Developments

Brief development Blog:

A novel three-dimensional grain-based (GB3D) approach was developed within the framework of GPGPU-parallelized HFDEM in 2022. The GB3D approach considers the microstructures of granular rocks, including grain morphology, to investigate the transgranular, intergranular and intragranular damages. Since GB3D modelling is more computationally intensive than conventional 3D modelling, efficient contact calculation algorithms such as the tetrahedron-to-point (TtoP) contact interaction algorithm and semi-adaptive contact approach (semi-ACAA) were incorporated to further speed up the GB3D modelling besides GPGPU parallelisation. The TtoP contact interaction algorithm discretizes the target triangular faces as Gauss points and is consequently approximately 1.5 times faster than the tetrahedron-to-triangular-facet (TtoF) contact interaction algorithm. Moreover, it is found that the grain-growth tessellation method generates rounder grains than Voronoi tessellation method. The rounder boundaries of the grains help increase the peak strength in the stress-strain curve, while the sharper boundaries of the grains result in more zigzag crack propagation during the post-failure stage, especially when the crack propagates along the grain boundaries. In addition, the GB3D modelling with a grain-growth tessellation method is 1.9 times faster than that with a Voronoi tessellation, since the better roundness of the grains generated using the grain-growth tessellation methods improves the mesh qualities. This development was published in Computers and Geotechnics.

In December 2021, GPGPU-parallelized HFDEM has been successfully compiled and run on CSIRO's high performance computer, Bracewell, which is equipped with 456 nVidia Tesla P100 GPUs

A novel and efficient semi-adaptive contact activation approach was implemented into HFDEM in 2021, which could adaptively activate contact calculations for continuum solid elements around the cohesive element just subjecting to shear softening while its softening function had just satisfied a prescribed threshold. The semi-adaptive contact activation not only overcomes spurious fracturing mode associated with FDEM simulations using the adaptive contact activation approach but also is more physically sound and computationally efficient compared with full contact activation approach prevalent in FDEMs with intrinsic cohesive zone model-based cohesive elements. The semi-adaptive contact activation approach was about 1.23 to 13.5 times faster than the full contact activation approach depending on the intensity of shear cracking in the simulations including mixed-mode cracking. This development was published in International Journal of Rock Mechanics and Mining Sciences.

GPGPU-parallelized HFDEM was redesigned in 2020 to become two versions, i.e. GUI (Graphical User Interface) and HPC (High Performance Computing). In HPC version of HFDEM, the GUI developed using Visual C++ and OpenGL was separated from the computational module, which was developed using standard C/C++ only for running on HPC.

An adaptive contact activation approach and a mass scaling technique with critical viscous damping were implemented into GPGPU-parallelized HFDEM, which proved that 3D modelling with the adaptive contact activation approach was 10.8 times faster than that with the traditional full contact activation approach and at least additional 25 times of speedups could be achieved by the mass scaling technique. These developments were detailed in our paper published in Computational Particle Mechanics in 2020.

Thanks to JSPS KAKENHI for Grants-in-Aid for Young Scientists and Australia-Japan Foundation, Dr Daisuke Fukuda was funded to visit Dr Hongyuan Liu's group at University of Tasmania for one year from 2017 to 2018, during which Y-HFDEM IDE2D/3D and DFPA3D were integrated and parallelized on the basis of GPGPU using CUDA C/C++ to develop a GPGPU-parallelized HFDEM. These developments in 2D and 3D were published in International Journal for Numerical and Analytical Methods in Geomechanics & Rock Mechanics and Rock Engineering, which indicated that GPGPU-parallelized HFDEM is more than 128 and 284 times faster than sequential HFDEM for 2D and 3D modellings, respectively.

Non-reflection boundary for modelling infinite rock, effect of loading rate on dynamic behaviour of rock, and detonation-induced gas expansion as well as flow through fracturing rock have been implemented into hybrid finite-discrete element method to simulate fracture and resultant fragment muck-piling in various blasting scenarios.

Dr Hongyuan Liu made a presentation "Hybrid finite-discrete element modelling of dynamic fracture of rock and resultant fragment muck-piling by rock blast" in the 8th Asian Rock Mechanics Symposium hosted in Japan in 2014, which was later published in Computers & Geotechnics and inspired Dr Daisuke Fukuda at Hokkaido University to work on FDEM. Later, Dr Daisuke Fukuda independently developed a three-dimensional Dynamic Fracture Process Analysis (DFPA3D) code using Fortran.

A both two-dimensional (2D) and three- dimensional (3D) Integrated Development Environment (IDE) of HFDEM, i.e. Y-HFDEM IDE2D/3D, was firstly developed by Dr Hongyuan Liu from 2010 to 2012 using Visual C++ and OpenGL on the basis of open-source Y library (Munjiza, 2004; Munjiza et al., 2011) of combined finite-discrete element method (FDEM) and Dr Liu's former RT2D (Liu et al., 2004) and Tunnel3D (Liu et al., 2008) of statistics and damage mechanics-enriched finite element method. Y-HFDEM IDE2D/3D was then presented in the 11th Australia New Zealand Conference on Geomechanics (ANZ2012), which was among one of 15 highly ranked papers and was later recommended by ANZ2012 to be included in a special issue of ANZ2012 of International Journal of Geotechnical Engineering.