An efficient low-dissipation hybrid weighted essentially non-oscillatory scheme

Xiangyu Hu, Bin Wang, Nikolaus A. AdamsJournal of Computational Physics 2015 · DOI · arXiv

Introduction

In the field of computational fluid dynamics (CFD), simulating complex flows that contain both sharp discontinuities (shocks) and fine-scale structures (turbulence or acoustic waves) remains a significant challenge. The Weighted Essentially Non-Oscillatory (WENO) scheme is widely considered the gold standard for shock-capturing due to its non-linear adaptive nature, which prevents oscillations near discontinuities. However, WENO schemes are computationally expensive and often introduce excessive numerical dissipation, which can "smear" delicate flow features in smooth regions.

This paper presents a novel Hybrid WENO (H-WENO) scheme designed to address these limitations. By combining the robustness of WENO with the speed and low dissipation of linear schemes, the authors provide a solver that maintains high-order accuracy while significantly reducing computational overhead.

Methodology

The Hybrid Architecture

The proposed H-WENO scheme operates by dynamically switching between two distinct reconstruction modes based on the local flow physics:

  1. The WENO Mode: A 5th-order characteristic-wise WENO scheme used exclusively in regions containing shocks or steep gradients.
  2. The Linear Mode: A 5th-order component-wise linear reconstruction used in smooth regions of the flow.

Innovation: The Discontinuity Detector

The core of the H-WENO scheme is a new non-dimensional discontinuity detector, denoted as $\sigma_s$. Unlike traditional sensors that rely on simple gradients—which often fail at "critical points" where the gradient is zero but the solution is complex—the $\sigma_s$ detector is based on the resolution limit of the linear scheme.

Crucially, the detector is normalized by the local Roe-average density ($\tilde{\rho}$). This normalization ensures the scheme remains robust across varying physical scales and prevents the mathematical breakdown of the sensor in low-gradient regions. By accurately identifying where non-linear stabilization is actually required, the solver avoids unnecessary WENO calculations.

Computational Efficiency

A major efficiency gain is achieved through "component-wise" processing. While standard WENO requires projecting variables into characteristic space—a mathematically intensive process involving complex matrix operations—the linear mode in this H-WENO scheme operates directly on physical components like density and pressure. This optimization allows the solver to skip approximately 5/6 of the most expensive operations in regions identified as smooth.

Numerical Experiments

The authors validated the H-WENO scheme using a comprehensive suite of classic CFD benchmarks designed to test different aspects of numerical performance:

  • Acoustic Preservation: Broadband sound wave tests were used to measure how well the scheme preserves delicate signals over time.
  • Shock-Tube Benchmarks: Standard tests (Sod, Lax, and 1-2-3 problems) evaluated the capture of simple shocks and expansion waves.
  • High-Energy Dynamics: Interacting blast waves and shock-density wave interactions tested the scheme's stability during violent wave collisions.
  • Complex 2D Flows: The Double Mach Reflection (Mach 10) and Viscous Shock Tube tests evaluated the scheme's ability to resolve complex structures like Mach stems and lambda-shocks in the presence of fluid friction.

Key Results and Discussion

High Local Efficiency

The most significant finding is that for the majority of complex flow problems, the expensive WENO logic is required at fewer than 2% of the grid points. In 2D viscous simulations, the "local efficiency"—the percentage of the domain handled by the faster linear mode—frequently exceeded 99.9%.

Improved Accuracy and Low Dissipation

In broadband sound wave tests, the H-WENO scheme significantly outperformed standard WENO-5. By utilizing the linear mode in smooth regions, the scheme avoided the numerical dissipation that typically causes standard WENO to smear wave peaks. This allows for the preservation of fine-scale features that are often lost in purely non-linear schemes.

Robustness and Stability

Despite relying on linear mathematics for the vast majority of the domain, the H-WENO scheme remained perfectly stable even in high-Mach number scenarios, such as the Mach 10 Double Mach Reflection. Furthermore, the authors demonstrated that any minor numerical artifacts created by the switching mechanism vanish as the grid is refined, confirming the mathematical consistency of the hybrid approach.

Conclusion

The H-WENO scheme represents a substantial improvement in the efficiency of high-order CFD solvers. By utilizing a smarter, density-normalized discontinuity detector, the scheme restricts expensive characteristic-wise calculations to the absolute minimum necessary for stability. This results in a solver that is not only faster but also more accurate for simulating flows where shocks and fine-scale structures coexist. Future work may involve extending this hybrid approach to unstructured grids to accommodate complex industrial geometries and exploring its performance in vacuum or near-vacuum conditions.

Summary updated

BibTeX
@article{hu2015efficient,
  author = {Xiangyu Hu and Bin Wang and Nikolaus A. Adams},
  title = {{An efficient low-dissipation hybrid weighted essentially non-oscillatory scheme}},
  journal = {Journal of Computational Physics},
  year = {2015},
  doi = {10.1016/j.jcp.2015.08.043},
}
Figure