Neural networks for inelastic distortional buckling capacity assessment of steel I-beams

An Artificial Neural Network (ANN) model is developed as a reliable modeling method for simulating and predicting the ultimate moment capacities for intermediate doubly-symmetric steel l-beams. The training and testing data for neural network are generated using Finite Element Analysis (FEA). In other word, an extensive numerical study was also undertaken to investigate the distortional budding behavior of simply supported compact steel I-beams. A series of nonlinear elasto-plastic FE analyses have been carried out to simulate the distortional buckling behavior of doubly-symmetric steel I-beams, and the effects of six independent parameters as input in a lateral-distortional buckling mode has been investigated. Moreover, unlike the existing design codes the model considers the effect of web distortions in a lateral-distortional buckling mode. Then a new formula based on the ANNs has been proposed to predict the member moment capacities of steel I-beams subjected to distortional buckling. The attempt was done to evaluate a practical formula considering all parameters which may affect distortional capacity. Then, a comparison has been made between the proposed formula and the predictions from the current design rules in some structural steel codes. The results show that the proposed formula is more accurate and applicable than existing design codes. Finally, a sensitivity analysis using Garson's algorithm has been also developed to determine the importance of each input parameters. (C) 2015 Elsevier Ltd. All rights reserved.