Error assessment of reconstructed 3D Digital Replica Models: From Computed Tomography data to pore-scale simulations

The application of flow simulations on porous media, reconstructed through Computerized Tomography (CT) scans, has emerged as a prevalent methodology for the computation of rock permeability. However, constructing a proper 3D model of a rock sample is areal challenge, mainly due to the lack of a unified procedure. Indeed, to ensure precise outcomes, specific prerequisites must be fulfilled. This paper proposes a methodology to assess the convergence and accuracy of computed solutions from CT data to pore-scale simulations. Starting from 3D volume data obtained by X-ray CT, we develop a workflow to investigate the effects of the reconstructed shape on the permeability of a granular porous medium composed of glass beads. Indeed, the choices of CT scan resolution and digital rock discretization can compromise the quality and computational cost of numerical results. Especially in configurations of porous media with high solid volume fractions and very narrow porous spaces, as observed in solid/solid contact zones, which can be either under or over-resolved depending on the numerical tools used. Highly resolved Direct Numerical Simulations (DNS) are conducted to solve incompressible Navier-Stokes equations through porous media. Body-fitted meshes are employed to resolve irregular shapes accurately, ensuring precise results even with coarser meshes. The methodology is validated with challenging simulations of flows through simple cubic close packing of particles, incorporating various geometric surface modeling techniques. A convergence of the results with respect to grid resolution is obtained for low- to moderate-Reynolds-number flows. The numerical results indicate that permeability calculation strongly depends on surface processing. Finally, we apply these recommendations to construct accurate digital replica models generated from CT data of our assembly of randomly arranged glass beads in a tube. The study of the pressure drop convergence demonstrates an excellent agreement with the empirical correlation.