Data-driven models for prediction of peak strength of R-CFST circular columns subjected to axial loading

Abstract

Reinforced concrete-filled steel tubular (R-CFST) columns are a relatively new member of the CFST family. They have been developed with the purpose of reducing the local buckling of steel tubes and enhancing both the load-carrying capacity and ductility of conventional CFST by configuring reinforcing bars in the concrete core. The study herein aims to develop consistent, functional, and robust analytical design models to predict the peak strength of the circular R-CFST columns. To this, an extensive database has been created by gathering up the 115 experimentally tested axially loaded R-CFST circular columns from 15 different studies existing in the literature. In this context, two famous and commonly used soft-computing methods known as artificial neural network (ANN) and gene expression programming (GEP) were employed to derive an analytical design model by means of the compiled experimental-based dataset. The analytical design models developed in this study were proposed based on their prediction performance that was determined by verifying and validating them with a testing dataset, statistically analyzing and comparing them with the existing design codes (Eurocode 4, AISC, ACI, and GB) and formulations proposed by some researchers, and performing sensitivity analysis. The results indicated that the developed ANN and GEP-based analytical design models have a superior prediction performance and considerably lower mean absolute error occurrence of 3.6 and 6.8%, respectively in comparison to the existing models.