Scholarlink

Floating structures are required to mitigate the issues of overcoming the land shortage in Hong Kong and other densely populated coastal cities. For this purpose, conventional steel pontoons need to be replaced by fiber-reinforced polymer (FRP) from the standpoint of durability and economy. Among various issues of FRP and concrete in coastal areas, the bond performance between FRP bars and concrete is critical in structural design and application. This study compiled 219 real pull-out test records with seven input parameters to investigate the bond performance between FRP bars and concrete under seawater immersion. Based on the dataset, three machine learning algorithms, namely Decision Tree, Random Forest, and XGBoost, were employed to construct predictive models for the bond strength between FRP bars and concrete under seawater immersion. Evaluation metrics, including R2, MAE, and RMSE, were exploited to calculate the performance of the developed models. Due to the superior predictive performance of the Random Forest model, this study further employed SHAP analysis to investigate the effects of influencing parameters on the bond strength. The developed models and the SHAP analysis provided a reference for assessing and designing bond performance between FRP bars and concrete under seawater immersion conditions.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluation of Bond Strength Between FRP Bars and Concrete Under Seawater Immersion Using Explainable Machine Learning Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Heng Cao,
  • Xiao Lin Zhao,
  • Mudassir Iqbal,
  • Daxu Zhang,
  • Pei-Fu Zhang

摘要

Floating structures are required to mitigate the issues of overcoming the land shortage in Hong Kong and other densely populated coastal cities. For this purpose, conventional steel pontoons need to be replaced by fiber-reinforced polymer (FRP) from the standpoint of durability and economy. Among various issues of FRP and concrete in coastal areas, the bond performance between FRP bars and concrete is critical in structural design and application. This study compiled 219 real pull-out test records with seven input parameters to investigate the bond performance between FRP bars and concrete under seawater immersion. Based on the dataset, three machine learning algorithms, namely Decision Tree, Random Forest, and XGBoost, were employed to construct predictive models for the bond strength between FRP bars and concrete under seawater immersion. Evaluation metrics, including R2, MAE, and RMSE, were exploited to calculate the performance of the developed models. Due to the superior predictive performance of the Random Forest model, this study further employed SHAP analysis to investigate the effects of influencing parameters on the bond strength. The developed models and the SHAP analysis provided a reference for assessing and designing bond performance between FRP bars and concrete under seawater immersion conditions.