FEM Analysis of Dropped Container in Offshore Platform Operations with Nonlinear Fitting Based on PLS-Based Neural Network

Offshore platform has long been in a complex and time-varying marine environment, would be inevitably influenced by environmental load and accidental loading, and result in the occurrence of various kinds of accidents or even disasters. The risk assessment for ocean platform has become an important issue in marine engineering field. However, the quantitative calculation of risks in the marine engineering industry is extremely difficult. The reason may attribute to the fact that there are few relevant statistical databases; and the mechanism of collision damage of marine structures is complex. The numerical simulation of impact caused by falling object is carried out by nonlinear finite element method (FEM). The PLS-RBF network is used for nonlinear interpolation purpose by combining the partial least square regression with the radial basis function (RBF) neural network. Therefore, the trained network can be implemented to obtain accurate collision results under various conditions by nonlinear interpolation. The feasibility and effectiveness of the proposed mapping approach for nonlinear finite analysis are testified by the simulation validation.