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.
swept mesh ..
Posted 1 janv. 2010, 16:50 UTC−5 7 Replies
Please login with a confirmed email address before reporting spam
Hi,
I'm having some difficulties with swept mesh generation. I have 3D object created by extrusion of 2D sketch. It consists of many adjacent subdomains of same length. I mesh the the sketch face of whole extruded geometry with free triangle mesh in order to provide the swept mesh algorithm with the starting 2D mesh. Then I try to generate the swept mesh for all the subdomains. On some it works with no problem. On other it fails and shouts :
Error 4141
Invalid starting mesh for this meshing operation.
-Subdomain : 40
This error is not listed in help(!). I tried re-meshing, without effect.
If you have an idea what could possibly be the cause of the problem, I'll be very, very grateful for help I'm really lost here.
Hans
I'm having some difficulties with swept mesh generation. I have 3D object created by extrusion of 2D sketch. It consists of many adjacent subdomains of same length. I mesh the the sketch face of whole extruded geometry with free triangle mesh in order to provide the swept mesh algorithm with the starting 2D mesh. Then I try to generate the swept mesh for all the subdomains. On some it works with no problem. On other it fails and shouts :
Error 4141
Invalid starting mesh for this meshing operation.
-Subdomain : 40
This error is not listed in help(!). I tried re-meshing, without effect.
If you have an idea what could possibly be the cause of the problem, I'll be very, very grateful for help I'm really lost here.
Hans
7 Replies Last Post 2 févr. 2010, 11:55 UTC−5
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi
CAnnot tell like that, but you have a clue by the "subdomain 40" indication, identify the subdomain 40 and see how it looks, it could be that it's touching or is adjacent to a couple of meshed items, and then perhaps the algorithm does not manage to fit all the adjacent meshes.
Hope it helps
Ivar
CAnnot tell like that, but you have a clue by the "subdomain 40" indication, identify the subdomain 40 and see how it looks, it could be that it's touching or is adjacent to a couple of meshed items, and then perhaps the algorithm does not manage to fit all the adjacent meshes.
Hope it helps
Ivar
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi Ivar,
its exactly as you say. The domain I've been trying to mesh is adjacent to meshed subdomains. Some are made by simple extrusion, some by extrusion with twist, ergo their resulting swept meshes cant match on sides. I missed that. Is there a chance to get over this ? For example is there a chance to create a special face in the model where 3D meshes, which are not compatible because of the shape of their elements, could be adjacent to this face(2D) without any problem? Some kind of face which would perform the interpolation from one 3D mesh to another ? Some kind of mesh-independent continuity of solution ?
its exactly as you say. The domain I've been trying to mesh is adjacent to meshed subdomains. Some are made by simple extrusion, some by extrusion with twist, ergo their resulting swept meshes cant match on sides. I missed that. Is there a chance to get over this ? For example is there a chance to create a special face in the model where 3D meshes, which are not compatible because of the shape of their elements, could be adjacent to this face(2D) without any problem? Some kind of face which would perform the interpolation from one 3D mesh to another ? Some kind of mesh-independent continuity of solution ?
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi
A difficult case I believe, as quad mesh do not mix with tets. One way could be to mesh all parts possible, then transform everything to tets and try a free mesh on the reminding. Or try to mesh all edges remainig in a regular way, and then the volumes
Good luck
Ivar
A difficult case I believe, as quad mesh do not mix with tets. One way could be to mesh all parts possible, then transform everything to tets and try a free mesh on the reminding. Or try to mesh all edges remainig in a regular way, and then the volumes
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi,
yea. I also thought about changing the swetp mesh elements to tets and than meshing the rest with free, but unfortunatelly my case is the small induction machine(the one we can meaure in our lab) and it has so small air gap(the critical area between the straight and twisted 2D->3D geometry extrusion meshed with sweep mesh - stator, rotor respetivelly) that the free mesh algorythm fails anyway(and if it made it, it would probably produce extremly low quality elements). So im gonna have to go with free mesh on whole geometry, which results in 2e6 DOFS :(
anyway, thanks x 1000 for your time
Hans
yea. I also thought about changing the swetp mesh elements to tets and than meshing the rest with free, but unfortunatelly my case is the small induction machine(the one we can meaure in our lab) and it has so small air gap(the critical area between the straight and twisted 2D->3D geometry extrusion meshed with sweep mesh - stator, rotor respetivelly) that the free mesh algorythm fails anyway(and if it made it, it would probably produce extremly low quality elements). So im gonna have to go with free mesh on whole geometry, which results in 2e6 DOFS :(
anyway, thanks x 1000 for your time
Hans
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi
Then my only recommendation is to try to simplify via symmetry, it's complicates, and one cannot do everything, but still, it gives results that with a little tweaking can mostly be extrapolated, FEM is in no way exact, I say always to some 10% precision only
Good luck
Ivar
Then my only recommendation is to try to simplify via symmetry, it's complicates, and one cannot do everything, but still, it gives results that with a little tweaking can mostly be extrapolated, FEM is in no way exact, I say always to some 10% precision only
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi,
I have spent some time thinking about simplifying the geometry, but then I found out that the best way is to leave the geometry unchanged (mainly arcs) and leave it to comsol. There are many arcs in my geometry and still the best way to deal with them is to leave it up to comsol. When meshig, COMSOL changes smooth arcs to straight, flat faces in an impressively smart way. Resulting mesh is less complicated than when I change arcs to straight shapes in geometry stage by hand. Unfortunatelly removing other geometry features would result in too rough approximation.
Hans
I have spent some time thinking about simplifying the geometry, but then I found out that the best way is to leave the geometry unchanged (mainly arcs) and leave it to comsol. There are many arcs in my geometry and still the best way to deal with them is to leave it up to comsol. When meshig, COMSOL changes smooth arcs to straight, flat faces in an impressively smart way. Resulting mesh is less complicated than when I change arcs to straight shapes in geometry stage by hand. Unfortunatelly removing other geometry features would result in too rough approximation.
Hans
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Im stupid, I could have used the assembly to prevent the mesh compatibility problems.