[bram94]

Hello,

I'm trying to simulate flow through corrugated channel. I have calculated reynolds number and the flow is turbulent. I'm using k-epsilon model.

I have refined mesh near the wall. I'm getting average y+ value of 2.0, does that mean that I'm resolving viscous boundary layer to a good level?

I read in a few forums that k-epsilon model works only 30<y+<300. Does that mean that I must change my turbulence model to capture the physics ?

Thank you.

[LuckyTran]

You don't need to change turbulence models. What model would you even change it to?


For resolving the boundary layer, what's more important than the y+ of the wall adjacent cells is how many cells you have within the boundary layer. Of course this is correlated to y+ of the 1st cell.

The k-epsilon model with wall functions (in fluent) is quite robust and works down to any y+.

Originally Posted by bram94

I read in a few forums that k-epsilon model works only 30<y+<300.

This has to do with which wall functions end up being used. In Fluent, the wall functions are blended anyway for low y+ and high y+ and this isn't an issue. It still is a good idea to either always have low y+ or always have high y+ so that you know better which regime and which functions are being used, but it is not a problematic issue.

Originally Posted by bram94

I have refined mesh near the wall. I'm getting average y+ value of 2.0, does that mean that I'm resolving viscous boundary layer to a good level?

Are you talking about resolving the boundary layer in general (which is obviously viscous and you are being redundant) or are you talking about resolving the viscous/laminar sublayer (a tiny fraction of the overall boundary layer)? The viscous sublayer occupies y+<5. A y+ of 2 means you only have ~2 cells in the viscous sublayer which is not enough to resolve the viscous sublayer.

[bram94]

Thanks for your response. I got an idea from your explanation. Sorry for the delay in replying.