Predictive maintenance has become the new grail for railway rolling stock manufacturers. The aim is to produce a maintenance plan adapted to the real needs of the rolling stock. The aim is to avoid getting trains in the workshop that do not need it or, conversely, to anticipate an unexpected entry into the workshop following the detection of a possible imminent breakdown. To achieve this, trains must become IoTs, the Internet of Things.
Until now, the maintenance of rolling stock has been essentially preventive: after X thousands of kilometres, a train must return to the workshop to check the wear of the axles. But are all axles worn out after an arbitrary number of kilometres? The answer is often yes, based on past experience and a number of analyses.
In the early 1970s, at the SNCF in France, the maintenance policy gave priority to the ergonomics of people at work by providing as much as possible for daytime and weekday work, even if it required more rolling stock.
This worker-centred policy came to an end in the 1990s. Every downtime is a cost factor, since during its time in the workshop a train no longer enters the commercial system, as the seats are not sold.
Of course, the trick is to replace the stopped train with another train fit for service. But the multiplication of this solution means that for a given service you need, for example, 20 trains, knowing that 3 are in the workshop.
But with predictive maintenance, you could imagine that you would only need 16 trainsets, with one or two likely to be in the workshop as a reserve in case of a sudden incident.
To avoid buying 4 extra trainsets is obviously a step forward for small operators, but it means that you must to have a very sophisticated maintenance programme to ensure maximum availability, which varies over time.
Manufacturers have therefore understood the importance of making their railway products more ‘digital’. But not only for the operational service of the operators.
The integration of predictive maintenance planned from the product design phase should above all also allow the manufacturer to collect data and optimise software developments by integrating the appropriate sensors.
Therefore, data collection allows the manufacturer to meet the guarantee it offers on its rolling stock, and to see any defects that it can correct instantly.
This is why many manufacturers offer a sales contract with maintenance. This is a win-win operation because the operator has less to worry about the availability of the trains, which is guaranteed by contract. And the manufacturer has direct access to vital technical data to improve its products.
With this definition, marketing is moving towards a new business model oriented towards ‘maintenance-as-a-service’, with the manufacturer playing the crucial role of ‘turnkey’ solution provider.
What does this predictive maintenance consist of? For example, monitoring of train batteries, remote monitoring of the condition of doors or monitoring of critical elements, such as axles, oil pressure or other important elements.
In Poland, for example, Alstom has installed the TrainScanner, which uses 3D cameras and lasers to automatically perform conditional and predictive maintenance on the wheels, brake pads and carbon strips of the pantograph, as well as on the subframe and body shells of Pendolino trains.
After automated inspection, the data is transmitted to Alstom’s HealthHub platform, which translates the raw data into actionable information, using rule-based algorithms, leading to the calculation of a health index for each asset.
In addition to the traditional multimeters and oscilloscopes, workers can access the readings of the many sensors present in each train, as well as their history, by connecting directly to the trains via their laptop: a valuable diagnostic aid! The software used allows the status of the rolling stock to be viewed when the trains are covered by Wi-Fi and can share data.
However, the profusion of sensors means that an astronomical amount of data is collected, which must be sorted out.
The major difficulty for railway companies is to improve the information system. At the SNCF, this involves a team of 27 data scientists and the installation of sensors that are easy to integrate so as not to disrupt the systems. ‘I realise that the more uses there are, the more optimisations there are to be made in our algorithms,’ Cyril Verdun, director of the Maintenance Engineering Department, told the Journal du Net, indicating that his department uses 16,000 algorithms to analyse train data.
The importance of trained staff
But as always, when it comes to the digital sector, a lot of education is needed. ‘Too many of our technicians still check a train visually,‘ said one operator.
Maintenance is also about trained staff. To have an ‘good’ technician, trained in a speciality and having assimilated all the specificities of railway maintenance, between 3 and 5 years of work are necessary, according to some sources. A little less according to others. The topic is sensitive…
If digital is often (too) perceived as a « loss job maker », it can on the contrary be a social lift, even if this raises fears among non-graduates. The simple worker no longer exists in maintenance. The sophistication of operations has transformed the profession into that of a technician, which is much more rewarding.
The digitalisation of trains is underway and will continue to increase. We might as well be prepared for it to make rail a transport tool of the future.
06/02/2022 – By Frédéric de Kemmeter – Railway signalling and freelance copywriter
Suscribe my blog