Study on the Extraction Method for Track-Side Acoustic Features Based on Cyclic Stationary Analysis

Abstract

Because of its non-contact measurement characteristics, trackside acoustic technology is now utilized for train bearing fault diagnosis. However, the collected acoustic signal produces Doppler distortions that can impact the accuracy of bearing fault diagnosis. Additionally, when a fault occurs in the train bearing, it is analyzed using cyclostationary methods. In this study, we combine bearing fault characteristics with Doppler distortion correction and cyclostationary analysis methods. The trackside acoustic test platform is employed to collect and test the fault signals from bearings. These signals are processed and analyzed using Doppler distortion correction algorithms and cyclostationary techniques. A comparison between time domain maps and power spectrum maps before and after correction reveals an increase in SNR (signal to noise ratio) and a more concentrated energy distribution within the fault signals—at least a 50% improvement is observed. To further validate our method’s effectiveness, we select existing TADS equipment from a depot to collect bearing signals for analysis and processing using our proposed bearing fault diagnosis method. Comparison of time domain maps and power spectrum maps before and after correction shows clearer overall images and amplitude increase of nearly 125%. Therefore, we have successfully developed a stepwise method for bearing fault diagnosis based on cyclostationary Doppler distortion correction.