Estimating permeability from thin sections without reconstruction: Digital rock study of 3D properties from 2D images

We present a new approach for predicting permeability of natural rocks using thin sections. Our approach involves two steps: (1) computing permeability of the thin sections for flow normal to the face, and (2) application of new robust 2D-3D transforms that relate thin section permeability to 3D rock permeability using calibration parameters. We perform step 1 using Lattice-Boltzmann and finite difference schemes, which are memory efficient. We discuss two models to perform step 2. Our two-step approach is fast and efficient, since it does not require reconstruction of the unknown 3D rock using 2D thin section information. We establish the applicability of this new approach using a dataset comprised of LBM-computed permeability of rock samples from various geologic formations, including Fontainebleau sandstone, Berea sandstone, Bituminous sand, and Grosmont carbonate. We find that for sandstones our approach predicts fairly accurate permeability with little calibration. Predicting permeability of carbonates from thin sections is more challenging due to microstructural complexity thus model parameters require more calibration. For general workflow, we propose to first calibrate the proposed models using the available 3D information on the rock microstructure (from microCT, SEM, etc.) and then predict the permeability for rocks from the same geological formation for which only 2D thin sections are available.