Evaluating structural response of concrete-filled steel tubular columns through machine learning

Abstract Concrete-filled steel tubular (CFST) columns are unique structural members that capitalize on the synergy between steel and concrete materials. Due to complexities arising from the interaction between steel tube and concrete filling, the analysis and design of CFST columns are both intricate and tedious. A closer examination to the provisions of American, European and Australian/New Zealand design guidelines shows how these building codes seem to diverge on a proper methodology to design CFST columns. This leverages naturally inspired machine learning (NIML) algorithms (namely genetic algorithms and gene expression programing) to derive compact and one-stepped predictive expressions that can accurately predict the structural response of CFST columns. These expressions were developed and validated using the results of 3,103 available tests carried out on CFST columns over the past few years. The outcome of this work shows that the NIML-derived expressions have superior prediction capabilities than those in currently used design codes.