[king lui]

Hello CFX Users,

I am doing some isothermal simulation to investigate the pressure loss in a cooling device. I am using the SST turbulence model with automatic wall treatment. My problem is now that for smaller flow rates I got converged results, but as I am doing a paramtric study, I could not achieve convergence at higher flow rates. So I generated another grid and the solution converged. However, as I have some results from literature it could be seen that for the first grid the results are quite realistic whereas by the second one the pressure loss is overestimated. In theory the automatic wall treatment should be relatively mesh insensitive. But I am sure the difference comes from the wall treatment. Does anybody know how CFX does calculate the wall shear stress?
By the way, my y+ values are for both grids smaller than one.

[ghorrocks]

Why do you say the difference is from the wall treatment? Is that a guess or do you have evidence?

I suspect your second mesh is coarser, at least in an area which matters. This will lead to easier convergence (as coarser grids are more dissipative and therefore easier to converge) but less accurate.

So I would look at the region which the second mesh coarsened and try to refien mesh there but improve mesh quality. That is you best bet to get convergence on a grid with adequate resolution to be accurate.

[king lui]

Hey Ghorrocks,

thanks for your reply.
You are right, my second mesh is coarser. Maybe I was too much focused on the first inflation layer in my post-processing. I observed that my wall shear stress and shear strain rate are much higher for the second grid with smaller first layer thickness. I thought this leads to the overestimation of the pressure loss. Probably my y+ in the second mesh is too small. Do you have any experience about too small y+ values?

For both grids the number of inflation layers is equal, however as the second one is more fine, probably I can´t capture the entire boundary layer inside the inflation layers. But does this lead to the far higher wall shear stress?

Thank you in advance

[ghorrocks]

Yes, you can have problems when y+ gets too small. Often it is because of numerical round off issues. Double precision can help here.

It can lead to the problems you are seeing, but I cannot guarantee that is the cause.

[king lui]

Alright, I did all simulations with double precision but I will have a look on that if this is the problem. Many thanks for your reply.

[ghorrocks]

You can still have problems with round off with double precision numerics - it just requires an even finer mesh.