Prototype Assembly

I've been quite busy assembling and setting up this prototype, among other things. Quite a bit of news:

First, I've retrieved the case from our outsourced water-jetter. He'd decided that lasercutting would be the more efficient process for the material, given the thinness. All told it was a hefty bill, but the components came back beautiful:

I'm quite pleased for my first time working with sheet metal or doing sheet metal design. I've already found a number of improvements and cost-saving measures to take on any future iterations, including eliminating four components and simplifying the design of the inventory chutes. I can also greatly reduce the amount of hardware needed in this way.








After a few nervous hours in the machine shop, carefully bending the super-expensive steel all to hell, I cranked out something vaguely resembling a condom vending machine:

 This picture doesn't really do it justice -- the parts look stunning honestly, the steel is very nice. Here I've got everything bent and spot welded, and ready to be assembled.

 These are the two inventory chutes, and three of the four brackets which the chutes mount with. I think I can eliminate the brackets from the design, I'll try to incorporate that in the next revision.

 Bearing bracket. I was really pleased with how this worked, the sheet metal origami really worked exactly as hoped, it's quite sturdy.

This is the (slightly inaccurate) cutout for the coin mechanism. I forgot to dimension a small slot at the top of the mechanism, which I've now got to cut out of the panel somehow -- the stainless steel is repelling most tools, and broke one pair of wire clippers already. I may have to break out a hacksaw. I'm also considering switching to Aluminum in the future -- it's easier to bend, far lighter, and approximately the same cost. Also easier to cut, which might mean I could use a less powerful (i.e. readily accessible and no labor fees) laser cutter.

 I've also been assembling the crankshaft, which is pretty simple -- slip the cogs onto a section of Al-6061 rod, press the bearings into the mounts, and bolt it all together. It's installed here.

Closeup of installed crankshaft. I have to adopt some unorthodox methods to tighten a few of the bolts, this is an improvement to the design I can implement -- quite a few of the bolts aren't hard to get to, but they aren't pleasant. That's one of those things which is hard to anticipate in CAD modeling, but becomes very apparent in real life.







I've since finished assembling the device, new pictures and info to come shortly. A little creative persuasion was needed to get a few things like they should be, more on that later. Things look good though, and with a few last steps, I should be ready for actual testing very very soon.

A Professional Opinion, Things Keep Rolling

This Thursday, I'll be meeting with Daniel Walczyk, PE, and David Banks, to review the current design of the prototype. Dan's offered his advice, and given that the design is a conglomeration of a few semesters and a few engineers at this point, it sounds like an excellent idea. Still waiting on the waterjet, but everything else is ready to go. I'm hopeful that, barring finding a major issue with the design on Thursday, I'll have something assembled here within the next two weeks inclusive.

Prototype Construction, cont'd.

I've actually kept quite busy for the last week. I've been sourcing and producing parts for the first working model, which is quite exciting. For this first model, whose primary purpose is to ensure everything in the design works as expected, I've chosen to use advanced production techniques not available in Ghana. This is mostly to reduce cost and effort, but also for the important fact that this prototype is meant to validate the design, and deviations in dimensions and quality could affect those results. If the design works with everything precisely in tolerance and properly assembled, I can start replacing high-tech parts with simpler ones, and start looking into what in the system can be replaced in the field with whatever's around.

So far, I've produced the large gear, both cams, and the two bearing mounts using 3-D printing, from an ABS plastic honeycomb. They're all within +-0.003" tolerance, according to the machine and measurements. I've also sent all sheet metal part blueprints to the water-jet, which has about a one week lead time, with a +-0.012" tolerance. I've also acquired all hardware and miscellaneous bits from McMaster-Carr, including a 6ft section of 7/16" Al-6061 rod which I'll machine six axles from.

Many of the parts -- specifically, the nuts, bolts, axle, and a few others -- are best purchased in quantity and amortized across multiple units. I've yet to put together an exact price estimation per unit, but I expect below $50 -- the sheet metal is the biggest unknown expense for production, and may drive costs up.

Below in Fig. 7 you can see the collection of 3-D printed parts -- anything which I've printed can just as easily be made from wood or even aluminum with only a bandsaw, file, and drill, which should ease production in the field. I understand the large gear may also be available from the same company which supplies our coin mechanisms, which would improve lifespan over a plastic or wooden component -- the necessary equipment to machine the specific pitch and modulus of gear needed is very uncommon, and not available in the easy-to-get-to shops.

