Unroll surface won't match developpable surface

[hugo.thibaudeau.2]

Hi,

I’m working on a ruled surface pavillon that will be built using CNC milling. The geometry is like a truncated irregular cone. There is a custom waffle structure built on which there will be screwed pannels. The waffle structure is tested and works. (figure 1) Problem is that when I unroll the surface built from a developpable loft and I cut it, it is warped and not at all the right surface. (figure 2)

Here’s what I tried to solve the problem:

I’ve tried solution in Rhino 5 and Rhino 6

-In Rhino, I analysed the gaussian curvature, which is 0 or 7 e-10 depending on the command (it should be a ruled or developpable surface)
-In Rhino, I built my surface from Sweep2, rebuilt curves and normal loft.
-In Rhino, I tested the commands Smash & Flatten, they give the same result
-In Rhino, I tested the command Squish, it gave a different result but not the right one.
-In Rhino, I built my surface from the ground up using higher tolerance (0,001mm)
-In Rhino, I unrolled the entire surface and not the single pannel.

-In Grasshopper, I built my surface from the ruled surface battery, with the same unrolled result
-In Grasshopper, I used unrolled Battery from: Fabtools, Peacock, Lunchbox,
-In Grasshopper, I used custom solutions (from old discussions) by @TomTom @Mahdiyar, @Tudor_Cosmatu

Here is the picture of the waffle structure assembled (figure 1)

Here is the picture of the unrolled pannel milled (figure 2)

Here is the .3dm with the unroll, flatten and squish profile.There is also the manual profile that is our goal (we measured hands-on the dimensions needed).
2019_04_29_Pavillon_Forum.3dm (5.2 MB)

Thanks in advance for your time

[pascal (Pascal Golay)]

Hello -I am confused, I guess - why is it that the none of the edges on the ‘manual’ version have the same length as the edges of the 3d polysurface? UnrollSrf comes pretty close.

-Pascal

[JoergH]

Hallo,
RH6 (unrollsrf) reports me a deviation of more than 15% !?

Can you post a RH5 File?

Best regards

[davidcockey]

The file is no longer available.

-In Rhino, I analysed the gaussian curvature, which is 0 or 7 e-10 depending on the command (it should be a ruled or developpable surface)

Guassian cuvature can be misleading because it has dimensions of 1/length^2. It’s magnitude depends on the units used and size of the object, not only on the shape. A better method to check if a surface is close to developable is to use the Curvature command and examine the surface on multiple locations. Once principal curvature should be zero or very close to zero everywhere. The associated curvature circle should be a straight or very close to straight line.

-In Rhino, I analysed the gaussian curvature, which is 0 or 7 e-10 depending on the command (it should be a ruled or developpable surface)
-In Rhino, I built my surface from Sweep2, rebuilt curves and normal loft.

All developable surfaces are ruled surfaces but some ruled surfaces are not developable.

Sweep2 and Loft usually do not produce developable surfaces.

[hugo.thibaudeau.2]

Thanks you for your help,

Here is a rhino 5 file
2019_04_29_Pavillon_Forum_Rhino5.3dm (5.3 MB)

[hugo.thibaudeau.2]

Thanks for your input, curvature is less than 0,00015 at every place I checked

Here is the file 2019_04_29_Pavillon_Forum.3dm (5.2 MB)

[hugo.thibaudeau.2]

First I’d like to thank you for your fast answer,

Actually the 'manual version was measured from the assembled structure (which explain the possible difference). The main problem is more about the distortion in the angles which cause the pannels to be skewed from the original as you can see in figure 2.

Pretty close (± 40mm) unfortunately just don’t make it with our digital fabrication process, as there will be many of the sections illustrated assembled next to each other.

[davidcockey]

I agree it looks like it should unroll.

Explode the surface, rebuild each part with less points, join, and unroll the simplified surfaces. They should unroll within tolerance. I rebuilt the center part with 10 points in the v direction and the resulting surface was within 0.04 mm of the original. Unless you are doing high precision work in metal that should be sufficiently close.

[pascal (Pascal Golay)]

Yeah, I used FitSrf and got unrollable surfaces.

-Pascal

[JoergH]

Hi,
we produce similar constuctions, but from aluminium with tolerances of +/- 0.2mm. Your surface is not great, but might be good enough for your tolerances. It still might be better to design proper developable surfaces. If your process works on this panel, does not mean that this is true for all panels.
But if you look at this screenshot there seems to be something wrong.
I think you just have too many problems in physical building as well as designing that construction.
If your “manual” template deviates this much from your 3d model, and you can see that even your lamellas do not conform to your 3d model you should check every step on building your 3d model as well as your physical mockups. If everything is designed and build correctly you should not see deviations greater than 2 or 3 mm in real world. And leave some mm space between panels for tolerance compensation and dilatation. pavillion1|690x381

[BrianM]

Have you allowed clearance in the slots to allow for the interlocking panels at the various angles. The pic looks as thought the panels are contorted to fit the cnc straight-cut slots, so maybe the framing is out and not the unrolled sheet. Did the cnc cut framing fall into place or forced?