Week 3

Hi everyone! This week, I continued building on my AI research, this time zooming out from bridges specifically to look at how artificial intelligence is being applied across transportation systems more broadly.

I looked at various journals and case studies on the application of artificial intelligence technology to railways, highways, and transit systems. A good resource for me was a survey article published in Sensors called “A Survey of Artificial Intelligence Predictive Maintenance Methods for Railway Infrastructure.” It reviewed machine learning and deep learning techniques (including neural networks, support vector machines, and random forests) for analyzing things like the geometry of the track, vibration levels, and images of the infrastructure. What stood out to me most after reading these articles was the similarities among the many predictive maintenance methods as they relate to the infrastructure on railways; the fundamental principles for measuring conditions of railway infrastructure are very similar to those for measuring bridge conditions.

I found an article in Scientific Reports about an explainable machine learning framework developed for the Metro of Porto, Portugal, which created a real-time predictive maintenance pipeline that processes sensor data, flags anomalies, and provides natural language explanations of these outputs to make machine decision-making understandable to engineers and not just data scientists. I was fascinated by the concept of explainability. A model that can tell you why it flagged a structure as at-risk is far more useful to a real engineer than one that just spits out a number.

What all of these articles reinforced is something I keep coming back to: the lack of visibility into asset condition, sparse monitoring, and manual analysis hinders traditional maintenance practices. Artificial intelligence will eliminate all three of these obstacles and allow for continuous, data-driven monitoring of all types of maintenance assets, including but not limited to track and bridge assets.

Outside of research, I have begun to look into potentially volunteering at transportation museums or getting direct access to bridges to better understand the physical side of infrastructure maintenance. There are many excellent sources of transportation/infrastructure exhibits in transportation museums throughout New York City, such as the New York Transit Museum, the New York City Department of Transportation’s historic archives, and the Intrepid Sea, Air & Space Museum. These museums would provide a tremendous resource to develop my familiarity with the physical presence of infrastructure, as well as how to work with and understand the actual environment (scale, how each component relates to one another) that inspectors encounter while working on the infrastructure. Access to these structures would only improve my ability to carry out my analysis and make my final framework a true representation of reality. 

Work Cited

Gihub.org. “Sensors and Machine Learning for Predictive Maintenance.” Global Infrastructure Hub, 2020, www.gihub.org/infrastructure-technology-use-cases/case-studies/sensors-and-machine-learning-for-predictive-maintenance/.

Meira, Jorge, et al. “Data-Driven Predictive Maintenance Framework for Railway Systems.” Integrated Computer-Aided Engineering, vol. 30, no. 3, 2023, journals.sagepub.com/doi/abs/10.3233/IDA-226811

Pérez-Pérez, Enrique Jose, et al. “A Survey of AI-Enabled Predictive Maintenance for Railway Infrastructure: Models, Data Sources, and Research Challenges.” Sensors, vol. 26, no. 3, 2026, www.mdpi.com/1424-8220/26/3/906.

Veloso, Bruno, et al. “An Explainable Machine Learning Framework for Railway Predictive Maintenance Using Data Streams from the Metro Operator of Portugal.” Scientific Reports, 2025, www.nature.com/articles/s41598-025-08084-1.