Method for Tracking the Evolution of Fatigue Cracks in Rail Turnouts Using Ultrasonic Guided Waves

Rail turnouts are an essential component of railway systems, and rail bottom is prone to fatigue cracks under long-term exposure to heavy loads, which can lead to structural failure and pose a serious threat to train operation safety. Since fatigue crack propagation is nonlinear and concealed, early detection is of great significance. Ultrasonic guided waves, due to their advantages of long propagation distance, low attenuation, and ability to detect concealed areas, are widely used in structural health monitoring. However, during the early stages of crack propagation, the cracks are extremely small, and the signal features can be easily obscured by noise, significantly complicating reliable detection. This paper proposes a virtual sensor-based anomaly feature extraction method, in which ultrasonic guided wave signals are segmented into sliding window subsequences serving as multiple virtual sensors. By combining temporal features and fluctuation patterns, the anomaly weight of each subsequence is evaluated to identify potential abnormal regions. Validation through a switch rail fatigue loading experiment demonstrates that the proposed method can effectively track crack development trends at the early stages, adapt to complex signal scenarios, and significantly improve the accuracy and reliability of structural health monitoring.