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.

Interface between a shell and a 3D domain element

Please login with a confirmed email address before reporting spam

I'm curious if anyone has any experience with the preferred way to couple the physical mechanical stress and strains between a 3D domain and a shell.

This MEMs problem consists of a very thin Si3N4 plate spanning a large trench etched in silicon. Early attempts which treated the span as a domain were of limited success so the next approach is to treat the Si3N4 as a span and interface it with the subdomain of the silicon substrate at teach end of the trench.

The 3D geometry model forms the silicon substrate with a very long and wide trench. A workplane is then added to create a geometry rectangle that becomes a shell.

Two Physics sets are added. The first is the 'Shell' and the second is a 'Solid Mechanics'. The difficulty I'm having is the coupling the displacements and forces from the silicon substrate domain to the shell. After some study the approach that seemed most viable was to create an 'assembly' rather than a 'union' and to create Identity Pairs for the needed physical coupling. It would seem that using 'Shell/Pairs/Continuity' would be the approach but the Identity pairs are not available for selection.

So, looking to the SolidMechanics/paris/continuity I find that only one end of the pairs is available.

I'm certainly open to suggestions on a better approach for this coupling.



Thanks,
Marty

12 Replies Last Post 2015年12月31日 GMT-5 01:11
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 2011年2月28日 GMT-5 15:11
Hi

I have not coupled shells over large areas, rather "but ended" between 3D shell (or beam) and some 3D structure. For that I need to define some kind of usefull rotation from the 3D displacements to the theta rotation dependent variables of the shell/beams. I ususlly use the curl or the small displacement = rotations from the 3D theory (but I lack a good "large displacement link"). Another way is to have the surface "slightly" overlap over a few mesh elements and leave the shell/beal rotations forally free, this overstiffens the overlapping region but this can often be neglected. Loads are thus also to be transferred via approapriate two way couplings or a weak formulation.

I find this rather tedious and error prone, and not trivial to validate, so I do not use it that often.

For 3D structures we really need more advanced boundary "thin film" and "stacked -multiple thin film" BC to add a set of thin layers on a 3D surface. Particularly for MEMS.

Just as we need some reduced anisotropic domain model to better and simpler simulate the numerous corragulated or perforated surfaces used today to make thin Si MEMS membranes lighter and stiffer

--
Good luck
Ivar
Hi I have not coupled shells over large areas, rather "but ended" between 3D shell (or beam) and some 3D structure. For that I need to define some kind of usefull rotation from the 3D displacements to the theta rotation dependent variables of the shell/beams. I ususlly use the curl or the small displacement = rotations from the 3D theory (but I lack a good "large displacement link"). Another way is to have the surface "slightly" overlap over a few mesh elements and leave the shell/beal rotations forally free, this overstiffens the overlapping region but this can often be neglected. Loads are thus also to be transferred via approapriate two way couplings or a weak formulation. I find this rather tedious and error prone, and not trivial to validate, so I do not use it that often. For 3D structures we really need more advanced boundary "thin film" and "stacked -multiple thin film" BC to add a set of thin layers on a 3D surface. Particularly for MEMS. Just as we need some reduced anisotropic domain model to better and simpler simulate the numerous corragulated or perforated surfaces used today to make thin Si MEMS membranes lighter and stiffer -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2011年2月28日 GMT-5 16:16
Thanks Ivar,
This amplifies on an entry you had made last fall which will be helpful.

I'm comfortable with the shell end overlap as you suggested, even though stiffer, because this is how the structure is actually fabricated. I guess I was just anticipating that these overlapped regions could be defined as Identity Pairs and that would immediately couple the independent parameters at the common nodes. Yep, always
learning.

If I have success I'll let you know what I learn..


Thanks for your continued involvement in this forum.

Marty
Thanks Ivar, This amplifies on an entry you had made last fall which will be helpful. I'm comfortable with the shell end overlap as you suggested, even though stiffer, because this is how the structure is actually fabricated. I guess I was just anticipating that these overlapped regions could be defined as Identity Pairs and that would immediately couple the independent parameters at the common nodes. Yep, always learning. If I have success I'll let you know what I learn.. Thanks for your continued involvement in this forum. Marty

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 2011年2月28日 GMT-5 16:40
Hi

check/validate carefully your models it's rather quickly done to get larger errors with these type of "tweakings"

--
Good luck
Ivar
Hi check/validate carefully your models it's rather quickly done to get larger errors with these type of "tweakings" -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2011年3月3日 GMT-5 10:41
I got some good advice from Supratik at COMSOL Support regarding this interface. His complete response is included below but in a nut shell one establishes a physics interface for the 'shell' and a separate one for the 'Structural Mechanics' 3D domain. To couple these two, one does not use identity pairs as I had attempted but rather as Ivar had suggested, the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions.

