STL conversion to surfaces won't stitch

[hicksanGZHFL]

I am working on an STL, downloaded from Thingiverse, of a USA 1955 GM bus. The model has been repaired in Netfabb, loaded as a mesh, reduced slightly and then converted to BREP form and then Validated into a set of surfaces. The problem now is the main object surface appears to be in some sense corrupted in that I cannot get it to stitch, even perfectly fitted patches.

What is clear is that the STL was generated based on a 3D scan of the vehicle, giving an interior and exterior, and this has produced numerous surfaces that are folded over on themselves, with layers folding though each other and masses of tiny slivers of faces, and that this complex geometry is (apparently) causing errors. Unless I am doing something else wrong. Attempts to repair the mesh resulted mostly in huge holes appearing. The mesh body has an interior as well as exterior, and most of the problems occur where these two collide or intertwine, such as the front and rear windows.

I have deleted the entire interior - I want to reduce the model to a closed shell that can be made into a solid, as it will be printed in a very small scale. I have also deleted most of the complex corrupted folded surfaces - mostly concentrated in only small areas of the model - reducing it from 35,000 faces to only 7,000 as a result (and most of that was the right windscreen which got wrapped around a handrail by the door into a dense column of shard faces). I intend to use sketches, patches or lofting to fill these holes, and have already created some patches for the simplest closed flat triangles - but I cannot get these to stitch. Fusion hangs for several hours then returns all the original surfaces - body and patches - renamed as "body...". Attempts to repair or adjust it while in mesh form were similarly fruitless, with long pauses and then no action - even on simple tasks such as merging two adjoining triangles into one.

My question is: how to solve this. I am pretty new to Fusion and am treating this as a learning exercise, but it is beyond anything I have found in any tutorials. So rebuilding the whole thing from scratch is not an option at this stage: I am trying to figure out what is wrong with this file, how to locate the errors and how to fix them. I am guessing that there is still something wrong with the main body surface; some illegal geometry that renders it unworkable, but I am unable to figure out how to locate it and how to fix it, beyond deleting the offending faces and then patching the holes. Validate has been useful but no longer reports new errors. I have attached the F3D file. Please can someone help? Thanks very much.

Andrew.

[hicksanGZHFL]

Here is the F3D file. It wouldn't send earlier.

[TheCADWhisperer]

As you have observed - stl is generally rubbish in terms of geometry.

---

@hicksanGZHFL wrote:

My question is: how to solve this.

---

I would use this only as reference in creating proper smooth geometry.

[hicksanGZHFL]

Thanks Cadwhisperer, but I'm afraid telling me the file is garbage doesn't help do what I am trying to do; which is understand what, exactly, is wrong with this file and how to use Fusion fix it. Under normal circumstances I would abandon it and make my own, as you advise, but that isn't going to help me learn in this case.

I have looked at the geometry as closely as Fusion allows. The holes in the model are where I have deleted tens of thousands of sliver sized triangles, where the sides have folded in on themselves. Some of the slivers are so small that I am unable to get Fusion to select them for deletion at any magnification. It is these error faces that pushed the face count so high, and probably render the main body somehow corrupt so it won't stitch. I understand it only has triangles. Isn't that supposed to be the whole basis of STL files? So I would be surprised to find anything else. At the scale I wish to print this (1:220) that will not be an issue.

I got the face count down to 5,066 but then validate stopped working. It highlighted that there were a further 121 illegal faces plus 202 intersecting/overlapping, and then deleted and restored them - leaving things back where they started with 59 error messages saying that removing those surfaces would render the model corrupt. From that point I could go no further.

I am asking if it is possible to fix this using Fusion 360, and if so how; not whether it is worth doing. But thanks very much indeed for taking the time to look at the file. I genuinely appreciate your attention and advice. If Fusion is not up to the job then that is fine: at least I then know that it is not that I am doing anything wrong. Fusion will have shown its limitations and I will know to work around them. That is not a criticism of Fusion, which is frankly superb software. Some things are just not possible and others can be overcome with sufficient effort. I would like to know which this is, please?

[TheCADWhisperer]

Pinging @TrippyLighting 

Where is your thread on working with Mesh geometry?

[TrippyLighting]

I am assuming you are referring to this thread on working with irregular and organic shapes.

