Mapping internal deformation in 4695 batteries through a combination of X-ray computer tomography and machine learning

Understanding the internal stress and deformation in lithium-ion batteries (LIBs) is critical for battery design and management, especially for 46XX silicon-graphite batteries with significant swelling. X-ray computer tomography (CT) is an effective non-contact measurement for internal stress based on electrode thickness variation. However, X-rays severely attenuate, resulting in limited measurement resolution and accuracy in large-size 46XX LIBs. Meanwhile, it is difficult to establish the complex internal stress distribution within the spirally wound jelly rolls. In this study, we propose a machine learning-based approach to map the deformation of positive and negative electrodes in a 4695 LIB using CT images. We have identified two structural details. The first is a uniform protrude in the 100° direction of all electrode windings, deviating from the ideal Archimedean spiral. Secondly, both positive and negative electrode ends cause slight deformation of surrounding electrodes, leading to an increase of local stress. The deformation mapping of the jelly roll reflects the simultaneous radial expansion and lateral translation in the 100° direction after charging. The mapping of electrode stack thickness has unveiled the radial distribution of stresses along the stacking direction and the maximum stress located in the 4th region or the 16th∼20th turn. These findings offer crucial insights for optimizing the spiral structure in the manufacturing process of LIBs. In addition, this method can be extended to analyze the complex deformation of LIBs with various configurations and may hold potential for in-situ monitoring based on fast and low-resolution CT imaging. •Analyzing CT images with clustering algorithms to quantify LIB deformation.•Positive and negative electrode ends exacerbate the surrounding deformation.•Expansion and translation of jelly rolls during lithiation are accurately quantified.•Thickness distribution of electrode stack reveals inflection of radial stress magnitude.