An Improved Marginal Index Method to Diagnose Poor Welded Joints of Heavy-haul Railway

Vertical acceleration of the axle box can usually reflect the high-frequency impulse of poor rail welded joints. To effectively diagnose poor welded joints of Heavy-haul railway, we propose a new method by combining wavelet packet decomposition (WPD) with adaptive synchro-squeezed short time Fourier transform (ASSTFT). Firstly, we decompose axle box acceleration of Heavy-haul railway by WPD and obtain a series of sub-signals. ASSTFT is utilized for each sub-signal and all of the time-frequency distributions are added together to obtain the overall time-frequency distribution of axle box acceleration. Next, we calculate the moving effective marginal spectrum of filtered time-frequency distribution. Finally, the marginal index of welded joint (WJMI) is obtained by normalizing moving effective marginal spectrum, and then we can give a threshold at 10 by statistical analysis of cumulative distribution. WJMI method is able to effectively detect and locate poor rail welded joints, and the filtering range of poor welded joints for heavy-haul railway is from 300Hz to 800 Hz. The proposed approach is proved to be valid and stable enough to diagnose poor welded joints.