Dragonfly CFD Validation and Uncertainty Quantification

This work demonstrates an application of Computational Fluid Dynamics (CFD) model validation and model-form uncertainty quantification for the NASA Dragonfly lander in its Preparation for Powered Flight (PPF) configuration. A multi-fidelity CFD workflow is employed, utilizing both mid-fidelity Reynolds-Averaged Navier-Stokes (RANS) and high-fidelity Improved Delayed Detached Eddy Simulation (IDDES) models. Surrogate models are trained from CFD simulations to ingest operational condition uncertainty and propagate uncertainty to loads of interest. An area metric validation quantity is used to establish model-form uncertainty. FZ area metric values based on RANS CFD results alone indicate high discrepancy, >12 N, near pure descent orientations, high descent velocity, and high lander rotor RPM. These errors are reduced to <4 N through targeted IDDES simulations. MZ error is consistent between the different model fidelities considered. The results demonstrate the effectiveness of the proposed procedure in quantifying model-form uncertainty and identifying need for increased model fidelity for downstream vehicle performance simulations.