Hi Alex, so it is the negative z axis thing.

The difference between these files is that "LEG01 EXPORT SLD.DXF" has the normal for each arc's plane pointing in the +z world axis direction.

The file LEG01.dxf has some plane normals in the +z world axis direction and some in the -z world axis direction.

It's a fairly common bug for DXF reading code to not do the AutoCAD "arbitrary axis algorithm" properly on a normal that is pointing in the negative z world axis direction. It's likely that the pipe bending software has this bug in it.

It's possible to flip any arc with a -z plane normal to go in +z instead but this will have a side effect of reversing the direction of the arc after it's flipped.

Some CAM workflows need the arc direction to stay undisturbed so that's why MOI doesn't flip them by default.

But you can go to Options > General > "Edit .ini file" button and inside moi.ini in the "[DXF Export]" section set:
FlipNegativeZAxisCirclesAndArcs=y

That should then flip any arcs that have a plane normal in the -z axis direction to the +z axis direction instead and avoid the bug in the bend software.


Does it work ok once that has been set?


For Rhino there's an export option "Flip arc normals to +Z" that you can turn on.

- Michael

Hi Michael,
Just tried that change but still no go its like it cant read the center line for profile. Strange as if I import the same DXF profile center line into solidworks and just lock tangents and re export it all imports correctly. It would be great to have the complete workflow coming from MOI as it wins hands down in modelling time. At this stage I am doing all the prep in MOI then export the individual master profiles into solidworks for export from there which works.

If not an easy fix all good Ill keep doing the same workflow just made me wonder if tangency did have a finite point in the transition from arc to arc.

Hi Alex, can you please post the new .dxf file you generated from moI with the FlipNegativeZAxisCirclesAndArcs=y set so I can double check it?

> just made me wonder if tangency did have a finite point in the transition from arc to arc.

It doesn't seem like that's the problem, the arcs seem to be tangent to a very tight tolerance. It's more likely to be some difference in ordering or directions.

- Michael

Thanks for the exploration here is the test file - I have marked the start curves with the MOI script when I draw i start at one end and manually trace it out from the same direction.

Hi Alex, I'm kind of confused about the different file formats being used.

I thought the recent thing I was checking was the difference between a DXF file exported out from SolidWorks vs a DXF file exported out from MOI.

There was a difference between them, the SolidWorks one had all arc plane normals pointing upwards in +z. After setting that moi.ini flag then the DXF file exported from MOI should also have all arc plane normals in +z axis direction.

But then your last message has a .3dm file not a .dxf file?

- Michael

Hi Michael,

So many programs so many workflows sorry it gets confusing - here is how the parts get into Pipebend.

With the pipebend software it imports 3d STP files - so I start with the governing center line I create the path curve and shell that to import as a STP. ( No Splines - All arcs and polylines only)
To move between MOI and Solidworks I export the DXF file only which I then lock the arc nodes in SW's and then shell followed by exporting the 3d STEP to import into Pipebend.
Somehow Pipe must read the outside surface radius and then construct a new Centerline from which to CNC control the motions. No dxfs are imported directly into pipebend not sure why they chose this workflow but the PIPEBEND software is made in China I have tried to contact them but no avail.

Here are their only 2 videos on Youtube - not alot of traffic but from what i gather all the tube bending machine makers use their software.

https://www.youtube.com/@yishansoft1510/videos

Many thanks for your continued assistance - appreciated.

Hi Alex, thanks. It seems like it will be difficult to figure out what is going on.

I don't understand why they would want to take a 3D model and attempt to reverse engineer the centerline from that instead of just taking in the centerline directly.

I think that you're lucky that there is any method that works reliably.

- Michael

Yep I agree I would have thought a DXF whether 2d or 3d imported would be the way to go but that is how these machines work - all good MOI exported STP files work 80% of the time I its just when there are multiple arcs back to back that it tends to fail and that is where the SW lock tangent gets them over the line.

Thanks for looking into it for me.

Does that error message come upon initial import of the STP file or is it generated by the post processor engine generating GCODE in pipebender?

The files to compare are the stp files not dxf.

