A Decoupled Approach to Multidisciplinary System Analysis Under Uncertainty

Abstract

This paper presents an efficient probabilistic approach for uncertainty propagation in multidisciplinary system analysis (MDA) under aleatory uncertainty (i.e., natural or physical variability). To enhance computational efficiency, a decoupled method is employed to separate the MDA from the probabilistic analysis. Initially, the paper introduces a moment-matching technique to estimate the first four moments of the coupling variables. This technique utilizes Taylor series expansion, direct tensor product quadrature, sparse grid numerical integration, and univariate dimension reduction methods. After quantifying the uncertainty in the coupling variables, system-level uncertainty propagation is carried out using methods similar to those used in single-discipline problems. The proposed methods are suitable for both sampling and analytical approximation-based reliability analysis techniques. The effectiveness of these methods is demonstrated through a mathematical problem and a practical engineering problem.