Building a replica FERMIAC mechanical computer

Similar projects worth following
The FERMIAC is a mechanical computer from the 1940's which was used to run monte carlo simulations of neutron transport inside nuclear reactors. I want to build one.

This is a bit of a challenge since Fermi only built one of them and little documentation is available.

If you haven't read the hackaday article about the FERMIAC, you can find it here.

In the 1930's Enrico Fermi and the Via Panisperna boys discovered that slow moving neutrons were much more likely to interact with atoms, enabling the later discovery of controlled nuclear fission. In the process Fermi developed statistical sampling techniques for analyzing complex atomic physics experiments.

In the late 1940's Stanislaw Ulam and John von Neumann were inspired to resurrect these techniques using the world's first general purpose computer, ENIAC to create what became known as the Monte Carlo method of statistical analysis.

Unfortunately, ENIAC was taken down for maintenance, leaving the world without any electronic computers for a few months. In the down time Fermi invented a mechanical computer to help perform Monte Carlo sims which was eventually nicknamed FERMIAC.

FERMIAC was actually a pretty apt name, the AC in ENIAC stood for And Computer, so the full name would really be Fermi And Computer. The FERMIAC device itself was only a small part of the simulation which required scale mechanical diagrams, tables of computed values and other calculations that were preformed by hand. The device wasn't really automated; Fermi used it as a tool in a much more complicated system for a few years until faster computers became available.

The only FERMIAC ever built eventually made its way into the Bradbury Science Museum in Los Alamos after being accidentally discovered in an office in 1966 as described in the Oct 1966 edition of the Los Alamos magazine The Atom (PDF, pg 9-11), where it remains to this day. In 2013-14 I became interested in finding out more about how it worked and how it was used but there was little information available (the wikipedia article was a couple of paragraphs long and listed only one source). I would occasionally google for more information but nothing would ever turn up; that is until a few weeks ago when I discovered the Museo Storico della Fisica e Centro Studi e Ricerche “Enrico Fermi”, an Italian museum had built their own replica in 2015. Fabrizio Coccetti, senior research technologist at the Enrico Fermi center in Rome published a paper about it as well as slides (pdf) from a talk at the 101st National Congress of the Italian Physics Society.

The slides contained a high-resolution image provided by the Bradbury Science Museum of a blueprint for the FERMIAC which was created after the re-discovery of the device by someone named D.F. Sterner at Los Alamos National Laboratory.

Finally, I had almost everything I needed to complete my project...

  • 1 × 3D Printer Series 1 Pro from Type A machines
  • 1 × Laser cutter Whatever they have at Metrix
  • 1 × Freecad For 3D design
  • 1 × Librecad For 2D design

  • Redline Updates

    Jeremy Ruhland10/27/2016 at 07:51 0 comments

    I pushed some small alterations I made during construction back into the repo, there was some cosmetic tweaking to the thumbscrew and slight thinning of the wheels to allow them to fit the smallest drum settings (I made my drums a little tight after gluing pt 1 & 2).

    The new files also correctly order the F/S indicator for neutron speed and extend the wheel C arms to give the read-line a little more room against the o-ring.

    Everything has also been cross-posted to thingiverse.

  • Together At Last, FERMIAC #3

    Jeremy Ruhland10/24/2016 at 03:07 0 comments

    Backdated Oct 19th, 2016:

    Everything is finally assembled! As far as I know this is the 3rd FERMIAC to have ever been constructed and the only one outside of a museum. The device is shown here on top of a mockup diagram of a nuclear fuel rod.

    The original FERMIAC is almost entirely brass and must weigh quite a bit. This version is very light which makes it quite back heavy, it tips up unless you've got your hand on the thumb screw. I might add some weight to the front to prevent that from happening during use.

    I've yet to use it to preform a monte carlo simulation because there are still a few unknowns about the process. I need to find out what the proper scale is for the diagram and I also need tables to select the proper distance measurement settings on drum F. The nature of the random values I need to feed in are also unknown, they follow a distribution that matches certain physical properties of neutrons in different materials, ie absorption probability, scattering potential, etc. More research is required before I can test it out but I'll be filming the process so others can see how its done.

  • Slicey Dicey

    Jeremy Ruhland10/24/2016 at 02:50 0 comments

    Backdated Oct 19th, 2016:

    Parts cut out at Metrix in Capitol Hill, Seattle.

