Statistical Analysis of Cathode Chemistry and State-of-Charge Effects on Thermal Runaway Behavior of Type 21700 Lithium-Ion Cells

Thermal runaway is a critical safety concern in lithium-ion batteries, particularly because cathode chemistry and state of charge directly influence the severity of thermal abuse behavior. The present study performs a multivariate statistical analysis of thermal abuse behavior for commercial 21700-format cylindrical cells for five state-of-charge (SOC) levels (0%, 30%, 60%, 80%, 100%) and cathode chemistries (LFP, NMC, NCA-HEI, NCA-HEII, NCA-HP) using accelerating rate calorimetry data. The extracted thermal metrics are maximum temperature, thermal runaway onset temperature, self-heating onset temperature, and maximum heating rate. Two-way ANOVA with partial eta-squared (η_p^2) effect sizes assert significant main effects of SOC and chemistry, alongside a dominant chemistry and SOC interaction for thermal runaway onset temperature (η_p^2 = 0.99, p < 0.001). Non-parametric Kruskal–Wallis tests confirm the parametric findings, indicating robustness to distributional assumptions. A novel data-driven Thermal Risk Index (TRI) is developed employing principal component analysis-derived weights, with the first two principal components together explaining 94.1% of the total variance. The TRI shows a moderate monotonic dependence on SOC (Spearman's ρ = 0.480, p < 0.001), supporting its physical interpretation. Under fully charged conditions, NCA-HEI cells show the highest mean TRI (78.6), markedly exceeding LFP (33.4), a 2.4-fold difference between the most and least hazardous chemistries. These results quantify the chemistry-dependent safety hierarchy under operational extremes and propose an interpretable framework for comparative thermal risk assessment.