Intelligent prediction framework for axial compressive capacity of FRP-RACFST columns

Fiber reinforced polymer confined-recycled aggregate concrete filled steel tube (FRP-RACFST) not only proposes excellent load-bearing capacity and corrosion resistance, but is also an environmentally friendly structure due to the incorporation of recycled aggregate concrete (RAC). This paper proposes a machine learning-based framework for predicting the axial compression capacity of FRP-RACFST. Firstly, a database containing 217 samples was established, and the criterion of input feature selection was verified using the variance inflation factor (VIF) method. Based on this, the applicability of six standard design methods was evaluated. Then, eight machine learning (ML) models were trained, with hyperparameters optimized using five fold cross-validation to ensure model generalizability. The results indicated that the machine learning models outperformed existing computational methods, with the extreme gradient boosting (XGBoost) achieving the highest accuracy, R2 values of 0.994 and 0.980 on the test and training sets, respectively. Additionally, a global and local interpretability analysis based on the SHapley Additive exPlanation (SHAP) method was used for the XGBoost model, revealing the contributions and potential impact mechanisms of eight design parameters on the axial compression capacity of FRP-RACFST. Finally, a user-friendly graphical user interface (GUI) was developed for the rapid assessment of the axial compression load-bearing capacity of FRP-RACFST.