Digital methods for the production of fuel cells and electrolysis sys-tems

The rapid advances in machine learning methods offer the potential to revolutionize the production of fuel cells and electrolyzers. Precise data analysis can be used to optimize production processes, detect production mistakes early, and significantly extend the life of these technologies. Machine learning enables the identification of critical parameters that increase production efficiency and make operations more reliable and cost-effective. In this way, these key technologies of the energy transition can be used more sustainably. The integration of these digital methods into the production of fuel cells and electrolyzers is therefore a very promising research topic. Fuel cells and electrolyzers share similar components and processes. Therefore, synergies can be exploited in their manufacturing. In addition, data analytics and machine learning methods are already widely used in the development and production processes of electrochemical energy storage systems such as lithium-ion batteries.

Production of electrochemical energy converters

The production of electrochemical energy converters, such as fuel cells and electrolyzers, involves highly complex processes that require precise quality requirements and sophisticated expertise. The central component of each of these technologies is the stack of the system. In the stack the electrochemical redox reactions take place in each of the individual cells. The assembly of this module involves the joining of precisely manufactured components such as bipolar plates (BPP), catalyst-coated membranes (CCM) and gas diffusion layers (GDL). The manufacturing processes for these components involve numerous manufacturing steps and control parameters, the interactions of which are critical to the quality of the end products.

Digital and connected production

Further research focuses on the detailed analysis of these processes using digital methods. The method not only aims to understand and optimize the relationships within individual manufacturing steps. Rather, the goal is to investigate which parameters influence the quality and function of the individual components and the overall system across the different production steps. With this a deeper understanding of how to further increase the efficiency and reliability in production of fuel cells and electrolyzers should be realized. Machine learning methods can be used to comprehensively evaluate the large amount of data generated during the production of fuel cells and electrolyzers. This data must be labeled in order to be utilized to create a machine learning model under supervised learning. With the data obtained in the ongoing production process, the model will be able to predict the quality of the manufactured component. Of course, this depends on the type of model used and the information learned. In this way, the model provides information not only to optimize the process parameters of individual production steps, but also to better understand the interactions between different manufacturing processes. This data-driven analysis can be used to identify critical parameters that influence the quality of components and the overall system. For example, with this digital and connected production it is possible to trace differences in the stack quality to the variations in the production of bipolar plates, gas diffusion layers, or electrodes. This means that adjustments can be made at an early stage of manufacturing to further increase the efficiency and reliability of the end products.

The challenges of the digital production

Integrating digital data analysis and machine learning methods into fuel cell and electrolyzer manufacturing offers many benefits. However, this technological shift also presents challenges, particularly in the areas of data quality/availability, investment, and system integration.

Data quality and availability are the key challenges in applying machine learning techniques in any field of study. On the one hand, this data must be collected from sensors and production monitoring systems, and on the other hand, it must be assigned to the corresponding manufactured system component (labeled). One of the main difficulties is ensuring that the data collected is accurate, complete and consistent. Incorrect or incomplete data can lead to faulty models and therefore inaccurate predictions.

Another challenge is the high investment associated with a digital manufacturing strategy. Equipment for analyzing, storing, and evaluating the data to be collected represents a significant investment that comes with risks.

A final challenge is the integration of digital machine learning into existing production structures. Depending on the age of the production facilities, they may not be compatible with modern data analysis applications without increased effort.

Conclusion

From these statements it can be concluded that machine learning methods offer immense advantages despite the challenges they present. Through detailed data analysis, not only individual production steps but also complex relationships between different processes can be identified and optimized. Even with the high initial investments, this leads to a significant increase in efficiency in the production of fuel cells and electrolyzers.



Quellen:

Achim Kampker, Heiner Hans Heimes: Elektromobilität Grundlagen einer Fortschrittstechnologie (2024)

Jay Lee, Hossein Davari, Jaskaran Singh, & Vibhor Pandhare: Industrial Artificial Intelligence for Industry 4.0-based Manufacturing Systems (2020)

Julia Meiners, Arian Fröhlich, Klaus Dröder: Potential of a machine learning based cross-process control in lithium-ion battery production (2022)

shoplogix: 6 Key Challenges of Digital Transformation in Manufacturing (April 2024)

Newsletter

Hier informieren wir regelmäßig über alle wichtigen Entwicklungen im Bereich des Wasserstoffs, der Brennstoffzelle und des Elektrolyseurs.

Downloads

Hier informieren wir regelmäßig über alle wichtigen Entwicklungen im Bereich des Wasserstoffs, der Brennstoffzelle und des Elektrolyseurs.

Whitepaper: In-line Viscosity Measurement

Download a white paper on innovative solutions in the areas of real-time process monitoring and quality assurance