Pore-Network Modeling vs. Direct Numerical Simulation: a Comparative Study

Pore-scale models of flow and transport fall into one of two broad categories: (a) direct numerical simulators (DNS) and (b) pore-network models (PNM). The former is more fundamental as it solves the governing equations on the "actual" pore space geometry obtained through some kind of imaging technology (e.g., µCT). Its drawback is that it is computationally very expensive. PNM, however, reduces the complex pore-space geometry into a "ball-and-stick" network representation, which makes it highly efficient. But geometric simplifications are accompanied by secondary simplifications in the flow and transport physics, which result in a loss of predictive accuracy. We perform one-to-one comparisons between PNM and DNS simulations (i.e., on the same porous media) to assess the impact of such simplifications on macroscopic single-phase transport dynamics. DNS simulations are performed using the popular OpenFOAM software, while our PNM utilizes a particle-tracking approach. The influence of order and disorder in the pore space morphology on the accuracy of PNM predictions is discussed.