HDMR在多维随机变量结构全局灵敏度分析的应用  

Application of High Dimensional Model Representation in Global Sensitivity Analysis of Multi-Dimensional Random Variable Structure

作　　者：陈杉杉 CHEN Shan-shan(Tongji University,Shanghai 200092,China)

机构地区：[1]同济大学结构工程与防灾研究所,上海200092

出　　处：《佳木斯大学学报（自然科学版）》2018年第2期192-195,共4页Journal of Jiamusi University：Natural Science Edition

摘　　要：针对多随机变量作用的框架结构体系,采用构造近似模型替代传统的大规模计算任务,并获取结构体系的全局灵敏度信息进行分析。结果表明,HDMR可以在一定样本的条件下准确构造近似模型并获取灵敏度信息,对结构优化与安全评估具有重要意义。High dimensional model representation can reduce the complexity of the high-dimensional model problem from polynomial level by exponential growth to the polynomial level by using the hierarchical property of the function,and reduce the computational cost,and it has the function of revealing the output of the system with respect to the multi-input variable.Therefore,the HDMR construction approximation model is used to replace the traditional large-scale computation task for the frame structure system with multi-random variables,and the global sensitivity information of the structural system is analyzed.The results show that HDMR can accurately construct approximate model and obtain sensitivity information under certain sample conditions,which is important for structural optimization and safety assessment.

关 键 词：高维模型表示 全局灵敏度 多随机变量结构 MonteCarlo采样 

分 类 号：TU318[建筑科学—结构工程]