Answers are grounded in this paper and cite their sources.
Predictive maintenance has become an essential strategy for improving the reliability and safety of critical systems, particularly those dependent on lithium-ion batteries in applications such as electric vehicles, aerospace, and renewable energy storage. This study presents a data-driven framework for predicting the end-of-life (EOL) of lithium-ion batteries by analyzing both charging and discharging cycles from the NASA battery dataset. Engineered features were extracted using quartile segmentation and min–max normalization to capture key patterns in voltage, current, and temperature. Multiple classification algorithms, including XGBoost, Logistic Regression, Support Vector Classifier, K-Nearest Neighbors, and Gaussian Naïve Bayes, were evaluated based on accuracy, precision, recall, and F1-score. XGBoost demonstrated superior performance with 99% accuracy for charging and 97.3% for discharging cycles. To improve interpretability, explainable AI techniques were applied, namely feature importance and SHAP, to uncover the most influential predictors of battery degradation. The analysis revealed that the instability of the early cycle voltage was a key driver in the prediction of EOL based on charging, while the variability of the temperature was most significant during discharge. These insights support the development of transparent and reliable predictive maintenance systems for safety-critical battery applications. • A framework to predict lithium-ion battery EOL using NASA cycle data. • Model for feature extraction using quartile segmentation and min–max normalization for key signals. • XGBoost achieved 99% accuracy for charging and 97.3% for discharging predictions. • Explainable AI employed to identify critical predictors, enhancing transparency.
No takes yet. Share an insight, caveat, or question.
Alqaryuti et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: