Lax-Wendroff Flux Reconstruction on adaptive curvilinear meshes with error based time stepping for hyperbolic conservation laws

Arpit Babbar, Praveen Chandrashekar

TL;DR

This paper extends the Lax-Wendroff Flux Reconstruction (LWFR) method to handle complex, adaptive curvilinear meshes for solving hyperbolic conservation laws. It introduces an error-based time-stepping scheme that automates time step selection, reducing manual tuning, and demonstrates its effectiveness on challenging fluid dynamics problems.

What Problem Does This Paper Address?

Solving hyperbolic conservation laws, which describe phenomena like fluid flow, often requires numerical methods that can handle complex geometries and adapt to rapidly changing solution features (like shocks). Traditional methods struggle with:

  • Curvilinear Meshes: Standard methods are often designed for simple rectangular grids. Adapting them to curved boundaries and complex mesh structures is challenging.
  • Adaptive Mesh Refinement (AMR): To efficiently capture localized features (e.g., shocks), meshes need to dynamically refine in some areas and coarsen in others. This creates non-conforming interfaces (hanging nodes) that complicate conservation and accuracy.
  • Time Stepping: Explicit numerical methods require small time steps to remain stable, often dictated by a Courant-Friedrichs-Lewy (CFL) condition. For curvilinear meshes, determining an optimal CFL number is difficult and requires manual tuning. This is especially problematic when dealing with complex flow phenomena.
  • Admissibility Preservation: Numerical solutions must remain physically realistic (e.g., non-negative density and pressure). Schemes need to prevent non-physical oscillations, especially near discontinuities.

This paper aims to provide a unified, robust, and efficient numerical framework that addresses all these challenges for hyperbolic conservation laws.

Methods

The authors extend the Lax-Wendroff Flux Reconstruction (LWFR) method, a high-order, single-stage, quadrature-free approach, by incorporating several key advancements:

  • Curvilinear Grid Extension: The LWFR scheme is reformulated to work on general curvilinear, body-fitted grids by transforming the governing equations to a reference element. This leverages existing techniques for handling metric terms and ensuring free-stream preservation.
  • Adaptive Mesh Refinement (AMR): The Mortar Element Method (MEM) is employed to handle non-conforming interfaces arising from AMR. This method ensures conservation across these interfaces.
  • Subcell Blending Limiter: The subcell-based blending limiter, previously developed for Cartesian grids, is extended to curvilinear meshes. This limiter is crucial for capturing shocks and preserving admissibility (physical realism) by adding controlled dissipation where needed.
  • Error-Based Time Stepping: A novel embedded error-based time stepping method is developed for LWFR. Instead of relying on a CFL condition, it uses the difference between two approximations of the time-averaged flux to estimate local error and automatically adjust the time step. This significantly reduces manual tuning.

Datasets

Not discussed in the paper. The paper uses standard test cases from the fluid dynamics literature to evaluate the method.

Key Insights and Ideas

  • Unified Framework: The paper successfully integrates LWFR with curvilinear meshes, AMR, and an automatic error-based time-stepping scheme, offering a powerful tool for complex fluid dynamics simulations.
  • Error-Based Time Stepping Automation: The developed error-based time stepping method is a significant contribution, automating the often tedious and problem-dependent process of selecting stable and accurate time steps, especially on curvilinear grids where CFL analysis is unreliable.
  • Admissibility Preservation on Complex Meshes: The extension of the subcell blending limiter to curvilinear AMR meshes demonstrates that physical realism can be maintained even in highly complex computational scenarios.
  • Free Stream Preservation on Curvilinear Grids: The theoretical and numerical verification of free-stream preservation on curvilinear grids, even with AMR, is crucial for ensuring the accuracy of steady-state solutions.

Key Results

  • Validation on Euler Equations: The proposed method is tested on the 2D compressible Euler equations, a standard benchmark for fluid dynamics.
  • Successful AMR Implementation: The AMR strategy effectively refines meshes in regions of interest (e.g., near shocks or small-scale structures) while maintaining conservation.
  • Effective Error-Based Time Stepping: The error-based time stepping scheme demonstrated superior performance compared to CFL-based methods in terms of reduced manual tuning and often fewer total time steps, especially for curved meshes. For instance, in the supersonic flow over a cylinder test, it achieved a 1.51x speedup with tolE=1e-6 and a 2.08x speedup with tolE=5e-6.
  • Robustness: The scheme successfully handles problems with shocks, discontinuities, and complex flow phenomena like Kelvin-Helmholtz instability and Mach reflections.
  • Accuracy: The method shows optimal convergence rates on domains with curved boundaries and meshes.

Limitations and Future Work

  • 3D Extension: The paper notes that some issues related to free-stream preservation with AMR might arise in 3D, which is beyond the scope of this 2D study.
  • Admissibility at Non-Solution Points: While the method ensures admissibility at solution points and element means, ensuring it at all points within an element (not just solution points) is an ongoing area of research, addressed by a scaling limiter.
  • Computational Cost of AMR: While AMR is efficient, the refinement and coarsening process itself adds computational overhead.
  • Future Work: The authors suggest exploring further improvements in error estimation, potentially extending the method to other types of PDEs, and investigating its performance in 3D.

Summary updated

Abstract

Lax-Wendroff Flux Reconstruction (LWFR) is a single-stage, high order, quadrature free method for solving hyperbolic conservation laws. This work extends the LWFR scheme to solve conservation laws on curvilinear meshes with adaptive mesh refinement (AMR). The scheme uses a subcell based blending limiter to perform shock capturing and exploits the same subcell structure to obtain admissibility preservation on curvilinear meshes. It is proven that the proposed extension of LWFR scheme to curvilinear grids preserves constant solution (free stream preservation) under the standard metric identities. For curvilinear meshes, linear Fourier stability analysis cannot be used to obtain an optimal CFL number. Thus, an embedded-error based time step computation method is proposed for LWFR method which reduces fine-tuning process required to select a stable CFL number using the wave speed based time step computation. The developments are tested on compressible Euler's equations, validating the blending limiter, admissibility preservation, AMR algorithm, curvilinear meshes and error based time stepping.
BibTeX
@article{babbar2024lax,
  author = {Arpit Babbar and Praveen Chandrashekar},
  title = {{Lax-Wendroff Flux Reconstruction on adaptive curvilinear meshes with error based time stepping for hyperbolic conservation laws}},
  journal = {Journal of Computational Physics},
  year = {2024},
  doi = {10.1016/j.jcp.2024.113622},
}
Figure