Railway fastener system damage detection method based on convolutional neural network

Abstract

The invention discloses a railway fastener system damage detection method based on a convolutional neural network. Based on a train-rail coupling dynamics calculation and analysis model, spring stiffness reduction is utilized to simulate fastener damage, and simulation calculation is carried out to obtain vibration acceleration responses of the steel rail under different irregularity excitations,different damage positions and different damage degrees so as to construct a big data set. A one-dimensional convolutional neural network is designed, the established network is trained by using the data set, and cross validation and parameter adjustment are carried out. The trained network is subjected to performance test on a test set, and the test result shows that the detection method has highdetection precision and robustness. Furthermore, a dynamic experiment of a key section of the target monitoring line is carried out, an actually measured big data set of system damage is constructed,and the big data set is used for carrying out transfer learning on a pre-trained one-dimensional convolutional neural network model.

Description

technical field [0001] The invention belongs to the technical field of rail transit, and in particular relates to a method for detecting damage of a railway fastener system based on a convolutional neural network. Background technique [0002] In recent years, the scale of my country's railway construction has grown rapidly, and real-time monitoring of its health status is critical to the safe operation of high-speed railways. Facing such a large-scale line network, traditional manual inspections are time-consuming and laborious. Therefore, it is necessary to develop a method for automatically detecting the health status of the railway network, especially for the damage detection of key components of the line structure in some special sections. At present, most of the intelligent detection methods for structural damage are based on computer vision. The limitation of this method is that the early damage and invisible defects of the structure cannot be identified, such as the...

AI Technical Summary

Problems solved by technology

Feature extraction performs signal processing on the original data, such as Fourier transform, wavelet packet transform, etc., to extract data features; on the basis of feature extraction, the features most sensitive to structural damage are selected as the input of the classifier; although this type of method Partial success has been achieved, but its recognition accuracy is greatly affected by the features obtained in the original data; on the other hand, manual extraction and selection of features is time-consuming and laborious, and requires a lot of prior knowledge

Images