Generative Design - Previewer fails when adding Starting Shape

[joedvin]

Hi all,

The problem seems trivial, but I am a complete beginner, and cannot move forward.

I have two preservative geometries which I want to connect with a starting shape in order to give a more desirable result. Without the starting shape, the previewer can run without a problem. (See no-starting-shape.png)

When adding the starting shape, which is a hollow body build from a 3D-scan of a leg, the previewer fails. (See w-starting-shape.png) The starting shape is cut with the the perservative geometries, and meets the surfaces perfectly.

Let me know if more info is needed, and thanks in advance. 

Best,

Edvin

Autodesk Fusion 360 (2.0.7421)

Windows 10 (1909 18363.628)

[I_Forge_KC]

I don't know of any reason why you would get an error with that setup. That said, try a couple changes and see if they give a different behavior as a diagnostic exercise.

1. Make sure the three bodies can actually be combined. Sometimes if bodies have odd interfaces the combine function can fail. Enter the edit model space and just see if you can combine them. Don't keep them combined, though. Rerun the previewer and see what happens.

2. Maybe there is an issue with the interface because the voxels have a shared interface there and bad things happen at the solver level. Try shifting the resolution slider in either direction to change the voxel interface locations throughout the part. Rerun the previewer and see what happens.

3. Remove the cuts to the starting geometry. Your loading conditions ensure that the area of interest is between the two preserves. Any geometry outside of that imaginary tube will eventually be removed (if given enough iterations and no obstacles between them). Even if you had a whole human body as the scan, you wouldn't have an issue getting to where you're interested in - though it may take a while. Try letting the starting shape pass into/through the preserves. Rerun the previewer and see what happens.

If you'd like to share the file, I'm more than happy to troubleshoot for you.

[Ben-Weiss]

In addition to @I_Forge_KC's comments above, it looks to me like your starting shape is kind of odd...like it used to be an STL and got converted by some tool or other into a solid body so you could use it in generative. Does Previewer still fail if you give it a simple cylinder or rectangle instead of the starting shape you chose? If so there may be something wonky with the solid body you're using as a starting shape. If using a different starting shape geometry does fix the problem, and you're willing to share the setup giving you issues, I'll take a look and see if I can see what's going wrong in the solver...

[Ben-Weiss]

@I_Forge_KC  I could be wrong here, but I don't think the resolution slider affects the Preview behavior. We crank the preview resolution as coarse as possible to try to make it almost-interactive on your laptop.

[joedvin]

Hello,

First of all, thank you both for taking your time to look at my problem. It is much appreciated.

As answers to your remarks:

@I_Forge_KC

1. Using the “Combine” -> “Join” tool on the three bodies successfully combines the bodies.
2. I have been trying to get an understanding on voxel size impact previously but am still a bit uncertain in what I can do you adjust this when modeling. However, if you are referring to slider under “Study Settings” -> “Synthesis Resolution”, I can confirm adjusting it to either of the extremes (max course and max fine) still does not let me run the previewer.
3. Thank you for giving me a clarification on this, I have been wondering about this, that is good information. The cut has been a result in trying to simplify to model in order to help the previewer run. Unfortunately, letting the starting shape geometry completely run through the preservative geometry does not help the previewer. See “uncut-starting-shape.png” for particular setup.

@Ben-Weiss

I have been suspecting the actual geometry of the starting shape as well. This has been created mostly through trial and error, and the process in short have been:

1. A 3D-scan downloaded as STL and opened in Fusion, which creates a mesh.
2. Convert to BREP and patch any openings.
3. Use fluid volume to create a body.
4. Scale a copy of the new body so I can use the cut tool resulting in making it a hollow body.
5. Due to the unsuccessfulness of the previewer, the model was simplified through the “Reduce” tool under the Mesh-tab, then repeating above steps.

However, I can confirm, using the starting shape which was created by using “Loft” between two ellipses, as seen in “simple-starting-shape.png”, I am still out of luck in running the previewer.

I’d be more than happy to share the complete set up and files, if that would be of interest anyway, let me know. In that case, how do I do this preferably?

Best,

Edvin

[I_Forge_KC]

File > Export

Attach the F3D file here and we'll take a closer look!

[joedvin]

Hello again,

I am attaching a public link to the file. Just to let you know, as you probably can see, some of the preparation of parts were done inside other files and then imported to this. Let me know if this info is needed as well. 

Once again, thank you a lot for considering troubleshooting with me, it is much appreciated.

Link to problematic file:

https://a360.co/2uCMonJ

Best,

Edvin

[joedvin]

Hi,

I had not seen you response when posting the link. Here is an export as well.

Best,

Edvin

[Ben-Weiss]

Hi Edvin,

I played with your study some and I believe the issue here is related to thin sections in your preserve bodies. When I try to generate, the generative solver complains that the shape it was provided isn't connected (that is, there are several different pieces of geometry, some floating in space; they refunded my cloud credits, though). The generative solver is looking at a voxelized version of the geometry, and sometimes thin pieces of Fusion solid bodies are too small to be represented by a single voxel-wide wall. This is probably why the previewer was failing. You have a few options:

• Increase the synthesis resolution. You've already maxed that out on the study you posted. This won't help the previewer but will reduce the chances of a failure when you generate. The higher your synthesis resolution, though, the longer the solve takes.
• Thicken your preserve bodies, particularly the top one.
• Extend your starting shape so it covers the preserve bodies and creates a thicker initial structure. This might still fail later in the optimization process, if the solver tries to remove material until only the preserve is remaining, but the solve at least starts.

I'm running the last two options now, and a dozen iterations in they look alright; I'll report back when they complete.

[joedvin]

Hi,

Once again, I am very grateful for the time and effort you both spent on this problem.

As suggested, I can confirm increasing the thickness of the upper preservative body (socket) and adding copies of the preservative geometries to the starting shape does let the previewer to run, but only with the simplified cylinder. Switching back to the original starting shape, the scanned leg, while following the same steps, the previewer once again fails. (see failing-starting-shape.png) I assume this indicates the "leg"-geometry also is part of the problem?

I am also attaching a newer version of the file, with the failing set up after modifications mentioned above, if it would be of interest.

Best,

Edvin

[Ben-Weiss · Accepted solution]

Hi Edvin,

Even though Preview fails for the file you posted, I tried it and it seems to be generating fine. The starting shape you're using has some thin sections, and I think the coarse resolution supported by the previewer is unable to represent the shape adequately.

Here are the results from the studies I ran yesterday:

One quick comment - you don't need to both make the starting shape cover the preserve bodies and thicken the preserves themselves. Either method ensures that there's enough geometry there at the start of the optimization process for it to not immediately break in pieces. The thicker starting shape with the original preserve definitely did have trouble representing the preserve body in the solver once the starting shape started to erode. Here's a snapshot from an early iteration of this case; the preserve body is only partially represented by the solver.

[joedvin]

Hi Ben,

I did try to run the study as well even though the previewer failed. While the results were not satisfactory with a stop at 2 iterations, the generation process did start.

I will mark your last response as a solution, which points out the thin geometries as a root to the problem, both for running the previewer and the results. It also confirms results can be generated even though the preview fails, due to the course resolution in the previewer.

I will continue my work with thicker volumes, and I believe with what I have learnt from this and previous post, I will be able to find  more satisfactory solutions. 

Once again, thanks for all the time and effort, much appreciated.

Edvin