Spatial artificial neural network model for subgrid-scale stress and heat flux of compressible turbulence

The subgrid-scale(SGS)stress and SGS heat flux are modeled by using an artificial neural network(ANN)for large eddy simulation(LES)of compressible turbulence.The input features of ANN model are based on the first-order and second-order derivatives of filtered velocity and temperature at different spatial locations.The proposed spatial artificial neural network(SANN)model gives much larger correlation coefficients and much smaller relative errors than the gradient model in an a priori analysis.In an a posteriori analysis,the SANN model performs better than the dynamic mixed model(DMM)in the prediction of spectra and statistical properties of velocity and temperature,and the instantaneous flow structures.

Constraine d large-e ddy simulation of turbulent flow over inhomogeneous rough surfaces

In this work we extend the method of the constrained large-eddy simulation(CLES)to simulate the tur-bulent flow over inhomogeneous rough walls.In the original concept of CLES,the subgrid-scale(SGS)stress is constrained so that the mean part and the fluctuation part of the SGS stress can be modelled separately to improve the accuracy of the simulation result.Here in the simulation of the rough-wall flows,we propose to interpret the extra stress terms in the CLES formulation as the roughness-induced stress so that the roughness inhomogeneity can be incorporated by modifying the formulation of the constrained SGS stress.This is examined with the simulations of the channel flow with the spanwise alternating high/low roughness strips.Then the CLES method is employed to investigate the temporal response of the turbulence to the change of the wall condition from rough to smooth.We demonstrate that the temporal development of the internal boundary layer is just similar to that in a spatial rough-to-smooth transition process,and the spanwise roughness inhomogeneity has little impact on the transition process.

Flow structures, nonlinear inertial waves and energy transfer in rotating spheres

We investigate flow structures,nonlinear inertial waves and energy transfer in a rotating fluid sphere,using a Galerkin spectral method based on helical-wave decomposition(HWD).Numerical simulations of flows in a sphere are performed with different system rotation rates,where a large-scale forcing is employed.For the case without system rotation,the intense vortex structures are tube-like.When a weak rotation is introduced,small-scale structures are reduced and vortex tubes tend to align with the rotation axis.As the rotation rate increases,a large-scale anticyclonic vortex structure is formed near the rotation axis.The structure is shown to be led by certain geostrophic modes.When the rotation rate further increases,a cyclone and an anticyclone emerge from the top and bottom of the boundary,respectively,where two quasi-geostrophic equatorially symmetric inertial waves dominate the flow.Based on HWD,effects of spherical confinement on rotating turbulence are systematically studied.It is found that the forward cascade becomes weaker as the rotation increases.When the rotation rate becomes larger than some critical value,dual energy cascades emerge,with an inverse cascade at large scales and a forward cascade at small scales.Finally,the flow behavior near the boundary is studied,where the average boundary layer thickness gets smaller when system rotation increases.The flow behavior in the boundary layer is closely related to the interior flow structures,which create significant mass flux between the boundary layer and the interior fluid through Ekman pumping.

Numerical investigation of plane Couette flow with weak spanwise rotation

Direct numerical simulation of rotating plane Couette flow(RPCF) at Re_w= 1300 and Ro = 0.02 was performed with different mesh resolutions and different sizes of computation domain. Our results showed that a grid resolution in wall units with ?x^+=8.51, ?z^+= 4.26, ?y^+|_(min)= 0.0873 and ?y^+|_(max)= 3.89 is fine enough to simulate the problem at the present parameters. The streamwise length Lxand spanwise length Lzof the computational box have different impacts on the flow statistics, where the statistics were converged if Lxis longer than 8πh, while no converged results were obtained for different Lz. More importantly,our results with very long simulation time showed that a state transition would happen if L_x≥ 8πh, from a state with four pairs of roll cells to a state with three pairs of roll cells with L_z= 6πh. Each state could survive for more than 1500 h/U_w, and the flow statistics were different.