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 The information provided may be out of date.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.

Induction heating problem

Please login with a confirmed email address before reporting spam

I am trying to add the physics to heat the central wire using the coil, by supplying the voltage.

Please see and let me know where is mistake.


--
ishant.jain@tatasteel.com

3 Replies Last Post 28 mag 2012, 03:28 GMT-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 mag 2012, 09:15 GMT-4
Hi

first advice, start with one model at the time, get them running and then join them in into one new combined model.

In MEF you need to have some conductivity on all materials, use 1 or 10 S/m for the air, that will help. Then as you are already quite high in frequency, you should use a boundary layer at the surface of your Cu conductor, most of the current will naturally flow at the wire tube external surface and not within the bulk conductor. What is tricker is that square tubes tend to give singularitites, ideally they should be filleted, but that is not fully trivial in COMSOL.
I would also suggest to use a Terminal and not an Electric Potential boundary on "76", it will give you more handy postprocessing variables

You should use a boundary layer also on the steel outer surface where the eddy curents would run

In the HT I would suggest to remove the heat flux cooling from the two ends, and also on the internal thin air sections as there will be no cooling from inside, only heating of the air

And use a normal or fine mesh only, else you need some 64Gb of RAM

Finally, develop the default solver sequence and add a STORE solution and move it in between the Stationary and the Compile equations: Time dependent (if COMSOL has not added it by default)

Then it runs for me but that does not mean that the rests are correct, for that further validations are required, your voltage is probably a bit high, or the corresponding current in the 50kA region perhaps a bit high ;)

--
Good luck
Ivar
Hi first advice, start with one model at the time, get them running and then join them in into one new combined model. In MEF you need to have some conductivity on all materials, use 1 or 10 S/m for the air, that will help. Then as you are already quite high in frequency, you should use a boundary layer at the surface of your Cu conductor, most of the current will naturally flow at the wire tube external surface and not within the bulk conductor. What is tricker is that square tubes tend to give singularitites, ideally they should be filleted, but that is not fully trivial in COMSOL. I would also suggest to use a Terminal and not an Electric Potential boundary on "76", it will give you more handy postprocessing variables You should use a boundary layer also on the steel outer surface where the eddy curents would run In the HT I would suggest to remove the heat flux cooling from the two ends, and also on the internal thin air sections as there will be no cooling from inside, only heating of the air And use a normal or fine mesh only, else you need some 64Gb of RAM Finally, develop the default solver sequence and add a STORE solution and move it in between the Stationary and the Compile equations: Time dependent (if COMSOL has not added it by default) Then it runs for me but that does not mean that the rests are correct, for that further validations are required, your voltage is probably a bit high, or the corresponding current in the 50kA region perhaps a bit high ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 mag 2012, 11:12 GMT-4
Hi Ivar,

Thanks for your response. Earlier there was some mistake, we didn't put the frequency value ( initially) in MEF and I have changed the mesh size to normal meshing, now it is running( before your reply came) but it is not giving the expected result.

I need your help to tune the model so that I can have the expected result, please help me to understand :

What is the need of boundary layer at the surface of the Cu conductor ( It is coil basically that is hollow from inside to provide water jacket for the purpose of cooling)

Conductivity ? You must be talking about electrical conductivity. I have selected the properties from the library domain itself. Tell me the significance ?

Please explain this : "square tubes tend to give singularities, ideally they should be filleted, but that is not fully trivial in COMSOL"

Please clarify the use of terminal instead of boundary?

Boundary layer over the steel surface : Considering the significance of skin depth but the wire is helical one, so how it would be possible to apply the same ?

You are right , the voltage is considerably right, It is 8 V for the single turn. Could you please check the same model and revert back.

Rgrds,
Ishant

Hi

first advice, start with one model at the time, get them running and then join them in into one new combined model.

