Sepehr Alemzadeh
Time-Series Anomaly Detection.
Rel. Paolo Garza. Politecnico di Torino, Corso di laurea magistrale in Data Science And Engineering, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (18MB) | Preview |
Abstract
Modern railway systems produce large amounts of sensor data that can be analysed to detect abnormal behavior, identify possible faults early, and support predictive maintenance. In the traction system, abnormal behaviour can affect train propulsion, operational reliability, and, in more serious cases, passenger safety. Detecting these anomalies in real industrial data is not straightforward. The signals are high dimensional, operating conditions change over time, and abnormal behavior may appear either as a sudden disturbance or as a gradual change. Another major challenge is the lack of reliable anomaly labels, which makes supervised learning difficult to apply. For this reason, this thesis focuses on unsupervised anomaly detection using data collected from a CRRC railway traction system in collaboration with BLUE Engineering.
The dataset includes 75 daily recordings from 88 sensors, sampled every second between February 2024 and June 2025
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
