基于降阶模型的叶片气动弹性预测模拟数据

针对气弹模拟时域法效率偏低的问题，提出基于神经网络模型的叶片非定常气动力降阶建模技术。数据集基于开发的非定常气动力降阶模型进行非定常气动力预测，数据来源于课题1开发的叶片气弹模拟降阶模型，该模型采用两种建模方法：基于LSTM的变种神经网络和基于图卷积神经网络的Graph Wavenet模型。数据集主要包括各气动/结构插值点上的三维气动力数据，并耦合结构有限元软件实现气弹模拟，输出叶片的位移响应数据。数据量约600MB。

Aiming at the low efficiency of time-domain methods for aeroelastic simulation, a reduced-order modeling technique for unsteady aerodynamic forces on blades based on neural network models is proposed. The dataset is utilized for unsteady aerodynamic force prediction based on the developed unsteady aerodynamic force reduced-order model. The data originates from the reduced-order model for blade aeroelastic simulation developed in Task 1, which employs two modeling methods: a variant of Long Short-Term Memory (LSTM)-based neural network and the Graph Wavenet model based on graph convolutional neural networks. The dataset primarily includes three-dimensional aerodynamic force data at each aerodynamic/structural interpolation point, and is coupled with structural finite element software to implement aeroelastic simulation, outputting the displacement response data of the blades. The overall data volume is approximately 600 MB.