ABSTRACT
This paper discusses the prediction of the bearing capacity of welded columns using artificial neural network. The built-up network is based on a finite amount of buckling load analyses carried out using a simplified welding simulation approach to take into account directly the welding influence. The recent research showed that this approach combining simplified welding simulations and nonlinear buckling analysis can achieve a good balance between increased accuracy of the results and computational efficiency. Nevertheless, a large-scale parameter study of welded column remains challenging, which is due to the large number of parameters in the welding simulation and the correlation between the different parameters. In this study, a set of numerical results data will be analysed and used to train the optimized artificial neural network with the genetic algorithm and then the predicted results will be tested and verified. Finally, the potential and advantages of this new approach are discussed based on the error analysis.
