Objective To accurately assess the in-service safety of ship structures, it is essential to clarify how welding-induced residual stress relaxation and cumulative plastic damage affect their load-carrying capacity and to develop a rapid method for predicting residual ultimate strength.
Method Finite element models of stiffened plates incorporating initial imperfections are developed using the finite element software ABAQUS. The ultimate strength and failure mechanisms of stiffened plates subjected to cyclic loading are investigated. Based on these analyses, a physics-informed neural network model incorporating the evolution laws of load-carrying capacity is established.
Results Under low-amplitude cyclic loading, the relaxation of welding-induced residual stresses can increase the ultimate strength by up to approximately 9.6%. In contrast, under high-amplitude cyclic loading, the progressive accumulation of plastic damage can reduce the ultimate strength by more than 73%. Strain amplitude and the number of loading cycles are the most influential factors contributing to ultimate strength degradation, accounting for 26.1% and 17.9% of the degradation, respectively. Approximately 93% of the samples in the database have absolute relative prediction errors below 5%. Furthermore, the model achieves a coefficient of determination (R2) of 0.984 on the independent test dataset, demonstrating its high predictive accuracy and good generalization capability.
Conclusion Residual stress relaxation and cumulative plastic damage significantly affect the ultimate strength of stiffened plates subjected to cyclic loading. The proposed prediction model provides a theoretical basis for assessing the strength of ship structures.