.png){kind=icon}
.png){kind=icon}
Return
Extended octree pattern-based mesh generation and massively parallel computing with the scaled boundary finite element method
.png){kind=icon}
OA
AI
Y
A
J
Y
E
C
DOI：10.1016/j.compstruc.2025.108051.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This work presents a fully automatic high-performance numerical framework for solving large-scale engineering problems. We propose an extended octree meshing paradigm to discretize the problem domains into polyhedral elements of a limited number of patterns. These patterns extend conventional octree meshes by enabling boundary conformity through a set of pre-defined trimming templates. The polyhedral elements are formulated using the scaled boundary finite element method. In the proposed framework, various formats of geometrical models such as STL and point clouds can be robustly and efficiently meshed without manual repair and surface reconstruction. The element matrices are computed from those of the element patterns, reducing memory access for iterative solvers and improving efficiency in parallel computing environments. The robustness as well as the accuracy are demonstrated by five numerical examples of elastostatics and elastodynamics, including complex geometries and heterogeneous problems. The results highlight the scalability and notable speedup for very large-scale simulations up to 1.2 billion degrees of freedom and 12,288 computing cores.
Keywords:
Mesh generation
Polyhedral elements
High-performance computing
Extended octree mesh
Scaled boundary finite element method
Heterogeneous domains
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
C
IF:
4.8
Papers:
6.2K
Citations:
1.7W
Organization
Cited Papers