Magnetic-mechanic coupling with non-conforming interfaces in 3D
Summary
In magnetic-mechanic simulation of a 3D problem with non-conforming interfaces the results are incorrect. The simulation time is also dramatically increased compared to the same setup with conforming mesh (1h for non-conforming vs 1min for conforming for the tested model).
Example Input or way to reproduce
The problem was encountered during a harmonic simulation of a transformer active part submerged in oil. The transformer active part is comprised of two cylindrical coils with isolating material between them and an iron core. One of the coils is under current. Due to rather complex geometry the non-coforming interfaces between active part and oil domain were introduced. Since the results seemed incorrect a conforming model has been also tested. The input files for both simulations are attached.
The simulation with non-conforming mesh (incorrect results):
- xml input mag_non_conf.xml
- mesh mesh_non_conf
The simulation with conforming mesh (correct results, used as reference):
- xml input mag_conf.xml
- mesh mesh_conf
Material file:mat.xml
What is the current bug behavior?
The simulation runs without errors, however the computation time is much longer than for the conforming mesh. The resulting magnetic flux vector field is not oriented as expected and has random spikes on individual elements. Furthermore the values of the flux overall are close to zero what contradicts the expectations. The picture of the resulting magnetic flux is attached.
What is the expected correct behavior?
The expected result would be a magnetic flux flowing through iron core as shown in the attached picture.
The same setup, but with conforming interfaces yields plausible results and runs approximately 60 times faster for the tested model.