Comparative analysis on pore-scale permeability prediction on micro-CT images of rock using numerical and empirical approaches

Varieties of pore-scale numerical and empirical approaches have been proposed to predict the rock permeability when the pore structure is known, for example, microscopic computerized tomography (micro-CT) technology. A comparative study on these approaches is conducted in this paper. A reference dataset of nine micro-CT images of porous rocks is generated and processed including artificial sandpacks, tight sandstone, and carbonate. Multiple numerical and empirical approaches are used to compute the absolute permeability of micro-CT images including the image voxel-based solver (VBS), pore network model (PNM), Lattice Boltzmann method (LBM), Kozeny-Carman (K-C) equation, and Thomeer relation. Computational accuracy and efficiency of different numerical approaches are investigated. The results indicate that good agreements among numerical solvers are achieved for the sample with a homogeneous structure, while the disagreement increases with an increase in heterogeneity and complexity of pore structure. The LBM and VBS solver both have a relative higher computation accuracy, whereas the PNM solver is less accurate due to simplification on the topological structure. The computation efficiency of the different solver is generally computation resources dependent, and the PNM solver is the fastest, followed by VBS and LBM solver. As expected, empirical relation can over-estimate permeability by a magnification of 50 or more, particularly for those strong heterogeneous structures reported in this study. Nevertheless, empirical relation is still applicable for artificial rocks.