equationchapter1section1多输出性能下的重要性测度指标及其.docVIP

多输出性能下的重要性测度指标及其求解方法 徐立扬1，吕震宙1，王飞1, 肖思男1 西北工业大学 航空学院，西安 710072)摘要：Moore-Penrose（MP） 关键词：中图分类号：文献标识码：A Global sensitivity analysis for multiple outputs and their solutions XU Liyang 1 , LU Zhenzhou 1, WANG fei 1, XIAO Sinan 1 （1.School of Aeronautics, Northwestern Polytechnical University, Xi’an, 710072,China） Abstract: For the popular multiple outputs system in the engineering, it is very meaningful to analyze the comprehensive influence of the uncertainty of input variables on multiple outputs. The effect of conditional input variables on the integrated performance of multiple outputs can be measured by Mahalanobis distance, from conditional mean vector to unconditional mean vector in multiple outputs space, however, which has the case that the covariance matrix inversion doesn’t exist. And when the output variable and input variable have the same dimension and the linear function relation, the importance measure based on the Mahalanobis distance can’t properly identify the significance of input variables. Aiming at the existing drawbacks of indices of the Mahalanobis distance, an importance measure based on the Moore-Penrose (MP) Mahalanobis distance weighted by spectral decomposition is proposed. Through building the generalized matrix inversion of covariance matrix of multi-output and the spectral decomposition, the problems that the covariance matrix can’t be inversed and misidentification for lacking the adequate consideration about the relation among the multiple outputs are solved. Thus, the limitations of indices of Mahalanobis distance are overcome. The numerical examples and engineer instance show that the proposed importance measure can accurately get the effects of input variables on the integrated performance of multi-output structure system, thus, it can provide effective information for reliability design. Key words: Multivariate output，Importance measure， Mahalanobis distance，Generalized inverse matrix，Weighted spectral decomposition 在结构系统可靠性设计中