0

I am modeling a composite cube block with spherical particle and spherical interphase around particle. I can simplify two spheres to two cubes. I am trying to model particle inclusion and interphase in a composite material. I am taking matrix as a cube and particle as a small cube at the center of matrix cube. Interphase layer will be another cube at the center of matrix cube but bit bigger than particle cube. Matrix, interphase, matrix all have different properties. But problem is I do not know how to model a cube inside a cube. When I assembled three parts, they can get 'separated' during simulation. I cannot model all 3 cubes as a single part. If I do boolean, the innermost geometry gets deleted. Importantly, I also want to make different sets for different parts.

Any suggestions on how to model this?

user3215074
  • 103
  • 1
  • 2
  • 7
  • Sounds like a numerical proof of Eshelby inclusion: http://micro.stanford.edu/~caiwei/me340b/content/me340b-lecture02-v03.pdf. Very interesting, but I don't think it belongs here. It's not a programming question; more about geometry and modeling. – duffymo Feb 11 '14 at 14:27
  • 4
    This question appears to be off-topic because it is about finite element modeling and geometry, not programming. – duffymo Feb 11 '14 at 14:27
  • I am seeking some ideas with this finite element modeling. – user3215074 Feb 11 '14 at 15:35
  • Yes, I get it. That was clear from the start. – duffymo Feb 11 '14 at 16:02
  • Thanks duffymo for your answer. I made some comments on it. Can you please see them and explain your answer? – user3215074 Feb 11 '14 at 16:55
  • perhaps a better place for this https://groups.yahoo.com/neo/groups/ABAQUS/info. I think the boolean approach should work if you check the "retain boundaries" box..by the way. You may need that if you try to tackle the sphere problem. – agentp Feb 12 '14 at 12:55

1 Answers1

1

You don't say what kind of elements you plan to use.

If they're 8 node linear bricks or 20 node quadratic bricks it's an "easy" problem: just create cubic regions for each different material and create your mesh.

I'm guessing that the interphase/interface layer is "thin". You'll have some aspect ratio/transtion problems to sort in the matrix and inclusion models.

This can also be done with a spherical inclusion with a shell interface/interphase between inclusion and matrix, but it'd be easier with tetrahedral elements and an automatic octree meshing solution.

You can exploit planes of symmetry: three in this case. Symmetry/antisymmetry boundary conditions can help you combine different loading methods.

You should not be using boolean operations to create this. Just create a cube with regions for each material. It's straightforward.

Can't show in 3D, but 2D will give you a hint:

+--------+---+---------+
| 1      | 2 | 3       |
|        |   |         |
+--------+---+---------+
| 4      | 5 | 6       |
+--------+---+---------+
| 7      | 8 | 9       |
+--------+-------------+
  1. Inclusion: region 7
  2. Interface: regions 4, 5, and 8
  3. Matrix : regions 1, 2, 3, 6, and 9.

Set material properties in each region appropriately.

Add boundary conditions to the left vertical and bottom horizontal surfaces. Loadings are applied to right vertical and top horizontal surfaces.

Now you should see what I'm getting at: the interface regions are probably thin, and their material less stiff than the inclusion or matrix. You'll have very small elements in that area in order to maintain good aspect ratios, which means lots of matrix and inclusion elements.

I'd recommend doing a simpler case in 2D, like the one I have here, to get a sense for the behavior before you try 3D. You'll be able to turn it around quickly and even check convergence before you tackle the bigger problem.

duffymo
  • 305,152
  • 44
  • 369
  • 561
  • Thanks for your comments! I am using 8 node linear brick element. Can you elaborate more how to create 'cubic regions for each different material and create your mesh.' I am assuming all this is to be done in a single part as making assembly will unnecessarily complicate the matter. I also don't want to use contact. When I model innermost cubic region and then outer layer of interphase, abaqus merges two regions. If I create two cubes and mesh inner cube by suppressing outer interphase, when I resume outer layer, abaqus deletes mesh from innermost layer. Basically I can't model this geometry. – user3215074 Feb 11 '14 at 16:31
  • I agree - contact isn't the answer, because you're assuming perfect adhesion between matrix, interface, and inclusion. You should not be using "boolean" operations when you create the geometry. Just create a cube with different regions for each material. – duffymo Feb 11 '14 at 17:05
  • The 2D image should make my suggestions clear. – duffymo Feb 11 '14 at 17:47
  • Hi duffymo, thank you very much for your succinct comments! You are really an expert of ABAQUS! Thanks! – user3215074 Feb 11 '14 at 18:18
  • I have a question here. Excuse me for naivety in ABAQUS. You said 'Just create a cube with regions for each material' So are you referring to create partitions or something? Also my cubic model has displacement constraints on two parallel faces of cube such that faces remain parallel even after deformation. if I use quarter symmetry model, can I do this? I am asking this because I did not do this before. Also if I switch to full 3D model, can I still use partitioning cube and assigning different materials? – user3215074 Feb 11 '14 at 18:25
  • I don't know what the word "partitions" means. I've never used ABAQUS, but I'm familiar with the founders of the company and their product. I've used other packages (NASTRAN and MARC). I haven't done FEA since 1995, so I'm relying on experience that's almost 20 years old now. I would have created geometry for my problem in a pre-processor and then applied a mesh to it in a separate step. Is that what you're doing? You can do what I described in 2D to a 3D model; the geometry is just more complex. – duffymo Feb 11 '14 at 18:35
  • "displacement constraints on two parallel faces of cube such that faces remain parallel even after deformation" - this would suggest that it's in pure shear. there's no need for a 3d model in that case. my 2d example would be sufficient. it's a slice parallel to your two parallel faces. if you look directly at it, i think you'll see that there's no variation perpendicular to that plane. – duffymo Feb 11 '14 at 18:48
  • Thank you once more. Apparently what I have to do is to create datum planes and cut the geometry with datum planes to get different 'cells' or partitions, as you showed in your 2D figure. By the way, you say 'but I'm familiar with the founders of the company and their product','I haven't done FEA since 1995' and still your comments are very relevant. I am thinking you must be a legend in this field. Anyway, thanks you very much!! – user3215074 Feb 11 '14 at 18:50
  • so kind of you to say. no, i'm a forgotten man who's moved on. i'm very glad that i was able to help you. do look at that paper about the eshelby tensor for inclusions. it's a beautiful result. – duffymo Feb 11 '14 at 19:01