Development of three-dimensional GPU DEM code–benchmarking, validation, and application in mineral processing

Laboratory-scale testing for particle dynamics is expensive and cumbersome; therefore, computational discrete element simulations (DEM) are a realistic option for aiding in the understanding of particle dynamics. GPUaccelerated simulation allows simulations to be conducted in a fraction of time by massively parallelizing compute-intensive elements of the calculations. The GPU’s highly threaded nature makes it an excellent choice for DEM simulations. This work develops a GPU DEM code for mineral processing applications such as hoppers and tumbling mills. To validate the developed DEM code, it was compared with other DEM simulation data and experimental data for various aspects of tumbling mill and hopper flow. The developed code was used to examine the influence of three lifter profiles and mill speeds on the granular dynamics and power consumption of a laboratory-scale tumbling mill. It was observed that for mill speeds less than 75% of the critical mill speed, the effect of various lifter profiles on particle dynamic behavior is minimal. This gives us the confidence to develop a significantly faster GPU-accelerated DEM code while maintaining reasonable accuracy at the microcontact level. Furthermore, the predictive capability of GPU-based DEM was demonstrated in an industrial-scale mill containing one million particles.