Design of a rail track irregularities measurement system for non-inspection vehicles

05/09/2025

Railway track geometry maintenance is crucial for the comfort, safety, and efficiency of railway operations. However, traditional inspection systems rely on costly specialised vehicles, limiting inspection frequency, and resulting in large intervals between maintenance activities. To address these limitations, a vehicle’s dynamic response-based methodology is developed to classify rail track irregularities based on a multibody (MB) model of the EM120 inspection vehicle and machine learning (ML) techniques. MB formulation is used to simulate realistic operating conditions of railway vehicles, helping to analyse correlations between the vehicle behaviour and track irregularities. The dataset to develop the ML models is built using data from 1014 km of track with known track irregularities, classified according to its condition, and bogie frame motion, including accelerations and angular velocities. Different Support Vector Machine (SVM) models are then trained, and evaluated with performance metrics. The results highlight the strong correlation between track irregularities and bogie frame dynamics, enabling accurate classification of track sections. The Gaussian SVM achieved the highest performance metrics across the tested case studies, showing its reliability and robustness. This research demonstrates the feasibility of continuous track monitoring using a sensor module mounted on the bogie frame to measure its dynamic behaviour, enabling the development of scalable, data-driven maintenance strategies.

Authors in the community:

• 

  J

  João Pedro Pereira Torres

  ist196411

Supervisors of this institution:

• 

  S

  Susana Margarida da Silva Vieira

  ist45346

• 

  H

  Hugo Miguel Pacheco Magalhães

  ist163344

mechanical-engineering - Mechanical engineering

por - Portuguese

Instituto Superior Técnico