Hybrid FE-XGBoost framework for strength prediction of CFRP-CFAAT columns

Concrete-filled aluminum alloy tubular (CFAAT) columns offer superior corrosion resistance but often suffer from limited load-bearing capacity due to the low elastic modulus of aluminum. While confining CFAAT with Carbon Fiber Reinforced Polymer (CFRP) enhances performance, existing empirical studies are limited by small sample sizes and lack unified design standards. To bridge this gap, this study presents a comprehensive framework combining validated finite element (FE) simulation with a data-driven machine learning approach. First, reliable FE models were validated against experimental results to conduct systematic parametric analysis on five key design factors. Subsequently, a substantial database of 226 specimens was compiled to develop a Bayesian-optimized XGBoost model for strength prediction. Results demonstrate that the proposed model achieves unprecedented accuracy (R²=0.986, MAPE=4.99 %), significantly outperforming existing design codes (EC4, AISC) and empirical equations. Furthermore, SHAP-based interpretability analysis was employed to reveal the underlying strengthening mechanisms, identifying the concrete strength and tube diameter as critical features. This study not only elucidates the synergistic effect of CFRP and aluminum but also provides a robust and intelligent tool for the precision design of CFRP-CFAAT members in marine infrastructure.