Microstructural issues in Li-ion batteries – like foreign inclusions, cracks, or delaminations – can affect a battery's performance, safety, and aging characteristics [1]. Detecting these types of defects in assembled cells requires high-resolution imaging, often generating thousands of images per cell for analysis. Previous research has demonstrated that deep learning techniques, such as convolutional neural networks (CNNs), can effectively classify micrographs [2] or individual defects [3] to identify these microstructural problems in Li-ion batteries automatically. Looking at the defect population of aged cells – as opposed to the previously examined pristine cells – it can be observed that new kinds of defects emerge. Those new defect types can differ vastly in terms of size, morphology and frequency when compared to the previously known defects, making the currently used approaches obsolete. In this work, we evaluate approaches from instance segmentation to detect and individually quantify these new kinds of defects and compare the results to object detection approaches.

References

[1] G. Qian, Cell Reports Physical Science, vol. 2, no. 9, p. 100554, 2021, doi: 10.1016/j.xcrp.2021.100554.

[2] O. Badmos, J Intell Manuf, vol. 31, no. 4, pp. 885–897, 2020, doi: 10.1007/s10845-019-01484-x.

[3] A. Jansche, FEMS EUROMAT, 2023.