Dors, Thomas M. R. et Groth, Clinton P. T..
2025.
« Large-eddy simulation of low- and high-speed turbulent flows using a high-order CENO finite-volume scheme ».
In Proceedings of the CSME-CFDSC-CSR 2025 International Congress (Montreal, QC, Canada, May 25-28, 2025)
Coll. « Progress in Canadian Mechanical Engineering », vol. 8.
Prévisualisation |
PDF
495 - Large-eddy simulation of low- and .pdf - Version publiée Licence d'utilisation : Tous les droits réservés aux détenteurs du droit d'auteur. Télécharger (973kB) | Prévisualisation |
Résumé
A density-based high-order Central Essentially Non-Oscillatory (CENO) finite-volume scheme in combination with minimally dissipative Riemann solver-based upwind numerical flux functions has been applied to both low- and high-speed canonical turbulent flow problems. In particular, implicit large-eddy simulation (LES) results were obtained by the numerical solution of the Favre-filtered form of the compressible Navier-Stokes equations for both the subsonic and supersonic Taylor-Green vortex (TGV) and decay of homogeneous isotropic turbulence (HIT) flows. The initial conditions for the latter were set to match those of previous low-speed wind tunnel experiments. DNS results for the HIT flow were also obtained through the implementation of a pseudo-spectral Fourier-Galerkin solver on a triply periodic cubic computational grid containing $2,048^3$ nodes which was used as a reference for the LES solutions. Published reference data was then used as a means of comparison for both of the TGV flow simulations considered in this study. For the subsonic TGV flow, the CENO scheme is shown to outperform a standard second-order finite-volume scheme by achieving a lower error in peak enstrophy at a reduced computational cost. Furthermore, the high-order CENO results obtained for the supersonic TGV flow show excellent agreement with the previous reference solutions. The high-speed TGV flow presents a unique challenge as it necessitates an accurate representation of small-scale turbulence while requiring sufficient numerical dissipation in the proximity of shocks. The CENO scheme addresses this challenge by employing a hybrid reconstruction procedure that transitions to a limited, low-order scheme for cells with under-resolved solution content. The latter are detected via a smoothness indicator, thereby ensuring that small turbulent flow scales in smooth regions are not excessively damped. Additionally, modified variants of the standard Roe flux are considered in this study and they are shown to provide significant reductions in numerical dissipation, as evidenced by the increased peak enstrophy for the TGV flow problem and a significantly improved prediction of the $-5/3$ slope in the spectral turbulence energy density compared with the results of the standard Roe flux in the case of the HIT decay problem.
| Type de document: | Compte rendu de conférence |
|---|---|
| Éditeurs: | Éditeurs ORCID Hof, Lucas A. NON SPÉCIFIÉ Di Labbio, Giuseppe NON SPÉCIFIÉ Tahan, Antoine NON SPÉCIFIÉ Sanjosé, Marlène NON SPÉCIFIÉ Lalonde, Sébastien NON SPÉCIFIÉ Demarquette, Nicole R. NON SPÉCIFIÉ |
| Date de dépôt: | 18 déc. 2025 15:09 |
| Dernière modification: | 18 déc. 2025 15:09 |
| URI: | https://espace2.etsmtl.ca/id/eprint/32386 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |