Simulation study of ionic current distribution and Li dendrite growth on the anode in lithium-ion batteries using support vector regression machine learning

Li-ion batteries (LIBs) could suffer damage such as contamination with foreign matter and non-uniformity of internal current density during transportation, which can cause serious accidents such as thermal runaway and explosions. Hence, a non-destructive monitoring system for evaluating their current densities must be developed. One such strategy is magnetic inverse analysis for detecting abnormal current distribution in the battery. Abnormal current density can be induced by the formation of Li dendrites, binder segregation, and/or electrode layer cracking. In particular, Li dendrites induce short circuits and explosions. In this study, we developed a non-destructive estimation model for the factors affecting Li dendrite growth at the anode of LIBs and evaluated the current distribution in the in-plane direction using support vector regression as the machine learning method. Two different cells with Ni-Co-Mn and LiCoO2 as cathodes and three separator structures (biaxially stretched, non-woven fabric, and foam structure) were evaluated. We also calculated Li dendrite precipitation, which is the starting point for degradation. Investigation of the local ionic current distribution revealed that the ionic conduction distance and overvoltage increased with increasing separator thickness. Furthermore, because of their structural morphology, foam-structure separators exhibited the most uniform in-plane ionic current and reaction distribution. Furthermore, the working potential revealed that the structure most sensitive to Li dendrite precipitation was biaxially stretched.