Explicit numerical simulation of fluids in reconstructed porous media

Substantial efforts are currently underway across the community to deploy in silico methods that can aid the development and validation of closure models for applied analysis of fluid transport and related phenomena in porous media and the determination of associated closure parameters. This paper represents a contribution to these efforts. We report, for the first time, a systematic comparison of a lattice-gas automata based explicit numerical simulation (ENS) tool for the flow of single-phase fluids in porous media with experiment. In particular, the study involved comparison of an experimentally determined permeability for a packed bed of spherical particles of mean size d _p = 464 μm with those determined using ENS on representative volumes of the bed taken from: (1) α ≈ 30 μm resolution binarized 3D NMR images of the packing; and (2) 3D volumes generated by the reconstruction method of Joshi-Quiblier-Adler (JQA) at resolutions of nα for n = 1.2 and 4. Permeabilities were underpredicted in all cases depending on the details of the representative volume. Then n = 2 and 4 resolution reconstructions on volumes of 125d³_p yielded permeabilities 2.4 and 1.7 smaller than experiment, respectively. The differences between the permeabilities predicted for the n = 1 reconstructions and the binarized NMR images and comparison of all the reconstructions with the binarized NMR images allowed these under-predictions to be attributed to the enhancement of surface roughness and fine structure and the loss of underlying collector shape in the JQA reconstructions.