Damage Detection in Railways Through Numerical Simulations and Machine Learning

This study presents a computational algorithm designed to simulate the dynamic interaction between railway vehicles and track infrastructure. The model, calibrated using experimental data, incorporates controlled stochasticity in its input parameters, enabling the generation of synthetic datasets representative of a wide range of operational and structural conditions. The algorithm was employed to simulate localized stiffness degradation in sleepers, and vertical acceleration signals were retrieved from the wagon axle to emulate onboard monitoring systems. The resulting synthetic data can be used to train machine learning models for detecting and localizing structural damage through pattern recognition in the simulated responses. The proposed approach highlights the potential of synthetic data to support the development of intelligent monitoring solutions within the framework of Structural Health Monitoring (SHM) methodologies