[Onur10]

Hey there;

I have the same question about the Nusselt Number. I have posted this onto someone elses topic too, but I think it will draw more attention if it's a separate topic by itself.

I am a bit new to Fluent, and I am currently working on a cylindrical pipe with turbulent flow.

Geometry: My pipe is 1 meter long and has a 8 mm diameter.

Working Fluid: I am using standart water as the working fluid, and compute turbulent flow with Re ranges of 6000 to 10000. Tried some random other fluids too after recieving bad results with water.

Boundary Conditions: I work for both constant wall temperature and constant heat flux boundary conditions.

Mesh: For 2D I use a standart mesh with bias on both ends to the wall to approach the Nu correctly, for 3D geometry I use Multizone mesh and inflation with a good mesh statistics.

Nu calculation:
I read q", Twall and Tbulk values from Fluent and calculate my Nu by hand.

Nu = h*D/k

h = q" / (Tw - Tb)

I get bulk Temperature by creating planes (or lines for 2D) at crossections of the pipe and getting surface integral / mass weighted Temperature average of them.

For constant heat flux, q" is already given, for constant wall Temperature, I read from post process - wall heat flux at the given position.

For constant Twall i use the given Twall, for constant q", I draw a line at the wall surface parallel to flow direction at post process, and read the inner wall temperature on that.

Other variables are given already at the start of the analysis.

First I calculate h from q", Tb and Tw. Then I calculate Nu from h, D and k.

Problem: For both 2D and 3D approach my Nu to a developed region Nu and stays constant there. However the approached Nu is very big compared to theoretical values.

As an example, theoretical (and experimental as well) values suggest Nu = 43 for a case, but I get both with 2D and 3D models I get Nu of 160 or something like that. Similarly for different cases the Nu in simulations is 3 to 4 times the actuel Nu.

What I have tried:
- Changed the mesh parameters, mesh type etc. Eliminated the possibility of mesh dependent solution

- Changed the problem geometry geometry

- Changed the fluid in consideration

- Changed boundary conditions (by having the same Re)

- Changed Re (flow velocity etc.)

- Changed simulation dimensions (both 2d and 3d)

* * *

I dont know what else I should do. I always get the same "much higher" Nu and there is a developed region there for sure. Do I use false calculations? I always approach a Nu about 3 times greater than the realistic Nu.

Anyone has any solution? It is a bit urgent, so I would really appreciate the help.

[avi031]

check your y+value and use turbulence model accordingly

[robboflea]

Your y+ when dealing with heat transfer should be less than 5 (better if less than 1). Consider also that the flow turbulence intensity may have some influences on heat transfer together with Re. Also have you tried to use different turbulence models? I never had to simulate flows in closed pipes but when dealing with Nu on a flat plate I found good agreement by using Spalart Allmaras model.

Hope this helps.

Cheers,

Rob

[sadir]

i think u cant use s. allmaras model for this problem. k-e turbulance model correct. u should make mistake to calculate nusselt number.there are lots of ways and theories to calculate its.maybe if u choose correct theory can get correct values

[yihe]

Did you slove the problem? I met the same problem with you, and I have checked my Y+ number, it's below 1.5. I also tried SA, kepsilon, komega models, both of these models gave a very high nusselt number.

[Shabi1992]

Originally Posted by yihe

Did you slove the problem? I met the same problem with you, and I have checked my Y+ number, it's below 1.5. I also tried SA, kepsilon, komega models, both of these models gave a very high nusselt number.

I have this problem as well. Can somebody help me?

[avi031]

this may have to do with incorrect post-processing.

[wales]

Slm,

Ensure firstly that your y+ is in the range of 1-5. From articles I have read I can say that for heat transfer simulations the SST turbulence model is held in high esteem. The nusselt number fluent gives is often deceptive for turbulent heat transfer simulations ( I am also dealing with one right now). Thus if your heat transfer coefficient goes with the theoretical one you can go on. But get the heat transfer coefficient by using empirical equations which result in good agreement with experimental studies. ( for example dittus-boelter equation)

Also check your convergence. Good luck