Parametric Study of Eccentrically Loaded Concrete Encased Steel Composite Columns Using Artificial Neural Networks

Abstract

This paper presents a parametric study to investigate the behavior of eccentrically loaded concrete encased steel composite columns (SRC). The artificial neural network (ANN) technique was adopted in this study by developing an efficient model to predict the behavior of such composite columns, depending on a total of 105 experimental tests for such composite columns with concrete rectangular section encased I-shape structural steel section and subjected to eccentric loads producing bending moment about one of the column section axes. The developed model was used to investigate the effects on the structural behavior of the eccentrically loaded composite columns owing to the steel contribution ratio, the axis of the applied bending, the concrete strength, and the structural steel yield stress by analyzing of 36 SRC specimens with different structural properties. Generally, it is shown that the effect of the axis of applied bending moment on the strength of SRC specimens is directly proportional to steel contribution ratio. It was observed, also, that in spite of the strength of the analyzed composite columns were increased with the increase in the strength of concrete, but the both effects, the axis of the applied bending moment and the increase of structural steel yield stress, are inversely proportional to the increase of concrete strength. The Predicted strengths of SRC specimens from ANN analysis were compared with that calculated using the EC4, giving good agreement reached to a ratio around 0.96.