QUESTION:
I have heard that the 3D magnetic vector potential (MVP) formulation used by SOLID97 fails to accurately predict the magnetic field at material interfaces across which there are large changes in magnetic permeability. This was the motivation for creating the edge flux potential formulation (SOLID117), which correctly predicts fields in such cases. However, a number of features available for SOLID97 (circuit coupling through CIRCU124, sliding interfaces via CEINTF, and companion infinite boundary elements) are not yet available for SOLID117. If I have an application requiring these features in a system with material interfaces, is there a workaround to correct the field error at material interfaces using SOLID97 so that I may use these features?
ANSWER:
An unsupported procedure has been known to work in at least some cases. A simple example (attached to this solution) in which a permanent magnet is placed along side a permeable bar is solved 4 different ways:
1) mdl02.inp: 3D MVP (SOLID97): field distribution and magnitude differs considerably from all other cases
2) mdl03.inp: SOLID117
3) mdl04.inp: SOLID98 magnetic scalar potential (MSP)
4) mdl02c.inp: SOLID97 using the workaround described below - field distribution and magnitude agrees well with cases 2 and 3. Input file calls macros mbdryMVP.mac and mbdryCP.mac.
The steps in the workaround may be summarized as follows:
a) rotate nodal degrees of freedom on the material interface so that the nodal x direction is normal to the interface
b) duplicate the iron mesh, discard the original mesh, retain the duplicate copy (no longer joined to the surrounding "air", but has duplicate nodes along the interface).
c) couple only tangential components (AY and AZ) of MVP on coincident pairs of nodes at the iron-air interface.
The attached presentation and input files illustrate that SOLID117 edge and SOLID98 MSP results agree and that SOLID97 results are incorrect unless the procedure summa
|
QUESTION: I have heard that the 3D magnetic vector potential (MVP) formulation used by SOLID97 fails to accurately predict the magnetic field at material interfaces across which there are large changes in magnetic permeability. This was the motivation for creating the edge flux potential formulation (SOLID117), which correctly predicts fields in such cases. However, a number of features available for SOLID97 (circuit coupling through CIRCU124, sliding interfaces via CEINTF, and companion infinite boundary elements) are not yet available for SOLID117. If I have an application requiring these features in a system with material interfaces, is there a workaround to correct the field error at material interfaces using SOLID97 so that I may use these features? ANSWER: An unsupported procedure has been known to work in at least some cases. A simple example (attached to this solution) in which a permanent magnet is placed along side a permeable bar is solved 4 different ways: 1) mdl02.inp: 3D MVP (SOLID97): field distribution and magnitude differs considerably from all other cases 2) mdl03.inp: SOLID117 3) mdl04.inp: SOLID98 magnetic scalar potential (MSP) 4) mdl02c.inp: SOLID97 using the workaround described below - field distribution and magnitude agrees well with cases 2 and 3. Input file calls macros mbdryMVP.mac and mbdryCP.mac. The steps in the workaround may be summarized as follows: a) rotate nodal degrees of freedom on the material interface so that the nodal x direction is normal to the interface b) duplicate the iron mesh, discard the original mesh, retain the duplicate copy (no longer joined to the surrounding "air", but has duplicate nodes along the interface). c) couple only tangential components (AY and AZ) of MVP on coincident pairs of nodes at the iron-air interface. The attached presentation and input files illustrate that SOLID117 edge and SOLID98 MSP results agree and that SOLID97 results are incorrect unless the procedure summarized above (I call it the "mu boundary workaround"), is implemented. |
||
|