Periodic BC length problem

[scabo]

Hi

I am using a periodic bc along the length of a 3D pipe which is semi-circular cross-section. I have no slip BC at all walls and a slip BC at the top surface. I saw that the results are vastly dependent upon the periodic length taken along the streamwise direction. If i take a bigger length the results are coming much better than if i take only one-cell or a very short length along the streamwise direction. Why is this? The results should be independent upon the streamwise length? Can someone please clarify, thanks. I am using 2nd order for mom, pressure and first order for Reynolds stresses, coupled and least square cell based.

[Rob]

What periodic settings are you using? Is anything else switched on?

[scabo]

So i am using pressure gradient as periodic BC in x direction only. It is a steady state with Pseudo time stepping switched on so that it converges properly. In the shorter/one-cell case, spurious results are coming, like asymetric contorus and streamlines , etc,

[Rob]

One cell? That may explain any odd behaviour. The option is intended to allow a shorter section to be modelled, but not that short. 

[scabo]

Yes i tried with both one-cell and a short length of 0.05m. In both cases results are coming bad but worse in one-cell. But when i increase to 0.5m then it matches with literature and results coming meaningful. My cell length is 5e-4 m. But there is a study that has used one-cell using periodic bc in other softwares so i thought it should work so that i can reduce cost.

[Rob]

It'll only work if the in-out effect doesn't set up some instabilities. I generally aim for a few diameters to get a meaningful size for post processing. 

[scabo]

What is meant by in-out effect? Is it related to the periodic bc?

 

[Rob]

Yes - if the periodic section is too short the solver may not have space to do much with the flow. 

[scabo]

 

The solver may not have much space:But doing periodic BC means simulating a fully developed section at an infinte distance. Does the solver need enough space for that though? It is just adding a BC on 2 surfaces. Isn't it? Can you explain a bit? Also, if i do second order upwing for momentum/pressure, if there is very less cells in streamwise dir, then it cannot find upstream cells to interpolate values from. Is this also a correct explanation of unphysical results?

 

[Rob]

Model a few diameters and, as you've said, it's working. 

[scabo]

Yes, so it can vary from software to software then-the limitations of different softwares.

[scabo]

Hi. I am coming back to this point after reading some literature. using a one-cell or very short lenght while using periodic conditions in Fluent creates bad results. Why exactly is this? Is this because the velocity/pressure gradients are not well resolved for a very short length? Another paper using OpenFOAM have successfully used a one-cell thick while doing periodic conditions. Can you please elaborate on this? thanks

[Rob]

No. But as you'll have noted from your literature review papers are not always to be trusted.