Discussion Forum

Hi All,

i got the curve of one periode of magnetizing in B vs H.

The BH Area could be caculated in primitiv way. comsol has rmm.Wm calculated. by using photoshop i can evaluate the relation (c=Ahyst/Amax) between closed loop area and the maximal area of B*H (instead of rmm.Wm).

Whyst [J/m³] = c * Wmax

Physt [W]= Whys * freq * Volume

any comments are welcome.

regards
akmal

Hi Akmal,

I do not know a lot about hysteresis models, but wouldn't it be possible to determine the the hysteresis losses by first doing an analysis using a material model without hysteresis, for instance by using a constant mur. And then using a material model that uses the HB curve and does model hysteresis.

If you use the coil current and the coil voltage to calculate the electrical power in both cases the difference between the analyses should be the hysteresis losses, right?

Regards,
Floris

Akmal,

I am also very interested in modeling hysteresis losses in magnetic materials. I had read that many people have successfully implemented the Preisach model in COMSOL, for example, they mention it in this paper:

www.comsol.com/papers/2604/download/Sergeant.pdf

I also am having trouble understanding how to model hysteresis losses in the simplest manner possble.
Any help is much appreciated!


Thanks,
Daniel

Hi All,

thx for your suggestion. have also consider separate modell only to evaluate hysteresis loss. B and H curve on stator tooth can be applied in model also. i'll try to share the modell only for hysteresis losses soon.

"If you use the coil current and the coil voltage to calculate the electrical power in both cases the difference between the analyses should be the hysteresis losses, right?"
No, the difference is not only hysteresis loss but also eddy current loss (induced current in the soft iron).

To Daniel Harburg, tell me about your model. it is a rotating machine or something else?

regards
akmal

Akmal,

Thanks for the reply. I am trying to model hysteresis losses in a laminated magnetic material layers which are separated by epoxy. I have attached the model that I have built so far. This model is of a small section of a toroidal inductor, so there are copper turns both above and below the magnetic material. We are trying to look at eddy current and hysteresis losses in the magnetic layers (which right now I am modeling just as soft iron).

I have implemented the BH curve in my model, but that does not produce any hysteresis loss since there is no coercivity information for the BH curve (or no BH loop with a lossy area).

I would be very interested to see your model which models hysteresis loss because I don't understand how I can model this type of loss for this model. Do I need to use the real and imaginary parts of permeability, or is there another way?


Thanks for all the help.

Daniel

Hi Daniel,

attached is hysteresis model using Jiles Atherton Method JAM.

regards
akmal

Hi Akmal,
Thanks for posting your model. I don't understand the expression for dMirrdt though. Could you explain it or give a reference?
Thank you!
Patrick

Hi Akmal,

Don't bother. I found A.J.Bergquist, IEEE trans. of Magnetics, vol 32. No5 1996, that explains it.

Thanks again for your instructive model.

Patrick

Hi Patrick,
Can you share that paper? I also need to study this.

Tung

Hello Tung,
There are lots of available documents on the web that cover the Jiles-Atherton Model. But one of the better papers is “An Inverse Jiles-Atherton Model to take into Account Hysteresis in Time-Stepping Finite-Element Calculations” by N. Sadowski, et. Al., IEEE Transactions on Magnetics, Vol. 38, No. 2, March 2002. With this document you can easily follow the hysteresis example attached above. I also modified the example to make it fully 2-D. I hope COMSOL will officially incorporate this type of scheme in their AC/DC Module.
Trevor

Hi,

I am working on an academic project, and trying to simulate magnetic hysteresis loss with the Jiles-Atherton method. I am using the model hysteresis_example_2D.mph as a basis. It's running fine with default parameters on Comsol 4.2a :
freq = 50Hz
1A<I<5A
a 560 A/m
Ms 1.6e6 A/m
k 1200
c 0.1
alpha 0.0007

But I get convergence problems (unsolvable system) when I modify parameters such as frequency and current amplitude.

I change frequency to 50kHz in parameters definition. I0_Coil is now defined with :
1[A]+2[A]*(t>0.04e-3)+2[A]*(t>0.08e-3)
I also adjust the time interval used for simulation like this.
range(0,5.0e-7,1.2e-4)
Now simulation is still ok. It runs fine.

But if I change current amplitude like that:
I0_Coil is now defined with : 1[A]+10[A]*(t>0.04e-3)+10[A]*(t>0.08e-3)
I assumed, no smoothing of signal was necessary because its a sine wave...

I got error :

Repeated error test failures. May have reached a singularity.
Time : 4.336672761686902e-005
Undefined value found.
Undefined value found in the stiffness matrix..
There are 6 equations giving NaN/Inf in the matrix rows for the variable mod1.H1x.
at coordinates: (-0.005,0.00885346), (-0.005,0.00942673), (-0.005,0.01), (-0.00461831,0.00963079), (-0.00423662,0.00926159), ...
There are 6 equations giving NaN/Inf in the matrix rows for the variable mod1.H1y.
at coordinates: (-0.005,0.00885346), (-0.005,0.00942673), (-0.005,0.01), (-0.00461831,0.00963079), (-0.00423662,0.00926159), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.
Last time step is not converged.

What can be done to solve this ? I tried reducing maximum time step in time-dependent solver but no success.

Thanks in advance.

D. Barthélémy

Hi, everyone. I am working in my PH degree and I need your help to implement the Jiles-Atherton model in Comsol.
I am trying to model ferromagnetic materials using the "hysteresis_example_2D.mph" example posted here, but I obtain "funny" results. My model is only a rectangle (ferromagnetic plate) and three circles (three copper conductors with balanced currents). I say I have "funny" results because is something strange that for balanced currents (shifted 0º, -120º and 120º) I obtain a "Failed to find consistent initial valaues" error, but if they are not shifted (0º, 0º and 0º) I have a correct solution. Please, any suggestions?

Thanks in advance.

Hi Trevor,

I am also trying to solve loss problem. I am going through your model. I have got few doubts,

1> Can you please elaborate the use of flshms and the expression
smoothRamp(Mrevx*d(Hex,t)+Mrevy*d(Hey,t),delta)/(c*k)*if(normMrev>0,Mrevy/normMrev,0)
2> When I see the ode expression it is being differentiated with respect to t,
d(mf.Mx,t)-(1-c)*dMirrxdt-c*d(Manx,t)

However In the Sadowski paper it is differentiated with respect to Be.

Can you please explain it for me.

With best regards,
Gaurav