Exploring the Impact of Particle Size Distribution on Crusher Sorting in SAG Mills via Machine Learning

  • Laudari, Sudip (The University of Sydney)

Please login to view abstract download link

SAG mills are extensively used in mining operations to grind and crush materials such as ores, metals, and cement. Although SAG mills offer significant benefits in improving the sorting and separation processes, they face criticisms such as lack of information in complex operational dynamics which makes difficulties in grain sorting. In this study, we aimed to investigate the potential impact of particle properties and particle size distribution (PSD) on the sorting process. To achieve this, we conducted two simulations in a SAG mill using DEM. In simulation 1, we used a mixture of larger (crushers) and smaller grains, while in simulation 2, we used a mixture of larger (crushers) and tiny grains. We employed a machine learning classifier to differentiate between the crushers based on various features, such as velocity, acceleration, and force. Our findings indicate that classifying crushers based on all their properties is only possible when the ratio(r) of the radius of the smaller(rS) and tiny particles(rT) is significantly large. This suggests that crushers exhibit statistically almost similar properties while grinding in SAG mills. However, when we used the acceleration of crushers from each time step as features, we were able to classify them successfully. This indicates that classification process is influenced by the pattern of grain flows, rather than solely relying on their inherent properties. Further, we explored the effect of smaller and tiny grains on crushers sorting which leads to linearity between ratio(r) and sorting accuracy(ε). Which implies that, (ε) depends on PSD. Moreover, we examined the effect of smaller and tiny particles on crushers with different drum rotation numbers, filling masses, and volumetric portions of crushers.