Accurate 3D representations of lithium-ion battery electrodes, distinguishing and labeling active particles, binder, and pore phases, can improve battery performance understanding. This work demonstrates a deep-learning methodology for reliable segmentation of volumetric electrode images where standard approaches fail due to insufficient contrast. A 3D U-Net architecture is implemented, and synthetic learning data, comprising realistic artificial electrode structures and their tomographic reconstructions, overcomes limitations of experimental imaging data to enhance network performance. The method accurately segments x-ray tomographic microscopy images of graphite-silicon composite electrodes, enabling statistically meaningful analysis of carbon-black and binder domain microstructural evolution during battery operation.

Simon Müller, Christina Sauter, Ramesh Shunmugasundaram, Nils Wenzler, Vincent De Andrade, Francesco De Carlo, Ender Konukoglu, Vanessa Wood