TU Vienna creates model to predict service life for tram infrastructure
Until recently, Viennese tram operator Wiener Linien had no centralized database containing net-wide infrastructure data. After implementing an infrastructure database, the operator is now working with the Technical University of Vienna to create a model to predict the service life for specific track segments. Associate researcher Johannes Kehrer from TU Vienna works on the project. “Right now, it is possible to predict how much renewal is necessary for each year for the overall network, but not for specific segments.”
The maintenance of the tram infrastructure is different from the maintenance of the railway tracks, Kehrer says. “The scope is different. First of all, the networks are way smaller.” In Austria, the entire railway network is about 5.600 kilometres long, whereas the tram network of Vienna is only 340 kilometres in total. “Secondly, lots of people use this small network. The ridership on the Viennese tram network is pretty much the same as on the Austrian railway network.” The track is crucial to optimally provide service to all passengers.
On top of that, there is limited time and money for maintenance. Track renewal efficiency is therefore of the essence. Regarding renewal, rail-wear is the decisive factor in a tram network. “The rails are really hard to exchange, mostly because of the track covering.”
There are plenty of challenges in the project to model predicted service life for each segment, Kehrer explains. One of the main issues is the fact that Wiener Linien does not have a lot of collected track-measurement data yet, because they only started using the current measurement system on a net-wide basis two years ago.
“The track measurement with the current system on a net-wide basis has just been established in 2016. And the network is not that big, so there aren’t many references or comparisons possible. So we came up with the idea to work with both the age of the segment and the current condition, combined with the influences and characteristics of each segment.”
Another challenge is the fact that is is not possible to make comparisons to other networks. “It is hard to learn from other tram networks in other cities and countries. Each track is too unique.”
So they have to work with limited information. But TU Vienna is also working on collecting and creating their own data. “There are 14.400 segments in the network. We try to learn which parameters influence the segment and in what way. Besides others, geometrical and operational factors such as load characteristics and track geometry are critical input parameters. So we can find a specific service life for the segment.”
They use the centralized information to create an algorithm. “We are working on an algorithm that learns how to make a prognosis for a segment, to find the service life. Since we have no time series to extrapolate for every single segment, one segment has to learn from the other segments.”
The model is supposed to serve another purpose as well, Kehrer says. “Once we have the expected service life for each segment, we can combine it with life cycle costs. We can see how these develop over time, to optimise track renewal. Wiener Linien is able to be more strategic with track maintenance.”
This is one part of a larger project of the TU Vienna and Wiener Linien. The model concerning the infrastructure, which is being created by Kehrer and his team, will be combined with a model to determine the operational consequences of the closing of specific segments. “Ultimately, the aim is to create a model that calculates the cost of each segment, and the cost of operational hindrances and the need for supplement services. Together, they will create a maintenance optimisation model.”
TU Vienna started with the project to predict service life for track segment based on centrally available infrastructure data in early 2018. The project is in the final stages of data processing. The intention is to finish the project in the spring of 2020.
Kehrer thinks that the railway sector can learn from the TU Vienna project. “There is very limited data available about this tram network, but we’re still attempting to create a model that can be used for predictive maintenance. That might be applicable to other railway networks where measurement data does not go back a long way. If it works for a relatively small network, it might work for big ones too.”