Structural behavior of steel-concrete partially encased composite columns containing demolished concrete lumps under axial compression

•The viability of using DCLs in partially encased composite columns is examined.•PEC columns’ stiffness and ultimate strength are less influenced by the incorporation of DCLs.•Post-peak behavior is negatively affected by DCLs, thus closer link spacing is needed.•A uniaxial constitutive model is proposed for the concrete core of PEC columns. Steel-concrete composite members incorporating demolished concrete lumps (DCLs) have been devised to provide an innovative alternative means of recycling old concrete. Field applications have demonstrated those members’ potential. In this study the aforementioned recycling method is applied to steel-concrete partially encased composite (PEC) columns. To evaluate the viability of this new solution, a comparative experimental campaign was implemented to investigate the behavior of eleven large-scale PEC columns containing different content of DCLs under axial compression. The key parameters considered were replacement ratio of DCLs, steel flange thickness, spacing of transverse links, the source of DCLs. Overall, the initial stiffness and axial load-carrying capacity of the PEC columns containing up to 33% DCLs were found to be comparable to those of the columns cast with new concrete alone. Nevertheless, a more rapid load drop right after the peak was observed for the columns containing DCLs when the link spacing equaled or exceeded half the column section depth. To remedy this, a maximum link spacing limit of 0.3d (d = the column section depth) is recommended for the columns containing DCLs. An analytical model is then developed to reproduce the full-range axial response of the PEC columns. The model accounts for the gain in strength due to concrete encasement and the loss due to flange buckling. It also attempts to capture the columns’ descending branch with recourse to the concept of compression fracture energy. Applicable to both types of PEC columns (i.e. containing DCLs or not), the proposed model exhibits a good accuracy, evidenced by comparing the prediction results against a set of test data compiled in this study.