No abstract

The turbulent structure in plane Couette flow at low Reynolds numbers is studied using data obtained both from numerical simulation and physical experiments. It is shown that the near-wall turbulence structure is quite similar to what has earlier been found in plane Poiseuille flow; however, there are also some large differences especially regarding Reynolds stress production. The commonly held view that the maximum in Reynolds stress close to the wall in Poiseuille and boundary layer flows is due to the turbulence-generating events must be modified as plane Couette flow does not exhibit such a maximum, although the near-wall coherent structures are quite similar. For two-dimensional mean flow, turbulence production occurs only for the streamwise fluctuations, and the present study shows the importance of the pressure-strain redistribution in connection with the near-wall coherent events.

System rotation is known to substantially affect the mean flow pattern as well as the turbulence structure in rotating channel flows. In a numerical study of plane Couette flow rotating slowly about an axis aligned with the mean vorticity, Bech & Andersson (1996a) found that the turbulence level was damped in the presence of anticyclonic system rotation, in spite of the occurrence of longitudinal counter-rotating roll cells. Moreover, the turbulence anisotropy was practically unaffected by the weak rotation, for which the rotation number Ro, defined as the ratio of twice the imposed angular vorticity Ω to the shear rate of the corresponding laminar flow, was ±0.01. The aim of the present paper is to explore the effects of stronger anticyclonic system rotation on directly simulated turbulent plane Couette flow. Turbulence statistics like energy, enstrophy and Taylor lengthscales, both componental and directional, were computed from the statistically steady flow fields and supplemented by structural information obtained by conditional sampling.The designation of the imposed system rotation as ‘high’ was associated with a reversal of the conventional Reynolds stress anisotropy so that the velocity fluctuations perpendicular to the wall exceeded those in the streamwise direction. It was observed that the anisotropy reversal was accompanied by an appreciable region of the mean velocity profile with slope ∼2Ω, i.e. the absolute mean vorticity tended to zero. It is particularly noteworthy that these characteristic features were shared by two fundamentally different flow regimes. First, the two-dimensional roll cell pattern already observed at Ro=0.01 became more regular and energetic at Ro=0.10 and 0.20, whereas the turbulence level was reduced by about 50%. Then, when Ro was further increased to 0.50, a disordering of the predominant roll cell pattern set in during a transient period until the flow field settled at a new statistically steady state substantially less affected by the roll cells. This was accompanied by a substantial amplification of the streamwise turbulent vorticity and an anomalous variation of the mean turbulent kinetic energy which peaked in the middle of the channel rather than near the walls. While the predominant flow structures of the non-rotating flow were longitudinal streaks, system rotation generated streamwise vortices, either ordered secondary flow or quasi-streamwise vortices. Eventually, at Ro=1.0, the turbulent fluctuations were completely suppressed and the flow field relaminarized.

As in the laminar case, the turbulent plane Couette flow is unstable (stable) with respect to roll cell instabilities when the weak background angular velocity Ωk is antiparallel (parallel) to the spanwise mean flow vorticity (-dU/dy)k. The critical value of the rotation number Ro, based on 2Ω and dU/dy of the corresponding laminar flow, was estimated as 0.0002 at a low Reynolds number with fully developed turbulence. Direct numerical simulations were performed for Ro = ±0.01 and compared with earlier results for non-rotating Couette flow. At the low rotation rates considered, both senses of rotation damped the turbulence and the number of near-wall turbulence-generating events was reduced. The destabilized flow was more energetic, but less three-dimensional, than the non-rotating flow. In the destabilized case, the two-dimensional roll cells extracted a comparable amount of kinetic energy from the mean flow as did the turbulence, thereby decreasing the turbulent kinetic energy. The turbulence anisotropy was practically unaffected by weak spanwise rotation, while the secondary flow was highly anisotropic due to its inability to contract and expand in the streamwise direction.

No abstract