On the Development of Artificial Neural Network Models of the Critical Moment of Tapered Beams

Abstract

The design of steel structures using elements with non‐uniform geometry is common in practice, both for aesthetic reasons and material optimization. Tapered beams are one of such examples which are widely used. This work is focused on the development of artificial neural networks to predict the critical moment of tapered beams subject to bending moments applied at the extremities since no analytical methods are currently available. Almost 70k data samples were calculated using the Finite Element Method (FEM) with shell elements and were used to train and validate the neural network models. The definition of the data samples was based on a random generation procedure with constrained geometric proportions to reduce their number. The accuracy of the developed model is further illustrated on some particular cases and against the FE results of other authors. Finally, it is demonstrated that neural network models are fast and reliable assessment tools that can be employed during the search of optimal structural solutions in the early design stages.