Tuesdays...

Now that yesterday's published.

We arrived at the machine shop at eight this morning, with sheet metal haphazardly strapped to the roof of our ancient shuttlebus. Papa and Michael again met us there, and brought with them a few more students from KNUST -- Phillip, a welder and son of a welder, and a few of his friends whose names I'll remember as soon as this publishes. Aimee and I were able to jigsaw-puzzle three units onto our single half-sheet of stainless, which cut incredibly easily with the aviation shears I brought. Papa and I set to cutting parts, which Aimee then hauled over to the drill press.



Phillip and I were able to set up a jigsaw station and he took over cutting the interior cuts -- such as the vending area in the lid, and the coin mechanism location in the case. Before I realized it, we had an entire machine shop humming out vending machine cases.

Phillip hard at work.

All these hands made light work, and I found myself wandering aimlessly a few times -- So I took pictures of big shiny machinery. Some of the equipment in the KNUST shop is truly impressive, obviously the biggest ones are the coolest.

Massive German drill press --  Aimee for scale. The main column is 12" diameter solid steel.

Partially uncovered 12 foot bed lathe. For scale, the 4-jaw chuck mounted is 18" diameter. The workpiece is near eye level for me at 6 foot.

A truly massive vertical mill -- both tables are 2'x4'. I estimated the capacity around 5 foot cubic. The top of the head towers around 12 feet off the ground.

By mid morning, we'd cut and drilled all three sets of components, and were faced with our largest challenge yet -- how to fold the cases without a sheet metal brake. As I pantomined the device I wished we had here in Ghana, Phillip realized that there was a fifty year old brake in the shop next door.

First case cut in Ghana -- this was a good moment.

Fruits of our labors -- a bunch of finished parts

Sheet Brake!

Whoa, those actually look like cases.

With the sheet brake and the number of hands on the project, we went from sheet of steel to three cases in about four hours. I assembled a coin mechanism in the case to demonstrate, and asked Phillip to run a 50 Peswa through. I think this was the moment that the KNUST students suddenly realized how this all works, and they ate it up -- they were incredibly excited to work the mechanism, the ratcheting click of a coin going through is quite satisfying.

Somewhere during the morning, an unidentified entrepreneur contact of Ron's who I know only as "George" stopped by the shop. He was equally, if not even more excited as Ron described the purpose of the project. He's very interested in continuing the experiment after we leave, and I'm sure Ron knows more about that, as will I in the future.

After how well everything has gone, I'm convinced there's a nightmare somewhere up ahead -- Murphy works on this side of the Atlantic as well. Tomorrow we'll be hunting for and cutting wooden components, so there'll be plenty of opportunities to knock on wood, as the saying goes.

Mondays...

This post coming to you live from Kumasi, on the evening of Monday -- oh wait, no it's not, because my T430 decided to fry in the middle of writing it. The flashing light pattern indicates a fried systems board, of course, when i'm 5500 miles from Lenovo support. Fortunately we brought extra netbooks for another project, and I've been able to borrow one of those.

As I was saying yesterday...

We got an early start today in the KNUST machine shop, crossdrilling axles and cutting out sheet metal components. Upon arriving, we were met by two Ghanaian students, Papa and Michael, who immediately took over the drill press in a merry fashion. This got a nice assembly line going, and we were able to finish all ten axle assemblies, as well as cut and drill all ten interface plates by lunchtime.

Aimee hitting metal things with another piece of metal

Engineery looking things

Papa and Michael hard at work. It's wonderful to converse in the universal language of machine shop.

Lots of holes.

In the afternoon, we piled into a KNUST van and careened through Ghanaian traffic to track down sheet metal and additional components for various projects. Our destination was a district known as "the magazine," where we found essentially a square mile chop shop. Roads were packed with small shops and customers alike, with each area taking on a distinctive flavor -- at one end, we found dozens of tractor-trailers and heavy equipment in various states of disassembly. On another street, the road was covered in 1" steel cable being sorted by dozens of Ghanaians. Yet another corner sold massive AC/DC motors, some the size of a keg -- you can see them in the corner of the photographs below.

Looking down one street in the maze of shops

Unsuspecting engineers in their natural environment, also giant motors.

Also a street

Might be a street. Not sure.

At one of the new-product shops, we purchased about 20 square feet of 1mm Steel, which we'll use tomorrow for the cases and lids. I'm optimistic that we can finish a few units by the end of this week, and then produce the remainder quickly, as most parts will already be prepared in quantity. Tomorrow, we'll head back to the machine shop bright and early, and begin cutting the first two cases.

Arrival

After two days of traveling by bus, plane, and train, we've arrived safely at KNUST in Kumasi. We spent last night in Accra at a beachfront resort, the Afia Beach Hotel, where I discovered that my stomach doesn't agree with Ghanaian food -- or perhaps it's just food poisoning from Delta. One way or another, hopefully the queasiness will go away by tomorrow.

We're all very eager to get to work, and I believe we'll visit the machine shop Monday. In the meanwhile, I've unpacked and settled in. We strolled the open-air markets this afternoon, attempting to source e-waste type materials for further development on the project, and have a few leads to follow up next week.

Pre-Departure

We've been cleared to leave for Kumasi, Ghana tomorrow, following a regrettable but necessary delay beyond the field expedition's control. As such, we'll only be on the ground for about 18 days, instead of the anticipated ~28. While this will make the project quite challenging, I'm looking forward to the opportunity. I've used the last week to continue stateside production of components, and work out a few small details like mounting locations and hardware.

