Progressive collapse response and ultimate strength evaluation of stiffened plates with welding residual stress under combined biaxial cyclic loads and lateral pressure

Stiffened plates used in bottom structures of ships are commonly subjected to multiaxial loading in harsh sea state. Thus, understanding their collapse behaviour under combined loads is highly valuable, and numerical simulations can provide accurate insights into this topic. The current research is the first to deal with the ultimate strength characteristics and assessment of initially deflected stiffened plates with welding residual stress (WRS) under combined biaxial cyclic loads and lateral pressure. Finite element analysis is conducted extensively to obtain the numerical results of longitudinal cyclic ultimate strength considering geometric and material nonlinearities with the aid of secondary development of ABAQUS software. It is found that the cyclic ultimate strength is governed by the coupling effects of structural dimensions, longitudinal and transverse residual stresses as well as transverse cyclic compressive/tensile loads and lateral pressure. A series of significant findings observed from this paper are presented in detail aiming to improve the ultimate limit state (ULS) design of ship structures. Based on sufficient sample points, four different machine learning (ML) models are trained maturely to predict the cyclic ultimate strength, including back propagation neural network (BPNN), support vector regression (SVR), random forest (RF) and radial basis function (RBF) network. Performance of these models is compared objectively using statistical metrics.