Automatic coal mine roof rating calculation using machine learning

Abstract

The Coal Mine Roof Rating (CMRR) is an essential input parameter for roof support design. The current manual calculation process brings evident subjectivity in selecting rock mechanical properties, roof unit classification, and fracture spacing calculation, which is needed for CMRR calculation. In this paper, state-of-the-art machine learning and computer vision techniques were applied to provide a repeatable and consistent data-driven solution to reduce the subjectivity of CMRR calculation through integration of geophysical logging data, drill core images, and laboratory data. Firstly, the group-based machine learning approach, which has demonstrated better performance than the conventional machine learning methods, was introduced to predict the uniaxial compressive strength (UCS) of roof strata. Then, an advanced computer vision model trained by transfer learning technique was adopted to extract the dimensions of the core pieces for the automatic rock quality designation (RQD) and fracture spacing calculation based on the drill core images. The geotechnical units within the roof bolted interval were classified in terms of the adjacent sedimentary layers, which share similar geotechnical properties. The predicted quantities (UCS, RQD, and fracture spacing) by the machine learning methods were converted into the corresponding ratings, which were further used for the unit ratings and machine learning based automatic CMRR calculation. The automatic CMRR values from machine learning models show a promising linear correlation (R2 = 0.78) with the manually calculated CMRR values, demonstrating that this new approach has the potential to be applied as an alternative method, which delivers a benchmark of CMRR values to enhance its calculation confidence and objectivity.