Another important piece of advice that Supratik offered were has to do with the solver configuration. Adding two physics interfaces resulted in a segrated solver arrangement. His steps below create a coupled solution instead.

Again, thanks to both Ivar and Supratik for the advice.

Marty

From COMSOL Support:
Thank you for contacting COMSOL Support. You have indeed correctly used
the combination of shell and solid mechanics for modeling your device.

We do not need to use identity pair. We can connect the solid mechanics
and shell interfaces by prescribing a displacement in the shell interface
and calling the displacement variables from the solid mechanics interface
to setup the displacement continuity condition at the interface boundaries.

COMSOL by default will try to solve this model using a segregated solver
because this model uses two interfaces but this default approach is not
suitable. Hence you would first need to right-click on the Study node and
click on Show Default Solver to generate the default solver setting. You
can then right-click on the Study 1 > Solver Configurations > Solver 1 >
Stationary Solver 1 branch and select Fully Coupled. You can then compute
the model as usual.
You can plot the displacement from both the shell and solid mechanics
interfaces in the same plot group. Please see how I have setup 3D plot
group 1.

I hope this information will be helpful for you.

Best Regards, Supratik, COMSOL Support Team
I got some good advice from Supratik at COMSOL Support regarding this interface. His complete response is included below but in a nut shell one establishes a physics interface for the 'shell' and a separate one for the 'Structural Mechanics' 3D domain. To couple these two, one does not use identity pairs as I had attempted but rather as Ivar had suggested, the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions. Another important piece of advice that Supratik offered were has to do with the solver configuration. Adding two physics interfaces resulted in a segrated solver arrangement. His steps below create a coupled solution instead. Again, thanks to both Ivar and Supratik for the advice. Marty From COMSOL Support: Thank you for contacting COMSOL Support. You have indeed correctly used the combination of shell and solid mechanics for modeling your device. We do not need to use identity pair. We can connect the solid mechanics and shell interfaces by prescribing a displacement in the shell interface and calling the displacement variables from the solid mechanics interface to setup the displacement continuity condition at the interface boundaries. COMSOL by default will try to solve this model using a segregated solver because this model uses two interfaces but this default approach is not suitable. Hence you would first need to right-click on the Study node and click on Show Default Solver to generate the default solver setting. You can then right-click on the Study 1 > Solver Configurations > Solver 1 > Stationary Solver 1 branch and select Fully Coupled. You can then compute the model as usual. You can plot the displacement from both the shell and solid mechanics interfaces in the same plot group. Please see how I have setup 3D plot group 1. I hope this information will be helpful for you. Best Regards, Supratik, COMSOL Support Team

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2011年3月3日 GMT-5 12:20
Thanks Marty for sharing the information.

I am using shell and coupled with solid elements, same way as the support suggested, it works for my problem.

Since I am using an imported FE mesh, I have to use version 4.0a, since version 4.1 doesn't support shell interface for
imported FE mesh. Although this sounds strange, since even version 3.5a support imported FE mesh for shell.
Hope this will be fixed in version 4.2.

Thanks.
XLi
Thanks Marty for sharing the information. I am using shell and coupled with solid elements, same way as the support suggested, it works for my problem. Since I am using an imported FE mesh, I have to use version 4.0a, since version 4.1 doesn't support shell interface for imported FE mesh. Although this sounds strange, since even version 3.5a support imported FE mesh for shell. Hope this will be fixed in version 4.2. Thanks. XLi

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年4月25日 GMT-4 11:26
'We can connect the solid mechanics and shell interfaces by prescribing a displacement in the shell interface and calling the displacement variables from the solid mechanics interface to setup the displacement continuity condition at the interface boundaries.'
Anyone know how to do this? How to 'calling the displacement variables from the solid mechanics interface'


I got some good advice from Supratik at COMSOL Support regarding this interface. His complete response is included below but in a nut shell one establishes a physics interface for the 'shell' and a separate one for the 'Structural Mechanics' 3D domain. To couple these two, one does not use identity pairs as I had attempted but rather as Ivar had suggested, the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions.

Another important piece of advice that Supratik offered were has to do with the solver configuration. Adding two physics interfaces resulted in a segrated solver arrangement. His steps below create a coupled solution instead.

Again, thanks to both Ivar and Supratik for the advice.

