考虑上游振动作用的串列方柱表面风压分布智能预测OA

To accurately predict the surface wind pressure distribution of a tandem square column structure under vibration feedback,a wind pressure prediction framework based on K-nearest neighbor dynamic mode decompo-sition(KDMD)and multi-scale convolutional neural network(MSCNN)was proposed.First,modal decompo-sition on the flow field data around the vibrating tandem square column structure was conducted using KDMD.The key typical modal features in the flow field,including low-frequency modes,main vortex shedding modes,forced vibration modes,and second-order main vortex shedding modes,were identified and extracted.Then,the modal features of different scales extracted from the flow field were used as the input of MSCNN to predict the surface wind pressure distribution of the vibrating tandem square columns.And the comparison was conducted against ten mainstream machine learning algorithms.The research results indicate that the prediction accuracy of the proposed method is the highest in the prediction of the wind pressure distribution of the vibrating tandem square columns.The mean square error,root mean square error,and mean absolute error between the predicted wind pressure coefficient and the measured value are 0.004 6,0.067 7,and 0.038 3,respectively,which are re-duced by at least 40%compared with those of other mainstream machine learning algorithms.