Generating a roof stucture based on a surface

Hello everyone,

I would like to get some advice how to generate a roof structure based on a surface created in rhino. I have already created a draft roof structure in Rhino - but as it was just a grid projected on the plane it doesn't look very accurate (especially where the surface sinks).

I have attached some screen shoots and an example of how more or less I would like the structure to look like.

Any help would be much appreciated!

Thanks!

Attachments:

Screen Shot 2015-10-18 at 09.28.19.png, 4.1 MB
Screen Shot 2015-10-18 at 09.28.00.png, 634 KB
example.jpg, 254 KB

Replies to This Discussion

Permalink Reply by peter fotiadis on October 18, 2015 at 1:53am

Hmm ... this is tricky (but not for the obvious reasons):

Ideally you should apply the following steps to that:

instead of this:

But ... well ... for doing it is not that simple (I have C# code that does this but is rather very complex).

Anyway ... let's forget the quad mesh for the moment:

1. Create some patched brep (i.e. a "surface" with trimming info: this [a bit confusingly, he he] is NOT a surface ... because it's ... er ... a brep).

2. GH Patch is very temperamental and frequently omits trimming the brep especially against the inner loop curves. Nothing too serious mind (I have a C# that deals with this issue).

3. Using MeshMachine create the simplest tri-mesh possible.

4. Use K1 (or K2) for "relaxing" the mesh.

Note: steps 3-4 are covered with this very simple demo case attached (load R file first).

5.Now ... the yielding mesh is a tri-mesh ... and that's not what we want: Quadrangulate (what a word) the mesh (THIS is very tricky - I use also C# for that) and using mesh connectivity data (mesh faces/vertices are NOT following the "divide-a-surface" ordered way of doing business) make some geodetic truss.

more soon.

Attachments:

Kangaroo_VS_patch_V2.3dm, 535 KB
Kangaroo_VS_patch_V2.gh, 43 KB

Permalink Reply by Asi on October 18, 2015 at 4:28am

Hi Peter,

Thanks a lot for your tips. Will try them out tomorrow ans see how it goes...

Permalink Reply by Joseph Oster on October 18, 2015 at 5:20am

Unless you just want Peter to do the work for you, or are comfortable working in C#, you will learn nothing about Grasshopper from Peter's code.

Permalink Reply by peter fotiadis on October 18, 2015 at 5:23am

On the other hand (and IF we are after quads "well organized" in the sense that the divide a surface pts DataTree is "well organized") ... we don't need meshes at all.

Here's a very old case miraculously escaped from the massacre (I've replaced all my old defs that use components with ones with C#). This case has a certain potential for 2 reasons:

1. If (with some changes in the def) we make a quad mesh (instead of that Delanuey thingy used) we can get rid of all the issues related with tri-meshing. If we replace quad mesh with a grid (see 2) ...

Here's another very simple test with regard grids:

2. If we can get rid of the mesh completely (and deform a grid > see def) ... then things start having a rather purposeful meaning.

In fact I have stuff that does exactly this but is solely via C# code (including K stuff since K2 is open sourced). Bad news: It's rather tedious to the max to "translate" them using native components.

Attachments: