How to create nodes/bone structure like this?

Hello I have recently came across the "Entry Paradise Pavilion" in Germany and was wondering whether someone could point me in good direction on how to model it using grasshopper? I would like to be able to move centerpoints of the nodes as well as their thickness and length.

Thanks in advance for any suggestions

Replies to This Discussion

Permalink Reply by Wiktor Kidziak on January 8, 2011 at 5:01am

The error says that two consecutive edges have the same value (0). And therefore Weaverbird failed to produce a valid mesh. Does anyone know how can I fix it?

I noticed that although I am using very simple arrangements of boxes very often when I want to boolean everything together, some holes appear very randomly or half of the structure disappears. Could someone also suggest why is that happening?

Attachments:

grasshopper006.3dm, 481 KB
stretchedfabric001.ghx, 172 KB

Permalink Reply by Daniel Piker on January 8, 2011 at 5:15am

The problem is arising because you are starting with a non-manifold mesh.

I know mesh booleans can be a little problematic at times. Sometimes it helps to boolean smaller sections first rather than trying to do it all at once. It's also a good idea to use the Rhino CheckMesh command to make sure you are not creating something non-manifold.

In this case you might be better off building your starting block structure by placing all the blocks next to each other rather than these multiple overlaps.

Permalink Reply by Wiktor Kidziak on January 8, 2011 at 11:19am

Thank you for the suggestions. I have never come across non-manifold meshes. I noticed that in Rhino 5 there is a very useful mesh repair wizard that shows you where the non-manifold meshes are and can fix it by separating it from the rest of the mesh but it doesn't really solve a problem in this case.

I started again placing all the boxes next to each other but it seems like Rhino is very bad at booleans...

Permalink Reply by Giulio Piacentino on January 8, 2011 at 10:27am

Hi Arystoteles,

I've looked at the example, thank you for posting it. The problem arises from some very thin and long triangles (ABC) that are created after the boolean union. They are not invalid, they are just very thin (for example, A~B), because of how the parallelepipeds are placed, with some similar and coplanar faces.

When the mesh is subdivided, these triangles collapse completely (the distance was obviously very small and becomes smaller at every subdivision). But this time, because the numbers are so similar, in the thin triangles above we have exactly A=B. A triangle made of AAC points is not valid any more, and that error is shown. See the what-happens.ghx file.

So, how can we protect ourselves against this?

One way might be to start not booleaning boxes with faces that are similar and coplanar, but even in that case, for this definition, there is a simple solution. We might pull apart these thin triangles, effectively making them valid. The Laplacian component can help with that. See the stretchedfabric002.ghx file.

I hope this helps.

- Giulio

Attachments:

Permalink Reply by Wiktor Kidziak on January 8, 2011 at 11:21am

I appreciate your very clear and visual explanation! thank you

Permalink Reply by Wiktor Kidziak on January 8, 2011 at 6:41pm

Could you please suggest how I could use grasshopper to divide and unroll the mesh so it could be fabricated?

Permalink Reply by Mårten Nettelbladt on January 8, 2011 at 11:09pm

What material will you use?

Permalink Reply by Wiktor Kidziak on January 9, 2011 at 1:18am

stitched and tensioned heavy duty fabric

Permalink Reply by Daniel Piker on January 10, 2011 at 5:25am

Good question. Once you've got the right form, this can still be a difficult step.

I guess just how difficult depends a lot on the material and how stretchy it is. Using something like thin balloon rubber (or tights-like material like in your original reference image) it wouldn't matter too much the precise geometry of the flat pattern(as long as you have the right topology) because it can stretch and contract a long way - something like our virtual form-finding material.

At the other end, using very stiff canvas, the range of size change under tension is much smaller, so to achieve the same form it needs to be very carefully patterned.

I believe geodesics on the surface are usually used for seam lines because of their relative straightness when flattened.

You'd have to decide a topology for the seams that divides the surface into simply connected patches. Then you could draw them on the (now fixed) form-found surface and relax them (minimize their length so they become geodesic).

I'll be adding a ConstrainToMesh function to Kangaroo soon, which would work for this.

There's still the question of how small/thin the patches/strips you divide it into have to be to be within the possible stretch of the fabric. I guess this will be about the ratio between the curved and flattened area. How to actually calculate the flattening is another question.

Then there are more subtle issues of how the warp/weft threads are aligned, shear and the anisotropy of the fabric.

The work of Tristan Simmonds and others on Anish Kapoor's Marsyas installation is a good precedent for this sort of thing (and I think where I first encountered these ideas of dynamic relaxation):

http://www.arup.com/_assets/_download/download73.pdf

and here's a reference that might be useful as a rough general overview

http://www.meliar.com/design.pdf

I'd be very interested to hear from more people that have actual experience in patterning tensile structures, and about how it is done in other software packages, as I think improving capabilities for this could really increase the usefulness of Kangaroo.

Permalink Reply by Ivan Kiryakov on January 10, 2011 at 6:39am

I`ve tried do build structure like this out of paper just to test if it would work out of bent flat material and fairly good result came out.the form is generated with kangaroo,than the mesh was optimized to reduce faces and achieve irregularity in the faces.Next thing to do is 1:1 model :)

cheers

ivan

Permalink Reply by Daniel Piker on January 12, 2011 at 6:53am

Nice!

It would be interesting to see the flat pattern for this.

I think there might be ways to more automatically break it into strips. These strips could be triangulated, or if I can get it to generate singly-curved strips that could be really nice.

Permalink Reply by Mårten Nettelbladt on January 12, 2011 at 7:14am

The technique described in Figure 17 of that publication looks very interesting! (Geodesic D-strips) Perhaps not so easy to achieve in an "automatic" process?