Mooring Line Failure Detection Using Artificial Neural Networks: An Application to Field Data Including Artificial Failure Cases

Abstract

Abstract Alternative methods to using physical sensors for monitoring the integrity of mooring lines on floating offshore platforms are of great interest in the Offshore O&G industry. These alternative methods can be used in parallel with the physical sensors at the start of the service life of the asset, measurements from the physical sensors can be used to validate these methods, and the validated methods can be vital tools when the physical sensors are no longer working. Technip Energies has presented ALANN (Anchor Lines monitoring using Artificial Neural Networks) system in a 2021 OTC paper to detect mooring line failure using a dry monitoring system. The system requires only monitoring the positions and headings of the vessel, and it requires information on the draft of the vessel for an FPSO. The system combines status-based detection and event detection to determine mooring line condition, whether they are intact or there is mooring line failure. Artificial Neural Networks (ANN) play a very important role in the status detection with the ability to detect subtle shifts in patterns of vessel motions from intact lines condition to a mooring line failure condition. Numerical algorithms are used for status detection that complements ANN for benign environmental conditions and for event detection. An ALANN system has been developed, including training ANN models using numerical simulations, for a spread moored FPSO in West Africa. The system has been tested using field data that include time series of easting and northing positions of the vessel, vessel headings and vessel drafts. In addition, field data have been modified to include artificial mooring line failure cases, and these altered field data are used to further test the system. ALANN system performs very well for measured data in good quality independently from the FPSO loading condition, and it is able to detect all the artificial failure cases. The tests confirm and demonstrate how components of the ALANN system contribute to and improve the robustness of the overall solution. Important lessons learned, including challenges that have been encountered due to typical real-life sensor and communication issues (leading to gaps, drift, spikes and/or varying sampling within the data), along with comparison of measured and simulated data are presented in this paper.