What is the best software for creating Hybrid Meshes? 1

[feajob]

Hi,

I am looking for more information about Hybrid Meshes. I heard that Ansys has pyramid elements useful for interfacing hex and tet meshes in a 3D model. I don't use Ansys (I use Patran & HyperMesh), but I would like to know what is the best tool for this purpose. Please share your experience with me.

Regards,
AAY

 

[gwolf]

Try modelling a cantilever beam half along the length in 8 noded bricks, half in 10 noded tets. Equivalence the nodes of four 8 noded bricks arranged in a square with two 10 noded tets. Put two or four bricks through the thickness and respectively one or two tets through the thickness.

Yes, that's right, you end up equivalencing the mid-side nodes of the tets with corner nodes on the bricks!!

*********
* * **
* * * *
* ** *
*********
* ** *
* * * *
** * *
*********

You may be pleasantly surprised by the results.

 

[gwolf]

ps, my diagram seems to have distorted during the font change - oh well, I'm sure that you can work it out

 

[feajob]

Thanks gwolf. In order to use quadratic elements (hex20 & tet10) I think that we need to use some special elements (I heard about pyramid elements in Ansys)? Do you have any experience in this regard?

AAY

 

[feadude]

I do not use patran but IDEAS uses MPCs to connect hex and tet elements together. Here is a brief from IDEAS help.

"To use this capability, you must first define the mapped mesh and then define the free mesh. You can use the following combinations of elements to mesh adjacent volumes:

linear hexahedral elements and linear tetrahedral elements
linear hexahedral elements and parabolic tetrahedral elements
parabolic hexahedral elements and parabolic tetrahedral elements
When you're meshing adjacent volumes with linear hexahedral and parabolic tetrahedral elements, or parabolic hexahedral and parabolic tetrahedral elements, the software must create new nodes on the tetrahedral elements to properly connect them to the hexahedral elements.

The software uses multi-point constraint equations to connect the nodes on the tetrahedrals and hexahedrals together so that the nodes' three global degrees of freedom (U, V, and W) match up correctly at the interface. These multi-point constraint equations give the average displacement at the midside nodes.


Understanding the Parabolic Hexahedral-Parabolic Tetrahedral Interface
When the software creates an interface between parabolic hexahedral and parabolic tetrahedral elements, the tetrahedral elements share the corner nodes and the mid-side nodes with the hexahedral elements. This leaves the mid-face nodes on the tetrahedral elements unattached.

With a parabolic hexahedral-parabolic tetrahedral interface, the software creates an additional node on each tetrahedral element's mid-face. This mid-face (center) node is connected to the four corner nodes and to the four mid-side nodes with one set of MPCs."


good57morning@netzero.com

 

[corus]

Abaqus will automatically tie two surfaces of dissimilar meshes together using some interpolation method. It's best to use this away from areas of interest in the model though.

corus

 

[Drej]

> In order to use quadratic elements (hex20 & tet10) I think that we need to use some special elements (I heard about pyramid elements in Ansys)? Do you have any experience in this regard?

ANSYS does use these, but then any mesher/pre-processor worth its salt (Hypermesh, Patran, ICEM) with elements having degenerating capability will do the same. There is no magic or mysticism about them, they're just automatically created at the tet/hex interface. In ANSYS you need to change the degenerate elements into the more efficient 10 noded tets after the meshing operation. As mentioned in the IDEAS notes above, you also need to make sure you hex mesh the part first, then free mesh (with the degenerate elements) the adjacent part.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[gwolf]

The trick which I describe applies to NASTRAN and ABAQUS which both have very good 8 noded bricks, I don't know about ANSYS.

I have recently been dealing with a model where I have *HAD* to equivalence 20 noded bricks and 10 noded tets in a very small area of a model. This is where you match two Tets to one brick. So you end up with one of the tet mid-nodes unequivalenced. The results were a *LOT* better than I thought they would be.

 

[feadude]

gwolf
I made a cube 1"x1"x1". I meshed it with 1 element 20 noded brick. Then I deleted the mesh. Then I meshed it with 10 noded tets. It took 12 elements to fill up the same cube. The length of the element was 1" in both cases. How did you match 2 tets to one brick? For me it took 12 tets to match one brick. Both quadratic elements.


good57morning@netzero.com

 

[gwolf]

fedude,

Equivalence the boundary nodes on a single 20 node brick 1x1x1 with an identical adjacent 1x1x1 tet mesh with two triangular facets on the boundary between the two cubes. I don't know how well this will work when loaded, I only used it in a quiet area of a model.

 

[feajob]

feadude,

Thanks for your response. Based on your experience, attaching hex20 to tet10 elements can be considered as a quick procedure? AS you wrote, IDEAS uses multi-point constraint equations to connect the nodes on the tetrahedrals and hexahedrals together so that the nodes' three global degrees of freedom (U, V, and W) match up correctly at the interface. I guess that this step is done automatically.

At the same time, I would like to thank Drej and gwolf, as you (both) mentioned, other softwares can also do this task, but I understood that there are more human interventions and it is not very practical when you have to deal with this technique in some regions of your model.


AAY

http://www.geocities.com/fea_tek/

 

[feadude]

yes the interface transition is done automatically. Hex volume should be meshed first.

good57morning@netzero.com