Figure 7: 3-D Printed Parts and hardware (Bearings visible on the right)

More to come over the coming weeks, particularly once the water-jet parts are returned -- from there, assembly should take only a day or two. I'm hopeful for a working device by the first of November. From there, I can start determining which parts can easily be made in the field, and playing with different methods of manufacturing the system in less-than-optimal conditions.

Breakthroughs -- hopefully

I've received some tremendously helpful work from the previous engineer on this project, Ari Munic. It appears there was a much newer revision of the design available than I had access to, which essentially is what I'd been slowly working towards -- except the model is finished, and best of all, many of the mechanisms have been tested.

With Dr. Eglash's approval, I've moved on to the prototype production phase far ahead of schedule. A few 4am Solidworks sessions have made me familiar with the software, and I've been able to make the few last necessary tweaks to the design. All that's left now is to submit vector files to the waterjet faculty, acquire a few bearings and bolts, and make a gear.

I've elected to produce the initial prototype using advanced methods such as waterjet cutters, vertical mills, and machine lathes. I understand that many of these tools are not readily available to our team in Ghana, but I've ensured that all parts could as easily be cut by hand with a file and a jigsaw, or similar more readily-accessible techniques.

Repair of the system with replacement parts manufactured onsite should be easy enough, and particularly, many parts cut on the waterjet (the one unit without any real equivalent in Ghana) are structural members which are unlikely to be damaged or wear out -- the most probably issues will revolve around the crankshaft assembly, which can be more easily replicated with a machine lathe and hand tools.

Below (Figure 6) is an image of the current (nearly final) (hopefully) design, which I should be able to begin production on this week or next. More to come.

Figure 6: Zany Penguin 1.2

First Construction, Lessons Learned

I made it to Albany Steel today and picked up some stock for the tech demo unit. Went down to the shop later on, and made both straps (The W-shaped components which hold the Inventory Tubes to the mount). I constructed them from 0.060" Steel (16ga). I realized halfway through that 16 gauge is overkill. The design of the straps, while elegant and easily replaced with on-site manufactured components, is actually quite a pain to bend on the sheet-metal equipment here at RPI. I didn't realize we didn't have a pan-head bender available, which would make some of the middle bends much easier. Thus, I'm going to reproduce the components from 0.030" steel, redesign the rest of the system's components for this thickness, and attempt to improve the design of the straps to ease production.

I also picked up a small section of 3" square aluminum tube for the Inventory Tubes, and realized again that it's overkill. I'm using a 3/16" wall thickness component, which looks ludicrously overbuilt. I'm looking into finding 1/8" or thinner square tube, or perhaps a sheet metal alternative. The shape of the tubes isn't conducive to sheet metal bending, but perhaps I can work a solution.

After speaking with John, the shop foreman, he recommended having most of the sheet metal components cut using a water jet cutter. This should ease construction, and reduce the amount of layout required -- which we both realize will be a huge pain. Unfortunately, the laser cutter (which has more than ten times the precision of the water jet) is too low wattage to cut plate steel. I'm looking into alternative laser-cutters, I've heard rumors of a 85W unit (compared to the 45W in Johnsson) somewhere on campus, so this will be a possibility.

In the meantime, I'm going to source 0.030" steel, and source and cut the wooden mockup components. I'm also in the process of assembling supplies for the scraperod.

Experimental testing, technology demonstrator for scraperod

I've decided the easiest, quickest way to test the scraperod design is to simply build a mockup. I've incorporated parts for the current prototype design where necessary, and simplified others. This will allow me to reuse parts and materials where possible, reducing prototyping cost -- but also the use of mockup dummy parts will speed up building the demonstrator, and reduce wasted time if it doesn't work.

Parts shown below in Figure 5 are produced from a variety of materials -- those I've colored brown will be wood or MDF, for simplicity's sake. Parts shown in orange are either Al-60xx or 16ga. steel. I picked up some materials today (10/3) from a local metal distributor, which should allow me to build this weekend.

Figure 5: Technology Demonstrator

The exact construction of the scraperod remains a question. I'm currently planning on building the first prototype from 1/4" threaded rod, sheet steel, and bad welds. Depending on how that goes, I might refine the rod-and-weld method, or perhaps redesign the shaft running from the "foot" to the axle to simplify construction.Some kind of set screw design would be quite nice.

