Abstract
The challenges facing the mining industry are varied not the least being trends toward undercover and deeper exploration as the rate of viable near-surface deposit discoveries declines, especially in well-established and -endowed mining jurisdictions. Other challenges facing the industry include increasing mining complexity as mining operations deepen and grades for key commodities such as copper decline and requirements to demonstrate strong environmental, social and governance principles. At the same time, the industry is also facing the challenge of meeting material growth forecasts as the world economy transitions to the emerging ‘green economy’. As the industry increases its undercover exploration efforts, to help support the generation of new targets and ultimately new discoveries, new technologies that integrate with more traditional approaches are being developed, adapted and/or evaluated.
One of these approaches is the use of machine learning (ML) in prospectivity analysis, which promises an objective approach that minimises human bias along with the ability to rapidly interrogate large and disparate data sets. As a result, ML has attracted a lot of interest, in recent times in particular, in the application to mineral exploration. In real world applications, however, prospectivity analysis is subject to a number of other biases that can be attributed to the data used, the complicated and tenuous relationship between these data and the mineral system proxies they aim to map and the sometimes equally tenuous relationship between proxies and the mineral system processes of interest. Ill-informed or a perceived lack of links between data, proxy and process is a weakness that the case study presented here has shown can be minimised with a nuanced combination of geoscience and data science.
BHP, SRK Consulting and DeepIQ have partnered in an initiative to tailor a combined approach to improve prospectivity modelling. By combining advanced data processing, knowledge compilation, geological interpretation and ML, the team investigated multiple approaches to support the discovery of large porphyry copper deposits (PCDs) in a well-explored region in the northern Chilean Andes. The worksflow utilised a mineral system approach to address the available data. Based on global and local geological knowledge, the main exploration vectors and criteria for PCDs were tabulated over the study’s Area of Interest (AOI). All valuable information was extracted from available regional datasets using a wide range of interpretational and processing methods and algorithms. Where regional datasets were considered lacking sufficient robustness, these were reworked to ensure they were fit for ingestion. This information was translated into evidential layers and the multi-modal data sets were ingested into a cloud data platform and integrated to form an analytics-ready dataset.
This data was analysed using a combination of knowledge-driven algorithms and advanced data-driven ML and deep learning algorithms, with various strengths and weakness of each incorporated into an iterative feedback loop. The data analysis outputs needed to identify known PCD locations with a high level of accuracy as the first step in demonstrating a credible and potentially robust approach to exploration targeting. Relative mineral potential was estimated in both the 2D and 3D environment using distributed and scalable cloud computing frameworks. The best results were merged using a Fusion approach to delineate focus areas. An overview of the approach, the different steps and of the successes and challenges encountered are presented. Implications for ongoing development and modern mineral exploration are briefly discussed.
