TL;DR
This paper introduces a novel hybrid Discontinuous Galerkin (DG) and Finite Volume (FV) numerical method for solving the 2D Euler equations. This approach achieves third-order accuracy while significantly reducing computational cost (CPU time and memory) compared to traditional DG methods, making it more practical for complex fluid dynamics simulations.
What Problem Does This Paper Address?
Simulating compressible fluid flows, especially those involving complex phenomena like turbulence or aeroacoustics, often requires high-order numerical methods to accurately capture fine flow structures and minimize numerical diffusion. While methods like Discontinuous Galerkin (DG) offer high-order accuracy and geometric flexibility, they suffer from substantial computational costs (memory and CPU time). Traditional Finite Volume (FV) methods are more efficient but generally limited to lower orders of accuracy. This paper aims to bridge this gap by developing a hybrid DG/FV method that leverages the strengths of both approaches to achieve high-order accuracy with improved computational efficiency.
Methods
The core innovation lies in the "hybrid reconstruction" strategy. The authors extend their previously developed concept of "static" and "dynamic" reconstruction. In their hybrid DG/FV scheme:
- Dynamic Reconstruction (DG-based): Lower-order derivatives of the piecewise polynomial solution within a cell are computed using the traditional DG method. This is done locally within each cell.
- Static Reconstruction (FV-based): Higher-order derivatives are then reconstructed using information from the cell itself and its neighboring cells, similar to FV methods. This "static" reconstruction is applied after the dynamic reconstruction of lower-order terms.
This hybrid approach uses Taylor basis functions for the DG part and a Green-Gauss formulation for the FV-based static reconstruction of higher-order derivatives. The method is applied to solve the 2D Euler equations on unstructured and hybrid grids.
Datasets
Not discussed in the paper. The paper focuses on numerical method development and validation using standard fluid dynamics test cases.
Key Insights and Ideas
- Hybrid Reconstruction Synergy: The combination of "dynamic" (local, DG) and "static" (neighborhood-based, FV) reconstruction effectively balances accuracy and computational cost. By computing only the lower-order terms dynamically, the method avoids the full computational burden of a pure DG scheme.
- Efficiency Gain: The static reconstruction of higher-order derivatives, leveraging information from neighboring cells, significantly reduces the need for extensive local computations, leading to substantial savings in CPU time and memory.
- Compactness Preservation: Despite using information from neighboring cells for static reconstruction, the method aims to retain a degree of "compactness," which is crucial for efficient parallelization.
- Taylor Basis Advantage: The use of Taylor basis functions is highlighted for its ability to handle arbitrary element shapes and facilitate hierarchical approaches, which are beneficial for complex geometries and adaptive meshing.
Key Results
- Third-Order Accuracy: The hybrid DG/FV method consistently achieves the desired third-order accuracy across various test cases and grid types (regular, irregular, and hybrid).
- Significant Cost Reduction: Compared to traditional third-order DG methods (DGM3), the hybrid DG/FV-1 scheme shows substantial reductions in computational cost. For example, in the vortex evolution problem, memory and CPU time were reduced by approximately 30% and 50%, respectively.
- Accurate Flow Feature Capture: The method successfully captures complex flow phenomena, including vortex evolution, shock-vortex interaction, subsonic flow past cylinders and airfoils, and transonic flow past an airfoil, demonstrating its robustness and accuracy.
- Improved Dissipation Handling: For long-time simulations (e.g., vortex evolution), the hybrid method, like DGM3, shows significantly less dissipation error compared to the second-order DG method (DGM2).
Limitations and Future Work
- Limiter for Strong Shocks: The paper notes that for typical cases with weak shocks, a simple Barth limiter is sufficient. However, for flows with strong shocks, a more robust limiter for higher-order reconstruction will be needed.
- Extension to Navier-Stokes: The authors suggest extending the method to Navier-Stokes equations, which would involve handling viscous terms.
- Implicit Methods and hp-Multigrid: For accelerating convergence, especially in future work on implicit methods and hp-multigrid approaches, further study is recommended.
- Three-Dimensional Applications: The authors anticipate the hybrid DG/FV scheme to be even more efficient for three-dimensional problems and suggest parallel computation as a necessary step for such applications.
- Alternative Reconstruction: While the Green-Gauss formulation is used for static reconstruction, the paper mentions that least-square approaches have also been explored and could be adopted.
Summary updated