‘Regulations block optimal use of big data’
Much more can be done with data than is currently the case, but regulations get in the way, says Sytse Bisschop from Dutch infrastructure manager ProRail. Data can only be used optimally if regulations keep up with technological progress.
Bisschop has been working for ProRail for six years. He is now the organisation’s system specialist in monitoring and analysis, and in this role focuses on track geometry. “What I often encounter is that we collect a vast amount of data, but a great deal of it is not used,” he explains. “Just from the measuring train, which runs on the whole rail network twice a year, we get mountains of data. This train measures 150 different aspects every 25 centimetres.”
Similarly, all switches on the track are regularly checked. “We scan all switches twice a year with lasers. In the past, we did that manually: then, we made a measurement every three metres. Now, with the measuring train, we measure every two to three centimetres.” That delivers a lot more data, but almost all of it is disregarded, says Bisschop. “Our regulations are not aligned to this.”
The regulations, which were established by ProRail itself at the European level, set out certain norms, for example about the type of data, and how much data is needed. They still state that data from every third metre is sufficient. As a result, the rest of the data is surplus to requirements and so is not used. “That is such a waste. You even start to question whether the investment in the measuring train was worth it.”
That is why it is important that regulations stay apace with all digital developments. Bisschop belongs to a standards committee at the European level. This working group decides, among other things, what a measuring train must measure.
That is not to say that measurements should be taken at random. Bisschop is calling for a practical approach, and also cautions against the ‘wild growth’ of data measurement. “You can go everywhere and say: ‘Yes, great, let’s measure!’ But if that does not align with what you need, or what the market needs, then it’s pointless.”
One of the other reasons that data remains unused is that it is inaccessible. “Or it does not fit in with what is being studied.” For example, there is often a focus on the data from incidents, but Bisschop says that all other data is also interesting. “We must know what the data looks like when nothing is wrong, in order to foresee incidents.”
Furthermore, when it comes to incidents, it should not only be about incidents that happened in recent weeks or months, but also those that took place years ago. Then predictions can be made about what needs to be renewed in a particular year. Bisschop names ‘Performance-Oriented Maintenance’ (‘Prestatie Gericht Onderhoud’, or PGO) as an example: “With PGO you really need to know what must be renewed seven years in advance. This allows you to estimate the costs for the contractor as well as possible.”
Data is often used to carry out predictive maintenance, and the next challenge is to see if this is being done too early. Here too, data, especially when collected over a long period of time, can help. “A rail has a lifetime of approximately 40 years. We don’t often replace a rail too late, but we do often replace one too early. With data, you can see if the top surface has been worn down more or less than normal.” Then you can decide to put the maintenance work back by a couple of years.
Moreover, it must also be effective. “With data, you can see that a train bumps on the track, but if you can’t send anyone to this track with the exact information, it’s useless.” Then it is necessary to know, for example, if the data means that there is something wrong with a wheel, or with a rail, and also what is precisely amiss, so that technicians can take the right material to the track straightaway. It boils down to the following: “How can we make this workable in practice? How does this help the man or woman on the track?”
Rail inspection vehicles
Bisschop is also realistic. Only recently has it become technologically feasible to take the next steps. Take video rail inspection vehicles, for example. These trains travel across the entire network to take video footage of everything.
“ProRail often receives complaints that it does not know where everything is,” says Bisschop. The video trains can help to counter this. “One of the cameras points straight ahead. Most images are now looked at manually, but if you could use them in a system with image recognition software, you could go through everything much more quickly and even make a combined picture. But image recognition software is only just becoming developed enough to do that.”
‘European rail sector can learn a lot from Scandinavian rail companies’
The European rail sector can learn a lot by looking at Scandinavian rail companies, says Peter Boom from Voestalpine Signaling. “In Scandinavia, they like to use new technologies and are often ahead of other countries.” The Intelligent Rail Summit will take place from 27 until 29 November 2018 in the Swedish city of Malmö. During… Read more ›
TU Graz is using acoustic waves to detect track failures
Graz University of Technology wants to use acoustic waves to detect track failures and worn components. In many countries, distributed acoustic sensing (DAS) is already being used to identify the location of trains. Ivan Vidovic is Research Associate at the Institute of Railway Engineering and Transport Economy at Graz University of Technology in Austria, and is working… Read more ›
‘Use machine learning to predict track maintenance’
Machine learning is an important asset to predict where track maintenance will be needed, Dirk Bothof of Asset Insight says. Machine learning is an artificial intelligence technique that gives computers the ability to learn. Asset Insight has developed a method with which track geometry can be predicted so that timely maintenance can be carried out…. Read more ›