Abstract
The mechanical behavior of corroded steel members is of interest due to the increased age of steel structures around the world. Especially, for the assessment of safety and development of the better maintenance plan, there is a compelling need for evaluating the residual strength. FEM analysis with solid elements is expected to evaluate the residual strength with high accuracy, however, it is computationally expensive. For this reason, FEM analysis with shell elements of which thickness is average thickness is often employed, but there is a problem that the result tends to be on danger side. This research proposes the method to evaluate the effective thickness by the convolutional neural network. The accuracy and effectiveness are verified by comparing FEM analysis with solid elements and shell elements of corroded steel plates and H-beam.
