Experimental and numerical investigation of corrosion effects on the seismic performance of shear links

• Conducted quasi-static tests on four corroded shear links. • Developed a FE model considering non-uniform corrosion effects. • Established overstrength prediction models based on multiple regression. Four corroded shear link specimens were prepared using accelerated corrosion techniques, and 3D laser scanning was employed to capture their surface morphology and corrosion depth distribution. Subsequently, quasi-static cyclic loading tests were conducted to investigate the effects of corrosion and web thickness on the failure modes, hysteretic behavior, energy dissipation capacity, and overstrength of shear links. Based on experimental findings, a finite element (FE) methodology was developed. This methodology stochastically assigns thickness reductions at nodal points on steel plates to accurately represent the surface morphology changes caused by corrosion. A parametric analysis of 245 corroded shear links was conducted to evaluate the influence of corrosion level, web height-to-thickness ratio, and flange-to-web ratio on overstrength. Finally, a predictive model for the overstrength factor of corroded shear links was established using nonlinear regression analysis. The results indicated that the corrosion depth distribution in steel plates approximately follows a log-normal distribution. Failures of all corroded shear links were primarily attributed to severe shear buckling-induced web tearing. As corrosion level increased and web thickness reduced, significant deterioration was observed in the bearing capacity, deformation capacity, energy dissipation, and overstrength of shear links. Experimental validation demonstrated that the developed FE methodology effectively captures the adverse mechanical effects of non-uniform corrosion, achieving high predictive accuracy. Furthermore, the prediction error of the proposed model for the overstrength of corroded shear links, based on the parametric analysis, was maintained within 10%.