基于证据理论和代理模型的QMU分析  

Quantification of Margins and Uncertainties Approach Using Evidence Theory and Surrogate Model

作　　者：谢朝阳[1] 李贵杰 彭忠明[1] 黄洪钟[2] 

机构地区：[1]中国工程物理研究院总体工程研究所,四川绵阳621000 [2]电子科技大学机械电子工程学院,成都611731

出　　处：《电子科技大学学报》2018年第1期66-72,共7页Journal of University of Electronic Science and Technology of China

基　　金：中国工程物理研究院-国家自然科学基金委联合基金(NSAF U1330130);总装技术基础项目(2013ZK1.2)

摘　　要：针对结构可靠性或性能评估中的随机和认知不确定性同时存在的情况,根据性能裕量与不确定性量化的概念,提出了基于证据理论和Kriging代理模型的性能裕量与不确定性量化分析计算方法。该方法首先对随机变量进行抽样并通过优化求解证据焦元内结构性能响应极值分布并生成训练样本空间,通过最大置信水平期望提高加点准则构建并更新Kriging代理模型,提高裕量与不确定性量化分析过程中不确定性传播的效率和精度,在此基础上通过计算置信因子实现结构可靠性或性能评估度量。最后通过算例比较研究了基于置信因子度量结构可靠性和结构非概率可靠性之间的差异。The structure reliability or performance assessment of complex systems with both epistemic and aleatory uncertainties is a challenge problem for engineering systems. In this paper, an implementation framework of systems reliability assessment based on the quantification of margins and uncertainties （QMU） methodology is proposed. The description of QMU concept is introduced first, then the Dempster-Shafer Theory of Evidence is used to present the presence of aleatory and epistemic uncertainties in the proposed QMU implementation framework. To alleviate the computational costs, the Kriging model has been implemented as the surrogate for the structure response. Then the structure reliability was presented by a QMU metric in terms of confidence factor（CF）. The technique is demonstrated by a numerical example to account for the computational efficiency. The difference between QMU metric and non-probabilistic reliability methods is discussed

关 键 词：证据理论 KRIGING模型 混合不确定性传播 裕量与不确定性量化 

分 类 号：TM315[电气工程—电机] TH114[机械工程—机械设计及理论]