  • Ready To Go

    Jeremy Ruhland10/24/2016 at 02:42 0 comments

    Backdated Oct 19th, 2016:

    The 2D design has been completed, ready for fabrication.

    You can see a small collision between wheel C and its arm support strut, this is because I duplicated wheel E's arms and translated them over to support wheel C. The actual wheel C arm support strut has its tab slightly lower down.

    The two square parts are going to be cut from acrylic and the remainder from 3/16" plywood. Green is my etching layer. For some reason the text on the far right turned white during the last copy/paste but its since been corrected. The angles of the etched lines in the acrylic were precomputed by Enrico Fermi to be equal to the angles that neutrons scatter off atoms and the directions that daughter neutrons may take after fission events.

    I'm using the old makerbot inspired tab/slot/t-nut right angle joint to hold the acrylic to the side supports. This design should be more mechanically sound than the original FERMIAC design which used four screws in close proximity to hold the acrylic to brass support bars and another set to hold the acrylic plates apart between brass standoff bars. The arms and struts on my design will be screwed and glued as necessary.

    The original FERMIAC kept its axles in place with washers retained by linchpins that passed through a hole in the end of each axle. On my version I'm simply sizing the length of each axle to sit flush with the surface of the side supports and arms and placing a piece of tape over the holes such that they axle won't slip out.

    As a funny note: apparently the S/F neutron speed indicators on Fermi's original FERMIAC were accidentally ordered backwards and corrected later with a piece of tape. I found out about this too late and made the same mistake on mine. Coccetti corrected this on his 2015 replica.

  • ...And Drums

    Jeremy Ruhland10/24/2016 at 02:10 0 comments

    Backdated Oct 16th, 2016:

    All the required 3D printed parts have been finished! The image below shows all 7:

    Drum D, pt 1Drum F, pt 1
    Wheel EWheel C
    Drum D, pt 2Drum F, pt 2
    Thumb screw

    The two parts of each drum fit together loosely and will have to be shimmed with some tape before being glued. The holes are sized for M3 or 3/16" rod. The large openings (visible on pt 2 of each drum, hidden on the underside of pt 1) are sized for 603 bearings. I was only able to find 3/16" rod which won't fit through my bearings so I ended up leaving them out. The PLA I printed these parts out of is slippery enough that they move freely on the smooth brass rod without bearings.

    I also started on the 2D design for the arms and supports that will hold the printed parts.

    You can see the overhead projections of the drums in red, along with top views of the wooden supports. White lines represent parts to be cut.

  • Wheels...

    Jeremy Ruhland10/24/2016 at 01:57 0 comments

    Backdated Oct 11, 2016:

    I printed out my first part: wheel C, which contacts Drum D and aids in the calculation of neutron time-of-flight. Its shown here with its o-ring attached.

    Since the neutron under study is traveling at a known speed and its path is being drawn on a scale diagram of the system under test the operator can keep track of of the neutron's time-of-flight using indicators inscribed on this wheel. The ratios between the radii of this wheel and drum D (two settings are available, fast and slow) provide the coefficient to determine time using the following equation:

  • Starting Out

    Jeremy Ruhland10/24/2016 at 01:36 0 comments

    Backdated to Oct 8th, 2016:

    Armed with my FERMIAC blueprint I started modeling wheels and drums, the main calculating components. The original components were manufactured on a lathe but I'll be using a 3D printer. Due to the overhangs in the design I decided to split each drum into two segments which will eventually be glued together.

    The following is an image of both components of drum F, which is used to calculate the distance a neutron has traveled.

    In the actual operation of the machine a neutron under study is given a certain amount of "distance points" which are deducted at different rates depending on the material its traveling through, these rates correspond to the ratios established by the multiple steps of the drum. When the neutron runs out of distance points a collision has occurred and it might trigger a fission reaction.

View all 7 project logs

Enjoy this project?



199123923 wrote 08/11/2019 at 19:20 point

Amazing work! Did you think of contacting Fabrizio Coccetti and sharing findings, asking more details etc.?

Regarding handling/functioning:

In the article "The Beginning of the Monte Carlo Method by N Metropolis" it reads:

"When a material boundary was crossed, another choice was made appropriate to the new material. The device could accommodate two neutron energies, referred to as "slow" and "fast." Once again, the Master had just the right feel for what was meaningful and relevant to do in the pursuit of science.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates