生成对抗网络在二维网格生成上的探索
Exploration of Generative Adversarial Networks on 2D Mesh Generation
蒋成义 1肖素梅2
作者信息
- 1. 西南科技大学 制造科学与工程学院,四川 绵阳 621010
- 2. 西南科技大学 制造科学与工程学院,四川 绵阳 621010;西南科技大学 外国语学院,四川 绵阳 621010
- 折叠
摘要
流体力学中的网格生成至关重要,因为它是计算机求解的首要步骤.但这项工作既耗时又费力,因此提高网格生成速度已成为研究的重点.基于此,文章设计了一种基于神经网络的网格生成方法,利用生成对抗网络来生成相应的网格.首先,通过现有的二维平面网格数据提取出相关信息,作为训练数据;然后,利用生成对抗网络的生成器和判别器的相互迭代训练得到最优的结果;最后,通过生成对抗网络生成需要的网格数据,展示相应结果.实验表明,对于简单二维平面网格可以使用神经网络进行智能化处理,并生成合适的网格数据,生成对抗网络在网格生成上具有良好的应用.
Abstract
Grid generation in fluid mechanics is crucial because it is the first step in computer solving.However,this work is time-consuming and laborious,so improving the speed of grid generation has become the focus of research.Based on this,this paper designs a method of grid generation based on neural network,using generative adversarial network to generate the corresponding grid.First,the relevant information is extracted from the existing two-dimensional planar grid data as training data;then,the best results are obtained by using the mutual iterative training of the generator and discriminator of the generative adversarial network;finally,the required grid data are generated by the generative adversarial network,and the corresponding results are displayed.This experiment shows that for simple two-dimensional planar meshes can be intelligently processed using neural networks and generate suitable mesh data,and that generative adversarial networks have good applications for mesh generation.
关键词
神经网络/网格生成/生成对抗网络Key words
neural network/grid generation/generative adversarial network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据