Indirect generation of quadrilateral and hexahedral-dominant meshes: a hybrid approach based on cross-field guidance, topological labelling and patterns by Christos GEORGIADIS

IMMC

May 26, 2021

16:30

Pour l’obtention du grade de Docteur en sciences de l’ingénieur et technologie

Numerical simulation is a powerful tool to analyze complex physical phenomena for scientific analysis and engineering design. Mesh generation is an essential step of numerical methods, providing a subdivision of the physical domain in question to a finite number of simple elements.

Quadrilateral and hexahedral meshes offer certain advantages compared to triangular and tetrahedral meshes, such as reduced number of elements and alignment with problem-specific directions. Nevertheless, while there exist mature methods to generate the latter, there are no conclusive algorithmic solutions for the former.

The objective of this thesis is to provide a framework for the robust and efficient generation of high-quality quadrilateral and hexahedral-dominant meshes. A hybrid approach is followed by borrowing elements from engineering techniques for mesh generation, graph theory, and the computer graphics community. Cross-fields are used to place points across the domain optimally. A two-step method using local mesh modifications is developed for surface meshing that enables the robust generation of meshes with cross-field aligned triangles. The output cross-field aligned triangular mesh can be converted to an all-quadrilateral mesh in an indirect fashion by merging pairs of triangles using a novel bipartite labeling scheme. Finally, the quality of output quadrilateral meshes can be further optimized by locally remeshing with topological patterns. The same indirect approach is followed for volume meshing. A complete pipeline for hexahedral-dominant meshing is developed, and the use of integrable frame-fields is investigated.

The developed pipelines are used to produce meshes on a large number of models with diverse characteristics in order to test the robustness limits. The algorithms are efficient since they are mainly based on local operations. Overall, the approach presented offers a pragmatic solution for the mesh generation of “real-life”, complex models.

Jury members :

  • Prof. Jean-François Remacle (UCLouvain), supervisor
  • Prof. Thomas Pardoen (UCLouvain), chairperson
  • Prof. Vincent Legat (UCLouvain), secretary
  • Dr. Jonathan Lambrechts (UCLouvain)
  • Prof. Koen Hillewaert (Université de Liège, Belgium)
  • Dr. Franck Ledoux (CEA, France)
  • Prof. Cecil Armstrong (Queen’s University, N. Ireland)
  • Dr. Gaëtan Compère (Siemens, Belgium)

Pay attention :

The public defense of Christos Georgiadis scheduled for Wednesday 26 May at 4:30 p.m will take place in the form of a video conference Teams

Categories Events: