Session: 37-12 LES, DES and Scale Resolving Methods

Paper Number: 80315

80315 - On the Use of Kinetic-Energy Balance for the Feature-Based Mesh Adaptation Applied to Large-Eddy Simulation in Complex Geometries 

Keywords : Feature-based mesh adaptation, Large Eddy Simulation, kinetic energy, Complex geometries, statistical convergence

Large-Eddy Simulation (LES) is a numerical method based on the explicit resolution of large scales structures and the modelling of the influence of the smallest ones by a subgrid-scale (SGS) model. For the past decades, LES has proven to be a valuable tool for the prediction of unsteady turbulent flows. Nevertheless, a significant amount of time is often spent by the user designing a suitable grid. The mesh resolution of this grid is important as it implicitly imposes the cut-off frequency between the large resolved scales and the modelled ones. Static and dynamic feature-based mesh adaptation, i.e. h-refinement methods, have gained momentum on the last decade as they recently allowed to improve results by refining areas to capture specific flow features. But to the authors knowledge, these methods do not always guaranty that the budget of the kinetic energy of the mean field is mesh independent as expected. In the present paper, a framework which uses the assessment of the kinetic energy budget to refine or coarsen the mesh locally is presented.

An iterative methodology is designed in which each term of the mean velocity kinetic energy budget equation is targeted one after another. This methodology starts from a computational domain with a poor resolution, that has shown to be independent of the final grid. A first run is performed to reach convergence and to evaluate the kinetic energy balance. This initial grid is then adapted in two successive loops. The first loop is designed to ensure a constant interpolation error of the mean velocity gradients across the whole domain and to guarantee that the viscous dissipation of the mean velocity kinetic energy is independent of the mesh. The second loop refines the mesh until the overall turbulent production term is also mesh independent. The production is due to the explicitly resolved turbulent scales and the SGS model.  In each loop, additional constraints can be imposed: the non-dimensional wall unit y+ or the minimum cell size to avoid too strong time-step restrictions. The compatibility of these constraints with the satisfaction of the kinetic energy budget is verified a posteriori. The final adapted mesh satisfies two main assumptions specific to LES: the correct resolution of mean velocity fields and gradients, and a balance of mean velocity kinetic energy independent of the mesh size. The convergence of the mean and rms quantities is of tremendous importance in the proposed methodology. Thus, a hierarchical analysis of time-varying volume-averaged quantities is proposed to ensure the convergence of the flow. Finally, to improve parallel performance, the number of processors is automatically adapted during the iterative process depending on the number of cells of the grid. The whole process converges towards a final grid with minimal inputs from the user.

The whole methodology is applied and assessed on three different turbulent flows. First, an academic turbulent free jet at a Re=6,000 is studied. Once mesh convergence is reached, results are compared with DNS and experimental data. The same process is applied to two other configurations representative of industrial applications: a row of 3 jets impinging on a surface at Re=23000 and a nozzle-to-plate distance H/D=5 and the non-reacting PRECCINSTA burner. For all of these cases, flow field quantities, including velocity, velocity rms and heat transfer fields on the final grids, are in good agreement with experimental data.

Presenting Author: Adrien Grenouilloux C.O.R.I.A

Presenting Author Biography: I am a PhD candidate working for Safran Nacelles at CORIA lab (France). <br/><br/>My PhD Topic is the high fidelity simulation of the aerodynamics of a turbofan thrust reverser (TR). <br/><br/>The goal of my PhD is to develop a method to perform Large Eddy Simulations (LES) of complex industrial aerothermal configurations. <br/><br/>Main topics: Heat transfer, Mesh adaptation, Compressible flows

Authors:

Adrien Grenouilloux C.O.R.I.A
Guillaume Balarac LEGI (Laboratoire des Écoulements Géophysiques et Industriels)
Julien Leparoux SAFRAN TECH
Vincent Moureau C.O.R.I.A.
Ghislain Lartigue C.O.R.I.A.
Pierre Bénard C.O.R.I.A
Paul Ferrey Safran Nacelles
Renaud Mercier Safran Tech