DEM Simulation of Mill Charge in 3D via GPU Computing

Discrete Element Method simulation of charge motion in ball, SAG and autogenous mills has become a standard for design of mill internals, power draft evaluation, etc. Both two-dimensional and three-dimensional codes are being used. The two-dimensional code dominates the user market since the code completes a simulation in one or two hours. On the other hand, three-dimensional codes have been in the domain of specialists who can handle the complexity of the code and have access to parallel computing hardware. For the first time, this barrier is overcome via computing with graphic processing units (GPU). Commodity graphics cards now allow parallel computation with hundreds of processors. It is shown that the computational time has been reduced by a factor of fifty using GPU computing. Hence, SAG charge simulation results in three dimensions are achievable in a few hours using affordable desktop machines. The GPU computing shown here enables the three dimensional simulation into a manageable daily task for mill designers and researchers. On the technical side, the simulations can be elevated to a higher level of complexity to simulate breakage and transport. First, the speedup factor of fifty is established for the Nvidia GeForce GTX 580 graphics card. Next, four simulation tasks, consisting of AG, SAG and ball mills are reported. The simulation of AG and SAG mills are completed in less than 8 hours. The ball mill simulation with 1.25 million spheres is completed in 27 hours. In the near future, as the graphic card hardware advances the compute time can be halved.