A geomechanics study was conducted to examine the seismic risks associated with late stage mining at a case study mine in Canada. Recent large scale pillar extractions and increasing seismicity at the site identified the need to consider seismic risk as the mine extends deeper and sill pillars are thinned and removed. Due to a lack of available quantitative data, a 3D FDM model was developed and the rock mass strength and behaviour was qualitatively calibrated to various geotechnical data sources. With the installation of a mine-wide seismic array, as well as measures of "depth of damage" during drilling and observed damage recorded throughout the mine during several field visits, sufficient data was accumulated to qualitatively calibrate a plastic model of the mine. The challenges in implementing a mine-wide inelastic model calibration with data acquired late in mine life are discussed.
Numerical methods are utilized in geomechanical engineering to improve understanding of physical processes, to determine stress or strain at specific instances in space or time and to predict rock mass response to excavation, construction or support. Some of the key advantages of numerical methods in mine design include the timely prediction of rock mechanics processes for the duration of a mining project, experimentation with design and construction options (e.g. stand-off distances and sequencing strategies), and performance of sensitivity analyses. Numerical analyses also allow engineers to gain a qualitative understanding of the complex geomechanical system through quantitative evaluation. The largest challenges faced in numerical modelling of geomechanical problems originate from lack of complete knowledge of the true rock mass properties, fracture networks, stress conditions and groundwater influences for the entire region of interest (i.e. project site). Rock masses are discontinuous, anisotropic, inhomogeneous, and for inelastic mine-scale simulations; the determination of material properties or in situ stress conditions is typically difficult over the full extent of a region being modelled.