Image-based simulations of absolute permeability with massively parallel pseudo-compressible stabilised finite element solver

摘要

We apply an accurate parallel stabilised finite element method to solve for Navier-Stokes equations directly on a binarised three-dimensional rock image, obtained by micro-CT imaging. The proposed algorithm has several advantages. First, the linear equal-order finite element space for velocity and pressure is ideal for presenting the pixel images. Second, the algorithm is fully explicit and versatile for describing complex boundary conditions. Third, the fully explicit matrix-free finite element implementation is ideal for parallelism on high-performance computers, similar to lattice Boltzmann. In the last, the memory usage is low compared with lattice Boltzmann or implicit finite volume. We compute the permeability of a range of rock images. The stabilisation parameter may affect the velocity, and an optimal parameter is chosen from the numerical tests. The steady state results are comparable with lattice Boltzmann method and implicit finite volume. The transient behaviour of pseudo-compressible stabilised finite element and lattice Boltzmann method is very similar. Our analysis shows that the stabilised finite element is an accurate and efficient method with low memory cost for the image-based simulations of flow in the pore scale up to 1 billion voxels on 128-GB ram workstation and on distributed clusters.