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 The information provided may be out of date.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.

Material properties normal to a surface?

Please login with a confirmed email address before reporting spam

Let's say I try to simulate a compressive force on an ellipsoid tubular made of composites (ex.: bicycle frame).

For the moment I'm beginning with a very simple model, a 2D section profile of that ellipsoid shape (like a "0") with a force coming from the top and pressing down.

If I assume that I have about 10 layers of evenly oriented composites sheet (no 90/45/0 directionality yet, I'm starting simple) that have been formed to shape, how do I apply my (x,y) material coordinates so that they follow my surface profile (i.e. y-coord would stay normal to the surface instead of globally vertical)?

Do I need to create a Mapped System for that or is there a simple way to define a material as a "formed sheet" where y-coord properties are always normal to the profile surface regardless of the shape I use?

(I'm just beginning with COMSOL so I still have a lot to learn!)

Thanks.

10 Replies Last Post 24 apr 2013, 20:31 GMT-4
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 4 feb 2011, 07:52 GMT-5
Hi

you have a very interesting question there, specially the use of COMSOl for composites, I hope there are others too that can give some insights in the best approach, and their advice, as this is a domain of high interst, also for me ;)

In v4.1 you have the "layer" on CAD solids, this makes one or more "onion" layers around your shapes, these can also be made via most external CAD tools.

However, I believe COMSOL is limited to one specific surface physics per boundary, so how to overlap 2-3 or more physics-layer / composite-sheets on the same boundary and to have them "stacked correctly" ?

--
Good luck
Ivar
Hi you have a very interesting question there, specially the use of COMSOl for composites, I hope there are others too that can give some insights in the best approach, and their advice, as this is a domain of high interst, also for me ;) In v4.1 you have the "layer" on CAD solids, this makes one or more "onion" layers around your shapes, these can also be made via most external CAD tools. However, I believe COMSOL is limited to one specific surface physics per boundary, so how to overlap 2-3 or more physics-layer / composite-sheets on the same boundary and to have them "stacked correctly" ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 feb 2011, 11:26 GMT-5
Ok, I think I'm on the right track but not there yet.

I created a "Mapped Coordinate System" with a frame type for Material.

In my Linear Elastic Material Model, I use that new Mapped System as my Coordinate System.

Now that my material model is link to that mapped coordinate system, all I need is to link that mapped system to my domain (which has a mapped mesh following the curve of my 2D profile).

I'm not sure but can I do something like this for the coordinate mapping of this new system:

Coordinate Expression (m)
x1 nx (normal to my domain? Or to my mesh?)
x2 tx (can we call the tangent that way?)
x3 0 (in 2D)

I'm not so sure on how to call for a specific domain or mesh (something like "model.domain.nx" ???)

Thanks for any inputs.
Ok, I think I'm on the right track but not there yet. I created a "Mapped Coordinate System" with a frame type for Material. In my Linear Elastic Material Model, I use that new Mapped System as my Coordinate System. Now that my material model is link to that mapped coordinate system, all I need is to link that mapped system to my domain (which has a mapped mesh following the curve of my 2D profile). I'm not sure but can I do something like this for the coordinate mapping of this new system: Coordinate Expression (m) x1 nx (normal to my domain? Or to my mesh?) x2 tx (can we call the tangent that way?) x3 0 (in 2D) I'm not so sure on how to call for a specific domain or mesh (something like "model.domain.nx" ???) Thanks for any inputs.

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 7 feb 2011, 15:47 GMT-5
Hi

in V4 you have the three different frames: spatial, material and mesh. By default spatial is different from material = mesh the latter referred to in upper cas X,Y,Z, see the doc.

But even if the material properties are anisotropic and defined in a specific coordinate system (w.r.t the geometry) the mesh is independent, or ?

a free thet mesh is as good (or about) as a mapped quad, there I do not follow your explanations

--
Good luck
Ivar
Hi in V4 you have the three different frames: spatial, material and mesh. By default spatial is different from material = mesh the latter referred to in upper cas X,Y,Z, see the doc. But even if the material properties are anisotropic and defined in a specific coordinate system (w.r.t the geometry) the mesh is independent, or ? a free thet mesh is as good (or about) as a mapped quad, there I do not follow your explanations -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 feb 2011, 16:21 GMT-5

But even if the material properties are anisotropic and defined in a specific coordinate system (w.r.t the geometry) the mesh is independent


That's exactly what I would like to do: "define orthogonal material properties in a specific coordinate system w.r.t. the geometry of my domain" (i.e. E1, E2, v12, G12 where 1 is the direction always tangent to my curve or domain and 2 is the direction always normal to it).

