Problem Shell Pair

[cmora]

Hello

1.- When using Detect Shell Pair it does not detect the filletes (R1.0 and R2.0)

The surface is not completed.

What is the problem ???

The whole model is made in PTC CREO 4.0

2.- By using Shell Pair manually and with the option to extend adjacent surfaces, the average surface sought is formed.

By using the AUTOGEM Review Geometry, it shows error in the edges

I have adjusted the absolute and relative precision variables, to be able to form those fillets (R1.0 and R2.0).

I assume that some variable needs to be adjusted in order to achieve the mesh of the average surface.

Regards

Carlos Mora L.

[SweetPeasHub]

The problem is with modelling.

My understanding is that auto-detect only detects thin solids, shelled solids, and thickened quilts.  Your rounds are not made by these.

In my opinion a good modelling practice would be to include the fillets in the revolved cross section even if not simulating.

Thin solidAutomatic Shell

 

Also useful is [REFINE Model]  Review Geom

Consider whether you need to model the small fillets.

[gfraulini]

Hello,
I don't think it's a good practice modeling thin radius with shells. The curvature is too high and you have to use solid elements.
I suggest you to insert volume regions "on the radius" in the way you extrude the volume along the length.
I don't remember the right curvature value beyond which you can use shells...

[SweetPeasHub]

Thanks gfraulini.

According to this documentation, the curvature is OK but getting close.

http://support.ptc.com/help/creo/creo_pma/usascii/index.html#page/simulate/simulate/modstr/props/reference/sh_thickness.html

Note: Use the midplane surface for the curvature analysis.

This design is curvature 0.667 * thickness 1.0 = 0.667 < 1.5 so it is OK.

According to this the minimum radius would be 0.17 (0.67mm radius at the midplane)

Curvature Measurement

I personally would not use shell pairs, I would use a quilt with a shell property.  That would allow a mapped mesh with only quads and a much better mesh and faster simulation. The number of elements and whether I included the small fillets would depend on what I was trying to accomplish.  If this ends up as a wheel with spokes on a full model of a bike I would include less details.  If I was only interested in deflection/stiffness/mode shapes I would omit the rounds.  If I was doing a design sensitivity study I might check the difference with and without rounds and then decide whether to include them for a multiple-run type of study. 

mapped quad shell mesh

[gfraulini]

Are you able to make a 2D-3D mesh where shell elements touch bricks or wedges which derived from a mapped control?
I tried but it didn't work...
The shells were made at the mid-plane and the mapped control created nodes in the thickness middle so that the two kind of elements can share nodes.
Have you any suggestion?

[SweetPeasHub]

I am not sure whether i have tried a mixed mesh with mapped elements. however you will need some connections more than the edge of the midplane because solid nodes do not transfer rotations.

[gfraulini]

I can't mesh with mepped elements, but I can with tetra. Why?

[SweetPeasHub]

Mapped meshes are tricky.  The internal nodes will not connect to neighbor elements even though they appear to match.  The nodes/split must be in the geometry and used by the mapping.  The thick lines show the separate mappings.  I split your model into 4 volume regions as shown.  You will need to increase the # elements along the length.  50 worked for me.  I used a conic as a middle ground between the radius and the sharp.

must be split at midplane.

Review model

Also, here are some visuals for a cantilever beam with mixed mesh, all mapped.

Final mesh

Review Model

Volume region splitting

 Also it would be a good idea to look at some articles on mapped meshing like mine.  

[gfraulini]

Thanks a lot! I never understand how one can use really this kind of control. What I missed is "The nodes/split must be in the geometry and used by the mapping"

[SweetPeasHub]

Glad I could help  It took me a while to understand mapped meshes.  

You must understand solid - shell connections.  Why they need links. Go Here.

This is a different software but so similar it is worth looking through to get some beginning understanding of being mappable.  Creo currently is more difficult, it requires more splitting and handling of transitions.

This youtube video.

and this PDF.

For CREO.

Also this tutorial.

And an answer I gave here.

regards,

[cmora · Accepted solution]

Thank you very much to all

Unfortunately i had to use Ansys for the analysis.
I am currently modifying an existing Tank so that it has 03 compartments.
Also use PTC CREO SIMULATE but there are singularities that I could not eliminate in the end.
Personally really like PTC CREO.
Maybe i'm lacking in practice in the mesh methods that PTC / CREO uses, 
Thank you very much for your help.