Discussion Forum

Hi

What I see is a "thin" 3D Solid model, rather complex in shape, but with a close to perfect 2D shell symmetry in the geometry, and in the BCs, no.

Then to get the HT (which is a diffusion problem) to converge easily, you must respect a certain mesh density that in your 3D model gives quite some elements, and will be unnecessarily heavy. Your material with a good electric conduction, also has a good thermal conduction and the heat diffusivity is such that across the thin layer your temperature gradient will rapidly drop to 0 (the gradient hence the T is constant across a section normal to your shape. This means that a 2D shell element would have some 1/500 or even less number of elements, hence will solve at least 500 times quicker and better in 2D Shell than in 3D solid.

You must estimate the heat diffusivity alpha=k/rho/Cp of your material and check the time constant to reach a constant T for a thickness corresponding to your shell. Then you need at least 3-5 elements across your shell thickness in 3D. The pulse shape of your current will decide if the current is flowing in the surface skin or in the bulk, compare that pulse time to the time to homogenise the half shell thickness and you will see if you have any effects in the thickness, I suspect not. => this means that you can just model the thin shell as a surface and use shell elements.

To understand this better start with a simple 2D thin rectangle with a ratio 1:20 and play with the heat transfer until you master the time constants: heat pulse travelling from one end to the other or from one large surface to the other. Check your mesh dependence on the results.

Then you might play with the EC model alone to see how a pulse propagates across a line/strip of similar shape, when you increase the frequency (i.e. AC current modus) you will see the current will concentrate along the surfaces, that is also where it will deposit its energy, hence heat up the bulk conductor, this heat will flow to the middle of the conductor, and some will leave by the external heat exchange (air convection, external cooling or whatever you have. Again, check your mesh density and relation to correctness of results.

Then mix everything in a full 3D model but taking into account the mesh density lessons learned above, you will need a high density in the thickness to resolve the flow, and along the length where the current is flowing, and towards the surface where you might have skin effects, but probably you can solve correctly with long elements transverse to the current flow (another indication that 2D could give as interesting results as a 3D model).

In any case its difficult to say exactly what is wrong (specially that I do not have access now to my COMSOL workstation to run the model) but my suspicion is the mesh density first of all ...
--
Good luck
Ivar

Thank you so much.

You are the best!