Bayesian optimisation for low-noise aerofoil design with aerodynamic constraints

Abstract

This paper presents an optimisation approach for designing low-noise Outlet Guide Vanes (OGVs) for fan broadband noise generated due to the interaction of turbulence and a cascade of 2-dimensional aerofoils. The paper demonstrates the usage of Bayesian optimisation with constraints to reduce the computation cost of optimisation. The prediction is based on Fourier synthesis of the impinging turbulence and the aerofoil response is predicted for each vortical modal component. A linearised unsteady Navier-Stokes solver is used to predict the aerofoil response due to an incoming harmonic vortical gust. This paper shows that to achieve noise reductions of 0.5 dB the penalty on the aerodynamic performance of 33% is observed compared to baseline aerofoil. Hence, the geometry changes such as thickness and nose radius can’t reduce broadband noise without effecting aerodynamic performance.