Ultrasonic Guided Waves-Based Monitoring of Rail Head: Laboratory and Field Tests

Recent train accidents have reaffirmed the necessity for developing a more effective rail defect detection system than what is currently in use. This paper focuses on a rail inspection prototype that employs ultrasonic guided waves and noncontact probes for automatic damage detection of defects in rail heads. The prototype utilizes an algorithm based on wavelet transform and outlier analysis to enhance inspection quality. Specifically, the discrete wavelet transform is implemented to denoise ultrasonic signals, generating a set of damage-sensitive data, which is then combined into a damage index vector used in an unsupervised learning algorithm for identifying anomalous rail conditions. The first part of the paper presents the operational prototype tested on a railroad track mock-up at the University of California, San Diego, which contained both surface and internal defects. Results from three experiments are discussed, highlighting the significance of feature selection for optimizing system sensitivity. The second part provides field testing results conducted in southeastern Pennsylvania, further validating the potential of this ultrasonic guided wave technique for practical applications in rail monitoring.