Modeling the subgrid-scale kinetic energy using artificial neural network: comparison with a dynamic subgrid-scale model

A deep neural network (DNN) is trained and employed to predict the subgrid-scale (SGS) kinetic energy, 𝑘 SGS. To train the DNN, direct numerical simulation (DNS) data of plain turbulent channel flow at wall friction Reynolds number 𝑅 𝑒 𝜏 =۳۸۱ is employed. The DNS data is further filtered using a box filter in the flow homogeneous directions. The trained DNN is successfully tested to predict the subgrid-scale kinetic energy, 𝑘 SGS. A close correspondence is found between the DNN-predicted 𝑘 SGS and filtered DNS data of turbulent channel flow. Furthermore, the probability density function (PDF) of 𝑘 SGS and its skewness are well-predicted by the DNN. The DNN predictions are further compared with those of a dynamic SGS model (DSM) that is commonly used in the literature for the prediction of the 𝑘 SGS.

کلیدواژه ها:

artificial neural network ، subgrid-scale kinetic energy ، dynamic procedure