Last week we introduced this series on Artificial Intelligence (AI), focusing on the opportunities and challenges this technology brings to the industry. Today I will examine why mining companies need smart machines, detailing the various approaches to using AI for resource estimation, what early results tell us, and the critical barriers to implementation.
The objective is to predict lithology and ore quality in technical logs based on geophysical information, and then reduce the number of samples requiring physical testing and analysis.
Mining is a long-term proposition. And that’s just for mineral extraction. Before that can happen, information must first be extracted to locate, evaluate, validate, quantify, and ultimately gamble on were to dig. And even before that, a resource geologist must factor fluctuating commodity prices, how to pursue hard-to-reach ore reserves, and capital investments and operating costs both direct and indirect. How to improve the odds of accurate estimation?
Artificial Intelligence (AI) and Machine Learning (ML) can help by quickly interpreting oceans of data no single person or team can manage alone. It’s about pattern recognition across disparate databases from wide-ranging sources: government geological records, site equipment sensors, borehole samples, satellite imagery, magnetic field readings, topographical maps, and field reports. Machine Learning can learn correlations among geoscience data to make predictions, which saves time and money while improving geological and resource modeling.
Faster Models, Less Drilling
For mining resource estimation, a geologist must create, coordinate, review and help prepare resource models. These tasks encompass interpretation of geology and mineral controls, estimation of ore qualities and validation of models. The geologist must be able to use reporting codes and mineral classifications to appropriately quantify risk and help improve technical understanding. At the same time, the geologist must reduce the time it takes to perform a resource classification, purge human bias, address variable drilling densities, and properly consider deposit undulation.
Mining companies typically use two exploratory drilling methods to get information on lithology, stratigraphy and to collect pristine samples for bio stratigraphic, geochemical, geochronological and mineralization studies:
- Open hole - or reverse circulation - drilling involves crushing the cuttings to surface. Cheap and quick, this method is, however, not particularly reliable.
- "Diamond" - or "core" - drilling involves withdrawing small diameters of core rock from an orebody for examination and analysis. While many geologists prefer this method, it is more costly and core analysis can take up to 14 weeks.
Machine Learning software, however, uses a geology database containing geophysical, lithological, and quality data to build structural and quality models to predict ore quality. The method improves the accuracy of quality prediction, and companies can cut drilling costs by reducing the number of costly core boreholes traditional exploration requires.
Moreover, exploration time is reduced because models can be created from instant predictions, which benefits short-term mine scale models by skipping lab analysis. Machine Learning drastically reduces the time from drilling to lab analysis. Human bias is reduced, enhancing the reliability and consistency of the geology. Accurate predictions of geotechnical properties also reduce blasting expenditures and create stable mine slopes.
Making the Case for ML
How to explain Machine Learning to non-technical executives? Machine Learning technology can generate faster, more accurate assessments than traditional methods. That’s important because major discoveries of near-surface mineral deposits are declining globally. Deep subsurface mineral deposits are getting harder to reach and smaller in size.
Machine Learning, with greater accuracy and reduced error, can be used to quickly develop accurate models to predict rock type and the economic viability of drilling. This leads to faster decisions with less risk, higher return on investment, improved health and safety, higher efficiency and reduced environmental impact.
In our next post, we drill deeper into various approaches data scientists are exploring to develop Machine Learning tools for resource estimation.
Min Liang is a Geostatistician and a Data Scientist. She holds a PhD in Geostatistics from the École Polytechnique de Montréal and records 6 publications on geospatial modelling and simulation. Before joining Dassault Systèmes in 2019, Min worked 1 year in Data Sciences and 3 years as an Environmental Consultant
Access all technical articles:
machine learning resource estimation optimization AI_for_resource_estimation series
