A-Priori Tests of Reynolds Stress Transport Models in Turbulent Pipe Expansion Flow

Abstract

The database that is obtained from a direct numerical simulation (DNS) of turbulent sudden pipe expansion flow with an expansion ratio of ER=1.2 has been used to analyze various Reynolds stress models. The inflow conditions were provided by a DNS of fully developed turbulent pipe flow. The Reynolds number at the inlet corresponds to that in the pipe and is 360 based on diameter and friction velocity. Various models for pressure strain correlations, turbulent transport, and viscous dissipation were evaluated using the DNS database. It was found that none of the investigated redistribution models is able to predict the extraction of fluctuating kinetic energy from the wall normal velocity component within the reattaching flow at least qualitatively correct. Improved models are needed to predict the turbulent transport of the wall normal stress component for separated flows accurately. The derivation of most of the so-called single-point turbulence models is based on the assumption that the turbulent flow is in local equilibrium. The time-dependent velocity vector in the inflow plane was obtained from the DNS of fully developed pipe flow during the entire simulation process. Radial velocity fluctuations can achieve higher values in a free-shear layer being unaffected by any kinematic constraint. An increased Reynolds stress in turn produces higher axial velocity fluctuations. In more complex flow regimes, separate modeling of turbulent diffusion and pressure diffusion seems necessary. A key to more satisfactory predictions could be their improved modeling.