Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Heat Transfer - Natural convection electrical wire wrapped around mandrel

Please login with a confirmed email address before reporting spam

Hi all,

I am trying to solve a simple thermal problem in which I want to see the temperature profile in wires in a coil configuration (i.e. wire wrapped around a mandrel). Power is dissipated through the wire via the flow of electrical current. The temperature is effected by natural convection (hot air rises).

I setup a model based off on the tutorials (Convection Cooling of Circuit Boards—3D Natural Convection, see attached). The one major difference between my model is that I set it up as a 2D axi symmetric.

For some reason I keep getting a convergence error. I think it has something to do with the direction of gravity in my model. For some reason it doesn't like that I have gravity (volume force) defined in the negative r direction and then I have an open boundary condition. I get convergence when gravity is in the z direction.

I have attached my model. Any suggestions?

Thanks!


9 Replies Last Post 2 juil. 2012, 11:11 UTC−4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 juin 2012, 02:05 UTC−4
Hi

Yes a have a few suggestions:

For me it seems that your PVC material data is missing some variables (rho and Cp) or is it the translation into 4.3 that failed ?

Then you need to define a fluid domain node for your air, if you want fluid transport, else you have "just" static conduction, finally if you have a radial gravity in a 2D-axi model that means you are rotating your model so that the centrifugal force is along r. Normally gravity can ONLY be along Z in a 2D axi problem.

Then you have a heat flux boundary but your air fluid will not exchange by convection like that (a solid yes) it rather an oupen boundary where heat can exist but then air will also enter so you need to define the entry temperature.

I would try with a larger air area around

--
Good luck
Ivar
Hi Yes a have a few suggestions: For me it seems that your PVC material data is missing some variables (rho and Cp) or is it the translation into 4.3 that failed ? Then you need to define a fluid domain node for your air, if you want fluid transport, else you have "just" static conduction, finally if you have a radial gravity in a 2D-axi model that means you are rotating your model so that the centrifugal force is along r. Normally gravity can ONLY be along Z in a 2D axi problem. Then you have a heat flux boundary but your air fluid will not exchange by convection like that (a solid yes) it rather an oupen boundary where heat can exist but then air will also enter so you need to define the entry temperature. I would try with a larger air area around -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 juin 2012, 11:40 UTC−4
Thanks for your advice Ivar.

Couple of quick questions.

If I can only have the gravity along the Z axis, how can I modify my geometry so that I keep the same relationship between the wire orientation and gravity? I would like to keep the same orientation as you saw in my original file. Would I have to go to a 3D model? If so, how does one create a wire geometry (i.e. a cylinder that is bent)?

Also, I am not sure what you mean when you say a fluid domain node. Are you referring to a Point Constraint? In that COMSOL pdf that I attached, would you recommend I add one?

Thanks again!
Thanks for your advice Ivar. Couple of quick questions. If I can only have the gravity along the Z axis, how can I modify my geometry so that I keep the same relationship between the wire orientation and gravity? I would like to keep the same orientation as you saw in my original file. Would I have to go to a 3D model? If so, how does one create a wire geometry (i.e. a cylinder that is bent)? Also, I am not sure what you mean when you say a fluid domain node. Are you referring to a Point Constraint? In that COMSOL pdf that I attached, would you recommend I add one? Thanks again!

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 juin 2012, 13:50 UTC−4
Hi

if you align the gravity along any oher axis than Z, for a 2D-axi you are violating the 2D-axi symmetry condition, so you then have only full 3D that can be used.

2D-axi means valid for any phi the loop angle, and then your gravity vector is not invariant w.r.t to the phi angle.

