A Hybrid AI and Machine Learning Framework for Predictive Modeling, Fault Diagnosis and Intelligent Decision-Making in Mechanical and Advanced Battery Energy Systems

A hybrid artificial intelligence and machine learning (AI–ML) framework is developed and evaluated for predictive modeling, fault diagnosis and intelligent decision-making for mechanical and advanced battery energy systems. A data-driven quantitative approach was adopted with the use of 1200 observations of mechanical operating parameters, battery performance indicators, severity of the fault and maintenance requirements. The proposed framework is a combination of Linear Regression, Random Forest, Gradient Boosting, Logistic Regression, Decision Tree, and Support Vector Machine algorithms for battery health prediction, remaining useful life estimation and multi-class fault classification. Descriptive statistical analyses, Pearson correlation analysis, assessment of feature importance, and predictive performance metrics were performed. The correlation between the operating hours and the battery state of health (0.912) and the battery capacity and the battery state of health (0.964) were high. The model of Linear Regression achieved the R² of 0.847 for battery health prediction and the model of Random Forest Regression achieved the R² of 0.931 for remaining useful life estimation. The best fault classification accuracy was obtained by Support Vector Machine with 62.8% while Gradient Boosting gave the best macro-F1-score 0.516. Maintenance decision agreement (reached) = 60.0%. The results show some potential to predict, but also some drawbacks for critical fault identification and decision reliability. Further validation of independent operational measurements is needed to establish practical applicability.