Turbulent boundary layer, a natural phenomenon occurring at the vicinity of bounding surface for almost any flow condition, is important and remains a challenging engineering problem in many domains such as transportation, energy or health. Recently, the large scale coherent structures have been shown to play a significant role in the statistics of friction drag. Moreover, dynamic of such structures are essential in an attempt to control such turbulent flows. The detailed description of 3D structures in physical space are mainly accessible from large Direct Numerical Simulation (DNS). Few DNS at significant Reynolds number have already been performed. Schlatter and Orlu (2010) have shown that rigorous approach to set up simulation is substantial since different simulations (for same canonical flow case) give inconsistent measures even for basic quantities. Sillero et al. (2011) investigated large scale structures in detail and compared with channel flows. However, the statistics of very large scales can be affected by the simulation parameters.

In the present study, DNS of zero pressure gradient turbulent boundary layer up to Reθ = 2550 is conducted with the code "Incompact3d" on a simulation domain of size 600δ_inlet × 40δ_inlet × 20δ_inlet. Transition to turbulence from the inlet Blasius profile is forced at Reθ=300 with a tripping suggested by Schlatter and Orlu (2010). The advantage of this method over recycling methods is to avoid some spurious streamwise correlations that could act upon large scale turbulence statistics for not long enough simulation domain. Up to 500, 3-dimensional velocity and pressure fields have been collected within 20 characteristic time based on uτ and δ at the middle of the simulation domain. In addition, time resolved data in four planes perpendicular to the flow at Reθ = 922, 1522, 2063 and 2365 also recorded in order to compare space and time statistics. The statistics of the current DNS is in good agreement with other state of the art TBL simulations.

The aim of the present study is to investigate large scale structures by using the same methodology used with data acquired from particle image velocimetry (PIV) on a large field of view at higher Reynolds numbers (Srinath et al. 2017). The statistics of structures extracted from 2-dimensional and 3-dimensional data, later differences between those are investigated. Besides, large scale structures are extracted from both spatial and time resolved data. Low and high momentum regions are extracted by applying a threshold on streamwise velocity fluctuations and morphological treatments. Moreover, the 3-dimensional examination is strengthened with quadrant analysis of Reynolds shear stress. Aspect ratio for different directions are presented and compared to similar results obtained in channel flow by Duran et al. (2014).

This study notifies the difficulties to extract statistics of large scale motions and to propose detection procedures which allow a fair comparison of statistics of different wall turbulent flows between 3-dimensional numerical data at moderate Reynolds numbers and high Reynolds experiments data.