A Deep Learning Approach to Optimize Airfoil Shape for Reduced Trailing Edge Noise in Urban Air Mobility Rotors

Abstract

The objective of this research is to optimize the airfoil shape of UAM VTOL rotors for trailing edge noise reduction. Our study focuses on employing deep learning techniques for airfoil shape optimization with an emphasis on noise reduction. We implement a neural network that combines a Variational Autoencoder and Generative Adversarial Networks to learn the representation of airfoil shapes. The generated airfoil shapes are then evaluated for aerodynamic performance using XFOIL and trailing edge noise using acoustic diffraction model. The optimization process employs a Particle Swarm Optimization algorithm to find optimal latent features for airfoil generation. The results show that the optimized airfoil shapes have improved aerodynamic performance and reduced noise levels compared to the airfoil VR-12 used in NASA's conceptual UAM. These findings demonstrate that our optimization process has the potential to reduce trailing edge noise while ensuring enhanced aerodynamic performance, which warrants further investigation.