Multivariate time-series modelling for monitoring underground transport infrastructure

Singapore's MRT carries millions of passengers a day, and reliability depends on the maintenance work behind the scenes. Track condition degrades over time, but conventional inspection is labour-intensive and reactive. The challenges compound: a large network, a short maintenance window each night, and heavy reliance on manual on-site protocols.

NTU, working with LTA and SBS Transit, replaces that pattern with AI-driven predictive maintenance. Custom smart sensors mounted around track infrastructure stream vibration patterns and acoustic signals. The AI model spots early signs of fastener loosening, wheel flange wear and track defects, then predicts potential failures so maintenance can be scheduled before service is hit. Data moves over 4G and 5G for real-time analysis. The team plans to expand deployment across more sections of the Singapore MRT network.

Caption language: en · Fetched: 2026-05-02

[Music] Singapore's MRT system serves millions daily, ensuring smooth and safe transport. But what goes on behind the scenes to maintain its reliability? Rail track conditions degrade over time, posing risks to safety and efficiency. Traditional inspections are labor intensive and reactive. The key challenges include managing extensive network, the limited maintenance window available each night and a significant dependence on venue inspection protocols. NTU in collaboration with LTA and SPS transit is revolutionizing rail monitoring with AI powered technology. AI enhances real infrastructure monitoring by transitioning maintenance from periodical schedule to a predictive approach.

AI provides insight into track conditioning on a daily basis, allowing for early detection of anomalies such as fastener loosening, wheel flash, and track defects. The AI model works by continuously analyzing data from the customized smart sensing system installed around the real infrastructure. It processes real-time inputs such as vibration patterns and acoustic signals to detect early signs of wheel or track defects. Using machine learning algorithm, the model identifies patterns and predicts potential failures, facilitating proactive maintenance. With the potential to be deployed across Singapore's MRT system, this AI powered solution detects rail defects in real time. With 4G and 5G connectivity, data is transmitted instantly for analysis, allowing timely intervention.

This technology enhances computer experience in Singapore by improving operational efficiencies and safety. Early anomaly detection helps identify potential risks and issues before they cause service disruptions, ensuring smoother, more reliable journeys for commuters. We look forward to scanning the system and expanding its deployment to cover more areas of the Singapore MRT network, further enhancing the efficiency and the safety of the entire transit system. With AIdriven rail monitoring, we are shaping the future of transportation smarter, safer, and more efficient. NTEU's innovation ensures a seamless journey for millions every day.