Benchmarking axi-symmetric pipe flow with sudden expansion
|
I am solving for flow in an axi-symmetric sudden expansion. Essentially I have a small pipe attached to a larger pipe. I am mainly interested in re-attachment points over a range of Reynolds numbers. I need best practice guidelines for which model to use in what Reynolds number Regime. 1. I am doing a series of runs - Re 200, 500, 800, 1000, 2000, 3000, 4000 and 5000. 2. I am using a 2d axisymmetric steady model - all calculations are well converged. 3. I am comparing my results with two papers - Back and Roschke, Journal of Applied Mechanics, 1972 and Furuchi et al, Experiments in Fluids, 2003. For the turbulent regime, I am using a kw-SST model and kw-SST with low RE corrections. For Re > 3000, I find that kw-SST gives good comparisons, For Re 2000 to 3000, kw-SST with low RE correction gives good comparison of reattachment lengths. In the laminar regime, Re < 300 gives good comparison. BUT, in the laminar regime 300 < Re < ~1000, my reattachment lengths are way off (200-300% off) what it should be according to the experimental papers. Fluent seems to be very much over-predicting the reattachment regions. I think that there is an inherent unsteadiness in the flow in this range of Re which cannot be captured by a steady model and by an axisymmetric model. Can you comment on this. Thanks, -Sree Hello Sree, here my comments to your CR. First of all, the kw-SST turbulence model, which you decided to use, is the best choice for fully turbulent pipe flow. As guideline I suggest to use kw-SST for the fully turbulent cases and no turbulence model at all in the laminar regime. In the laminar regime no turbulence needs to be modeled. In general pipeflows are turbulent above approx. Re>2300 and laminar for approx. Re<1000. In the region between transition occurs and a transition model has to be used. In your case the soft bounds between turbulent, laminar and trasition region differ because of the sudden expansion where turbulence is induced. The kw-SST model offers such a transition switch. But, be careful with this transition model. The correlations of our transition model were validated on free boundary flows (e.g. flow over a plate). Your setup is not such a flow condition. So, use the transition model having this fact in mind. Low Re number correction has basically influence onto the near the wall treatment. The sharp peak in turbulent kinetic energy near the wall is therewith resolved. I think resolving the sharp peak is in your application not so important, as in your case the separation point is given by the corner of the sudden expansion. In other cases, e.g. a diffusor, the turbulent kinetic energy near the wall may influence the location of the flow separation point. Comments regarding your assumption of transient effects in the transition regime while using steady simulation: It is possible and likely that transient effects exist in transition. An indicator for transient effects in a steady run is the convergence history. A slow convergence which even may oscillate from time to time in a steady run may indicate such a transient effect. The solution switches thereby between different transient solutions. For further information regarding Transition please have a look into the powerpoint library at TeamAnsys Sales Portal: there exist a ppt file with audio support: Transition Model - Technical Overview with Audio Support Hope this helps Best regards, Gerhard |
||
|