Issue with simple heat flux into cylinder

[AlfredBerger]

Dear Sirs,

I have simulated a simple steel cylinder and have given just a heat flow on its lateral surface (in W/m2).
I did check the heat flux with SaveScalars and the value (in W) was O.K.
But: providing this certain heat flow would create distinctively higher temperatures (more than an order higher than received values), which I cannot understand.
In addition, the initial temperature is 20 degree, the minimum Temperatures is about -20 degree after a couple of seconds. Any smaller temperature than 20 does not make physically any sense as there is only one BC with heat flow and this is positive and constant.

Attached you will find the mesh, case.sif and SaveScalars results.

Any help would be appreciated.
Alfred

[raback]

Hi

I would rather prescribe "External Temperature" and "Heat Transfer Coefficient" than given heat flux. It may be that you ask such a flux that the only way to enable it is to create a negative temperature below the skin. The standard finite element method does not ensure monotonicty, the equations are enforced only weakly.

When you turn the "Heat Transfer Coefficient" higher you will see what is the maximum heat flux that you can pump to the system.

-Peter

[AlfredBerger]

Hello Peter,

Thank you for the very fast / nightly and clear answer. Reading your answer I feel ashamed of myself that I could not answer it myself.
Only issue is, that the values are typical ones for induction heating, so I did not try to push temperature in too fast. 150mm steel bar to be heated up 300 K in about 3 minute (plus homogenization time) is industrial standard. I checked with the induction system supplier and they confirmed that the 400 W/cm2 heat flux is doable. So I sticked to the 400W/cm2 but did not introduce it as "heat flux" value but via the "heat transfer coefficient". I ended up with the same situation (much too low temperature, negative temperatures under the outer shell). Any numeric tricks to enforce the equations ?

So if you are correct, then still there is the question, why - with the high heat flux - does the material not heat up accordingly in AVERAGE? So we do have a body, we see a certain heat flux (18kW) but the mean temperature does not show up with corresponding values ?

By the way, is there any specific book on CFD that you would recommend before starting - in my case - compressible laminar CFD studies with Elmer ?
For my first trial (air filled pipe) I could not even set-up a mesh with two bodies. What is the recommended (if possible OS) preprocessor for CFD studies ?

Thank you again for your support on this excellent OS program,
Alfred

[AlfredBerger]

Hallo Peter,

Addition to the above:
I have used smaller values (1/10) to get results, without negative temperatures. The evaluated heat flux (W) seems Ok, but the temperature rise definitely not. I have double checked it with my Finite Difference Temperature Program.

I do not find my fault.

In principle I wanted to see the vonMises stresses caused by the heating. Do you think that the other active solver is messing here up the Heat solver ?

Alfred.

[raback]

Hi

The implementation should be consistent. The it is about convergence with respect to time and space resolution. I would start with smaller timesteps, or possible add some nice boundary layer, or introduce higher order elements.

-Peter