An efficient parallel mesh generation method for finite element based analysis of large complex architecture

Abstract

The Finite Element Method (FEM) is renowned for its exceptional computational accuracy and adaptability to diverse, complex geometric configurations, making it a powerful tool for assessing the structural stability of historical edifices. By leveraging structural mechanics analysis through FEM, a more proactive approach to the conservation of ancient structures can be pursued. Nonetheless, the analysis of many extant ancient architectures, characterized by their considerable scale and structural complexity, demands a vast quantity of finite element meshes to enable high-precision calculations—demands that traditional serial processing methods fail to meet. This paper introduces a parallel large-scale mesh generation algorithm designed for cluster computing environments, which facilitates the mesh generation for large-scale complex structures. The proposed algorithm enhances the continuity and integrity of mesh information across subdomains by projecting volume elements located on the interfaces of adjacent subdomains onto each other based on global barycentric IDs. This projection ensures that finite element analysis can proceed independently within each subdomain. Experimental results demonstrate that our approach not only generates a substantial number of meshes more efficiently but also ensures that the finite element analysis results for each subdomain closely approximate those of a global finite element analysis, as well as reducing the communication time among subdomains and improving the efficiency of the subsequent finite element solution. This advancement significantly boosts the time-efficiency of high-performance preventive monitoring and protection for largecomplex heritage architectures, thereby coenhancing the efficacy of preservation efforts.