Meshing a Triangle with Tri elements

[RGX124]

Hello,
I am trying to mesh a triangular shell (100x100x100) with tri elements. (see the images in the Zip file)

1. When I specify a size of 3.125 per edge (Free mesh control), the resulting mesh is uniform and structured. See image free-size3125.jpg

2. When I specify a size of 2.0 per edge (Free mesh control), the resulting mesh is not uniform anymore. See image free-size2000.jpg


3. When I specify a size of 3.125 per edge with a structured mesh, the resulting mesh is "strange" . See image structured-size3125.jpg


Questions:
Can I control this problem (the non uniform mesh) without having to partition my structure?
What are the limits of CAE according to meshing techniques?

Thanks for your reply

Regis

 

[vumat721]

Have you tried specifying number of elements per side instead of size?

 

[RGX124]

Hi,

I have just tried it.
It gives the same (bad) results.

R

 

[corus]

What is 'bad' about the meshes? In all cases they look fine to me. Use the mesh verification tool to see if any would give warnings or errors.

corus

 

[RGX124]

Hi Corus,

According to the mechanical response, it is not a "bad" (consider the "") mesh. It even minimise the number of elements.
However, if i need to control that integration points should be on a line parralel to the triangle edge, it would be easier to get a structured (or mapped) mesh.
The only way I found to have a structured mesh was to partition the triangle in smaller triangles.

R

 

[rstupplebeen]

I agree with corus. Only for trivial meshes can you get an absolutely uniform mesh like the first example. As long as your aspect ratios are good the mesh is good. I would mesh this with quads instead since triangles are stiff and 2nd order triangles increase solve time. I hope this helps.

Rob

 

[corus]

In the 3rd example you can see that Abaqus has partitioned the region into 3 quadrilateral regions, rather than triangular regions. It has done this by taking a line from the mid-point of each side to the centre. I'm not sure why you would need the integration points to be on a line though. It seems to be taking the problem to a fine end of a .. whatever. Quads are better elements as rstubblebeen suggests, and in my view, give smoother results than triangular elements.

corus