CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   problem from k epsilon to k omega (http://www.cfd-online.com/Forums/openfoam-solving/116213-problem-k-epsilon-k-omega.html)

 NJG April 15, 2013 08:24

problem from k epsilon to k omega

4 Attachment(s)
Foamers,

I have been running a modified convective heat transfer solver (buoyanrBoussinesqPimpleFoam with the addition of CO2 mass fraction) to model a person breathing out in a closed room. Basically the person-like pillar breathes out hot CO2 rich air at a constant rate and his "ears" are constant pressure BCs that allow a mass balance in the room.

I started with a k epsilon based approach (using k and epsilon values at the inlet I borrowed from another model without much thought, just to kick the tires a bit, which where 0.015 in both cases) and I was getting good results consistent with previous work we have done on this. Note the two k epsilon attached photos. "Profile" shows the person profile and you can see the breathing out effect and "backview" is a view from the back so you can see the "ears" doing their mass balance thing.

This was all good, but over time the k epsilon scheme showed instability at the "ears". My k and epsilon values would sometimes suddenly exploded to huge values, crashing the model.

So, I switch to k omega to try and fix this, and so far I have had no instability issues. BUT, the behavior of my plume is much different. Here I actually calculated a k and omega based on the breath out velocity of 0.06 m/s, turbulent intensity of 10%, and the length scale of the "mouth" which is 0.04 m (so I have k 5.55E-5 and omega 0.186).

You can see the results of this in my k omega pictures attached below. Note how much more diffuse the plume is. Also, note how my back view shows a disturbing lack of symmetry on the effects of my ears.

So now my questions:

1) It seems to me, if all I changed was from k epsilon to k omega, through the selection of better values of k and omega, I should be able to produce the exact same results. Is this not correct?

2) Why would I be getting this non-symmetric behavior if I have a fully symmetric domain and BCs and had fully symmetric result in a k epsilon scheme?

3) If anyone has any tips on how to perhaps just make k epsilon more stable that approach might be better as k omega is not giving me good results.

4) Any other insight into this that anyone has would be most appreciated.

-NG

 All times are GMT -4. The time now is 12:59.