Dr Chi-Young Jung's research team from the Hydrogen Research & Demonstration Center at the Korea Institute of Energy Research (KIER) said the new method is 100 times faster than existing methods.
Carbon fibre paper, which helps to facilitate water discharge and fuel supply in fuel cell stacks, is composed of materials including carbon fibres, binders (adhesives), and coatings. Over time, the arrangement, structure, and coating condition of these materials change, leading to a decline in the performance of the fuel cell. For this reason, analysing the microstructure of carbon fibre paper has become an essential step in diagnosing the condition of fuel cells.
According to the team, real-time analysis of the high-resolution microstructure of carbon fibre paper has been impossible until now because obtaining accurate analysis results requires a process in which the carbon fibre paper sample is damaged and then subjected to detailed examination using an electron microscope.
To address the limitations of existing analysis methods, the research team developed a technology that analyses the microstructure of carbon fibre paper using X-ray diagnostics and an AI-based image learning model. Notably, this technology enables precise analysis using only X-ray tomography, eliminating the need for an electron microscope. Consequently, it allows for near real-time condition diagnosis.
MORE FROM ARTIFICIAL INTELLIGENCE
The research team extracted 5,000 images from over 200 samples of carbon fibre paper and trained a machine learning algorithm with this data. The trained model was able to predict the 3D distribution and arrangement of the key components of carbon fibre paper - including carbon fibres, binders, and coatings - with an accuracy of over 98 per cent. This capability enables the comparison of the initial state of the carbon fibre paper with its current state, allowing for the immediate identification of performance degradation causes.
The conventional analysis method takes at least two hours to complete. In contrast, the analysis model developed by the research team can identify the degradation, damaged areas, and extent of damage in the carbon fibre paper within a few seconds using X-ray tomography equipment.
In addition, the research team utilised data from the developed model to identify how design factors such as the thickness of the carbon fibre paper and the binder content affect fuel cell performance. They said they had also extracted optimal design parameters and proposed an ideal design plan aimed at improving the efficiency of fuel cells.
In a statement, research lead Dr Chi-Young Jung said: "This study is significant in that it enhances analysis technology by combining AI with virtual space utilisation, and clearly identifies the relationship between the structure and properties of energy materials, thereby demonstrating its practical applicability. We expect it to play a significant role in related fields such as secondary batteries and water electrolysis in the future.“
This study was conducted with the support of the Korea Institute of Energy Research's (KIER) research program and was published online in October 2024 in Applied Energy.
Radio wave weapon knocks out drone swarms
I hope they have assessed how easy it is to shield a drone against the attack. Hopefully the shielding would add too much weight.