In MEF you need to have some conductivity on all materials, use 1 or 10 S/m for the air, that will help. Then as you are already quite high in frequency, you should use a boundary layer at the surface of your Cu conductor, most of the current will naturally flow at the wire tube external surface and not within the bulk conductor. What is tricker is that square tubes tend to give singularitites, ideally they should be filleted, but that is not fully trivial in COMSOL.
I would also suggest to use a Terminal and not an Electric Potential boundary on "76", it will give you more handy postprocessing variables

You should use a boundary layer also on the steel outer surface where the eddy curents would run

In the HT I would suggest to remove the heat flux cooling from the two ends, and also on the internal thin air sections as there will be no cooling from inside, only heating of the air

And use a normal or fine mesh only, else you need some 64Gb of RAM

Finally, develop the default solver sequence and add a STORE solution and move it in between the Stationary and the Compile equations: Time dependent (if COMSOL has not added it by default)

Then it runs for me but that does not mean that the rests are correct, for that further validations are required, your voltage is probably a bit high, or the corresponding current in the 50kA region perhaps a bit high ;)

--
Good luck
Ivar





--
ishant.jain@tatasteel.com
Hi Ivar, Thanks for your response. Earlier there was some mistake, we didn't put the frequency value ( initially) in MEF and I have changed the mesh size to normal meshing, now it is running( before your reply came) but it is not giving the expected result. I need your help to tune the model so that I can have the expected result, please help me to understand : What is the need of boundary layer at the surface of the Cu conductor ( It is coil basically that is hollow from inside to provide water jacket for the purpose of cooling) Conductivity ? You must be talking about electrical conductivity. I have selected the properties from the library domain itself. Tell me the significance ? Please explain this : "square tubes tend to give singularities, ideally they should be filleted, but that is not fully trivial in COMSOL" Please clarify the use of terminal instead of boundary? Boundary layer over the steel surface : Considering the significance of skin depth but the wire is helical one, so how it would be possible to apply the same ? You are right , the voltage is considerably right, It is 8 V for the single turn. Could you please check the same model and revert back. Rgrds, Ishant [QUOTE] Hi first advice, start with one model at the time, get them running and then join them in into one new combined model. In MEF you need to have some conductivity on all materials, use 1 or 10 S/m for the air, that will help. Then as you are already quite high in frequency, you should use a boundary layer at the surface of your Cu conductor, most of the current will naturally flow at the wire tube external surface and not within the bulk conductor. What is tricker is that square tubes tend to give singularitites, ideally they should be filleted, but that is not fully trivial in COMSOL. I would also suggest to use a Terminal and not an Electric Potential boundary on "76", it will give you more handy postprocessing variables You should use a boundary layer also on the steel outer surface where the eddy curents would run In the HT I would suggest to remove the heat flux cooling from the two ends, and also on the internal thin air sections as there will be no cooling from inside, only heating of the air And use a normal or fine mesh only, else you need some 64Gb of RAM Finally, develop the default solver sequence and add a STORE solution and move it in between the Stationary and the Compile equations: Time dependent (if COMSOL has not added it by default) Then it runs for me but that does not mean that the rests are correct, for that further validations are required, your voltage is probably a bit high, or the corresponding current in the 50kA region perhaps a bit high ;) -- Good luck Ivar [/QUOTE] -- ishant.jain@tatasteel.com

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 28 mag 2012, 03:28 GMT-4
Hi

indeed I was talking about the electric conductivity, with MEF you solve for A and V, the V means voltage and for voltage to flow you need some conductivity (ES physics solve for E electric field, not directly V and derive V from E, that doees not need a conducting air). So add a little conductivity, typically 1E-6 times your highest conductivity, for you ithe highest s copper so 1 to S/m conductivity in the air should do. you will get uA current flowing in tha air, but this is "peanuts" compared to the 50 kA in the coil, so your results remain valid, to the precision one can exect from mumerical simulations. But if you choose much less than 1E-6 you will get into numerical errors and the solver might not manage to find any solution due to numerical underflow, the binary representation of number has holes and leaks, I neveer trust anything some doze times smaller, in ratio, than sqrt(eps), or even (eps)(1/3) for higher discretisations. (check what sqrt(eps) give in your Parameter field).

