FE mesh problems - simple silo model

FE mesh problems - simple silo model

Anonymous
Not applicable
976 Views
10 Replies
Message 1 of 11

FE mesh problems - simple silo model

Anonymous
Not applicable

I would like to ask how to use the built-in tools to generate a simple silo model with a consisted mesh along the edges and densified on the sidewall and in the funnel when both are modeled by extrude+scalling ?

Is it possible to set individual parameters for FE mesh generation for these elements to satisfy user expectations ?

Maybe I don't know something or meshing in Robot  is still its weak point and has not changed from 20 years ???

 

What are the other ways to generate any tank model with a correct mesh along edges and with a given density ?

 

Bez tytułu.jpg

 

0 Likes
977 Views
10 Replies
Replies (10)
Message 2 of 11

Krzysztof_Wasik
Autodesk Support
Autodesk Support

Hi @Anonymous 

If model is defined precisely (equal number of  division for cover and tank side shell and corner points defined in the same position) decreasing mesh size for top cover should help. Set mesh size for cover, equal or bigger than side shell mesh size. 



Krzysztof Wasik
0 Likes
Message 3 of 11

Anonymous
Not applicable

Dear Krzysztof;

Sorry, but what are we talking about?
About a primitive mesher and primitive approaches that everyone knows, or about a professional engineering program?
The truth is, the mesh engine hasn't been touched since two decades.

I will show you a short test below.

A different approach without extruding a 2D arc or using other solid modeling options.
First, I created 4 arcs (divided into 10 parts) for the top and bottom levels of the funnel, added vertical lines for the edges of the panels.
Then I defined the appropriate panels. For the funnel, I took Coons method to generate a mesh with user-edge divisions of 20/20 elements.
The FE mesh looks smooth and corresponds to the arc divisions. In the case of a funnel, all division methods (triangles with rectangles or trapezoids work well).
I also took Coons for the sidewall and tested the mesh generation with all the options available. Neither of them are working properly!
Robot cannot generate the mesh that corresponds to the division of an arc edge. I wonder why this is possible for funnel cone walls but not for a straight cylinder wall.
FE mesh elements do not "stick" to the surface and nodes extend beyond the surface, deplanning the shape of the shell. The FE nodes at the junction of the wall and funnel do not match!
Delunay method also doesn't work - at funnel level every 2-nd node of wall matches funnel node, at the free edge on top of a cylinder FE nodes are outside of the component lines of the arc.

Please find attached model and answer me like a Professionalist how to solve the problem?
Why does Robot have such basic meshing problems for curved surfaces?
It cannot be a constant "battle" with Robot, I lost 2 hours...and nothing !

Other programs don't have any problems (whole operation time modelling + mesh generation took 2 minutes)

 

funnel-side wall joint with Coons.jpgother program.jpg

0 Likes
Message 4 of 11

Krzysztof_Wasik
Autodesk Support
Autodesk Support

Hi @Anonymous 

You have not mentioned about your first problem (deck meshing) so I understand that solution with changing mesh size works.

Regarding your second test (cone to shell meshing) I would recommend definition of arc division, (20) the same for cone and side shell by editing  Object properties.

See attached video



Krzysztof Wasik
0 Likes
Message 5 of 11

Anonymous
Not applicable

Hi Krzysztof;

 

In my first post, I described the "general" mesh incompatibility problem for objects generated as extruded using the solid modeler built-in tools (Geometry menu> Objects> Extrude & Edit menu> Modify Substructure> Modify Object)
and others simple objects (i.e. flat panels).

I did not continue this approach because of the silo will only be a small part of my complex 3D model.

I think any modification of the adjoining structure will result in errors in the mesh connections in the future.
The generated funnel and cylinder objects are "locked" in Robot and cannot be modified. This is a "one-time" operation and the mesh parameters do not work for such generated objects.

It's like creating a mesh manually by deliberately dividing an object - this is not today's method.

This should be done by an automatic generator for all user-set parameters.

 

In my second post, I would like to highlight the issue and ask why the standard approach to creating panels with edges from simple geometric objects (arcs in this case) leads to incoherent meshes along common edges.

Why cannot Robot handle problem with mesh generation between ordinary cone and cylinder surfaces?
This is such a simple geometric case that I could write a macro to generate FE nodes from rectangles and trapezoids or import from other program - but this is not an expected solution - rather Robot should be able to do it for any 3D model.
Therefore, I also gave an example from another FEM system that was done in seconds and correctly!

 

PS.
Thanks for the video !

 I'll try to divide the edges like you did, but I'm worried if it works in my 3d model and about the size of the task  (extracting the results for the entire model).

0 Likes
Message 6 of 11

Anonymous
Not applicable

@ support TEAM

Please add into the Robot newest manual a note that the arcs must be discretized on at least 20 elements - for the meshing to work properly. It seems that this is a knowledge only for the "insiders" or code authors.

It will save a lot of time for users frustrated by similar problems.

0 Likes
Message 7 of 11

Krzysztof_Wasik
Autodesk Support
Autodesk Support

Hi @Anonymous 

This solution works for smaller number of division as well (at least in your case). See attached video for number of divisions=5.



Krzysztof Wasik
0 Likes
Message 8 of 11

Anonymous
Not applicable
Thank you very much for your excellent support on Sunday! 🙂

Now tell me what are the rules because it can't be roulette?
The 10-part split doesn't work, but the 20's and 5's work ... some magic!
0 Likes
Message 9 of 11

Krzysztof_Wasik
Autodesk Support
Autodesk Support

Hi @Anonymous 

All magic is in arc discretization activation. It works for division number =10 as well. See video

 

 



Krzysztof Wasik
0 Likes
Message 10 of 11

Anonymous
Not applicable

I'm not sure - better zoom on the common edge, because it looks good from a distance only. 

What does the value "10" mean after the coordinates of the arc nodes - not the number of divisions?

When I defined arcs, I had this value set to "10"

 

Please look at results:

 

arc_1_10.jpg

 

Please upload all modified models to check if it works in each case.

 

0 Likes
Message 11 of 11

Anonymous
Not applicable

After the tests, a small conclusion: the number of FE elements on a common arc cannot exceed the number of its divisions. And such a value must be set at most in the mesh parameters to the Coons method

0 Likes