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 The information provided may be out of date.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.

Singular Matrix and revolved mesh?

Please login with a confirmed email address before reporting spam

Hello All,

After revolving my rather simple geometry and applying my subdomain/boundary conditions, I am clicking the Solve button and receiving a Singular Matrix error. This is an acoustics / structural mechanics coupled problem.

Since I am always getting the same 'Singular revolutions not supported for triangular elements' (4123) error message, I am using Quad elements while revolving my mesh. I am ticking on the Symmetric check box and applying my subdomain and boundary settings afterwards. After clicking on the solve button, I can follow in the text region below, Comsol is complaining about avoiding some inverted elements. Unlike my other cases (other 3D or axi-symmetric models), the repeated lines have different number of inverted elements. And, I am receiving a Singular matrix (7111) error.

I have checked the 'Singular Matrix' section of the knowledge database and I have checked the explanation of the error number 7111. I am sure that my element quality is over than 0.005 as indicated by Comsol and my problem doesn't fit any of the explanations in the knowledge base.

Has anybody struggled with this problem before?
/Onur

17 Replies Last Post May 29, 2012, 10:04 a.m. EDT
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 18, 2009, 3:58 p.m. EST
Hi

I havn't got into the same problems, yet, but I do see that on the rotation axis you can easily get "singular" values and problems because of mesh geometry shapes. everything converges on this axis so your prisms are not very "prismatic". There are a few comments about this somewhere in the documentation, I have forgotten exactly where now.

On trick/way around is to leave a small "hole" in the structure on the axis, but this is not always compatible with the models.

It is correct that tri's cannot be revolved to tets, but you can always get back to tet's by appplying the "prism to tet" conversion in the meshing options, once the part is revolved

In axisymmetric mode the singularity effets are minimised because COMSOL uses the trick to multiply several dimensions by "r" to avoid the 1/r for r=0 asymptote.

Hope this can help you on the way, good luck
Ivar
Hi I havn't got into the same problems, yet, but I do see that on the rotation axis you can easily get "singular" values and problems because of mesh geometry shapes. everything converges on this axis so your prisms are not very "prismatic". There are a few comments about this somewhere in the documentation, I have forgotten exactly where now. On trick/way around is to leave a small "hole" in the structure on the axis, but this is not always compatible with the models. It is correct that tri's cannot be revolved to tets, but you can always get back to tet's by appplying the "prism to tet" conversion in the meshing options, once the part is revolved In axisymmetric mode the singularity effets are minimised because COMSOL uses the trick to multiply several dimensions by "r" to avoid the 1/r for r=0 asymptote. Hope this can help you on the way, good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 5:20 a.m. EDT
Dear Onur ,
Can you please let me know how u could solve your problem?
I have a same error with using free quad mesh.
Please help me.
Best Wishes,
Ali
Dear Onur , Can you please let me know how u could solve your problem? I have a same error with using free quad mesh. Please help me. Best Wishes, Ali

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 5:22 a.m. EDT
Dear Ivar,
I have a same problem and i can't solve it. Please help me.
Best Wishes,
Ali
Dear Ivar, I have a same problem and i can't solve it. Please help me. Best Wishes, Ali

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 9:24 a.m. EDT
Hi

try avoid a quad mesh on the axis of rotation, it could be an issue (but it might be accepted by COMSOL by now in the newer versions)

--
Good luck
Ivar
Hi try avoid a quad mesh on the axis of rotation, it could be an issue (but it might be accepted by COMSOL by now in the newer versions) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 9:31 a.m. EDT
Dear Ivar,
Thanks your kind attention.
But I have a very simple geometry that u can see in the attached file. Meanwhile I use a new version on COMSOL.
Would you please give me another solution?
Best Wishes,
Ali
Dear Ivar, Thanks your kind attention. But I have a very simple geometry that u can see in the attached file. Meanwhile I use a new version on COMSOL. Would you please give me another solution? Best Wishes, Ali


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 9:43 a.m. EDT
Hi

start using COMSOLs default mesh probably it will work easier

--
Good luck
Ivar
Hi start using COMSOLs default mesh probably it will work easier -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 9:48 a.m. EDT
Dear Ivar,
Thanks your quick reply.
I know about default mesh, but i should use quad mesh for my geometry.
I have to use quad mesh or i should present reason why it is not possible.
Could you please help me?
Best Wishes,
Ali
Dear Ivar, Thanks your quick reply. I know about default mesh, but i should use quad mesh for my geometry. I have to use quad mesh or i should present reason why it is not possible. Could you please help me? Best Wishes, Ali

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 9:57 a.m. EDT
Hi

He has probably a good reason, but I'm not sure its possible through the full system with COMSOL, play a little with the swwep mesh too and try, probbaly he wants you t find the correct way out

--
Good luck
Ivar
Hi He has probably a good reason, but I'm not sure its possible through the full system with COMSOL, play a little with the swwep mesh too and try, probbaly he wants you t find the correct way out -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 26, 2012, 10:01 a.m. EDT
Dear Ivar,
Thanks your help.
I have tested with swept meshing and i can solve for regular structures. But when my structure is irregular it is not possible to use swept mesh. I don't know what I should do now.

