A Methodology for Fast and Efficient Geometallurgy Units Definition Utilizing Statistical Tools

Managing and analysing the vast amounts of data collected in a mining project effectively is crucial for enhancing the project's economic feasibility. The efficient utilization of this data is critical to minimising the time spent on projects and increasing project value. Process engineers can effectively integrate geometallurgical data into plant operations by employing analytical tools like Python, significantly optimizing the mine-to-mill process. This practical approach is a game-changer in the mining industry. Classifying a geomet dataset into units of similar characteristics (properties of the ore) is an important outcome of a geomet study. This approach simplifies data classification, allowing for class-by-class process modelling, which improves the decision-making process in mine planning and development. Classifying a geomet dataset involves using grade, lithology, and test work data (if available) to create blocks of similar characteristics, known as ‘geomet units.’ Compared to conventional methodologies, these units provide context and empirical evidence that can be more accurately mapped onto block-by-block process flowsheet development and net present value (NPV) estimation. This robust classification method reduces a project's technical risks and influences the initial mine planning sequence to maximise profitability. Defining robust geomet units with data sets with similar material is challenging due to the vast amounts of data available and the database's high heterogeneity. The methodology's purpose is to minimise time-consuming low-value-adding aspects of geometallurgy. The approach is different for a greenfield and a brownfield study, as all the valuable data in a brownfield process can be utilised to define the geomet units. Python has various libraries providing multiple machine-learning algorithms to assist in classification. The k-means algorithm is used for clustering, and the SHapley Additive exPlanations (SHAP) method is used to quantify the effect of each feature on the clustering results. However, even with automation, the methodology still requires a keen eye to account for the nuances that appear from project to project.