Analysis of FPSO dropped objects combining Monte Carlo simulation and neural network-genetic approach

Dropped objects is one of the most hazardous accidents of floating production storage offloading (FPSO) due to the potentially severe consequences, therefore, the quantitative calculation of dropped objects impact risk, which is featured with uncertainties, is essential. Monte Carlo Simulation (MCS) is a method for this intention with flexibility and efficiency. However, wide calculating resources are needed for MCS-based accident analysis, especially when multiple random variables are concerned. Moreover, dropped objects collision damage models are perplexed and usually pronounced by the Finite Element Method (FEM), which is of a computational complexity. To address this issue, this paper considered a combined methodology: the Artificial Neural Network (ANN) adjusted by genetic algorithm (GA) is united with MCS to analyze dropped objects collision failure probability quantitatively. A time-dependent progressive simulation is employed to forecast the structural response caused by dropped objects collision. Then the ANN-GA is trained based upon the collision data and used as an alternative for the computational FEM performs with the MCS methodology. The risk level for the dropped objects collision of FPSO is assessed by comparing the proposed method and the DNV rules, and a qualitative uncertainty analysis is also conducted in this quantitative failure analysis.