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.
Eddy_Currents_In_PM
Posted 11.03.2016, 09:46 GMT-5 Low-Frequency Electromagnetics, Studies & Solvers Version 4.4, Version 5.1 3 Replies
Please login with a confirmed email address before reporting spam
Dear all COMSOL experts,
I am simulating a Synchronous Surface Mounted Permanent Magnet Machine. My aim is to estimate the eddy currents within the permanent magnets placed on the rotor surface, since my machine operates at high frequency and runs at high speed (17000 rpm). I have four different permanent magnets radially magnetized.
Since my simulation is a 2D one I have to force the following condition in order to reflect fairly the reality: the net current, i.e. the integral over the magnet cross section, must be zero for each time step. I tried to enforce it establishing a scalar global equation defining the current computed as an integral over the magnets cross section. After this step I impose a surface current density exactl equal to the opposite of the current divided by the magnet area.
I performed two simulations: firstly the permanent magnets are demagnetized (just a conductive material attached to the rotor) and the winding is excited with a current (just checking the eddy currents generated solely by the armature reaction on the stator); secondly I magnetize the magnets and "switch off" the windings (no load test).
It turns out that in the first case the simulation is working fine, i.e. it converges without problems, whereas in the second case there is no way the time stepping converges. If I choose an adaptive time step (free) the solver computes several time steps (very small compared with the actual time step subdivision I impose in the range visualization of the time stepping settings) and after some hundreds it stops and signal a convergence error. If I force a manual time step it is not even capable to compute the first time step, even if it is able to converge in the first stationary simulation which is imposed before the time stepping one.
I also tried the Single Turn Coil imposing zero current instead of defining a scalar equation for the current and imposing an external current density. No way it works.
In the time stepping solver configuration I tried to relax some error conditions, like posing Exclude algebraic instead of include. In this way, utilizing the Single Turn coil it is able to compute the time stepping. Nonetheless the results are not reliable since the current density integral over the airgap is not null anymore.
Basically I can see that when I magnetize the magnets (the machine is "magnetically charged") the time stepping is not able to go on in the computation. I tried several non linear materials iron curves, even with hundreds of points for definition. No way it works.
Has anyone else encountered the same issue before?
If you have any suggestions about this point I will be immensely grateful.
Sincerely,
Nicola
I am simulating a Synchronous Surface Mounted Permanent Magnet Machine. My aim is to estimate the eddy currents within the permanent magnets placed on the rotor surface, since my machine operates at high frequency and runs at high speed (17000 rpm). I have four different permanent magnets radially magnetized.
Since my simulation is a 2D one I have to force the following condition in order to reflect fairly the reality: the net current, i.e. the integral over the magnet cross section, must be zero for each time step. I tried to enforce it establishing a scalar global equation defining the current computed as an integral over the magnets cross section. After this step I impose a surface current density exactl equal to the opposite of the current divided by the magnet area.
I performed two simulations: firstly the permanent magnets are demagnetized (just a conductive material attached to the rotor) and the winding is excited with a current (just checking the eddy currents generated solely by the armature reaction on the stator); secondly I magnetize the magnets and "switch off" the windings (no load test).
It turns out that in the first case the simulation is working fine, i.e. it converges without problems, whereas in the second case there is no way the time stepping converges. If I choose an adaptive time step (free) the solver computes several time steps (very small compared with the actual time step subdivision I impose in the range visualization of the time stepping settings) and after some hundreds it stops and signal a convergence error. If I force a manual time step it is not even capable to compute the first time step, even if it is able to converge in the first stationary simulation which is imposed before the time stepping one.
I also tried the Single Turn Coil imposing zero current instead of defining a scalar equation for the current and imposing an external current density. No way it works.
In the time stepping solver configuration I tried to relax some error conditions, like posing Exclude algebraic instead of include. In this way, utilizing the Single Turn coil it is able to compute the time stepping. Nonetheless the results are not reliable since the current density integral over the airgap is not null anymore.
Basically I can see that when I magnetize the magnets (the machine is "magnetically charged") the time stepping is not able to go on in the computation. I tried several non linear materials iron curves, even with hundreds of points for definition. No way it works.
Has anyone else encountered the same issue before?
If you have any suggestions about this point I will be immensely grateful.
Sincerely,
Nicola
3 Replies Last Post 18.05.2017, 04:08 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
8 years ago
Hi Nicola,
are you doing a stationary study to provide good initial values for the time-dependent study?
Are you ramping up the rotational speed or are you starting at full speed? Ramping up can help.
Cheers
Edgar
--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
are you doing a stationary study to provide good initial values for the time-dependent study?
Are you ramping up the rotational speed or are you starting at full speed? Ramping up can help.
Cheers
Edgar
--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Please login with a confirmed email address before reporting spam
Posted:
8 years ago
Hi Edgar,
yes. I solve a stationary magnetostatic problem and I use the results of magnetic vector potential as initial value of the time stepping simulation.
Unfortunately is still not working. I do not really do how to cope with this issue.
Sincerely,
Nicola
yes. I solve a stationary magnetostatic problem and I use the results of magnetic vector potential as initial value of the time stepping simulation.
Unfortunately is still not working. I do not really do how to cope with this issue.
Sincerely,
Nicola
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Hi Nicola,
Have you fixed this problem? I also have this problem when calculating eddy current losses in permanent magnets. I have tried both single-turn-coil and external-opposite-current-density to keep the net current in one magnet zero. As you mentioned, the time step became very small and after the computation time of some microseconds, it did not converge. Do you already have a solution to your own problem? If yes, it may help my model as well. Could you please share your solution with me?
Regards,
Dong
Have you fixed this problem? I also have this problem when calculating eddy current losses in permanent magnets. I have tried both single-turn-coil and external-opposite-current-density to keep the net current in one magnet zero. As you mentioned, the time step became very small and after the computation time of some microseconds, it did not converge. Do you already have a solution to your own problem? If yes, it may help my model as well. Could you please share your solution with me?
Regards,
Dong