FLUX3D - magnet segmentation

Ruud

Dear all,

I am trying to simulate a segmented magnet in Flux3D. To this extent, I have divided the magnet into two volumes, and assigned a (different) volume region to each of the volumes. The properties of both volume regions are the same (solid conductor region with magnet material; linear Br and relative permeability; linear resistivity). Further, I defined a face region at the face that interconnect the two magnets, in order to impose an insulating layer. I have tried different settings for this face region (air gap region, magnetic non-conducting region with mu_r of 1, perfect insulator in conduction medium), but in all cases I get an error message:

"The coupling with neighbouring volume regions of faces:
27
of face region REGIONFACE_1 is forbidden."

for airgap and magnetic non-conducting region and:

"The electric insulator surface region REGIONFACE_1 , on faces:
is not edge of volume regions with M*3TWSCRHJ or M*3TWOM
formulations"

for perfect insulator region. I have tried the same approach in Flux2D (with air gap region), where it worked perfectly. Can you explain why I am getting this error message in Flux3D? Is there another method to include the insulation layer between the magnet segments without modeling it as an actual volume region?

Thanks in advance for your reply!

Best regards,

Ruud

SimonGuicheteau

Hello,

the correct way should be with the insulation layer. As you explained, it needs a separate volume region on each side and the face(s) between these regions should be an insulation. In case of multiple segmentations, it should be one separate insulation per couple of touching magnets.

In your case, I don't know what happens. Maybe the insulation is touching the boundary of the domain ?

Ruud

Hi Simon,

Thanks for your reply. The edges of the insulation face region are touching on a non-conducting magnetic region with a low permeability value. Could this be a problem?

Best regards,

Ruud

Ruud

The problem only seems to happen when I apply an insulating face region on the boundary face between two conduction regions. When I change the volume region description of one of the magnets to 'non-conducting magnetic region' instead of 'solid conductor region', the model solves without issues. Apparently, it is not allowed to define an insulating face region on the boundary face between two solid conductor region, but i do not understand why.

Alexandru-Ionel Constantin

Hi Ruud,

 

There is a method to include the insulation layer between the magnet segments defined as solid conductor regions without modelling it as an actual volume region between them.

 

You need to change the Approximating functions type for vector potential from nodal finite elements to Edge finite elements in the Application settings, under Formulation model.

 

Best regards,

Alexandru

Ruud

Hi Alexandru,

Thanks for your reply. I have tried setting the approximating functions type to edge finite elements but it does not solve the issue.

In fact, I found out this morning that I have to force Flux3D to use the magnetic vector potential formulation (instead of magnetic scalar potential) to obtain correct results for both the magnetic flux density and the induced current density in an unsegmented magnet. However, with this formulation selected is seems to be impossible to create a face region on the boundary between the two magnet regions. As shown in the picture below, Flux3D does not seem to have any suitable formulation for the definition of a face region when the magnetic vector potential is used as model formulation. Do you know a solution to this issue?

Best regards,

Ruud

SimonGuicheteau

Hello,

you should not have to change any of the formulation to use segmented magnets. All automatic options which should be edge elements and scalar formulations in 3D should be ok.

Can you show how the model is set up ? what region is touching what ? Where is the periodicity if there is one ? 

Best regards

Ruud

Hi,

To make things more clear, I have included the model to this post. The model solves correctly using the magnetic vector potential formulation (both the resulting flux density distribution and the induced current density seem correct). If I switch to scalar potential formulation, the induced current is not calculated correctly (will be practically zero).

As you can see, the electric insulation between in magnet segments is currently not taken into account, as I cannot find a way to define it in the magnetic vector potential formulation case. Any help to overcome this issue is appreciated.

Thanks in advance and best regards,

Ruud

Ruud

Btw, the purpose of this model is to isolate the magnet loss calculation from the full motor simulation. The time variation of the flux density is taken into account through a time-varying Br value for the magnet, whereas the high permeability layer around the magnet region is required to force the magnetic flux density in the magnet to the same value as the applied Br (effectively short-circuiting the magnet).

SimonGuicheteau

I confirm my previous answer. If you switch back the settings to automatic formulations, and define an insulation layer in the face between the two magnets, it works perfectly fine.

You should not have to play with formulations.

Ruud

Indeed, it solves perfectly when you switch back to the automatic formulation and add the insulation layer. However, the amplitude of the induced current in NOT correct in this case. Take a look at the 3D model results below, where current density is just a few A/m2 (corresponding magnet loss is in order of 1E-11).

I performed a 2D simulation for an axial slice of the same geometry, which results in the following induced current distribution:

As you can see, there is 5 orders of magnitude difference between the 2D and 3D simulation. I know that there will be some difference between 2D and 3D results, but it should not be this much. Further, the magnet losses calculated by the 2D model are in a realistic order of magnitude, so I can only conclude that the induced current calculated by the 3D model is incorrect.

Ruud

Also, if I switch to the magnetic vector potential formulation in the 3D model (ignoring the segmentation for now), the order of magnitude of induced current is the same as for the 2D simulation: