Condition monitoring and predictive maintenance for Copenhagen driverless metro

Every year, more than 73 million passengers travel with Copenhagen metro. The system operates 24/7, making maintenance a great challenge. Intervention schedules on the metro assets are time-based. Most components are checked or replaced when their remaining useful life is still acceptable. Maintenance costs are rising every year and the number of passengers is increasing. Given these trends, the market is gradually demanding more and more solutions regarding the optimization of maintenance activities. Data acquisition and real-time data analysis are gaining popularity, since they contain information about asset health (Condition Monitoring) and can be exploited to understand when interventions should be performed (Predictive Maintenance). This thesis is part of an early-stage study conducted by Hitachi Rail. The objective is to introduce a case study on the asset management digitization of the Copenhagen metro, starting from how maintenance is currently done, which systems are most impacted by it, and eventually formulate proposals for condition monitoring and predictive maintenance systems. The methodology includes 6 parts. The first one is “Selection of case study”, analyzing Hitachi Rail driverless solutions; “Maintenance activities overview” includes observations on components and maintenance procedures; “Data collection” discusses about the acquisition of relevant data in a two months on-field experience in Copenhagen; “Data analysis” deals with pre-processing, features extraction and machine learning algorithms; “System architecture” defines a scalable platform, able to manage both simulated and real-time data; eventually, “Data visualization” designs a front-end solution with a responsive dashboard built in Vue.js. This thesis lays the foundations for implementing a condition monitoring and predictive maintenance system, which in the future can improve the maintenance plan, based on the real use of the asset and not with a fixed schedule, reducing maintenance cost by intervening only at the right time. Furthermore, this system can optimize the first assessment in case of a failure event, having the correct maintenance team ready for each specific task, with the necessary tools and spare parts.