Core databases often span multiple decades and can represent thousands to hundreds of thousands of meters of core samples. Significant effort and cost go into collecting this core, yet historical datasets are often underutilized in advanced exploration and operational settings.
The increasing accessibility and power of machine learning workflows allowing mining practitioners to look at this historical data in ways not previously possible. By leveraging these technologies, SRK Consulting has developed innovative workflows to improve the utilization of core photo databases.
Over the past four years, SRK has been implementing automated classification of core imagery to increase data availability and spatial resolution for geological characterization. The specialist team includes structural geologists, resource geologists, and geotechnical engineers, bringing a diverse set of skills and industry experience to support their clients.
In addition, SRK has worked collaboratively with its clients to gain an appropriate understanding of the geological context. In some cases, certain parameters, such as a geological unit, alteration style, foliation or veining density, were not collected systematically or reliably because their importance was not clear at the time. Once the limitations of the current models are understood, a set of pragmatic classifications is proposed for implementation on existing core imagery. The resulting solutions range from typical logging classifications to custom-tailored workflows that capture the most relevant aspects of a deposit.
To date, the team has processed over 1.5 million meters of core across many deposit styles. In many cases, the outcomes have directly supported downstream study work undertaken by SRK, further improving product quality for clients.
LeRiche, A., Tims, S., Saunders, E., and Mohr, P. (2022). Geotechnical Domaining for the Aktogay Porphyry Deposit Supported by Machine Learning Techniques. In Proceedings of: International Slope Stability 2022 Symposium, Tucson, US
