面向复杂几何模型的并行四面体网格生成方法

面向大规模工程计算等数值模拟领域，提出了一种支持复杂几何模型的大规模四面体网格并行生成方法。该方法以复杂几何模型作为输入，首先采用串行网格生成方法生成初始四面体网格，然后通过两级区域分解方法将初始网格分解为多个子网格并分配到相应的进程中，进程间并行地提取出子网格的表面网格，并基于几何模型对面网格进行贴体加密，最后对加密后的面网格采用Delaunay方法重新生成四面体网格，该方法可以更好地适应高性能计算机体系结构，较好地克服了并行方法中并行性能和网格质量不能兼顾的问题。对三峡大坝模型进行测试和验证，证明该方法具有良好的并行效率和可扩展性，可以在数万处理器核上并行生成数十亿高质量四面体网格。

A parallel generation method of tetrahedral mesh for complex geometric models

A parallel generation method of large-scale tetrahedral meshes for models in complex geometric shapes is proposed for the numerical simulation of large-scale engineering computation.Firstly, the initial tetrahedral mesh is generated by a sequential mesh generation method for a model of complex geometric shape.Secondly, a two-level domain decomposition method is applied to decompose the initial mesh into multiple sub-meshes and then distribute sub-meshes to the corresponding process.Finally, the surface meshes are extracted from the sub-meshes and refined by uniform refinement method in parallel.A new tetrahedral mesh is generated by sequential Delaunay mesh generation method.This method solves the problem that parallel performance and mesh quality cannot be both achieved in general parallel methods.Through the test and verification of the Three Gorges dam model, a good parallel efficiency and scalability are demostrated, and billions of high-quality tetrahedral meshes can be generated on thousands of processor cores in parallel by the proposed method.