Fraunhofer Institute for Production Technology IPTResearch project "KIMEA"
The MEA: The Heart of Electrolyzers and Fuel Cells
The KIMEA project was launched to address the growing importance of green hydrogen by developing high-performance membrane electrode assemblies (MEAs) for this purpose. MEAs are the central component of electrolyzers and fuel cells; therefore, their quality determines the efficiency, service life, and cost-effectiveness of the systems. Current quality assurance methods—especially the measurement of complete I-V characteristics, i.e., the characteristic curves of current versus voltage—are time-consuming, lab-bound, and impractical for mass production. This leads to quality uncertainties, scrap, and increased costs.
Inline Quality Assurance Using Artificial Intelligence
The project will develop an AI-supported inline quality assurance system that automatically predicts the future electrochemical performance of an MEA based on process, sensor, and machine data. Substandard products are then detected and rejected during production. The partners will implement quality assurance using an edge device—a local industrial computer at the production line—that can be integrated not only into new but also into existing systems, regardless of the control system manufacturer or system type.
Support from an integrated language model
Scientists designed a LLM-based language model to analyze the complex process data and support machine operators by providing actionable recommendations on demand. Since it offers easy-to-understand guidance, the system simplifies operation in the production environment.
Validation of the AI with real machine data
To develop the AI, Fraunhofer IPT will generate extensive, structured datasets on a test facility and link them to actual measured performance values. After the AI and language model has been trained and integrated, a step-by-step validation process will take place—first on the institute’s test facility, then in the real production environment at project partner Laufenberg.
Transferability to different production conditions
The project partners are systematically investigating the extent to which the AI models can be transferred to other plant configurations, production conditions, and ink types. This results in a flexible and widely applicable quality assurance system.
Fraunhofer IPT is coordinating the KIMEA project and providing the central infrastructure for data acquisition with its MEA production facility. It is developing the data acquisition environment, integrating new sensor technology, and conducting automated test series to generate high-quality training data for the AI models. In addition, Fraunhofer IPT is training, evaluating the AI models and integrating them, along with the language model, into its own test facility. In the next steps, it will comprehensively validate the entire system and integrate the edge device on-site at Fraunhofer IPT. Furthermore, its researchers are investigating how the models can be transferred to new inks and production conditions.
Project Consortium
- Fraunhofer Institute for Production Technology IPT, Aachen
- Machine Tool Laboratory WZL at RWTH Aachen University, Aachen
- MT Analytics, Aachen
- Laufenberg GmbH, Krefeld
- Siemens AG, Leipzig
Funding
The KIMEA research project is funded by the European Union and the state of North Rhine-Westphalia as part of the ERDF/JTF program NEXT.IN.NRW.
Funding code: 20801734
Grant Program: NEXT.IN.NRW
Project sponsor: PTJ Projektträger Jülich