[Ria]

Dear Foamers,

I am trying to change simpleFoam in order to use cyclic boundary condition for a simulate fully developed turbulent flow in a channel. To do it I mainly used:

http://www.cfd-online.com/Forums/ope...implefoam.html


My problem is that I do not understand why I get different velocity values along the cyclic patch (as showed in the picture) also if I give uniform velocity as initial condition.


Thanks a lot,
Ilaria

[tom ato]

Hej Ria,

this is the example I promised in my message. A more detailled description of BC etc. you will find there. As you can see, the values of Uy and Uz are VERY small (especially if you compare it with the values of Ux), so you can neglect them as numerical errors.

Kind regards

Tom


p.s. the posted images were removed, since I don't want people I don't know and I don't trust to use them. These examples were supposed to help Ria, but the strange entry of "dichvusuachua" make me feel, that unreliable people are entering this discussion without the intention to help.

@Ria: if you want to have a look at the exapmle again, please give me an serious e-mail address e.g. the mail account of your university or your company. I will send it to you by mail.

[samiam1000]

Hi Tom Ato (great nick!),

pardon the question, but..
..how can it be considered a numerical error?

I would agree if there was a random behavior, but I can clearly see a symmetrical behavior. What does this mean, to you?

Also, if Ria would like to simulate an infinite cylinder (and not a 2D one) both in x and z, I guess that the BC are correct, aren't them?

[tom ato]

Hi Samuele,

yes indeed, if Ria wants to compute an infinite channel both in x- and z-direction, cyclic BC in both directions have to be set.

A comment to the quality of numerical errors: They don't have to show random behavoiur, since interpolation errors are numerical errors as well. By using the same BC at top and bottom, the only difference between the treatment of both patches is the direction of the normal vectors (this why the sign of the error is changing). Since the wireframe of the posted computation is fined to the walls in a symmetric way, the quality of the error is symmetric. The slight difference in the max. values is the numerical error due to round off errors.
From a physical point of view, you can see, that the max. values of Uy is very close to the wall, so it's in the viscous sublayer of the boundary layer. But boundary-layer theory makes the assumption, that the wall-normal velocity is supposed to be 0. So you would expect from your computation, that Uy is zero in this section of the flow. Do you agree?

Anyway, I can't tell, if this is Ria's problem (and I don't think that this is her problem, because if it would be her problem, she would not have a problem at all...) since she posted no color legend.

Kind Regards

Tom

[samiam1000]

Hi Tom,

I agree whit what you are saying: it seems to be reasonable.

Thanks a lot,
Sam