Non-Probabilistic Uncertainty and Correlation Propagation Analysis Methods Based on Multidimensional Parallelepiped Model

Abstract

In engineering practice, the uncertainty and correlation may coexist in the input parameters, as well as in the output responses. To address such cases, several methods are developed for the non-probabilistic uncertainty and correlation propagation analysis in this study. In the proposed methods, the multidimensional parallelepiped model (MPM) is introduced to quantify the uncertainty and correlation of input parameters. In the uncertainty propagation analysis, three methods are presented to calculate the interval bounds of output responses. Among the methods, the Monte Carlo uncertainty analysis method (MCUAM) is firstly presented as a reference method, and then the first-order perturbation method (FOPM) is employed to promote the computational efficiency, and the sub-parallelepiped perturbation method (SPPM) is further developed to handle the correlated parameters with large uncertainty. In the correlation propagation analysis, the Monte Carlo correlation analysis method (MCCAM) is proposed based on the MPM and Monte Carlo simulation, which aims to compute the correlation among different output responses. The uncertainty domains between any two responses can also be constructed by the MCCAM. The effectiveness of the proposed methods on dealing with the uncertainty and correlation propagation problems is demonstrated by three numerical examples.