An efficient large-scale DEM model initialization procedure

TY - GEN

T1 - An efficient large-scale DEM model initialization procedure

AU - Ciantia, Matteo

AU - Shire, Tom

N1 - This study was undertaken as part of an Imperial College London Junior Research Fellowship Research Grant.

PY - 2017

Y1 - 2017

N2 - In civil engineering, the discrete element method (DEM) is a numerical tool often used to reproduce soil behaviour at the elementary level. In general, small cubic or cylindrical samples are generated in a representative elementary volume (REV) to assess the micromechanical behaviour of the virtual soil. As the response of particulate media is strongly dependant on the initial state (stress and porosity), sample generation is the most crucial part of a DEM model and different procedures have been established to prepare homogeneous REVs. Thanks to the technological improvement of both hardware and software, in recent years the DEM has started to be used to simulate large-scale boundary value problems (BVPs). However, the use of any of the existing procedures employed to prepare REV samples is unsatisfactory in terms of computational cost and sample homogeneity. In this work a simple but very efficient procedure to initialize large-scale DEM models is presented. Periodic cells are first generated with a sufficient number of particles (enough to consider the cell an REV) matching the desired particle size distribution (PSD) and equilibrated at 100 kPa isotropic stress state at the desired porosity. Once the cell is in equilibrium, it is replicated in space in order to fill the problem domain. Once the BVP’s model is filled, only a small number of mechanical cycles are needed to equilibrate the large domain. The result is an equilibrated homogeneous sample at the desired porosity in a large volume. Results of shearing test simulations on such samples are presented and compared to simulations using the smaller REVs alone.

AB - In civil engineering, the discrete element method (DEM) is a numerical tool often used to reproduce soil behaviour at the elementary level. In general, small cubic or cylindrical samples are generated in a representative elementary volume (REV) to assess the micromechanical behaviour of the virtual soil. As the response of particulate media is strongly dependant on the initial state (stress and porosity), sample generation is the most crucial part of a DEM model and different procedures have been established to prepare homogeneous REVs. Thanks to the technological improvement of both hardware and software, in recent years the DEM has started to be used to simulate large-scale boundary value problems (BVPs). However, the use of any of the existing procedures employed to prepare REV samples is unsatisfactory in terms of computational cost and sample homogeneity. In this work a simple but very efficient procedure to initialize large-scale DEM models is presented. Periodic cells are first generated with a sufficient number of particles (enough to consider the cell an REV) matching the desired particle size distribution (PSD) and equilibrated at 100 kPa isotropic stress state at the desired porosity. Once the cell is in equilibrium, it is replicated in space in order to fill the problem domain. Once the BVP’s model is filled, only a small number of mechanical cycles are needed to equilibrate the large domain. The result is an equilibrated homogeneous sample at the desired porosity in a large volume. Results of shearing test simulations on such samples are presented and compared to simulations using the smaller REVs alone.

KW - DEM

KW - Periodic Cells

KW - Sample preparation

KW - BVP

UR - http://ukacm2017.ukacm.org/wp-content/uploads/2017/04/Proceedings-UKACM2017-compressed.pdf

M3 - Conference contribution

SP - 187

EP - 190

BT - Proceedings of the 25th UKACM Conference on Computational Mechanics

A2 - Faramarzi, Asaad

A2 - Dirar, Samir

PB - University of Birmingham

CY - Birmingham, United Kingdom

T2 - 25th UKACM Conference on Computational Mechanics

Y2 - 11 April 2017 through 13 April 2017

ER -