Artificial neural network modeling of the stability behavior of stainless steel I-beams with sinusoidal web openings

Perforated steel beams are advantageous compared to plain-webbed beams. However, steel beams with sequential web openings are more prone to instabilities, requiring special design consideration to prevent potential failure modes. This study investigates the stability behavior of perforated beams with sinusoidal openings made from different grades of stainless steel using numerical simulations in ABAQUS software and uses Artificial Neural Network modeling to propose a data-driven design approach for these members. The study provides insight into the global stability behavior of these elements by developing 9720 finite element models under different types of loads. The results indicate that current standards do not accurately represent the behavior of members that exhibit lateral-distortional buckling and interactions between local and global failure modes, which is a significant design concern. Additionally, in the work, it was considered the simulation of a newly developed stainless steel grade, the S600E high-strength stainless steel. High-strength members are more susceptible to interaction-governed failure modes than conventional-yield-strength members. Finally, some design codes may fail to correctly represent the behavior of members loaded outside the shear center due to the destabilizing effect of loading on these structures. The neural network model developed is highly effective in predicting the behavior of the studied structures, considering the data interval that was treated in this study. A computer program was developed to enable the application of the trained model.