Talk · January 1, 2025

Uncertainty Analysis of Directional Eddy Current Probes in In-Line High-Speed Rail Defect Detection

Abstract

Machine learning has become a widely accepted tool for analyzing non-destructive testing (NDT) data, but its application must transcend best practices to fundamentally advance the field. Given the high safety requirements and potential catastrophic consequences of undetected defects, quality data is crucial for classifying rolling contact fatigue (RCF) types of local rail damages. A directional eddy current (EC) probe was integrated into a composite cutter device to simulate an in-line inspection environment. Task-specific uncertainty, based on the VDA 5.1 standard for in-line metrology evaluation, was adopted. The EC inspection results demonstrated that the developed probe could detect the targeted defect with a signal-to-structural-noise ratio (SNR) of approximately 10, up to a speed range of 1 ms. The uncertainty analysis indicates that higher-speed EC inspection affects the sensor signal, necessitating the grouping of feeding data for machine learning at each speed.

Non-Destructive Testing TechniquesRailway Engineering and DynamicsUltrasonics and Acoustic Wave Propagationhigh-speed railrolling contact fatiguerail defect detection

Uncertainty Analysis of Directional Eddy Current Probes in In-Line High-Speed Rail Defect Detection

Abstract

More by Meirbek Mussatayev