H2SkaProMo

In order to achieve the climate targets adopted by the German government, alternatives to fossil fuels are needed. Here, hydrogen can play a key role as an energy carrier. Green hydrogen makes it possible to reduce CO₂ emissions in both industry and the transport sector significantly. At present, however, the production and use of green hydrogen is not yet economically viable. A key factor here is to increase productivity in the manufacturing of fuel cells. Due to the currently still low demand and the dynamic fuel cell stack development, there is an increased need for flexibly scalable production systems. This is where the ”H2SkaProMo” project comes in. Three different approaches of production lines are to be developed on prototype level: A fully manual, a partially automated as well as a fully automated line. Manual assembly processes achieve high product variance with high flexibility (pre-series). Semi-automated assembly processes are used to meet rising but volatile demands (small series), while automated assembly processes enable mass production to the highest quality standards (large series).

All three prototype lines have the ability to map the entire assembly process from separation, stacking and compression to clamping and the subsequent leak test and final load test. Along the process, changes to the physical product, process and equipment are to be synchronized with the help of a digital twin, in which product- and process-related data are recorded, processed and made available throughout. The transformability is ensured by the development and integration of innovative mechanical, electrical and software modules. The use of relevant information contained in the data is to be improved by AI-based ”quality maintenance” and ”predictive maintenance” services. 

The main task of Fraunhofer IZFP within the project is the development of methods for nondestructive quality inspection. One focus is the optical control of the stacking process including the development of automated segmentation and classification algorithms based on neural networks. Here, the position and orientation of the bipolar plates as well as the membrane electrode units for the gripper system must be determined very precisely so that the components can be placed with appropriate accuracy during stacking. Each step must also be monitored during stacking in order to avoid stacking errors through inline readjustment. The challenge here lies in the high stacking speed (target value 1 Hz), the low tolerance of 200 µm in the process chain ”alignment of the individual components” and ”stacking”, and the resolution of the optical measurement technology. The problem is to be solved by combining conventional methods and deep learning approaches such as ”semantic wireframe” and ”keypoint detection” algorithms.

The second research focus is on the development of methods for leakage testing of fuel cell stacks. Current methods only allow a global statement on the leak tightness of the entire stack. Local localization is currently only possible by scanning methods, which are usually very time-consuming. By further developing spectral thermography, a two-stage process is to be developed that detects leaks globally and localizes them precisely in a second test step. This will enable not only the detection and rejection of leaking stacks in the process, but also a repair strategy based on the results of the leak test as part of sustainable and resource-efficient production.

Here, too, the experimental technical phase is accompanied by the development of machine learning algorithms, with which test results can be independently classified and information transmitted to upstream and downstream process steps.

At the end of the project, the prototypes of the assembly lines will be built and evaluated. The aim is to systematically process the obtained results and to derive further optimization potential. In the sense of an ”Open Lab Factory”, the transfer of the results to industry is also planned after project completion (December 2024).

To this end, the demonstrators are to be integrated into a test factory in order to make them available to all project partners for product tests such as leak testing of fuel cell prototypes.

Project partner:

• HYDAC Accessories GmbH
• INNOCISE GmbH
• Munzinger Maschinenbau GmbH & Co. KG
• Schaeffler Technologies AG & Co. KG
• SUSI & James GmbH
• Umwelt-Campus Birkenfeld
• Universität des Saarlandes, Lehrstuhl für Montagesysteme
• URT Utz Ratio Technik GmbH
• XENON Automatisierungstechnik GmbH