The maintenance of the railways is of paramount importance for safe and reliable transport. Eddy Current Testing (ECT) provides high-resolution time-series signals that capture subtle anomalies on the rail surface. This paper expands on previous analyses by combining classical time-frequency methods (short-time Fourier transform and continuous wavelet transform) and estimation of fractal dimensions with advanced feature extraction approaches, including wavelet sub-band decomposition, Hilbert–Huang transform, peak analysis and entropy metrics. Subsequently, a Random Forest classifier is applied to each set of characteristics, and we report comparative accuracy results on a dataset comprising rail segments with joints, welds, or squats. Experimental findings reveal that the Hilbert–Huang transform features yield the highest accuracy (93.28 %), while simpler features, such as peak counts, are less discriminative (46.93 %). These results underscore the effectiveness of using multiple signal-decomposition strategies and advanced analytics to robustly detect and categorize surface defects for better rail-maintenance decisions
Advanced Feature Analysis of Eddy Current Testing Signals for Rail Surface Defect Characterization / Quercio, M.; Santoro, L.; Sesana, R.; Riganti Fulginei, F.. - In: IEEE ACCESS. - ISSN 2169-3536. - 13:(2025), pp. 1-15. [10.1109/ACCESS.2025.3597079]
Advanced Feature Analysis of Eddy Current Testing Signals for Rail Surface Defect Characterization
Quercio M.;Santoro L.;Sesana R.;
2025
Abstract
The maintenance of the railways is of paramount importance for safe and reliable transport. Eddy Current Testing (ECT) provides high-resolution time-series signals that capture subtle anomalies on the rail surface. This paper expands on previous analyses by combining classical time-frequency methods (short-time Fourier transform and continuous wavelet transform) and estimation of fractal dimensions with advanced feature extraction approaches, including wavelet sub-band decomposition, Hilbert–Huang transform, peak analysis and entropy metrics. Subsequently, a Random Forest classifier is applied to each set of characteristics, and we report comparative accuracy results on a dataset comprising rail segments with joints, welds, or squats. Experimental findings reveal that the Hilbert–Huang transform features yield the highest accuracy (93.28 %), while simpler features, such as peak counts, are less discriminative (46.93 %). These results underscore the effectiveness of using multiple signal-decomposition strategies and advanced analytics to robustly detect and categorize surface defects for better rail-maintenance decisions| File | Dimensione | Formato | |
|---|---|---|---|
|
Advanced_Feature_Analysis_of_Eddy_Current_Testing_Signals_for_Rail_Surface_Defect_Characterization.pdf
accesso aperto
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Creative commons
Dimensione
5.6 MB
Formato
Adobe PDF
|
5.6 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/3002376
