Friday, December 15, 2017

Figure Sculpture Fabricated in Paper

This week my son and I finished fabricating a digital figure sculpt I created three years ago. Well it was a long time coming. But I'm delighted it's done!

It was sculpted from two live models in an embracing pose.

I used ZBrush as I usually do and the result was a high resolution mesh. This was greatly reduced in complexity and unfolded. It was then plotted on paper and cut out. Then the parts were folded up and glued together. The full details of how that was done are shown below. 

Here's a video of the result: 

Fabrication Details

I wanted to make this out of folded paper. This means the higher resolution ZBrush sculpt needed to be decimated down to far fewer triangles. Here's the sculpt at 1000 polygons:

ZBrush decimates nicely. But the arrangement of the triangles is not necessarily aesthetically pleasing. To improve that, but not alter the form much, I used the simple mesh editing tools inside Rhino. The main tools were SwapMeshEdge and SplitMeshEdge.

This takes the form from as on the left to as on the right. The difference is subtle but the flow of the polygons is better at matching the flow of the human body.

Next I wanted to colorize the triangles. I did this using two simple Grasshopper definitions:

The upper one randomly assigns grayscale values to the triangles of the male figure. The lower one assigns a color gradient to the female figure. The lowest triangles get the start of the range and the top most in height get the end of the range.

By changing the gradients you can quickly alter the look and feel:

I settled on this scheme:

To fabricate the parts need to be unrolled and cut out, then reassembled. To unroll we used Pepakura Designer 4. The parts look like this when unfolded in Pepakura, then broken up into smaller sections, then imported into Rhino:

The black lines are cut lines with tabs. The blue are mountain folds (convex portions of the form) and the red are valley folds (concave).

Coloring the Laid Out Parts

Once they are laid out flat the color information is lost. In order to match the laid out parts up with the colored 3D model the area of the triangles are used. This is done with Grasshopper as well. There are two Python nodes to make the overall definition simpler. One converts a triangle mesh into a list of meshes, each which has a single face. These become colored surfaces. The other Python node finds the laid out triangle with the same area.

Triangular surfaces are build with the correct colors. This gets nested on a sheet so they can be plotted at Taubman College. There are cross-hair marks on the corners of the sheet. This allows the sheet to be oriented on the Zund  Knife Cutter bed to match the cutting file.

A simple test was done cutting with a utility knife and securing the tabs with Krazy Glue.

Okay... it works. Onward.

The Zund

The idea was to cut the parts from the sheet on the Zund Knife Cutter. This CNC machine holds a knife (think X-Acto blade) which it can rotate to align with the cut. It's very fast and is perfect for cutting out all the parts... in theory.

The plot is aligned on the Zund using a laser (you can see the small red dot in the image below). I had boxes draw in the corners of the sheet I could use for test cuts to make sure it was aligned well. I cut a few test on opposing corners. It was not perfectly dead on... but was very close. I thought it would come out well.

The result were... drum roll please... horrible! Zoom in and check out the terrible alignment issues:

In addition to alignment issue there were issues with the parts being pulled loose. At maximum vacuum they still came up enough to tear. I'd tape them back down and hope to get through more of the sheet. But all in all it was a disaster:

The corners lined up well. I dunno man, I dunno. I think it might work if I used sheets of plastic. It would be thicker, and have much greater resistance to tearing. That of course still doesn't fix the alignment issue.

Cut and Assembled by Hand

So, the parts were cut by hand: Robots - 0, Humans - 1. Here my son, Kirk, cuts and scores the parts. It took 32 hours to cut and assemble both figures!

This work is challenging and I'm glad he was helping me. 

Almost finished with the assembly:


Here are the figures fully folded and glued.


It's actually quite challenging assembling forms of this complexity. I think about 500 triangles per figure is about the limit of what's doable while maintaining sanity, or at least quality of life. As a first go I thought it turned out well. But both Kirk and I learned some things about how to break up the parts into pieces which are easier to fold. And also some methods of sequence of assembly which make the whole process go easier. Our next attempt will be better!