XGBoost algorithm based estimation of near surface mounted FRP rod-to-concrete bond strength and failure mode

Using fiber-reinforced polymer (FRP) in reinforced concrete (RC) structures can mitigate the colossal repair costs due to reinforcing steel corrosion. Hence, FRP rod/bar is gaining wider applications in diverse RC structures as a partial or full replacement for steel rebar. The FRP rod-to-concrete interfacial bond is pivotal in transferring stresses from concrete to FRP rods. This study develops a novel prediction model to estimate the near surface mounted FRP rod-to-concrete bond strength as well as the failure type using five machine learning (ML) algorithms, namely, linear regression, decision tree, gradient boosting tree, random forest, and extreme gradient boosting (XGB). The performance of the developed models was compared with that of four bond strength design guidelines and one analytical model. A database comprising 416 experimental datasets was constructed and used for model training and validation. Based on statistical performance metrics, the precision of the XGB algorithm was superior to that of the other ML models, design guidelines, and analytical model. Feature importance analysis based on the SHapley Additive exPlanations theory and partial dependence plot was performed. The results show that the bond length had the most significant influence on the bond strength, followed by the tensile strength of the FRP composite, diameter of the FRP rod, compressive strength of concrete, and elastic modulus of the FRP composite. A graphical user interface was developed and offers a user friendly, free access, and simple tool for estimating the bond strength and failure type of FRP rods in concrete.