Learning Forum

[damis93]

Hi all,
 
I'm doing simulations in turbulence by using the Relizable k-eps model with standard wall functions. The goal of my simulation is to detect a temperature profile in my channel and compare it with my experiments.
 
My concern is if it is ok to use the wall standard functions. By looking at the web, they suggest to use the STANDARD functions when the first cell adjacent to the wall(yplus > 30). The log law is in fact applied when yplus is > 30. From the manual of Ansys, it is explained that when yplus is greater than 11.125 the log law is applied, but it is important to have a yplus greater than 30 to not have a deterioration of the log law. In the case the yplus is lower than 11.125, Ansys applies the laminar stress strain-relation (u*=y*), as said in the manual of Ansys. Now, in my case, the yplus is in the order of one. Then, according to Ansys, the laminar stress-relation is applied. My doubts are:

• The linear stress relation is applied for the first cells (if yplus is about to be 1), but what does Ansys do for the upper cells in the domain where the distance is not anymore in the viscous sublayer? (yplus > 5)


• Can the standard wall functions be used if I use an inflation layer such as my first cell has a yplus in the order of 1?

 
I know that there are also other functions (like enhanced wall functions) and also other models supposed to use a specific treatment very close to the wall, but I found  the standard wall functions to predict very well all the experiments! 

[Kalyan Goparaju]

Hello, 

Our recommendation, especially for wall bounded flow that you are talking about, is to use the K-Omega SST model. It is a y+ insensitive model and for that reason, you will not see a deterioration of results either with a refined or a coarsened grid. Standard wall functions should not be used if your yplus is of the other 1

Thanks, 

Kalyan

[DrAmine]

You should not consider the the boundary layer resolution from the yplus point of view only. Think about flows with low to moderate Reynolds Number: yplus maximum can be not larger than 300 or even lower. For High Reynolds Flow the upper limit can be as large as 5000.

Now if you put your yplus at 30 for low-RE or moderate Re case  or more there would not enough cells to resolve the boundary layer. One would require at least 10 to have a good resolution. Another consideration to take place when you work with high Pr numbers where thermal layer is very thin. 

Another aspect standard wall functions do assume zero pressure gradient and an equilibrium between production and dissipation and this almost always wrong.

Summary: SST (omega based)  model is what we recommend for almost vast range of industrial applications. 

[Karthik Remella]

Hello,

Firstly, I believe the Fluent Theory guide is referring to y* (not y+) > 11.225 in Ansys Fluent when the log-law is being applied. Also, your statement is true about the laminar stress-strain relationship for y* < 11.225.

Also, to answer your other question - in a general manner of speaking, wall functions are less reliable when the cell thickness is in the viscous sub-layer region. This is because, in the viscous sub-layer, your turbulent production is much smaller than the dissipation while in the log-law region, these terms are very similar.

I'd strongly recommend using an appropriate near-wall treatment which is able to accurately account for y+ ~ 1 meshes (especially, if temperature if your final goal).

Thanks.  

[damis93]

Hi all,

thanks for your answers. Yes, near wall treatment is ideally the best. Another model who gave good results was the Transition k-kl-omega(indeed in my case I have a transition to lower Reynolds). 

The matter is that relizable k-eps with standard wall functions gave the best prediction for ALL the exps I runned. In the specific, I was wondering why I got this good agreement. From Ansys manual, it is mentioned that standard wall function deteriorate if you are using a y*>11.125 and y*<30 (buffer layer), because there the log-law (used in the turbulent layer) deteriorates. 

What I don't really understand is:

If I use an inflation with y*=1, then I am in the viscous sublayer and Ansys will apply the laminar stress relation, which is in principle correct to me. So in this case, where is the deterioration? Is it in the buffer layer(in the upper cells of the domain) ?

Thanks for your answering.

 

[Kalyan Goparaju]

Hello, 

The laminar strain relationship you are talking about is for velocity only. The real issue is with the turbulence quantities (k,epsilon). The deterioration mentioned in the help manual is regarding these variables. 

Thanks, 

Kalyan

[damis93]

Ok, so the manual says that the production of k (G_k) is based on the logarithmic law (which I suppose is inserted into the wall stress tau_w). 

What about the Non_Equilibrium wall function then? They also gave good agreement with the experiments.

 

Thanks.

[DrAmine]

hey we already recommended the SST model. Non equilibrium are just an extension and I would not stick to it. Ensure you have your boundary layer resolved.