An intelligent predictive maintenance framework for floating offshore wind turbine based on structural damage prediction

Predictive maintenance for floating offshore wind turbines (FOWTs) presents significant challenges due to the need for flexible integration of structural damage prediction and maintenance decision-making under uncertainty. In particular, it requires accurate estimation of damage progression and adaptive planning of maintenance actions that balance safety, reliability, and cost across the lifecycle. To address these challenges, this study proposes an intelligent predictive maintenance framework that couples a damage magnitude prediction model with a reinforcement learning-based decision-making module. The prediction model estimates damage magnitude, quantifies uncertainty, and evaluates failure probability within inspection intervals, while the decision module selects optimal preventive actions to maintain structural integrity and economic efficiency. The framework is validated using a high-fidelity simulated FOWT dataset. The results demonstrate that prediction uncertainty decreases as damage severity increases, indicating greater model confidence in critical conditions. Furthermore, the reinforcement learning module adaptively balances risk and operational cost, yielding near-optimal maintenance schedules even under cost uncertainty. Overall, proposed framework reduces operation and maintenance costs, enhances safety, and supports sustainable FOWT operation.