A Machine Learning Model for Evaluation of Chain Pillar Stability in Deep Longwall Workings in India - Mining, Metallurgy & Exploration (2023)

The size of chain pillars in the Indian geo-mining conditions is primarily decided in accordance with the regulatory provisions. These provisions are of little help for deep longwall workings, exceeding the cover depth of 360 m. Furthermore, these provisions are primarily meant to support pillars in bord and pillar workings. They may not be suitable for the chain pillars owing to a significant difference in layouts and functional requirements. Hence, it is imperative to develop a method to evaluate the stability of the chain pillar considering the complex loading and geo-mining conditions. To this end, a machine learning-based model was developed in this study to assess the stability of the chain pillars under high depth of cover considering the field-representative conditions. The data for this work was generated by conducting a parametric study using a field-validated numerical model. The strength and deformability parameters of the caved, fractured, and continuous deformation zones were established by calibrating the numerical model outcomes against the site-specific field observations. The developed ML model was used to evaluate the influence of pillar width, pillar height, cover depth, abutment angle, face length, coal strength, moduli ratio of the roof and floor strata to the coal seam, and the elastic modulus and unit weight of the overburden on the chain pillar stability. An interesting finding of this study showed that the factor of safety of the chain pillar increased with the increasing abutment angle, which may be indicative of the confining effect of the abutment angle on the pillar. Eventually, the ML model was verified by comparing its outcomes with the calibrated numerical model for a typical longwall working from Indian geo-mining conditions.