结构非线性动力分析显式积分并行算法

Parallel algorithm for explicit integration method in nonlinear dynamic structural analysis

对同时具有几何非线性、材料非线性和边界条件非线性的结构动力分析问题的并行性进行了讨论。提出了能有效用于该问题并行计算的双重区域分解法 ,重新规划了各处理机与宿主机间的任务分配。采用了数据整体传送技术 ,不需要处理子域交界面上的“影响单元”,减少了算法的复杂度 ,降低了通信开销。对 3 2 3 0 4 5阶自由度、 62 5 0 0时间步的大规模冲击接触问题进行了并行计算 ,在网络机群环境下采用 8个处理机并行计算时的并行效率为 84 .6%。结果表明 。

This paper analyzes the parallelism of dynamic structural analyses with geometric nonlinearities, material nonlinearities and nonlinear boundary conditions. A Double Domain Decomposition Method (DDDM) was used to provide high parallel efficiency for the numerical simulation. The computing loads were reallocated between master and slave processors. The global data communication technique was used, so the special processing method for “interface elements” was no longer required which reduced the complexity of the parallel algorithm and the communication requirements. For a large scale contact impact problem with 323 045 degrees of freedom and 62 500 time steps used as an engineering example, the parallel efficiency was satisfactory with an efficiency of 84.6% with 8 CPUs. The results show that the parallel algorithm is suitable for the parallel computation of large scale nonlinear problems in cluster environments.