基于PVM的网络并行计算在结构优化设计中的应用

Parallel Computing Network Application in Structural Optimization Based on PVM

结构优化涉及大量的目标函数、设计变量和设计约束,分析工作量大,求解非常费时.一个替代的办法是把整体结构优化分解为两层优化问题:顶层为协调层优化,底层为若干个子系统层优化.提出了一个改进的负载平衡算法.以10杆和52杆桁架为例验证了二层结构优化的可行性和收敛性,计算结果表明,问题粒度越大越能获得更好的加速比和并行效率.

Optimization of structural systems is characterized by many components. A large design variable space, a significant number of design constraints and tremendous analysis leads to time-consuming procedures. The original optimization is decomposed into subsystem level optimization and coordination level optimization, a two-level hierarchy. The load balance algorithm is improved. The feasibility and convergence are investigated through a 10-bar truss and a 52-bar truss in two-level structural optimization design. The tests show that the larger the problem granularity is, the higher the speedup and the parallel efficiency are.