Contribution
A Fast Machine Learning Enhanced Approach for Aircraft Noise Prediction
* Presenting author
Abstract:
Propeller aircraft noise is a key feature that has to be taken into account during aircraft design and operation. In this study, we present a fast prediction methodology that is targeted to be used for acoustic mission planning. For this purpose, multiple noise footprints on ground need to be calculated in basically real-time to provide an acoustically optimized trajectory to the pilot or aircraft operator. As basis, we use a low-fidelity prediction methodology, which is then trained and tailored to a specific aircraft platform with measurement data and/or high-fidelity simulation data. We demonstrate the prediction improvement using the machine learning approach on a use case of a twin-turboprop aircraft (Beechcraft King Air 350). In this case, the machine learning model was trained with a limited set of measurement data of an acoustic flight test campaign. The prediction results are compared with measurement data that were excluded from the training. The comparisons show a significant improvement in tonal and broadband noise prediction using the machine learning approach over the low-fidelity approach.