Property evaluation by numerical modelling based on voxelized images – Accuracy versus resolution

Abstract

3D imaging techniques such as Micro-CT and MRI provide insights into the sample inner structure, therefore attract increasing attention. It is common practice to conduct numerical simulations for physical properties based on 3D voxelized images. Although errors caused by voxelization in the 3D structure are inevitable, voxel scale mechanisms have not been well explained or explored. This study assesses such errors by simulations of physical processes in an ideal porous media made of simple cubic packing spheres, with known analytical solutions or well-understood properties, and its voxelized counterpart. Physical properties of the porous media including thermal conductivity, stiffness, electrical conductivity, and permeability are extracted, meanwhile considering different conditions such as cementation, freezing, and partial water saturation. Results reveal that the error for most physical properties is solely determined by the relative voxel size defined as voxel size divided by particle or pore size, rather than actual voxel size or resolution. The error does not always decrease with the increase in resolution. Errors of different properties vary several orders of magnitude and are dominated by different geometry features of representative units. Suggestions whether the evaluation of a certain physical property could be credibly conducted are given based on the relative error.