“Predictive maintenance is on the way: using data from sensors, we’ll be able to take trains offline just before a door stops working. Downtime will be shorter and much more efficient, and it will cost much less,” SNCF CEO Jean-Pierre Farandou said recently.1 For over six years, we’ve installed sensors on hundreds of trainsets in the Paris region and used their data to spot future breakdowns as early as possible. For you, that means continuing improvements in reliability and on-time performance.
Today, we’re taking predictive maintenance to a whole new level by rolling out this innovative technology on our TER regional trains. SNCF is a global leader in predictive maintenance for rail: we’re now equipped to analyse over 2,000 variables from 300 trains simultaneously—in real time.
We went behind the scenes with Cyril Verdun, head of the Rolling Stock Engineering Cluster West2 in Saint-Pierre-des-Corps, near Tours.
SNCF has been outfitting trains with predictive maintenance solutions for nearly six years now. Where are we now?
Right now we’re making progress with two different groups of trains. The first group doesn’t have predictive maintenance equipment built in, so we install IOT devices3—connected sensors—that collect the data we need. That’s called remote diagnostics. The second group already has built-in data networks and 3G or 4G SIM cards that can carry out predictive maintenance operations. With these models—our Regio 2N and Francilien trainsets, for example—we’re expanding the system to monitor new parts of the train.
What difference does it make when you expand predictive maintenance to new areas?
With remote diagnostics and predictive maintenance, we can eliminate half—even two-thirds—of breakdowns, so this is no gimmick. Put simply, it makes day-to-day transport better for our passengers. One good example is pantographs4, the flexible arms that transfer power from catenaries to trainsets. With predictive maintenance, we don’t have to climb on top of the trains to check the pressure they exert on the catenaries, which means we don’t have to take the train offline. Batteries are another example. Thanks to a number of new data points, we know exactly how much power they have left and exactly when to change them. Bottom line: we can keep more trains in service because we don’t need to do preventive battery maintenance every X number of years.
2000 variables from a single train are analysed in real time
Which trains are fitted with these new devices?
We’ve already got SIM cards or IOT sensors in over 300 Regio 2N and Francilien trains running on Transilien and RER commuter lines in the Paris region, and our TER regional trains have recently made the leap to predictive maintenance. And over the past 18 months, our teams have been working independently to deploy remote diagnostics on several models. For example, SNCF TER in the Grand-Est region is carrying out a comprehensive remote-diagnostics experiment with on a TER 2N NG trainset. It began running in mid-May, and we now have enough usable data to zero in on water levels in the toilets and network data. Experiments on a TER Régiolis and a Regio 2N are also underway.
In short, TER is a new front in your rollout of remote diagnostics and predictive maintenance. Why now?
Regional transport organizing authorities have realized that predictive maintenance is a game-changer that can improve on-time performance and reduce the number of TER breakdowns. With the domestic market set to open up to competition, our expertise in this area gives us an undeniable edge. What’s more, our TER 2N NGs and high-capacity AGC trains are now reaching mid-life. When they hit the 20-year mark, we do a complete overhaul, and that gives us the opportunity to suggest adding new IoT options and installing sensors on the components that are most likely to cause breakdowns, such as doors. Once installed, a sensor can tell us how often the door opens and closes, in how much time, so we can solve problems before they shut the train down. But we also use predictive maintenance on the tracks.
With remote diagnostics and predictive maintenance, we can eliminate two-thirds of breakdowns.
Cyril Verdun, head of Rolling Stock Engineering Cluster West
We’ve installed a fully equipped, modular maintenance bench on the tracks at our Châtillon TGV Technicentre near Montparnasse station in Paris. This enables us to analyse the condition of mechanical parts on the outside of the TGV, such as the axles and the brake linings and discs.
How does it work?
Modules with cameras, lasers and microphones are positioned in the ground, and they analyse the components as the train goes by. We use artificial intelligence and algorithms to interpret the measurement data and the images recorded by the bench, and that tells us when a component is worn or defective. The price tag for these set-ups is high, but they allow us to automate maintenance and optimize use of both our fleet and our maintenance resources. Some modules are still in the experimental or fine-tuning stage.
1 Excerpt from remarks by Jean-Pierre Farandou before the French National Assembly’s Commission on Sustainable Development and Regional Planning Commission, 2 October 2019.
2 Our Engineering Cluster West complex is part of SNCF Voyageurs. Its four sites are home to seven Rolling Stock Engineering units with a total of 300 engineers and technicians.
3 The “Internet of Things” (IoT) is the network of physical objects that are connected to the Internet.
4 A hinged device mounted on an electric train to collect power through contact with the catenary.