Deep Learning Segmentation of Operando X-Ray Microtomography of 18650 Lithium-Ion Batteries for Electric Vehicles

Abstract

The US Advanced Battery Consortium goals for 2023 call for low cost, fast charge, electric vehicle (EV) batteries with a 15-minute charge time at 80% pack capacity.[1] This milestone would have a huge impact on the universal market adoption of EVs. Unfortunately, fast charging at rates above 1C, defined as the current at which it takes one hour for a battery to charge/discharge to full capacity, imposes extensive deleterious effects on current lithium-ion battery (LiB) chemistries. High charging rates aggressively accelerate degradation mechanisms owed to structural damage due to both increased cycling temperature and inhomogeneous Li ion mass transport properties. The induced electrode damage can cause void formation between the active material and conductive-binder matrix and delamination from the current collecting foil. The now electronically isolated active material results in large costs on the capacity retention of the battery.In this study, we report measurements via operando X-ray microtomography of cylindrical cells used in EVs under simulated fast charge and drive cycles. Three batteries were measured during cycling after antecedent fast charging cycles, the 3rd, 5th, and 81st - 82nd cycles were analyzed to track morphological damage at different battery life points. To track the spatial and morphological degradation of both the anode and cathode structures, we employed a deep learning segmentation method using the U-Net convolutional neural network (CNN). Using a Euclidean Distance Mapping method, void formation is tracked spatially in 3-dimensions within the electrode coating. Insight into how fast charging induces structural damage will better inform research into fast-charge protocols and new battery chemistries for electrolytes, electrolyte additives, and novel electrode architectures. [1]Liu, Y.; Zhu, Y.; Cui, Y. Challenges and Opportunities towards Fast-Charging Battery Materials. Nat. Energy 2019, 4 (7), 540–550. https://doi.org/10.1038/s41560-019-0405-3.