Artificial neural network model to assist in design of ship stiffened plates considering ultimate strength

Abstract

The ship structure is exposed to combined loading conditions during operational exposure to waves due to the hull weight, cargo weight, and other loads. Because the primary structural members of the ship structure are stiffened plates, securing structural safety during the ship design stage is critical. However, in the current design method, the results are inaccurate and inefficient in terms of the required time or cost owing to the poor preparation of the physical model or assumptions regarding the boundary conditions. To resolve this, the present study performs a structural analysis on 3000 stiffened plates with included initial imperfections and subjected to axial compression, which causes the buckling loads on the ship. Deep learning database is constructed based on the finite element analysis results. Then, deep learning model is developed to propose an effective method for predicting the ultimate strength of the stiffened plates. The prediction results of the deep learning method showed similarity with the FE analysis results. During the initial design stage, the structural designer can use this model to determine the geometrical properties of the stiffened plate within a reasonable time and cost. In addition, these results can potentially be used in the reverse engineering of vessels.