Enhancing the structural stress assessment of distorted lightweight ship deck structures

Pursuing enhanced ship performance has driven lightweight structural solutions into modern cruise ship design. Among available strategies, the employment of thin steel plates in welded superstructure decks appears achievable, more sustainable and economically feasible. However, thin plates are susceptible to complex welding-induced distortions, which cannot be disregarded in the fatigue and limit state analysis of the welded structure. Since the effect of those distortions is not entirely considered by ship design rules, its evaluation requires full-field scanning of welded plates to be modelled in costly numerical analyses. This thesis investigates computationally efficient structural stress assessment approaches on buttwelded 4 mm-thick plates in stiffened panels from actual shipyard production, resulting in average to severe initial distortions according to classifications in the marine structures community. The distortion measurement and characterisation are followed by the 3D geometrically non-linear finite element analysis (GNL-FEA) of the panels under tension, simulating the effect of hull girder bending on the superstructure decks. The 3D model is validated against uni-axial tensile tests on the panels. Thereafter, a gradual scale reduction from 3D to 2D and 1D models is performed numerically and analytically, where the von Kármán kinematic assumption accounts for the geometric non-linearity. As a last step, a beam model is developed for a simple half-sine curvature and considering the effect of weld rigidity. In characterising the distortions, both amplitude and slope parameters need to be considered. For multi-buckled shapes with amplitudes below the plate thickness, a 2D analytical model neglecting the geometric discontinuity due to the weld can predict global structural stresses over the panel plate field; however, the weld cross-section must be considered in the local structural stress assessment of the welded area. For the latter, the 1D GNL-FEA of a distorted longitudinal profilelocated within 60% of the plate width results in less than 10% error. The 1D finite element model remains reliable if the distortion is included up to its first buckle from the weld location. Consequently, the analytical beam model can be adapted to the butt-welded area between thin plates in stiffened panels. Such an adaptation requires future research on the geometric parameters and boundary conditions in beam modelling.