Ultimate strength assessment of rectangular plates subjected to in-plane compression using a statistical model of welding initial deflection

Stiffened panels in steel ships and offshore structures are usually fabricated by fillet welding to attach stiffeners to a plating. Thermal contraction due to the welding causes angular distortion of the plate along the stiffeners, and the plate initially deflects. This welding initial deflection significantly affects the ultimate strength and collapse behavior of the stiffened panels, and it is one of the important uncertainty factors in ultimate strength assessment. This study proposes a methodology to quantify the uncertainty of the ultimate strength of rectangular plates due to the welding initial deflection. This paper has twofold. One is the development of a statistical model of the initial deflection shape based on measured data in past studies. Bayesian statistics is adopted since it can consider uncertainty accompanied by the insufficient number of measured data. A regression model of the initial deflection is selected as the best model from several candidate models based on an information criterion. The other is the assessment of probability distributions of the ultimate strength of simply supported rectangular plates. The probabilistic distributions of the ultimate strength are calculated by Monte Carlo simulation using a nonlinear finite element method. Initial deflection samples are generated from the developed statistical model and applied to finite element models of the rectangular plates. The analysis result indicates that the probability distribution of the ultimate strength depends on the collapse mode which is changed by the initial deflection shape. As a usage example of the proposed methodology, existing formulas of the ultimate strength are evaluated and adjusted to give safe side estimations for design. •A statistical model of welding initial deflection of rectangular plates, which is a Bayesian regression model constructed based on measured data in actual ships, is newly proposed.•Probability distribution of ultimate strength subjected to in-plane compression is assessed by using the proposed regression model and nonlinear finite element method.•Existing formulas of ultimate strength are evaluated and tuned by considering the obtained ultimate strength distribution.