论文部分内容阅读
随着岩土工程规模的不断扩大、复杂性的增加以及计算参数的多样化和计算精度的提高,人们对于计算机计算能力的要求越来越高,然而单处理器无法满足这类大规模计算.从数据输入、区域分解、线性方程组的迭代求解、后处理等方面详细阐述高性能计算平台上并行有限元求解大规模岩土工程的关键问题.提出了利用MPI2的新特性进行海量数据的分段并行读入,采用ParMetis软件并行地进行区域分解,实现了前处理过程的完全并行化;采用基于Jacobi预处理技术的预处理共轭梯度法(PCG)进行线性方程组的并行迭代求解;采用Paraview软件实现了后处理的并行可视化.在深腾7000系统上对某隧道工程的三维开挖过程进行了数值模拟,对其并行性能进行了分析和评价,验证了采用的区域分解算法和系统方程组的求解方法的可行性,并且具有较高的加速比和并行效率.
With the continuous expansion of geotechnical engineering, the increase of complexity and the increase of computational parameters and the improvement of computational accuracy, people are demanding more and more computational power. However, single-processor systems can not meet these large-scale computations. The key problems of large-scale geotechnical engineering solved by parallel finite element method on high-performance computing platform are elaborated from the aspects of data input, area decomposition, iterative solution of linear equations and postprocessing, etc. The new feature of MPI2 is proposed to analyze the mass data The parallel readings were carried out by using ParMetis software and the parallelization of the preprocessing process was realized in parallel. The preconditioned conjugate gradient method (PCG) based on Jacobi preprocessing was used to solve the parallel iterative equations of the linear equations. Paraview software realizes parallel visualization of postprocessing.The 3D excavation process of a tunnel project is simulated in Shen Teng 7000 system and its parallel performance is analyzed and evaluated.The regional decomposition algorithm and system equation Group solution method is feasible, and has a higher speedup and parallel efficiency.