High-order mesh generation and adaptation for complex industrial geometries
Author(s)
S Kirilov, Kaloyan
Type
Thesis or dissertation
Abstract
High-order methods have been a standard for large eddy (LES) and direct numerical simulations (DNS) in academia and are now emerging as prime candidates for high-fidelity industrial CFD. A key bottleneck for this industrialization are the high-order meshing tools that automatically and robustly generate coarse high-quality body-fitted curvilinear meshes with sufficient resolution for resolving the flow features. Therefore, this thesis addresses the challenge through two complementary approaches: the robust generation of coarse curved meshes and mesh modification techniques allowing for efficient Adaptive Mesh Refinement (AMR).
In the first part, we propose a method for high-order mesh generation from third-party meshes. At its core is an approach for reconstructing the link between the mesh and the CAD model, which allows to leverage the robustness and flexibility of third-party straight-sided mesh generators with in-house state-of-the-art high-order mesh curving and optimisation techniques. We further address regions with reduced quality for very complex geometries through high-order mesh modifications and reformulate the long standing isoparametric prism layer splitting approach to allow generic boundary layer meshes. With these, we demonstrate robust generation of coarse curvilinear meshes for complex geometries such as F1 front wings and a full aircraft in high-lift configuration.
In the second part, we present a novel implementation of high-order conformal h-adaptation with predefined in the reference space maps in 2D and 3D. Through canonical examples, we demonstrate that it maintains mesh curvature and validity and often improves mesh quality. By combining the h-adaptation with a feature-based error indicator, we show that a very coarse, underresolved mesh can be adapted and capture the correct transition mechanisms. Finally, we demonstrate the potential of a true spectral/hp method, resolving accurately two inviscid test cases with combinations of h, r and p AMR, with the final h-r-p adaptation requiring hundreds of times fewer DoF than uniform refinement.
In the first part, we propose a method for high-order mesh generation from third-party meshes. At its core is an approach for reconstructing the link between the mesh and the CAD model, which allows to leverage the robustness and flexibility of third-party straight-sided mesh generators with in-house state-of-the-art high-order mesh curving and optimisation techniques. We further address regions with reduced quality for very complex geometries through high-order mesh modifications and reformulate the long standing isoparametric prism layer splitting approach to allow generic boundary layer meshes. With these, we demonstrate robust generation of coarse curvilinear meshes for complex geometries such as F1 front wings and a full aircraft in high-lift configuration.
In the second part, we present a novel implementation of high-order conformal h-adaptation with predefined in the reference space maps in 2D and 3D. Through canonical examples, we demonstrate that it maintains mesh curvature and validity and often improves mesh quality. By combining the h-adaptation with a feature-based error indicator, we show that a very coarse, underresolved mesh can be adapted and capture the correct transition mechanisms. Finally, we demonstrate the potential of a true spectral/hp method, resolving accurately two inviscid test cases with combinations of h, r and p AMR, with the final h-r-p adaptation requiring hundreds of times fewer DoF than uniform refinement.
Version
Open Access
Date Issued
2025-08-01
Date Awarded
2026-08-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Peiró, Prof. Joaquim
Sherwin, Prof. Spencer J
Moxey, Prof. David
Sponsor
European Commission
Grant Number
No.955923
Publisher Department
Department of Aeronautics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)