PLS-Based RBF Network Interpolation for Nonlinear FEM Analysis of Dropped Drum in Offshore Platform Operations

Abstract

Offshore drilling platform plays an important role in the exploitation of offshore natural resources. Safety is the top priority in offshore platform operations. Among various risks, dropping objects is a major source of risk that threatens personal safety, platform structure and environment safety. In this paper, simulations are performed using finite element simulation software. As the custom objects in platform crane operations, the oil drum is used as the research object in the simulation of damage caused by dropping objects on the drilling platform deck structure at different contact angles. Through the analysis of the simulation test results, the relationship between the angle of the dropping object and the energy impact of the deck is obtained. As the result of the impact and the contact angle is a highly nonlinear mapping, the radial basis function neural network based on partial least squares is implemented for interpolation purposes. The approach of PLS-RBF (Partial Least Square-Radial Basis Function) method takes advantage of the RBF network and PLS regression method can obtain high generalization accuracy for nonlinear system mappings. The results are compared with other approaches to illustrate its effectiveness.