ann-based modeling of shear behavior of reinforced concrete columns under constant axial loads

Many older reinforced concrete (rc) buildings designed under outdated seismic codes exhibit inadequate shear capacity, leading to brittle column failures during earthquakes. accurate prediction of shear strength is therefore essential for nonlinear seismic assessment. this study develops an analytical–computational framework using artificial neural networks (anns) to model the nonlinear flexural–shear behavior of rc columns subjected to constant axial loads. a fiber-based flexural model was formulated, while shear strength was estimated through a mohr's circle–based approach enhanced with a ductility-dependent degradation parameter. an ann trained on 164 experimental column tests provided highly accurate shear predictions, outperforming existing analytical models. the framework was validatated against independent experiments confirmed its reliability. the proposed ann-based approach offers a practical tool for seismic performance evaluation and retrofit design of deficient rc columns.