Learning Forum

[sagardeprem]

I am performing a high speed simulation with a normal shock of about Mach 1.5 (Model - k-w sst). I need to resolve the boundary layer close to the wall in order to predict flow separation.

I have an experimental velocity profile at a certain location that needs to be matched. I first did the simulation with zero roughness height and the y-plus was about 0.9-1 (which I believe is what I need), but I couldn't get the same experimental profile.  By experimenting with different non-zero values for roughness height (same mesh) I was able to somewhat match the profile, but the wall yplus was above 100. Decreasing the first wall height didn't make a difference. And by this time I had realized, my roughness height is way higher than the first cell height (which I found out was non-physical)

So I read the documentation that a non-zero roughness height would shift the wall to half the roughness height. If that is so, from where is the yplus being calculated? The actual wall or the shifted virtual wall? 

Prem

[Raef.Kobeissi]

Hello The law of the wall changes slightly when roughness height is added. Please read this from Fluent help document:
“Note that it is not physically meaningful to have a mesh size such that the wall-adjacent cell is smaller than the roughness height. For best results, make sure that the distance from the wall to the centroid of the wall-adjacent cell is greater than $K_s$.”

I believe your Y plus value has increased because the roughness height is now determining it rather than your first cell height .

[sagardeprem]

Thank you for your reply! That statement was the sole reason I had to ask this.

Experimentally, there's a small separation bubble close to the wall and the size of this is smaller than the roughness height that matched the velocity profile at a location upstream. So I can't increase my first cell height above the roughness height, else I cannot capture the separation. 

So in order to make sure my BL is correctly predicted, would it help to calculate the yplus from the actual wall and keep it about 1? I couldn't find anything that would imply this is valid in the fluent document.

Prem

 

[sagardeprem]

So I calculated the integral of the wall shear stress for both the cases and I think I found the culprit. The average friction velocity (u*) were in proportion to the yplus for both the cases. So I know, the velocities are not wrong, but the wall shear stress is wrongly predicted by setting first cell height below the wall roughness height.

Also, I gathered that yplus is calculated from the original wall itself (link below), which seems correct now because it was the high friction velocity (u*) that increased the yplus.   

 https://www.cfd-online.com/Forums/fluent/165170-simulation-surface-roughness-fluent.html

[Raef.Kobeissi]

Excellent!

[jonheb]

Hi there I was wondering if you could help me out with a similar problem.

I am modelling a micro channel with k omega SST turbulence and there is a constant heat flux also added at the wall of my micro channel. 

I have modelled the micro channel as a smooth flow initially and have been getting a Y+ of 1 for Reynolds of 40k approx., but when I add ks of 9.84e-5 the Y+ is now 200+. I don't understand what I need to do in order to get it back to Y+=1 again, I have played with the bias of the mesh but nothing changes.

I have read the law of the wall modified for roughness but it doesn't really make sense to me.

[LorenzoMazzei]

Hi jonheb,

I think is wrong to seek necessarily the y+ value of 1 in case of rough surface. It's better to stay in the log-wall region and model the near-wall flow with a rough wall function.

[LorenzoMazzei]

Dear Raef,

I don't understand if there's a sort of Automatic wall treatment also in case of rough wall. How can I modify the first layer height each time depending on the roughness height? Is there a kind of scalable wall function also for rough walls?

Lorenzo

[DrAmine]

There us an automatic wall treatment and yplus should be kept as revommended for smooth walls even if roughness destroys the viscous sublayer. Just use SST turbulence model with.

[LorenzoMazzei]

There us an automatic wall treatment and yplus should be kept as revommended for smooth walls even if roughness destroys the viscous sublayer. Just use SST turbulence model with.

Dear abenhadj,

thanks and sorry for my late reply, but I didn't receive any notification. I am performing several tests and what I am seeing is that for high roughness values there is a strong sensitivity to mesh, while I would have expected less impact.

I am simulating a simple straight channel with 10, 20, 40 elements per hydraulic diameter and 5, 10, 20 prisms on the wall. Despite the very fine mesh, I find it difficult to obtain an insensitive result in terms of both friction and heat transfer. 

 

[Rob]

Depending on the size of the channel what else could be going on?  

[LorenzoMazzei]

Depending on the size of the channel what else could be going on?  

Hi rwoolhou,

what I mean is that I found difficult to obtain a mesh insensitive result in case of high roughness (up to 25% differences in f and Nu between the coarsest and the finest mesh), whereas it is more straightforward to do it with a smooth wall (5% max error).

I am still trying to figure out why.

Lorenzo

[DrAmine]

Hi if you have high roughness or the surface roughness is large relative to the near wall mesh please use the new High Roughness Icing Models

Investigate, verify, validate and compare and publish.

 

[LorenzoMazzei]

Hi abenhadj,

this is exactly what I am doing. I am performing simulations against exp data of rough AM test cases, but I can't get a mesh insensitive result. If you have something in mind I am accepting suggestions.
I'll make additional tests using also the icing models as suggested.

Thanks,

Lorenzo

[AhmedCfx]

Hi there

How can I change wall roughness to gravel has different diameters (10-20)mm ?

[Rob]

Work out the roughness parameters using the documentation and theory.