######## WORK NOTES COPIED FROM INCIDENT 349043 ########
--------------------------------------------------------------



QUESTION:
I modeled a multiply connected magnetic region with SOLID96 and SOLID117 (C core with a permanent magnet in the airgap). In both models, all boundaries were made flux parallel. Results differs substantially. The SOLID96 MSP predicts circulation of flux around the core and through the PM. The SOLID117 edge-flux predicts flux emerging from one PM pole doubling bacK through adjacent "air" into the other PM pole, with NO circulation in the core. A 2D MVP model agrees with the 3D SOLID96 model. Why does the SOLID117 predict these seemingly counterintuitive results? Models are attached to this solution.

ANSWER:
In Emag problems with edge-flux elements (SOLID117),
the "flux parallel" boundary condition (AZ=0) can do more than set a local
condition that B is parallel to the boundary. If applied on surfaces forming
a multiply connected region, it also creates a global condition that there
is zero net flux through any area for which the boundary lies entirely in
those surfaces. The reason is obviously that setting AZ=0 for a set means the
line integral of A over any chain of nodes belonging to the set is zero.

DETAILS:

In the attached input file:

method=1 ! set method=0 to use edge-flux, 1 to use scalar potential
multconn=1 ! set multconn=1 to make hole in air (and make the boundary multiply connected)

! The following combinations give the results an uninformed user would expect:
! method=0 , multconn=0
! method=1 , multconn=0
! method=1 , multconn=1

! The combination method=0, multconn=1 gives "incorrect" (or at least misleading) results
! since it's actually a very different physical system that gets modelled.


NOTES:
Not yet clear what action, if any, shall be taken regarding element formulation.


######## WORK NOTES COPIED FROM INCIDENT 349043 ########
--------------------------------------------------------------



QUESTION:
I modeled a multiply connected magnetic region with SOLID96 and SOLID117 (C core with a permanent magnet in the airgap). In both models, all boundaries were made flux parallel. Results differs substantially. The SOLID96 MSP predicts circulation of flux around the core and through the PM. The SOLID117 edge-flux predicts flux emerging from one PM pole doubling bacK through adjacent "air" into the other PM pole, with NO circulation in the core. A 2D MVP model agrees with the 3D SOLID96 model. Why does the SOLID117 predict these seemingly counterintuitive results? Models are attached to this solution.

ANSWER:
In Emag problems with edge-flux elements (SOLID117),
the "flux parallel" boundary condition (AZ=0) can do more than set a local
condition that B is parallel to the boundary. If applied on surfaces forming
a multiply connected region, it also creates a global condition that there
is zero net flux through any area for which the boundary lies entirely in
those surfaces. The reason is obviously that setting AZ=0 for a set means the
line integral of A over any chain of nodes belonging to the set is zero.

DETAILS:

In the attached input file:

method=1 ! set method=0 to use edge-flux, 1 to use scalar potential
multconn=1 ! set multconn=1 to make hole in air (and make the boundary multiply connected)

! The following combinations give the results an uninformed user would expect:
! method=0 , multconn=0
! method=1 , multconn=0
! method=1 , multconn=1

! The combination method=0, multconn=1 gives "incorrect" (or at least misleading) results
! since it's actually a very different physical system that gets modelled.


NOTES:
Not yet clear what action, if any, shall be taken regarding element formulation.





Show Form
No comments yet. Be the first to add a comment!