Off on a tangent: Making tangential force application work

The task of moving condoms from tangential force application is not an easy one. To tackle it, I've broken down the process into a few steps. Firstly, applying force to move the condom from below only. Second, ensuring that only the bottom condom is vended. Thirdly, real-world testing, since condoms aren't exactly a nice clean form factor, which will doubtlessly cause many issues.

My initial attempts were to model different materials for the pushrod, or scraperod. It quickly became apparent that this was an issue of surface area more than material, and perhaps a combination of the two would fix things. I played with the profile of the scraperod in a very non-scientific way, and observed that a non-regular radiused, long surface made huge improvements. The irregular radius -- a simple parabola offset from the centerline (Fig. 4) allows more pressure to be applied at the end of the axle rotation, ensuring the condom is in fact completely moved out of the inventory tube.

Figure 4: Scraperod Design

By using this design, I believe I successfully met my first step. The next will be to address the issue of moving only the bottom condom. This may involve applying extra downward force from a spring and plunger assembly in the inventory tube, or perhaps by redesigning the inventory dispensing slot with a flap of rubber to prevent distribution.

Below, video of simulation of the scraperod system. I've modeled the coefficients of friction very roughly from internet data, and obviously the condoms are still approximated as square boxes. I'm quite pleased, it seems to function better than the previous pushrod design.

First Steps: Analysis and review of previous designs, customer needs

I've gotten started by glancing through data compiled by previous engineers on this project. Various issues have prevented most of them from succeeding, and I hope to build off their successes, and failures. Much preliminary work is out of the way, leaving me the fun parts. Out of five or six designs suggested previously, most work recently has concentrated on the "Zany Penguin" design. While reminiscent of a linux distribution, it's surprisingly lacking in kernels. What it does have, however, is the basic theory of a pusher/grabber mechanism removing condoms one at a time from the bottom of a stack. This isn't the traditional method -- Typically, gas-station style machines (I'll refer to them as that from now on, i think) rely on a helix coil with condoms placed within -- similar to those big black and glass snack vending machines that are ubiquitous in every office building ever constructed. While this method was carefully considered -- if it ain't broke, don't fix it -- consensus of my predecessors was that the system has a few irreconcilable problems:

• Mechanically, it's complex. And difficult to source parts for -- complex gear drive, specifically designed helix inventory rack, etc.
• From an end-user perspective, it's fairly time-consuming and troublesome to maintain. Each time it's refilled, condoms have to be individually added to each "slot" in the helix, and lined up just so to ensure it vends reliably. Our goal is to get condoms to people as easily and cheaply as possible, rather than for profit, and those stocking the machines will be local business-owners and individual volunteers, not Pepsi Salesmen. 

Other designs were interesting -- there's some novel approaches, involving circular drums and such, but the one which shows the most potential remains Zany Penguin.

I was initially hopeful upon viewing the current (when I joined) CAD models: Things looked pretty well fleshed-out and ready for prototyping. However, as I examined the model more deeply, I noticed a few major oversights. (Fig. 1)

 

 

 

Fig.1: Zany Penguin, alpha build

What you're looking at is four rows of sheet metal square 'tubes', which I've taken to calling inventory tubes. The bottom of each tube is positioned above the axle, which rotates, engaging the bottom condom with a large, square block. This pushes the condom forward, and out to the customer.

Clearly this design isn't as finished as it first appears. Firstly, there's no system for getting the condom from the axle to the customer -- some kind of chute is needed. Also, the coin mechanism still needs to be integrated, a case design finished, etc. These are minor details, however, compared to the big issue -- will the device actually move condoms from the inventory tubes to their final destination?

One of the most valuable things an engineer knows is that, while every gadget works perfectly in his head, unfortunately, his mind does not encompass the entire world. The large problem here lies with physics, and typically involves friction. Both of these concepts were developed by physicists and scientists, and, as usual, it's their fault, not ours. However, we still have the problems to deal with.

A really nice tool for dealing with this problem is mathematics, specifically the kind done by a piece of silicon, not the engineer. I'm referring to computer-aided simulation. While the condom-inventory tube-axle mechanism is rather complex, some reasonable simplifications can get it to the point of easily tweaking with it and seeing what happens for various designs. The largest issue with the Zany Penguin Alpha is one of friction. A large stack of condoms wishes very much to stay a large stack of condoms, due to the weight pushing on the bottom condom. When the axle encounters the condom and suggests that it, and none of its friends, leave for the cold cruel world, the condom takes offense. Either it tries to bring ten of its closest condom-friends and jams the mechanism, or it politely declines the axle's suggestion, and the axle slips right past, to try again a rotation later.

