Session: 02-06-01 Data-driven Models for Marine Structures

Submission Number: 156367

Detection of Failed Mooring Lines of Spread-Moored FPSO Using Platform Motions and Artificial Neural Networks 

This study explores the application of an artificial neural network for monitoring mooring failures in a spread-moored FPSO (Floating Production Storage and Offloading) vessel from platform motion data. Time-domain numerical simulations were conducted to generate a comprehensive dataset. Specifically, a coupled system consisting of a generic FPSO, twelve mooring lines, and a single riser was modeled. The mooring configuration included four groups, each comprising three spread mooring lines with a 5-degree interval between lines. Environmental conditions for the simulation were derived from four years of wave, wind, and current data collected from ERA5 and HYCOM databases. The simulations encompassed both intact and various single-line failure conditions, capturing six degrees of freedom in platform motion. This dataset was subsequently used to train and optimize artificial neural networks through a systematic hyperparameter tuning process.

In the ANN-based detection algorithm, statistical parameters of platform motion served as input, while the output classified conditions into either the mooring group containing a failed mooring line or the exact failed line number. The model achieved a 100% accuracy rate in identifying the group with a failed mooring line. However, accuracy dropped to 84% when pinpointing the exact line number. The confusion matrix revealed that the algorithm tended to misidentify failed lines as their adjacent lines rather than as lines from other groups. This limitation likely arises from the subtle motion differences between neighboring lines, given their close spacing at a 5-degree interval.

Despite this, the algorithm's consistent success in identifying the group containing the failed mooring line underscores its potential as a reliable, practical tool for offshore operations, offering an efficient means of monitoring mooring integrity in real time.

Presenting Author: Omar Jebari Florida Institute of Technology

Presenting Author Biography: Omar Jebari is currently Ph.D candidate in the Department of Ocean Engineering at Florida Institute of Technology.