Data-Driven Fatigue Reliability Evaluation of Offshore Wind Turbines under Floating Ice Loading

Floating ice loading is a severe natural hazard for offshore wind turbines (OWTs) operating in cold sea regions. The long-term ice-induced vibration resulting from winter drift ice presents a significant threat to the fatigue reliability of OWTs. Machine learning not only exhibits robust data-driven capabilities but also efficiently handles complex relationships among multiple factors. Through effective learning and utilization of diverse features, a comprehensive assessment of the fatigue reliability of structures becomes achievable, consequently resulting in increased efficiency, accuracy, and adaptability of the assessment. This study conducts ice-induced vibration simulations on a 5-MW monopile OWT. Employing orthogonal experimental methods, the study investigates the influence of three critical ice parameters, i.e., ice thickness, ice velocity, and ice crushing strength, on the fatigue damage caused by ice loading on OWTs. Additionally, it explores the effects of ice loading exceedance probability on extreme fatigue damage values of OWTs under combined ice and wind loadings. Furthermore, a surrogate model based on support vector machine (SVM) is developed to effectively capture the intricate mapping relationship between fatigue damage and various random environmental parameters. By conducting a comprehensive evaluation that considers the probabilistic characteristics of ice loading, wind loading, and the S-N curve, this study assesses the fatigue reliability under combined ice and wind. Moreover, this study analyzes the impact of operational duration during winter drift ice on the fatigue reliability index of OWTs. The findings of this study can provide a theoretical basis for devising operational strategies for OWTs operating in icy sea regions.