Rotational and reflectional equivariant convolutional neural network for data-limited applications: Multiphase flow demonstration
Rotational and reflectional equivariant convolutional neural network for data-limited applications: Multiphase flow demonstration
复制标题
DOI：
10.1063/5.0066049
复制
发表时间：
2021-08
影响因子：
4.6
通讯作者：
B. Siddani;S. Balachandar;R. Fang
中科院分区：
文献类型：
--
作者：
B. Siddani;S. Balachandar;R. Fang
本文利用卷积神经网络(CNN)对离散多相流中固定颗粒在随机分布的影响下的稳态颗粒分辨流体流动进行了数值模拟。所考虑的问题涉及关于平均速度(流向)的旋转对称性。因此，这项工作使用$\mathbf{\extbf{SE(3)-等变}}$来加强这种对称性，它是平移和三维旋转等变的3维特殊欧几里得群，CNN结构。本研究主要探讨SE(3)等变网络的泛化能力与效益。谨慎应用对称性感知数据驱动方法，得到了雷诺数和颗粒体积分数组合在[86.22,172.96]和[0.11，0.45]范围内的精确合成流场。
This article deals with approximating steady-state particle-resolved fluid flow around a fixed particle of interest under the influence of randomly distributed stationary particles in a dispersed multiphase setup using Convolutional Neural Network (CNN). The considered problem involves rotational symmetry about the mean velocity (streamwise) direction. Thus, this work enforces this symmetry using $\mathbf{\textbf{SE(3)-equivariant}}$, special Euclidean group of dimension 3, CNN architecture, which is translation and three-dimensional rotation equivariant. This study mainly explores the generalization capabilities and benefits of SE(3)-equivariant network. Accurate synthetic flow fields for Reynolds number and particle volume fraction combinations spanning over a range of [86.22, 172.96] and [0.11, 0.45] respectively are produced with careful application of symmetry-aware data-driven approach.