Discussion: Evaluating the effects of loading protocol on the strength and deformation capacity of Flexure-Shear critical concrete columns

The increasing reliance on ATENA software-based finite-element simulations for modelling columns necessitates transparency in parameter selection and calibration. While the study by Dolati et al. (2023) provides an extensive investigation of flexure-shear critical columns, some aspects of the modelling strategy and data reporting require clarification to ensure the validity of the conclusions.

The following points about the discussed paper necessitate clarification.

• Modelling parameter and calibration. The onset of axial strength degradation was assumed to occur at an axial shortening of 1% of the column clear height. However, this threshold was not reached in the experimental programmes used for model calibration (LeBorgne, 2012; Lynn et al., 1996; Sezen and Moehle, 2006; Sokoli and Ghannoum, 2016). This criterion raises concerns about the validity of the ATENA material calibration. A particularly critical issue concerns the selection of the concrete softening parameter (w_d) of 50 mm, which is notably higher than the range recommended by the ATENA software developers (0.5–6 mm) (Cervenka et al., 2014). Such a choice delays damage localisation, thereby artificially increasing the concrete resistance, particularly in columns with low confinement levels (Cervenka et al., 2014). In addition, the volumetric parameter (⁠$β$⁠) was fixed at 0.2 for all the specimens, despite recommendations to increase $β$ for higher confinement levels (Cervenka et al., 2014). Instead, for modelling the most heavily confined column (CS60), the unloading factor $fU$ was increased from 0.3 to 0.6 and attributed to ‘relatively high levels of confinement’ (Dolati et al., 2023: pp. 9, 17). However, the primary reason for increasing $fU$ in modelling CS60 is its highest stiffness, as $β$ governs the confinement level. Finally, no reduction in compressive strength of concrete due to cracking (⁠$fcr$⁠) was applied, despite a default value of 0.8 and the recommended range of 0.45 to 1 (Cervenka et al., 2000). This choice may overestimate the residual axial capacity, especially in modelling columns under cyclic loading, therefore it requires justification.

• Interpretation of simulation results. The simulation results complicate the interpretation of some of the stated conclusions. For instance, it is concluded that the columns experience larger drifts under fewer loading cycles; however, the maximum drift ratio of column CS60 is approximately 9% for two cycles of loading (figure 17 in the paper by Dolati et al. (2023)), while it is only 5% for one cycle of loading (the left-hand side of figure 24 in the discussed paper). Furthermore, the residual lateral strength at the axial degradation for specimens with the lower confinement ratios was higher than for the highly confined column CS60 (table 3 of Dolati et al. (2023)), which is not aligned with the stated conclusion. Further explanation is required to ensure consistency between the results and conclusions, as these variations may stem from simplifications introduced in the numerical modelling. Notably, the transverse and longitudinal reinforcement ratios for specimen CS60 were simplified to 0.0124 and 0.045, respectively, deviating from the original experimental values of 1.5% and 4.7% (Sokoli and Ghannoum, 2016). Such simplifications undermine the validity of direct comparisons between the numerical and experimental results. Additionally, the applied axial load ratios on specimens CS60 and 3CMD12 were reported as 0.3 and 0.26, respectively, differing from the original values of 0.27 (Sokoli and Ghannoum, 2016) and 0.35 (Lynn et al., 1996), respectively. Given the strong influence of axial load ratios on shear failure and drift capacity, these variances require justification.

• Data consistency. The data of some of the simulated columns were reported differently within the discussed article. For example, specimen 4, tested monotonically, has a maximum experimental drift of around 6% in figure 3 of Dolati et al. (2023) and both experimental and simulated drifts of approximately 6% in figure 16 (left) of Dolati et al. (2023), whereas table 9 of Dolati et al. (2023) lists a simulated drift of 11.6%. These discrepancies hinder the interpretation and require clarification.

Clarification addressing the reported inconsistencies in modelling assumptions, parameter calibration, result interpretation and data reporting is essential to justify the conclusions. Addressing these issues would enhance the transparency and reliability of ATENA-based simulations for flexure-shear critical reinforced concrete columns under seismic loading.