Hello cfx-users

Is it possible to generate a atmospheric boundary layer in a wind tunnel in CFX? I would like to generate an atmospheric boundary layer with a atmospheric roughness height of 0.05m and a friction velocity Uf=0.06m/s. The atmospheric velocity distribution I what to generate is defined as

U=(Uf/k)ln(z/z0) (1)

where U is the velocity tangent to the terrain, Uf is the friction velocity, z is the height over the terrain, z0 is the atmospheric roughness height, k is karmans constant=0.41 and ln is the natural logarithm.

When I fit the velocity distribution i (1) to the equations used to control the flow near the wall in CFX (Eq 400 page 101 in theory manual) I get an equivivalent sand grain roughness equal 1.17m. I have assumed that the friction velocity is the same in both equations.

This means that first cells off the wall in the grid has to have a height of 1.17m*2=2.34m. This is a too coarse grid to get accurate modelling of the flow near the wall.

What can I do?

Regards Morten

Morten

In a previous life this is a problem we had a lot of the time - how to generate a good ATM BL. As you said you don't want that first cell to be too large as it then compramises the fidelity of the model, however if you then drop the size of the first cell this then can impact on your BL profile.

We were quite luck as we were often looking at flow over urban terrain with the building of interest often being surrounded by other obstructing buildings. So we would set up a mesh of coarse resolution up to these first outer buildings, then refine the mesh from there on in. Not ideal but a way round the problem.

If your object of interest is sat by itself then that may be more tricky. I would set up a pseudo 2D model (quick and easy to run) of an empty duct (eg wind tunnel) and test different ideas. If you used a coarse mesh that then when closer to the object of interest became more refined, what would be the impact to the velocity / turbulence profiles ?

Hope this helps? another option is to call the CFX guys, as they have some very switched on guys. We used to consult a lot with them on issues like this. Usually they have experience or some one who has had this before.

Stuart

Hi,

This doesn't answer your question regarding roughness height, but you will have additional problems if you are trying to get a decent ABL profile. First, the log-law of the ABL is not expressed in the 'traditional' form (specifically, y+ or z+ in your case includes the roughness length) that most wall functions are based on. This requires some rather tricky manipulation of the wall functions; I've seen a couple of papers on this but not implemented it myself.

Second, the turbulence model will have an effect, specifically linear eddy-viscosity models used comonly in industrial environments(k-epsilon, k-omega) will not represent the stress anisotropy at all.

Have you considered these issues, and, if so, come to a solution?

Oli

Ks=27*Z0 (Ks it is a roughness height for fully rough wall)

Z0=0.05 => Ks=1.35 hm-m... yo Ks=1.17 ?

But first cells off the wall in the grid has to have a height of 2*Z0 not 2*Ks, so 0.05*2=0.1