Hello,

I have been doing many calculations now with Fluent and the same question pops up every time I have to make conclusions about it. What should I think about this y+ that ranges (generally) from 1 to 2000 knowing that I use the k-omega model?

Generally, I study pressure effects on a quite important surface or volume so I guess the fact that the boundary layer is not well calculated is not that important. So, two questions :

- is it that bad that the y+ is that high and/or not variant?

- are my conclusions about the pressure where I mesure it still valids?

Thank you in advance for answering my mail.

Eduardo Mendoza

Take care of y+ near the wall. That will give better result. Make it in range of 100.

Thank you for your answer.

I already knew that trying to keep a y+ close to 1 was the right thing to do but...

if you consider a calculation that takes about 5 days and you don't necessarily have the time and financial means to do your calculation all over again, what can you say about the calculation if the y+ is ranging from 1 to 2000 and that you don't especially work with the boudary layer but more with global mesures such as pressure on a big surface or volume.

Eduardo,

without knowing too much about your specific problem and by assuming that you are using a wall function to resolve the boundary layer region, I would say the following:

"- is it that bad that the y+ is that high and/or not variant? "

A y+ range of 1 to 2000 is definitely no good news. You ought to aim for 30 <= y+ <= 250, maybe even 30 <= y+ <= 500 but definitely no higher than that. Also, you shouldn't go lower than 30 (or 11.5, check Fluent manual) either.

"- are my conclusions about the pressure where I mesure it still valids? "

Well, that's the big one, but I don't want to go too deep into your specific problem as I simply don't know enough (how many characteristic lengths is the region of interst away from the wall, what are Re, Ma, etc). One thing though, which might help you, is to stop thinking of a flow field as two separate unconnected layers (boundary layer and free stream) because they are not. The boundary layer has a profound influence on the main stream and vice versa so they do influence each other!!

However, I suggest you do an inviscid calc. This is definitely quicker than a full blown viscid one and it should allow you to assess the effects of the wall on your region of interest.

Regards

Shubidu

Hello Shubidu or else,

thank you for taking time to answer. I beleive the k-omega model takes into account the y+ it measures and has a wall function law or not depending on the value it measures.

My question might need to be reformulated : how bad can influence the y+ result (let's say it is 2000) on the pressure considering the friction force due to viscosity on the wall is certainly dozens of time less important than other forces such as pressure force generated by an engine.

E.M.

With a y+ of 2000 you might as well not run viscous, because your turbulence model will not give you a good measure of the boundary layer. If (and since I don't know the details of your calculation I say if) your problem is not critically dependent on the boundary layer, then your approximation is probably reasonable. If boundary layer effects are important, then your y+ is way too large. I think this is what Shubidu is also referring to above. If such things as flow separation are important in your flow then you will need to reduce your y+.

I think the real question is why such a large range in y+? Are you using a HEX or TET/PRISM mesh? If you are using either you should be able to control the first node off the wall and reduce the large range.

Also, why not consider a KEps turb model? It will allow you to have a larger y+ than a Komega.

Derrek

Even with a k-eps model that value of y+ is too large to get an accurate simulation of the boundary layer, even with wall functions. With a y+ that large you might as well run inviscid.

Agreed.. But KOmega is very limiting. At least with KEps you can go up to ~100.

Cheers,

Derrek

Hello,

Well, the debate seems to go on utterly different direction than the one I intended to give in the first place but anyway, it is interesting as well.

Well, from what I red so far, k-w is always a better option than k-eps, is that wrong? Or is it just that k-w is always better than k-eps when in runs in its validation field?

As for my original problem, I have a neophyt question then, what does it change concretly to have a y+ of 2000. Does it mean that the value of the pressure at the end of the wall cell is not correct?

Thanks everybody for all the answers!

".... what does it change concretly to have a y+ of 2000. Does it mean that the value of the pressure at the end of the wall cell is not correct? "

Yes. You should not forget that turbulence modelling is by its very nature already quite a daunting task as it attempts to model a chaotic system through mathematical equations and empirical relations. These correlations are exactly that, empirical, and should therefore be adhered to as vigorously as possible.