Key points are not available for this paper at this time.
The bond strength and failure mode of fibre-reinforced polymer (FRP) bars embedded in ultra-high-performance concrete (UHPC) and ultra-high-performance seawater sea sand concrete (UHPSSC) are significant to ensure the structural integrity of FRP-UHPC/UHPSSC structures. Given the multiplicity influencing bond interaction, conventional methods face challenges in developing predictive models and equations. For precise prediction and comprehensive investigation of bond performance in FRP-UHPC/UHPSSC, this paper employed an effective approach utilising extreme gradient boosting (XGBoost) technique to predict bond strength and failure mode. Two XGBoost predictive models were established based on a dataset comprising 542 data points with 14 input parameters. The developed models demonstrated reliable performances through an exhaustive assessment. SHAP (SHapley Additive exPlanations) analysis was employed to study the influences of input variables considering the bond mechanism. Specifically, the bonded-length-to-diameter ratio is pivotal in bond strength prediction, followed by FRP surface properties, concrete compressive strength and cover-to-diameter ratio; while cover-to-diameter ratio, bar diameter and concrete compressive strength stand out in failure mode prediction. Furthermore, explicit SHAP-based bond strength predictive equations were derived for concrete splitting and pullout failures. The developed model and equation demonstrate a closer approximation to experimental results compared to that in ACI 440.1R-15.
Building similarity graph...
Analyzing shared references across papers
Loading...
Pei-Fu Zhang
Shanghai Jiao Tong University
Xiao‐Ling Zhao
Hong Kong Polytechnic University
Daxu Zhang
University of Electronic Science and Technology of China
Composite Structures
Shanghai Jiao Tong University
Hong Kong Polytechnic University
Shanghai Ocean University
Building similarity graph...
Analyzing shared references across papers
Loading...
Zhang et al. (Wed,) studied this question.
synapsesocial.com/papers/69dd4f2a0a7b4bc8c41015f8 — DOI: https://doi.org/10.1016/j.compstruct.2024.118437