I'm looking at the design of a contraction in front of a wind-tunnel. The contraction has an inlet/outlet area ratio of about 6 and hence gives a strong accelaration of the flow. This strong accelaration leads to a reducion of stream-wise turbulent fluctuations and an amplification of corss-stream-wise turbulent fluctuations. There are correlations developed by Bradshaw to estimate this effect.

I would like to do a CFD computation including the contraction and I'm a bit hesitant if the turbulent decay/amplification in this type of strongly accelarated flows can be captured with a RANS based turbulence models. Obviously eddy-viscosity models like k-epsilon or k-omega can not capture the anisotropic effects. But do this type of models have any chance to capture the total effect on turbulence intensity (I doubt it myself)? Will RSM model it correctly? Or should I skip all RANS based model?

Hi,

In fact your looking at turbulence subject to strain and RSM should be able to model that, but perhaps not all models are very accurate. The article of Sjogren & Johansson (J. Fluid Mech. 374, pp. 59) is perhaps interesting for you. In your case you have a quite strong contraction, so the effect of strain is possibly well predicted by rapid distortion theory (RDT), see Pope (2000) Turbulent flows, or Townsend (1976) Turbulent shear flows or something like that.

Tom

Thanks for the references Tom, I'll check them asap.

I'm also leaning towards using RSM here. We have an RSM simulation running now and it will be interesting to compare the results with Bradshaws' correlations for wind-tunnel contractions (k-epsilon fails quite badly).

Jonas.

I don't know if this helps, but it seems to me that the production term due to streamwise irrotational straining is -u2dU/dx and this will be negative in strongly accelerated flow. However, an eddy-viscosity closure will always produce a positive contribution, and so a RSM will at least get the sign right.