(Ex.: when you fold a sheet of fiber/matrix composite with different longitudinal and transversal properties.)

I know that for a mapped coordinate system, I can give specific equations for u, v, w but isn't there a way to give a normal/tangent coordinate system regardless of the shape of the curve?

The "boundary coordinate system" includes tangent/normal parameters but it's not applicable to material properties.

It seems simple enough yet difficult to implement at the same time.


[QUOTE] But even if the material properties are anisotropic and defined in a specific coordinate system (w.r.t the geometry) the mesh is independent[/QUOTE] That's exactly what I would like to do: "define orthogonal material properties in a specific coordinate system w.r.t. the geometry of my domain" (i.e. E1, E2, v12, G12 where 1 is the direction always tangent to my curve or domain and 2 is the direction always normal to it). (Ex.: when you fold a sheet of fiber/matrix composite with different longitudinal and transversal properties.) I know that for a mapped coordinate system, I can give specific equations for u, v, w but isn't there a way to give a normal/tangent coordinate system regardless of the shape of the curve? The "boundary coordinate system" includes tangent/normal parameters but it's not applicable to material properties. It seems simple enough yet difficult to implement at the same time.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 9 feb 2011, 16:52 GMT-5
Now I'm lost...

See in attachment.

My model is basically an arch using a cylindrical coordinate system.

I'm trying to simulate the effects of many composites layers one over the other but for the moment they do not do anything special.

My material model is using "user-defined" orthotropic properties simulating the longitudinal strength of fibers (hence why I don't have any material added under "Materials".

Seems that my model has "lost" any way to calculate results.

I've reviewed all my parameters and I can't understand why I get 0 displacement and 0 Von Mises all over....

Any thought?

Thanks.
Now I'm lost... See in attachment. My model is basically an arch using a cylindrical coordinate system. I'm trying to simulate the effects of many composites layers one over the other but for the moment they do not do anything special. My material model is using "user-defined" orthotropic properties simulating the longitudinal strength of fibers (hence why I don't have any material added under "Materials". Seems that my model has "lost" any way to calculate results. I've reviewed all my parameters and I can't understand why I get 0 displacement and 0 Von Mises all over.... Any thought? Thanks.


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 10 feb 2011, 03:17 GMT-5
Hi

yes indeed your model is bizarre I tried to set up a body load: no effect, switzed to isotropic material idem.

I would redefine the model from scratch, perhaps with only 1-2 layers for first tests (by the way you have the "layer" definition in the new 3D CAD of COMSOL).

For me your model approach for anisotropic material should work OK, if you havent missed the directions?
The tensor notation is so easy to get up/down (note PZT notation is different from standard structural, COMSOL follow the IEEE convention which crosses the xy-yx values ;)

By the way in structural, for your user coordinate systems be sure you work with the correct "Frame" use the material frame to define your material properties

--
Good luck
Ivar
Hi yes indeed your model is bizarre I tried to set up a body load: no effect, switzed to isotropic material idem. I would redefine the model from scratch, perhaps with only 1-2 layers for first tests (by the way you have the "layer" definition in the new 3D CAD of COMSOL). For me your model approach for anisotropic material should work OK, if you havent missed the directions? The tensor notation is so easy to get up/down (note PZT notation is different from standard structural, COMSOL follow the IEEE convention which crosses the xy-yx values ;) By the way in structural, for your user coordinate systems be sure you work with the correct "Frame" use the material frame to define your material properties -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 10 feb 2011, 10:10 GMT-5
Yes, I started from scratch and now at least I get some calculations done.

The only thing I find weird is that if I put my longitudinal elasticity modulus (E11) in the "phi" direction (i.e. following along a layer as if it was following a fiber) and my transversal elasticity modulus (E22) in the "r" direction (i.e. passing through the layer, normal to it) I see really weird stresses that seem to be stuck between layers and mesh.
(see PICTURE 1 in attachment). My E11 is about 15 times higher than my E22 so I would thought that more stress would develop tranversaly and not along the fibers.

If I reverse these (put longitudinal E11 along "r" and tranversal E22 along "phi") I get a nicer stress pattern (see PICTURE 2 in attachment) but I don't think these are the right E11/E22 directions.


"Layers" in COMSOL: Are they only a geometry building tool or do they have a specific function? I am using SolidWorks to create my geometry so it's much easier (Are there any topology aspects inherent to the LAYER function of COMSOL?).

