A research group at CSEM has developed a system for monitoring production facilities that uses just one ultrasonic microphone instead of a multitude of different sensors. What makes this showcase remarkable is not only the product, but also the story behind it.
Picture: CSEM
SMEs in the manufacturing industry are increasingly faced with the challenge of making sure they retain their know-how. The reason for this is a changed work culture that leads to employees only staying at a company for four or five years – and no longer for ten, twenty or more. This makes holding onto ‘implicit knowledge’ challenging for companies. Implicit knowledge means know-how, something that a person can do without being able to explain what they do or how. Applied to the maintenance of production facilities and especially the recognition of deviations from standards, it is often experiential knowledge that is hard to find in textbooks or manuals. The experienced specialist intuitively knows when something is wrong.
Another factor is the intensifying global competition. This results in price pressure, which in turn means that the time during which production facilities’ machines are at a standstill is becoming increasingly expensive. Thus, such times must be reduced to a minimum.
Production-facility monitoring systems that determine whether the facility is functioning correctly, or whether the produced parts are of the right quality, respond to these two challenges. Such systems belong to the field of ‘evidence-based maintenance’ and are part of the Industrial Internet of Things (IIoT), as well as Industry 4.0: the vision of fully digitalised and networked production. In recent years, there has been more and more talk of ‘predictive maintenance’, the aim of which is to identify possible failures in effectively supervised machinery and take the necessary steps before a machine actually breaks down.
In most cases, the production facilities to be monitored are equipped with various sensors to detect deviations. Such sensors measure pressure, temperature development, movements or vibrations, for instance. Cameras are also used. This data is then automatically analysed – which, these days, usually involves a neural network. One of the main challenges in projects involving implementation of these monitoring systems is to reliably detect, within the data, those anomalies that actually impact the functioning of the production facility or on the quality of the parts produced.
However, the vast amounts of sensor data are also a challenge, as all this various information has to be coordinated and evaluated.
Together with its industrial partners Aurovis AG, KNF Flodos AG, Maxon Motor Ltd and Schurter Ltd, CSEM in central Switzerland has developed such a system, which continually monitors the state of production machinery. What makes CSEM’s system special, is that the multitude of sensors normally used for such monitoring systems are replaced by one ultrasonic microphone. Mario Russi, Project Manager and Senior Engineer, compares it to the following phenomenon: “All drivers immediately hear when something is wrong with their car’s engine and it needs servicing”. Ultrasonic microphones are actually not all that different; they register the sound waves caused by a machine. However, ultrasonic microphones are able to register frequencies far beyond the range of human hearing. While humans can only hear up to about 20 kilohertz, these microphones can record frequencies up to 150 kilohertz.
In terms of readiness, the CSEM system is at prototype level. It is able to determine whether the machine it is monitoring is working smoothly or whether there is a problem somewhere. Even though the system is not quite gold standard yet, it can currently identify 80 percent of faults and the general approach is promising: Existing production facilities do not have to be retrofitted and the microphones are much easier to install than many other sensors.
At present, the research group is working on a follow-up project to find out whether it is also possible to localise the faults detected by the system: Much like how humans or animals with two ears can locate sounds in a space and hear where they are coming from, a network of several microphones should also be able to determine where a sound source is situated. There are some evident difficulties when it comes to implementation. For instance, sound waves can be reflected or dampened by walls. Here, the crux of the matter is that sound waves behave differently according to the frequency range: While a sound with frequency A is swallowed by a wall, frequency B is amplified. It is correspondingly difficult to triangulate the various microphones so precisely that they detect minimal differences in volume and timing accurately enough to indicate the exact location of a fault.
CSEM is an internationally recognised Swiss technology innovation centre, jointly funded by the Swiss Confederation, the Cantons and a number of industrial companies. Its goal is to develop groundbreaking technologies with high societal impact and to transfer them to industry. Most projects are carried out in cooperation with companies. This particular project was set up to run a physical demonstration system to demonstrate the current technology’s capabilities to interested companies and show them how such a project can be approached in concrete terms.
While various sensors were being evaluated for this demonstration project, Philipp Schmid, Head of Research and Business Development for Industry 4.0 and Machine Learning, happened to run into Marco Gumprich in the neighbourhood. Gumprich is Managing Director at Elekon AG, the company that manufactures these microphones. As is customary in neighbourly conversations, they caught up on each other’s daily life, work and current problems. Marco Gumprich talked about microphones that are used on wind farms to classify bats and allow recordings far beyond the spectrum heard by humans. When asked how these microphones are tested, as it would probably be difficult to keep bats in an electronics lab, Gumprich said that zips and train movements, where metal rubs against metal, served this purposebecause they sometimes produce very high frequencies. During this conversation, Schmid realised that such a microphone would also provide information about undesirable states in production facilities. Schmid took the microphone with him. “There have been attempts to use microphones to identify machine status since the 1980s,” says Philipp Schmid. “With conventional microphones though, this has only worked to a limited extent in real production environments with a lot of noise interference”. The results with ultrasonic microphones are promising, as shown by the project’s progress to date.
This anecdote illustrates three aspects that are important in all digitalisation projects: Firstly, innovation cannot be planned. Instead, it tends to be a result of serendipity – of coincidences with consequences. Secondly, there are factors that make innovation more likely. These include, for instance, curiosity and an ability to think beyond ready-made solutions. Thirdly, this anecdote shows how important and momentous open conversations can be. They are essential to any innovation process because they stimulate new ideas and facilitate knowledge transfer.
