| Popis: |
Safe and productive offshore operations are of utmost importance, with monitoring the integrity of mooring lines on floating offshore platforms being one of the key factors. The conventional method uses sensors installed on mooring components, which may fail over time and can be costly to replace. Alternative methods using dry and non-intrusive monitoring systems offer a lot of potentials to the industry. An alternative method that uses only Differential Global Positioning System (DGPS) data has been proposed by Sidarta et al. (2018, 2019), and it does not require any information on environmental conditions. This alternative method is based on monitoring shifts in the low-frequency periods and mean yaw angles as a function of vessel positions, mass and added mass. The method utilizes Artificial Intelligence, specifically Artificial Neural Network (ANN), for the detection of mooring line failure, which is a pattern recognition and classification problem. The ANN model learns to recognize and classify patterns of intact mooring lines and those of a broken line. One of the proposed models is a group identification model, in which the model identifies the mooring group that has a broken line. This paper shows that an ANN model can be quite robust and tolerant in dealing with conditions that are somewhat different from its training. As an example, an ANN model for detecting mooring line failure on a spread moored FPSO has been trained using MLTSIM hydrodynamic simulations with quasi static model of the mooring lines and risers to significantly reduce the computational time to generate the ANN training data. The trained ANN model can properly function when tested using fully coupled OrcaFlex hydrodynamic simulations with environmental conditions that are not included in the training. Moreover, although the ANN model has been trained using simulations with a completely removed line, the trained model can still function for a line broken at the bottom. This ANN model is an ANN-based status detection model, which is one of the key components in the ALANN (Anchor Lines monitoring using Artificial Neural Networks) System. The system also composes of an ANN-based system evaluation model, an algorithm-based status detection program and an event detection program. A series of fully coupled dynamic simulations have been used to test the ALANN System. Most of the simulations have a single mooring line failure that occurs randomly during simulation, and the failed line varies for different simulations. Each simulation lasts for six hours. The ALANN System uses a two-hour time window at a time and moves every 20 minutes. The tests demonstrate how each component of the ALANN System contributes to and improves the robustness of the overall solution. |
