AI-driven digital twin for uncertainty-aware structural health monitoring of offshore wind turbines considering biofouling effects and reliability prediction

Abstract

Marine biofouling on offshore wind turbine substructures poses major challenges to structural integrity and the dependability of vibration-based structural health monitoring (SHM) because it drastically changes mass distribution, decreases structural stiffness, and increases hydrodynamic loading. Conventional SHM methods often misdiagnose biofouling effects as structural damage. To address this limitation, the present study introduces an AI-driven digital twin framework that integrates artificial intelligence (AI), real-time Internet of Things (IoT)-enabled monitoring, and advanced numerical modeling to enhance damage detection and reliability assessment. The framework combines finite element analysis, computational fluid dynamics, and AI-based predictive analytics using convolutional neural networks, XGBoost, and Bayesian inference models to evaluate the dynamic behavior of four-legged jacket and tripod-type platforms under both clean and biofouled conditions. Real-time sensor data—including vibration, strain, and environmental measurements—are processed through machine learning models for accurate damage localization and predictive maintenance. Validation against real-world data indicates that biofouling, which increases structural mass by approximately 1,350  kg/m3 , causes a 6–12% reduction in natural frequencies and distorts mode shapes, complicating conventional SHM interpretation. The proposed AI-enhanced modal strain energy approach, supported by Bayesian uncertainty quantification and frequency compensation techniques, improves damage detection accuracy by 15–25% and reduces false positives by 25%. Moreover, an IoT-based biofouling detection system further increases SHM reliability by 18%. A  cost-benefit analysis reveals that AI-guided predictive maintenance strategies reduce inspection costs by 22%, decrease unplanned operational downtime by 60%, and accelerate damage detection by 30%. These findings demonstrate the potential of AI-integrated SHM systems to optimize offshore wind farm management, reduce operational risks, and extend structural service life.