The AIPM workshop, held in the week of 27 September 2024 at the Italian Railway Network (RFI) headquarters in Rome, marks a significant milestone as it served as the last face-to-face gathering before the project comes to an end in the fourth quarter of 2024. Hosted by the Italian State Railways (FS) and RFI, the workshop brought together key stakeholders to explore the latest initiatives in artificial intelligence (AI), particularly in terms of addressing and mitigating risks associated with seismic and hydrogeological challenges, with FS/RFI presenting their current projects.
In collaboration with RFI, the University of Bologna shared insights from a data-driven research project on the structural health of bridges, which used various machine and deep learning models, including random forest, Gaussian Bayes, K-Nearest Neighbours, Support Vector Machine, logistic regression, ridge/lasso regression, elastic net, and gradient boosting for the former, and feed-forward neural networks and Convolutional Neural Networks for the latter. The presentation highlighted both the lessons learned and the challenges faced throughout the research.
The Spanish railway infrastructure manager ADIF showcased its in-house economic optimisation of maintenance planning through AI, emphasising the economic benefits derived from integrating AI into predictive maintenance. AIPM members then engaged in discussions regarding the economic advantages of using AI in predictive maintenance for decision-makers, underscoring the growing recognition of AI’s potential in enhancing operational efficiency and cost-effectiveness.
The agenda for the workshop included a variety of presentations and discussions. On the first day, the welcome session featured UIC colleagues and an introduction to RFI. Presentations included the use of drones for inspections, an overview of RFI systems as created by SISMA and EWS, and an AIPM group activity focusing on the economic benefits of AI. The day concluded with reflections on the lessons learned from the project and a social dinner.
On the second day, attendees listened to a presentation on the RFI Diagnostic Train Diamante 2.0, followed by discussions on AI in bridge monitoring systems, led by experts from the University of Bologna. The workshop wrapped up with AIPM group activities focusing on the work carried out in 2024 and plans to continue the project from 2025 to 2027 with plans to delve deeper into the scalability, transferability, and explainability of AI models. Additionally, the scope of use cases will be expanded to include generative AI applications.
The workshop represented a critical opportunity for participants to exchange ideas, share experiences, and pave the way for future advancements in AI applications within predictive maintenance. The collaborative spirit and innovative discussions are expected to significantly influence the direction of AI initiatives in the transport sector.
