AI, possible solution to Swiss rail problems

Switzerland’s rail network is facing numerous problems in optimising train movements and avoiding collision risks. The management of the state-owned railway company in the Song Country wants to solve some of these problems with Artificial Intelligence.

Swiss rail problemsAfter a collision between a train and railway machinery in the canton of St. Gallen, the Swiss train drivers’ union is pushing for tougher rail safety specifications.

In mid-July last year, a train operated by Südostbahn was travelling through Walensee towards Sargans in the canton of St. Gallen. The Interregio 35 train of the cantonal railway company was approaching Unterterzen station and all signals were green. In fact, the train with 25 passengers on board risked a collision with two railway machines on the tracks at the station entrance.

Only the driver’s presence of mind prevented a catastrophe: he applied the brake when he saw the obstacle in front of him and the Interregio train stopped less than 200 metres from the machines. According to Hubert Giger, president of the Union of Locomotive Engineers (VSLF), it was extremely fortunate that there was no serious rail accident. After several incidents similar to the one at Unterterzen had occurred in the meantime, union officials intervened with the Confederation authorities, demanding stricter railway instructions.

Swiss network faces security problems

From the trade unionists’ point of view, the near collision at the entrance to Unterterzen station illustrates the safety problems facing the Swiss railway network. Devices installed along the route registered the presence of railway machines on the line, and on the central monitors of the CFF (Chemins de fer fédéraux suisses) the section was red, signifying that crossing was forbidden. Furthermore, the traffic controller contacted the site’s head of security, who assured him that there were no further obstacles on the tracks. Following this telephone conversation, the traffic officer cleared the way, putting the signal on green, whereas in reality, on the spot, the site machines were still on the tracks.

According to the Locomotive Engineers’ Union, this is not the first time that open lines have been announced and marked in green in the area of railway repair works and machines are still on the tracks. There are many reasons why this happens, from misunderstandings to high pressure on the sites. Most of the time, everything ends well, but the risk remains high, explain the trade unionists.

A head-on collision between a train and a railway construction machine at high speed would be devastating and many people could lose their lives. In 2019, in Thalwil in the canton of Zurich, a SBB fast train avoided a collision with a forgotten construction machine on the track in time. And then it was also the locomotive driver’s attentiveness that led to the catastrophe being avoided: when he saw the machine, he braked suddenly. And it’s not the only incident of this kind, if we go back a little further in time, and not all of them had happy endings. In 2016, near the Sihlbrugg railway station between the cantons of Zug and Zurich, a steam locomotive collided with a wagon, injuring two train drivers and 18 passengers.

In this context, from the point of view of the Locomotive Engineers’ Union, a simple phone call to the site managers is not enough. And that’s because the risk of an error is too great, as previous incidents have shown.

If electronic devices indicate that a line is busy, much stricter rules should apply, unionists argue. In such cases, train drivers should run on open track, i.e. slow down so they can brake in time in case of emergency. According to union leaders, this would only cause a ten-minute delay, but could considerably improve rail safety. These stricter rules should apply in particular, they say, to the first train to cross a section that has been upgraded or repaired and reopened to traffic.

Artificial intelligence used by CFF to solve Swiss rail problems

CFF officials have considered the unionists’ proposal, but have not given a pro or con response, saying only that they welcome any proposal that could lead to improved rail safety. Instead, another change is taking place in the way Switzerland’s rail network is managed.

Faced with the impossibility of expanding infrastructure in a country with difficult and saturated geography, CFF is optimising train movements with the help of Artificial Intelligence. Jochen Decker, Chief Information Officer (CIO) at CFF, is involved in the project, which is looking at projects already under way to organise flows or predictive maintenance.

It’s hard to find more difficult land for railway construction anywhere else than Switzerland. The Cantons are made up almost entirely of mountains, many of them already bored by tunnels and crossed by bridges. It is the densest rail network in Europe, making expansion impossible. All suitable land has already been used.

“Under these circumstances, there is only one solution left: to optimise,” said Jochen Decker recently at the IT Strategy Days in Hamburg. This is all the more necessary given that the state-owned railway company expects passenger numbers to increase by 30-40% by 2034 compared with the current figures.

