Comsol moving boundary+ Heat Transfer

AI Thread Summary
The discussion focuses on using COMSOL software to model heat transfer in a 2D geometry with a moving boundary condition. Users suggest employing the Arbitrary Lagrangian-Eulerian (ALE) application mode or parameterized geometry to handle the heat-induced boundary changes. Challenges include managing mesh inversion during simulations and ensuring convergence with varying material properties. Recommendations include defining moving subdomains, using larger mesh elements, and remeshing from existing geometry for larger displacements. The modeling guide is referenced for examples and further guidance on implementing these techniques effectively.
Skyeagle
Messages
15
Reaction score
0
Hi guys,

I have a question in using comsol software, particularly on heat transfer.

Its a 2D geometry. A rectangle whereby heat is applied on the top surface at the particular point and the 3 sides are thermally insulated.

As more heat is applied over time, a hole will be created. A small portion of the upper boundary gets 'removed' over time.

How do I create such 'moving boundary' condition in comsol?

Rgds
 
Engineering news on Phys.org
Hi Skyeagle, have you tried the ALE (moving mesh) application mode or the parametrized geometry application mode, those would probably suit your problem?
 
PerennialII,

How do I go about doing it? Or would it be possible to send me a comsol file of how the ALE works?

I am having trouble on setting the parameters...
 
The model library I believe has some examples (and the modeling guide has 2, remember seeing one moving water surface model somewhere in there at least), but we can sure work on something else if need be (if you've your basis 2d case mph file for example?). At 1st sight would try solving it by giving it a boundary displacement (for example) which would depend on the parameters of your problem (like in a simple case, deflect the mesh if the temperature goes beyond melting temperature ... give it a boundary velocity or displacement). The problem may turn out to be numerically "challenging", so using the simpler parametric geometry application mode to get started might be a way to go. And test something "simple" at first, like simply moving a boundary as a function of time and then implement more problem specifics which would model, for example, a melting process.
 
PerennialII,
Hi, I am currentlyworking on creating a model in Comsol multiphysics and I am using the ALE mode and fluid structure interaction (FSI). I've noticed that the arbitrary eulerian lagrangian method only works for small displacements due to the inversion of the mesh surrounding the moving object.

My question to you is, do you know how, or know of any information that I can get my hands on, which will tell me or give an example of how to save geometry after running my simulation and using that geometry for a re-mesh. Then I can re run the simulation with the new geometry and get larger displacements.

I know code can be written with MATLab, but right now I am only interested in figuring this out with Comsol. Any advice is appreciated.

Regards,
Nick Bruno
Northern Arizona University
Graduate Student - Mechanical Engineering
 
I have tried to use ALE in comsol. The 2D fluid_interaction model has an elastic blade normal to a liquid flow stream. The solution shows it bending. The Youngs modulus for the blade is 200kPa. To get familiar with the routine I tried changing E to 100kPa and it failed converge. I changed it to 150kPa and it failed again. It will work at 175 kPa. If this example written by the experts is so touchy, the method is either flawed or badly coded.
 
I have actually recently improved my understanding on this moving mesh application and it seems the problem with this ALE method is that the mesh elements "invert" or have a negative area is not configured correctly. There are ways around this.

One way is the define moving subdomains which smash or re-scale your elements to allow for minimal overlapping.

http://comsol.com/support/knowledgebase/970/

The example you are talking about uses one subdomain for the fluid, and one for the structure. If you modify the fluid subdomain into two parts (say at the top of the structure) and define the boundary to move with the same velocity at that point, you may be able to get much larger displacements with a beam that has a much lower flexural modulus.

Another idea could be to use larger elements. This will give you poorer results, but the mesh won't invert as often.

Another idea could be to run the simulation for a shorter period of time, remesh from existing geometry, and run the simulation for the rest of the time. Then compile the moving in the end. (read chp 16 in the modeling guide)
 
Last edited by a moderator:
Also, to answer the question in the top of this thread, there is an example of exactly what you are trying to accomplish in the modeling guide, chapter 16
 
Thanks a lot! Back to the keyboard.
 
Back
Top