Abstract
he growing deployment of offshore wind turbines (OWTs) demands efficient structural health monitoring (SHM) strategies to ensure safety and reliability under multi-hazard conditions. OWTs experience complex dynamic interactions from wind, waves, and seismic forces, making real-time predictive capabilities essential for condition-based maintenance. However, high-fidelity simulations such as finite element models (FEM) are computationally intensive for real-time use. This study presents an ensemble machine learning framework as a fast surrogate for FEM-based simulations, enabling rapid prediction of key structural responses including hub displacement, acceleration, and pile head rotation. The model captures intricate relationships between key system parameters, including wind turbine capacity, offshore environmental conditions (wind speed, wave height, wave period, and ground acceleration), and sites-specific factors (water depth, soil properties). Model performance and generalisation are systematically assessed using a statistical learning workflow. The proposed surrogate model significantly reduces computational cost while retaining predictive accuracy, supporting real-time SHM integration and facilitating informed decision-making. This approach advances AI-driven SHM for offshore renewables, providing scalable, data-efficient solutions to enhance resilience, operational efficiency, and sustainability.