Learning Forum

[scabo]

Hi

I am solving flow in a semi-circular pipe with slip BC at the top wall. When i am refining the mesh beyond a certain point the solution is diverging. I am using RSM BSL model and inflation layers with 3um as first layer thickness. I am also using pseudo transient, coupled solver with pressure, mom as 2nd order but turbulence 1st order. What could be the issue? Do i need to reduce the time scale factor or under-relaxation? etc,, any help is appreciated

thanks

[Rob]

The cell height may be an issue in that eventually some of the terms will tend to zero or gradients to infinity. The RSM model tends to be less stable than k-e and will nearly always find transient flow features. 

Does having a 3micron cell make sense physically? Ie how good is the surface in the experiment to require that level of resolution? How is the flow field looking before the last few levels of mesh refinement? 

[scabo]

Hi,

i have used 3um for a case and it performed well and the results looked good. But when i am refining the mesh even further the solution is divergng and the y+ is shooting up from 0.1 to 400. So i reduced the first layer height in the next run and it converge. But this problem is happening when i am using the mass flow rate specification in a periodic setup but not the pressure gradient specification. It seems its a bit problematic. Any suggestions pleas?

[Rob]

If the pressure method is working what mass flow are you seeing for the estimated pressure drop? 

[scabo]

Yes so i am seeing a sensible mass flow 0.598, but the exact should be around .6104. and the results look good.

[Rob]

And if you switch to pressure option now the flow is established? Do NOT initialise, just change the bc and iterate. 

[scabo]

I have not done that, but in some cases i only know the mass flow rate not the pressure gradient. But for this specific case i can do it as u say. But why the y+ shoots up to 400 when i refine the mesh? i am keeping the first layer height same in all cases. is it something to do with RSM?

[Rob]

If the solution is beginning to diverge chances are the solution is giving a nonphysical velocity: check the flow as well as looking at y+

Assuming you have a rough idea of the mass flow you're aiming to improve the initial condition rather than estimating precisely. 

[scabo]

Yes the velocity is badly diverging 1e20 and that is because of inflation settings with RSM. But why do u think the pressure gradient is working but not the mass flow rate

[Rob]

The two methods are using slightly different maths, and given the limitations https://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v242/en/flu_ug/flu_ug_sec_periodic.html it's possible that pressure gradient is more stable in this case. Without seeing any flow results I can't comment beyond that. 

[scabo]

Thanks- do u think the way i am initialising is making any difference? I am doing standard init. I can do hybrid or fmg? Will it make any difference? Also i have attached a picture of the vel contour. All the other variables are also coming very high. thanks

[scabo]

Also why with 3um first layer height, the previous mesh was working but the new more refined mesh is failing both with mass flow rate?

[Rob]

I'd generally favour Hybrid for most applications unless you have an existing interpolation file. The above has already failed, but I'd also check the cell size adjacent to the inflation looking at the shading.