Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
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.
Meshing and convergence problems
Posted Mar 28, 2013, 11:08 a.m. EDT Low-Frequency Electromagnetics, Geometry, Mesh, Modeling Tools & Definitions, Parameters, Variables, & Functions, Studies & Solvers Version 4.3a 8 Replies
Please login with a confirmed email address before reporting spam
Dear Comsolians,
I was finally able to draw the problem in 3D in Comsol :), am also a beginner, lol.
The way I have modeled the coils is that they do not form a complete closed loop, thus I assigned "External current density" as a form of excitation. Does this work?
When I was reading the documentation about "Multi-turn Coils" it said that the coils must terminate on external boundary? What is an external boundary and do I need to model closed loops for the currents to get a solution, because currently they are just open to the air?
Then I just added the mesh node and solved the problem, but unfortunately after 4days the problem did not converge :(.
NB: PC properties - Intel Core(TM) i5-2400 CPU@3.10Ghz 3.10GHz
Memory: 4GB on a 64 bit Windows 7 OS with graphics card NVIDIA Quadro 600.
Initially I had tried to mesh the thin soft iron region which is 0.3556mm in the middle with a normal mesh and the poles which are 20mm in depth with also a normal mesh, and the surrounding air with a coarse/coarser mesh, and I had the same convergent issues.
I think the transition from a thickness of 0.3556mm to 20mm introduces meshing issues.
I also have a lot filleted edges and this makes these regions to have a high element density.
So how do I model the coils and mesh the problem?
Regards and please find attached the model
John Wanjiku
I was finally able to draw the problem in 3D in Comsol :), am also a beginner, lol.
The way I have modeled the coils is that they do not form a complete closed loop, thus I assigned "External current density" as a form of excitation. Does this work?
When I was reading the documentation about "Multi-turn Coils" it said that the coils must terminate on external boundary? What is an external boundary and do I need to model closed loops for the currents to get a solution, because currently they are just open to the air?
Then I just added the mesh node and solved the problem, but unfortunately after 4days the problem did not converge :(.
NB: PC properties - Intel Core(TM) i5-2400 CPU@3.10Ghz 3.10GHz
Memory: 4GB on a 64 bit Windows 7 OS with graphics card NVIDIA Quadro 600.
Initially I had tried to mesh the thin soft iron region which is 0.3556mm in the middle with a normal mesh and the poles which are 20mm in depth with also a normal mesh, and the surrounding air with a coarse/coarser mesh, and I had the same convergent issues.
I think the transition from a thickness of 0.3556mm to 20mm introduces meshing issues.
I also have a lot filleted edges and this makes these regions to have a high element density.
So how do I model the coils and mesh the problem?
Regards and please find attached the model
John Wanjiku
Attachments:
8 Replies Last Post Apr 5, 2013, 7:47 p.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
John,
Current loop in the coil must be closed. In your model, current is injected from the bottom of cylindrical domain and exits top surface of the same domain – there is no closed loop. It means your setup violets Maxwell equations.
There are several models in Comsol library showing how to model coils. For example:
www.comsol.com/showroom/gallery/13777/
www.comsol.com/showroom/gallery/15/
Regards,
Sergei
Current loop in the coil must be closed. In your model, current is injected from the bottom of cylindrical domain and exits top surface of the same domain – there is no closed loop. It means your setup violets Maxwell equations.
There are several models in Comsol library showing how to model coils. For example:
www.comsol.com/showroom/gallery/13777/
www.comsol.com/showroom/gallery/15/
Regards,
Sergei
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks Sergei.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
check your geometry: you have a central overlapping area (double) with some very small edges joining them, as well as some lateral side domains, the are very thin and give you the meshing issue.
Use the View selection list to run through the domains
--
Good luck
Ivar
check your geometry: you have a central overlapping area (double) with some very small edges joining them, as well as some lateral side domains, the are very thin and give you the meshing issue.
Use the View selection list to run through the domains
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello Sergei,
I closed the loops and I initially made an electromagnetic pole wound with two coils at quadrature.
Then I assigned the Multi-turn Coil Domain to each coil and injected the current according to the procedure for Numeric coil type.
The above procedure is OK for a very low number of coils. I have 16 poles, which implies I have 32 coils to work with. So to ease the process I duplicated a pole with its coils 16 times, but when I plotted the arrow volume to show me the direction of current it looked really weird.
