Akademik Veri Yönetim Sistemi
3rd International Conference on Optimization and Data Science in Industrial Engineering, ODSIE 2025, Hybrid, Istanbul, Türkiye, 20 - 22 Kasım 2025, cilt.2854 CCIS, ss.577-592, (Tam Metin Bildiri)
The reliable estimation of bond strength degradation caused by corrosion is a major concern in reinforced concrete research, as it directly affects service life and structural safety. This study develops a machine learning framework to predict bond behavior using 254 experimental literature-based pull-out test data. Six regression algorithms—Random Forest, Extra Trees, XGBoost, AdaBoost, Decision Tree, and Ridge regression—are trained and validated through 5-fold cross-validation to ensure accuracy and robustness. Comparative results indicate that ensemble-based methods provide the highest predictive capability, with Extra Trees achieving a coefficient of determination of 0.923 and a root mean squared error of 2.43 MPa on the test set, followed by Random Forest with a coefficient of determination of = 0.917. Feature importance analysis identifies corrosion level and compressive strength as the dominant parameters, contributing 39.9% and 29.6% to the overall prediction, respectively. Concrete cover and bond length exert moderate influence, while steel diameter and type have marginal effects. The findings demonstrate that machine learning provides a reliable data-driven means of quantifying the relative importance of physical parameters and predicting corrosion-induced bond deterioration. This framework enhances structural performance assessment and supports service life prediction, maintenance planning, and rehabilitation design in reinforced concrete structures.