A Strongly Consistent Finite Difference Scheme for Steady Stokes Flow and its Modified Equations

Yury A. Blinkov, Vladimir P. Gerdt, Dmitry A. Lyakhov, Dominik L. Michels

Abstract

This paper introduces a novel second-order finite difference scheme for steady Stokes flow that achieves "strong consistency" (s-consistency). By leveraging symbolic computer algebra—specifically Janet and Gröbner bases—the authors construct a discrete system that algebraically preserves the underlying mathematical structure and hidden integrability conditions of the original partial differential equations (PDEs). Evaluation against the standard Marker-and-Cell (MAC) method demonstrates that this s-consistent scheme maintains high accuracy with significantly coarser grid spacings, offering a more robust and efficient approach for simulating creeping flows in complex geometries.

Introduction

In fluid dynamics, Stokes flow (or creeping flow) describes regimes where viscous forces dominate over inertial forces, such as the movement of honey or oil through porous media. While the governing equations are linear, translating them into a discrete finite difference form for numerical simulation presents a significant challenge: consistency.

Most numerical schemes are "weakly consistent," meaning they approximate individual equations as the grid spacing ($h$) approaches zero. However, they often fail to be "strongly consistent" (s-consistent). S-consistency ensures that the discrete equations inherit the algebraic properties of the original system, such as the pressure Poisson equation. Inconsistent schemes can introduce "phantom" constraints or physical errors that do not exist in the continuous model.

Methodology

The researchers developed a five-step pipeline that integrates numerical analysis with symbolic computer algebra to ensure s-consistency:

  1. Completion to Involution: Using the JANET Maple package, they identify the "Janet basis" of the PDEs to uncover hidden integrability conditions.
  2. Integral Conversion: The PDEs are transformed into an equivalent integral form over a defined control volume.
  3. Integral Relations: Exact relations for derivatives are added to link grid values together.
  4. Numerical Integration: Standard midpoint and trapezoidal rules are applied to evaluate the integrals.
  5. Difference Elimination: The LDA (Linear Difference Algebra) package is used to eliminate intermediate variables, resulting in a symmetric finite difference scheme.
The 3x3 grid stencil used for integral conversion and the construction of the strongly consistent finite difference scheme.

A critical component of this methodology is the use of Gröbner bases. This algebraic tool allows the authors to prove that the "ideal" (the set of all possible equations) generated by their discrete scheme perfectly matches the "ideal" of the original PDEs, ensuring algebraic fidelity.

Results and Evaluation

The scheme was evaluated using a numerical benchmark simulating fluid flow through a complex geometry representing porous media. Since no analytical solution exists for such geometries, ground truth was established using an extremely fine grid ($h \to 0$).

Key Findings:

  • Accuracy: The scheme is confirmed to be second-order accurate ($O(h^2)$).
  • Efficiency vs. MAC: The s-consistent scheme is significantly more robust than the industry-standard Marker-and-Cell (MAC) method. To maintain a relative error below 15%, the new scheme allows for a grid spacing 1.7 times larger than the MAC method, leading to substantial computational savings.
  • The Trap of Compactness: The authors discovered that "compact" schemes (using fewer grid points) can actually break strong consistency. Such s-inconsistency introduces artificial constraints on external forces, potentially leading to simulation failure in physically valid scenarios.

Discussion and Conclusion

This work demonstrates that numerical analysis should focus not only on Taylor series expansions (truncation error) but also on the algebraic inheritance of a system's properties. By extending the "method of modified equations" to non-evolutionary (steady-state) systems, the authors provide a new framework for verifying the accuracy of stationary flow solvers.

While currently focused on 2D steady-state flow, the next logical step is extending this s-consistency framework to 3D models and the full, non-linear Navier-Stokes equations. Future work will also aim to automate these symbolic derivations into a "black-box" tool for broader engineering applications.

Summary updated

BibTeX
@inproceedings{blinkov2018strongly,
  author = {Yury A. Blinkov and Vladimir P. Gerdt and Dmitry A. Lyakhov and Dominik L. Michels},
  title = {{A Strongly Consistent Finite Difference Scheme for Steady Stokes Flow and its Modified Equations}},
  booktitle = {CASC},
  year = {2018},
  doi = {10.1007/978-3-319-99639-4\_5},
}
Figure