My question is do I need to treat each coil independently?
Regards
John Wanjiku
I closed the loops and I initially made an electromagnetic pole wound with two coils at quadrature.
Then I assigned the Multi-turn Coil Domain to each coil and injected the current according to the procedure for Numeric coil type.
The above procedure is OK for a very low number of coils. I have 16 poles, which implies I have 32 coils to work with. So to ease the process I duplicated a pole with its coils 16 times, but when I plotted the arrow volume to show me the direction of current it looked really weird.
My question is do I need to treat each coil independently?
Regards
John Wanjiku
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
have you checked the coil group nodes, and the related docs ?
--
Good luck
Ivar
have you checked the coil group nodes, and the related docs ?
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar,
I think it is mix up of the current densities directions when I duplicate and rotate the initial geometry that I use to create my model. Note I pre-assign all the conditions before duplicating to ease the work. Does it work that way? Because creating an insulator boundary and defining the input boundary for the currents is tedious for 32 coils, that is why i opted for 2 coils that make the initial geometry.
When I check the nodes everything seems ok including the current input boundary for each of the coils, but the results??
Regards and pls find attached the model
Wanjiku
I think it is mix up of the current densities directions when I duplicate and rotate the initial geometry that I use to create my model. Note I pre-assign all the conditions before duplicating to ease the work. Does it work that way? Because creating an insulator boundary and defining the input boundary for the currents is tedious for 32 coils, that is why i opted for 2 coils that make the initial geometry.
When I check the nodes everything seems ok including the current input boundary for each of the coils, but the results??
Regards and pls find attached the model
Wanjiku
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I'm working here with a colleague on 20 coils arranged along a sphere, indeed it's tedious but the only way.
Note that the orientation of the coil will change the direction of the edges hence the direction of the natural + current flow in the coil definitions, so you must turn on the View node: line ID and line arrows, thn set the EDGE view and look at the coil loops and find the direction of the highest edge ID (it seems to be the one giving the current flow direction for single or multiple coils BCs). Or once done get an arrow plot of the current directions and check the loop around in the correct directions.
Probabla if you observe well you can find the patten for the current direction, it would change I believe for the 4 quadrants
--
Good luck
Ivar
I'm working here with a colleague on 20 coils arranged along a sphere, indeed it's tedious but the only way.
Note that the orientation of the coil will change the direction of the edges hence the direction of the natural + current flow in the coil definitions, so you must turn on the View node: line ID and line arrows, thn set the EDGE view and look at the coil loops and find the direction of the highest edge ID (it seems to be the one giving the current flow direction for single or multiple coils BCs). Or once done get an arrow plot of the current directions and check the loop around in the correct directions.
Probabla if you observe well you can find the patten for the current direction, it would change I believe for the 4 quadrants
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello Ivar,
I had to treat each coil independently, it was the easiest way out and was able to see the flow of current by the arrow plot. Thanks for your advice.
I also resulted to circular instead of numeric since it doesn't require me to specify the input boundary.
I have noticed my model has symmetry, i.e. I can model a quarter or half the geometry. Its will be much more faster in terms of time and PC resources. But from experience with the coils, I shouldn't touch them, lol. I have managed to get a repeating geometric section, that repeats after 180 degs. In the quarter section I have to cut the coils and they will be open loops which will be bad for the physics, hence convergence.
I checked out the documentary, most of it is just definitions.
The examples given have the sources of the fields being at the boundary but am not sure I want to do that with the coils?
So how do I go about assigning BCs to give me a repeating section with a rotation of 180 degs.
Regards.
Wanjiku
I had to treat each coil independently, it was the easiest way out and was able to see the flow of current by the arrow plot. Thanks for your advice.
I also resulted to circular instead of numeric since it doesn't require me to specify the input boundary.
I have noticed my model has symmetry, i.e. I can model a quarter or half the geometry. Its will be much more faster in terms of time and PC resources. But from experience with the coils, I shouldn't touch them, lol. I have managed to get a repeating geometric section, that repeats after 180 degs. In the quarter section I have to cut the coils and they will be open loops which will be bad for the physics, hence convergence.
I checked out the documentary, most of it is just definitions.
The examples given have the sources of the fields being at the boundary but am not sure I want to do that with the coils?
So how do I go about assigning BCs to give me a repeating section with a rotation of 180 degs.
Regards.
Wanjiku