Modeling the subgrid-scale kinetic energy in a turbulent channel flow using artificial neural network

Document Type : Research Article

Authors

Faculty of Mechanical and Energy Engineering, Shahid Beheshti University, Tehran, Iran

Abstract

Modeling the subgrid-scale (SGS) kinetic energy, K_SGS, has important applications in large-eddy simulation (LES), including LES using the lattice Boltzmann method and formulation of advanced SGS models. In this study, a deep neural network (DNN) is specifically developed to predict K_SGS for LES of turbulent channel flow. To obtain the training data for the DNN, a direct numerical simulation (DNS) of turbulent channel flow at the friction Reynolds number Re_τ=381 is performed using an existing highly accurate pseudo-spectral method. The impact of the DNN configuration on its predictions is studied by examining the mean, probability density function (PDF) and skewness of K_SGS. Moreover, the correlation with the filtered DNS data, its relative and root mean square error are also examined using a priori analysis. Appreciable improvements in the predictions were observed with increasing the number of neurons in the hidden layer, up to 64. Increasing the number of hidden layers to two and three showed small improvements in the predictions of the DNN.

Performance of the DNN is also compared with a dynamic SGS model (DSM). The comparison reveals that the DNN predictions reach correlation coefficients higher than 90% with the filtered DNS data, whereas the DSM predictions only reach up to about 50%. Also, a closer agreement with the filtered DNS data was observed for the DNN predictions of K_SGS, compared with the DSM.

Keywords

Main Subjects

AUT Journal of Mechanical Engineering
Volume 8, Issue 1
March 2024
Pages 2-2

Files

Share

How to cite

Statistics