A Holistic Framework for Uncertainty Quantification and Resilient Aircraft System Design

In modern aircraft system development, uncertainty plays a critical role in shaping performance, safety, and certification outcomes. These uncertainties emerge across all stages of design and operation—ranging from material properties and aerodynamic loads to structural dynamics and control behavior. Addressing this complexity requires a structured, interdisciplinary approach that integrates quantitative analysis with engineering expertise. We propose a holistic, feedback-driven Uncertainty Quantification Framework (UQF) for aerospace systems engineering. The UQF couples \emph{six computational stages} with \emph{two governance layers} (expert-review gates and provenance/archiving with feedback), closing the loop by feeding insights back into models and data. A key differentiator is continuous collaboration with domain experts at each gate. A central innovation is the integration of provenance models, based on W3C PROV-O and implemented through \emph{provtool}, an open-source toolkit. Provenance containers trace inputs, tools, transformations, and agents, supporting digital continuity and certification-grade traceability. We demonstrate the UQF on a permanent-magnet synchronous motor (PMSM) within an electromechanical actuator. In this case, the UQF revealed an interaction-dominated structure and, via direct quasi--Monte Carlo propagation, verified the “ready-to-lock” speed requirement with substantial margin: no violations in $M=2^{20}$ samples (95\% Wilson upper bound ${\le} 3.7\times10^{-6}$) and a reliability-calibrated frontier near $430~\mathrm{rad/s}$ for a $10^{-3}$ risk target. All artifacts were captured as W3C PROV/SHACL bundles by \emph{provtool}, demonstrating auditability and reuse. While applied here to a single actuator subsystem, the framework is method- and use-case-agnostic and is scalable across aircraft domains. Ultimately, the UQF establishes a reproducible, modular, and certifiable methodology for design under uncertainty by embedding human expertise within a structured UQ process and enabling transparent data traceability. The results presented here were produced as part of the DLR internal project AdViCe (\textit{Advanced Methods for Virtual Certification}).