Application of XGBoost Model Optimized by Multi-Algorithm Ensemble in Predicting FRP-Concrete Interfacial Bond...

Area of Science:

• Civil Engineering
• Materials Science
• Computational Mechanics

Background:

• Accurate prediction of fiber-reinforced polymer (FRP)-concrete interfacial bond strength is vital for the safety and longevity of strengthened structures.
• Traditional empirical models often lack the necessary accuracy for complex structural designs.
• Developing robust predictive tools is essential for advancing the application of FRP in civil engineering.

Purpose of the Study:

• To develop a highly accurate and interpretable predictive model for FRP-concrete interfacial bond strength.
• To overcome the limitations of existing empirical approaches through advanced machine learning techniques.
• To identify the key parameters governing FRP-concrete bond strength.

Main Methods:

• Utilized the extreme gradient boosting (XGBoost) machine learning algorithm.
• Enhanced XGBoost model performance through global hyperparameter optimization using the Nevergrad framework with seven integrated optimizers.
• Employed a five-fold cross-validation strategy for robust model generalization.
• Validated the model using 855 single-lap shear test datasets.

Main Results:

• The optimized XGBoost model achieved superior prediction performance on the test set with R² = 0.9726, RMSE = 1.8745, and MAE = 1.3857.
• Demonstrated a 22.3% improvement in R² and significant reductions in RMSE (63.4%) and MAE (61.8%) compared to the best existing empirical model.
• SHAP interpretability analysis identified FRP width, thickness, elastic modulus, and bond length as critical factors influencing bond strength.

Conclusions:

• The developed XGBoost-Nevergrad model offers a significant advancement in predicting FRP-concrete interfacial bond strength.
• The model provides a reliable and interpretable tool for engineers, enhancing the design process for FRP-strengthened structures.
• This intelligent approach combines high predictive accuracy with valuable insights into material behavior and design parameters.