Efficient first-order reliability analysis of multidisciplinary systems

This paper develops two algorithms that address the reliability evaluation of complex multidisciplinary engineering systems. In particular, systems with feedback coupling – a common characteristic of many multidisciplinary analyses – are considered. In such analyses, iterative convergence loops are needed to resolve inconsistencies in feedback variables. Assessing the reliability of such systems with a traditional 'black box' or fully coupled approach requires Multidisciplinary Analysis (MDA) convergence loops nested inside iterative loops for probabilistic analysis. The resulting computational effort is unacceptable for most high fidelity analyses. Therefore, this paper proposes two first-order reliability analysis methods that efficiently apply probabilistic analysis to multidisciplinary systems with feedback using a decoupling approach. The first method uses a First-Order Second Moment (FOSM) technique to characterise intermediate variables while applying more rigorous reliability analysis on the system as a whole. The second algorithm gives a specific solution to a decoupled first-order reliability analysis formulation as an optimisation problem. Each method is applied to an illustrative mathematical model and compared to otherwise equivalent coupled approaches with respect to accuracy and computational effort.