【摘要】 针对浅水环境下的污染物迁移问题,提出了一种基于非结构网格的全隐格式可扩展并行算法。首先,对计算区域进行非结构网格建模;然后,对污染物迁移方程时间方向采用二阶向后差分隐式格式、空间方向采用有限元方法进行离散,并使用基于区域分解的NKS并行算法对离散系统进行求解。最后,使用该算法对珠江前航道和珠江出海口的污染物扩散过程进行了测试模拟,在"天河二号"超级计算机上,最高使用2 400个处理器核时获得了64%的并行效率。测试结果表明NKS算法能有效地用于复杂区域大规模问题的数值模拟。更多还原

【Abstract】 An unstructural-grid-based fully implicit scalable algorithm was proposed for the problem of contaminant transport in shallow water environment. Firstly, the computational domain was meshed with unstructured tetrahedral elements, and then the second-order backward differentiation formula in time direction and the finite element method in space were applied to discretize the contaminant transport equations. The NKS(Newton-Krylov-Schwarz) parallel algorithm based on domain decomposition was used to solve the discrete system. Finally, the contaminant transport process of the Pearl River front channel and the Pearl River estuary were simulated, and a parallel efficiency of 64% was achieved for 2 400 processor cores on the "Tianhe 2" supercomputer. The experimental results show that the NKS algorithm can be effectively used for numerical simulation of large-scale problems in complex regions.更多还原