Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
On the Prediction of Propeller Tonal Noise with Machine Learning
Submission Author:
Pedro Henrique de Carvalho , TO
Co-Authors:
Pedro Henrique de Carvalho, Rafael Cuenca, Filipe Dutra da Silva
Presenter: Filipe Dutra da Silva
doi://10.26678/ABCM.COBEM2023.COB2023-1717
Abstract
From a historical perspective, propeller noise has been a community concern, mainly for areas with high air traffic such as airports and urban areas. Recently, the popularity of drones and other unmanned aerial vehicles (UAV) for in-city missions, such as package deliveries and surveillance, has motivated the investigation of propeller noise and the development of quieter devices and noise reduction systems. Aiming to investigate the noise impact of UAVs on community areas located on flight paths, a multilayer perceptron (MLP) neural network for regression was trained to model the propellers performance, using the Advanced Precision Composites (APC) database. The machine learning models were coupled with Hanson's load and thickness tonal noise prediction method to evaluate the noise for a ground observer. The repeated-kFold cross validation technique was implemented to evaluate the mean squared error (MSE) of 16 different MLPs configurations and the chosen one resulted in a MSE of the order 1 x 10E-5. Finally, the performance model was validated through comparisons with experimental data and, after that, the noise prediction model was coupled to it. Acoustics results showed good agreement with experimental data, highlighting the accuracy of the final model.
Keywords
Propeller noise, Aeroacoustics, machine learning, reduced order model, Propeller performance