Marty

From COMSOL Support:
Thank you for contacting COMSOL Support. You have indeed correctly used
the combination of shell and solid mechanics for modeling your device.

We do not need to use identity pair. We can connect the solid mechanics
and shell interfaces by prescribing a displacement in the shell interface
and calling the displacement variables from the solid mechanics interface
to setup the displacement continuity condition at the interface boundaries.

COMSOL by default will try to solve this model using a segregated solver
because this model uses two interfaces but this default approach is not
suitable. Hence you would first need to right-click on the Study node and
click on Show Default Solver to generate the default solver setting. You
can then right-click on the Study 1 > Solver Configurations > Solver 1 >
Stationary Solver 1 branch and select Fully Coupled. You can then compute
the model as usual.
You can plot the displacement from both the shell and solid mechanics
interfaces in the same plot group. Please see how I have setup 3D plot
group 1.

I hope this information will be helpful for you.

Best Regards, Supratik, COMSOL Support Team

'We can connect the solid mechanics and shell interfaces by prescribing a displacement in the shell interface and calling the displacement variables from the solid mechanics interface to setup the displacement continuity condition at the interface boundaries.' Anyone know how to do this? How to 'calling the displacement variables from the solid mechanics interface' [QUOTE] I got some good advice from Supratik at COMSOL Support regarding this interface. His complete response is included below but in a nut shell one establishes a physics interface for the 'shell' and a separate one for the 'Structural Mechanics' 3D domain. To couple these two, one does not use identity pairs as I had attempted but rather as Ivar had suggested, the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions. Another important piece of advice that Supratik offered were has to do with the solver configuration. Adding two physics interfaces resulted in a segrated solver arrangement. His steps below create a coupled solution instead. Again, thanks to both Ivar and Supratik for the advice. Marty From COMSOL Support: Thank you for contacting COMSOL Support. You have indeed correctly used the combination of shell and solid mechanics for modeling your device. We do not need to use identity pair. We can connect the solid mechanics and shell interfaces by prescribing a displacement in the shell interface and calling the displacement variables from the solid mechanics interface to setup the displacement continuity condition at the interface boundaries. COMSOL by default will try to solve this model using a segregated solver because this model uses two interfaces but this default approach is not suitable. Hence you would first need to right-click on the Study node and click on Show Default Solver to generate the default solver setting. You can then right-click on the Study 1 > Solver Configurations > Solver 1 > Stationary Solver 1 branch and select Fully Coupled. You can then compute the model as usual. You can plot the displacement from both the shell and solid mechanics interfaces in the same plot group. Please see how I have setup 3D plot group 1. I hope this information will be helpful for you. Best Regards, Supratik, COMSOL Support Team [/QUOTE]

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 2012年4月25日 GMT-4 12:08
Hi

one of the issues to check is that you do not get a "mechanism" when you link a solid (solving for u,v,w) and a shell item, solving for both displacements and 3 rotations. You must ensure you link in the rotations too, for example (for small displacements) via the angles solid.curlUX, Y, Z

--
Good luck
Ivar
Hi one of the issues to check is that you do not get a "mechanism" when you link a solid (solving for u,v,w) and a shell item, solving for both displacements and 3 rotations. You must ensure you link in the rotations too, for example (for small displacements) via the angles solid.curlUX, Y, Z -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年4月25日 GMT-4 21:23
Thanks, Ivar. I still don't know how to link the displacements of a solid and a shell. The only way I know to link two items is using contact pairs or identity pairs. Obviously, we cann't use contact pairs or identity pairs in this problem. Attached is my model. Can you please have a look and give me some comments? Any help is appreciated.

The model includes femur and cartilage. The femur is modeled as shell, and cartilage is modeled as solid. A prescribed displacement is applied to the top surface of the femur to compress the entire geometry.

Thanks.

Hongqiang


Hi

one of the issues to check is that you do not get a "mechanism" when you link a solid (solving for u,v,w) and a shell item, solving for both displacements and 3 rotations. You must ensure you link in the rotations too, for example (for small displacements) via the angles solid.curlUX, Y, Z

--
Good luck
Ivar


Thanks, Ivar. I still don't know how to link the displacements of a solid and a shell. The only way I know to link two items is using contact pairs or identity pairs. Obviously, we cann't use contact pairs or identity pairs in this problem. Attached is my model. Can you please have a look and give me some comments? Any help is appreciated. The model includes femur and cartilage. The femur is modeled as shell, and cartilage is modeled as solid. A prescribed displacement is applied to the top surface of the femur to compress the entire geometry. Thanks. Hongqiang [QUOTE] Hi one of the issues to check is that you do not get a "mechanism" when you link a solid (solving for u,v,w) and a shell item, solving for both displacements and 3 rotations. You must ensure you link in the rotations too, for example (for small displacements) via the angles solid.curlUX, Y, Z -- Good luck Ivar [/QUOTE]


