FreeCAD Forum

balrobs wrote: ↑Sat Sep 04, 2021 2:28 pm BTW the interested object isn't a wall but a beam created with the BIM workbench, isn't it?

Yes(structural wall/beam).

paullee wrote: ↑Sat Sep 04, 2021 3:49 pm Not only par-004 flip in direction, I had once revised the Height of par-004, and it also change direction ! But can't reproduce that.

I did this test and only par-003 changed.

paullee wrote: ↑Sat Sep 04, 2021 3:49 pm Note both par-003 and par-004 have Axis-z = -1

Other beams also have Axis-z = -1 and do not change.

thomas-neemann wrote: ↑Wed Sep 08, 2021 8:32 am if you need a quick transitional solution, you could make a simple copy in part before the export and export it. it is enough to get the geometry.

Sorry, I do not understand.
Use Part WB / Create simple copy?
This is not the case, as Arch/Ifc modeling is lost.
The issue here is not the solution to this use case.
Imagine the situation:
A giant project, a structural project received from a designer through an IFC file, positively matched with the architecture by the project sector and will be sent, via IFC, to a subcontractor for construction, and at the time of export an extremely beam important for the structural stability of the building undergoes alteration and is constructed wrongly.
The question is:
Is this behavior a bug or an error using the Create Beam tool?

walpa wrote: ↑Wed Sep 08, 2021 12:58 pm

if the file changes without notice, this is a security problem. that must not be. i use a workflow where it doesn't happen because the models are stable. In this case, the instability is apparently not caused by the ifc export, but appears beforehand, e.g. by recalculation / update

In my judgement, this does not have anything to do with IFC export. Just force recomputation with App.ActiveDocument.Structure027.touch() and you will see the wall flip. Alternatively edit any value (position or the like) by entering the same value again. To me it looks like an outdated (out of sync) shape got saved with the file somehow without the dirty flag being set. This results in the wrong shape (which looks correct but is wrong with regard to the model) being displayed and as soon as a recompute is triggered (for example when exporting IFC) this is corrected.

If you think about it, the flipped position conforms to the model: With rotation angle 0, the wall extends in x direction, you expect it to extend in y direction. So if you rotate about (0,0,1) by 90° or about (0,0,-1) by -90°, you end up with the expected result. With 90° about (0,0,-1) or -90° about (0,0,1) you will end up with the flipped direction. To correct, fix the axis or angle of the rotation.

What seems strange to me, is the behaviour of the ArchStructure for negative coordinates. If I look at the nodes in data or by switching them on in view tab, I see the second node at (-4010,0,0), but the structure extends to (4010,0,0). Everything stays the same though when changing the node coordinates to the positive x-value.

Thinking again about the previous observation: If the structure would extend towards (-4010,0,0), it would end up in the right place. See screenshot below with rotation set to 0. Node coordinates are (0,0,0) and (-4010,0,0), but shape endpoints are at (0,0,0) and (4010,0,0). Thus the given rotation seems to perform correct for node coordinates, but just appears wrong for the shape of the structure.
This looks like a regression in ArchStructure in how the shape is created from the nodes for this particular case. I couldn't identify a suspicious change. Maybe someone more familiar with the code has an idea? @Yorik?

I think, when a beam is created, nodes are converted to the internal model with second node at (length, 0,0) and rotation, hence beam axis rotated to local x-axis. Then subsequently, only length and rotation are used for shape generation. This file looks as if at one point, shapes where generated differently.

Curiously, when node coordinates are changed directly, length and rotation are not updated, which I find strange. Is there a reason for this? Maybe recalculation to the internal model should also happen on file load to fix this issue?

hlg wrote: ↑Wed Sep 22, 2021 3:07 pm ....What seems strange to me, is the behaviour of the ArchStructure for negative coordinates....

I already had problems with that. But in reality there are no components with negative lengths, so I've only worked with positive lengths since then and everything has worked

Can help to file a bug report at Mantis BugTracker?

Thanks

Sure, done: https://tracker.freecadweb.org/view.php?id=4751