Rock Pillars are important structural elements that provide temporary or permanent support for overburden formations or seams in underground hard rock mines. Unstable pillars can result in rock sloughing, roof collapse, and even rock burst; hence its prediction is of great importance. This paper presents a novel application of Logistic Model Tree (LMT) to predict underground rock pillar stability. Five parameters —Pillar width, Pillar height, Ratio of pillar width to height, Uniaxial compressive stress of rock, Average pillar stress— are adopted to construct tree model. The LogitBoost algorithm is employed to learn from database of rock pillars case histories from six hard rock mines in Canada, South Africa and Sweden. Finally, the constructed model is validated with 10fold cross-validation and with another new group of case histories that have not been employed during training work. Results suggest that the Accuracy value of proposed LMT is the highest among the methods employed in the previous literature, which could be a guidance on the underground rock pillar stability prediction in the future.
Pillars are important structural elements that exist in rock masses during underground excavation. Without extra support and reinforcement, those remaining portions of rock masses can provide temporary or permanent support for overburden formations or seams, and endure continuous stress redistribution and concentration induced by tunnelling and mining work (Brady and Brown 1985; Zhou et al. 2011; Zhou et al. 2015). To realize the maximum extraction rate of mineral resources and sustainable mining, underground pillars must be confined to a certain range of size (width and height) while still keep stable for safe working condition throughout the entire life. Unstable pillars can result in rock sloughing, lead to collapse of roof, and even cause rock burst when the pillar bumps with large amount of accumulated elastic energy release (Dou et al. 2009; Zhou et al. 2015). With gradual increase of excavation depth, pillar instability and failure events or accidents occur frequently due to ambient stress or more complex structural stress in deep mining. It is therefore expected that new multi-disciplinary methods to estimate rock pillar stability during mining and other underground activities should be discussed and developed.