Best Wishes,
Ali
Dear Ivar, Thanks your help. I have tested with swept meshing and i can solve for regular structures. But when my structure is irregular it is not possible to use swept mesh. I don't know what I should do now. Best Wishes, Ali

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 4:29 a.m. EDT
Hi

In fact neither do I for a swept mesh in your geoemtry case.

But what I still would do, is to solve it with thets, and validate that the results are still reasonable, then read carefully the doc about uqad and swept mesh, make a few examples and explain why you could NOT use a quad mesh. And perhaps give a short introduction to all the mesh types in COMSOL: tri, quad, thet, brick ...

In this way you will be more familiar with the mesh types and when they apply, nothing better than a few examples for any learning process, I still do it regularly when entering a new field, or revisiting one where I have not been for some time, even 30+ years after my studies ;)

--
Good luck
Ivar
Hi In fact neither do I for a swept mesh in your geoemtry case. But what I still would do, is to solve it with thets, and validate that the results are still reasonable, then read carefully the doc about uqad and swept mesh, make a few examples and explain why you could NOT use a quad mesh. And perhaps give a short introduction to all the mesh types in COMSOL: tri, quad, thet, brick ... In this way you will be more familiar with the mesh types and when they apply, nothing better than a few examples for any learning process, I still do it regularly when entering a new field, or revisiting one where I have not been for some time, even 30+ years after my studies ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 6:21 a.m. EDT
Dear Ivar,

Thanks your kind guides.
I have tried to find scientific reason for why free quad cannot solve my simulation even for very simple geometries but i don't find the answer yet. :(
What is your idea about the question?
Best Wshes,
Dear Ivar, Thanks your kind guides. I have tried to find scientific reason for why free quad cannot solve my simulation even for very simple geometries but i don't find the answer yet. :( What is your idea about the question? Best Wshes,

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 6:24 a.m. EDT
Hi

for me its the way COMSOL generate the mesh, it is not possible (from my udnerstanding) in v4.2 to use swept or structured mesh on such uneven geometries.It's only a mesh generation issue not a FEM issue.

Perhaps in the new v4.3 where the mesh generation has been improved, it will be posible ?

--
Good luck
Ivar
Hi for me its the way COMSOL generate the mesh, it is not possible (from my udnerstanding) in v4.2 to use swept or structured mesh on such uneven geometries.It's only a mesh generation issue not a FEM issue. Perhaps in the new v4.3 where the mesh generation has been improved, it will be posible ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 6:31 a.m. EDT
Dear Ivar,

I don't have a new version of COMSOL and I can't check it. :(

But if you are sure about it maybe i can ask my university to buy it.

and about free quad you think reason is COMSOL's mesh ability?

Best Wishes,
Ali
Dear Ivar, I don't have a new version of COMSOL and I can't check it. :( But if you are sure about it maybe i can ask my university to buy it. and about free quad you think reason is COMSOL's mesh ability? Best Wishes, Ali

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 6:57 a.m. EDT
Hi

if your university is wise (and rich enough) it has a maintennce running then you will receive the new version soon.

To test meshes, make a simple rectangle/cube and try out the different meshes, then add a circle/sphere in the middle and try again, test the sweep mesh too in 3D and 2D

--
Good luck
Ivar
Hi if your university is wise (and rich enough) it has a maintennce running then you will receive the new version soon. To test meshes, make a simple rectangle/cube and try out the different meshes, then add a circle/sphere in the middle and try again, test the sweep mesh too in 3D and 2D -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 28, 2012, 7:59 a.m. EDT
Dear Ivar,

Thanks your helps.

I checked it but still solving free quad meshing is singular.

However i thanks your good comments.

Best Wishes,

Ali
Dear Ivar, Thanks your helps. I checked it but still solving free quad meshing is singular. However i thanks your good comments. Best Wishes, Ali

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 29, 2012, 6:25 a.m. EDT
Dear Ivar,
Sorry for my bothering.
I have a another question that if i can answer it maybe i can solve my problem. You saw my geometry that two cube have a intersect with each other. I want to use surface instead of cubes because physically it is possible. I tried to use rectangular shapes but in the 3D it is not possible. Would you please let me know what i can do?
Best Wishes,
Ali
Dear Ivar, Sorry for my bothering. I have a another question that if i can answer it maybe i can solve my problem. You saw my geometry that two cube have a intersect with each other. I want to use surface instead of cubes because physically it is possible. I tried to use rectangular shapes but in the 3D it is not possible. Would you please let me know what i can do? Best Wishes, Ali

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago May 29, 2012, 10:04 a.m. EDT
Hi

I do not believe so (but I might be wrong :)
you need the 3D domains as you have no true symmetry in your geometrical model

--
Good luck
Ivar
Hi I do not believe so (but I might be wrong :) you need the 3D domains as you have no true symmetry in your geometrical model -- Good luck Ivar

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.