Discrete element simulation of mill charge in 3D using the BLAZE-DEM GPU framework

Abstract

The Discrete Element Method (DEM) simulation of charge motion in ball, semi autogenous (SAG) and autogenous mills has advanced to a stage where the effects of lifter design, power draft and product size can be evaluated with sufficient accuracy using either two-dimensional (2D) or three-dimensional (3D) codes. While 2D codes may provide a reasonable profile of charge distribution in the mill there is a difference in power estimations as the anisotropic nature within the mill cannot be neglected. Thus 3D codes are preferred as they can provide a more accurate estimation of power draw and charge distribution. While 2D codes complete a typical industrial simulation in the order of hours 3D codes require computing time in the order of days to weeks on a typical multi-threaded desktop computer. This paper introduces a 3D GPU code based on the BLAZE-DEM framework that utilizes the Graphical Processor Unit (GPU) via the NVIDIA CUDA programming model. Utilizing the parallelism of the GPU a 3D simulation of an industrial mill with four million particles takes 1.16 hour to simulate one second (12 FPS) on a GTX 880 laptop GPU. This new performance level makes 3D simulations a routine task for mill designers and researchers. Furthermore the shorter compute time can elevate computations to a higher level wherein ore particle breakage and slurry transport can be included in the simulation. In this paper we verify our GPU code by comparing charge profiles and power draw obtained using the CPU based code Millsoft and pilot scale experiments. Finally, we show computations for plant scale mills.

Blaze-DEMGPU: Modular high performance DEM framework for the GPU architecture

“…2(a) shows the benefits of Blaze-DEM in terms of increasing the number of particles in a simulation. We see that current DEM simulations did not correctly capture the behavior of what was being simulated, while Blaze-DEM gives a very good match to what is seen in reality [12,14]. Fig.…”

“…In [12,14] we applied the code to industrial ball mill and silo simulations respectively. The results obtained were in good agreement to experiment.…”

“…2(a) the lack of particles in existing DEM simulations yielded results that are not correct. Hence using the code DEM can be used as an effective tool for the prediction of many [9] Clumped High 2.56 × 10 5 1.49 × 10 6 Radake et al [10] *Sphere High 20 × 10 6 20 × 10 6 Nvidia SDK (2014) [11] *Sphere Low 2.5 × 10 5 125 × 10 6 BLAZE-DEM [12] *Sphere High 60 × 10 6 100 × 10 6…”

“…Recent advances in numerical algorithms for nearest neighbor searching have made it possible to simulate larger number of particles on mainly multi-core central processing unit (CPU) architectures. The CPU architecture limits the potential to further increase the number of particles to be simulated, due to the limited number (typically [8][9][10][11][12][13][14][15][16] of high performance computing cores [2][3][4]. Several DEM codes take advantage of modern parallel processing capabilities through shared or distributed systems.…”

Multi-layer kinematics and collision energy in a large-scale grinding mill-the largest semi-autogenous grinding mill in China

“…In addition, because of the relative tangential velocity between the mill and the steel balls, the total collision energy calculated using the classic theory should consider the relative tangential velocity. 11,12 Currently, the kinematic characteristics of steel ball have been numerically simulated using discrete element methods [13][14][15][16][17][18][19] and experimentally investigated through sensor technology. [20][21][22][23][24][25][26][27] It was found that the operating parameters had a profound effect on the charge characteristics, 28 and the effect of operating variables on the optimum performance of the mill still needed to be addressed to enhance energy utilization.…”