Announcements
Autodesk Community will be read-only between April 26 and April 27 as we complete essential maintenance. We will remove this banner once completed. Thanks for your understanding

Bug Report: Surface Patch not functioning properly on 3D splines

OceanHydroAU
Collaborator Collaborator
726 Views
10 Replies
Message 1 of 11

Bug Report: Surface Patch not functioning properly on 3D splines

OceanHydroAU
Collaborator
Collaborator

It's missing some points out, so the patch doesn't match the sketch properly:-

 

Screen Shot 2020-11-28 at 9.19.53 am.png

 

This spline is slightly curved around an imaginary cylinder - if I project the spline to a flat plane, the patch works happily:

 

Screen Shot 2020-11-28 at 9.23.51 am.png

 

but that's no good for me of course - I need that wrap, and this bug should be squashed anyhow.

 

Here is the sketch in question:

https://a360.co/2Va2dfh

 

Here is the screencast:  https://autode.sk/37cka2o

 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0 Likes
727 Views
10 Replies
Replies (10)
Message 2 of 11

jeff_strater
Community Manager
Community Manager

I can reproduce this, and created bug FUS-75503 for this issue.  Not quite sure what the issue is, since an Extrude of that same curve does follow the curve fine.

Screen Shot 2020-11-27 at 5.07.09 PM.png


Jeff Strater
Engineering Director
0 Likes
Message 3 of 11

OceanHydroAU
Collaborator
Collaborator

I found something else which might be the same problem, but which manifests itself differently.

 

These 2 near-identical profiles here: Sketch is here: https://a360.co/2KQJZ0z - the one on the left refuses to "patch" at all.

 

I made a screen-cast of it before I realized the problem was one of the profiles (I thought it was a loft problem at first): https://autode.sk/39CdjSX

 

I've successfully patched several hundred of almost exactly the same thing today - it's just a few that give problems, and mostly (only?) ones with greater amounts of difference in the Z of their points.

0 Likes
Message 4 of 11

OceanHydroAU
Collaborator
Collaborator

This might be the same bug:

 

https://forums.autodesk.com/t5/fusion-360-support/bug-report-fusion-360-3d-surface-patch-getting-clo...

 

Or if not, would be convenient to fix at the same time.

0 Likes
Message 5 of 11

TrippyLighting
Consultant
Consultant

I would not call this pe-se a bug. Fusion 360 creates a trimmed NURBS surface in cases like this. I verified this in another application by looking a the CV distribution.

For curves that lie flat on a surface that is straight forward and only the simple surface with 4 control points and a trimming loop are needed

However, the airfoil curve does not lie on a plane. It is double curved, so to speak. As such the simple surface that needs to be trimmed is not simple whatsoever. 

The airfoil curve is densely and unevenly populated with too many control points (both are deadly when working with NURBS surfaces) , which results in bad curvature. 

 

Now Fusion 360 will have a hard time trying to match that mess with a NURBS surface.


EESignature

0 Likes
Message 6 of 11

OceanHydroAU
Collaborator
Collaborator

Hi - "too many control points" is not the meaning I expect you're trying to convey.

 

As you may know, badly spaced control points can lead to curious mathematical loops - which is why it is important to have *more* points in areas of tight curvature, but not to suddenly have too few thereafter (which would trip the loop conditions).

0 Likes
Message 7 of 11

TrippyLighting
Consultant
Consultant

Check the curvature of the splines with the curvature comb tool in the inspection menu. That provides you with all the info you need. The curvature is garbage!

Then perhaps check out the Autodesk Alias Theory builders to learn the basics how technical surfacing works.

 

Then, at some point in time the practical aspects of technical surface generation may catch up with your mathematical knowledge.  


EESignature

0 Likes
Message 8 of 11

OceanHydroAU
Collaborator
Collaborator

The curvature appears to be irrelevant.  This bug still crops up on "perfect" splines in about 5% of cases.  It might be related to the angle that the starting and ending points arrive together at perhaps?  It's so random it's hard to work out why 95% work and why those 5% fail.

 

Worse, in about 10% of those 5% - it crashes Fusion 360 dead (locks up forever)

 

Screen Shot 2020-12-08 at 10.29.09 am.png

 

Screen Shot 2020-12-08 at 10.29.25 am.png

Screen Shot 2020-12-08 at 10.26.04 am.png

 

Try surface-lofting these 3 examples to observe the problems - one of the inner splines on the top-left prop blade is broken, and the tip of the bottom-right is too.  https://a360.co/2VOwKiW

 

Fusion 360 crashes for me after I've done the top and middle blade, then try to do the last one.

0 Likes
Message 9 of 11

TrippyLighting
Consultant
Consultant

This is a different problem than initially presented.

I believe in this case the difficulty for FUiosn 360 is to determine the proper location of the twist control handles.

It's a closed spline so might need some user input to work, if it works at all.

 

In respect to your previous comments, the software you are using is used for technical surfacing in the area of product and industrial design. I had linked to the Autodesk Alias Theory builders to provide you with a background on how that usually works. That is the background for my earlier statements of too many spline control points. Less is more!

 

Perhaps it's time for your practical knowledge of technical surfacing to catch up with your purely theoretical musings.


EESignature

1 Like
Message 10 of 11

OceanHydroAU
Collaborator
Collaborator

I can't use less points, because Autodesk will not tell us the exact algorithm they're using for fit-point splines and their end conditions, so I'm forced to convert from my internal 17-point control-point format into sufficient points to faithfully reproduce the right shapes.

 

As you might know, control-point splines don't wrap around cylinders neatly (plus the projection-only implementation makes them frustrating for users to try and work with).

 

In this broken case, I moved the tail point around slightly, and the problem went away. I gave up trying to work out exactly which move steps solved it, because it took too long (Fusion360 kept locking up).

 

Can you give me a clue about what might be missing in my practical knowledge of surfacing [is that even the right term - this is a "bent" 2D spline]?  I did read the entirety of the Alias Theory Builders reference you suggested, but keep in mind that I'm doing engineering: the majority of that reference is for artists.  Pretty joins and Class-A surfaces etc don't interest me: perfect behaviour in a flow does.

0 Likes
Message 11 of 11

OceanHydroAU
Collaborator
Collaborator

It looks like they fixed this in todays update? - I can no longer reproduce these bugs...

0 Likes