Percorrer por autor "Risca, Diogo Ferreira"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Lifelong predictive maintenance for railway faultPublication . Risca, Diogo Ferreira; Figueiredo, Ana Maria Neves de Almeida BaptistaThe integration of advanced sensor technologies with machine learning algorithms has revolutionized fault diagnosis in railway systems, particularly at the wheel-track interface. Although numerous models have been proposed to detect irregularities such as wheel out-of-roundness, they often fall short in real-world applications due to the dynamic and nonstationary nature of railway operations. This thesis explores the challenges and opportunities of applying continual learning for predictive maintenance in railway systems, where the model’s ability to share knowledge between domains is critical to improving performance over time. By allowing the model to continuously learn and adapt as new data become available, continual learning overcomes the issue of catastrophic forgetting, which often plagues traditional models. The model retains past knowledge while improving predictive accuracy with each new learning episode, leveraging knowledge sharing mechanisms to adapt to evolving operational conditions, such as changes in speed, load, and track irregularities. Techniques such as experience replay and regularization-based strategies enhance model performance across multiple domains, making it particularly suitable for complex real-world environments. The methodology is validated through comprehensive simulations of train-track dynamic interactions, which capture realistic railway operating conditions. The proposed model demonstrates significant improvements in identifying wheel defects and other irregularities, establishing a reliable sequence for maintenance interventions. Future work will focus on field trials to assess the robustness of the approach in real-world railway environments, including challenges posed by track environments such as bridges and tunnels.