Analysis of the Kozeny–Carman model based on pore networks

Abstract

Abstract The Kozeny–Carman (KC) model is commonly used to predict permeability (k) but sometimes presents obvious predictive deviations. Comparatively, k prediction based on pore networks could more effectively describe the dependence of k on porous structure and accordingly shows high accuracy and reliability. This triggers us to examine the rationality of the KC assumptions about the pore structure and analyze the model applicability according to the pore networks. Two glass bead packs, a sintered glass bead plate and a Berea core are measured using computed tomography imaging and their pore parameters are accordingly calculated. Their pore parameters are obviously distinct, generally reflecting the progressively stronger roles of particle size, compaction-alike sintering and weak cementation in reducing porosity (ϕ), k, pore and pore throat radii (rp and rt). When correlating the pore parameters of the KC model to those of the pore networks, it is found that the specific surface area (av) has no specific relation to rp and rt and that pore structures could be distinct despite the equivalent av. Thus, av is insufficient to distinguish the differences in pore geometry and reflects their influence on k. An analysis of the modified KC equations of our cores and the ϕ−k relationships of some relative homogeneous reservoir rocks (e.g. silty sandstone and Fontainebleau sands) indicates that the variety of the ϕ−k relationships induced by such factors as particle size and pore radius could not be fully predicted by the KC model in some cases, especially when the ϕ is relatively low.