Predicting Permeability through 3d Pore-space Images Reconstructed Using Multiple-point Statistics

Pore-scale network modeling can predict multiphase flow properties with arbitrary wetting conditions if the network represents the geology of the sample accurately. Such pore-scale modeling uses topologically disordered networks that realistically represent the pore structure. To generate the network it is first necessary to have a three-dimensional voxel-based pore-space representation that is constructed by either a direct imaging technique such as micro-CT scanning, stochastic methods, or object-based approaches. Micro-CT scanning is the most promising among these three approaches since it is the most direct. However, its resolution – a few microns – means that for many rocks, particularly carbonates, significant porosity cannot be imaged. Furthermore, alternative approaches, such as reconstruction through simulating the geological processes by which the rock was formed, such as sedimentation and diagenesis, may be problematic for many materials whose depositional and diagenetic history is uncertain or complex. Statistical reconstruction is more general and is not limited by the pore size. Statistics of the pore space are obtained from readily available experimental data such as thin-section images. Using only single and two-point statistics in the reconstruction often underestimates the pore connectivity, especially for low porosity materials.