Detection and monitoring of breathing cracks in high-speed train wheelsets using low-order vibration harmonics

Undetected axle cracks in in-service train wheelsets may lead to axle fracture and catastrophic consequences. Vibration components at integer multiples of the rotational frequency (n × Rev components) have been demonstrated as effective indicators for revealing the presence of cracks, since they originate directly from the nonlinear modulation introduced by crack breathing behavior. However, existing studies on these indicators rarely consider the test with health baseline established from long-term wheelset wear, nor do they provide sufficient validation under high-speed operating conditions. In this study, an experimental campaign was conducted to establish a long-mileage health baseline (14,500 km) under full-scale wheelset test bench laboratory conditions, and three-axis acceleration signals were acquired from sensors mounted on axle-box of the wheelset within a wide speed range (100–300 km/h). The effectiveness of n × Rev indicators for crack detection was evaluated. Furthermore, a non-dimensional n × Rev indicator is proposed to mitigate the influence introduced by long-term wear on crack indication. Through a data-driven implementation case, the advantages of joint decision-making based on multiple n × Rev indicators are demonstrated. The results provide both qualitative and quantitative support for the development of wheelset crack monitoring tools based on on-board sensor nodes.