Battery aging can follow multiple degradation pathways, which influence future aging due to self-amplification, self-limitation, and interactions among mechanisms. This phenomenon is known as path-dependent aging. Understanding path dependency is crucial for reliable lifetime prediction and requires identifying distinct degradation pathways. In this study, k-means clustering is applied to aging data from 48 commercial lithium-ion batteries (LIB), cycled under 24 combinations of temperature and C-rate. Key degradation metrics, including capacity fade, pulse resistance, and degradation modes, are used to construct path-indicator spaces. Clustering with degradation modes reveals three distinct degradation regimes, characterized by proximity in both path-indicator and stress-factor space. These regimes are further validated using microscopic analysis of the negative electrode and distribution of relaxation times analysis. Based on the findings, general guidelines are proposed for designing dynamic usage schedules to test path dependency in LIB aging. Therefore, the methodology presented in this study provides a generalizable framework for characterizing battery degradation with a multi-dimensional feature space and introduces an unsupervised approach for identifying distinct degradation pathways. Additionally, the proposed method can help in building dynamic test protocols that trigger distinct degradation pathways and aid in the development and validation of lifetime prediction models.
Ramasubramanian et al. (Wed,) studied this question.