Delete random points in a geodestic dome

Hey guys,
I am really new to Grasshopper and build a geodestic dome with some triangle panels.
I extruded circles into the triangle points of the geodesic dome and removed it from each other with the help of ‘trim solid’. (like the fly’s eye dome from buckminster fuller)

Now I would like to remove the random points that do not match the panel endpoints (red marks in the next picture).

I think the points are created by trim solid between the dome and the cylinders. When the circles are extruded into a cylinder (green), a like-line curve (red line in the next picture) is created that leads to the additional point.

Is it possible to filter out the points? I have already tried a list, but unfortunately I couldn’t do it because the points are randomly distributed.

Maybe someone can help me.
Thanks! (124.6 KB)

Well … I can’t read your def (I don’t have the add-ons used) but given the opportunity a couple of things:

  1. Geo domes come in real-life as meshes. A mesh is way faster than a Brep (Point3f VS Point3d) and you can take advantage of the also fast Mesh connectivity data (in terms of trees). Mesh connectivity is used when you want, say, a truss out of a given dome mesh (either of type W or “flat”) and/or some sort of envelope etc etc (for instance an EFTE one if you use the inner W truss “layer” of struts that outline hexagons or pentagons in most of the cases).
  2. Given a mesh and a face to vertices [FV] connectivity Tree (or vertex to vertex [VV] connectivity) you can achieve what you want on a per triangle face basis OR on a per vertex basis … without using any ccx Methods (is way faster that way). For instance (case: polyline openings) using a VV Tree get a vertex, sort the neighbors and for each neighbor define a point (at D with regard the vertex-neighbor line) then sample all the vertices in a polyline, add the first in order to get a closed one, store the polyline in a Tree etc etc.
  3. A slightly different technique is used when you are after the opaque “panels”: you are working with faces and a FV connectivity. This is slower because you do Breps from closed Curve loops. Compared to the preview mode (see options in the attached) the opaque mode is about 100 times slower … thus the preview can serve a thing or two prior doing the final result.
  4. Purely for fun play with the attached (is 100% code thus may be useless for you - unless you have future plans to learn coding). Used a couple of domes of varying frequency (deformed in X, Y, Z or not) as test cases. Domes are done with my main dome generator C#. (187.4 KB)