Prediction of the Bending Collapse of Thin-Walled Rectangular Tubes

A method based on Artificial Neural Networks is presented to estimate the load-displacement behavior of thin-walled rectangular tubes under pure bending. Besides deep plastic collapse the maximum bending moment (limit load) is important for many tubular structures in industry. The accuracy of the new idea has been proven to be better compared to an existing analytical method. An implementation on finite element models using beam elements is presented as well to show the superior computational efficiency of this approach. The method thus is especially useful in early design stages were many design alternatives and expression should be investigated to find a “best of” initial structure, or for numerical optimization or for validation of more detailed models, like shell or solid element dominant simulation models.