Very high divergence at outlet boundary

[stek...@gmail.com]

Hello everyone,

I have a cylindrical outlet boundary next to two rotating walls and I get very high divergence at the outlet, next to the walls (see attached screenshot). The region of high divergence also radiates back into the domain for a short distance. There is no backflow, the fluid flows towards the outlet everywhere.

This causes a WARNING: DIV(V)-QTL too large!. In the rest of the domain, the flow is laminar with very low divergence.

Does someone know why that happens or how to prevent this? Somehow the rotating wall seems to be incompatible with the outlet boundary. Usually you ignore the outlet region anyway but it would be nice to know an explanation.

Thanks!

[Emmanuel Gillyns]

Hi,

I'm not sure I understand your case properly, but a high divengence is a common problem near the outflow boundary.

A solution that worked well for me is to locally accelerate the flow at this boundary : just add the following lines in userchk :

real d(lx1,ly1,lz1,lelt),m1(lx1*ly1*lz1,lelt), rq,uin

rq = 1.5 !flowrate ratio : uout/uin (typically 1.5)

uin = 0. !will compute the exact flowrate if =0.

call turb_outflow(d,m1,rq,uin)

Hoping this helps you,

Emmanuel

[stek...@gmail.com]

Hi Emmanuel, Thanks for your help, that sounds exactly like what I'm looking for, however I tried it with different values for rq (1...10) and it made the problem worse.

As far as I understand, the purpose of turb_outflow is mainly to prevent turbulent flow from locally recirculating back into the domain, but in my case the flow isn't even turbulent and there is no backflow, yet there is still high divergence at the outlet. My theory is that the boundary conditions are not compatible. The moving wall is implemented via a 'v  ' boundary with imposed velocity parallel to the outlet surface. This seems to conflict with the no stress 'O  ' bounday (there actually is stress due to the moving wall). Maybe its possible to use different types of boundary conditions, but at the moment I don't know which ones.

Thanks,

Stefan