Train information has been collected and analysed using big data tools for six months to monitor the operation of doors, compressors and air conditioning. This has made it possible to respond more quickly to the first symptoms of failure. The aim is to avoid over-maintenance resulting from mileage-based maintenance, as well as to make savings on maintenance actions.
The 5000 series trains used for line 5 are the first trains in the Barcelona metro system to benefit from the implementation of a digital platform to manage their maintenance.
The platform is LeadMind, a project forming part of the digitisation of maintenance coordinated by the Metro Rolling Stock Engineering department and carried out jointly with the railway company CAF.
The aim of the project, which started six months ago, is to improve the maintenance strategy to both reduce its cost and ensure the availability of the service and the safety of passengers using L5, the second busiest line on the network.
For this reason, during the last few months LeadMind has been facilitating the monitoring of the fleet’s health status, establishing behaviour patterns for the trains’ variables that can lead to alarms to detect equipment malfunctions before they develop into full breakdowns.
In addition, algorithms to maintain the equipment based on its condition, thus avoiding unnecessary maintenance visits to the garage, are being created.
Detection of abnormalities
During these months of activity, various indicators to detect abnormalities and anticipate breakdowns that could affect the service have been proposed, assessed and validated.
The platform’s algorithms have been able to detect abnormalities in the operation of doors, compressors and air conditioning very accurately.
On several occasions, the data generated has made it possible to take action quickly following early symptoms of failure. The foundations for condition-based maintenance that avoids the over-maintenance resulting from mileage-based maintenance and leads to savings in maintenance work, have thus been laid.
Collecting big data
The project, which started in late August 2020, is based on the collection of big data from 8 units out of the 37 that make up the line’s rolling stock.
The L5 maintenance team chose the most critical aspects, such as air conditioning, doors and air generation, to analyse their behaviour and see what the results of applying predictive maintenance would be.
The trains have been fitted with computers that can read all the variables of the on-board computer system (over 10,000 constantly updated variables) and send the data to the cloud over 4G for further processing.
Once the patterns are detected, checks are carried out to confirm if the equipment that triggered an alarm in the application really does have a malfunction on the train. If this is not the case, the algorithm is reviewed again until the results appearing in the application match the actual situation.