26 10 2008

I’m working toward producing a set of vectors for the joints between the tori. I just need to calculate where the these line segments intersect:



this approach is proving difficult to actually accomplish in code, mine or others.


23 10 2008

Our unit dodecahedron does not lineup with out unit icosahedron:

(±1, ±1, ±1)
(0, ±1/φ, ±φ)
(±1/φ, ±φ, 0)
(±φ, 0, ±1/φ)

(0, ±1, ±φ)
(±1, ±φ, 0)
(±φ, 0, ±1)

We need to scale and rotate the dodecahedron, not sure of the specifics.


23 10 2008

Further progress towards a working chassis:

this is what it would look like before inserting the coils.

here is a closeup:


23 10 2008

I think these are beautiful shapes that may make for good logos (with some framing of course!):

First Success

21 10 2008

Check this out. A truncated dodecahedral Polywell rendered in CAD.

I created this using ruby to pass draw instructions to mged (the main command line tool for BRL-CAD):

require 'matrix'
phi = (1+Math.sqrt(5))/2
icosahedron = Matrix[
[0, +1, +phi],
[0, +1, -phi],
[0, -1, +phi],
[0, -1, -phi],
[+1, +phi, 0],
[+1, -phi, 0],
[-1, +phi, 0],
[-1, -phi, 0],
[+phi, 0, +1],
[+phi, 0, -1],
[-phi, 0, +1],
[-phi, 0, -1]

icosahedron.row_vectors().each_with_index do |v,index|
`/usr/brlcad/bin/mged -f -c test3.g 'in torus#{index}.s tor #{v[0]} #{v[1]} #{v[2]} #{v[0]} #{v[1]} #{v[2]} 1.0 0.125'`

This basically iterates through the vertices of the icosahedron, and draws a torus normal to the origin. Now we are tantalizingly close to having a CAD file we can render in metal.


21 10 2008

First off, I’m all about open source software. Free, unencumbered, agile, fast.

For CAD I’m using BRL-CAD which is based on constructive solid geometry

For glue code I’ll use Ruby or Python. I know ruby I don’t know python.

SAGE for a mathematics package.

I havn’t looked closely at this yet, but here is a list of robots software packages

My laptop is a mac, I use linux on the server side, and will likely use EC2 for cloud computing.

Fabrication Technology

21 10 2008

This is my favorite part.

Lets start with Direct Metal Deposition. This video breaks it down:


This approach can create fully melted seamless structural elements out of stainless or tool steel. Incredible! I see using this for the structural elements such as the walls of the core.


For delicate welding there is electron beam welding. EMC used this to weld the toruses without damaging the the coils inside.


For an additive process with a fine structure we have  Electron Beam Melting. Here is a cool pic from Arcam’s brocure:

Electron Beam Melting

Electron Beam Melting

However this approach requires 1) a level work surface 2) high vacuum 3) covering the work surface with metal powder. Arcam’s current size limit is 200mm X 200mm X 350mm, so too small for the core. However the technology could possibly be adapted for a larger format. I image EBM could be used to fabricate superconductors in-situ (wild speculation, more later).

Finally, we have old fashioned robotic milling, as featured recently in Apple’s making of the macbook video:

Robotic milling and polishing allow you to achieve high tolerances and high finish even while using an additive process like Direct Metal Deposition.

What is so exciting about these fabrication technologies is that taken together they allow for a direct to metal workflow. A CAD file goes in one side and a finished product comes out the other (with hours of WTF in-between).

The Core

20 10 2008

The Polywell reactor has a core. This is the core of WB6 from EMC2:





This is what it looks like opened up:



the core exposed

the core exposed, you can see the coils of the electromagnet


So lets start by modeling the shape of the core. To make it more interesting, we’ll use the truncated dodecahedron. Supposedly the truncated dodecahedron has advantages over the truncated cube:


Truncated Dodecahedron

Truncated Dodecahedron rendered by ephi


This is a simple dodecahedron:




To find the midpoint of the dodecahedron’s planes, we can use the Icosahedron:



So now we can generate a torus at each of the vertices of the Icosahedron, orthogonal to the line connecting the vertices to the origin. But first lets take a look at some of the tool and technologies we might use.

The Big Idea

20 10 2008

The truth is, I don’t really understand the physics of the Polywell reactor. Fortunately I’m not alone. An often expressed sentiment on the forums is that you really need to “build it and see”. I couldn’t agree more. 

So I will attempt to build it. 

How do you build a device when you don’t understand the principles of it’s operation?

Trial and error.

Now you won’t get very far with trial and error if you are making a handful of attempts. But you _might_ get somewhere if you are making a large number of attempts. I’m suggesting applying brute force to the problem.

What this means in practice is a fully automated fabrication and testing cycle. Computers and robots will fabricate, assemble and operate all of the candidate reactor designs. This approach would allow for the fabrication and testing of a large number of reactor designs, potentially in parallel.

Essentially I am describing finding the solution using an evolutionary process. Although the candidate reactor designs can come from human designers or from genetic algorithms.

Now, this all sounds pretty expensive no? I’m sure it will be! So how does a broke loner like me make any progress?

Turns out quite a bit can be done in the pre-production that doesn’t cost money, although is does cost time (lots of it).

It all starts on the computer.

%d bloggers like this: