Roberto Rolo, MsC, PhD.
Use of Machine Learning and Geostatistics to Optimize Ore Control Models in Mining Operations
In the mining industry, the final destination of each mining block is often determined using geological models called production, grade, and ore-control models. The available information from diamond or reverse circulation drilling often does not provide the required resolution. For this reason, we proposed a hybrid machine learning and geostatistical workflow that integrates available grade data from production blastholes with its operational logs taken during drilling, ensuring the required resolution in the grade control models.
The step of geological logging can be either replaced or enhanced by machine learning models for classifying blastholes into lithologies and hardness classes. An automated workflow then uses all the available data to construct a short-term geological model with estimated rock type, ore grades, and hardness. This workflow was applied in a world-class iron ore mine in Brazil and was able to reduce ore loss and dilution, and improve mining predictability. By extending the life of the mine and reducing waste, the workflow contributes to a more sustainable mining industry.
Mining Engineer graduated from the Federal University of Ouro Preto (Brazil), MsC and PhD in Geostatistics from the Federal University of Rio Grande do Sul (Brazil). He has expertise in: implicit geological modeling and uncertainty assessment in geological models; workflows for estimating and simulating grades in accordance with the main international codes for declaring mineral resources and reserves; Python programming and Machine Learning applied to geosciences. He is currently a mineral resources consultant at Geovariances.