Meshing repeats along boundary and scaling/round off error?

[mdgiovinazzo]

I don't have any scripts to illustrate this right now, but I wanted to frame the problem for now to motivate me to do so.

Consider the following repetitive geometry:


When we mesh it using separately instanced yarns, it works fine:


When we mesh it using yarn repeats, it goes weird (matrix hidden):



I really wish I had the script used to produce these files, but I made them quite a while ago.


The second problem uses the same geometry as before, but just one unit cell:


If we mesh it using certain dimensions, it works great (matrix hidden):


However, when we divide all dimensions by 100, the mesh gets shot off beyond the boundary:

[louisepb]

Hi,

I'm currently working my way through the meshing code, trying to get my head around how it works. If you can give any more information with your examples I might be able to see what's happening. I know that there have been issues with the meshing so any examples to illustrate problems will help in getting it working well.

The example with yarn repeats looks very strange! Was it rendered ok in the user interface?

When you divided the dimensions by 100 did you alter the seed size and merge tolerance? It could possibly be floating point arithmetic issues.

Do you have the tg3 file for the textile? If you could send it to me then I could take a quick look and see if anything obvious springs to mind.

Just a few thoughts,
Best regards,
Louise

[louisepb]

Hi,

I've managed to replicate the error with repeats and think I have an idea of what the problem is. May take me a bit longer to work out how to solve it!

Louise

[mdgio]

Hi, this is Michael Giovinazzo again. It seems like the password reset email is having trouble reaching me.

I'm sorry for not answering your question about a different tolerance on those two meshes. I believe they were different, but the most important part seemed at the time to be the numerical dimensions. You can observe that the thresholds were both reasonable (maybe a little too low in the first case) by looking at the edges of the yarns as they overlap. Shown here:



I don't have those old files, unfortunately. I would like to try to make some example scripts showing the problems I've observed but my time has been limited.

I've found some different meshing problems recently. I really wish I could share the scripts with you but my hands are tied on them. This makes me want to make example scripts to reproduce them even more.

The basic geometry is as follows. It's a 3D weave generated using CTextileWeave3D() and PushUp(), PushDown() with a modification to the major yarn nodes by using InsertNode(). It is bounded by AssignDefaultDomain() or a manual AssignDoman(). The dimensions are about 0.4 x 0.4 x 0.15.



I haven't seen these types of meshing problems before, so it may be new in the June 2010 version. I'll be trying some old versions to see if I can fix the problem and I'll be sure to let you know if I've fixed it by doing so. The second picture with the four spiked edges is somewhat similar to the previous problem in this thread, though.

They are meshed using various combinations of SetQuadratic(), SetPeriodic(), SetProjectMidSideNodes() True and False. They are meshed with SetSeed(.0075) and SetMergeTolerance(0.001) or slight variations thereof. Various combinations of geometric scales, positions and boundaries have been attempted to fix the problem without success.

The second picture's problem is more prevalent than the first.




Kind Regards,
Mike

[louisepb]

Hi Mike,

If you find any errors that have been introduced in the new version I would be grateful if you could let me know about them. I don't think that there were any changes to the meshing in the newest version so it's not obvious where they might have been introduced.

As far as the meshing of stacks of yarns goes there is an issue with the way that the extents for meshing are calculated in the z direction when there is a z axis repeat. The best way to resolve this at the moment is to assign different yarn numbers to each yarn in the z direction and then the yarns will be meshed individually. It would need some fairly major rewriting of the code to sort this out and is not one of my high work priorities at the moment so I'm afraid it's a case of duplicating yarns in the z direction rather than assigning repeats for now.

I hope that helps,
Louise

[mdgio]

I wanted to update that I resolved the meshing problem I was having (on Jul 23, 2010) by using a resolution of 20 instead of 32. It did not seem to be a result of any version changes, and everything is working well now.

I agree that the z repeat issue is not a big deal as there is a simple workaround for it.