结构非线性有限元分析的有效隐式并行算法

An effective implicit parallel algorithm for structural nonlinear finite element analysis

利用单元接单元的预处理线性共轭梯度法对非线性隐式有限元结构分析的粗粒度并行算法进行研究。采用二级区域分解,第一级粗区域级分解为每个处理器的并行计算,创建一个负载平衡的区域;第二级分解为将每个区域分解成同类单元块(相同本构模型、积分级数、单元类型),以使每个处理器中进行细粒度的并行计算。采取Hughes-Winget(HW)单元接单元的预处理子。将非结构相关图和一种新的平衡图着色算法结合,来实现区域内和区域间的并行计算,并与对角预处理子进行比较。结果表明:本文二级区域分解的HW预处理子算法可提高运算速度,显示良好的并行性能,适合分布式储存体系的机群。

A coarse-grain parallel algorithm of a linear preconditioned conjugate gradient solver using an element-by-element and preconditioner for nonlinear implicit finite element analysis is studied. A dual-level mesh decomposition is used, i.e.a coarse, domain-level decomposition creates a load-balanced domain for each processor for parallel computation, while a second level decomposition breaks each domain into blocks of similar elements(same constitutive model, order of integration, element type) for fine-grained parallel computation on each processor. The Hughes Winget(HW) element-by-element preconditioner is employed. The implementation couples an unstructured dependency graph with a new balanced graph-coloring algorithm to schedule parallel computations within and across domains. The diagonal preconditioner is compared in examples. The example illustrate that the proposed algorithm improves overall execution speed and shows good parallel performance. The algorithm is suitable for workstation clusters with distributed memory architectures.