Need help on solving! Warning: linear system solving probably did not converge. Low convergence of Newton-raphson, check the results.
I am trying to solve a surface mounted permanent magnet motor model for torque in 3D transient magnetic environment and my model is a 1/6 slice of full model with full axial length (no symmetry but with anticyclic boundary conditions). One thing which is not normal is that this model only has one phase real end-winding connection while for the other faces, they are modelled by a rectangular volume region. All the 'conductor regions' are assigned to solid bar conductors with eddy current taken into account.
The problem occurs when I tried to solve this model. It can be run from the start, but will end in the middle of solving scenario (solving with mechanical set, full step is 120, and this solving scenario usually stops in step number 40) with the warning: 'Linear system solving probably did not converge. Low convergence of Newton-raphson, check the results. Low convergence of the solving process. If necessary, deactivate the looseness of the non-linear threshold in the solving options.'
Thanks for your help!
Answers
-
Hi Yangyu,
Did you try to deactivate the looseness of the threshold as suggested by the error message? check the link below for information how t change this
Do you know what linear solver used during solving. If it use an iterative solver (ICCG) you can try to force a direct solver (MUMPS). Although you need more memory to run with MUMPS, sometime it converge better than ICCG.
Otherwise, it will be difficult to debug the issue without the model. If you can please create a support ticket and one of Altair AE can help more with this.
Thanks
0 -
Mohammed Elamin_22169 said:
Hi Yangyu,
Did you try to deactivate the looseness of the threshold as suggested by the error message? check the link below for information how t change this
Do you know what linear solver used during solving. If it use an iterative solver (ICCG) you can try to force a direct solver (MUMPS). Although you need more memory to run with MUMPS, sometime it converge better than ICCG.
Otherwise, it will be difficult to debug the issue without the model. If you can please create a support ticket and one of Altair AE can help more with this.
Thanks
Thank your for you reply, I will have a try of deactivating the looseness of threshold. Also, I think there is something wrong with physical setting or geometry and I will have a look at it.
0