Castellated beams are a new type of structural member fabricated by cutting the web of an I-section or H-sectionmember along with the polygonal lines and then welding the upper and lower halves together. Compared to the traditional solid-web beams, castellated beams exhibit the advantages of a high strength-to-weight ratio, high in-plane stiffness, and economic efficiency. Due to the increased depth and the presence of openings, castellated beams are more susceptible to buckling. Studies have shown that the distribution of residual stresses across the section significantly influences their overall stability behavior. In this study, the residual stress distribution of two solid beams, four beams with low-height web plates, and four castellated beam specimens was measured using the sectioning method to investigate the influence of the cutting and welding processes on the residual stresses in the castellated beam sections. Test results indicate that the shape of the flange residual stress distribution remains largely unchanged after cutting and welding, but the tensile residual stress at the flange-web junction increases significantly, and it can achieve the yield strength after welding. The web cut region in the shallow and deep web specimens exhibit high levels of tensile residual stress. The residual stress distribution in the T-section of castellated beams is similar to that of the short and deep web specimens, while the web post section in castellated beams exhibits changes due to welding, with tensile residual stresses reaching yield strength at the welds. Based on the analysis of test results and existing models, a simplified model for longitudinal residual stress distribution was proposed for castellated beams, which could provide a reference for the stability analysis and design of castellated beams.

To investigate the distortional buckling performance of castellated beams, six castellated beam specimens were subjected to static load at the mid-span. The corresponding failure modes, load-vertical displacement responses, and web deformed shape curves of specimens were obtained. The results indicate that all six specimens failed in a distortional buckling mode. Specifically, the compressive flange of the entire specimens exhibited an S-shaped curve upon failure. Meanwhile, out-of-plane bulging deformation occurred in the web post section at the mid-span loading location. With the increase of castellated beam length, both the initial stiffness of the load-vertical displacement curve and the critical distortional buckling capacity reduced. Results of the distortional buckling strengths obtained from the experiments were further compared with the predications calculated using the modified method suggested by AS4100, AISC, and Nethercot and Trahair. The comparison results show that the predications of the AS4100 method and the Nethercot and Trahair approach are in good agreement with the test results.