It's the same for flow in a pipeline (axis horizontal with "vertical gravity") you cannot simulate that as is in 2D-axi as the gravity components is NOT 2D-axi symmeric

You still might want to simualte everything except the fluid gravity in 2D axi, and then project the heat flow onto a 3D fluid simulation, but then you will NOT couple back the fluid temperature induced changes into the 2D axi heat generation model, its a one way coupling, not flly representative

--
Good luck
Ivar
Hi if you align the gravity along any oher axis than Z, for a 2D-axi you are violating the 2D-axi symmetry condition, so you then have only full 3D that can be used. 2D-axi means valid for any phi the loop angle, and then your gravity vector is not invariant w.r.t to the phi angle. It's the same for flow in a pipeline (axis horizontal with "vertical gravity") you cannot simulate that as is in 2D-axi as the gravity components is NOT 2D-axi symmeric You still might want to simualte everything except the fluid gravity in 2D axi, and then project the heat flow onto a 3D fluid simulation, but then you will NOT couple back the fluid temperature induced changes into the 2D axi heat generation model, its a one way coupling, not flly representative -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 juin 2012, 18:51 UTC−4
Thanks Ivar, your explanation is very helpful.

Based on your advice, I tried to setup a 3D model. For some reason I am now getting a weird meshing error. It looks like the Mesh engine cannot handle the curvature of the 'wire'. I tried both a helix structure and a torus, but I get the same error with both. I have attached both my model files.

Do you have any advice? I have never seen such an error before. I really appreciate your help.

Thanks!
Thanks Ivar, your explanation is very helpful. Based on your advice, I tried to setup a 3D model. For some reason I am now getting a weird meshing error. It looks like the Mesh engine cannot handle the curvature of the 'wire'. I tried both a helix structure and a torus, but I get the same error with both. I have attached both my model files. Do you have any advice? I have never seen such an error before. I really appreciate your help. Thanks!


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27 juin 2012, 09:20 UTC−4
Hi
meshing a helix is delicate, I have never really seen that work nicely before also for other software than COMSOL, its' not so much making a single helix mesh, but as for your case there are several embedded helix's+ the air around that gets a weird mesh or as you see no full mesh at all.

One way around is to make 3 loops (torus's and not an helix) and simulate that, you can then even cut up into 1/8 by symmetry (but that would not be compatible with a Y gravity vector, hence only 1/4). Or what I often do, cut into 1/4 slices as then you do not have full torus topologies and the mesh is often easier made, not to say you can use mesh mirror commands

furthermore for your case with joule heating at term for your coils you should ideally add some straight cables going all the way out to the boundary, and loop the current around in the outer air boundary to preserve current conservation, but that is another story

--
Good luck
Ivar
Hi meshing a helix is delicate, I have never really seen that work nicely before also for other software than COMSOL, its' not so much making a single helix mesh, but as for your case there are several embedded helix's+ the air around that gets a weird mesh or as you see no full mesh at all. One way around is to make 3 loops (torus's and not an helix) and simulate that, you can then even cut up into 1/8 by symmetry (but that would not be compatible with a Y gravity vector, hence only 1/4). Or what I often do, cut into 1/4 slices as then you do not have full torus topologies and the mesh is often easier made, not to say you can use mesh mirror commands furthermore for your case with joule heating at term for your coils you should ideally add some straight cables going all the way out to the boundary, and loop the current around in the outer air boundary to preserve current conservation, but that is another story -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27 juin 2012, 13:17 UTC−4
Hi Ivar,

Thanks so much again for your advice.

It looks like taking a torus and splitting it into 4 parts helps in terms of meshing. Just to make things simple I am looking at one loop of wire. Also, I am not trying to model Joule heating. I already know the power dissipated, so I am just interested in strictly a thermal problem.

For some reason, now I am getting a convergence issue. I get the error:

Failed to find a solution.
Very ill-conditioned preconditioner.
The relative residual is more than 1000 times larger than the relative tolerance.
Returned solution is not converged.

I wonder if the issue is with my boundary conditions. Right now all the boundaries on my box (i.e. surrounding air) is Open Boundary. Do you think this is a good idea? I tried to make some of the Thermal insulation, but that didn't really help.

I have attached my model file for your convenience. Sorry to keep bothering you, but it's easy to get frustrated. Your advice is much appreciated!