"Frame Type": In my Cylindrical Coordinate System, I use the "Material (X,Y,Z)" frame type and not the Spatial (x,y,z) one. That way, it is available under my "Linear Elastic Material Model".



I'm wondering if I can model at all longitudinal layers of fibers that way at all... :(

Thanks for all your comments Ivar, it's really appreciated!
Yes, I started from scratch and now at least I get some calculations done. The only thing I find weird is that if I put my longitudinal elasticity modulus (E11) in the "phi" direction (i.e. following along a layer as if it was following a fiber) and my transversal elasticity modulus (E22) in the "r" direction (i.e. passing through the layer, normal to it) I see really weird stresses that seem to be stuck between layers and mesh. (see PICTURE 1 in attachment). My E11 is about 15 times higher than my E22 so I would thought that more stress would develop tranversaly and not along the fibers. If I reverse these (put longitudinal E11 along "r" and tranversal E22 along "phi") I get a nicer stress pattern (see PICTURE 2 in attachment) but I don't think these are the right E11/E22 directions. "Layers" in COMSOL: Are they only a geometry building tool or do they have a specific function? I am using SolidWorks to create my geometry so it's much easier (Are there any topology aspects inherent to the LAYER function of COMSOL?). "Frame Type": In my Cylindrical Coordinate System, I use the "Material (X,Y,Z)" frame type and not the Spatial (x,y,z) one. That way, it is available under my "Linear Elastic Material Model". I'm wondering if I can model at all longitudinal layers of fibers that way at all... :( Thanks for all your comments Ivar, it's really appreciated!


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 10 feb 2011, 15:02 GMT-5
Hi

but what "smoothing" are you using in your display ? and do you really have more than 1 element per layer ?, if not you will not really see a correct gradient of stress across your layer thickness.

I'm not sure what are the specificities of the "layer" in V4.1 CAD, but they are handy to generate PML regions, or for laers in structures like yours.


For the frames, I can only suggest that you take a close look, as when you do structial, particularly of largely deforming bodies, you get strange results if you do not master correctly the Frames and their use, as in v4 the spatial frame is "on" by default (in contrary to 3.5). These are powerfull tools, often allowing you to avoid setting up an ALE system for coupled physics too.

I do not see any other way to implement anisotropic material, but take car, I know by experience that setting up the tensor correctly is something I manage often to mix up ;) So I expect its easy for other too. But with the new cooridnate arrow displays in the postrpocessing plot "More" section yu ave now at least some usefull helping items

--
Good luck
Ivar
Hi but what "smoothing" are you using in your display ? and do you really have more than 1 element per layer ?, if not you will not really see a correct gradient of stress across your layer thickness. I'm not sure what are the specificities of the "layer" in V4.1 CAD, but they are handy to generate PML regions, or for laers in structures like yours. For the frames, I can only suggest that you take a close look, as when you do structial, particularly of largely deforming bodies, you get strange results if you do not master correctly the Frames and their use, as in v4 the spatial frame is "on" by default (in contrary to 3.5). These are powerfull tools, often allowing you to avoid setting up an ALE system for coupled physics too. I do not see any other way to implement anisotropic material, but take car, I know by experience that setting up the tensor correctly is something I manage often to mix up ;) So I expect its easy for other too. But with the new cooridnate arrow displays in the postrpocessing plot "More" section yu ave now at least some usefull helping items -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24 apr 2013, 20:31 GMT-4
Dominique

Could you manage to define composite stiffness matrix in different directions? I have a L shape composite, and I need to define stiffness matrix correctly. Results show that it does not follow the local coordinates, and I am totally clueless how to define local coordinates for stiffness matrix.

Thanks,

Shawn
Dominique Could you manage to define composite stiffness matrix in different directions? I have a L shape composite, and I need to define stiffness matrix correctly. Results show that it does not follow the local coordinates, and I am totally clueless how to define local coordinates for stiffness matrix. Thanks, Shawn

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 24 apr 2013, 20:31 GMT-4
Dominique

Could you manage to define composite stiffness matrix in different directions? I have a L shape composite, and I need to define stiffness matrix correctly. Results show that it does not follow the local coordinates, and I am totally clueless how to define local coordinates for stiffness matrix.

Thanks,

Shawn
Dominique Could you manage to define composite stiffness matrix in different directions? I have a L shape composite, and I need to define stiffness matrix correctly. Results show that it does not follow the local coordinates, and I am totally clueless how to define local coordinates for stiffness matrix. Thanks, Shawn

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.