Silk: a NURBS workbench

[microelly2]

I'm not realy a friend of the surface object. It creates good approximations for the input data but the outcome is not stylish:
the poles grid in this case does not have a relation to the border faces and the face is a patch of the bspline surface.

bp_792.png (105.86 KiB) Viewed 510 times

the waviness of the surface can be smoothed in a second step but it is the result of a brutal interpolation.
I think the task is to find a method which creates for a triangle gab definde by 3 nurbs surfaces a filler consisting of 3 unpatched nurbs.
in my example I used 3 ruled surfaces (along bezier curves) as borders (brown/pink ) and created inside the hole 3 nurbs red, blue green

bp_790.png (91.58 KiB) Viewed 510 times

bp_791.png (466.99 KiB) Viewed 510 times

because there is only a small number of poles and because of the simple bezier curves the continuity is not anywhere 2 but the isocurve structure is naturally. The grid of the inner faces is scaled by factor 2 to show the balanced correlation between the knots of all faces.
for statistics:
the border curves have each 7 poles (two connected bezier curves)
the inner area has 36 poles - each segment is a 4 x 4 bspline surface of degree 3, so all edges are simple bezier curves.
here is my example file
https://www.dropbox.com/s/i6n5sas1ky5c6 ... fcstd?dl=0

[emills2]

I think the task is to find a method which creates for a triangle gab definde by 3 nurbs surfaces a filler consisting of 3 unpatched nurbs.
in my example I used 3 ruled surfaces (along bezier curves) as borders (brown/pink ) and created inside the hole 3 nurbs red, blue green

that's the same strategy i talked about a year ago:
https://forum.freecadweb.org/viewtopic. ... 20#p187547

and that is precisely what i am comparing the filling surface to in my posts above

Silk_model_74_15.png (195.05 KiB) Viewed 486 times

july 29 2017

Silk has a system that handles N sided patches and gets G2 at the OUTER edges of each 'NStar' (but the center has a slight pinch, and it seems to get worse the more branches the star has)

Silk_model_113_15.png (337.34 KiB) Viewed 486 times

september 17 2017

they are lightweight and fast, but in my estimation, G1 filling from the surface workbench will often be better overall.

i used N 36 control point patches in N 6X6 cubic NURBS surfaces, and that is as far as i want to push it personally. The inner point placement can be refined, but i spent too much energy already in that. the long term goal for me was always to get the final surface using the OCC filling method. Trimmed fitted surfaces are generally considered the state of the art for odd hole filling, and if we keep scratching, we will uncover how to make the built in G2 functionality work.

Silk_model_74_11.png (199.71 KiB) Viewed 486 times

july 29 2017

how did you generate your '3 ruled surface example'? can you run it on my 3 border surfaces in the model so we can compare side by side all 3 versions?

[PrzemoF]

Nice progress.

definitely but i certainly don't claim any credit. all i did was sit back and wait for someone else to implement OCC filling.

i think the kayak models can go a lot further now!

I hope it will happen soon!

[microelly2]

that's the same strategy i talked about a year ago:
https://forum.freecadweb.org/viewtopic. ... 20#p187547

and that is precisely what i am comparing the filling surface to in my posts above
Silk_model_74_15.png
july 29 2017

Silk_model_74_11.png
july 29 2017

Yes I see it, nice to have already an implementation.
I aks me now how to make it easy to use.

how did you generate your '3 ruled surface example'? can you run it on my 3 border surfaces in the model so we can compare side by side all 3 versions?

Here is my script

ecke_aus_drei_nurbs.py

(5.43 KiB) Downloaded 27 times

phpBB [video]

[emills2]

ping

thanks for the script. it's a lot to digest, so could you load the final surfaces for your 3 piece corner? i'd like to look at them in a bit more detail before trying to understand how your script works.

[microelly2]

https://www.dropbox.com/s/dr0i21dwvmbr6 ... fcstd?dl=0

[Kunda1]

Silk WB still needs a logo for the add-on Manager
https://github.com/edwardvmills/Silk/issues/10