Research Of Flow Field Data Postprocessing Methods Based On Deep Learning

The flow field data is the data that records the motion of fluid particles.Computational fluid mechanics uses computer technology to solve and simulate fluid motion.The large amount of computational fluid dynamics makes the calculation speed an important factor restricting its development.In addition,because of the special form of the flow field data,it takes a lot of effort to visualize the flow field data.Therefore,this paper proposes an image processing method for flow field data and an image regression calculation model based on deep learning.They are applied to the post-processing of flow field data to increase calculation speed of computational fluid dynamics and to simplify the visualization of calculation results,which accelerate the post-processing of the entire flow field data.The aerospace field is a typical application scenario of flow field data calculation and processing.This paper uses the aircraft design process as an example,and applies the proposed regression calculation model to the calculation process of the aircraft flow field data,and validates the proposed method and model.Feasibility.The main research content of this paper is as follows:(1)Image processing method of flow field dataStudy and analyze the structure of flow field data.Taking the finite element mesh data in Tecplot software as an example,the method of extracting the characteristics of the flow field data grid and the method of converting the 3D flow field data into 2D images are proposed.(2)Regression calculation model of flow field dataAn image regression model based on full convolutional network is proposed to complete the pixel-level regression calculation of the flow field data.The network structure and training method of the regression model are studied and modified for the particularity of the flow field data.(3)Simulated regression calculation of flow field dataUsing the proposed method for image data processing and image regression model for regression analysis of flight data,the regression model was tested for regression error and time performance,and the generalized performance of the regression model on different data and grids was tested.