Aeroacoustic Airfoil Shape Optimization Enhanced by Autoencoders

• DOI: 10.2139/ssrn.4236021
• Publication Date: 2022-10-06T06:59:55Z
• Authors: Jiaqing Kou; Laura Botero-Bolivar; Roman Ballano; Oscar Marino; Leandro de Santana; Eusebio Valero; Esteban Ferrer
• Keywords: UT-Hybrid-D; Fluid Dynamics (physics.flu-dyn); Física; FOS: Physical sciences; Autoencoder; Physics - Fluid Dynamics; 7. Clean energy; 01 natural sciences; Aeronáutica; Amiet theory; Machine learning; 0103 physical sciences; Aeroacoustics; Optimization design

We present a framework for airfoil shape optimization to reduce the trailing edge noise design of wind turbine blades. Far-field is evaluated using Amiet's theory coupled with TNO-Blake model calculate wall pressure spectrum and fast turn-around XFOIL simulations evaluate boundary layer parameters. The computational first validated NACA0012 at 0 angle attack. Particle swarm used find optimized configuration. multi-objective minimizes A-weighted overall sound level various angles attack, while ensuring enough lift minimum drag. compare classic parametrization methods define geometry (i.e., CST) machine learning method variational autoencoder). observe that autoencoders can represent wide variety geometries, only four encoded variables, leading efficient optimizations, which result in improved optimal shapes. When compared baseline geometry, NACA0012, autoencoder-based reduces by 3% (1.75 dBA) (with decreased across entire frequency range), maintaining favorable aerodynamic properties terms