| Autor: |
Sven Linden, Hans Hagen, Andreas Wiegmann |
| Rok vydání: |
2014 |
| Předmět: |
|
| Zdroj: |
Scopus-Elsevier |
| ISSN: |
2214-4609 |
| DOI: |
10.3997/2214-4609.20140834 |
| Popis: |
We introduce a novel approach to compute the permeability on very large CT images of core samples. The LIR-tree is used for spatial partitioning of the geometry. The images are coarsened in areas where the velocity and pressure do not vary much while keeping the original resolution near the solid surfaces. In addition, solid regions do not occupy computational memory. Variables are arranged in a way such that each cell is able to satisfy the Stokes-equations independently from its neighbor cells. Pressure and velocity are discretized on staggered grids but instead of using one velocity variable on the cell faces we introduce two velocity variables. They discretize the two one-sided limits at the center of the cell surface. The discretization of the momentum and mass conservation equations yields one small linear system per cell. This structure allows to use the block Gaus-Seidel-algorithm as iterative solver. We compare our method to three other solvers on a large complex Berea sandstone dataset. The method is 3-6 times faster, scales well for up to 32 processors and has very low memory requirements. An additional benefit of this work is that the range of permeabilities from the benchmark can be narrowed down significantly. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
