Environmental parameter design for jack-up platforms based on FEM-BPNN-JPA coupled approach

Abstract

This study proposes a novel coupled approach, integrating the non-linear finite element model (FEM), BP neural network (BPNN), and joint probability analysis (JPA), to conduct the ocean environmental parameter design for jack-up platforms. Two bottom constraint methods are involved in the FEM, namely the contact surface model and the hinge constraint model, to conduct the structural response analysis. Subsequently, based on the calculated structural responses from FEM, the BPNN is employed to predict the jack-up response under varying combinations of wind speed, wave height and current velocity. This FEM-BPNN coupled method proves to be an effective and efficient tool for determining the jack-up responses under different environmental load combinations for JPA. Based on the optimal univariate probability distributions of each environmental parameter, multi-dimensional copulas are adopted to construct the conditional and joint probability distributions. These distributions, in turn, yield environmental design parameters under different design criteria and corresponding horizontal displacements of the jack-up platform. Results show that environmental design parameters and corresponding horizontal displacements are smaller compared to those obtained from the univariate probability distribution for the same return period. The FEM-BPNN-JPA coupled approach proves to be effective in estimating environmental design parameters by considering parameter correlations.