Difference in velocity prediction by kepsilon and komega
I have simulated a single phase water flow with both kepsilon and komega models in CFX/12.1.
However, there is significant difference in velocity predicted by both the models.
k-epsilon model's prediction seems to be incorrect as the boundary layer development seems incorrect as compared to komega model.
Please refer attached velocity magnitude plots and velocity vectors for both the models.In the figures, fluid inlets from the left side at a Re of around 600. Velocity magnitude plots also show the mesh used. I have used a fine uniform mesh. However, I have not ensured any particular y+ at the wall due to use of scalable wall function for k-epsilon model and automatic wall treatment for k-omega model.
Any idea why kepsilon model prediction is so different and seemingly incorrect?
Although inlet Re<1000 calls for use of laminar model as per CFX manual, laminar model does not converge for this case (a possible reason being that the water flows ahead into a porous medium where Re based on pore size is >1 and thus inertia is important in that part of the computational domain which the laminar model is not able to handle).
Thanks for your inputs!
Glen already answered this in your other thread. k-epsilon is inappropriate for laminar flows.
Thanks for your reply.
Yes, Glen had replied to a similar query from me previously. But I did not expect there will be so much difference in the results. The boundary layer really looks weird.
Do you think that any mesh change can improve the results?
Actually my problem is that I have to simulate conjugate heat transfer in this geometry for various flow rates and water inlet temperatures. The results I show here are for the lowest Re where both k-epsilon and komega models converge. I also tried SST and it also converges giving results nearly similar to komega.
Hoevere, at the highest flow rate and highest water inlet temoperature (where inlet Re is the highest and ~4000), only kepsilon converges and komega and SST do not converge. Though I have not shown it here, for kepsilon for this high Re also, the boundary layer development does not seem to be correct and is similar to the low Re case.
In essence, kepsilon converges at both the extremes of inlet Re for my simulations but seems to give incorrect results at both extremes !
As I said in the other thread, forget about ANY turbulence models for these low Re flows. You have a problem with convergence on your laminar model so fix that.
|All times are GMT -4. The time now is 17:22.|