Embedded geometry is a family of techniques dedicated to engineering problems involving remeshing on the fly during the simulation, or mesh generation in complex geometries. Embedded geometry has gained attention over the recent years due to its ability to shift a significant part of the burden of the geometrical description and/or discretization from the CAD tool to the simulation tool. The work presented in this thesis has led to a “purely” mesh-based technique to handle immersed geometries in Computational Fluid Dynamic problems. The proposed technique relies upon anisotropic mesh adaptivity in the vicinity of the embedded interface. The mesh adaptation is driven by the Hessian of the flow solution, and it contributes to improve the accuracy of the computation. An overview of existing methods for handling embedded geometry is given in the first part of the thesis. The state-of-the-art is discussed and existing trends for the treatment of immersed boundaries and immersed domains are reviewed and analysed in details, in order to pave the way for the proposed technique. In the context of the advocated ”nearly” body-fitted mesh approach, the use of very-large anisotropic elements in the vicinity of the interface results in an approximated interface, playing the role of an embedded boundary during simulation. Error analysis shows that an optimal rate of convergence can be recovered by a careful mesh size choice in the direction normal to the interface. The last part of the thesis deals with the application of the method to a number of Computational Fluid Dynamics problems, with the emphasis on immersed no-slip walls. The obtained results are shown to be in good agreement with reference studies by other authors, indicating that the proposed technique is an accurate, robust and convenient approach to solving CFD problems with embedded geometries.
Affiliations
UCLouvainSST / IMMC / MEMA - Applied mathematics and mechanics
Citations
APA
Chicago
FWB
Quan, D. L. (2014). A mesh-based technique for solving CFD problems with embedded geometries : anisotropic adaptive “nearly” body-fitted mesh approach. https://hdl.handle.net/2078.5/192491