Failed elements after tet4 split into hexa

Altair Forum User

Hi,

I wonder if someone ever encountered the problem of failed hexa elements (negative Jacobian) after splitting tets4 into hexas.

I usually mesh solids with tets , directly without previous surface meshing, and split them into 4 hexas each. Often comes the message,that some node could not be placed on surface.

The tet mesh is very good however hexas often fail. Not always the same hexas (20 out of 200,000) their locations depend on tetra meshing parameters.

Thanks

Altair Forum User

Why you need HEXA mesh by this way?

Altair Forum User

Thanks for the comment,

Two reasons:

-Solid to mesh is not mappable

-large plastic strains occurring... Solver prefers hexes in order to establish 2overlays of tetras internally for compensation

Summary: I have to use hexas, their shape is not such a problem for the solver.

No one bumped into the same problem?

If I have well shaped tetras in the mesh, why does the splitting generate such hexas, with Jacobian negative?

Altair Forum User

I will come up with some more details:

 

I picked out a tetra before:

 

<?xml version="1.0" encoding="UTF-8"?>

 

and after splitting:

<?xml version="1.0" encoding="UTF-8"?>

 

shape parameters are shown below each picture. it seems like the mid surface nodes are not placed correct on the surface of the tetra and so the hexa is bad shaped.

 

EDIT:

 

I added a hm file with one of the tetras in it. If I split this tetra element, 2 out of 4 hexas are bad shaped. I added the file so you can understand what I mean.

 

Interesting fact is that there is no influence from other neighbour elements, since this element is isolated.... and resulting in bad shaped hexas /emoticons/default_angry.png' alt=''>

 

PLEASE HELP!!!!

 

EDIT 2:

 

just a side note: is it normal that when I go to Tool ->normals and then choosing the element and clicking 'display normals' HM says: 'Not able to calculate the normal of element' ?

Unable to find an attachment - read this blog

Altair Forum User

Altair Forum User said:

Thanks for the comment,

Two reasons:

-Solid to mesh is not mappable

-large plastic strains occurring... Solver prefers hexes in order to establish 2overlays of tetras internally for compensation

Summary: I have to use hexas, their shape is not such a problem for the solver.

What solver do you use?

I think alot of FEA solver work very well with TET meshing model.

From point of view of FEA quality, I would like to work with a good TET model, not a 'bad' HEX ones /emoticons/default_smile.png' alt=':)' srcset='/emoticons/smile@2x.png 2x' width='20' height='20'>

Splitting TET to get HEX mesh is possible, but that's NOT the best method to get good results.

 

Good luck!

tinh

Altair Forum User said:

Thanks for the comment,

Two reasons:

-Solid to mesh is not mappable

-large plastic strains occurring... Solver prefers hexes in order to establish 2overlays of tetras internally for compensation

Summary: I have to use hexas, their shape is not such a problem for the solver.

No one bumped into the same problem?

If I have well shaped tetras in the mesh, why does the splitting generate such hexas, with Jacobian negative?

 

 

Hello

please check your mesh is associated on any surface? if so, when you split, new node generated will be associated to the surf also and it may not be at exact position

to break association, just delete surf, or translate it + then -

no association, your hexa could be better shape

 

 

I have same ideal with Mr Q.NGUYEN-DAI. why don't you use TETRA10 instead of HEXA8?

Altair Forum User

Altair Forum User said:

What solver do you use?

I think alot of FEA solver work very well with TET meshing model.

From point of view of FEA quality, I would like to work with a good TET model, not a 'bad' HEX ones /emoticons/default_smile.png' alt=':)' srcset='/emoticons/smile@2x.png 2x' width='20' height='20'>

Splitting TET to get HEX mesh is possible, but that's NOT the best method to get good results.

Good luck!

Thanks for the info. I can only use hex8 or tet4 elements. I would prefer as well rich elements. U can believe me. have you tried the spkitting on the file I attached?

Altair Forum User

Altair Forum User said:

Hello

please check your mesh is associated on any surface? if so, when you split, new node generated will be associated to the surf also and it may not be at exact position

to break association, just delete surf, or translate it + then -

no association, your hexa could be better shape

I have same ideal with Mr Q.NGUYEN-DAI. why don't you use TETRA10 instead of HEXA8?

Thank you for the hint. I tried it. However the failed split even occurs with only one tet element in the workspace( file attached)

Have you tried splitting? With the same result?

Altair Forum User

Altair Forum User said:

Thanks for the info. I can only use hex8 or tet4 elements. I would prefer as well rich elements. U can believe me. have you tried the spkitting on the file I attached?

 

My question remains without response  : 'What solver do you use?'

Altair Forum User

Altair Forum User said:

My question remains without response  : 'What solver do you use?'

Ah yes. The solver is a in house fdm code we are actually doing some tests with it. Only using the nodes and elements for it.

However IMHO, the type of solver does not clarify or find the answer to the issue, that the tets are not splitted correctly.

Any proposals?

tinh

Hello

I tested, jacobian minimum is 0.22

new nodes are at exact position (COG of tria sides)

Altair Forum User

Thank You tinh,

 

With your aid I figured out, that even in split operations, solids and surfaces have to be eliminated to avoid any influence on the splitting. I tried it with the one tetra, but I had the solids still in the file, later I uploaded the file without solids, so it was working fine, I should have double checked that.

 

problem solved, thank You!