Methodology for Probabilistic Analysis of Distributed Multidisciplinary Architecture

You are here: Homepage > Search

Methodology for Probabilistic Analysis of Distributed Multidisciplinary Architecture

Ghosh, Sayan / Mavris, Dimitri N.

Uncertainty based multidisciplinary analysis (UMDA) is a methodology to quantify uncertainty on the system-level quantity of interest by propagating uncertainty within and among multiple disciplines of a system. In an industrial setting, the design of complex systems involves the collaboration of multiple and diverse teams involving high-fidelity disciplinary tools and experts. Under such scenarios, engineers prefer UMDA methods based on distributed and coordination based procedures which preserve some disciplinary autonomy for uncertainty quantification and propagation. But these methods have limitations in capturing the dependence in coupling variables, which can lead to less accurate results in the system-level uncertainty. In this Paper, a methodology for probabilistic analysis of distributed multidisciplinary architecture is demonstrated in which disciplines can carry out uncertainty propagation autonomously and concurrently while accurately quantifying the dependence of the coupling variables. The method is applied on an analytical function and a supersonic jet design problem to demonstrate its capability of handing dependencies and non-Gaussian distributions of coupling variables and accurate estimation of the system-level quantity of metrics.