摘要
非定常流场的高维非线性动力学特征使快速准确预测流体运动规律成为计算流体力学(Computational Fluid Dynamics, CFD)研究的难点。针对CFD求解首先需要划分网格将求解区域离散化,采用基于网格数据的图神经网络来捕捉非定常流的时空特征,能适应网格的离散化并模拟非定常流体的流动过程,实现流场变量的快速预测。文中输入自适应网格和初始流场变量数据,通过训练后的模型就能快速准确预测未来时刻的流场变量,同时自适应网格可兼顾预测精度和效率,与图神经网络有较好的适配性。结果表明,采用图神经网络预测的流场数据与CFD数值计算的流场数据具有较好的一致性,并能明显缩短计算时间,为流场变量的预测提供了一种新的解决方案。
The high-dimensional and non-linear dynamic characteristics of unsteady flow field make it difficult to predict fluid movement quickly and accurately in Computational Fluid Dynamics(CFD).First, CFD computation needs to discretize the solution region into mesh.then we adoptgraph neural network for the mesh-based data to capture the temporal and spatial characteristics of unsteady flow. the method can adapt the meshdiscretization and learn the unsteady flowsimulationsto predict field variables rapidly. By feeding adaptive mesh and initial flow field variables data, the trained model can predict the future flow field variables quickly and accurately.At the same time, the adaptive mesh can strike favorable trade-offs between accuracy and efficiency, adapting the graph neural network well. The results show that the flow field data predicted by the graph neural network is in good agreement with the numerical calculationin CFD, and the calculation time can be shortened obviously, which provides a new solutionfor the prediction of flow field variables.
作者
李治龙
周恒安
翁俊辉
欧洺余
朱宏娜
Li Zhilong;Zhou Hengan;Weng Junhui;Ou Mingyu;Zhu Hongna(School of Physical Science and Technology,Southwest Jiaotong University,Chengdu,China;School of Information Science and Technology,Southwest Jiaotong University,Chengdu,China)
出处
《科学技术创新》
2023年第3期104-108,共5页
Scientific and Technological Innovation
基金
中央高校基本科研基金(2682021GF018)
四川省科技厅计划项目(2020YJ0016)。
关键词
计算流体力学
网格
图神经网络
非定常流场
预测
computational fluid dynamics
mesh
graph neural network
unsteady flow
prediction