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.
Modeling a Linear Electromagnetic Generator : Moving Mesh - Inverted Mesh Problem
Posted Nov 28, 2010, 8:20 a.m. EST 3 Replies
Please login with a confirmed email address before reporting spam
Hi all!
I'm modelling a linear electromagnetic generator with moving mesh and perpendicular induction currents (VP) applications modes which are both transient. The geometry of the problem is in the attachments.
When displacement of magnet exceeds a critical value, solver gives inverted mesh errors which leads to convergence problems.
First i modelled the boundary between magnet and air as a pair but this technique gave wrong solutions and mesh in air domain didnt follow magnet domain.
So i modelled without any pairs between air and magnet.
For small displacements (in which magnet didnt go near the 3rd teeth(pole) in the generator core), results are ok. But when modeling the system as in the reality which magnet moves to the 3rd pole and go back 2nd one, solver didnt solve the problem as i mentioned before.
I tried refining meshes between magnet and the core, but it didnt solve the problem.
I tried giving the solver a stop condition about mesh quality and after the stop of solver i remeshed and started simulation from that time step. But it didnt solve my problem either.
So as you see, I'm stucked! : /
Any help, idea or clue for that problem?
Thank you!
I'm modelling a linear electromagnetic generator with moving mesh and perpendicular induction currents (VP) applications modes which are both transient. The geometry of the problem is in the attachments.
When displacement of magnet exceeds a critical value, solver gives inverted mesh errors which leads to convergence problems.
First i modelled the boundary between magnet and air as a pair but this technique gave wrong solutions and mesh in air domain didnt follow magnet domain.
So i modelled without any pairs between air and magnet.
For small displacements (in which magnet didnt go near the 3rd teeth(pole) in the generator core), results are ok. But when modeling the system as in the reality which magnet moves to the 3rd pole and go back 2nd one, solver didnt solve the problem as i mentioned before.
I tried refining meshes between magnet and the core, but it didnt solve the problem.
I tried giving the solver a stop condition about mesh quality and after the stop of solver i remeshed and started simulation from that time step. But it didnt solve my problem either.
So as you see, I'm stucked! : /
Any help, idea or clue for that problem?
Thank you!
Attachments:
3 Replies Last Post Nov 29, 2010, 10:04 a.m. EST
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
inverted mesh on moving mesh ore often tricky to solve for large displacements. There are a few parameters in the settings that one can tweak to improve the unifromity of the displacement, or cut the domains to procide "mesh" sliders. you could also try with a rather coarse mesh for air. if you do a lot of moving mesh it's really worth to go to the advanced meshing course of COMSOL, there you learn and exercice a lot of tricks
Now if I understand your model correctly your magnet moves up and down (only), or in all directions ?
if only up and down you can cut your model in two vertically in the middle of the air gap and use assembly mode to link the physics along a sliding boundary (take the 2D rotation generator example of 3.5 and generalise it to linear)
if in any direction, note I do not know if it works, but for the fluidics they use the "level set" method and "virtual-true" material defined by a boolean expression in space. Perhaps your magnet could be simulated that way, then you can forget the moving mesh and inverted element issues (not sure it works, just a crazy idea, but from the principle, what's different?)
--
Good luck
Ivar
inverted mesh on moving mesh ore often tricky to solve for large displacements. There are a few parameters in the settings that one can tweak to improve the unifromity of the displacement, or cut the domains to procide "mesh" sliders. you could also try with a rather coarse mesh for air. if you do a lot of moving mesh it's really worth to go to the advanced meshing course of COMSOL, there you learn and exercice a lot of tricks
Now if I understand your model correctly your magnet moves up and down (only), or in all directions ?
if only up and down you can cut your model in two vertically in the middle of the air gap and use assembly mode to link the physics along a sliding boundary (take the 2D rotation generator example of 3.5 and generalise it to linear)
if in any direction, note I do not know if it works, but for the fluidics they use the "level set" method and "virtual-true" material defined by a boolean expression in space. Perhaps your magnet could be simulated that way, then you can forget the moving mesh and inverted element issues (not sure it works, just a crazy idea, but from the principle, what's different?)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar!
Magnet in my model only moves in up and down direction.
I tried two different things as you suggested.
First i remeshed air coarser. It didnt solved my problem.(Inverted mesh errors decreased as predicted.)
Second i remodelled the system with five air domains(geo figure is in the attachment) I describe a pair between moving air and not moving without imprints. Now solver find a solution which i'm not sure if it is correct. (for small displacements, results are still OK) And in log there is a new warning as
" Warning: New constraint force nodes detected: these are not stored !".
I increase mesh number but they are still there.
Are they really harmless as described in previous posts in the forum? Or what should I do to remove them?
Thank you Ivar!
And I searched about COMSOL Advanced Meshing Trainings but there are not available training near or in my country.. Webinars are the only source that i can reach i think.
Magnet in my model only moves in up and down direction.
I tried two different things as you suggested.
First i remeshed air coarser. It didnt solved my problem.(Inverted mesh errors decreased as predicted.)
Second i remodelled the system with five air domains(geo figure is in the attachment) I describe a pair between moving air and not moving without imprints. Now solver find a solution which i'm not sure if it is correct. (for small displacements, results are still OK) And in log there is a new warning as
" Warning: New constraint force nodes detected: these are not stored !".
I increase mesh number but they are still there.
Are they really harmless as described in previous posts in the forum? Or what should I do to remove them?
Thank you Ivar!
And I searched about COMSOL Advanced Meshing Trainings but there are not available training near or in my country.. Webinars are the only source that i can reach i think.
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
that warning is less important, it has to do with the moving and crossing of flux lines, the best eay to check is to see if the images moves coherently (contour flux lines are not interrupted along the moving edge
Often it's best o debug the case in a simple goemtry and then apply it to the desired one. I beleive this has been discussd a year or two ago for 3.5, or 3.5a probably, try a search on the forum
--
Good luck
Ivar
that warning is less important, it has to do with the moving and crossing of flux lines, the best eay to check is to see if the images moves coherently (contour flux lines are not interrupted along the moving edge
Often it's best o debug the case in a simple goemtry and then apply it to the desired one. I beleive this has been discussd a year or two ago for 3.5, or 3.5a probably, try a search on the forum
--
Good luck
Ivar