分布参数的主观不确定性对结构平均输出性能方差影响的降维分析方法  

A dimensional reduction method for analysis of the impact of the epistemic uncertainty of distribution parameters on a structural system's variance of average perfarmance output

作　　者：张强[1] 吕震宙[1] 

机构地区：[1]西北工业大学航空学院,西安710072

出　　处：《高技术通讯》2014年第9期963-970,共8页Chinese High Technology Letters

基　　金：国家自然科学基金(51308459)资助项目

摘　　要：为了有效控制结构平均输出性能,研究了分布参数主观不确定性对平均输出性能方差的影响。考虑到分布参数的主观不确定性对平均输出性能的方差贡献与分离主、客观不确定性后分布参数的主观不确定性对输出性能的方差贡献相同,并针对传统Monte Carlo方法效率低、计算量大的缺点,首先采用乘法降维方法求解基于平均性能方差的全局灵敏度,该方法对功能函数的调用次数远远小于传统的Monte Carlo方法;其次将主、客观分离方法与乘法降维方法相结合,求解分布参数的主观不确定性对平均输出性能方差的影响,该方法在保证精度的同时,进一步提高了计算效率。To achieve effective control of the average performance output of a structural system, the influence of the epistemic uncertainty of distribution parameters on the average output variance was analyzed. In view of the fact the conventional Monte Carlo analysis method has the short comings of lower efficiency and large computational cost, a muhiplicative version of the dimensional reduction method （M-DRM） was employed to compute the average output variance-based global sensitivity to reduce the functional function＇ s usage greatly compared to the conventional Monte Carlo method, and the method of separating the aleatory and epistemic uncertainties, combined with the MDRM method, was used to resolve impact of the epistemic uncertainty of contribution parameters on the average output variance to further improve the computional efficiency while keeping a higher precision.

关 键 词：主观不确定性 全局灵敏度 乘法降维方法 

分 类 号：TB114[理学—概率论与数理统计]