Assessment of Multi-fidelity Surrogate Models for High-Altitude Propeller Optimization

In this paper, the performance of propellers is evaluated with 3D RANS and Vortex Theory for a wide range of geometries and advance ratios at high-altitude conditions. For this purpose, the results of a high-altitude propeller optimization are used as a CFD database and discussed briefly. A comparison is made between the predictions of 3D RANS and Vortex Theory with a goal to examine the agreement between the two models. Moreover, a variance-based sensitivity analysis is performed to determine geometrical or operational features of the propeller designs that influence the error between the two models. In the second part, the data from the two physical models are used to train different multi-fidelity surrogate models, using the data produced from 3D RANS, as the high-fidelity dataset and Vortex Theory, as the low-fidelity one. Different performance metrics are used to evaluate the predictive capabilities of the multi-fidelity and single-fidelity surrogate models in new propeller geometries. The predictions of these data-driven models are finally compared to the predictions of Vortex Theory.