DeepTech Octonion announced the launch of a new artificial intelligence feature in its Brainium Edge AI solution. Focused on sound analysis directly from a microcontroller, it allows manufacturers to optimize the upstream detection of anomalies and guarantee a better efficiency of their machines to their customers.

Initially designed to analyze movements and vibrations, Octonion, whose mission is to bring intelligence to industrial products, integrates for the first time acoustic analysis at the microcontroller level in its Brainium Edge AI solution, in the service of conditioned monitoring for industrials with predictive maintenance.

Acoustic analysis to detect anomalies

In addition to vibration measurements, the sound analysis proposed by Octonion makes it possible to truly put machines or industrial equipment “under continuous listening”. These two intelligent analyses, sensitive respectively to mechanical vibrations and movements and to the sounds produced by them, make it possible to detect the state of the machines and potential operating anomalies.

Sound analysis is performed using acoustic analysis, which is one of the three most reliable recognized methods for predictive maintenance, along with thermography and vibration analysis. Indeed, for certain types of frequencies, sound analysis allows industrialists to optimise the upstream detection of anomalies, which vibration analysis does not allow in certain cases or over a longer period of time.

This monitoring system then makes it possible to listen to the acoustic frequencies, compare them and define the “normal” behaviour of the machine. If an abnormal behaviour is detected, the system warns the operator, by message (sms, email, …) or other predefined signal. If the noise increases again, the system can trigger an alarm which allows to stop the machine. Alerted automatically, the industrialist can then avoid a malfunction, which allows him to optimize his maintenance costs while limiting his production stops and incidents.

Cases of use

Octonion is currently working on three use cases with industrials in acoustic analysis: human presence detection to activate/deactivate the necessary equipment in a building (e.g. heating activation for smart building); activity detection for elevator doors to identify failures during their operation; predictive maintenance for boilers.

The sound analysis according to the artificial intelligence model specific to each machine or equipment is performed either by machine learning supervised by humans, or without supervision thanks to the “AI Studio” developed in January 2019 by Octonion, its autonomous machine learning solution integrating inference functions embedded in microcontrollers.

The AI Studio at the disposal of industrials

The AI Studio allows manufacturers of industrial products to create and operate their own Artificial Intelligence models for their machines. The sounds analyzed can be of different types such as slamming, percussions, shocks, blows, water leaks…

Finally, thanks to Octonion’s Edge IA solution, the user’s data is completely secure as it is neither shared with a third party nor sent to the cloud. Indeed, the acoustic data is analyzed at the edge (i.e. close to the source) that generated the sounds, thanks to the power of the learning machine and inference, rather than being sent directly to the cloud or to a central server located elsewhere.

For Ongan Mordeniz, co-founder of Octonion:

“Industrial maintenance is becoming part of the factory of the future. The addition of this new acoustic analysis functionality reinforces our artificial intelligence technology, in addition to vibration and motion analysis. Octonion’s priority is to enable Industrials to make their equipment Intelligent in order to know how to best anticipate breakdowns for their customers, which generate additional costs, production delays due to line stoppages and, consequently, lower productivity.

With these vibration and acoustic analyses at the heart of industrial equipment, we give manufacturers access to reliable, pre-processed data, as well as the technology to manage this intelligence from start to finish. With this new data that can be used in their factories as well as at their customers’ sites, manufacturers will have a real field feedback of the state and use of their machines in order to improve them in the future. »

Translated from Maintenance prédictive : Octonion lance une fonctionnalité d’analyse acoustique pour les industriels