[mauricio]

Hello, i was wondering if any of you knew any resources, papers, bibliography of some guidlines into setting inflation layers for walls. I'm running a sensibility analisis on a Tet mesh for slug flow, created on ICEM Cfd, and i was wondering on what should i base how much layers of inflation i choose, how much total thickness, etc.

Since it's a 8m long pipe and i'm not running this in a cluster I can't go for y+=1 but more than that i'm interested on what to base how much layers of inflation i use or else i just have to pick 3 or 4 and just run the sensibility analysis on the meshes that come from there changing general seed size on tets, which is what i've been doing so far.
Thanks in advance.

[lindner]

If you can´t have y+=<1 then you have to choose a k-epsilon model witch accepts 30 < y+ < 300. The first layer you have to adjust after a few iterations checking the y+ value.
The number of layers you can regulate to have a smooth transition between the last inflation and first cell after that.
Try not to exceed 1.25 expansion ratio in the inflation.
This is the general rule for mesh, but every case is a different case.

[mauricio]

thanks lindner,
I'm using K-w for the turbulence, does that work well with y+>30? Does anyone know if ICEM cfd has any tools regarding y+?

edit1: oK I checked this http://www.cfd-online.com/Wiki/Y_plus_wall_distance_estimation
Now i'd have to ask, how do you decide on a desired boundary layer length?

Edit2: Great as this says: http://www.cfd-online.com/Wiki/Near-...k-omega_models
I don't have to go for y+=1 since my Re goes beyond transitional... i still wonder about the boundary length though. Also since there are two fluids, which one would you use in properties for the layer calculation? the water or the air? or the mixture? I'm guessing that if i'm using one instead of the mixture i'd have to set the superficial velocity as the "freestream velocity" there....?

[ghorrocks]

If you can´t have y+=<1 then you have to choose a k-epsilon model

Rubbish. Many other turbulence models work fine at y+>1. The SST model is a key example, it is the turbulence model to use in most cases and can handle any reasonable value of y+.

I'm using K-w for the turbulence, does that work well with y+>30?

No it does not. It requires integration to the wall and therefore y+ of no more than about 1.

lso since there are two fluids, which one would you use in properties for the layer calculation? the water or the air?

What multiphase model are you using? Free surface? Lagrangian? Eularian?

Unless you have a good reason to do otherwise I strongly recommend you use the SST turbulence model. It can handle any sensible value of y+, can automatically transition from integration to the wall to wall functions, and is easily extended to include curvature effects, transition and several others if required.

And do not be obsessed with getting y+=1. Wall functions, with y+>20 or so, when used correctly, are just as accurate and often more accurate than integrating to the wall. If you do not need the detailed boundary layer profile, just need the correct wall shear stress I recommend using a y+>20 mesh which then runs using wall functions.

[mauricio]

Thanks glenn. I'm using eulerian-eulerian with the mixture model for interphase drag.

[ghorrocks]

OK thanks. Be aware that turbulence and near wall modelling for multiphase flows is not a mature science - you should expect larger errors compared to single phase models.

Oh yes, and to address your original question - a good source of information about near wall meshing is the CFX documentation. I think it is in the Solver Modelling guide, the section "Advice on Flow Modelling".

[mauricio]

Thanks. Oh yeah another thing, in the calculation of the first layer thickness for y+, what should i set as the boundary layer length?

[ghorrocks]

The CFX documentation has another way of estimating the y+ value which does not need the boundary layer length. Have a look for it.