Predicting Propeller Tonal Noise with AI Trained First-Principle Models: A Novel Methodology

Abstract

The development of propeller-driven small aircraft and Advanced Air-Mobility vehicles poses significant challenges in urban environments making accurate noise prediction crucial in the preliminary design phase. This research aims to develop a semi-empirical model combining physics based and empirical findings using Machine Learning to predict tonal noise from isolated and installed propeller configurations.A propeller noise source model is selected by coupling the Unsteady Panel Method code with the FW-H equation based solver. Preliminary results demonstrate good agreement between tones obtained from UPM-APSIM and a traditional BEMT model coupled to HANSON model, thus validating the UPM-APSIM�s ability to capture tonal noise characteristics accurately for isolated propellers. For installed configurations, the UPM-APSIM predicts tonal noise including wing-wake interaction noise in terms of higher harmonics using vortex particle method. Building on these findings, a generic case of propeller installed with wing is defined, and parameters influencing noise are identified. An AI dataset is created, to define a mathematical function for predicting tonal values. The final paper will validate the AI predicted tonal values against a well-defined test case. This research aims to mitigate noise impact on urban environments through data-driven modeling and simulation, utilizing AI-trained first-principle models for accurate tonal noise prediction.