Simulating Flow in Porous Media Using the Lattice Boltzmann Method: Intercomparison of Single-Node Boundary Schemes from Benchmarking to Application
Xuhui Meng,Liang Wang,Weifeng Zhao,Xiaofan Yang
DOI: https://doi.org/10.1016/j.advwatres.2020.103583
IF: 5.361
2020-01-01
Advances in Water Resources
Abstract:Lattice Boltzmann method (LBM) is widely adopted in simulating flow and reactive transport in porous media due to its easy treatment of boundaries and high computational efficiency. There have been various boundary schemes developed in the LBM due to their vital roles when dealing with fluid-solid interface. In the present work, we aim to review and intercompare all the single-node boundary schemes (including the existing schemes and our newly-developed schemes) in the LBM through a series of numerical experiments from simple benchmarking cases to realistic applications. Both the accuracy and computational efficiency of all the schemes are evaluated and analyzed. Among the existing schemes, it is known that the bounce-back scheme (BB) is commonly employed at the boundaries in the LBM. However, the BB is of only first-order spatial accuracy when zero-velocity is not exactly located in the middle of the fluid and solid nodes, which cannot be guaranteed in porous media flows due to the complex pore structures. Therefore, fine grids are usually generated to obtain satisfactory results yet computationally expensive. Recently, a second-order single-node boundary scheme (SSN), which is more flexible than the BB while retaining the locality, has been proposed (Zhao et al, J. Comput. Phys., 329:1-15, 2017 Zhao and Yong (2017)). Unfortunately, the SSN fails to provide a viscosity-independent permeability, i.e., the predicted permeability is related to the fluid viscosity, which is unphysical. Based on the SSN, we first developed a magic second-order single-node boundary scheme (MSSN) to overcome this shortcoming based on the Chapman-Enskog analysis of the two-relaxation-time lattice Boltzmann equation (TRT-LBE) model. Comparative study of the BB, SSN/MSSN, and another single-node boundary scheme, i.e., the improved bounce-back scheme (IBB)/the magic improved bounce-back scheme (MIBB) is then performed both theoretically and numerically. Numerical simulations of porous media flows in both two and three dimensions confirm that the BB, MIBB, and MSSN can obtain the viscosity-independent permeability, while the IBB and SSN fail to do so. First of all, as benchmarking tests, flow through a 2D fracture is studied using different schemes, the results of which show that the IBB and SSN cannot provide viscosity-independent permeability while such drawbacks can be overcome by the BB/MIBB/MSSN. Furthermore, simulations of flows through a hexagonal array of cylinders and a 3D sphere packing demonstrate that the MSSN is more accurate than both the BB and MIBB. Finally, for porous media samples generated from the micro-CT images where the locations of the boundaries are not exactly known, the MSSN is more flexible to locate the positions of the fluid-solid interface than the BB and MIBB, which in turn can provide more accurate descriptions of the fluid flow. The current intercomparison study will benefit the community by providing comprehensive review and evaluation of the performance when using different single-node boundary schemes in the LBM.