Prediction of noise of commercial aircraft based on itself specifications by using machine learning methods

Abstract

The concerns related to aircraft noise have come to light due to the increase in commercial aircraft activities. Forecasting aircraft noise with high accuracy is of high importance for helping attempts regarding noise mitigation, which is an important concern for people living in the environment of the airports. In this study, the noise of commercial aircraft is predicted for lateral, flyover and approach points based on maximum take-off mass (MTOM), maximum landing mass (MLM) and engine take-off thrust. For this study, the data of more than 12000 is filtered to 3528 due to existing repeated data and the prediction is performed by employing two machine learning methods such as Random Forest (RF) and Long Short-Term Memory (LSTM). Moreover, the analysis of feature importance is carried out for three cases where the modeling is established. According to analysis results, noise is predicted with between about 0.96 and 0.97 of R2 through three points by RF where mean absolute error (MAE) changes 0.043-0.049. On the other hand, LSTM achieves noise modeling with higher accuracy, which provides more than 0.99 of R2. Namely, MAE is obtained to change between 0.0085 and 0.023 for all phases. Lastly, MTOM has the highest importance for prediction of noise with 82.58%-94.48% whereas it is followed by engine take-off thrust, which has 12.5% importance at flyover phase. This study shows that aircraft noise can be forecasted with relatively low model error using three known specifications of any aircraft-engine pairing. To predict aircraft noise with high accuracy helps the designers to observe the effects of changes in aircraft weight and power of the engine on aircraft noise due to the retrofitting of new technologies.