A DNN-based approach to predict dynamic mooring tensions for semi-submersible platform under a mooring line failure condition

Abstract

The station-keeping system of offshore platforms is susceptible to complex metocean environments. The failure of the mooring line is of utmost concern due to uncertainty in the mooring line's behavior and the severe aftermath of the accidents. Once the mooring line breakage occurs, the remaining mooring lines should provide sufficient tension to warrant the continual production operations and personnel safety on the platform. This paper proposes a deep neural networks (DNN) approach to predict the dynamic mooring line tension under one mooring line failure condition. Firstly, the tension change of mooring lines induced by mooring failure is investigated in two hydrodynamic models, and then select the tensions on the mooring line with maximum sensitivity to failure as the output objective. Secondly, using the responses of floating structures and mooring tensions to construct the two datasets of different hydrodynamic models, then train two DNN models and utilize grid search to determine the optimal network structure. Finally, some case studies with different mooring arrangements and sea state conditions are employed to verify the feasibility and adaptability of two established DNN models and compare their accuracy. According to the unseen data test results, the DNN-II model shows a better prediction performance compared to the DNN-I model in all test cases. Based on the results of the two types of the correlation coefficient, the main control variables in mooring tensions prediction are believed to be significantly increased sway and surge motion induced by mooring broken. This study provides an idea for practical engineering, through combining with the mature and stable Global Positioning System (GPS) in a platform or FPSO, the remaining mooring line tensions under mooring failure conditions can be monitored and controlled in real time.