ABSTRACT

The capability of an artificial neural network-based model in prediction of shear strength of trapezoidal corrugated steel webs (TCSWs) with pitting corrosion was investigated in this study. H-shape steel girders with TCSWs were design and elastic shear buckling strength and ultimate shear capacity were analyzed. Finite element models of H-shape steel girders with TCSWs were developed by using ANSYS, different pitting corrosion parameters, including the corrosion zone height, pitting depth and the pitting diameter, were taken into account and effect of corrosion parameters on shear strength of TCSWs were obtained. Considering the excellent machine learning capabilities of artificial neural networks (ANN), prediction system of shear strength of TCSWs with pitting corrosion based on a 3-lever ANN was proposed. The shear strength of TCSWs database based on the results of 200 finite element (FE) models was used to train and test the ANN. The nonlinear mapping relationship between the affecting variables and the shear strength were established. the ANN trained by these FE results were found to provide close predictions of the shear strength under different pitting corrosion parameters. it can be concluded that properly trained and well calibrated neural networks can be reliable for predicting shear strength of TCSWs with pitting corrosion.