Abstract Manager

AtkinsRéalis is committed to exploring the use of cutting-edge technology to optimize the Operations and Maintenance of the Canada Line. This presentation will provide a high-level overview of the Canada Line Digital Twin project, with a deep dive into our switch machine predictive performance module that was developed to support true predictive maintenance practices.

The Canada Line fully automated, driverless LRT, with automated switch movements. Operational failures of switch machines can cause significant delays to the travelling public, and often occur without indication from typical leading indications such as asset age, maintenance history, or component wear.  In the development of the switch machine predictive performance module, we sought to determine whether we could predict operational failures using machine learning algorithms applied to the available near real-time and historical data on switch operations and performance. In doing so we identified a significant correlation between switch swing time volatility, and near-term failure.  Using this data as input, we tested various machine learning algorithms against historical data and failures to determine an optimal digital model. 

Now, this module is using the current champion digital model to analyse switch swing movements in near real-time to determine a failure probability value, which has allowed our maintenance staff to move from a scheduled/reactive to a predictive maintenance model, reducing both maintenance effort and operational disruptions.