Bond Strength Between <scp>FRP</scp> Bars and Concrete Under Seawater Corrosion Environments: Data‐Driven Machine Learning Analysis

ABSTRACT This study explores the bond strength between fiber‐reinforced polymer (FRP) bars and concrete under seawater corrosion using data‐driven machine learning (ML). A database of 295 specimens was established, and six ML algorithms were employed to develop predictive models for bond strength. The Shapley additive explanations (SHAP) method was applied to assess input feature importance. Among the models, the extreme gradient boosting model demonstrated the highest prediction accuracy, yielding a coefficient of determination of 0.93, a root mean square error of 1.56 MPa, and a mean absolute error of 1.22 MPa, followed by artificial neural networks model. Support vector regression and k‐nearest neighbors models showed moderate performance, while linear regression and decision tree models performed the least accurately due to their simpler algorithm structures. SHAP analysis identified concrete compressive strength and FRP tensile strength as the most influential factors, followed by bar diameter and rib characteristics, with seawater corrosion time and temperature having minimal impact. These results highlight the robust bond behavior of FRP‐concrete interfaces in marine environments. Future work should expand the dataset to include more diverse bar dimensions, bond lengths, and corrosion durations to enhance model performance. The developed ML models not only enable accurate, low‐cost prediction of bond performance but also provide insights for selecting optimal FRP configurations and bond lengths in the design of corrosion‐resistant marine concrete structures.