Performance analysis of concrete-filled aluminum tubes confined with aramid fiber sheets under axial loading: a combined numerical and machine learning approach

Abstract This study presents a comprehensive investigation into the axial performance of concrete-filled aluminum tubular (CFAT) columns externally confined with aramid fiber-reinforced polymer (AFRP) sheets, using an integrated finite element analysis (FEA) and machine learning (ML) framework. While CFAT columns offer significant advantages such as reduced weight, high corrosion resistance, and architectural appeal, their structural performance is often limited by the lower stiffness and yield strength of aluminum. To overcome these limitations, this research explores the use of AFRP confinement to enhance load-bearing capacity and ductility. A validated FEA model was developed in ABAQUS based on 23 experimental CFAT stub column tests, showing strong agreement with results (where PEXP/PFEA, the ratio of experimental to FEA-predicted load capacities, ranged from 0.85 to 1.14, confirming model accuracy). A detailed parametric analysis investigated the effects of AFRP layer count, concrete strength, and tube geometry (D/t ratio, the diameter-to-thickness ratio indicating slenderness), revealing that AFRP confinement significantly improves performance—particularly in thin-walled columns. Additionally, four ML models (SVR, RF, ANN, Meta-ANN) were trained on 113 datasets generated from numerical simulations to predict ultimate axial load capacity. The Meta-ANN model achieved the highest accuracy with a MAPE of 2.02% and R of 0.99. To interpret the predictions, SHAP (SHapley Additive exPlanations) analysis was used, identifying column diameter and concrete strength as the most influential parameters. This dual numerical–data-driven approach demonstrates the potential of combining AFRP confinement with AI-based prediction tools for the design and optimization of advanced composite columns.