Relax puzzles for Jon

Replies to This Discussion

Permalink Reply by Jon Mirtschin on December 26, 2011 at 2:06pm

Hi Peter,

Seasons greetings to you.  The model looks great.  Here's my initial reply to your questions/comments:

2.  Yes, you are correct.  If all the mesh vertex are naked (located on the perimeter of the topology) then with no degrees of freedom nothing will compute.  I should add a check to identify and report this situation back to the user.  

3.  As few as possible quads isn't something I've dabbled in, I've put my hand a little to meshing, resizing etc but typically very specific circumstances when I've need to apply in particular circumstances.  I just figure there must be mathematicians and programmers in various research/study positions that must be more knowledgeable than me in this area.  If there is an existing algorithm/publication that I can enable into GH I'm happy to advise about enabling it for Grasshopper.

4.  http://geometrygym.blogspot.com/search/label/Mesh%20Relaxation gives some indication of the example models etc that the force density tools have been developed in mind, also this project http://www.grasshopper3d.com/photo/albums/ballito-lifestyle-centre  Typically this is akin to minimizing discontinuities in grids (British Museum Great Court Roof) or deriving equal mesh edge lengths.  The output is a mesh, it can be exploded out to vertex and faces (or face edges).  If you wish to create trusses, I'm assuming you want geodesics or similar?  In an object with polar symmetry like yours, I'd tend to generate a single and rotate around.  Do you have a sketch with what you desire?

5.  The mesh routines only work with the subset of "untrimmed" nurbs surfaces, thus the number of useful applications is diminished.  I haven't really coded this, although it shouldn't take that much time (it would compute an underlying UV mesh).  

I hope this helps explain a little more, keep posting other questions and observations.

Cheers,

Jon

Permalink Reply by peter fotiadis on December 27, 2011 at 10:57am

Hi Jon

Seasons greetings to you my friend.

"Do you have a sketch with what you desire?"

Well, I'll try to make something meaningful in Microstation and I'll post it here (hopefully "translated" to Rhino speech - God help us, he he).

In the mean time I'm trying to figure out how the Weave thing works on branches ( see stupid questions: http://www.grasshopper3d.com/forum/topics/weave-on-branches-stupid-... ).

The problem with quad trusses on certain surfaces (like the one provided as sample in the above thread) is that you need to mastermind an evolving truss "pattern/density" (so to speak) in order to cover dense U/V areas with fewer truss members and - depending on tube size criteria etc - to start duplicate the nodes used etc etc. Say start with a "wide spread" grid on U=1 and gradually end-up in a dense grid in U=n.

So in a way (if we forget the mesh approach that makes things impossible : single branch stuff) this is kinda an evolving weld quad mesh logic puzzle.

Best, Peter

Permalink Reply by peter fotiadis on December 28, 2011 at 7:19am

Here's a sketch (imported from Microstation - no Gen Comp used) to roughly outline the "truss relax" puzzle:

Imagine 2 surfaces (see layers inside Rhino file) that define the top/bottom (or out/in if you prefer) skin of a future to be truss (of quad type). If we divide the classic way these surfaces we end up with a dense grid near by the inner limits (no good - useless) and a sparse grid near by the exterior limits (no good - stupid).

So the ideal solution could be a transition schema from sparse (in) to dense (out). Obviously with no mesh around (useless for further work). Kinda like a fractal logic so to speak.

But...as I'm getting(?) more familiar(???) with GH...er...hmm...read my notes inside file. In a nutshell it's hard to imagine doing complex real-life AEC work with GH with the current state of things/capabilities (problematic clusters, one Canvas, one way control : from Canvas to Viewports), no real-life visual tree explorer/manager, no bake management of any kind, no preview management of any kind etc etc etc).

I mean (back to original definition) that I spend more time trying to preview portions of the whole solution (10% completed - imagine what's next...he he) and/or to figure out what belongs where ... than adding some new logic.

It's the law of entropy that is always proportional .... etc etc, I guess.

Keep in mind that this (as a typical AEC thing) is not predefined : it's not like applying a Voronoi grid into a blob...you work by trial-and-error, you add stuff, you see results, you change approach, go back again etc etc - meaning that the amount of "help" items present in a given solution is probably ten times the amount of the real (i.e. final) things. This is also a big problem in Microstation (blame that stupid old-times Level way to organize things, forget un realistic Cells and Planet Utopia Refs - at least with regard the current state of things).

It's quite explainable why CATIA dominate in the Plant market segment (and has serious plans to invade in more humble AEC segments as well).  

best, Peter

Attachments:

• 170_TERMINAL23C-OUTLINE.3dm, 937 KB

Permalink Reply by peter fotiadis on December 28, 2011 at 7:24am

PS: there's not even a Zoom Window capability available in Canvas - a small thing that could make life 100000 times easier. Like Luxology Modo : they designed the best GUI available (bar none) and forgot the most important View Nav tool (but I do love lab people, he he).

Permalink Reply by peter fotiadis on December 28, 2011 at 7:40am

Here's the faulty "twisted" sweep2 surface previously mentioned (wrong section deployment I would say)