Active learning with a multi-point enrichment strategy for Multi-Location Fatigue Assessment of offshore wind turbines

Long-term fatigue damage estimation is essential for designing offshore wind turbines but requires thousands of multi-physics simulations, making it highly time-consuming. Despite efforts to reduce computational demands, an efficient solution is still needed. This study proposes an active learning Kriging method to estimate long-term fatigue damage across multiple locations in offshore wind turbine subsystems. The Kriging models predict fatigue under diverse wind-wave conditions, and an active learning approach is tailored for multi-location fatigue assessment. To enhance efficiency, a practical multi-point enrichment strategy based on wind speed intervals is introduced, enabling parallel computation during the active learning process. The method is tested on two IEA 15MW offshore turbine types (bottom-fixed and floating) at four different offshore sites. Results demonstrate that the proposed approach effectively estimates fatigue damage across various subsystem locations, achieving over 20-fold reductions in computational cost with a maximum error of less than 5%. The multi-point enrichment strategy further minimizes active learning iterations. This approach and enrichment strategy together have significant potential to reduce both the time and costs associated with offshore wind turbine design, making it a valuable tool for engineers in this field. • Long-term fatigue damage assessment of offshore wind turbines. • An active learning approach for multi-location fatigue assessment. • A practical multi-point enrichment strategy based on wind-speed intervals. • Two types of offshore wind turbines and four different offshore sites. • Increase efficiency by more than 20 times with an error of less than 5%.