Research Of Rail Fastener Loosening Detection Based On Wavelet Packet Energy Spectrum

The rail fasteners were one of the important components in the structure of track. Under the repeated load, the fastener might be loosened, which may cause the train derailment accident. With the development of high speed train, it is important to detect the loosen fastener, however, the manual inspection method and the method based on computer vision can't detect the fastener state changing and the fastener loosening degree in real time. When the rail fasteners were loosened, the dynamic characteristic parameters of the track would be changed, which could be presented by the dynamic response of the structure, and we can detect the rail fastener status by using the vibration signal differences before and after fastener loosening conditions. Due to wavelet packet analysis method having the advantages of multi-resolution analysis and good time-frequency localization, this paper was analyzed the acceleration signal collected by the sensor nodes combined with wavelet packet theory, vector similarity theory and statistical theory to study the rail fastener loosening conditions. The following several aspects of research have done:(1) The traditional wavelet packet algorithm defect aspects'reasons were analyzed and the improved algorithm was proposed. The band aliasing and band interleaving of the wavelet packet algorithm were analyzed. The improved algorithm was proposed to eliminate the frequency aliasing phenomenon by the Fourier transform and inverse Fourier transform theory, and to eliminate the frequency band interleaving phenomenon by adjusting the order of wavelet filter banks. The simulation test was used to prove the effectiveness of the improved wavelet packet algorithm.(2) The maximum dissimilarity index and the energy accumulation variation index were used to detect the rail fastener state, and was verified the effectiveness of the method by the rail loading test platform. The position index combined with several measurement nodes was put forward to achieve the position of the loosen fastener. The relationship between the energy accumulation variation index and the rail fasteners loosening degree was studied under single conditions. Finally, the effect of recognition results of the loose degree was affected by fasteners loosening degree, fasteners loosening numbers and noise level.(3) The maximum dissimilarity and the energy accumulation variation index were separately used to analyze the measured vertical acceleration data. The variation range of the maximum dissimilarity and the energy accumulation variation index of the rail fastener under unloosening conditions was obtained.