Font Size: a A A

Research And Implementation Of Hybrid Multi-granularity Parallel Tetrahedral Mesh Generation Algorithm

Posted on:2023-05-21Degree:MasterType:Thesis
Country:ChinaCandidate:F R HaoFull Text:PDF
GTID:2530306833988959Subject:Engineering
Abstract/Summary:PDF Full Text Request
Finite Element Analysis is an important basic analysis method in computational electromagnetics,computational fluid mechanics and computational solid mechanics.As the pre-processing of numerical simulation represented by Finite Element method,the process and result of mesh generation will greatly affect the post-processing.In practical application,as the calculation model adopted by various numerical simulation methods becomes more and more complex,the demand for generated mesh cell has been achieved with hundred million for the unit scale,and traditional serial algorithms can no longer meet the requirements of large scale mesh generation in memory space and time efficiency,the mesh generation technique become a hot issue in engineering research,the mesh generation algorithm has gradually developed from serial algorithm to more efficient parallel algorithm.With the rapid development of supercomputers in recent decades,the research on parallel algorithms for supercomputers to improve the speed and efficiency of mesh generation is now a key issue,and there are a lot of low-quality mesh cells in the process of mesh generation.In view of the above problems,this paper has the following specific work:(1)Starting with Delaunay algorithm,a common method for constructing unstructured meshes,various algorithms of tetrahedral mesh generation and mesh optimization are analyzed.An optimization strategy based on the combined parallel optimization algorithm is proposed to improve the mesh quality.The experimental results of the test model show the effectiveness of this strategy: the proportion of mesh quantity(Volume-length Ratio)less than 0.25 decreases from 7.45% to 1.35%,and 84.62% of mesh quality after optimization remains above 0.5.(2)Based on the idea of “divide and conquer”,a coarse-grained parallel tetrahedral mesh generation algorithm for distributed memory architecture is proposed.Three geometric models are tested using a workstation consisting of five compute nodes,and the experiments show that the acceleration ratio obtained by four computing processes is about 3.Then,by combining two thread-level fine-grained parallel triangulation algorithms and using MPI+Open MP programming model,the hybrid multi-granularity parallel tetrahedral mesh generation algorithm for distributed + shared memory hybrid architecture is implemented.The experiments show that the hybrid algorithm is feasible and the speed of model mesh generation is significantly improved.(3)Based on the combination of thread-level parallel algorithm and construction parallel optimization strategy,a simple Delaunay tetrahedral mesh generation system is implemented,which is different from the common client-based single software system,but based on the Browser/Server architecture,with certain convenience and scalability.
Keywords/Search Tags:Mesh Generation, Combined Mesh Optimization, Parallel Algorithms, Delaunay Algorithm
PDF Full Text Request
Related items