression, decision trees and probabilistic neural nets, were considered and evaluated to establish correlations and associations between stability and the other variables. Two of the more successful techniques were Decision Trees and Multinomial Logistic Regression. The analyses showed that a number of factors impacted stability:• Overburden thickness (impact on stress concentration magnitudes)• Initial opening area (impact on stress concentration magnitudes)• Sulfur content (depositional environment & impact of mechanical properties, leaching effects)• Intersection type (impact on stress concentrations)• Gob distance (impact on stress concentrations)• Presence or proximity of a particular sandstone (impact on mechanical properties)• Roof bolt type and diameter (impact on reinforcement)Some factors showed no statistically significant relation to stability. These included:• Total area• Supported areaSeepage was a more problematic variable in assess. Seepage is associated with instability, but it is not clear whether seepage occurs as a consequence of instability, for example, through the creation of tension cracks, or whether the intersection has been excavated where water already exists and has degraded the rock prior to excavation, leading to greater instability.Seepage appears related to the geology of the roof rock mudstones, not the sandstones. Tension cracks in the intersections occur preferentially where the Sulfur content is below 0.9%. The Sulfur may relate to mechanical differences in the roof rock, and where the Sulfur is low, the roof is less stable, there are more tension cracks, and seepage is greater. This suggests that the weaker mudstones may be more prone the development of tension cracks, not because these rocks are more brittle, but possibly weaker.2-way and 3-way intersections appear to have less seepage than 4-way, although very few intersections overall have any seepage. This is similar to the roof stability, in which the 2-way and 3-way are more stable. 2-way and 3-way intersections have fewer-than-expected tension cracks, while 4-way have more than expected. These results suggest that four-way intersections are weaker and more prone to tension crack development, also suggesting that increased seepage results from weaker roof conditions, rather than the converse. Thus seepage appears to be mostly a consequence of instability rather than a cause and is higher where there are 4-way intersections, mudstone geology and external loading factors like depth of cover and proximity to gob that promotes crack development. The study also indicated that a better understanding of the mudrock facies could reduce uncertainty."