Discussion Forum

So, now I've tried to rebuild a 2D-Model i've found in 3D, to see if the PDEs still work, but I always get the Error Message: "Failed to find consistent Initial Values". I have attached both the 2D and the 3D Model. It would be great if someone could tell me what I'm doing wrong.

regards
Rook

Hi Rook,

I just loaded "hysterese_example.mph" into versions 4.3.2.189 and 4.4.0.195 and it
works for me right away. See picture. I'am running Comsol on Linux.

Your PDE is simply f=0, right? Maybe you could try the "dode" interface.

Regards,

Jens

Hi Jens,

hysterese_example.mph is a model I found in another Dicussion here, so of course it works ;)
My Problem is, that I have to build my model in 3D, but if I do so, I get error Messages or the Hysteresis-Curve looks weird. To test, if I can use the Differential Equations from hysteresis_example.mph in 3 Dimensions, I built hystersis_example_3D (which is basically the same as hysteresis_example.mph) but it doesn't work... Did you take a look at hysteresis_example_3D too? Does it work for you?

Also, I tried to use the "dode"-Interface just now, but it didn't work either.
I got the Error-message:

Nonlinear solver did not converge.
Maximum number of Newton iterations reached.
There was an error message from the linear solver.
The relative error (1.6e+006) is greater than the relative tolerance.
Time : 0
Last time step is not converged.

Hi Rook,

there are still some issue but I got running for ~75 time steps now:

Main points:
+ when using the direct solver you need "Gauge Fixing for A-field" under the mf-node
+ for transient calculation you need el. conductivity != 0 in all domains (for instance 1 S/m)

Also I gave the nonlinear solver more iterations.

The simulation is still somewhat bumpy and you will need to fine tune solver parameters
but I hope this gives you a start.

Regards

Jens

Hi Jens,

Thank you so much for your help!
It's still not perfect, but It basically works now :)

I attached the Hysteresis Curve I produced :)

Regards
Rook

So, now I've run into a new Problem. Since the Model I actually want to study doesn't have a Zylindrical symmetry, it doesn't really make sense to use zylindrical Coordinates like in Hysteresis_example_3D.mph.
But when I try to rebuild this Model in karthesian Coordinates, the simulation doesn't work. (I actually don't even get an Error Message, the Progress-Report simply stays at 0 %). My guess is, I might have done somthing wrong with the PDEs, but I'm not sure...

I attached the new Model. It would be great if someone could help me figure out what I did wrong this time :)

Regards
Rook

Hi

it is working for me, but convergence is slow. The time steps are so small that the
progess meter stays at 0% for a very long (wall clock) time. But if I look at
the log-window I see progress.

Regards

Jens

Hi

That's exactly whats confusing me. Why does the Model with the cartesian Coordinates take such a long time, while the Model with the zylindric Coordinates is done in half a day? Is this normal, or does it mean theres some mistake in the cartesian Model? I don't really want to run a week long simulation, just to get a nonsense-result due to a falty model.
Is there a way, to make a Simulation run faster, or at least to check, if it works the way you intended to?

regards
Rook

Hi Rook,
I have the same problem with 3D cartesian model with magnetic hysteresis effect. Did you reach any success on your convergency issue?
Regards,
Parnia

Hi Parnia,

unfortunately, I did not have any succes so far. By now im pretty sure, that there is somthing wrong with my equations. I have actually also tried different variations of the JA-Model, but so far non of it works.

Do you use a similar Model as I am, or have you tried a different approach?

regards Rook