Optimization and modeling of axial strength of concrete-filled double skin steel tubular columns using response surface and neural-network methods

Abstract The main objective of this study is to optimize and model the ultimate strength of axially loaded concrete-filled double skin steel tubular (CFDST) composite columns having a circular hollow section . By performing the multi-objective optimization analysis through the response surface methodology (RSM), an ultimate strength-based optimal design solution for CFDST composite columns was proposed. Besides, artificial neural networks (ANN) together with RSM techniques were used to develop design formulations that can be involved in the prediction of the ultimate axial strength of such types of composite columns. An experiment data repository compiled from the studies available in the literature was used in both optimization and modeling analysis. The aim of the optimization was to achieve the optimum values for the design parameters based on the maximum capacity. However, in the modeling, attaining a design formula having an accurate and reliable prediction performance was the purpose. Furthermore, a verification study in which the developed models were compared with the modified design formulas of the AISC and Eurocode 4 and some existing models proposed by the researchers was carried out. It was observed that the generated models gave a good estimation capability with a considerably high R-squared value and fewer errors. After statistically analyzing the outputs of the models, it was proved that the models proposed in the current study yielded more reliable, accurate, and consistent prediction performance than the existing formulas. Besides, it was indicated the optimization analysis has a desirability value of 0.942, meaning an almost perfect optimization.