A Data Science Approach to Identifying and Quantifying Causes of Dilution at Cannington Mine

A data science based dilution study was conducted to identify the causes of dilution at South32’s Cannington Mine (Cannington), a large open stoping operation. In terms of industry benchmarks for unplanned dilution, Cannington is already performing very well, and so opportunities for improvement are more difficult to identify. The study is based upon back analysis of nearly 150 case studies from the mines reconciliation database, and has quantified the relative effect and significance levels for the following dilution parameters; time until paste filling, faulting and cable bolt reinforcement. Specifically, significant contributions were associated with faults and stope open duration. Quantifying the effect of faults on stability graph accuracy enabled a site specific adjustment to improve the predictive ability of the Cannington stability graph (unstable stope specificity improved from 63 per cent to 74 per cent). Despite Cannington’s already high levels of stope performance compared to industry benchmarks, data science methods have identified a complex interaction between identified dilution factors (ie stope open duration, presence of faults and cablebolting). Specifically, a third order Factorial ANOVA (Analysis of Variance) provided evidence that stope open duration reduces the effectiveness of cable bolts in faulted ground (average dilution = 2.1 m, p-value = 0.13). These results are consistent with site observations of a relationship between the stope open duration and the extent of failure when unravelling on a fault, even when cables are present.CITATION:Maher, K, Stewart, P C and Robotham, S, 2016. A data science approach to identifying and quantifying causes of dilution at Cannington Mine, in Proceedings Seventh International Conference and Exhibition on Mass Mining (MassMin 2016), pp 491–500 (The Australasian Institute of Mining and Metallurgy: Melbourne).