Optimal design of circular concrete-filled steel tubular columns based on a combination of artificial neural network, balancing composite motion algorithm and a large experimental database.

This article proposes a novel design methodology for circular concrete-filled steel tubular (CFST) columns, combining a large experimental database, an artificial neural network, and the balancing-composite-motion-optimization algorithm. First, the experimental database consisting of 1245 compression tests was compiled from the available literature, involving five input parameters: cross-sectional outer diameter, thickness of steel tube, length of column, yield strength of steel and cylindrical compressive strength of concrete, and one output variable: the measured axial capacity of columns. Second, the artificial neural network model was trained and validated using that database, in combination with various error measurement criteria: mean-squared-error, mean-absolute-error, and coefficient of determination. The optimization problem to find the optimal design of circular CFST columns was then formulated as a function of material cost and the geometrical parameters of the cross-section. Finally, balancing-composite-motion-optimization was employed to solve the aforementioned nonlinear optimization problem. The proposed model's performance was compared with the widely used standards EC4, AISC, and DBJ, especially for scenarios exceeding the limitations of individual material strengths. Results showed that the proposed equation exhibited the best predictive performance. For practical application, we have derived an explicit prediction equation and implemented it in Excel, a user-friendly interface (the Excel file is appended to this paper). Thus, the proposed method in this study can assist existing codes in finding the optimal design of CFST columns, as the current standards have limitations as to the characteristic strength of the materials. [ABSTRACT FROM AUTHOR]