Effect of X-ray CT resolution on the quality of permeability computation for granular soils: definition of a criterion based on morphological properties
In this study, the quality of soil permeability estimation based on computational fluid dynamics is discussed. Two types of three-dimensional geometries were considered: an image of Fontainebleau sand obtained from X-ray computed micro-tomography and a virtual pack of spheres. Numerical methods such as finite difference or lattice Boltzmann can conveniently use the image voxels as computational mesh elements. In this framework, the image resolution is directly associated with quality of the numerical computation. A higher resolution should promote both a better morphological description and discretisation. However, increasing the resolution may prevent the studied volume from being representative. Here, each sample was scaled and analysed at five resolutions. The dependence of soil properties with respect to the image resolution is discussed. As resolution decreased, the permeability and specific surface values tended to diverge from the reference value. This deterioration could be attributed to the shift of the pore size distribution towards badly resolved pores in the voxelised geometry. As long as granular soils are investigated, the volume fraction of pores smaller than six voxels in diameter should not exceed 50% to ensure the validity of permeability computation. In addition, based on an analysis of flow distribution, the volume fraction of pores smaller than four voxels should not exceed 25% in order to limit the flow rate occurring in badly discretised pores under 10%. For the Fontainebleau sand and virtual pack of spheres, the maximum voxel size meeting this criterion corresponded to 1/14 and 1/20 of the mean grain size respectively.