Mesh pattern related to Warpage Accuracy?

[dkosheba]

Is there any study that concludes the mesh needs to be random in pattern and not highly aligned on surface to provide a better warpage solution? 

Using Moldflow to mesh CAD generates a random tria pattern which I was told is good to provide a more stable solution.  The flip side are highly aligned mesh from third party software that looks like a very structured quad mesh, and then split to trias. All of the elements are aligned in rows based on the geometry of the part, and look very neat.

The reason I ask, I have part that was meshed very neat and the warpage prediction is wrong.  I then mesh the native CAD in Moldflow and the warpage prediction is correct on direction now, and only off on magnitude.  This is the same on the Dual domain part, and 3d meshed part.  

I suspect the high alignment of elements in rows is causing a problem in performing the structural analysis aspect on warpage, and with the trias aligned, it can artificially bend on the rows of nodes, and in this case, the wrong direction.

Don K.

[jim.mcguire]

Don,

This is an important question, and one I'm surprised that no Autodesk personnel have attempted to answer.  We can all explore this, but I think that Autodesk should give us a definitive "best practice" guideline for this.

Could this be a case in which large deflection 3D analysis should be used?  If the answer is so sensitive to mesh, perhaps that is why. 

Very best regards,

Jim McGuire

[madhukeshwart]

yes...Right angle tria elments ( i.e create quad elements first and then split them in to two tria element, which create two right angle tria elements) are not preferrable for structural analysis/ i.e warpage analysis.

.But moldow tech team never ever commented on that to support this comment.

I may be wrong.. If they say same.

[madhukeshwart]

I.e equilateral triangles are preferred......

[madhukeshwart]

why right angled tria elements are not preferred in moldflow and why equilateral traingles are preferred?

Is there any one can answer and explain the theory behind this ?

[madhukeshwart]

why right angled tria elements are not preferred in moldflow and why equilateral traingles are preferred?

Is there any one can answer and explain the theory behind this ?

refer attached image

[madhukeshwart]

The aspect ratio of mesh elements can affect analysis performance. High aspect ratios can cause a slower analysis, and affect results.

In right angled traingles, with the longest side in the direction of flow, the end node of high aspect ratio elements will add an excessive resistance factor to flow front calculations.

we need to avoid very high aspect ratio triangles, which have their longest side in the direction of flow.

If very high aspect ratios cannot be avoided, the longest side should, if possible, be at right angles to the flow direction.

Probably moldflow recommended aspect ratio < 6 will take care the same.

[kameshg]

I did run a comaprison with both right tria and equilateral tria,and found the results to be very similar, i did use a third party software for meshing, so the warpage prediction was off at certain regions.I dont know if the type of tria can produce a significant difference in results unless you have different global edge length.

[aniq.uppal]

Don,

Take a look at Mesh statistics of the mesh models, compare the mesh match ratios. I believe Moldflow recommends mesh match ratios of 85-90% to get accurate results

Aniq

[sabarinathan1021]

Hello,

Mesh match % (more than 85%) is the key to get a warpage results closer to the reality.

If you could get a mesh from third party software with higher mesh match % (more than 85%) will get you a better results.

Equilateral triangles are better than R-trais.

Right angle triangle mesh will take more time to solve fill analysis & warpage may be results will be different.

So It is recommended aspect ratio < 6, use Equilateral triangles as much as possible, preferably in Base wall.

Some percentage of right angled triangles are ok in B-side feature like rib, boss, which contributes much for results.

[madhukeshwart]

Is there any study that concludes the mesh needs to be random in pattern and not highly aligned on surface to provide a better warpage solution? 

1. Mesh should be in random Pattern for better flow calculation.

2. If Mesh is highly aligned, then better they should be equilateral triangles

[kameshg]

hi,

i dont think random mesh pattern is good, not only for moldflow but for any FEA software, Since the results calculated based on the nodes,a uniform pattern would be ideal or atleast there should be a gradual change in pattern.

regards,

kamesh

[kameshg]

Hi Don,

can you share the image of those files, may be we are missing something.

kamesh