The main objective of this thesis is to observe numerically and to analyze the effects of intermittency in a high Reynolds number turbulent channel flow. To this end, the thesis is focused on characterization and stochastic modelling of the fluid acceleration in such a flow, with emphasis on long-range interactions across the channel. In the first part, the acceleration is studied using DNS for three Reynolds numbers (180, 590 et1000). It is observed that the norm of acceleration is log-normal whatever the wall distance is. The universal form of scaling law for the acceleration is proposed by dimensional analysis. In the second part, the acceleration is simulated stochastically, assuming the norm/orientation decomposition. The stochastic model for the norm is based on the fragmentation process in order to represent the long-range interactions across the channel. The orientation is simulated by random walk on a sphere in order to reproduce the relaxation towards isotropy with increasing the wall distance. This was observed preliminary in our DNS. These models were applied in the framework of LES-SSAM approach (Stochastic Subgrid AccelerationModel), which was introduced by Sabel’nikov, Chtab and Gorokhovski and assessed in the case of the box turbulence. Our computations showed that the mean velocity, the energy spectra, the viscous and turbulent stresses, as well as the non-gaussianity of acceleration statistics can be considerably improved in comparison with standard LES. The advantage of the LES-SSAM approach, which accounts for intermittency on subgrid scales, is demonstrated in the last part of this thesis. Here the transport of inertial point wise particles was studied by DNS and by LES-SSAM. The influence of wall structures on the particle’s dynamics is analyzed.
