Due to economic reasons the industry is seeking new lightweight solutions for large steel structures. However, when moving from traditional steel plate thicknesses, i.e. 5 mm or larger, to thinner ones, the fatigue design becomes challenging due to larger initial distortions caused by welding. The fatigue assessment methods used for thicker welded structures are not fully validated for thinner ones. This paper deals with the fatigue assessment of large thin-walled structures starting from the global response analysis of a whole structure to the stiffened panel and finally welded joint. A modern cruise ship is used as an example case, where traditional superstructure deck plate thickness of 5 mm is replaced by 3 mm. The influence of initial distortion at different levels of structural analysis is studied using geometrically nonlinear finite element (FE) analysis. For the lowest level of analysis, i.e. small welded joint, the experiments have been carried out including geometry measurements and fatigue tests. It is shown that for a large thin-walled structure the global response analysis can be carried out with acceptable accuracy using ideally straight plates and geometrically linear FE analysis. For intermediate level of analysis, i.e. stiffened panel, the analysis can also be geometrically linear, but the actual shape of the plates influences the structural stresses near welds significantly. When analyzing small welded specimens to define experimental fatigue strength, both the actual shape and the geometrically nonlinear FE analysis are needed in order to capture the straightening effect and to obtain the correct structural stress.
