Transit Predictive Maintenance: An AI solution for targeted preventative train maintenance

The Problem

Managing a vast array of transit infrastructure presents a costly balancing act between two operational extremes. On one hand, over-maintenance through overly frequent servicing ensures high safety and reliability but results in unnecessary expenses and wasted resources. On the other hand, under-maintenance reduces immediate costs but compromises system safety and reliability, drastically increasing the risk of catastrophic equipment failures that ultimately cost far more to repair than regular upkeep. The core challenge for the City is finding the optimal maintenance frequency to minimize overall operational costs without sacrificing safety.

The Solution

To address this challenge, the predictive maintenance project leverages sensor and historical failure data to forecast when existing maintenance schedules are falling short. By identifying these gaps before a breakdown occurs, ETS can transition to targeted preventative maintenance. This data-driven approach effectively prevents catastrophic, expensive equipment failures, while avoiding the unnecessary expenditures associated with over-maintaining the system. The project was built with open-source tools and existing departmental resources.

The Impact

The system is expected to support predictive maintenance, providing the underlying framework necessary to transition the LRT network from costly fixed-schedule maintenance toward proactive, automated track defect identification.

Lessons Learned

The primary lessons learned center on the critical role of data infrastructure. A significant early hurdle was the lack of a consistent data pipeline from the Light Rail Vehicle (LRV) data acquisition system.

Launch year: 2021