Originally Posted by SENER (Post 502367)

Hello everyone,

I am a new user in FLUENT. I have been investigate turbulence flow in rectangular duct. I tried to use lots of mesh type on model but I didnt adapt to Dittus-Boelter correlation for Avaraged Nusselt Number.

My model is 80x80x200 mm rectangular duct, Re:10 000 so that dittus boelter appx. 32 . But I found between 55-87 at same Re value.

Turbulence model is k-epsilon, y+ is changing between 0,1-1,2(y+<5) . FL thickness is 0,0498 mm. I tried to select between 10-35 layers in Inflation but not achieved.

Can you explain me how to decrease A. Nusselt Number to dittus boelter value?

Thanks.

Originally Posted by Goutam (Post 502485)

It's better to use Realizable k-e model for turbulent case. What about Gnielinski, petukhov or Notter-Rouse correlations? Did you try all these correlation as well? What is the boundary condition? Did you extract the wall temperature first and then calculate the local Nu and from that average Nu?

You can check these all !

Originally Posted by SENER (Post 502786)

Thank you Goutam.

I didnt try to compare Gilenski and Petukhov with D. Boelter before.OK.

Boundary conditions; Rectangular duct is heating constant heat flux from bottom and also inlet and outlet.

When I find a good result from model, I will attach some fin geometry in duct. What about number of boundary layers? How can I optimize number of layers for Avag. Nusselt Number?

Originally Posted by Goutam (Post 502789)

I think, you should use Gilenski, Petukhov and Notter-Rouse correlations for comparison. Also, try to use some experimental data.

I think, Boundary layer is not a big problem as long as y+ < 5. You can also try uniform mesh but dense mesh (along y direction means radial) which will force to keep y+ < 5.

First part is to check how your local Nusselt number behaviour. It's meaningless to check only average Nusselt number, because this will give you only a constant value.

Thanks