Missing Support


#1

I’ve oriented this model numerous ways, and automatic support generation misses extremely obvious spots where support is required. The only thing “wrong” with the model is a few inverted normals, otherwise it passes inspection.


#2

Hi Paul,

Which normals are inverted?
Normals are crucial in your support generation. In general: rubbish in means rubbish out.

You can send me the model over email/wetransfer so i could have a look for you.

Elco


#3

Hi Paul,

I’m not sure what was wrong with this file; but i get the feeling it’s entirely inverted of some sort.

When i imported in Rhino and then saved as .stl it does import correctly…
Please see attached file.

maza-fixed-2.stl (6.1 MB)

Elco


#4

As an add on:

formware reads the triangle faces from the stl including the normals.
Rhino probably reads only the triangle’s coordinates and then creates the normals according to the right hand rule.

It’s a bit of an ‘industry’ error that there is no solid guideline for this and in a way there is double information in each stl file.
Each program handles this differently…


#5

i forgot to paste in the image (perhaps for other readers)
The image on the right is the file you send me. The image on the left after getting it through rhino.
What this shows me is that all normals were not correctly defined in the file. (in formware 3d the entire shader and critical faces are inversed)


#6

Thank you very much! That is interesting. The very original file was not manifold (which formware picked up). I loaded it into meshmixer to make it manifold. Perhaps meshmixer messed it up while making it manifold, or perhaps it was messed up from the beginning and meshmixer just didn’t fix that. I’ll toy with processing with rhino next time I get this problem (or perhaps all STL’s from this source, which is Hero Forge)


#7

Still though, some are missed. A LOT less, for sure, but for example, you can see an “airprint” here (is that the right term for resin? probably not). It is a lot easier to fix up now, and I think that seemed to be the only one, but ideally something like this wouldn’t be missed. I toyed with sampling strategy/surface sampling but it only got worse, not better. It is a very very small feature, perhaps that is why it is being missed? (this was just loaded, auto-oriented (which has a bug on the number of tries…) and then auto-support with basically the defaults)


#8

mm. i see. This is a bit of an exception case.

The placement of columns generally has a lot of logic in determining intersections; finding alternative locations.

I think in this case the distance is very short.
Solution I see is that I add a very short/thin column for these scenarios… will add it to the todo list and use this model as a test :slight_smile: .