PLS-RBF Neural Network for Nonlinear FEM Analysis of Dropped Container in Offshore Platform Operations

Abstract

Accidents caused by accidental load not only lead to casualties and major economic losses, but also cause serious pollution and damage to the surrounding environment and marine ecology. The study of its safety of offshore platform in complex and extreme environments is a research spotlight in ocean engineering area. In addition to the normal work load and environmental load, the ocean platform is also threatened by accidental load, such as the collision of the ship and the impact of the upper part of the platform. It is necessary to evaluate the risk that the platform encountered. However, there is little statistics on the risk of offshore platforms in the industry of marine engineering, which has caused difficulties in quantitative calculation of risk. In addition, the mechanical mechanism of the marine structure injury caused by the collision itself is complex and it is not feasible to conduct large number of experimental simulations, so the numerical simulation using nonlinear finite element method (FEM) is implemented in this study. To achieve accurate impact result under arbitrary situations, nonlinear interpolation is made by means of PLS-RBF network. Simulation results demonstrate the feasibility and effectiveness of the nonlinear PLS-RBF mapping for nonlinear finite analysis.