MODEL-SI: Modeling and Simulation - Multi-fidelity surrogate model of an eVTOL for certification

With this paper we present the project MODEL-SI, which is funded by the Horizon Europe and managed by the European Union Aviation Safety Agency (EASA). It aims at assessing the goodness of state-of-art modelling techniques as they are applied to eVTOL aircraft and the potential savings in flight testing costs, if a digital twin is exploited during the certification process. The aim of the project could also be reformulated by saying that (i) eVTOL aircraft exhibit peculiar aerodynamics phenomena and may experience very interesting dynamic responses to perturbations such as wind gusts, and (ii) state of art modelling and multi-fidelity in particular may turn out to be an effective approach to the exploration of the flight envelope with all relevant configurations. As a matter of fact, we know that some fast, low-fidelity model is necessary to cover flight envelope and a number of configurations; moreover, we expect local, mid- and high-fidelity modelling to be indispensable to capture some of the flow mechanics patterns (typically, the interactions rotor-rotor and rotor-wing). Finally, we expect machine learning to play a role in our project or even to be instrumental. Trivially, we plan to exploit gaussian processes and / or neural networks to combine data from low-, mid- and high-fidelity physics-based models into a fast yet accurate surrogate model. However, pattern-recognition may also be exploited to acquire a deeper understanding of the physics involved; for instance, this can happen from CFD-generated or experimental data. The physics-based simulation model architecture is presented together with the workflow of the data-driven models and how they are integrated in a simulation environment.