Rail surface damage method based on time-domain feature values analysis of vibration signals

Traditional rail detection methods suffer from low efficiency and high costs. This paper proposes a lightweight rail surface damage detection model using vibration time-domain signals to address the inefficiency and high cost of traditional methods. The approach involves signal preprocessing (least squares fitting, DC component removal, and low-pass filtering), sensitive feature selection, and quantitative damage analysis under various conditions. With compact architecture and high accuracy, the deployable model enables fully automated, real-time, and precise inspection on embedded platforms, offering a reliable solution for railway safety maintenance.