[kingjewel1]

Hi there,

I'm hoping you can put me straight here:

I'm investigating particle deposition onto indoor surfaces. Therefore Reynolds numbers are generally less than 10,000. So I'm using a wall function which requires y+ in the first cell to be about 30. Mine are, however, often about 1.

Therefore this gives me incorrect results. However I need sufficient grid resolution for accurate particle deposition.

What are your recommendations?

Regards,

[calim_cfd]

Originally Posted by kingjewel1

Hi there,

I'm hoping you can put me straight here:

I'm investigating particle deposition onto indoor surfaces. Therefore Reynolds numbers are generally less than 10,000. So I'm using a wall function which requires y+ in the first cell to be about 30. Mine are, however, often about 1.

Therefore this gives me incorrect results. However I need sufficient grid resolution for accurate particle deposition.

What are your recommendations?

Regards,

i'd say turn off your wall functions. y+ = 1 puts you in the in the viscous sub-layer. I guess basically you'll be solving the boundary layers instead of modeling. Then i guess you'll need at least ~15-20 elements along the BL's height with a growth rate of 1.1-1.3.

i haven't rly done any turbulent simulations in openfoam w/o wall functions. Some say that you may need y+ as small as 0.1 to resolve the whole thing properly but i guess that depends on the turbulence modeling/intensity

edit:
guess that you'll have to use low-re models to resolve the BL... try LaunderSharmaKE ras incomp/comp model or the new transition incompressible one kklOmega

set zero gradient on walls and respective values at inlet and outlet!? can any1 with xp in this topic back me up? or maybe correct me?

sry kingj, but i cant help you any better

hope it helps somehow

[kingjewel1]

Originally Posted by calim_cfd

i'd say turn off your wall functions. y+ = 1 puts you in the in the viscous sub-layer. I guess basically you'll be solving the boundary layers instead of modeling. Then i guess you'll need at least ~15-20 elements along the BL's height with a growth rate of 1.1-1.3.

i haven't rly done any turbulent simulations in openfoam w/o wall functions. Some say that you may need y+ as small as 0.1 to resolve the whole thing properly but i guess that depends on the turbulence modeling/intensity

edit:
guess that you'll have to use low-re models to resolve the BL... try LaunderSharmaKE ras incomp/comp model or the new transition incompressible one kklOmega

set zero gradient on walls and respective values at inlet and outlet!? can any1 with xp in this topic back me up? or maybe correct me?

sry kingj, but i cant help you any better

hope it helps somehow

This is what I am coming to believe too. But what is odd though, that there are papers out there that seem to get good experimental/numerical comparison without using low-re models. I just don't see how they're doing it with wall functions. Clarification would be greatly appreciated.
Cheers,

[calim_cfd]

Originally Posted by kingjewel1

.. But what is odd though, that there are papers out there that seem to get good experimental/numerical comparison without using low-re models.

i guess that depends on the flow?

Originally Posted by kingjewel1

I just don't see how they're doing it with wall functions.

try setting usual turbulent setting on patches and zeroGrad on walls and c what you get!

if i have more time i'll try checking some references cuz i'm interested in that 2!!!!!

sry to be of no help

[calim_cfd]

ok i've looked a bit more into the topic and i confess this is one of the most puzzling topics in cfd for me...i'm still reading Pope's book of turbulence for better modeling theory but atm this is what comes to me from my current knowledge

let's rate:

1:we all know the scale issues related to HiRe lowRe numbers and bounded flows

2:you either have turbulence or you don't!

3:if you do, you have to choose between low or high Re turbulence modeling (lets stick with RAS)

4[edited]: removed

5[edited]: wallfunctions... kinda routine settings for hiRe.. BUT, to resolve a bl you need a lowRe Ras model: have a refined mesh with low y+=1, use calculated type for nut/mut (few say zeroGrad wont change much), zeroGrad for k/epis or very small values like 1e-10 on walls

==>>zeroGrad for k/epis or very small values like 1e-10 on walls: i guess that in the limit they should yield similar results? no? == any comments/corrections to this are most welcome!!! ==


6[edited]: removed

check this pdf,
http://www.google.com.br/url?sa=t&rc...Py7KIzWAuhcT1w

maybe it will point sth you may have missed

guess that's all i can try to help you with!


EDIT: sry about the editing but i corrected some assertions made previously