As the demand for sustainable energy storage increases, second-life batteries (SLBs) offer a promising route to extend the useful lifetime of lithium-ion batteries (LiBs). Their large-scale reuse, however, requires reliable, efficient, and non-invasive methods for screening and diagnostics. This paper presents a hybrid screening and diagnostic framework based on thermal signatures for the assessment of LiB degradation. The framework is validated using a comprehensive dataset obtained from eight cylindrical 18650-type NMC cells tested under different cycling conditions. The results show that thermal features can serve as reliable indicators of aging and state of health. A robust feature-selection strategy combining correlation analysis and machine learning is developed to improve diagnostic performance. Among the evaluated models, the extreme gradient boosting (XGBoost) regressor achieved the best state-of-health (SOH) estimation performance, with residual errors predominantly within ± 0 . 3 % . In addition, a Random Forest-based classifier reached an overall accuracy of 98%, demonstrating the ability of the proposed framework to distinguish batteries according to their degradation background. The proposed approach provides a practical basis for future smart, cloud-enabled battery management systems and digital battery passports. By enabling automated SLB screening and diagnostics, it can support the industrial-scale reuse of LiBs and help unlock their circular value. • A hybrid thermal-based framework for second-life battery screening and diagnostics is proposed. • XGBoost achieves ±0.3% SOH prediction error using only temperature-derived features. • The model identifies degradation categories with 98% accuracy. • An extensive dataset from eight NMC cells supports model validation across the first and second life. • Approach is scalable, non-invasive, and compatible with smart battery management systems and battery passports integration.
Azizghalehsari et al. (2026) studied this question.