Hi Burrman

The error comes up with the STP import before getting to Gcode stage so its definitely something in the tangent to tangent structure as the only change I make to the same underlying curve pathway in SW is to lock the tangents - which must mean there is a further degree of tangency that perhaps the PARASOLID engine enables for the PIPEBEND software to read the 3d part.

Thats just my current thinking.

Ok. I was trying to apply my CAM softwares thinking when it is parsing arc and such. The post processor has several setting for various machines and software outputs.

Break arcs into quadrants. Combine more than 180 degrees. Etc… etc… something about signed arcs and the processing into xyz coords that the machines controller reads. There are tons of controllers. Not all created equal.

Was taking a guess at what “lock tangents” is actually doing or setting in the step file with regard to arcs. Was trying to figure out if yyspipebender is actually trying to post process the geometry and needed one of these specific settings

Maybe if we can see the 2 stp files we can figure that out…

Thanks Burrman,
Here is my workflow - file created in MOI
Export dxf into SW
Lock Tangents in SW
Export Step
Import STP into PIPEBEND

Thanks for follow up Alex. I will certainly look over you outpit and locked tan files a bit later.

With my last post and just “looking at the screenshot” in your last post, it appears that last arc is greater than 180? (Havent verified this yet)

If it is, try adding a couple trim points to make it 2 or 3 smaller arcs and try it without the locked tan process to see if yyspipebender will accept that….

I’ll respond again after i get a look at the stp file output

Hi Alex,

I took a look at your file "3 TEST ARC SW.STEP"

Pipebend likes this file, correct?

Would you also post a STEP that Pipebend rejects?

- Peer

""""""Would you also post a STEP that Pipebend rejects?""""""

Take the MoI 3dm file and save it as a STP...

Thanks here are files

Hi Alex,

It looks like the problem I mentioned before: the Sweep command in MoI (which the pipe script is based on) has limited accuracy. I examined the surfaces in both STEP files in MoI using Construct > Curve > Iso and then checking the properties of the generated curves. The isocurves of the solid exported from SolidWorks are very close to being circles and so are recognized as circles by MoI which shows a diameter value for them in the details area in the upper right corner. Doing the same thing on the "TEST ARCS MOI EXP.stp" shows that the curves aren't as close to being true circles, and so MoI just shows bounding box dimensions rather than a diameter value:



Have you tried exporting a STEP from SolidWorks without doing "lock tangents"? Is Pipebend still happy if "lock tangents" is skipped?

Another way to see what's going on is by opening the STEP files in a text editor and doing a text search. The SolidWorks file is made up of TOROIDAL_SURFACE entities. The MoI file just contains general freeform surfaces.

- Peer

Thanks Peer,

Your CAD forensics are impressive - all makes sense I was wondering were PIPEBEND was collecting the bend info.
I did think there was an underlying structure difference between Parasolid and Open Nurbs and the way the surface was created.

I tried a part from SW without Tangent lock and same curve from Moi and that imported AOK into pipebend so yes that curve pathway is tangent - but have a look at the MOI Fillets file that was a series of polylines with fillets applied and then shelled. That solid imports into PIPEBEND aok so strange it structures the surfaces in some instances in same fashion and then in other style that PIPEBEND cant read.

Really appreciate the insights.

No help from this member of the peanut gallery on the problem at hand, but just to say that I'm looking forward to pics once your new sculpture begins to come to life Alex!

--Larry

Hi Alex,

Re:
> I did think there was an underlying structure difference between
> Parasolid and Open Nurbs and the way the surface was created.

I doubt this has anything to do with Parasolid vs OpenNURBS. Both are capable of analytic representation of a torus.

Re:
> That solid imports into PIPEBEND aok so strange it structures the
> surfaces in some instances in same fashion and then in other style that
> PIPEBEND cant read.

I'm guessing that Pipebend is smart enough to look at general surfaces and check whether they are sufficiently close to being a true torus. The key word here is "sufficiently" because if MoI and Pipebend use different tolerances for "close enough" then sometimes they are going to agree, and sometimes not.

One thing you could do is look through the Pipebend settings to see if there's an import tolerance setting somewhere. If there is, you could try setting it to a larger number like 0.03 mm.

- Peer