A pencil lead break-triggered, adversarial autoencoder-based approach for rapid and robust rail damage detection

Detecting early-stage damage is essential for railway maintenance, ruling out potential risks that could undermine railway ride comfort and safety. Ultrasonic testing methods, featuring high precision and non-destructive characteristics, have gained widespread use for on-site inspections in modern railway systems. However, current ultrasonic testing remains a highly complex technique that requires expensive ultrasonic devices and trained professionals for operation. This study presents a novel approach for rail damage detection utilizing a disposable mechanical pencil. By intentionally breaking the pencil lead on rail surface, the accumulated potential energy is released in the form of ultrasonic bursts which are acquired by sensors mounted on the rail. The rail damage diagnosis is empowered by an adversarial autoencoder (AAE) which learns representations of ultrasonic signals induced by pencil lead break (PLB). A damage-sensitive indicator is developed based on the Jensen-Shannon Divergence (JSD) between the AAE model output distributions of the baseline and an unknown signal, facilitating rapid and accurate damage diagnosis. Both laboratory experiments and on-site verifications were conducted to validate the proposed approach. The results demonstrate the effectiveness of the damage detection framework in identifying rail damage, exhibiting excellent robustness and reliability. Comparative studies are also conducted to demonstrate the adaptability and effectiveness of the proposed method against field testing environments. The research outcomes of this study will significantly contribute to the development of more efficient on-site inspection techniques for railway maintenance and sustainability.