Abstract

Steel reinforcement is essential to reinforced concrete (RC) structures, but chloride-induced corrosion can cause cracking,strength loss, and premature deterioration. This study investigates a sustainable glass–plastic powder coating for reinforcing steeland develops a Hybrid Ensemble-XDI predictive framework that combines extreme Gradient Boosting (XG Boost), Decision Tree(DT), and Isolation Forest (IF) through weighted averaging. The methodology uses data preprocessing, anomaly detection, andSpearman correlation-based feature selection to identify the most influential predictors of compressive-strength loss. Theexperimental program considers M43 concrete beams with uncoated, glass-coated, plastic-coated, and combined glass–plasticcoated reinforcement, cured for 7, 14, and 28 days and exposed to 10% NaCl solution for 100 h. The proposed model achieved98.55% accuracy, 98.28% precision, 98.30% recall, 98.47% specificity, and 98.28% F1-score. The results indicate that glass powdercontributes densification and plastic powder contributes hydrophobic behavior, while the hybrid model provides a robust data-driventool for durability assessment.Index Terms—steel reinforcement, chloride corrosion, glass powder, plastic powder, XG Boost, decision tree, isolation forest, hybridensemble, durability prediction