Development of 2-D Hybrid FEM/DEM Method Code Using GPGPU

Fukuda, D. (Hokkaido University / University of Tasmania) | Liu, H. (University of Tasmania) | Mohammadnejad, M. (University of Tasmania) | Chan, A. (University of Tasmania) | Cho, S. H. (Chonbuk National University) | Min, G. J. (Chonbuk National University) | Kodama, J. (Hokkaido University) | Yoshiaki, F. (Hokkaido University)

OnePetro 

Abstract

This paper introduces the Y-HFDEM code based on two-dimensional combined finite-discrete-element method (FEM/DEM) for numerical simulation of fracturing process in brittle and semi-brittle materials including rocks. The code has been successfully employed to simulate rock breakage under both quasi-static (e.g. uniaxial compression) and dynamic (e.g. rock blasting) loading conditions. However, the most challenging part in the application of the original Y-HFDEM code was its simulation time required to solve large-scale problems with massive number of nodes, elements and contact interactions. To overcome this limitation, this paper demonstrates the application of GPGPU (General Purpose Graphic Processing Unit) and CUDA (Compute Unified Device Architecture) C/C++ to parallelize the original sequential 2-D Y-HFDEM code along with related numerical algorithms using the GPGPU. Obtained results from verification examples demonstrate the capability of the proposed Y-HFDEM code in modelling larger scale problems in which massive computational effort is required.

1. Introduction

Understanding the mechanism of the fracture process in rocks is important in the field of civil and mining engineering. Numerical methods have been increasingly applied recently to analyze the fracture process of rocks. For a realistic simulation of the fracture process of rock, numerical techniques must be capable of capturing crack onset and arbitrary crack growth, correct crack length within a given time interval as well as the propagating directions. In recent years, increasing attention has been paid on the techniques which bring together the advantages of the continuum-based and discontinuum-based computational methods. The combined finite-discrete element method (FEM/DEM) proposed by Munjiza (2004) has been employed successfully to model problems dealing with transition process from continuum to discontinuum such as rock fracturing and fragmentation (Mohammadnejad et al., 2018). ELFEN(2D/3D) (Rockfield, 2005) and Y(2D/3D) code (Munjiza, 2004) are two main implementations of the combined FEM/DEM. Several attempts have been made to extend the Y code such as Y-GEO(Mahabadi et al., 2012), IRAZU(Mahabadi et al., 2016), SOLIDITY (Xiang et al., 2016), HOSS with MUNROU (Rougier et al., 2014) and authors’ Y-hFdEM (Liu et al., 2015). The principles of the combined FEM/DEM are based on both continuum mechanics, cohesive zone modelling and contact mechanics which make it computationally expensive. Therefore, developing a capable parallel computation schemes is important in order to deal with larger scale problems with massive number of nodes, elements and contact interactions.