Discussion Groups

Simulation CFD

Reply
New Member
daniel.foeger
Posts: 2
Registered: ‎10-03-2013
Accepted Solution

Strange chimney effect at simulations with radiation

201 Views, 3 Replies
10-04-2013 04:26 AM

Hi!

 

I am working on a thermal simulation for a luminaire. The usage of Radiation in the simulation-model yields a strange very strong airstream from the bottom to the top, when using an autosized mesh. This of courses reduces the temperatures on the luminaire dramatically.

Depending on the mesh resolution the effect is stronger or weaker:

 

Clipboard01.png

                      autosized mesh                                                                high resolution mesh

 

The chimney effect is also present on various other simulation-models.

 

The effect occurs e.g. by changing the thickness of a PCB:

thick.png

 

Or by changing the emissivity of the outer surfaces:

epsilon.png

 

  *   I don't have a clue where this effect comes from??

  *   How can I be sure that the simulation delivers the right temperatures, when not checking the vertical airstream?

 

Could you please help me?!

 

Please use plain text.
Product Support
Royce_adsk
Posts: 554
Registered: ‎08-24-2011

Re: Strange chimney effect at simulations with radiation

10-04-2013 08:42 AM in reply to: daniel.foeger

Hi Daniel,

 

The default automatic sizing is almost always never enough for a natural convection chimney model.  So, I am not surprised that you needed to refine the model to get the appropriate plume development over the luminare.

 

Regarding meshing guidelines, we have a very detailed writeup regarding how to approach setting up your mesh for lighting applications that has proved very successful with our user base.

 

http://help.autodesk.com/cloudhelp/2014/ENU/SimCFD/files/GUID-7BFD671B-B6BA-4E6A-A67D-5DA51A3BC0AD.h...

 

Regarding radiation, you are going to need to apply a fixed temperature to the 4 walls so that you aren't inducing more convective currents off the walls. You can see the increasing velocities along the 4 walls of the chimney that really shouldn't be there.  The temperature that you apply to the 4 walls is going to be the same temperature that you have assigned to your inlet or your ambient temperature for your model.

 

For the Air emissivity we generally suggest to modify the default Air material with an emissivity of 0.3 and that has shown to be a better value without any other information about the environment around the light.  This will increase the temperature of your system a few degrees compared to running with the emissivity of 1.0 (default for air).

 

Other solver changes that you should probably do for you model setup: Use ADV5 and Mixing Length or Laminar physics models.



Royce Abel
Please use plain text.
New Member
daniel.foeger
Posts: 2
Registered: ‎10-03-2013

Re: Strange chimney effect at simulations with radiation

10-14-2013 06:36 AM in reply to: daniel.foeger

Hi Royce,

 

I agree with you:

The automatic meshing seems to be never enough for natural convection chimney models.

 

The hint with the four fixed temperatures at the side walls improves the situation:

 

Clipboard01.png

 

This is very helpful for us, thanks!

 

Setting the default emissivity value of air to 0.3 is of course better than always applying a radiation boundary with 0.3.

Am I right with the assumption that than the assigned temperature at the outer surfaces has to be the same as the environment temperature of air?

 

I also tried to use Laminar with ADV5 and Mixing Length with ADV5 solver models. Both models work quite well, the temperature is increased by one degree (mixed) and two degrees (lam), else nothing significantly changed.

Could you explain me the advantages of these models compared to the default k-epsilon model?

Please use plain text.
Product Support
Royce_adsk
Posts: 554
Registered: ‎08-24-2011

Re: Strange chimney effect at simulations with radiation

10-15-2013 12:38 PM in reply to: daniel.foeger
Great! I am glad your results are looking significantly better.

Am I right with the assumption that than the assigned temperature at the outer surfaces has to be the same as the environment temperature of air?

Yes!

I also tried to use Laminar with ADV5 and Mixing Length with ADV5 solver models. Both models work quite well, the temperature is increased by one degree (mixed) and two degrees (lam), else nothing significantly changed.
Could you explain me the advantages of these models compared to the default k-epsilon model?

In the past we have noticed that the mixing length and laminar models were more accurate for chimney models and also more stable in some situations. The other advantage is that these models are simpler models so it should solve faster.

Best regards,


Royce Abel
Please use plain text.