Boundary layers: when you use AC current with higer frequencies (even about from 100 Hz) you current tend to flow only on the wire surface, because the e- are mutually pushing themselves to the outer edge due to the complex EM couplings (I current => B magnetic field => action on current vxB etc). So to make life easier use a boundary layer on the outer sourfaces of the copper-air interface OF YOUR CONDUCTOR. This als applies to the steel part as the induced eddy current will not penetrate the steel very far. Check thedoc about the skin depth, and run with different parameters with the model library ACDC models on skin effects, they are very instructive.With boundary layers in the critical regions, you can use a overall "normal" or even coarse mesh in the air. But try meshing the conductors first with normal or fine, then the air with coarse and then the boundary layers

For me your model solved, with your "square" tubes so runding iff is not really required, but if you want to study in detail, be aware that sharp edges make singularities, some dependent variables are defined as the spatial derivatives of the solved dependent variable, but such derivatives do not exist on sharp edges, they tend to infinity. THere is some true physics behind, if you connest a knife to a HV source you will see coronna effects principally emerging from the shaor edges and tip of the knife, but then you have air breakdown that limits the true voltage, in FEM you wil normally NOT see these coronna effects (deends onthe physics you add) but you will see the sharp edges that tend towards INF, and even making the mesh finer, you will see the edges with higher fields, but you will end up saturating the numerical solving process, again due to the limited resolution the IEEE binary number provides, we would need even higher resolution ALU in our chips, that will certainly come in a few years time too ...

With these changes it ould solve easily for you, try it out ;)


Start with a coarse or very coarse mesh, unti lit solves, then make the mesh finer and note how our results changes, eiterh until you hit your RAM limit, or until the overall results stabilises and do not change by more than 3-5% (apart from the singularities regions, but that again is whre you need the engineering "feeling" of the FEM user, and this you learn by doing ...
--
Good luck
Ivar
Hi indeed I was talking about the electric conductivity, with MEF you solve for A and V, the V means voltage and for voltage to flow you need some conductivity (ES physics solve for E electric field, not directly V and derive V from E, that doees not need a conducting air). So add a little conductivity, typically 1E-6 times your highest conductivity, for you ithe highest s copper so 1 to S/m conductivity in the air should do. you will get uA current flowing in tha air, but this is "peanuts" compared to the 50 kA in the coil, so your results remain valid, to the precision one can exect from mumerical simulations. But if you choose much less than 1E-6 you will get into numerical errors and the solver might not manage to find any solution due to numerical underflow, the binary representation of number has holes and leaks, I neveer trust anything some doze times smaller, in ratio, than sqrt(eps), or even (eps)(1/3) for higher discretisations. (check what sqrt(eps) give in your Parameter field). Boundary layers: when you use AC current with higer frequencies (even about from 100 Hz) you current tend to flow only on the wire surface, because the e- are mutually pushing themselves to the outer edge due to the complex EM couplings (I current => B magnetic field => action on current vxB etc). So to make life easier use a boundary layer on the outer sourfaces of the copper-air interface OF YOUR CONDUCTOR. This als applies to the steel part as the induced eddy current will not penetrate the steel very far. Check thedoc about the skin depth, and run with different parameters with the model library ACDC models on skin effects, they are very instructive.With boundary layers in the critical regions, you can use a overall "normal" or even coarse mesh in the air. But try meshing the conductors first with normal or fine, then the air with coarse and then the boundary layers For me your model solved, with your "square" tubes so runding iff is not really required, but if you want to study in detail, be aware that sharp edges make singularities, some dependent variables are defined as the spatial derivatives of the solved dependent variable, but such derivatives do not exist on sharp edges, they tend to infinity. THere is some true physics behind, if you connest a knife to a HV source you will see coronna effects principally emerging from the shaor edges and tip of the knife, but then you have air breakdown that limits the true voltage, in FEM you wil normally NOT see these coronna effects (deends onthe physics you add) but you will see the sharp edges that tend towards INF, and even making the mesh finer, you will see the edges with higher fields, but you will end up saturating the numerical solving process, again due to the limited resolution the IEEE binary number provides, we would need even higher resolution ALU in our chips, that will certainly come in a few years time too ... With these changes it ould solve easily for you, try it out ;) Start with a coarse or very coarse mesh, unti lit solves, then make the mesh finer and note how our results changes, eiterh until you hit your RAM limit, or until the overall results stabilises and do not change by more than 3-5% (apart from the singularities regions, but that again is whre you need the engineering "feeling" of the FEM user, and this you learn by doing ... -- Good luck Ivar

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.