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
22 apr 2013, 11:58 GMT-4
Hi
I'm not by my WS so I cannot take a look, but is you are new to COMSOL, I could imagine you have missed the fact that for AC analysis, the best is to use frequency domain studies, of fixed frequency 60Hz and then you add in the amplitudes of your current. If you have several phases, you need to use the phasor complex representation to get the relative phases correct.
For Joule heating you need to define the convection too, either it's a known constant "h" and you do not need to model the air, or you have more complex cases in which case modelling air too (conjugated heat transfer could be required. But my advice, start simple.
So far COMSOL is showing very good correspondance for me, but I'm working mainly in the small scale << 1 dm^3, but I do not see why it should not give good results also for macro-scale ;)
However it takes some time to get aquinted to the COMSOl way, that is rather different from other, older, types of FEM programmes, as here you mix all physics under the same roof, hence some mothodological difference in seting up the physics. But I'm sure once used to it you will love it, at least I have never beem able to simulate so many complex systems in one tool before, and with a full control and knowledge of what I'm doing !
--
Good luck
Ivar
Hi
I'm not by my WS so I cannot take a look, but is you are new to COMSOL, I could imagine you have missed the fact that for AC analysis, the best is to use frequency domain studies, of fixed frequency 60Hz and then you add in the amplitudes of your current. If you have several phases, you need to use the phasor complex representation to get the relative phases correct.
For Joule heating you need to define the convection too, either it's a known constant "h" and you do not need to model the air, or you have more complex cases in which case modelling air too (conjugated heat transfer could be required. But my advice, start simple.
So far COMSOL is showing very good correspondance for me, but I'm working mainly in the small scale
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
22 apr 2013, 12:06 GMT-4
Thanks Ivar,
I do have the convective option already added, with the 5 entered for the coefficient.
Can you explain a little better what you mean by "add in the amplitudes of your current"?
Right now, the method I"ve just been using is to have one surface a terminal, with a specified current, and the other a ground. Anytime I do a frequency domain analysis for joule heating (using the actual joule heating module), I get nonsense results. For instance, if i run that model, I simply get the entire assembly being room temperature, which is clearly not correct.
Thanks Ivar,
I do have the convective option already added, with the 5 entered for the coefficient.
Can you explain a little better what you mean by "add in the amplitudes of your current"?
Right now, the method I"ve just been using is to have one surface a terminal, with a specified current, and the other a ground. Anytime I do a frequency domain analysis for joule heating (using the actual joule heating module), I get nonsense results. For instance, if i run that model, I simply get the entire assembly being room temperature, which is clearly not correct.
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
23 apr 2013, 04:45 GMT-4
Hi
having taken a rapid look at your model, indeed
1) you should use Frequency - Stationary for JH else the thermal transfer is not done correctly and you will only see the average voltage drop. Check the derived values
2) the two first stationary and time dependent (do use a power of 2 type progression with intermediate time stepping, its cleaner) are DC current so the power is twice the value for the frequency domain one
3) I would suggest that you remove the convective cooling from the internal surfaces for the boltings
4) you are missing the temperature of the GND and the Terminal surfaces, these are connected to some copper wire, I would expect, and their temperature will influence greatly
5) in "reality" you have a thermal thin resistive layer at the adjacent surfaces but it's not evident what kind of heat resistance to use, that depends greatly on the surface finish, it can easily be 0.1-0.3 K/W
--
Good luck
Ivar
Hi
having taken a rapid look at your model, indeed
1) you should use Frequency - Stationary for JH else the thermal transfer is not done correctly and you will only see the average voltage drop. Check the derived values
2) the two first stationary and time dependent (do use a power of 2 type progression with intermediate time stepping, its cleaner) are DC current so the power is twice the value for the frequency domain one
3) I would suggest that you remove the convective cooling from the internal surfaces for the boltings
4) you are missing the temperature of the GND and the Terminal surfaces, these are connected to some copper wire, I would expect, and their temperature will influence greatly
5) in "reality" you have a thermal thin resistive layer at the adjacent surfaces but it's not evident what kind of heat resistance to use, that depends greatly on the surface finish, it can easily be 0.1-0.3 K/W
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
23 apr 2013, 10:00 GMT-4
Thanks for taking a look Ivar,
1)Understood.
2)Is there anyway I could use time dependant, but in the frequency/stationary studies? Part of our testing involves heat cycling (current on for an hour, current off for an hour etc etc), and that's something I was trying to see if we could get in comsol. But I suppose technically, the frequency/stationary solution should tell us the worst case end state, correct?
3) Done, and understood.
4) In this test, we're connected to some pads hooked up directly to a transformer. Everything starts at ambient temperature, so I was trying to sort of simulate everything from a cold start up, if that makes sense. In general, we do basically two different styles of current cycling tests, but the test articles can be vastly different, so again, I was trying to sort of create a generic "Template" on how to set the model up.
5)By adjacent surfaces, I assume you mean the interfaces where the parts are touching each other? Also, would that then also apply to the exposed pad surfaces that I'm using for terminal and ground?
Again, thanks for helping to answer my questions.
Thanks for taking a look Ivar,
1)Understood.
2)Is there anyway I could use time dependant, but in the frequency/stationary studies? Part of our testing involves heat cycling (current on for an hour, current off for an hour etc etc), and that's something I was trying to see if we could get in comsol. But I suppose technically, the frequency/stationary solution should tell us the worst case end state, correct?
3) Done, and understood.
4) In this test, we're connected to some pads hooked up directly to a transformer. Everything starts at ambient temperature, so I was trying to sort of simulate everything from a cold start up, if that makes sense. In general, we do basically two different styles of current cycling tests, but the test articles can be vastly different, so again, I was trying to sort of create a generic "Template" on how to set the model up.
5)By adjacent surfaces, I assume you mean the interfaces where the parts are touching each other? Also, would that then also apply to the exposed pad surfaces that I'm using for terminal and ground?
Again, thanks for helping to answer my questions.
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
23 apr 2013, 13:55 GMT-4
Hi
2) check "frequency domain - time dependent" solvers
4) the test pads have some mass and thermal inertia too, so they will strongly influence the temperature evolution, so either you can measure their temperature, or you need to include these parts too, material, geometry, interfaces ...
5) certainly, all surfaces in contact will have a loss in contact surface, polished finish might give a "perfect contact" this contact resistance is very highly pressure and surface finish dependent, difficult to estimate, almost easier to simulate from detailed experimental results
--
Good luck
Ivar
Hi
2) check "frequency domain - time dependent" solvers
4) the test pads have some mass and thermal inertia too, so they will strongly influence the temperature evolution, so either you can measure their temperature, or you need to include these parts too, material, geometry, interfaces ...
5) certainly, all surfaces in contact will have a loss in contact surface, polished finish might give a "perfect contact" this contact resistance is very highly pressure and surface finish dependent, difficult to estimate, almost easier to simulate from detailed experimental results
--
Good luck
Ivar