Lofting of fuselage crashes

cfsengineering commented 5 years ago

It's really strange. On my computer, tiglviewer can open and display the fuselage contains in the file simpleFuselage-ok.xml but not the simpleFuselage-bug.xml. The only difference between the two files is that in simpleFuselage-ok, I change the translation of the last sectionElement from 6 to 6.1. Tigl library seems not to have any problem to open both files.

rainman110 commented 5 years ago

? Can you please add some detail?

cfsengineering commented 5 years ago

done :)

rainman110 commented 5 years ago

I can confirm this bug. I'll have a look into it tomorrow.

joergbrech commented 5 years ago

Good find! I just ran your example in debug mode, it crashes here:

https://github.com/DLR-SC/tigl/blob/bf998a86a8e44744504becd406f980e328742565/src/geometry/CTiglPatchShell.cpp#L169

There still seems to be an issue with the loft generation sometimes, we will take a look at it.

rainman110 commented 5 years ago

@joergbrech Since this is your code, could you have a look into this issue?

joergbrech commented 5 years ago

Ok, so this is the situation:

In CTiglMakeLoft, the generated fuselage loft is cut at the profile parameters, so that the loft consists of patches exactly as it would if it were generated from a curve network. This makes sense because we want to have consistent shapes independent of the algorithm used to create them.
The function CutShellAtUVParamters calls CTiglBSplineAlgorithms::trimSurface, which is just a wrapper for the corresponding OpenCascade function + an OpenCascade conversion. https://github.com/DLR-SC/tigl/blob/bf998a86a8e44744504becd406f980e328742565/src/geometry/CTiglBSplineAlgorithms.cpp#L882 It cuts the input face at the desired parameters. So far, the parameters passed to the function seem to be correct, but one of the trimmed surfaces (the fourth segment in this specific example) calculated by this function is way too large.
As a consequence, faces of the loft overlap and we cannot make it a solid.

This is more of a note to self, I will check out what goes wrong exactly when I find the time later this week.

joergbrech commented 5 years ago

This seems to be a bug in OpenCascade to me, at least I can reproduce it in pythonocc (The input surface is still generated by TiGL though...)

TiGL

cutShellAtUVParametersBug

pythonocc

cutShellAtUVParametersBug_OCE

brep file and python script

fuselabe_bug.zip

rainman110 commented 4 years ago

I condensed this issue and posted it at the OCCT mantis: https://tracker.dev.opencascade.org/view.php?id=31402

It even exists with OpenCASCADE 7.4.0.

rainman110 commented 4 years ago

The issue should now be fixed with my workaround

rainman110 commented 3 years ago

Opencascade has finally fixed this bug:

https://tracker.dev.opencascade.org/view.php?id=31402

As we have a workaround in place, there is nothing we have to do right now. I am writing this just for documentation purposes.

DLR-SC / tigl