Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Simulation of current densities of an overhead line conductor wont't converge

Joachim Niemann-Larsen

Please login with a confirmed email address before reporting spam

Dear COMSOL community,

I've used COMSOL for some time, and this is the first time I’ve encountered a problem, hence this is my first post.

I am trying to model a stranded overhead line conductor of the type ACSR (aluminum conductor-steel reinforced) in 3D Magnetic Fields (mf), Frequency Domain, and I can't get my model to converge. For simplicity I've started out with a single steel core and one layer of helically stranded aluminum layer around the steel core. For the aluminum strands I use the Helix primitive in the Geometry. Form Union and Form Assembly have both been tried out (know the difference between these), and I've tried with and without any physically contact between the single strands.

For the coil domain, I've added coil Inputs to one end of the aluminum strands (see attached image) and Output to the other ends and added the necessary Coil Geometry Analysis under at first position under Study.

I've also tried including an air domain enclosing the simplified conductor (I guess this should be added for the magnetic flux to cause an induction in the steel core), but without any luck.

When performing the simulation, the computation takes +3 hours with only little progress, and when I cancel, I see a lot of warnings about MUMPS memory issues, even though my system has 16GB of RAM. Sometimes the simulation stops by itself with messaging it could not converge or same warning as above.

The meshing seems to be fine, and I've tried out various mesh configurations/divisions.

The problem seems to regard the Coil Geometry Analysis, which when run alone, seems to take hours without any convergence, so I'm guessing my definition of the Coil Domain might be wrong? It could also be the uninization in the geometry

I think my geometry is simple and since it is built in COMSOL I would expect it to converge, so I am doing something wrong in the creation of my model. I just want to see some induction of current in the steel core for now and was hoping to get it to work to build the complete conductor layout and examine the current density distributions of each layer.

Any help is greatly appreciated! Thanks.

My model is attached.



6 Replies Last Post 19.03.2018, 12:29 GMT-4
Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 15.03.2018, 16:40 GMT-4

Joachim,

  1. You don't have air domain enclosing you coil - so no physically maningfull results can be expected with Mahentic Fields physics interface without surrounding air.

  2. After adding air domain, make sure that you have closed loop for the current flow in your coil.

  3. Central steel wire is included in the coil domain but no current excitation is assigned to this wire, which doesnt'n make much physical sence.

Regards, Sergei

Joachim, 1. You don't have air domain enclosing you coil - so no physically maningfull results can be expected with Mahentic Fields physics interface without surrounding air. 2. After adding air domain, make sure that you have closed loop for the current flow in your coil. 3. Central steel wire is included in the coil domain but no current excitation is assigned to this wire, which doesnt'n make much physical sence. Regards, Sergei

Joachim Niemann-Larsen

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 15.03.2018, 17:03 GMT-4

Sergei,

Thank you very much for replying.

  1. I will include again the air cylinder around my stranded conductor.
  2. Should the length of the air domain be exactly the length of the conductor, so that I apply Input and Output at the cross-sections of the conductor right at the Magnetic Insulation (made up of the air cylinder of same length)?
  3. Nicely spotted; I will remove Coil Domain from center steel wire, however, I am sure that this could be a mistake in this attachment only (sorry), and that I have tried without the center wire included in the Coil Domain. Will re-run simulation.
Sergei, Thank you very much for replying. 1. I will include again the air cylinder around my stranded conductor. 2. Should the length of the air domain be exactly the length of the conductor, so that I apply Input and Output at the cross-sections of the conductor right at the Magnetic Insulation (made up of the air cylinder of same length)? 3. Nicely spotted; I will remove Coil Domain from center steel wire, however, I am sure that this could be a mistake in this attachment only (sorry), and that I have tried without the center wire included in the Coil Domain. Will re-run simulation.

Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 16.03.2018, 09:05 GMT-4

Joachim,

Yes, if you set the length of the air cylinder equal to the length of the conductor, then closing current loop will be formed by the Magnetic Insulation boundary condition which provides the surface current return path, as explaned here:

https://www.comsol.com/blogs/computing-the-inductance-of-a-wire/

In your original model, air conductivity is set to zero. Set it to some small number to get solver convergence.

Regards,

Sergei

Joachim, Yes, if you set the length of the air cylinder equal to the length of the conductor, then closing current loop will be formed by the Magnetic Insulation boundary condition which provides the surface current return path, as explaned here: https://www.comsol.com/blogs/computing-the-inductance-of-a-wire/ In your original model, air conductivity is set to zero. Set it to some small number to get solver convergence. Regards, Sergei

Joachim Niemann-Larsen

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 17.03.2018, 10:10 GMT-4

Sergei,

Thank you again!

I have performed those things now, however, my simulation still won't converge. When I compute, the simulation takes hours, and it will return with the following error:

Failed to find a solution. Out of memory during LU factorization. Returned solution is not converged. Not all parameter steps returned.

I can't really seem to figure out the problem with my model - everything seems to be correctly set up.

I've added my COMSOL model - maybe you can see some discrepancies within the model?

Again, any suggestions are highly appreciated.

Sergei, Thank you again! I have performed those things now, however, my simulation still won't converge. When I compute, the simulation takes hours, and it will return with the following error: > Failed to find a solution. > Out of memory during LU factorization. > Returned solution is not converged. > Not all parameter steps returned. I can't really seem to figure out the problem with my model - everything seems to be correctly set up. I've added my COMSOL model - maybe you can see some discrepancies within the model? Again, any suggestions are highly appreciated.


Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 19.03.2018, 08:54 GMT-4

Joachim,

Not sure if it is good ided to use Coil feature for multiple domains connected to each other along the edges. Attached is converged model using (mef) physics interface.

Regards,

Sergei

Joachim, Not sure if it is good ided to use Coil feature for multiple domains connected to each other along the edges. Attached is converged model using (mef) physics interface. Regards, Sergei


Joachim Niemann-Larsen

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 19.03.2018, 12:29 GMT-4

Sergei,

Okay, I see.

And your model seems to converge just fine, yes - thanks.

The magnetic flux density norm plot that you've added is implemented with log10 - why is that? And why can't I just chose mef.normB and it will plot similar to mf.normB (magnetic fields)? In mef it all just seems close to zero.

Sergei, Okay, I see. And your model seems to converge just fine, yes - thanks. The magnetic flux density norm plot that you've added is implemented with log10 - why is that? And why can't I just chose mef.normB and it will plot similar to mf.normB (magnetic fields)? In mef it all just seems close to zero.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.