Research on the application of neural network in the prediction of metro vehicle noise

Metro trains generate a lot of noise during operation, which adversely affects the surrounding environment and passengers' health. This paper analyses the correlation between the time and frequency domain characteristics of various types of axlebox acceleration and the train noise Second Program Loader (SPL) through examples, and demonstrates the feasibility of predicting the noise A-weighted SPL based on the vertical vibration acceleration of the axlebox. Subsequently, one-dimensional and two-dimensional convolutional neural network-based noise prediction models for train passenger compartments are designed and the prediction accuracies of the two models on the test set are 88.7% and 89.1% respectively. In conclusion, the prediction model based on convolutional neural network in this paper achieves a more accurate prediction of the noise level in the train compartment.