The following is an article from Benoît McFadyen, Australian Centre for Geomechanics, Canada which appeared in the Deep Mining 2024 Supplement.
Stope optimisation: new tools for understanding and improving mined geometries
By Benoît McFadyen, Australian Centre for Geomechanics, Canada
Introduction
Open stoping is a popular method for extracting ore in underground mines and is a method that has seen little change in the design process since the 1980s when the stability chart was introduced (Mathews et al. 1981). While the stability chart is a reliable approach at the feasibility stage to determine stable sizes for stopes, it offers little benefit at the operational stage for optimising stope performance (qualitative predictions, underbreak [UB] not considered, and limited parameters taken into account). Current mining practice relies heavily on the stability chart for stope design even though a large amount of data is collected throughout the different design, mining and reconciliation steps (Figure 1; Potvin et al. 2020). This is due to inefficient means, such as time-consuming manipulations, and personnel turnover that hinder the transformation of the data into information and the building of knowledge that can be fed back into the design process. Therefore, a new method and tools were developed by the Australian Centre for Geomechanics within the mXrap software to streamline the transformation of data into knowledge for operational mines so they can get the most insight out of their data to improve their stope performance (economically and in terms of overbreak [OB] and UB). An overview of the methodology and tools is given through this article.
The optimisation approach: overview
The methodology developed and implemented through the mXrap Stope Reconciliation and Prediction Suite aims to provide a path for operational mines to optimise stope performance by understanding the spatial distribution of OB and UB at a metre-scale resolution and using that information to predict the expected stope mined geometry (McFadyen 2024). The approach is divided into four key steps which are achieved through data quantified at an octree resolution, statistical analysis and a machine learning model (Figure 2).
Per octree is a finer resolution developed by Woodward et al. (2019) to characterise the spatial distribution of OB and UB along the design face by calculating the projected distance from the design to the cavity monitoring system (CMS) on a grid point basis (Figure 3).
Using the optimisation approach: the mXrap Stope Reconciliation and Prediction Suite
The mXrap Stope Reconciliation and Prediction Suite was developed with the objective to provide an efficient method for sites to build a stope database for quantifying stope performance, undertake root cause analysis, and predict stope performance, all while increasing the level of information and knowledge that sites are currently used to. There are three main parts to the suite: stope reconciliation, analysis, and predictions.
Stope reconciliation app
The stope reconciliation app allows the user to import design geometry, drill ring design, and CMS shape to quantify the different variables and stope performance. This is done at three levels of resolution:
1. per stope
2. per face
3. per octree.
The app is separated into different steps according to the stage the stope design or mining is at. First is the design step (Figure 4) which allows the user to import the design geometry and establish the different faces, octree data structure, and different variables (geometrical, geological, and geotechnical).
At the octree resolution, different variables can be currently quantified given the data available:
- geometrical variables
◊ characterises the design shape and size - geological variables
◊ characterises major structures with regard to the design - operational variables
◊ characterises drill design, undercutting of drives - geotechnical variables
◊ characterises the rock mass around the design using block model data.
This approach is not limited to these variables.
Additional variables which look to characterise stress, mining sequence, ground support, and backfill will also be available through further research and development. Site-specific variables can also be developed.
When the drill design is made, it can be imported to calculate the drill-and-blast variables (Figure 5).
Finally, when the stope has been mined, the design is reconciled with the CMS to obtain the stope performance (Figure 6). The information obtained gets added to the stope database after each step.
Analysis app
The analysis app allows the user to visualise the stope database and filter according to what they want to analyse. Stope performance can be assessed at three levels of resolution using univariate and bivariate charts (Figure 7). Through root cause analysis, trends can be analysed between variables and OB or UB to understand the mined geometry that is obtained and key variables to use for predicting the expected mined geometry can be identified.
Prediction module
The prediction module is nestled in the stope reconciliation app, just before the reconciliation step. It allows the user to run predictions for the design that has been imported (Figure 8). Predictions are made for each octree using a random forest model, allowing the user to build the expected mined geometry.
In addition, a standard error is calculated for each prediction allowing the user to establish a prediction interval within which the OB or UB most likely lies. Furthermore, given the octree resolution, a geological block model is used to map economic data to the predictions and determine the economic value of the OB, UB and overall predicted shape. This approach provides an assessment of the stability and economic performance which gives the necessary information for sites to assess optimisation needs. New predictions can be generated after modifying certain variables to assess how stope performance can be improved.
Summary
The mXrap Stope Reconciliation and Prediction Suite has proven to be a valuable asset for operational mines to get the most out of their data, transforming it into information and valuable knowledge for their stope design process. Having the whole process and database within one software system has enabled the process to be streamlined, allowing the user to reconcile and analyse the data within 15 minutes with a similar timeframe when predictions are run. The use of octrees has allowed users to dive deeper into the analysis of OB and UB and generate predictions that allow them to determine the predicted mined geometry, going well beyond qualitative prediction of the stability chart or the average predictions of equivalent linear overbreak sloughing (ELOS) charts (Figure 9). With the inclusion of economic data, the methodology and tools offer what is needed to optimise stope performance.
References
Mathews, KE, Hoek, DC & Stewart, SBV 1981, Prediction Of Stable Excavation Spans for Mining at Depths below 1000 Metres in Hard Rock, report to Canada Centre for Mining and Energy Technology.
McFadyen, B, Grenon, M, Woodward, K & Potvin, Y 2023, ‘Assessing stope performance using georeferenced octrees and multivariate analysis’, Mining Technology, vol. 132.
McFadyen, B 2024, Developing a New Methodology for Predicting Open Stopes’ Performance, PhD thesis, Université Laval, Québec.
McFadyen, B, Grenon, M, Woodward, K & Potvin, Y 2024, ‘Optimising stope design through economic and geotechnic assessments of predictions made at a meter scale resolution using the sites’ reconciled data’, International Journal of Rock Mechanics and Mining Sciences, vol. 178.
Potvin, Y, Woodward, KR, McFadyen, B, Thin, I & Grant, D 2020, ‘Benchmarking of stope design and reconciliation
practices’, in J Wesseloo (ed.), UMT 2020: Proceedings of the Second International Conference on Underground Mining Technology, Australian Centre for Geomechanics, Perth, pp. 299–308, https://doi.org/10.36487/ACG_repo/2035_14
Woodward, K, McFadyen, B, Potvin, Y & Wesseloo, J 2019, Probabilistic Stope Design, MRIWA Project No. M489, Australian Centre for Geomechanics, Perth.