Unlike other European railway groups such as Deutsche Bahn or SNCF, CFF is an integrated group that brings together all activities – passenger and freight transport, infrastructure, real estate – under the same umbrella, which has the advantage of facilitating the design and construction of investment objectives.

For several years, CFF has been implementing three optimisation programmes running until 2027, worth a total of EUR 1 billion. The first programme, for traffic management, aims to improve the operation of the lines, in particular by reducing the distance between trains. The second programme focuses on planning and production, to operate more kilometres with the same human and material resources, making sure that trains stay as short as possible and that train drivers spend as much time actually driving as possible. Finally, the third programme, the management of stocks, is designed to reduce wear and tear on materials and improve the operation of workshops.

Of the total EUR 1 billion allocated, EUR 20 million is earmarked for Artificial Intelligence. “This opens up all kinds of possibilities we didn’t have before,” says Jochen Decker.

AI technology interests the company both for its huge potential and especially for its low costs for infrastructure (track maintenance) and rolling stock (axle maintenance). By constantly monitoring wheel wear with cameras and sensors and then analysing the data obtained in this way, CFF is able to predict very precisely when a wheel needs to be replaced. The company’s managers pass this information on to the repair workshops, so that they can implement a genuine predictive maintenance system.

Under these conditions, the wheels of the rolling stock are replaced neither too early nor too late in the repair shop, which has the time and resources to deal with them at exactly the right time. “The prerequisite for this system to work is good quality data,” says Jochen Decker. But the system does not require a large budget allocation. The predictive maintenance project cost less than EUR 300 000.

The state-owned railway company in Switzerland does the same for track maintenance. A dedicated vehicle equipped with cameras drives along the tracks at 120 km/hour to assess the condition of the track. When a rail crack is detected on a route travelled by the machine, there is always the question of whether it is the same crack discovered the day before or a new one just a few centimetres away. In this case, the answer is Artificial Intelligence.

A third example of CFF’s use of AI is managing operations and optimising the use of routes. This could provide the solution to the problem raised by the Train Drivers’ Union. It is about answering the question of which train should be put into service and which should not. And this is also where the AI comes in, as there are many routes available to passengers if they want to cross Switzerland by rail.

Drawing the line, efficient planning of transport capacity using AI obviously costs much less than building new tunnels and new lines, which, as we have shown, is almost impossible due to overcrowding in the Cantons.

Flurina chatbot, a real success after six months of use at Rhaetian Railway

The issues surrounding the use of Artificial Intelligence are also of concern to the management of Swiss railway company Rhaetian Railway (RhB).

Last September, RhB started using an AI chatbot called Flurina, developed in collaboration with Microsoft and start-up ParetoLabs. Based on Azure OpenAI – a combination of Microsoft’s cloud computing resources and OpenAI’s AI knowledge and skills – the chatbot complements RhB’s existing service channels, making it easier for human staff to work with the increasing number of customer requests.

Since going live six months ago, RhB customers who have used the Flurina AI chatbot have already engaged in more than 12,000 sessions and asked more than 26,000 questions. “The data shows that it has been very well received. Using Microsoft Azure OpenAI for data processing in Switzerland has allowed the chatbot to be customized according to RhB’s needs to provide specific and up-to-date information. Customer demands are different now – they are international and need to chat 24 hours a day, seven days a week, which is why companies realise they need to use the latest digital technologies. AI helps them lighten their workload, giving them greater productivity and efficiency through more agile work processes,” said Primo Amrein, Head of Cloud at Microsoft Switzerland.

Since its launch, RhB benefits from the advantages of a productive chatbot, running successfully on the Azure OpenAI platform in a Swiss data centre, compliant with the strictest security, data protection and privacy rules. The chatbot has been enthusiastically received, thanks to the fact that it is always available and communicates with customers in German, English and Italian. This allows passengers to get essential information about bookings, travel routes and more.

Rhaetian Railway employees also benefit from the advantages of the new chatbot. Flurina allows Railservice employees to spend more time on phone and email enquiries and other projects. The digital assistant plays an important role in creating a modern workplace as part of the company’s digitisation strategy. This will help the transport operator to attract qualified people from the labour market.

Share on: