Abstract. The nonlinear analysis became a common tool to precisely assess the load-bearing behavior of steel beam and column members. The failure level is significantly influenced by different types of imperfections, among geometric also structural imperfections (residual stresses). Here are still gaps in the knowledge. Nowadays, 3-D welding simulation developed to a level where it could provide reliable estimation of weld-induced distortion and residual stresses. Nevertheless, modelling and computational effort are still in a less practicable range. In this study a simplified procedure to implement residual welding stresses in continuous large scale members is proposed and the influence on the ultimate limit state of slender members in compression is evaluated for two common structural steel grades. The results showed significant improvements in the utilization of load bearing capacity compared with simplified design methods. The comparatively general approach in this study offers potential for future optimization.
IntroductionWelded I-girders are used in different applications in steel construction (e.g. industrial buildings, bridges) due to either dimensions and/or efficiency through customizable plate thicknesses, shapes and/or materials. Many standards, including Eurocode 3 (EC 3) permit the use of non-linear finite element analysis (FEA) for the design of steel structures. The development in this field allows performing "experiments" in computing software instead of the laboratory or expensive in-situ experiments. Still, the implementation of imperfections (including residual stresses due to weld manufacturing) remains a major task in the performance of such analysis. Unlike geometrical imperfections, residual stresses (RS) are not standardized. For that reason, more or less founded simplified distribution functions as proposed by the Swedish design code BSK 99 [1] are used. Limitations on the applicability of this model are not reported except for the plate thicknesses which should not exceed 40 mm. However, the influence of RS seemed somewhat overestimated for many cases comparing conventional structural steel S355 and S690QL. As in the EC 3, direct proportionality of the tensile and compressive RS (the latter are the main interest in this study) and the yield strength is assumed. An opposite effect was recently noticed in [2,3] showing relatively reduced RS in high strength steel. A significant overestimation of RS was generally noticed if the chords are narrow. A lower limit is missing. Additionally, the correlation with the plate thicknesses only is doubted since no record is taken of actual welding parameters. Hence, a new approach was required. This contribution (as part of a national research project [2]) presents an improved general