Fraunhofer Institute for Microelectronic Circuits and SystemsIn the GenSATIOn-Edge project, self-learning AI models for condition diagnosis and prognosis based on vibration measurements are being developed in close collaboration with partners from research and industry. A key application is the prediction of remaining service life for tools used in machining. A specially developed edge sensor node enables resource-efficient implementation in embedded systems with local data processing for the first time, resulting in significant CO₂ savings compared to conventional AI.
Milling machine working on steel detail with lot of sparks.
GenSATIOn-Edge
Generative AI for adaptive training in industrial sensor systems
GenSATIOn-Edge
About the project
Motivation
At the beginning of July 2024, NRW State Secretary Silke Krebs (left) presented the official funding approval in the NEXT.IN.NRW innovation competition to the GenSATIOn-Edge team.
The metal industry is one of the most important sectors in North Rhine-Westphalia and faces the challenge of making its production processes more efficient, digital, and sustainable. Unexpected machine failures – for example, in milling machines – cause high costs and production downtime. Intelligent remaining service life prediction and predictive maintenance can minimize these risks and ensure the long-term competitiveness of the industry.
Goals
Example: Milling head for machining metallic materials.
The GenSATIOn-Edge project aims to make self-learning AI models, which continuously analyze machine status and detect wear and tear or impending failures at an early stage, usable on embedded edge sensor nodes in a resource-efficient manner. To this end, a novel, intelligent sensor node is to be developed and the complex neural networks optimized so that they can run resource-efficiently on the smallest microcontrollers. The practical application of this approach will be demonstrated by predicting the remaining service life of milling tools based on local data analysis using embedded sensor nodes in a productive industrial environment.
This increase in efficiency will result in significant CO₂ savings – estimated to be up to a factor of 400 compared to conventional AI.
Solution
Intelligent sensor node: Detection of vibrations on a milling machine and local data analysis using embedded AI.
Fraunhofer IMS is collaborating with GED Gesellschaft für Elektronik und Design mbH, Ruhr-West University of Applied Sciences, R&R-Formentechnik, and Formtec GmbH to develop an intelligent sensor system with local data analysis that can be attached directly to industrial equipment such as milling machines.
The system uses deep learning algorithms for self-monitored learning to monitor production processes in real time and detect wear at an early stage. A key challenge is to identify characteristics that indicate increased wear. To this end, the expert knowledge of machine operators is actively incorporated into the development of the models – a so-called human-in-the-loop pipeline.
The result is an adaptive, energy-efficient, and practical AI system that significantly increases the reliability of industrial manufacturing.
Funding
The project is co-financed by the EU and the state of North Rhine-Westphalia with funds from the ERDF/JTF program (EFRE-20800495).
