Multi-Fidelity Surrogate Model for Interactional Aerodynamics of a Multicopter

The growing interest in large electric multicopters (eVTOL aircraft) has prompted the search for methods that can accurately and efficiently predict their aerodynamic performance under different designs and operating conditions. The challenge is modeling the complex interactional effects of rotors operating in close proximity. This can be tackled with high-fidelity computational fluid dynamics (CFD) models, which capture the physics of rotor interaction from first principles. However, they are computationally demanding for performing studies over a range of parameters. On the other hand, lower-fidelity models are computationally inexpensive, but approximate the underlying physics and can be imprecise in predicting the fields of interest. In this study we present a multi-fidelity approach that inherits the accuracy of a high-fidelity model, while retaining most of the computational efficiency of a low-fidelity model. In this approach, the low-fidelity model is used to investigate the entire space of parameters and identify key parameter values to perform high-fidelity simulations. Thereafter, these high-fidelity simulations are used in a lifting procedure to determine multi-fidelity solutions at desired parameter values. We apply this strategy to determine the rotors' lift and drag distributions of a 2-rotor assembly in forward flight. The parameters considered are design variables, namely the longitudinal and vertical rotor-to-rotor separation, and operating conditions variables: forward speed and disk loading (DL). We conclude that over a large of parameters this approach yields results that retain the accuracy of the high-fidelity predictions at the computational cost of the low-fidelity model.