Application of Eurocode 4 design provisions and development of new predictive models for eccentrically loaded CFST elliptical columns

The present study mainly aims to propose design models for the prediction of the peak strength of the concrete-filled steel tubular (CFST) elliptical columns subjected to eccentric compressive loading. At that point, the study has gain importance since the design of specifically such elliptical columns is not covered by current design codes. For example, the current provisions of Eurocode 4 provide a simplified method for the design of composite columns having circular or rectan-gular/square cross-sections. To this aim, first of all, the applicability and assessment of Eurocode 4 to the CFST elliptical columns under eccentric loading conditions were performed. Following, a data repository covering the experimental results of eccentrically loaded CFST elliptical columns (having different geometry, material property, load eccentricity in y and z axes, and longitudinal reinforcement ratio) was constituted. This data repository was used in the evaluation of the design guidance in Eurocode 4. Besides, the same experimental results were employed to develop two new numerical models based on gene expression programming (GEP) and artificial neural network (ANN) approaches. Finally, a sensitive statistical analysis was made to compare the prediction performances of Eurocode 4, GEP, and ANN models against experimental results. It was observed that Eurocode 4 design provisions had sufficiently accurate prediction performance, however, the developed GEP and ANN design models yielded consistent and robust prediction performance than the current design provisions by providing a much smaller mean absolute percent error of 7.5 and 1.4, respectively.