Real-Time Monitoring of Clump Weight Integrity Loss in Floating Wind Turbines via Deep Learning

Abstract

Floating Offshore Wind Turbines (FOWTs) represent a promising avenue for harnessing wind energy in deep-water regions through the use of mooring systems. However, mooring lines account for over 20% of the total construction cost and require careful optimization to ensure both structural stability and economic feasibility. One effective strategy for enhancing mooring performance is the strategic placement of clump weights along the lines. While beneficial, these clump weights are vulnerable to damage or detachment under harsh marine conditions, potentially compromising the structural integrity and reducing the service life of FOWTs. To address this challenge, we propose a deep learning-based method for real-time detection of clump weight loss using a Long Short-Term Memory (LSTM) network. The training data for the LSTM model are generated using high-fidelity simulations conducted in the open-source software OpenFAST, based on the 5 MW OC3-Hywind spar-buoy FOWT model equipped with clump-weighted mooring lines. The proposed LSTM classifier achieved 86% accuracy on a held-out test split and 73% on a completely unseen dataset, demonstrating both effectiveness in detecting clump weight loss and generalization capability for real-time condition monitoring in FOWT mooring systems.