Towards a Two-Stage Multi-Fidelity Shape Optimization Approach to Noise Reduction in Urban Air Mobility Applications

We propose a two-stage optimization framework based on computational models of different fidelity. The framework integrates two sequential stages. Namely, an exploratory stage, based on a Bayesian approach, followed by the exploitative stage, executed using gradient-based method. First, the optimizer employs aerodynamic solvers of increasing fidelity to explore the entire design space and identify promising design candidates. After, the geometry is reparametrized and the design space expanded. An adjoint-based optimizer refines the design using a high fidelity model. Overall, the goal is to reduce the noise produced by an isolated propeller tailored to Urban Air Mobility applications. The objective is to automatically design a two-blade propeller, minimizing the tonal noise signature, evaluated using the hybrid Ffowcs Williams-Hawkings method, while meeting specific aerodynamic performance targets. The proposed optimization architecture surpasses standalone Bayesian and adjoint methods in designing a high-thrust propeller with reduced tonal noise emissions during forward flight, highlighting the potential benefits of incorporating a positively swept and dihedral tip region.