A point cloud deep neural network metamodel method for aerodynamic prediction
作者机构:School of Aerospace EngineeringBeijing Institute of TechnologyBeijing 100081China Department of Mechanical EngineeringImperial College LondonLondon SW72AZUK
出 版 物:《Chinese Journal of Aeronautics》 (中国航空学报(英文版))
年 卷 期:2023年第36卷第4期
页 面:92-103页
核心收录:
学科分类:0710[理学-生物学] 12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 080103[工学-流体力学] 080104[工学-工程力学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术(可授工学、理学学位)] 0801[工学-力学(可授工学、理学学位)]
基 金:supported by the National Natural Science Foundation of China(No.52175214) the Basic Research Program of Equipment Development Department(No.514010103-302)
主 题:Aerodynamic prediction Deep neural network Metamodel Point clouds Robust shape optimization
摘 要:Aiming to reduce the high expense of 3-Dimensional(3D)aerodynamics numerical sim-ulations and overcome the limitations of the traditional parametric learning methods,a point cloud deep learning non-parametric metamodel method is proposed in this *** 3D geometric data,corresponding to the object boundaries,are chosen as point clouds and a deep learning neural net-work metamodel fed by the point clouds is further established based on the PointNet *** network can learn an end-to-end mapping between spatial positions of the object surface and CFD numerical *** the proposed aerodynamic metamodel approach,the point clouds are constructed by collecting the coordinates of grid vertices on the object surface in a CFD domain,which can maintain the boundary smoothness and allow the network to detect small changes between ***,the point clouds are easily accessible from 3D *** point cloud deep learning neural network,which employs re-sampling technique,the spatial transformer network and the fully connected layer,is developed to predict the aerodynamic char-acteristics of 3D *** effectiveness of the proposed metamodel method is further verified by aerodynamic prediction and robust shape optimization of the ONERA M6 *** results show that the proposed method can achieve more satisfactory agreement with the experimental measurements compared to the parametric-learning-based deep neural network.