GPU-based Parallel Algorithm for Super-Quadric Discrete Element Method and Its Applications for Non-Spherical Granular Flows. (January 2021)

• Record Type: Journal Article
• Authors: Wang, Siqiang; Zhang, Qi; Ji, Shunying
• Is Part Of: Advances in engineering software. Volume 151(2021)
• Publication Date: 2021-01
• Subjects: Discrete element method -- GPU-based parallel algorithm -- Super-quadric element -- Granular flow; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2020.102931
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: The CUDA-GPU parallel algorithm is developed for non-spherical particles. Differently shaped particles are modeled with super-quadric elements. The effects of particle shape on computational efficiency are investigated. The speedup ratio of the GPU to the CPU for differently shaped particles is studied. Abstract: GPU-based parallel algorithms for large-scale DEM simulations of spherical elements have been well established. However, complex systems composed of non-spherical elements are common in nature and industry. The dynamic behavior and mechanical properties of non-spherical elements are significantly different from those of spheres on the macroscopic and microscopic scales. Considering the construction of non-spherical elements and the computational requirements of large-scale engineering applications, a CUDA-GPU parallel algorithm based on super-quadric elements is developed. In this method, the parallel-vector concept of spheres is employed, and the bounding box list and the Newton iterative list are added. To examine the applicability of the GPU parallel approach, four tests are performed. The first involves the generation of a large-scale non-spherical granular bed. The second consists of comparisons against the experimental flow processes of the non-spherical granular column. In the third, the influences of the particle shape on the calculation efficiency and the speedup ratio of the GPU to the CPU during the discharging process are investigated. The last test consists of the evaluating the mixing behaviors of large-scale non-spherical systems within a horizontally rotating drum. These studies demonstrate that the proposed CUDA-GPU parallel algorithms are applicable and reliable for the large-scale engineering applications of non-spherical granular systems.