Researchers from an EV commercial battery solutions provider, multiple universities, and two Department of Energy research labs have collaborated on a study addressing the evolution of battery degradation under fast charging conditions. In their work, written up in ACS Applied Energy Materials, high-resolution micro CT imaging was used to observe battery microstructure in operando for batteries subjected to rapid charging followed by discharging. Hundreds of electrode layers from numerous time points were automatically segmented by Dragonfly deep learning models. Those deep learning results revealed the magnitude and spatial distribution of void formation–which can lead to delamination failure modes. Further, Dragonfly distance mapping enabled the direct characterization of the reversibility of electrode dilation to determine aspects of cycling efficiency.
First author Eva Allen reports, “I discovered Dragonfly just as I was starting my analysis, and found that its Deep Learning solution automated an otherwise overwhelming analysis. Combined with the distance mapping technique, these tools were so easy to learn because of the outstanding online training videos. It enabled us to go directly to the valuable, quantifiable answers we knew we could extract from our imaging experiments to connect real-time microstructure evolution to battery performance while using the entire dataset.”
