Adaptive Consistency in Mimetic Finite Difference Schemes: A Residual-Based Criterion for K-orthogonal Cell Classification

  • Cha, Yoo Jin (Stanford University)
  • Duran, Omar (Stanford University)
  • Castelletto, Nicola (Lawrence Livermore National Laboratory)
  • Tchelepi, Hamdi (Stanford University)

Please login to view abstract download link

Mimetic Finite Difference (MFD) schemes ensure local conservation on general polytopal meshes but become computationally expensive when applied uniformly. In practice, many well-shaped cells satisfy the Two-Point Flux Approximation (TPFA) K-orthogonality consistency requirement, rendering full MFD stencils unnecessary on large portions of the grid. To improve efficiency, we present a residual-driven local consistency indicator for adaptive TPFA-MFD schemes. The proposed indicator evaluates operator consistency by projecting an analytical linear pressure field onto the discrete degrees of freedom and measuring the resulting cell-local residual of the MFD weak formulation. The approach is purely local, requiring no global assembly or auxiliary solvers; it relies solely on the geometry and absolute permeability associated with each cell. Validated on 2D and 3D polytopal meshes, the criterion reliably distinguishes TPFA-compatible cells from those affected by distortion or anisotropy. By enabling a priori cell classification, the methodology reduces computational overhead and improves matrix sparsity while maintaining accuracy on complex geometries, providing a scalable tool for large-scale porous media simulations.