By Stefanie Nowak and Katrin Pudenz
There are around 1,000 escalators at German railway stations. Again and again, systems grind to a halt. Sometimes they are out of service for a long time - because the failure is not immediately noticed and the repair is often very complex.
What if if system failures could be noticed earlier - or avoided completely? Technical staff at Deutsche Bahn (DB) are now receiving support to do just this. AIM, an internal start-up at DB subsidiary DB Systel, has installed microphones on escalators at various railway stations. Software constantly analyzes the operating sounds. "If a system starts to sound unusual, the system sends an alert," explains Jens Glöckner, the founder of AIM.
Improved availability and lower maintenance costs
AIM stands for Acoustic Infrastructure Monitoring. As well as the status descriptions "running" and "not running", the sound analysis also enables the "not running well" status description. Once alerted by a warning from the system, technical staff can examine the systems where there are problems in a targeted way, thus ensuring that they do not fail in the first place. "This saves operating and maintenance costs and ensures, for example, that fewer passengers are stranded with their suitcases in front of a broken escalator," says Glöckner.
"THE ALGORITHM AGAINST ABRASION"
AIM began life as a project by three DB Systel employees in early 2017. "Back then, I took a walk around Frankfurt Central Station with service technicians from DB Services. Suddenly, one of the technicians stopped dead and said: 'Listen, something isn't right.' He then readjusted the steps on the escalator - and the noise disappeared. I've never forgotten what the technician said to me: 'Systems cry for help when something is wrong,'" remembers Glöckner.
This experience gave him the idea of supporting all escalators with an expert ear. Together with his team, he developed small boxes in which microphones record the operating sounds and send them to a cloud. "First, we collected lots of audio recordings of different escalators. Experienced technicians then assessed the sounds," explains Glöckner. This data was used to develop and train the software that now analyzes the operating sounds automatically.
On the way to predictive maintenance
"At the moment, the installation reports that 'everything is OK' or 'something is wrong' through the system. It cannot yet say what is wrong. But we are working on it," reports Glöckner. When the technicians return after having completed the work the system directed them to, the AIM team asks them to describe the malfunction. This data is fed back into the system and will hopefully help to identify the cause of failure sounds in the future. "In this way, we hope to move from condition monitoring to predictive maintenance in the long term - and thus be able to forecast exactly when maintenance work will be required in the future," says Glöckner.
Further fields of use
As Glöckner explains, AIM is not only concentrating on escalators, but is also examining options for use in many other fields. "Our technique is actually of interest to all systems in which moving parts make noises." The company is currently conducting various use cases, for example regarding air conditioning and ventilation systems, signal boxes and ICE wheel sets. "Even in ticket machines, it is possible to record what it sounds like when a coin falls correctly into the slot - or when something is not right."
Glöckner emphasizes, however: "We do not do any of the checks that concern safety." Nor does the system intervene in control. "Because all we do is listen, our solution is non-invasive."
Since February 2018, AIM has been an internal start-up within DB Systel with 16 staff. At the moment, various options for using the acoustic early fault detection system are being tested in proofs of concept. The data acquired in the various fields of use is fed into an analysis platform, which will soon be used as the basis for offering services such as acoustic status monitoring for systems.
Opening for the market
AIM's customers have so far come from within the DB group, but Glöckner has bigger plans. "We want to offer our system as a service on the external market and have already held some initial talks." The target group is operators who use systems from multiple manufacturers in parallel, he says. "Although the manufacturers offer their own predictive maintenance systems, they often only work for their own products or even only for the latest generation." AIM is currently working on training the software to work with sounds from as many systems as possible, he says - from any manufacturer and year of construction.
Background: DB Systel
The DB subsidiary aims to drive forward the digitalization of DB, in part by promoting the ideas of its own staff in a targeted way. The Skydeck Accelerator Program is a standardized process with which ideas are developed and tested in practice within ten weeks, with staff pitching their ideas to a judging panel. If they get the go-ahead for the program, they first work with experts to compile a business case and trial a prototype with a customer (proof of concept). The decision is then taken as to whether to pursue the idea, for example as a project or start-up. AIM was born out of this process.
Profile: Jens Glöckner
- Apprenticeship as an energy electronics technician, specializing in systems technology, at a mechanical engineering company
- Degree in industrial engineering
- IT consultant at a management consultancy firm
- At DB Systel since 2014
Jens Glöckner talks to Katrin Pudenz about AIM.