A physics-based PSO-BPNN model for civil aircraft noise assessment

Abstract

With the increasingly severe problem of aircraft noise pollution around airports, it is urgent to explore accurate and effective aircraft noise assessment methods. This article proposes a physics-based PSO-BPNN model based on the European Civil Aviation Conference (ECAC) best practice model, backpropagation neural network (BPNN), and particle swarm optimization (PSO) to enrich the methodology system of aircraft noise assessment. The primary modeling process is as follows. Firstly, based on airport parameters, flight information, and other data, the theoretical noise levels of ground monitoring point are calculated using the ECAC dynamic model. Then, a dataset containing measured noise values, theoretical noise values, trajectory data, and meteorological data is constructed to train the physics-based BPNN model, in order to correct the theoretical noise level. Finally, the PSO algorithm is introduced to optimize the parameters of the BPNN model and the construction of the physics-based PSO-BPNN model is completed. By taking Hefei Xinqiao International Airport (HFE) as a research case, the experimental results show that the physics-based PSO-BPNN model, which combines best practice model and machine learning model, demonstrates better performance than the ECAC model and physics-based BPNN model because of its balance between stability and flexibility. In the validation set, the error of 74.77 % of the predicted results was within ±3 dB(A), and the coefficient of determination R2 between all predicted values and measured values reached 0.9450, which indicates the physics-based PSO-BPNN model a potential aircraft noise assessment solution.