多输出性能下的重要性测度指标及其求解方法
发布时间：2018-11-05 16:23
【摘要】：针对基于马氏距离的重要性测度存在的问题,提出了基于谱分解加权摩尔彭罗斯马氏距离的重要性测度指标,通过构造多输出协方差阵的广义逆矩阵以及谱分解的策略,有效解决了协方差阵求逆奇异情况以及由于未能充分考虑多输出之间的相互关系而导致的错误识别重要变量的问题,克服了基于马氏距离指标的局限性。数值算例与工程算例结果表明:所提重要性测度可以更加准确地获得输入变量对结构系统多输出性能随机取值特征贡献的排序,从而为可靠性设计提供充分的信息。
[Abstract]:In order to solve the problem of importance measure based on Markov distance, the importance measure index based on spectral decomposition weighted Moore Penrose Markov distance is proposed. The generalized inverse matrix of multi-output covariance matrix and the strategy of spectral decomposition are constructed. It effectively solves the problem of finding inverse singularity by covariance matrix and the error identification of important variables caused by failure to fully consider the interrelation between multiple outputs, and overcomes the limitation based on Markov distance index. Numerical examples and engineering examples show that the proposed importance measure can obtain more accurately the ranking of the input variables' contributions to the random characteristic values of the multi-output performance of the structural system, thus providing sufficient information for the reliability design.
【作者单位】： 西北工业大学航空学院;
【基金】：国家自然科学基金资助项目(NSFC51475370) 中央高校基本科研业务费专项资金资助项目(3102015BJ(Ⅱ)CG009)
【分类号】：TB114.3
，
本文编号：2312629
[Abstract]:In order to solve the problem of importance measure based on Markov distance, the importance measure index based on spectral decomposition weighted Moore Penrose Markov distance is proposed. The generalized inverse matrix of multi-output covariance matrix and the strategy of spectral decomposition are constructed. It effectively solves the problem of finding inverse singularity by covariance matrix and the error identification of important variables caused by failure to fully consider the interrelation between multiple outputs, and overcomes the limitation based on Markov distance index. Numerical examples and engineering examples show that the proposed importance measure can obtain more accurately the ranking of the input variables' contributions to the random characteristic values of the multi-output performance of the structural system, thus providing sufficient information for the reliability design.
【作者单位】： 西北工业大学航空学院;
【基金】：国家自然科学基金资助项目(NSFC51475370) 中央高校基本科研业务费专项资金资助项目(3102015BJ(Ⅱ)CG009)
【分类号】：TB114.3
，
本文编号：2312629
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2312629.html
