High-speed Railway Fastener Status Diagnostics Based On Vehicle Dynamic Response

As a major mode of transportation,railways are directly related to the economic and social development of a country.With the continuous progress of high-speed railway construction in China,various diseases in track structures are also constantly emerging.Among them,fastener failure is one of the important diseases that cause unstable operation of high-speed rail and endanger driving safety.How to quickly and accurately identify it is crucial to ensure the safety of high-speed rail operation.This article is based on the measured dataset of dynamic response of high-speed railway vehicles,and studies an intelligent diagnosis method for periodic faults of high-speed railway continuous multi segment fasteners that combines time-frequency analysis and machine learning.This article uses short-time Fourier transform for frequency domain preprocessing,maximum overlap discrete wavelet packet transform for filtering,generalized demodulation time-frequency analysis algorithm for signal decomposition,and sparrow search algorithm for optimizing support vector machine model for intelligent diagnosis of high-speed railway continuous multi segment fastener faults.Firstly,an acceleration sensor installed on a high-speed comprehensive detection train is used to collect the dynamic response signals of the vehicle in the normal and failure sections of the fasteners.Short time Fourier transform is used to determine the frequency domain range of the main concentration of fastener information in the signal.The maximum overlap discrete wavelet packet analysis method is used to filter out mixed noise and other irrelevant information in the vehicle dynamic response signal,and the phase function is fitted.Secondly,the signal filtered by the maximum overlapping discrete wavelet packet is subjected to generalized demodulation time-frequency analysis to obtain a series of Analytic signal components.In order to select the main signal components containing fastener information,calculate and compare each signal component with the effective value of the original signal,and screen out the signal components containing the main information of fastener status.Then,the effective values,energy contribution rate,and wavelength of the main signal components are used as characteristic indicators to characterize the status of fasteners.Evaluation criteria for the failure status of fasteners are formulated,and the sparrow search algorithm is inputted to optimize the support vector machine model for classification and diagnosis,achieving intelligent discrimination of fastener faults.Experiments have shown that this algorithm has high recognition accuracy.In order to verify the rationality of the recognition algorithm,traditional support vector machine models and PSO-SVM based classification detection methods will be used to classify and diagnose the status of fasteners.Compared with the proposed algorithm,it was found that the algorithm proposed in this paper outperforms the other two algorithms in terms of computational speed and accuracy.Finally,by drawing ROC curves and calculating KS and AUC values,as well as accuracy,recall,and F1 scores,the performance of the model is evaluated,further proving that the research results of this article have high practical value.