Finite Element Simulation of Axially Loaded Corroded Steel Tubes Considering the Intended and Unintended Imperfections

Steel tubes are commonly utilized in a variety of structural applications, including columns of steel buildings, wind turbines, and bridges. Corrosion is a serious type of damage that causes deterioration of these tubes. In order to numerically investigate the effect of corrosion on the axial behavior of steel tubes, an experimental program was first conducted to serve as a reference for comparison, which involved examining corroded and intact (non-corroded) steel tube specimens. The corrosion was simulated by reducing the thickness of the tube, t, by an amount equals to ∆t at a specific zone on the surface of the tube. Different levels of thickness reduction having ∆t/t=0.00,0.33,0.50 and 0.67 were adopted to simulate different levels of corrosion damage. The finite element (FE) simulation was then used to numerically investigate the behavior and facilitate future parametric studies. However, to conduct an FE simulation that accurately predicts the axial load behavior, two types of imperfections need to be considered. These are (a) the unintended initial imperfection to account for any initial imperfections found on the tubes before the introduction of corrosion damage, such as unleveled cutting of tubes, out-of-roundness, thickness variation, residual stresses, etc.; and (b) the intended imperfection made intentionally to simulate the corrosion damage. In this study, it is proposed to introduce an equivalent imperfection to account for the effect of the unintended imperfections. The shape and amplitude of the equivalent imperfection were selected by calibrating the load-carrying capacity of the FE model of the perfect steel tube with the load-carrying capacity of the intact lab specimens. The developed FE model, which accounts for the identified equivalent imperfection, became ready to introduce the intended imperfection that simulates the corrosion damage. Different levels of corrosion damage were then introduced to the FE models, which incorporated the equivalent imperfection. The developed FE models, after introducing the unintended and intended imperfections, were then analyzed. The obtained FE results were compared with the experimental results, verifying the accuracy of the proposed approach for FE simulation of this type of problem.