Numerical techniques for fast generation of large discrete-element models

Abstract

In recent years, civil engineers have started to use discrete-element modelling to simulate large-scale soil volumes thanks to technological improvements in both hardware and software. However, existing procedures to prepare ‘representative elementary volumes’ (REV) are unsatisfactory in terms of computational cost and sample homogeneity. In this work, a simple but efficient procedure to initialise large-scale discrete-element 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 and equilibrated at the desired stress state, porosity and coordination number. When the cell is in equilibrium, it is replicated in space to fill the problem domain, and when the model is filled, only a small number of mechanical cycles is needed to equilibrate a large domain. The result is an equilibrated homogeneous sample at the desired initial state in a large volume.

Effects of screw pile installation on installation requirements and in-service performance using the discrete element method

“…Rather than using a continuum to model the soil, DEM uses discrete particles that have the ability to interact as a soil body. With the application of an increased gravitational field, the DEM is able to act as a virtual centrifuge (Ciantia et al 2018;Sharif, et al 2019a) when properly calibrated (as detailed in Sharif et al (2019a)), with the added benefit of using a single soil bed (particle arrangement) which can be reset and used multiple times (Shi et al 2019).…”

“…The virtual soil beds for DEM analysis were created in accordance with the specification in Sharif et al (2019b). Three soil beds were created using the periodic cell replication method (Ciantia et al 2018) and particle refinement method (McDowell et al 2012), with each bed having a different relative density (D r ) , based upon the physical voids ratio of the sand modelled. The relative densities selected were 30%, 52% and 83%, with the densest bed being consistent with the physical modelling study conducted by Davidson et al (2019).…”

“…This is supported by the finding of Sharif et al (2020) who showed that there was negligible increase in mean effective stress at a radial distance of 13R and no increase D r a f t 12 at a radial distance of 20R when installing piles using the DEM. Additional information and detail of the formation process can be found in Sharif et al (2019b) and Ciantia et al (2018). To reduce the run-time of the simulation, a particle size distribution (PSD) scaling value (n i ) of 20 was used at the centre of the sample with a maximum n i of 96.5 at the boundaries.…”

Using discrete element method (DEM) to create a cone penetration test (CPT)-based method to estimate the installation requirements of rotary-installed piles in sand

“…DEM has been previously used to model a variety of different soil-structure interaction problems including pile plugging (Liu et al, 2019), cone penetration tests (Butlanska et al, 2014) and jacked piles in sand (Ciantia et al, 2019). With the application of an increase gravitational field, the DEM is able to act as a virtual centrifuge (Ciantia et al, 2018) which when properly calibrated, has the added benefit of using a single soil chamber which can be reset and used multiple times. This allows for direct comparisons to be made in parametric studies and D r a f t 7 potentially removes the reliance on specialist laboratory facilities or comparisons to expensive field studies, where soil variability can be an issue with interpretation.…”

Standard penetration testing in a virtual calibration chamber

“…To roughly mimic the effect of non-spherical particle shapes, particle rotation was inhibited, directly fixing the rotational degrees of freedom of the particles. This simplified approach, can be traced back to Ting et al [35] and was successfully applied in previous work with granular materials [21,36,37,38,39]. More refined consideration of particle shape effects may be obtained using rolling-resistance contact models (e.g.…”