• PhD,

PhD defense - Sajad Mozaffari - ED SPI

You are cordially invited to attend the Ph.D. defense by Sajad Mozaffari on the subject "Adaptive Grid Refinement for Hybrid RANS/LES".

on May 18, 2020

The defense will take place by videoconference on Monday 18 May at 2:15 pm (pending official authorisation from the Ecole Doctorale).

Committee
The thesis will be defended before the following committee:
- Rickard BENSOW, Professeur, Chalmers University of Technology (rapporteur)
- Kai SCHNEIDER, Professeur, Université d’Aix-Marseille (rapporteur)
- Paola CINNELLA, Professeur, Arts et Métiers ParisTech - ENSAM (examinateur)
- Eric LAMBALLAIS, Professeur, Université de Poitiers (examinateur)
- Marta DE LA LLAVE PLATA, Ingénieur de recherche, ONERA (examinateur)
- Michel VISONNEAU, Directeur de recherche CNRS, École Centrale de Nantes/CNRS (directeur de thèse)
- Jeroen WACKERS, Ingénieur de recherche HDR, École Centrale de Nantes/CNRS (co-encadrant)

Abstract
Taking advantage of the capabilities of RANS and LES models, hybrid RANS/LES models are suitable for the simulation of high Reynolds number flow with complex physical phenomena and geometries. However, due to the dependency of the behavior of these models on the local grid size, the generation of mesh with the right resolution is crucial. For a complex hybrid RANS/LES simulation, the mesh generation is a time- and effort-consuming step. An adaptive refinement process is an attractive alternative, but requires the consideration of mesh change effects on the performance of the model.

This thesis focuses on the development of adaptive grid refinement based on averaging and its integration in the ISIS-CFD flow solver. The aim is to obtain an adapted mesh which has a generally static topology based on the main flow features.

First, the dependency of hybrid RANS/LES models on the mesh changes and the effect of a refinement interface on the turbulence production and destruction are considered. Then, time-averaging of the instantaneous solutions over various intervals is carried out to filter the flow fluctuations in the mean solution based on the frequencies of the flow features. A refinement procedure is then developed based on two averaging strategies over instantaneous refinement criteria: the sliding window averaging, and the averaging over the whole computational time. The proposed adaptation process is assessed by performing a DDES-based simulation of a ship in drift in order to reproduce the solution on a reference fine mesh. The average-based adaptation successfully follows the main flow features and creates stable refined grids around the main vortices. Thus, the adaptive approach can be an alternative for the manual mesh generation process of hybrid RANS/LES simulations, especially with complex geometries.
Published on May 14, 2020 Updated on May 18, 2020