Covering Disruptive Technology Powering Business in The Digital Age

Home > DTA news > News > How big data helps German trains run on time
How big data helps German trains run on time
September 8, 2016 News

Sensors and analytics are making predictive maintenance work, for engineers and carmakers

For the past two years, a small group of Siemens employees has been working tirelessly to revive an out-of-date stereotype: that German trains run on time.

As long-suffering passengers are all too well aware, this has not been the case for years. According to the most recent European Commission report, for 2014, only 78.3 per cent of Germany’s long-distance trains were punctual — defined as arriving within 6 minutes of the published timetable. Of the 23 European countries studied, only two were deemed worse for delays: Lithuania, where only 74.8 per cent of trains ran on time, and Portugal, where it was 77 per cent.

Now, Siemens believes it can change all that, by combining two industrial disciplines: “big data” analysis and predictive maintenance. By fitting hundreds of sensors to its locomotives, which feed back data on the condition of components, it can ensure parts are replaced before they fail or cause delays.

In a joint venture with Renfe, the Spanish rail network, just one-in-2,300 Siemens-hauled journeys was more than five minutes late — a punctuality rate of 99.98 per cent, beating the country’s average 89.9 per cent and Finland’s exemplary 95.4 per cent. It seems German trains can indeed run on time … just mainly in Spain.

Siemens’ train initiative dates back to 2014, when the German engineering group realised that sensors and connected devices — the so-called “Internet of Things” — could allow it to provide customers with more than just hardware.

To maximise its use of data, the company transformed its locomotive-manufacturing site in Allach, just outside Munich, into a digital hub. It is there that 30 software specialists now sift through the reams of data generated by sensors on Siemens trains. They look for patterns and anomalies that indicate a part may need to be replaced, and ensure this happens during routine maintenance, avoiding unexpected breakdowns and delays.

In Spain, the system has proved so effective that Renfe is willing to offer customers a full refund if its high-speed service between Madrid and Barcelona is delayed by any more than 15 minutes.

Gerhard Kress, head of Siemens’ Mobility Data Services Center, says avoiding breakdowns is crucial for rail operators because just one obstruction can make several trains late. “London is one of these great examples,” he points out. “If you have one problem at rush hour at seven in the morning, you’re going to feel it until 12.”

Predictive maintenance, as a concept, is not new. It has been used on factory machinery and on assembly lines for at least 15 years, to avoid breakdowns and costly downtime. However, it is only recently that internet connectivity has extended the approach to consumer products and services.

Rolls-Royce and GE were early pioneers. Both realised that instead of simply selling aero engines for a one-off fee, predictive maintenance could let them to sell “power by the hour” service contracts, guaranteeing customers that engines could be in service for a set period.

Jason Kasper, an analyst at LNS Research, argues that this idea is revolutionising the way other hardware is sold and maintained. He points out that, in rail, it used to be the train operator’s responsibility to make enough money to fix locomotives and systems. “Now Siemens says, ‘I’m going to sell you the availability and guarantee you’ll have trains available when you need them’,” he explains.

Similarly, at carmaker Audi, the luxury unit of Volkswagen, predictive maintenance is being trialled to maximise the lifespan of parts. Audi’s latest models have sensors that can send data to its dealerships, enabling them to schedule services at optimal times.

Markus Siebrecht, senior director for After Sales at Audi, takes the example of brake pads. Today, if a customer books in a car for a routine service, and mechanics find the brake pads are 70 per cent used, they will change them because they might not last another 12 months. In future, though — once data privacy issues are resolved — dealers will be able to schedule appointments nearer the best time to change the parts, saving customers time and money.

“The whole digitisation of the car brings us closer to the actual usage of the customer,” Mr Siebrecht says — ensuring “maximum use from wear and tear parts,” and allowing customers to “drive as long as possible, safely, before [old parts] are swapped”.

Even elevator “out of service” signs could become a thing of the past. Last year,ThyssenKrupp, the engineering group, began installing sensors to track the acceleration, acoustics and vibration of its 1.2m elevators worldwide. As a result, it can now calculate a 10- or 20-day window when maintenance is likely to be needed. “We can make contact with the customers and say: ‘There is a risk — when would be the best time for you to take [the elevator] out of operation and maintain it, with the lowest impact to your operation?’” says Andreas Schierenbeck, chief executive of ThyssenKrupp Elevator Technology.

At present, an average elevator needs maintenance five to six times per year but, with better data analysis, this will fall to just three times, Mr Schierenbeck believes. This could save ThyssenKrupp “three-digit million” euros a year, he suggests.

This article was originally published on can be viewed in full