Abstract
The simulation of flows in porous media on the microscale level requires a high number of unknowns to resolve the geometry of the fissures in-between the sand grains. For the geometric representation of the sand grains we use fracture network-like structures in 2D and sphere packings in 3D. These geometries are created with an in-house scenario generator, based on a modified and extended version of the Lubachevsky-Stillinger algorithm in 3D. The incompressible Navier-Stokes flow solver of the PDE framework Peano is used to compute the overall flow through the reference volume. The throughput results allow, in combination with suitable criteria, for the estimation of the necessary resolution of a simulation setup. Highly parallel runs have been performed with several thousands of compute nodes on a HPC system showing good scalability results even for this kind of unbalanced geometry data.
Original language | English |
---|---|
State | Published - 2012 |
Event | 7th International Conference on Computational Fluid Dynamics, ICCFD 2012 - Big Island, United States Duration: 9 Jul 2012 → 13 Jul 2012 |
Conference
Conference | 7th International Conference on Computational Fluid Dynamics, ICCFD 2012 |
---|---|
Country/Territory | United States |
City | Big Island |
Period | 9/07/12 → 13/07/12 |
Keywords
- Adaptive grids
- Cartesian grids
- Computational fluid dynamics
- Geometry generation
- Hpc
- Porous media