Research on grid-dependence of neural network turbulence model

Turbulence machine learning based on deep neural network has become a research hotspot in turbulence modeling. Although most turbulence models have strict requirements on grid dependence, there are few analyses on grid dependence on the coupling of models and equations. In this article, a neural network turbulence model is constructed for the flow around airfoil with high Reynolds number, and the effects of wall-normal grid spacing on the calculation accuracy are studied. The results show that compared with the traditional differential equation turbulence model, the neural network turbulence model can break through the limitation of y(+) < 1 and relax the requirement of the normal density of the boundary layer grid while ensuring the accuracy.