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 2012年4月26日 GMT-4 11:14
Hi

it has to do with your geoemtry, as you have only imported mesh you do not have any traditional geoemtry with common boundaries and imprints etc. I'm not used to mesh import as iut did never work for me earlier, but it seems that in V4 now it has also improved, nice. You are better off asking directly "support" I believe, but pls report back as this is of great interest to us dealing with other softwares too and having issues interfacing our clients tools and our tools ;)

Normally I would have expected two domains and not a shell and a domain, why use a "shell" for the bone, I do not really find it "easier", at least not from the physical understanding side ;)

--
Good luck
Ivar
Hi it has to do with your geoemtry, as you have only imported mesh you do not have any traditional geoemtry with common boundaries and imprints etc. I'm not used to mesh import as iut did never work for me earlier, but it seems that in V4 now it has also improved, nice. You are better off asking directly "support" I believe, but pls report back as this is of great interest to us dealing with other softwares too and having issues interfacing our clients tools and our tools ;) Normally I would have expected two domains and not a shell and a domain, why use a "shell" for the bone, I do not really find it "easier", at least not from the physical understanding side ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年4月26日 GMT-4 14:36
Thanks, Ivar. I'll contact COMSOL support. I report back if get anything new.

Using shell for the bone is to reduce the element number, thus save computational time and memory. In my model, I only care about the surface of the femur.
Thanks, Ivar. I'll contact COMSOL support. I report back if get anything new. Using shell for the bone is to reduce the element number, thus save computational time and memory. In my model, I only care about the surface of the femur.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2013年6月23日 GMT-4 17:24
Hi,

it seems that I have a very similar problem. I read this thread over and over again. But I can't fix my problem. My geometry is based on a meander-shaped metal bellows (thin s. ) attached to thick wall pipe (solid) and finalized as an assembly. The thermal transition is a edge on a surface connection. I tried every contact pair ... it wont work.

Where can i find the already mentioned settings for this solution: "...the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions." ?

Maybe this is obvious, but as a newbie I need a more detailed description what I need to do. Or a link to tutorial. I found only one model related to this problem in the model library: disk_stack_heat_sink.mph. But compared to my model it is based on a union and as far as I know it was solved with a thin conductive layer ?

Thank you in advance for your effort.

Best regards,
Alex
Hi, it seems that I have a very similar problem. I read this thread over and over again. But I can't fix my problem. My geometry is based on a meander-shaped metal bellows (thin s. ) attached to thick wall pipe (solid) and finalized as an assembly. The thermal transition is a edge on a surface connection. I tried every contact pair ... it wont work. Where can i find the already mentioned settings for this solution: "...the displacement variables, u, v, w for the SM interface are implemented as the Prescribed Displacements of the Shell interface regions." ? Maybe this is obvious, but as a newbie I need a more detailed description what I need to do. Or a link to tutorial. I found only one model related to this problem in the model library: disk_stack_heat_sink.mph. But compared to my model it is based on a union and as far as I know it was solved with a thin conductive layer ? Thank you in advance for your effort. Best regards, Alex

Please login with a confirmed email address before reporting spam

Posted: 9 years ago 2015年12月31日 GMT-5 01:11
Hi Hongqiang

I am new with Comsol. I would like to model a simple concrete bridge which composed of deck (floor), beams and abutments. Shell physic will be assigned to deck/ floor element and solid for beams and abutments members.

I have a problem similar to your previous problem regarding to assign interface between shell and solid element. As I observed from this forum you have solved the problem properly. Therefore, I would like your explanation on how to assign the interface between shell and solid element. Also, how to link displacement between them.

I attached the file I made for you to see. Many thanks for your helps.

Regards


Maizuar
Hi Hongqiang I am new with Comsol. I would like to model a simple concrete bridge which composed of deck (floor), beams and abutments. Shell physic will be assigned to deck/ floor element and solid for beams and abutments members. I have a problem similar to your previous problem regarding to assign interface between shell and solid element. As I observed from this forum you have solved the problem properly. Therefore, I would like your explanation on how to assign the interface between shell and solid element. Also, how to link displacement between them. I attached the file I made for you to see. Many thanks for your helps. Regards Maizuar

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.