A GPU-accelerated CFD-VOF-DEM approach for modelling the wet grinding process in a SAG mill and the mill speed effects

Abstract

Semi-autogenous grinding (SAG) mills are essential for cost-effective rock beneficiation, primarily utilizing wet milling processes. Although the milling process has been extensively studied by DEM modelling, the effect of slurry is frequently neglected due to its complex interactions with the charge. This paper presents an efficient CPU-GPU framework for CFD-DEM modelling of the gas–liquid-solid flows during the wet grinding in a 34′ SAG mill. The model, which adopts a VOF approach to simulate the free surface between gas and liquid slurry phase, was comprehensively validated across four different cases, including particle sedimentation, phase conservation, and particle dynamics in both dry and wet drums. The validated model was used to simulate multiphase interactions in a SAG mill at various rotation speeds.

The milling process stabilized after 20 s, with the bed showing a kidney shape, characterized by the cascading zone and cataracting stream. The slurry flow, enveloped by the particle bed, exhibited a strong recirculating flow following the particle trajectories. The inclusion of slurry resulted in weaker and fewer particle collisions, with the fluid force accounting for half of the total force acting on particles. The effect of the slurry on rock particles was greater than on ball particles due to the density difference. Power consumption mainly resulted from interactions between the lifter sides and particles, while the slurry consumed minimal power. Both collision energy and frequency significantly increased at higher mill speeds, which are also expected to cause more severe liner wear and increased power consumption.

The model provides an efficient framework for simulating dense particles interacting with multiple fluid phases. The findings elucidate the multiphase interaction mechanisms in the SAG mill and underscore the necessity of including slurry in simulations to accurately model the milling dynamics.