Intelligent long-term prediction of wind-induced response of super high-rise buildings during typhoons based on wind tunnel tests

The vibration prediction of super high-rise buildings during typhoons represents a significant challenge in the field of structural wind resistance. Despite considerable research efforts, the longterm prediction of structural response during typhoons remains an unsolved problem. An intelligent method for predicting the long-term response of super high-rise buildings during typhoons was proposed. First, a neural network (NN) training set was constructed on the basis of wind tunnel tests by considering the impact of varying parameters, including wind speed, wind direction, structural frequency, and damping ratio. The structural response NN prediction model was developed based on the Cascade Forward-Backward Propagation Network (CFBPN). Subsequently, a linear typhoon wind field model was utilized combined with AI-based weather forecasting methods to predict meteorological conditions. The integration of the wind field from the meteorological prediction model with the response NN prediction model enabled the rapid prediction of the long-term responses of super high-rise buildings. The efficacy of the proposed method has been validated through a series of case studies examining the structural acceleration response of Kingkey 100 during Super Typhoon Mangkhut (1822). The results demonstrate that the CFBPN model is effective in accurately predicting the structural acceleration response, with a correlation coefficient exceeding 0.99. The long-term response prediction results, based on the measured wind field, demonstrate a correlation coefficient exceeding 0.89 with the actual measurements. By integrating the linear typhoon wind field model, long-term wind field predictions can be made. Furthermore, incorporating the structural dynamic characteristics enables rapid prediction of Kingkey 100's wind-induced response, validating the feasibility of using the intelligent meteorological prediction model for long-term structural response forecasting. This study explores the application of machine learning models to assess wind-induced responses in complex structures. It offers a reference for predicting structural vibration during typhoons, enhancing urban emergency response capabilities, and enabling a shift from reactive emergency measures to proactive decision-making.