Intelligent and Convenient Detection of Micrometer-Scale Rail Corrugation

In the process of long-term operation of the railway, rail corrugation seriously affects the safe operation of vehicles and greatly increases the maintenance costs, so it is very necessary to know the degree of rail corrugation in advance.A wavelet packet timeconvolutional neural network is proposed to detect micrometer-scale rail corrugation by using car body acceleration, which is a low-cost and fast detection method.By taking the car body acceleration as input, the recognition accuracy of wavelet packet time-convolutional neural network for micron-scale rail corrugation at different wavelengths and amplitudes is compared.The results show that there is strong robustness, superiority of the wavelet packet time-convolutional neural network.The recognition accuracy of the method is 96.60% ~ 99.10% at different wavelengths and 95.40% ~ 100% at different amplitudes, which is a great improvement compared with the traditional model.