Spatio-temporal analysis of high-frequency overload measurements offers a powerful means to monitor and predict railway track health. This thesis addresses the challenge of transforming raw telemetry data—captured by Lucchini RS’s Smartset system—into actionable insights for preventive maintenance. We develop an end-to-end analytical framework that (1) segments continuous strain‐gauge and GPS recordings into discrete journeys, (2) leverages OpenStreetMap geometries to snap each measurement to its precise location on the track, (3) normalizes dynamic axle loads using a Brimann‐derived model accounting for speed and curvature effects, and (4) applies a weighted DBSCAN clustering algorithm over a three‐dimensional feature space (chainage, timestamp, normalized load) to detect persistent overload hotspots. Chainage referencing and interactive mapping tools further enable intuitive visualization of both spatial degradation patterns and their evolution over time. Validation on real‐world Brenner–Innsbruck data demonstrates the ability to pinpoint segments with abnormally high and recurrent dynamic stresses—often coinciding with tight curves—thereby prioritizing areas for inspection. By integrating geospatial preprocessing, physics‐based normalization, and impact‐sensitive clustering, this work advances data‐driven railway maintenance planning, promising enhanced safety, reduced costs, and optimized intervention scheduling.
L’analisi spazio-temporale di misure ad alta frequenza di sovraccarico rappresenta uno strumento avanzato per il monitoraggio e la predizione delle condizioni delle rotaie ferroviarie. Questa tesi affronta la trasformazione dei dati grezzi—rilevati dal sistema Smartset di Lucchini RS—in informazioni utili per la manutenzione preventiva. Si propone un framework analitico completo che (1) suddivide le registrazioni continue di deformazione e posizione GPS in “viaggi” distinti, (2) sfrutta la geometria di OpenStreetMap per proiettare ogni misura nel punto esatto del tracciato, (3) normalizza i carichi dinamici mediante un modello Brimann che integra velocità e raggio di curvatura, e (4) applica un algoritmo DBSCAN pesato su uno spazio di caratteristiche tridimensionale (chainage, tempo, carico normalizzato) per individuare aree di sovraccarico persistente. Il sistema di coordinate di chainage e strumenti di mappatura interattiva consentono visualizzazioni intuitive delle zone degradate e della loro evoluzione temporale. I test su dati reali della linea Brennero-Innsbruck confermano l’efficacia nell’identificare segmenti con sollecitazioni dinamiche anomale e ricorrenti—spesso associate a curve strette—prioritizzando così le ispezioni. Integrando pre-elaborazione geospaziale, normalizzazione fisico ingegneristica e clustering sensibile all’impatto, il lavoro favorisce una manutenzione ferroviaria basata sui dati, migliorando sicurezza, ottimizzando costi e programmando interventi più efficaci.
Spatio-temporal analysis of overload data for track health monitiring: industrial application at Lucchini RS
NEMATPOUR, MILAD
2024/2025
Abstract
Spatio-temporal analysis of high-frequency overload measurements offers a powerful means to monitor and predict railway track health. This thesis addresses the challenge of transforming raw telemetry data—captured by Lucchini RS’s Smartset system—into actionable insights for preventive maintenance. We develop an end-to-end analytical framework that (1) segments continuous strain‐gauge and GPS recordings into discrete journeys, (2) leverages OpenStreetMap geometries to snap each measurement to its precise location on the track, (3) normalizes dynamic axle loads using a Brimann‐derived model accounting for speed and curvature effects, and (4) applies a weighted DBSCAN clustering algorithm over a three‐dimensional feature space (chainage, timestamp, normalized load) to detect persistent overload hotspots. Chainage referencing and interactive mapping tools further enable intuitive visualization of both spatial degradation patterns and their evolution over time. Validation on real‐world Brenner–Innsbruck data demonstrates the ability to pinpoint segments with abnormally high and recurrent dynamic stresses—often coinciding with tight curves—thereby prioritizing areas for inspection. By integrating geospatial preprocessing, physics‐based normalization, and impact‐sensitive clustering, this work advances data‐driven railway maintenance planning, promising enhanced safety, reduced costs, and optimized intervention scheduling.File | Dimensione | Formato | |
---|---|---|---|
2025_07_Nematpour.pdf
non accessibile
Descrizione: Thesis-Internship
Dimensione
3.45 MB
Formato
Adobe PDF
|
3.45 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/240559