Estimation of axial load-carrying capacity of elliptical concrete filled steel tubular columns using computational intelligence

Elliptical Concrete Filled Steel Tubular (ECFST) members exhibit superior aesthetics and improved structural performance in certain applications. However, accurately estimating their axial load-carrying capacity remains a challenge due to the complex interaction between geometric parameters and material strengths. This study presents a computational intelligence-based approach to predict the axial load-carrying capacity of ECFST members using hybrid artificial neural networks (ANN) and metaheuristic optimization techniques. A comprehensive dataset comprising 500 structural performance literature data has been initially collated. The dataset incorporates key geometry and material parameters, covering a wide range of material strengths and cross-sectional and member geometrical dimensions. A total of 1,555,200 different ANN architectures were trained using global optimization algorithms to deduce the optimum condition. The performance of the optimum model is also compared to current design standards, including European, American and Chinese codes. Assessment of the key parameters influencing the axial load-carrying capacity is also done using SHapley Additive exPlanations (SHAP). For practical application, a matrix form explicit equation and a Graphical User Interface are also derived based on the optimum prediction model and provided as supplementary material to the interested researchers.