TL;DR
This paper introduces a new high-order gas-kinetic scheme (HGKS-cur) designed for simulating complex, compressible turbulent flows in arbitrary geometries. By combining advanced numerical techniques with a robust gas-kinetic framework, the authors demonstrate its capability to accurately capture turbulent phenomena in challenging scenarios like channel flows and flows over periodic hills, paving the way for more reliable simulations in aerospace and other engineering fields.
What Problem Does This Paper Address?
Simulating turbulent flows, especially compressible ones with complex geometries, is a grand challenge in fluid dynamics. Traditional methods struggle with the vast range of scales involved and the need for high accuracy, particularly in applications like aerospace where understanding phenomena like shock-boundary interactions is critical. While Direct Numerical Simulation (DNS) offers the highest fidelity, its computational cost is prohibitive for most practical problems. Large Eddy Simulation (LES) offers a compromise, but its accuracy heavily relies on the underlying numerical schemes. Existing high-order schemes often face issues with stability, accuracy in curvilinear coordinates, or handling discontinuities. This paper aims to bridge this gap by developing a high-order gas-kinetic scheme that is accurate, robust, and applicable to complex geometries and compressible turbulent flows.
Methods
The core of this work is the development of a High-Order Gas-Kinetic Scheme in General Curvilinear Coordinate (HGKS-cur). Key methodological aspects include:
- Coordinate Transformation: The governing equations (Bhatnagar-Gross-Krook, BGK) are transformed from physical space to a computational curvilinear coordinate system.
- Geometric Metrics Reconstruction: For meshes defined by discrete points (not analytical transformations), geometric metrics are reconstructed using dimension-by-dimension Lagrangian interpolation to ensure geometrical conservation.
- High-Order Spatial Reconstruction: A multidimensional Weighted Essentially Non-Oscillatory (WENO) reconstruction is employed to achieve high spatial accuracy. This involves reconstructing cell-averaged Jacobian and Jacobian-weighted conservative variables.
- Two-Stage Fourth-Order Temporal Discretization: A robust two-stage fourth-order time-stepping scheme, previously developed for spatial-temporal coupled solvers, is used for temporal accuracy.
- Implicit Large Eddy Simulation (iLES): The HGKS-cur is implemented within an iLES framework. The inherent numerical dissipation of the scheme acts as the subgrid-scale (SGS) dissipation, eliminating the need for explicit SGS models.
- Gas-Kinetic Solver: The scheme leverages the gas-kinetic approach, which recovers inviscid and viscous fluxes from a time-dependent gas distribution function at cell interfaces, providing a unified treatment of different flow regimes.
Datasets
The paper uses established benchmark cases for validation:
- 3D Advection of Density Perturbation: Used for accuracy tests and validation of the geometric conservation law on both structured and unstructured meshes.
- Lid-Driven Cavity Flow: A standard benchmark for laminar flow simulations, used to test the scheme's ability to handle steady and unsteady laminar flows in 3D.
- Compressible Turbulent Channel Flow: A well-studied problem used to validate the scheme's performance for compressible wall-bounded turbulence, comparing results with DNS and other LES approaches.
- Compressible Turbulent Flow Over Periodic Hills: A challenging case involving massive flow separation, used to test the scheme's capability in handling complex geometries and separated flows.
No new datasets were introduced; instead, existing benchmark problems were used to validate the developed numerical scheme.
Key Insights and Ideas
- Geometric Conservation is Crucial: Accurately preserving the geometric conservation law (GCL) is vital for simulations on non-orthogonal, curvilinear meshes. The paper demonstrates that their Lagrangian interpolation approach for geometric metrics successfully maintains this.
- WENO Reconstruction for Accuracy: The use of WENO reconstruction is shown to be effective in achieving high spatial accuracy, even in the presence of complex flow features.
- iLES with HGKS-cur is Promising: The inherent numerical dissipation of the HGKS-cur acts as an effective SGS model for iLES, simplifying the simulation setup and demonstrating good agreement with established methods for turbulent flows.
- Prandtl Number Matters: For compressible turbulence, the Prandtl number significantly influences statistical quantities. The paper highlights the importance of using the correct Prandtl number for accurate simulations.
- Computational Domain Size: For turbulent channel flows, an adequate streamwise computational domain is essential for capturing the full range of turbulent structures and obtaining accurate statistics.
Key Results
- Accuracy Validation: The HGKS-cur achieved expected high-order accuracy (close to fifth order) in the density perturbation advection tests.
- Geometric Conservation: The scheme demonstrated excellent preservation of the geometric conservation law, with errors reducing to machine zero on refined meshes.
- Laminar Flow Performance: The scheme accurately simulated both steady and unsteady laminar flows in the lid-driven cavity benchmark.
- Turbulent Flow Agreement: For compressible turbulent channel flow and flow over periodic hills, the iLES results obtained with HGKS-cur showed good agreement with reference DNS and spectral/finite volume methods.
- Robustness: The scheme proved robust in handling complex geometries and compressible turbulent flows, including massive flow separation.
Limitations and Future Work
- Limitations:
- The paper notes that for the compressible turbulent channel flow, the smaller streamwise computational domain (case G2) slightly altered statistical quantities and obviously deviated root-mean-square values, suggesting the need for sufficient domain length.
- While the scheme is robust, the paper mentions that for supersonic turbulent channel flow, spectral methods still work well due to the absence of strong shocklets. The true advantage of HGKS-cur might be more pronounced in cases with stronger discontinuities.
- Future Work:
- The authors plan to investigate more challenging compressible wall-bounded turbulence problems, such as supersonic and hypersonic flat plate turbulent boundary layers.
- They also intend to explore higher Mach numbers and different flow configurations.
- Further investigation into the effect of the Prandtl number and computational domain size on turbulent statistics is implied.
Summary updated