Data‐Driven Modelling of Composite Recycled Aggregate Concrete‐Filled Steel Tube Columns

Abstract

Concrete‐filled steel tubes with recycled aggregate concrete are a sustainable composite system where the steel tube ensures strong confinement and improved column performance. Despite increasing experimental studies, reliable high‐fidelity numerical models for these members remain limited. This work presents a computational framework that integrates Bayesian optimisation with nonlinear finite element analysis in ABAQUS, fully automated through MATLAB–Python scripting. The procedure calibrates key parameters of the concrete stress–strain model to reproduce experimental axial response curves while accounting for confinement. The optimised parameters are further used to develop nonlinear regression equations and to support axial capacity prediction, including machine‐learning approaches. Validation against experimental results shows close agreement in terms of load–displacement response, and ultimate column strength. The model performance is additionally evaluated against two well established confined concrete models. The framework provides an efficient and transferable methodology for simulating composite column behaviour and demonstrates the potential of combining optimisation, numerical modelling, data‐driven and machine learning methods.