Thanks!
Hi Ivar, Thanks so much again for your advice. It looks like taking a torus and splitting it into 4 parts helps in terms of meshing. Just to make things simple I am looking at one loop of wire. Also, I am not trying to model Joule heating. I already know the power dissipated, so I am just interested in strictly a thermal problem. For some reason, now I am getting a convergence issue. I get the error: Failed to find a solution. Very ill-conditioned preconditioner. The relative residual is more than 1000 times larger than the relative tolerance. Returned solution is not converged. I wonder if the issue is with my boundary conditions. Right now all the boundaries on my box (i.e. surrounding air) is Open Boundary. Do you think this is a good idea? I tried to make some of the Thermal insulation, but that didn't really help. I have attached my model file for your convenience. Sorry to keep bothering you, but it's easy to get frustrated. Your advice is much appreciated! Thanks!


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27 juin 2012, 16:56 UTC−4
Hi

first of all there is a much easier way to split your volume in "Union" mode. You simply add a layer in your block on left and front side of half the total width length i.e. 40 here

Then you draw your full torus (the two you want)

and that is it

When you enter into the material mode, COMSOL unites the objects and split them up accordingly (this does not always work like that for helixes, that is also an issue ;)

Anyhow its does not seem to be enough the ration of your torus diameter to the large block diameter is such that the mesher has issues to stitch the mesh together, you need to use a very fine mesh, hence many elements, oyu migt ned to cut your air volume into toe in the thickness too, to help the mesher

--
Good luck
Ivar
Hi first of all there is a much easier way to split your volume in "Union" mode. You simply add a layer in your block on left and front side of half the total width length i.e. 40 here Then you draw your full torus (the two you want) and that is it When you enter into the material mode, COMSOL unites the objects and split them up accordingly (this does not always work like that for helixes, that is also an issue ;) Anyhow its does not seem to be enough the ration of your torus diameter to the large block diameter is such that the mesher has issues to stitch the mesh together, you need to use a very fine mesh, hence many elements, oyu migt ned to cut your air volume into toe in the thickness too, to help the mesher -- Good luck Ivar


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 juin 2012, 12:29 UTC−4
Hi Ivar,

Thanks for your advice again....especially for how to split volumes into multiple pieces.

I did what you recommended (splitting the air into 3 parts and increase mesh density) but my model still does not converge.

I looked at the screen capture you provided and it looks like your mesh is much more sophisticated than mine. Am I correct? How did you go about making your mesh? Or is the standard meshing that I used fine? I have attached my model again.

Thanks!
Hi Ivar, Thanks for your advice again....especially for how to split volumes into multiple pieces. I did what you recommended (splitting the air into 3 parts and increase mesh density) but my model still does not converge. I looked at the screen capture you provided and it looks like your mesh is much more sophisticated than mine. Am I correct? How did you go about making your mesh? Or is the standard meshing that I used fine? I have attached my model again. Thanks!


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2 juil. 2012, 11:11 UTC−4
Hi Ivar and others,

Any other thoughts on why model is having an issue converging? I have attached it again. I feel like it's either a mesh issue and/or some boundary condition.

Essentially, I want to see the temperature around a wire (with insulation) given a certain power dissipation; the main physics I am interested in is natural convection. I only have a simply model setup in the attached file. However, my end goal is to have a 30-40 turns and see how the spacing between the wires impacts the temperature.

I have setup this problem using a 2D-axi symmetric setup. Unfortunately, I am limited by the direction I can use for gravity with respect to the orientation of the coils. Thus, I have to go to a 3D model.

I would appreciate any thoughts anyone might have!

Thanks!
Hi Ivar and others, Any other thoughts on why model is having an issue converging? I have attached it again. I feel like it's either a mesh issue and/or some boundary condition. Essentially, I want to see the temperature around a wire (with insulation) given a certain power dissipation; the main physics I am interested in is natural convection. I only have a simply model setup in the attached file. However, my end goal is to have a 30-40 turns and see how the spacing between the wires impacts the temperature. I have setup this problem using a 2D-axi symmetric setup. Unfortunately, I am limited by the direction I can use for gravity with respect to the orientation of the coils. Thus, I have to go to a 3D model. I would appreciate any thoughts anyone might have! Thanks!

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.