Cerfacs Entrez dans le monde de la haute performance...

PhD defense: Johan DEGRIGNY – « Towards the Computational Prediction of Low-Speed Buffet: Improved Wall Modeling for the Lattice-Boltzmann Method »

  Mardi 9 novembre 2021 à 14h00

  CERFACS, Toulouse - Jean-Claude ANDRÉ meeting room    


CFD (Computational Fluid Dynamics) is a trusted and widespread tool in aircraft aerodynamics for predicting flows in conditions close to cruise design points. Accurately predicting unsteady aerodynamic phenomena involving massive flow separation, however, remains a challenge that often exceeds the capabilities of the classical RANS (Reynolds-Averaged Navier–Stokes) turbulence modeling strategy, which currently is the industry standard. Low-speed buffet—the mechanical excitation of some components (e.g. the tailplane) by the wake caused by local flow separation on an upstream component (e.g. on the wing)—falls into this category of flows. Scale-resolving simulations (which explicitly resolve the largest turbulent scales) are expected to fill that capability gap, thus enabling the more cost effective development of better aircraft with shorter design cycles.

The LBM (Lattice-Boltzmann Method) appears as a good candidate for meeting the expectations placed on scale-resolving techniques applied in the frame of aircraft development processes in terms of turnaround time and handling complex geometries. Besides turbulence modeling in the fluid near the wall, wall modeling—the modeling of the flow within the inner boundary layer — is crucial to the success of the LBM, even more so than for other numerical methods. Indeed computing fully resolved high-Reynolds-number boundary layers on Cartesian grids is exceedingly costly, even with RANS modeling.

Low-speed buffet simultaneously involves both separated and attached flow, yet the modeling of the latter is much less mature with the LBM, due to the use of Cartesian grids. The aim of this thesis is thus improving the wall treatment (wall model implementation) and the near-wall turbulence modeling in the ProLB LBM solver to expand its capabilities towards low-speed buffet.

Regarding the wall treatment, four mutually complementing features are introduced: the input data for the wall model is gathered without interpolating the near-wall velocity field, the LBM boundary condition takes the velocity gradient from the wall model into account for consistency between the modeled and computed parts of velocity profiles, some nodes located very close to the wall are eliminated to avert numerical issues, and a correction is applied to the definition of the wall-normal direction for overhanging nodes occurring at sharp edges. This wall treatment is calibrated and validated on RANS computations with a simple algebraic wall model on a grid-aligned flat plate without pressure gradients and on a NACA0012 airfoil. The smoothness of surface pressure and skin friction coefficients is greatly improved. Within the limitations of that simple wall model, the accuracy of the results is also enhanced.

Regarding near-wall turbulence modeling, the LES (Large Eddy Simulation) formalism provides a framework for scale-resolving turbulence modeling, but its application in realistic cases can be complex and lack robustness. Hybrid RANS-LES models such as DES (Detached Eddy Simulation) derivatives, which combine the advantages of both turbulence modeling strategies, are thus most promising for an industry setting. The recently published advanced ZDES mode 2 (2020) model is thus implemented, and its advantages are demonstrated on a high-lift airfoil and on a generic high-lift aircraft configuration.

Keywords: Wall treatment, wall modeling, Lattice-Boltzmann Method, Hybrid RANS LES, Low-Speed Buffet


Sébastien DECK ONERA Referee
Eric LAMBALLAIS Université de Poitiers Referee
Damiano CASALINO Delft University of Technology Member
Maria Vittoria SALVETTI Università di Pisa Member
Grégoire PONT Airbus Invited member
Pierre SAGAUT Aix-Marseille Université Advisor
Jean-François BOUSSUGE Cerfacs Co advisor







Assimilation de données

Du lundi 24 juin 2024 au jeudi 27 juin 2024






Apprentissage machine pour la science des données

Du lundi 7 octobre 2024 au jeudi 10 octobre 2024






Bases théoriques pour la compréhension et l’analyse de Simulations aux Grandes Echelles compressibles

Du lundi 14 octobre 2024 au mercredi 16 octobre 2024

  Formation       Cerfacs, Toulouse, France