What needs to be checked, is that the amount of force applied to slide the bottom condom out is enough to make it budge. However, it has to be applied just right, or else multiple condoms are going to start moving.

This is where the gas-station companies stopped, had a coffee break, and someone noticed a snack vending machine. The rest is history.

Of course, being stubborn and viewing this problem as a personal affront to engineers everywhere, I decided to wrestle with it and see what happened.

I started from scratch, for a few reasons. Firstly, I hate Solidworks. I've played with it a few times, and never really liked the taste it left in my mouth. I'm much more comfortable in Siemens NX, which, while a bit finicky and quirky, is actually much better at some things than Solidworks -- and pitifully terrible at most others. Of course, Murphy's Law also states that any CAD project must switch formats at least once throughout the design process.

Also, there's another issue with physics in Alpha design -- there's four inventory tubes, with an axle peg placed every 90 degrees around the axle. That means the device can only move through a quarter rotation per vend, and thus the axle stub can go from essentially -45 to 45 degrees, relative to a line normal to the face of the condom. I felt that it'd be likely that much more surface area than this would be required to contact in order to slide the condom out.

My first gross over-simplification, for the purposes of just feeling things out, was to treat condoms as a nice, homogenous, square box. This let me get a basic mechanism down, and proved it was feasible.
(Figs. 2 and 3: Views with case partially hidden/partially transparent)

Fig. 2: Front View

Fig. 3: Trimetric View

 

 

 

I've made a few changes to the overall design worth noting. Firstly, placing the inventory tubes offset from the rear of the device allows a longer crankshaft arm length -- providing a larger radius arc, and more tangential force application to the condom over a longer duration. I also played with placement and determined that offsetting the rod to the front slightly improved vending, as the boxes were fully pushed out -- with a centered rod, I ran into issues in simulation where boxes would get caught by their back corners under the next box as it fell.

I used mechatronics concept simulator for this first rough simulation, which shows just how well the geometry of the pushrod works out -- notice how the box is engaged, then the pushrod moves it forward, simultaneously lifting the stack above it up behind the vend slot. I'm quite satisfied with this design, and feel that it would work incredibly well with perfect little square boxes.

Actually, nice square boxes are one type of condom packaging. However, given the unpopularity of that form factor, we're designing for foil wrappers, which are much more common.

I experimented with varying the box dimensions, the "scatter" of the stack of boxes, and received very consistent good results. I then began experimenting with contacting only the bottom surface of the boxes, which will be a requirement if negligible- or variable-thickness, deformable foil packs are to be used. More on that next time.

Inventing a simpler wheel: Designing a condom vending machine capable of survival and repair in third world environments

I've recently joined Dr. Ron Eglash (Homepage) and David A. Banks (Relevant) in their endeavors to fight AIDS and STDs in Ghana through increased availability of low-profile options to purchase condoms. Our belief is that sociological issues prevent many from purchasing condoms at currently available locations, such as stores and health clinics. Field research in Ghana indicates that there's unfortunately a stigma associated with sexual activity among many Ghanaians, and that inroads to STD prevention can be made by providing discreet, readily available condom sources.

To that end, we intend to design a condom vending machine, with some specific qualities which differentiate it from those commonly seen here in the States.

While the ubiquitous "gas station restroom" condom vending machines are available for purchase, stocking, and shipping to Ghana immediately, they don't quite fit the needs of the Ghana market. Major issues with off-the-shelf units revolve around:

• Lack of readily available repair parts in Ghana -- Most designs are copyrighted, with schematics or replacement parts only available from a single, distant manufacturer -- quite an issue in a developing country. We intend to rectify this through gross simplification of the design, and by placing the design under open source/creative commons styled restrictions, distributing detailed information on replacement of components using available materials (more on that later), etc.
• Ruggedness concerns -- Again, the environment which we expect these devices to be exposed to will likely be far harsher than that expected by manufacturers of bathroom-wall units. This would render OTS units entirely cost-ineffective, they're surprisingly expensive.
• As part of something bigger -- Using the aforementioned public domain-styled design theory, we hope that our designs can be improved and distributed to far wider markets and saturations. One research team can only do so much, but the knowledge produced can do incredible things if properly disseminated.

My role in this project is that of Mechanical Systems Engineer. I've been tasked with analyzing previous designs, of which some have been prototyped to varying degrees of success, providing a workable design, building and testing prototypes, and developing a final production design from all this.