Advances in machine learning for wind-induced fluid-structure interaction of large-scale structures

With the rapid development of computational technology and data science, machine learning provides a novel research paradigm for addressing complex fluid-structure interaction problems in large-scale structural wind effects. This paper systematically reviews recent advances in machine learning applications for wind effects on large-scale structures, focusing on four key aspects: structural surface wind pressure prediction, wind-induced response analysis and modeling, intelligent identification of aerodynamic equations, and reinforcement learning-based structural vibration control. For structural surface wind pressure prediction, machine learning effectively captures complex nonlinear wind pressure characteristics on structural surfaces. In the analysis and modeling of structural wind-induced responses, machine learning techniques enables accurate identification and modeling of abnormal large-amplitude vibrations of large-scale structures. Regarding intelligent identification of aerodynamic equations, data-driven machine learning significantly enhances the automation and accuracy of nonlinear equation identification. For structural vibration control, reinforcement learning offers optimized real-time active control strategies. However, challenges persist in data fusion, model generalization, and physical interpretability. Future studies should integrate physical mechanisms with data-driven models to develop machine learning approaches characterized by high generalization, robustness, and physical interpretability, thus further advancing the intelligent development of structural wind engineering.