Discussion Forum

Kyle,

There is no electric field in the Magnetic Fields (mf) physics for the Stationary analysis type, meaning that EM heat source is zero.

Regards,
Sergei

Hi Sergei,

Thanks for looking into my problem. I added a time dependent step to my project, but I'm now getting an error when I compute, "Failed to find consistent initial values" and the solution does not converge. I believe my initial conditions are reasonable, but I would not at all be surprised if I configured something incorrectly, particularly with the heat transfer physics. I'm completely new to the thermal side of comsol. Any assistance with the proper boundary conditions etc would be appreciated.

Kyle

Kyle,

For Magnetic Fields (mf) physics in time domain or in frequency domain, air conductivity should be non-zero small value (say, 1 S/m) to get convergence (this is so called "Conductivity scaling effect" - solver is unstable in 3D if air conductivity is zero). Also, you are applying constant current at t=0 meaning that there is huge magnetic flux change rate at initial time instance - extremely difficult situation for solver. It is always good idea to avoid singularities and ramp up current continuously.

Based on your initial description of the problem you are trying to solve, seems that you want to calculate coil heating under steady-state or DC current. If this is the case, you don't need (mf) physics interfaces. You can use Electric Currents (ec) physics interface, apply Ground BC at one side and Terminal/Current at the other side of the of the coil. Than apply domain heat heat source Q0=ec.Qrh*Vf_copper, where Vf_copper is volume fraction of the copper wires. Attached file is implementation of this approach.

Regards,
Sergei

Thanks again Sergei. That all makes sense. Unfortunately, when I attempt to open your uploaded mph file, I get the following error "COMSOL Multiphysics model file is damaged or not valid."

I believe the problem is because you used 5.2a, I just have 5.2.

Kyle,

Attached are models in 5.2.

Regards,
Sergei

Thanks again. I'm understanding more now, but have a couple follow up questions, because I'm not totally confident in the results. I've also upgraded to 5.2a, so the version issue is no more.

Q1. Why are there no subnodes under the mulitphysics node? By default, if I just add Joule Heating to a model, I get three subnodes under multiphysics (Electromagnetic Heat Source, Boundary Electromagnetic Heat Source, and Temperature Coupling). Are these unnecessary? Have you compensated for their non-existence in some way?

Q2. I get no change in terminal voltage over time. This is not realistic, since as the copper heats up, resistance goes up, and with a fixed current source, I should be getting a voltage increase over time. I assume that comsol should be able to account for this. However, for some reason it isn't.

Q3. Something is up with the amount of power that gets dumped into the coil. My magnetic field simulations gave me ~13kW for NS turns each carrying III amps. Now when I feed NS*III current into the block, and multiply that value by the output terminal voltage, I get around 1.8kW of power. This results in lower temperatures than I would expect. I tried tweaking the model and put in a scale factor of 3 to increase the current to reach P = 16kW = Vout*III*NS*3, but I'm still not sure why the original didn't give the expected ~13kW.

The forum isn't letting my upload my latest file because of an "extension error"

Kyle,

Q1. Multiphysics coupling node applies heat source as mf.Qrh or ec.Qrh, depending on which physics you are using. Multiturn coil in (mf) physics takes into account potting material and calculates wire volume based on wire radius and number of turns. Terminal BC in (ec) physics applies current to the solid copper domain - that's why I choose to specify heat source manually taking into account wire volume fraction. Since the heat source is the only coupling predefined in the Multiphysics node (there is no temperature coupling in your model, because material properties you specified do not dependent on temperature), I disabled this node.

Q2. Yes, copper heats up but material properties of the copper remain the same in your model.

Q3. Not sure, where P=16 kW is coming from. I used Global Evaluation to calculate total power for the (mf) and (ec) and it turns out that power in both cases is almost the same (when factor Vf_wire is taken into account, as shown in the attached image.

Regards,
Sergei

Thanks again for your help. I think I have everything all figured out now.

Q1. Makes sense.

Q2. I believe I have changed this properly, and now copper resistivity is a function of temperature. I now get a changing voltage with time that makes sense.

Q3. Sorry for the confusion here, this was a mistake by me. I had the number 13kW in my head, because that was the total power dissipated in the full coil, not just the 1/4 symmetry coil. So the output from these power calculations in the heat model are 1/4 of the total power. Also, I refined the 0.8 factor to a slightly more accurate 0.65.

At the end of all that, after 20 minutes of applied current, I get 3.2kW dissipated, which when scaled for full symmetry gets me right near my expected 13kW, with a time varying voltage.

Thanks very much for your detailed help! It was greatly appreciated.