The Device Chronicle interviewed Rolf Härdi, CTIO, Deutsche Bahn. Rolf is a leading domain expert and can explain the trends in adoption of IoT in railways, and edge and cloud technologies across the rail industry.
Rolf begins the interview by saying that the digital transformation and IoT in railways is progressing well, and specifically for Deutsche Bahn in two ways. Rolf says Deutsche Bahn has embraced digital and that they are quite happy where we are at with digital working for what would have been traditionally “office” employees. Rolf continues “We have everything in the cloud. Covid has really accelerated things. We are reducing office space and increasing home working. In the last 19 months, a totally new working culture has been introduced (into Deutsche Bahn) and it’s not only on the technology but also on the culture side. So hierarchies have been reduced. People talk with each other. Everything is far more flat than it was before.”
Homologation for IoT in railways
Rolf then addresses the central role played by the process of homologation. This involves proving to the authorities that a train or a device is safe. Certification from the authorities is needed to run a train. Rolf says this is very difficult, very expensive, and it’s even more difficult with software. “So naturally, we are very reluctant to change software on a train. The train design is such that you make a hard distinction between “operational” software and “non-operational” software. And you have to make sure that those two cannot be mixed together. He says “So if you have a software just reading sensors and physical controls that influence a sensor, it is not as delicate as if you have a software which controls your brakes for instance.”
Digitalisation of rail assets
Rolf moves on to address the integration of digital technologies into the physical assets in rail. He says “If you look at the asset sites, which are trains and train stations, the digital transformation could be much faster but the challenge is that normally Deutsche Bahn uses these assets for up to 40 years and so each train has a long lifecycle.” Rolf provides an example of an ICE train that was built 40 years ago and is still in operation. In the rail industry, he says, you don’t change things very quickly. One of the main reasons for this patient approach is the regulations, especially the safety regulations. “Whatever you do, you have to prove it is safe. It’s not like a computer where you do a little bit of programming and then you try it out then it works, and maybe sometimes it doesn’t work. We can’t do that on trains. So everything that is digitised, we (Deutsche Bahn) have to first design and prove that it is really safe. And then you have to prove it is safe to your authorities, and this is an entirely different game for you to do something.”
Maintenance is an impactful IoT in railways use case
Rolf describes maintenance as a very important use case for IoT in railways. In rail, he says, there are fixed maintenance cycles. For example, maintenance must be carried out on a train door every 30,000 hours, or that many kilometres and so on. “Performance-based maintenance is a very impactful area so after so many miles, we are changing a bearing, oil or a rubber element just at the time before it’s not going to perform anymore. With digitalization, the appropriate sensors and systems, operational services can really be optimised and costs minimized.” Rolf says that other maintenance cases with digital detection can be more complex and difficult to address. Rolf provides the example of having to change the bearing on the wheel every 600,000 kilometers. This, he says, is a common occurrence. “So you install a sensor and it will tell you just one month before the bearing will actually turn bad. You have to change it now but how can you prove that the sensor is measuring the right signal all the time? And you can apply this dilemma to almost everything within the rail.”
Collaboration is key for Deutsche Bahn
Rolf says Deutsche Bahn is at a moment in its digital strategy, where implementing such processes and systems is being done in conjunction with third parties such as OEM suppliers. Rolf says “We (Deutsche Bahn) do have the technical knowledge to go into certain systems such as wheel sets. “On one of the projects we are working on, we can predict wheel wear and tear. And we can monitor wheels with wayside monitoring equipment. These are lasers every time the wheel is passing by, and they are going to measure that wheel and it’s not just the tear in the diameter that is measured. It also measures changes to the shape or the profile of the wheel which is very important for safety.” With algorithms, Rolf says his team can predict precisely when a wheel needs to be changed. This is very important given that wheel sets would have an order lead time of about 12 months. He says “We (Deutsche Bahn) can’t go to the local DIY shop and purchase the part. We actually have to order it 12 months in advance. So this is really where digitalization is our biggest use case.”