[wmhazzard]

I don't understand what you are trying to do in Fusion. If the original STL was repaired in Netfabb, why not just 3D print the repaired STL file. If you need to change the bus in some way then Blender or Meshmixer would be better programs to use than Fusion. 

[hicksanGZHFL]

Short answer: because Netfabb didn't work properly.

That is not to say Netfabb wasn't helpful. It fixed up the mesh enough to be able convert it to BREP in Fusion, and that was a big step forward. However it still left two big (huge) problems: (1) the model still includes an interior space that cannot be filled in Fusion - and I need a solid shell to print it - and (2) the surface geometry remains folded over on itself so badly that it would never print anyway.

I have tried printing models like these before: the walls are so thin that I know from experience it will not print properly. I am reducing it in scale to 1:220 (tiny) so I need to make it into a solid object, which means first deleting the interior so only the outside shell remains. Because the geometry is corrupt the 'boundary fill' command in Fusion doesn't work. In fact at first pretty much nothing works.

If you (or anyone) is interested I attach here the original and the 'fixed' versions of the STL. If you load either into Fusion they look - superficially - to be fine except for a strange frosting over some of the windows. However when you try and do anything - like 'convert to solid' - it ends up with huge holes where these folded faces have collapsed.

I have tried loading it into Fusion, Meshmixer, Blender and Instant Meshes but get the same problem with all of them: none of the smoothing tools out there is able to 'unfold' this fractured geometry.

The only solution I have found that works is to locate and delete any folds in the surface, in Fusion, using the Inspect tool in Meshmixer to show me where these are by exporting the broken surface to STL and loading it into Meshmixer. Validate (surface) in Fusion is also very helpful - removing hundreds of shards of fractured folds every time it works - but it doesn't always work. Often it gets stuck and returns an error message that repairing the body will make it even more corrupt. The way to get past these situations is more manual deletion based on Meshmixer diagnosis, and then if you are lucky and delete the problem areas, it starts again.

I have now reduced this mesh from 35,000 surfaces to 4,756 and now it is not only validating but finally is allowing 'stitch' to be used - which was the whole object of this query, so I can begin the long process of patching the holes and restoring the model. I'll post an update later to show progress in fixing this blasted file. Thanks very much to everyone for all the help and advice so far. I really appreciated watching the videos, guys. I am happy to sit at your feet and learn.

[TrippyLighting]

You did not attach any files 😉

[hicksanGZHFL]

Here are pics of the original STL in Fusion, as mesh and converted (35,981 faces), showing the frosting, along with my 'fixed' version (with only 4,436 faces).

[hicksanGZHFL]

This is what happens to the original in Meshmixer:

[hicksanGZHFL]

And here is the body, scaled to 1:220, successfully printed last night (@ 20 micron layer resolution). Solid object, hollowed out with 1mm wall thickness. Very pleased with that. Only detail lost was the wing mirror - and at that scale I expected it. It should paint up very nicely. Now for the wheels, which are corrupted even worse!

[hicksanGZHFL]

This is just to report that the wheels are now done - I took a cross-section profile off the broken original and extruded it into a wheel, made two copies - one for front and t'other for rear, so I could flatten the reverse side of each (they were back-to-back) - then sliced off the wheel nuts from the original and added them to the new front wheel. My first actual 3D object model. Also had to mirror the bus body as the person wants doors on the left, which took less than a minute. Anyway it all printed perfectly. Case closed and thanks very much for the suggestions, words of encouragement and most of all for your interest in reading this. Pics attached.

[hicksanGZHFL]

Just for reference: here are two screenshots of the original wheel - making obvious why it would never be printable (part of the inner rim has less than 0 thickness...).

If you look carefully you see two little "lightning bolt" shapes, one front, one rear, below the hub on each side. These illustrate an illegal geometry fold that renders this object un-editable. I can slice bits off it but cannot validate it or delete any faces, since doing so renders it a 'bad' object which Fusion does not allow. So even worse than the bus body - at least I could convert that into a surface and delete faces. My only option in this case was to create an entirely new object. Fortunately the geometry was simple enough that this was not difficult, even for me.

I also attach the STL of the original wheel, in case anybody wants to see for themselves, and suggest any way it might be made editable. This file has been 'fixed' by Netfabb, but remains corrupted. Situations like this I presume must be fairly common, so any solution would be very interesting.