search

realthunder / FreeCAD_assembly3

Experimental attempt for the next generation assembly workbench for FreeCAD
GNU General Public License v3.0
885 stars 74 forks source link

Surface creation #677

Open parnasist3 opened 3 years ago

parnasist3 commented 3 years ago

OS: openSUSE Leap 15.2 (KDE//usr/share/xsessions/default) Word size of OS: 64-bit Word size of FreeCAD: 64-bit Version: 2021.615.24301 +3747 (Git) AppImage Build type: Release Branch: LinkDaily Hash: ecbfba008c481db4cb17bf8bd514b2d12690a8e5 Python version: 3.9.4 Qt version: 5.12.9 Coin version: 4.0.1 OCC version: 7.5.2

Regarding surface creation:

Would it be possible for you to somehow create a way to make a multisection surface with start- plus end-tangency via a couple of common curves/lines? See attached picture. Multisection_creation It shall start with "Direction vector A2" and be tangent to "Plane A". Guide curve shall be "Blend curve A1-C1". It shall end with "Direction vector C2" and be tangent to "Plane C".

Part_Arbitrary_surface.FCStd.txt (remove trailing .txt)

parnasist3 commented 3 years ago

Any chance you can have a look at the illusive "Multi-section surface"?

Rabbit-sk commented 3 years ago

I don't know if that helps you, but here are som links about Curves workbench and Silk workbench: https://wiki.freecadweb.org/Curves_Workbench/en https://wiki.freecadweb.org/Curves_GordonSurface https://forum.freecadweb.org/viewtopic.php?f=3&t=22675

https://github.com/edwardvmills/Silk https://github.com/edwardvmills/Silk/wiki https://forum.freecadweb.org/viewtopic.php?f=3&t=60073 http://edwardvmills.github.io/NURBSlib_EVM/ https://github.com/edwardvmills/NURBSlib_EVM/blob/gh-pages/Tutorial%200.01%20ControlPoly4%20and%20CubicCurve4.md https://github.com/edwardvmills/NURBSlib_EVM/blob/gh-pages/Tutorial%200.02%20ControlGrid44%20and%20CubicSurface44%20-%20page%2001.md https://github.com/edwardvmills/NURBSlib_EVM/blob/gh-pages/Tutorial%200.03%20Point_onCurve%20ControlPoly4_segment%20ControlPoly6%20and%20CubicCurve6%20-%20page%2001.md https://forum.freecadweb.org/viewtopic.php?f=22&t=23243

They are not easy no use at first time, but they may be useful for you.

parnasist3 commented 3 years ago

Hi Rabbit-sk, the Silk workbench do get a step on the way to general surface creation. However, it does not handle the arbitrary ruling multisection surface principle which consists of just that, a multisection surface. Edwards version handles dual-/two-section ruling surface, but not multisection.

Curves workbench have recently got an upgrade to handle arbitrary ruling surfaces, but it relays on approximated curves as a base, which is a big no-no CAD-wise.

The "built in" Surface workbench also works with arbitrary ruling surfaces, but it requires the user to choose the ruling surfaces edge as base for resulting surface creation, which is a big no-no topo-naming wise.

So, as it stands right now it is possible to make semi complex surfaces, but it will not survive in a big and complex model.

microelly2 may have pulled it off with his Nurbs workbench, but I don't understand how he did it. Maybe someone do understand how and will make an instruction in the future?

The "only" thing missing is to make true multisection surface working (which includes start and end tangency) plus no surface edge selection, it shall be predefined curve selection and the curve shall not be approximated. However, I get the feeling it takes someone with higher skills to pull this off, hence the request.

parnasist3 commented 3 years ago

If a bold and brave code reader could dive deep in OCC's bruchy code swamp, I'm sure there are hints on how to make an ruled/governed surface according to the first post in this thread.

As an example, the "built in" surface creation tool forces the user to select a previously created surface's edge in order for the algorithm to figure out what surface to use as a ruling/governing/sovereign surface.

This implies the programmer/coder use that surface edge to look up what is its parent. I suggest to do the other way around, user first choose the curve(i.e. not surface edge) then a ruling/governing/sovereign surface. The algorithm should then check if the curve is a parent to the ruling/governing/sovereign surface.

The ruling/governing/sovereign surface shall ofcourse be made with the multisection tool and use the curve as a guide.

I'm pretty sure that a fill surface must somehow utilize direction vectors at the corner points and "poles"(which basically are end points on a local vector) along the curves. These "poles" are spread out along each curve in and equidistant way per curve. So the OCC code do approximate the original curve by dividing it into multiple short curves and placing one "pole" at each tiny curve segment start/end. It seams that, as of today, the OCC algorithm divide the curve in a fixed set of segments/poles (think it is 30?). Now that will pose a problem when the curve is "longer", read: road length.

So the general solution should also take that into account and implement some sort of "pole count" setting along the curve and show it in the 3D-view port.