Cycle-wise detrending method for strain-based aging characterization and feature development of batteries: Case study of 18650 NMC Li-ion cells

Accurate assessment of lithium-ion battery state of health (SOH) is essential for ensuring their safe and reliable operation, particularly under abnormal conditions such as overcharging. This study investigates the use of strain measurements to evaluate the SOH of a 3000 mAh Samsung 30Q cell subjected to standard cycling and an overcharge event introduced midway through its expected lifetime. Continuous strain gauge-based monitoring revealed distinct changes in residual deformation following overcharge, indicating irreversible structural modification. A strain-decomposition framework was developed to separate cyclic deformation into offset, slope, and residual components. The offset captured irreversible baseline shifts associated with cumulative degradation, while the slope reflected reversible, SOC-dependent expansion and contraction. The residual component isolated nonlinear intercalation-driven deformation, providing a sensitive indicator of evolving internal electrochemical conditions. Feature-based analysis of the residual strain incorporated distribution-shape features, in addition to both time-domain and frequency-domain metrics. Maximum strain and root mean square frequency (RMSF) exhibited the strongest cross-cell consistency, with average Spearman correlations exceeding 95% and median absolute differences below 4%. Principal component analysis further revealed distinct mechanical signatures between charging and discharging. Charging PC1 was dominated by waveform-shape descriptors such as skewness, shape factor, and crest factor, whereas discharging PC1 was governed primarily by strain-magnitude features, including RMSF, maximum strain, and mean strain, aligning more with feature correlation results. These complementary modes captured approximately 99% of the total variance across the first five charging and discharging components. Post-overcharge analysis displayed clear increases in residual strain and changes in slope, consistent with irreversible structural damage and aligned with observed reductions in electrical performance. This work establishes a foundation for integrating mechanical strain diagnostics into future battery management systems.