Research On Grid Quality Discrimination Technology Based On Convolutional Neural Network

With the continuous improvement of computing capacity,computational fluid dynamics(CFD)plays an increasingly important role in practical engineering applications.Grid quality determines the accuracy of simulation data.Therefore,the process of grid quality discrimination is particularly important,which indirectly determines the accuracy of later calculation work.The process of grid quality discrimination is usually performed manually.It will restrict the development of high-quality grid and therefore need automatic generation technology in the next stage.This thesis investigates the automatic grid quality discrimination technology.First,a scheme of CFD grid quality discrimination based on convolutional neural network is designed,and the flow of automatic grid quality discrimination is determined.Then,through the research of convolutional neural network identification technology,a set of optimization strategy for training and identification of CFD grid data is determined.Finally,based on the Tensor Flow machine learning framework,a CFD mesh quality automatic discrimination program is developed with Python programming language,and the two-dimensional airfoil mesh quality automatic discrimination is realized.In this thesis,the convolutional neural network recognition and classification method is applied to the intelligent discrimination of grid quality.Gradient grid data automatic generation program and grid quality feature extraction tool is researched and developed,The original grid data are made into data sets that can be trained for convolutional neural network model.The convolutional neural network model is trained,and then the final classification and recognition results are obtained.The optimal model is obtained by testing and evaluating.The results show that the method can automatically identify the mesh quality in two-dimensional airfoil mesh recognition with high accuracy and robustness.Compared with the manual process,the method proposed in the thesis can greatly improve the efficiency of grid quality discrimination and minimize the workload of grid quality discrimination.CFD grid quality automatic discrimination technology provides tremendous technical support for grid automatic generation and a bright prospect in the engineering research and application.