how to relax my form? Using Kangaroo

Hi i want to relax the form i have in the Rhino file attached, exactly like the images attached where the form is anchored to particular points and is weightless or relaxed using Kangaroo everywhere else. I have tried several tutorials using kangaroo but nothing seems to work, My rhino file is attached and the points i want the form anchored to are in one of the layers.

Attachments:

Grasshopper.3dm, 76 KB
image 1.png, 678 KB
image 2.jpg, 85 KB

Replies to This Discussion

Permalink Reply by Jon Mirtschin on August 31, 2015 at 6:55pm

Hi Chris,

Hopefully someone else can assist with how to achieve this with Kangaroo, but if you're interested in alternative solutions then Bullant can calculate a relaxed mesh. Refer to the attached.

Cheers,

Jon

Attachments:

150901 mesh relax.gh, 16 KB

Permalink Reply by chris krambias on September 1, 2015 at 5:23pm

thank you

Permalink Reply by chris krambias on September 1, 2015 at 6:16pm

jon im getting a request for a license signature when im trying to look at your file, do you mind just sending me the gh file? Thanks in advance

Permalink Reply by Nik Willmore on August 31, 2015 at 10:48pm

Issue: your top "plane" is a zero height box so a closed mesh cannot be created from your overall solid object.

Unless your points lie exactly at mesh vertices, I suspect you will get odd artifacts of tension membrane effects between them, so in using Kangaroo's MeshMachine, I deleted the points (some of which were duplicates) and used fixed edge curves instead of the points input.

I even used custom meshing settings to get a uniform mesh to relax from your NURBS surfaces.

However, MeshMachine fails no matter what setting I try. To get it to relax at all instead of really just remesh, I have to set Pull to zero, for a true tension membrane, but then the changes per timer step are so great that the mesh rips and big gaping holes suddenly appear everywhere before it suddenly collapses.'

It does the same thing when I simply join your NURBS surfaces into a solid and use that an MeshMachine input. The fixed curves barely work differently than the normal edges that are not fixed. Just not made for this sort of thing, it turns out. This is the frame where one face dies just before in the next frame it all collapses:

I've used it before many times with Pull (to the original form) set to 0 with great success but the changes here are just too extreme for that to work I guess.

All the extra mesh tweaking components I added didn't help.

MeshMachine is usually magic since it does nice triangular remeshing with sharp-feature adaptivity to give smaller triangles. With Pull set to 0 though it will usually collapse fully so I only run it a bit in order to relax, say, a Boolean union of a bunch of Rhino solids, in order to get a curvy relaxed form. It seems that Pull = 0 fails with edges constraints added though.

With some setting I can't reproduce, I happened to obtain this wild chaos remeshing:

Permalink Reply by Nik Willmore on August 31, 2015 at 11:27pm

Millipede's Minimal Surface is lame for this, giving the fixed point artifacts I suspected points would afford, but worse, it never really relaxes well everywhere else either:

Well, actually, with a whopping 6000 steps, it does relax correctly, but fixing the right points will be tricky:

Let's point populate the hell out of the fixed edges. Alas, even with 3000 points, Millipede falls off the branch, worse than it did with only a few points, and it lacks fixed curves like MeshMachine has.

Permalink Reply by Nik Willmore on August 31, 2015 at 11:37pm

You have two superfluous faces in how you are fixing your edges all around two of your flat faces, and with their edges fully fixed, they can be omitted from the calculation and added back later, which lets me use another Millipede component to make a minimal surface from edges which would fail with those other two faces included, which would result in just two flat faces, I assume if I included them in the assigned edges to make a minimal surface.

Permalink Reply by Nik Willmore on September 1, 2015 at 12:11am

I can get Millipede's Minimal Surface to work better by determining the actual mesh vertex points to fix for the two surfaces you have pinned the edges of. It's terribly slow now. In all these non-MeshMachine cases there is no remeshing so you also get odd meshing topology due to the original layout being retained. It took 8 minutes for 3000 steps, and failed to minimize fully while blowing silly holes in the mesh:

Permalink Reply by Nik Willmore on September 1, 2015 at 12:34am

The Kangaroo2 Soap Bubble goal works great to immediately minimize the unpinned vertices, but I can't get it to stop undergoing frantic Brownian-like motion that is just to noisy in time to afford a final smooth point array to then mesh:

I don't see any Kangaroo2 settings component, nor any damping force as a goal component. The Plastic Anchor component does help, which fixes points when they stop having so much force on them, but it's not ideal, and it makes even worse the history artifacts where more points are in the center since so many more can from the original distant corner areas. It doesn't just spread out, and in fact soap film is based on triangles so it does retain original topology, the same connectivity:

Attachments:

Kangaroo2 Soap Film TEMP.gh, 19 KB

Permalink Reply by Nik Willmore on September 1, 2015 at 1:14am

Oh wow, Kangaroo2 also has a Soap Film accessory component called Tangential Smooth that does mellow it out and affords a stable result:

So how do I convert back to a mesh now though? All I get out are points, not faces. Reconstruction using Construct Mesh from the points and the old face indices using Deconstruct Mesh gives a jumbled result. Initially there is a tube artifact left behind as it all collapses into that, which then slowly is pulled out into the surface until the tube is gone, quite oddly. Initially it actually gives two tubes, one for each distant corner, that then merge into one tube:

Attachments:

Kangaroo2 Soap Film B.gh, 24 KB

Permalink Reply by taz on September 1, 2015 at 3:19pm

Was curious about this so took a look... Since the geo is so regular I built a coarse mesh by hand (not dynamic) based around a sphere quad mesh.Then I looked at the soap film example that shipped with K2. It seems the mesh still needs to be deconstructed into a triangular curve network for everything to come out as a mesh on the other end. Result is fairly stable using your same goal values.

Attachments:

soapTest.gh, 15 KB

Permalink Reply by taz on September 1, 2015 at 3:51pm

That all being said from a preliminary Millipede or other quick study (as you've shown) one can intuit what the minimal surface form wants to be which can inform selection of a better initial coarse mesh. Using the two excluded faces MeshMachine works well and provides a more stable version of the same resulting soap film as the solution from the more complicated 5 surface mesh.

Attachments: