HTC in Pipe flow
 

Hi,

I am simulating a pipe flow an encountering problems obtaining correct values for the inner heat transfer coefficient.

My mesh has about 8million nodes and contains of hexa and prism elements. Growth is 1.04. The smallest thickness is in the prism layer at the wall with 2.3e-2mm. With a pipe diameter of 10mm this leads to a y+ of little below 5, which literature stated to be sufficient. I use the SST turbulence model with automatic wall functions. My boundary condition is a constant heat flux (though I already tried constant temperature and identical temperature for flow and wall which lead to the same restults) of 1000 W/m².
I cutted the pipe into half with a symmetry plane to reduce calculation time. I know that turbulent flows are three dimensional, but literature says that flows in annular pipes are an exception. With high resolution i let the calculation converge to about e-7 for all residuals, the average heat transfer was constant within the last 100 iterations.

My problem is, that CFX calculates a value of about 27000 W/m² which Gnielinski states is far too high (Gnielinski equations leads to about 17000 W/m²). I read that Tbulk for HTC must be set and tried that - without success. Anyhow I have a problem in understanding the setting:
CFX calculates a wall temperature of 298.185K and a bulk temperature of 298.15K (when stream is fully developed). With htc = q/dT this results in a htc of about 28000 - roughly the value cfx calculated before. I already checked the values for density, heat capacity etc. with values from NIST, but everything is correct. I use CFX12.

Since I assumed that a pipe flow with Re of about 90000 is a rather simple matter I am entirely confused, since i keep on rechecking any settings for two weeks now and can't find my mistake. I would really appreciate some help!
Thank you.

This is my first post, so feel free to ask, if I forgot some important information.

Subsemitonium

You cannot simply use the mesh recommendations from the literature. Half the time they do not check properly, half the time they use a different solver with different requirements. This FAQ has some general tips.

http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

But heat transfer modelling is far from accurate. Errors of 50% can happen. A big source of errors is the turbulence model, getting the heat transfer right is usually even harder than getting the bulk flow right.

Also, the point of a turbulence model is that a 3D transient flow can be reduced to a 2D steady flow. The turbulence model does the simplification by modelling the effects of the turbulent fluctuations on the bulk flow.

Thanks for your answer. Actually I don't think it's the mesh. I modified/refined it four times and always got exactly the same results. On my way to desperation i already conducted a transient simulation run and *tataaa* got the same 27000 W/M². What could be wrong, do you have any hints?

The FAQ discusses much more than just the mesh.

Hello I am new to star ccm+ and I am trying to simulate laminar flow of mercury through the pipe of diameter 1cm and flow avg velocity = 0.022m/s.
Re=2000
and constant heat flux of 25000W/m^2 through the walls. the pipe length is 2m.
flow inlet temperature 300K

what should I keep static temp of outlet ?

What boundary conditions shall I set to get appropriate solution ?
(Currently I am not at all getting proper results)
Please Help...

Try the star CCM forum: http://www.cfd-online.com/Forums/star-ccm/

I have simulated single phase tube heat transfer with CFX before and was within about half a percent from the Petukhov correlation. I used SST and had a Y+ of about 1. So it worked very well for me.

I see you say you have an annular pipe (tube in a tube), what correlation did Gnielinski give for a turbulent annulus? You describe only 1 diameter?

Are you looking at Hybrid values for wall temperature?

What did you pick for "bulk temperature" It should be a function of length along your pipe which you would calculate with a simple mass and energy balance.

I could be wrong, but I'm guessing there may be something wrong with the way you are calculating the HTC.