The project's online presence has spread -- while I'm still using the Google Drive repository for my personal development files, David has installed and hosted a wikipedia site for the project, as well as updating his personal page to reflect the project's current direction.

I'm slightly conflicted about the amount of manufacturing done here in the states. I'm keenly aware that, while the methods and tools I've used are comparable to those available overseas, there's no replacement for on-site development. I'm certain, however, that without prior manufacture, I'd be unable to deploy the devices during my brief time in Ghana.

From an engineering side, development has been frozen with Version 0.33. The manufacturing prototype exhibited excellent vending characteristics, with <2% jam rate and <4% double-vend rate. I'm attributing this primarily to improvements to the inventory chute, which also greatly simplified manufacture of the (previously "W" shaped) faceplate. Pictures of the updated design can be found on the wikipedia as linked.

Further updates from Kumasi as soon as I'm on the ground!

Assembly of 0.32, testing

I spent last evening assembling the latest revision of 0.32, the current production prototype. Below are some images from throughout the assembly process:

Cutting new interface plates from a sheet of steel

Cutting the first version of the new interface plate.

One of the major changes was the move of the interface plate from the inventory chute, to the axle/bracket assembly. Along with the front plate, which has been modified to a curved profile, these components help make the axle assembly far sturdier.

Completed "Magic Box," note the 2nd revision of the interface plate, using countersunk screws

 

 

Countersinks.

This technique requires using a thick (~2mm) steel (preferably) or aluminum sheet for the interface plate -- something rigid enough to hold up under the weight of the inventory, as well as smooth and flat enough to vend smoothly. Below, you can see the thickness difference between the thin casing sheet, and the interface plate's sheet steel.

Note the radius of PN008, the front plate. The interface plate, PN003, is resting edge-on in the vending tray lip.

I've also chamfered the corners of PN008, hoping to reduce the occurrences of condoms catching on the plate during vend. This hasn't been too helpful, as testing has shown. I'm optimistic that improvements could be found by tucking the plate under the interface plate -- a kind of tabbed arrangement, that would eliminate any and all edges from the process. I'd hoped at one point to convert PN003&PN008 to a single part, reducing the number of cuts to two cutouts for cam clearance, but requiring a sharp and accurate bend at the leading edge (where the two parts currently meet). Using the thick steel for the interface plate, however, prevents this -- so I've got to find a way to eliminate the corners between the parts which are causing the problem.

This revision also includes the new cam design, which I'm quite pleased with. I'm going to attempt to produce more of the cams over the next few days, and make sure this design will work for production in Ghana.

Upon assembly, I've collected about 100 data points of vending, 50 per chute. Below are the results:


I've also eliminated the gap designed into the inventory chute: by moving the interface plate to the "magic box" and doing this, the inventory chute is now symmetric -- with the exception of the interface bar, the bottom of the "W" shaped section on the face:

Inventory chute, old design. New CAD model to come soon. Note the W, however.

 I've experimented with flipping the inventory chute over, to determine if the interface bar is necessary for reliability. Below you can see the results.

Compare this with the results after reorienting the inventory chute to utilize the interface bar:

There's an appreciable number of double vends and a few non-vends. Those were all caused by a condom becoming trapped under PN008 during vend, which then necessitated a manual clearing of the jam after another failed vend -- thus a single jam shows as two error points. I'm optimistic that improving the PN008 design can eliminate this issue.

Hinge & Latch Design, Ver. 0.32

I've reconsidered the hinge mechanism, as well as some mechanical improvements to the interface plate and mounting system down there. Also set up P/Ns and hole patterns for security fasteners where necessary, and redesigned the axle mount blocks again. I'm too tired to describe in depth, but most modifications are clearly evolutionary and self-explanatory.

Full Version 0.32

Case hidden

Case assembly, fastener details seen below

These are standard tamper-proof screws, there's a variety of head designs available

The interface plate has been removed from the inventory chutes...

...and now forms the top surface of the newly-christened Magic Box.

As is the current custom, I've uploaded 0.32 to the public Google Drive Folder. This version is pretty much ready for production, barring the cam alignment and functionality issues that I'll fight with for the next month.

Cams

I spent a trying three hours in the Student Machine Shop today, attempting to cut out cams from a section of 3/16" pegboard. This turned out more difficult than I'd have expected, as I'd foolishly selected such a dense material that the laser-cutter was incapable of cutting at speed. I was able to cut about half of the laminate sections for one machine, which was enough to convince me that this wasn't saving any effort.

But hey, at least it looks cool, right?

I've got more than enough cams now, however, to experiment with different profiles over the weekend in a controlled fashion. Tuesday is the next opportunity to use the campus lasercutter, but I'll likely cut new cams tonight using traditional methods. The incorporation of the alignment peg has allowed me to improve the cam mounting system as well -- before and after models seen below.

Old cam mounting method, note the professional screw and cotter pin

New cam mounting method, cutaway view

Both cam halves visible here.

 This new design constrains cam motion better, allows installation and removal of cams without tools or permanent modifications, and reduces each cam location dimension to a single cotter pin mounting point -- it looks more complex, but simplifies the installation and construction of the cams greatly. I'm also very optimistic about a split cam, as opposed to a solid cross-section. I believe that this will increase pressure at contact points, increasing the likelihood of non-slip frictional contact there.