Self-learning sensor systems monitor industrial manufacturing processes

The work to date has focused on analyzing real machining processes. Together with the partners, the project team carried out extensive series of tests on CNC milling machines and recorded sensor data under different production conditions. On this basis, the researchers have already been able to identify initial correlations between vibration signals, process parameters and tool properties. The data obtained forms the basis for AI models that, for example, recognize tool wear or allow conclusions to be drawn about workpiece quality. Initial prediction models already show that tool wear can be reliably detected and classified over time based on sensor data. This means that critical conditions can be identified at an early stage and maintenance measures can be planned in a targeted manner before quality losses or unplanned downtimes occur.

Learning directly on the machine

"AI has to go where the data is generated, i.e. directly into production," says Dr. Sebastian Wirtz, project manager at Fraunhofer IMS. "With GenSATIOn-Edge, we are creating the basis for machines to understand their own condition and react to changes at an early stage. This not only makes industrial processes more efficient, but also significantly more robust." The project team set up an initial software and data infrastructure that combines sensor data, machine information and user input and stores it for the long term. This creates a structured data basis on which the AI systems can continuously learn and adapt to new production conditions.

Intelligent sensor technology is the key

Parallel to the data analysis, the project consortium is developing a new type of IoT sensor node. This records relevant process signals directly on the machine and prepares them for AI evaluation. Fraunhofer IMS is contributing its expertise here, particularly in data processing and feature extraction. The aim is to implement the necessary analyses as efficiently as possible on resource-limited hardware. This creates the basis for later industrial use.

From the lab to production

In the coming project phases, the approaches developed will increasingly be tested under real production conditions. Further test series are planned, both in the laboratory environment and in industrial production. The focus here is particularly on the transferability of the AI models: they should not only work under controlled conditions, but also deliver robust and reliable results in practice.

GenSATIOn-Edge is supported by an interdisciplinary consortium. In addition to Fraunhofer IMS as project coordinator, the partners GED Gesellschaft für Elektronik und Design, Ruhr-West University of Applied Sciences, R&R Formentechnik and Formtec are also involved. The project has a total volume of around 1.7 million euros and will run from July 1, 2024 to December 31, 2028. It is co-financed by the European Union and the state of North Rhine-Westphalia as part of the ERDF/JTF program (funding code: EFRE-20800495).