High-order ALE gas-kinetic scheme with WENO reconstruction

Liang Pan, Fengxiang Zhao, Kun XuJournal of Computational Physics 2020 · DOI

TL;DR

This paper introduces a novel high-order Arbitrary Lagrangian-Eulerian (ALE) gas-kinetic scheme for simulating inviscid and viscous flows. It overcomes limitations of traditional ALE methods by eliminating rezoning and remapping steps, and achieves improved accuracy and robustness on distorted meshes through an unstructured WENO reconstruction.

What Problem Does This Paper Address?

Simulating fluid dynamics often involves complex geometries and moving boundaries, which are challenging for standard Eulerian methods that use fixed grids. Lagrangian methods, where the grid moves with the fluid, can handle these situations better but suffer from mesh distortion and tangling. The Arbitrary Lagrangian-Eulerian (ALE) approach offers a compromise by allowing the mesh to move independently of the fluid, but traditional ALE methods involve computationally expensive rezoning and remapping steps to maintain mesh quality. Furthermore, achieving high-order accuracy in ALE schemes, especially on distorted meshes, is difficult because the mesh velocity variation can violate geometric conservation laws. This paper addresses these challenges by developing a high-order ALE gas-kinetic scheme that integrates an unstructured WENO reconstruction to handle mesh distortion and avoids explicit rezoning/remapping.

Methods

The core of this work is a high-order ALE gas-kinetic scheme built upon the Bhatnagar-Gross-Krook (BGK) model. Key methodological contributions include:

  • Two-Stage Fourth-Order Temporal Discretization: This method is adapted for the moving mesh framework, providing temporal accuracy without requiring explicit rezoning or remapping.
  • Unstructured WENO Reconstruction: A third-order Weighted Essentially Non-Oscillatory (WENO) reconstruction, originally developed for quadrilateral and triangular meshes, is extended to the gas-kinetic scheme. This is crucial for maintaining accuracy and robustness on distorted meshes.
  • Local Moving Coordinate Reconstruction: The WENO reconstruction is performed at each Gaussian quadrature point in a local moving coordinate system. This accounts for the variation of mesh velocity along cell interfaces, ensuring geometric conservation law is preserved even with significant mesh deformation.
  • Gas-Kinetic Solver: The second-order gas-kinetic solver is used for flux evaluation, recovering both inviscid and viscous fluxes from a time-dependent gas distribution function.

Datasets

Not discussed in the paper. The paper focuses on developing and validating a numerical method using established benchmark problems.

Key Insights and Ideas

  • Eliminating Rezoning/Remapping: The proposed scheme bypasses the need for explicit rezoning and remapping steps, which are computationally intensive and can introduce errors in traditional ALE methods. This significantly simplifies the implementation and potentially improves efficiency.
  • Robustness on Distorted Meshes: The integration of an unstructured WENO reconstruction is a key innovation that allows the scheme to maintain high-order accuracy and robustness even when the computational mesh becomes severely distorted.
  • Geometric Conservation Law Preservation: By performing reconstruction in local moving coordinates and accounting for mesh velocity variations, the scheme accurately preserves the geometric conservation law, which is critical for the stability and accuracy of ALE methods.
  • Unified Framework: The method provides a unified framework for both inviscid and viscous flows, and can handle both static and moving meshes, offering a versatile tool for CFD simulations.

Key Results

The paper presents numerical examples to validate the scheme's performance.

  • Accuracy Tests: Accuracy tests on stationary and moving meshes (quadrilateral and triangular) demonstrate that the scheme achieves the expected third-order spatial accuracy and fourth-order temporal accuracy. For instance, in Table 1, the L¹ and L² errors for the density perturbation test show convergence orders close to 3.
  • Moving Mesh Performance: The scheme performs well with moving meshes, including those with significant deformation (Type-I and Type-II). Tables 5 and 6 show that the expected accuracy is maintained.
  • Geometric Conservation Law: Table 7 shows that the geometric conservation law is well preserved, with errors reducing to machine zero for various moving mesh scenarios.
  • Benchmark Problem Validation: The scheme successfully simulates various benchmark problems, including Riemann problems (Sod, Lax, two-dimensional Riemann problems), a triple-point problem, a Sedov blast wave problem, and a viscous shock tube problem. The numerical results generally agree well with exact solutions or reference data. For the viscous shock tube, the predicted primary vortex heights are comparable to other advanced schemes.

Limitations and Future Work

  • Computational Cost: While rezoning and remapping are avoided, high-order reconstruction and temporal discretization can still be computationally intensive.
  • Development of Higher-Order WENO: The paper mentions that the current scheme is more dissipative than fifth-order WENO reconstructions on structured meshes due to the nonlinear combination of linear polynomials. The authors note that the development of higher-order WENO on unstructured meshes is ongoing.
  • Extension to More Complex Geometries: While the unstructured WENO is a significant step, further work might be needed to handle extremely complex or evolving geometries with very poor mesh quality.

The authors suggest that the ongoing development of higher-order WENO on unstructured meshes will further enhance the scheme's capabilities.

Summary updated

BibTeX
@article{pan2020high,
  author = {Liang Pan and Fengxiang Zhao and Kun Xu},
  title = {{High-order ALE gas-kinetic scheme with WENO reconstruction}},
  journal = {Journal of Computational Physics},
  year = {2020},
  doi = {10.1016/j.jcp.2020.109558},
}
Figure