Field validation of enhanced area estimation method for GPR-based subsurface fouling detection in operational railway ballast

Résumé fourni par la source

This study proposes an enhanced area estimation method for detecting ballast fouling using Ground-Penetrating Radar (GPR) A-scan data. A conventional method quantifies signal strength by integrating the full waveform but tends to overestimate strong reflections from near-surface regions, which can reduce attenuation effects at greater depths where fouling commonly progresses. To address this, the proposed method introduces a depth-specific analysis that segments A-scan signals and applies amplitude normalisation, allowing the computation of Area Percentage Error (APE) between reference and target waveforms across multiple integration ranges. GPR data were collected along a 132 m long operational railway testbed using a 350 MHz antenna, and APE-based fouling detections were performed at 10 representative sections. These detections were validated through direct excavation, and the method successfully identified all severely fouled zones, including those with significant fouling below the surface but without surface-visible fines. In addition, sections exhibiting surface-visible fines showed consistently high APE values, indicating strong agreement. To evaluate comparative performance under identical field conditions, both the conventional and proposed methods were applied to the same GPR dataset. The results showed that the proposed method more effectively captured subsurface fouling, especially at deeper layers, by reducing the bias introduced by strong reflections near the surface.