Digital twin at Deutsche Bahn
Rolf describes a strategy where Deutsche Bahn is building digital twins so they can “virtualize” their assets. This is being done in partnership with the OEM suppliers and the sub suppliers. For instance, Rolf says, manufacturer doors are notorious for failing on trains due to different reasons, including vandalism, wear and tear, effects from adverse weather conditions and so on. So with condition-based monitoring with a sensor, Rolf says, it is possible to actually predict a failure, as the door would begin to open more slowly and then more slowly over time, or more electric current would have to be used to actually close the door as performance degrades. With the result, these failures can be counteracted so the door will not fail and this is all in insights from the data. The data is imported into a digital twin for analysis and the insights are then used for maintenance.
Collaborative partnerships & IoT in railways
Currently, AWS is Deutsche Bahn’s main cloud provider. Microsoft and Google are also used depending on specific application needs. Rolf says that there are many data lakes within Deutsche Bahn. Different methods including machine learning and AI are used to help the analysts make sense out of the data. Rolf explains that Deutsche Bahn also has an engineering approach where an engineer would look at the data for a bogey and a certain type of metal for instance metal to understand how the combination would behave.
As referenced above, Deutsche Bahn is also cooperating with the OEMs and the sub suppliers. Rolf explains that through collaboration, Deutsche Bahn wants to optimize services. In an example shared by Rolf, the partner has detailed knowledge of the door. “The partner is the only stakeholder with the full organizational understanding of all the details on what’s happening there with the door. The best approach is to collaborate, share the data, and for both parties to benefit. Deutsche Bahn’s approach is this win-win situation. The partner gets all the experience of the Deutsche Bahn train door that gives them the chance to improve their products because they’re going to know exactly how it behaves in service and how it reacts to certain material impacts. Deutsche Bahn gets an improved operation of the doors and reduced maintenance costs.”
Rolf stresses that digitalization combines technical, cultural and business challenges. He says “Digitalization only works when we all work together and the industry is not really equipped for that. So we work with patterns where one stakeholder tries to shield their knowledge from other stakeholders as competitors. Each stakeholder is tied into supplier relationships and out of this culture, we are trying to encourage a more open collaborative culture yet in an appropriately controlled way. Digitalization only really works if we collaborate and share data.”
Data combined is more powerful than data in silo
Also, Rolf believes that data in isolation such as that about a train door doesn’t help you much, unless it can be combined with other datas such as environmental data, data on how many passengers go through, data on speed and so on. You must combine information from different sources so that you can be in a position to actually implement the use cases in condition-based monitoring and condition-based maintenance. No one stakeholder can get access to all the datas and do it on their own and that’s why the collaboration is needed.
Risk averse in IoT in railways
In general, train operators are averse to making changes to software on the edge side. This is something train operators don’t like to do that often. Rolf says “We (Deutsche Bahn) have machine learning but we only use it on the cloud side. We (Deutsche Bahn) have some algorithms on the edge side. I’m not quite sure if they qualify for machine learning. They are hard-coded. And also we don’t really have at this time a possibility to change software remotely. In OT security, IT security is one of the big concerns so access to the train beyond reading out information is almost impossible. So what we try at the moment is that if a train goes into the yard, the train in standstill goes into a maintenance mode and then certain elements of the train can be accessed for remote testing. For instance, remote testing of the air conditioning units, the brake unit and so on. None of this testing is actually during the operation of the train.
Operational level and the Comfort level
For mission critical updates for embedded devices on trains, the update to the device has to be done locally. Rolf puts this scenario in perspective: “I said before we have different levels of IT infrastructure. Deutsche Bahn has to function at an “operational level” and also the “comfort level.” This differentiates between “control” systems with firmware embedded; and passenger announcement systems, information screens which are Windows or Linux-based systems and which would be updated regularly. The updates on the mission critical systems in trains still have to be done locally with laptops and code signed USB. “Yes, that’s where we are at the moment, updates in mission critical systems are still done in a very traditional way.” Rolf concludes by saying “I was a commissioning engineer in the 1990s and it’s actually still the same process today as was used in those times.
Countering obsolescence with OTA software updates
Rolf concludes by highlighting an increasing challenge that could be addressed by OTA updates. This is product obsolescence. Rolf says the life cycle of hardware is not 40 years anymore. There is a desire to upgrade hardware (and software) to a certain extent over time. OTA updates could help address this challenge.”
We wish Rolf and his colleagues at Deutsche Bahn well as they continue on the digital transformation journey in rail.