Hello everyone,

I am trying to solve the distribution of gas in a rectangular vessel. Inlet and outlet of gas is relatively small compared with cross-sectional area of vessel. So gas flow is stable when gas velocity is small. But if gas velocity becomes high, gas flow will be unstable. And now I am trying to solve at high gas velocity. The residual does not converge and also does not diverge. I mean the plot of residual is jagging. Like this "/\/\/\/\/\/\/\". So the residual is almost flat.

Does anyone tell me about what this behavior of residual is meaning?

I do not use Engulish very much. Sorry for that.

Thanks,

AKR

Hi AKR,

ther is a VERY large discussion about oscillating residuals:

http://www.cfd-online.com/Forum/flue...cgi?read=34904

Hope, it helps

Ralf

akr, There are many issues but the simple thing if you are result oriented, check if the convergence achieeved is acceptable then the zig zag nature is not so significant

This basicaly a result of bad mesh (may be few cells)

or 1) try n plot th egas velocity if it is in compressible range 2) what are your urfs? 3) what is the descritisation scheme you are trying

Thanks, Ralf and sach.

I will check the discussion, Ralf.

sach, I could not get some words in your sentence. "check if the convergence achieeved is acceptable" is meaning that comparing with experimental result? What is "n plot" and "urfs"?

I am trying to solve the problem with laminar model. As I said, the apparatus which I am trying to model is rectangular vessel with relatively small inlet and outlet. The size of vessel is 0.3(length) by 0.258(height) by 0.1(width). Inlet and outlet is 0.08(width) by 0.01(height). I calculated reynolds number of gas. It is 2790 based on the cross-sectional area of inlet. So I think it is OK if I use steady state laminar model. There is no energy equation in the model. The fluid is only gas and incompressible.

The grid is 3D and meshed by using Tgrid. Actually, I wanted to mesh the geometry by using quadrilateral or hexahedral grid. But it was not possible, because there is some structure in the rectangular vessel. I used first-order upwind as descretization. It is default setting.

This is what I tell you so far. If you have any question, please let me know.

Thanks,

AKR

I mean if the convergence achieved, i.e. differance in mass flow rate inlet and outlet is less than e-3 I suppose this is comfirtable convergence you have and no need to get more

urfs means under relaxation factors you can reduce thenm to get faster convergence (in terms of residuals) I would suggest if you have low reynold no case better go for sst k omega model with transition regime on

Thank you for your help, sach.

I was setting urfs default values. I will try the reduced urfs. And also, sst k omega model.

Thanks,

AKR

If you are builiding it from scrach check the Y +, but for sst k omega the range is 5 to 10 and mesh adaption is also automaticaly done.