Numerical and Analytical Evaluation of Elastic Lateral-Torsional Buckling in Cellular Steel Beams

Steel cellular beams, characterized by the presence of circular openings along the web, are produced by periodically cutting the web of a hot-rolled I-section and subsequently displacing and welding the resulting parts to form profiles with increased height and sequential openings. This increase in the effective height of the profile, combined with the openings, makes this type of beam more susceptible to new failure modes and intensifies the collapse mechanisms already present. This study analyzes steel cellular beams through numerical analysis using Abaqus 2021 software, with an emphasis on the collapse mode known as lateral-torsional buckling (LTB), which occurs when a beam subjected to bending moments experiences lateral displacement combined with rotation about its longitudinal axis. The research also discusses the numerical methodologies used to assess this phenomenon, referencing the Brazilian standard ABNT NBR 8800:2024, which applies to solid-web beams, and the European standard EN 1993-1-1:2005, which addresses the specific characteristics of cellular beams. It is noteworthy that, after calibrating the cellular models, the elastic critical moment results obtained both by the analytical method of the Brazilian standard and through numerical modeling using Abaqus showed good agreement, demonstrating the reliability of the methods employed. The results highlight that cellular beams can support loads equivalent to or greater than those of solid-web beams, thus enabling structural optimization and cost savings in steel beam design. It was also observed that geometric and mechanical parameters influence the LTB resistance of the beams, with flange dimensions having a greater impact due to their significant relationship with bending moment failure. After calibrating the cellular models, the elastic critical moment results obtained through both the analytical method of the Brazilian standard and the numerical modeling in Abaqus showed good agreement, confirming the reliability of the applied methods.