Kotze, G, Santos De Santana, D, Mendes Machado, LR & Grix, Q 2024, 'Enhancing the calibration of elastic numerical models through stress measurements and observations of stress-induced overbreak', in P Andrieux & D Cumming-Potvin (eds), Deep Mining 2024: Proceedings of the 10th International Conference on Deep and High Stress Mining, Australian Centre for Geomechanics, Perth, pp. 739-756, https://doi.org/10.36487/ACG_repo/2465_46
(https://papers.acg.uwa.edu.au/p/2465_46_Kotze/)
Abstract: In deep mines, where high stress conditions can prevail, geotechnical engineers are required to conduct excavation stability assessments and predict excavation damage proactively. The geotechnical engineer usually undertakes a numerical modelling assessment that aims to quantify the causality between mining sequence, layout and damage. 
The reliability and value offering of an uncalibrated numerical model are probably similar to the reliability of parametric or sensitivity studies. It is therefore prudent to quantify the pre-mining stress state and to consider the rock mass responses with a view to enhancing the reliability of model results. Examples of responses include seismic event locations and source parameters, instrumentation data, time-dependent deformation of rock, damage in tunnels or pillars, and stress-induced sloughing.
Calibration of elastic models is based on the notion that excess stress is a direct predictor of expected plastic strain. The model calibration process considers the stress state at damage locations in a mine. The stress state can be described by pairs of major and minor principal stresses collected from the corresponding damage locations in the numerical models. To obtain a strength envelope, a curve is fitted through the pairs of data. A manual process of data appreciation is followed, where the fitted curves and interpretation of data are considered with a view to providing a simple criterion for predicting damage.  
At Caraiba mine in Brazil, an initial calibration was done using cavity mine surveys and the historically accepted stress tensor. Subsequently the model input parameters were updated using actual stress measurements and then followed by an update of the initial calibration. Elastic models are quick to set up and the observational method of model calibration is relatively simple to execute. This methodology is attractive to geotechnical engineers since they have to collect rock mass response data routinely and they have limited time to conduct numerical modelling. 

Keywords: stress-induced overbreak, model calibration, damage observations, elastic modelling