Tuesday, November 27, 2007

Simple bending brake

This is a simple 2' bending brake I built recently. The goal is to validate the overall design approach in preparation for building a 4' version.

The leaf and bed are 1x4 oak from Home Depot. The bending bar is made of two pieces of 1x2 hard maple from a local lumberyard, topped with a piece of 1x4 pine, and a 1/8" radius is created with a beading router bit. The bed is screwed down to a large 2x6 for stability; the latter is clamped my workbench during use. (The reason the bending bar is in several pieces is because I didn't want to buy a wide -- read, expensive -- single piece of maple. I'll splurge for the next brake.)

The construction is pretty obvious. I ensured proper alignment of the edges of the leaf and bed with the edge of the piano hinge by "match drilling" each side separately while fixtured to a base, like this:

For the #8 wood screws I used, I match drilled using an 11/64" drill for the unthreaded portion, ensuring reproducible alignment. I also drilled a deeper, pilot hole for the threaded portion. (Note that, in so doing, I ignored the pre-drilled holes in the hinge.)

One problem I encountered -- perhaps because my pilot holes were too small, or not perfectly centered, or whatever -- is that the (admittedly, small and questionable quality) wood screws would torque off at the junction between the threaded and unthreaded portions.

I added setback stops to provide a repeatable location for the bending bar:

I aligned the stops by putting some scrap into the brake, raising the leaf to the angle I expected I would need to get a 90 degree bend taking springback into account, then snugging the bending bar forward evenly:

This is the bend line being set up for making a 2" wide channel. I figured out from some experimentation that my parts came out 1/32" undersize, so I needed to "steal" 1/64" or so from each flange, which is why my alignment is just a smidge to the left of the line here:

After bending, this is as far as I could go towards 90 degrees. This is due to inadequate leverage in my leaf -- I need to attach a handle:

But a few taps of a mallet put me all the way to a right angle. Note that the long 3/8" dia. lag bolts that you see pointing upwards and engaging the wingnuts are just about to get in the way of the channel if I bend it far enough. Also, you can't see this, but it's really pretty painful to tighten the wingnuts without mechanical assistance (hence the pliers you see), so one might as well just use regular nuts. In any case, the result is that, at one end, my channel is pretty exactly 2" wide:

But it's about 1/32" too small on the other end:

This is within spec for my (forgiving) uses, so I'm happy, though I will run some more metal through this to see how reproducible (or improveable) these results are.

The following are the inspirations for this brake:
  1. Dave Clay's brake, made of steel angle sections; and
  2. Murray Johnson's "Home Depot" (wood) bending brake.
The following are the things I would/will do differently next time:
  1. More leverage for the bending leaf;
  2. Make all 3 working surfaces (leaf, bed and bar) out of maple;
  3. Use larger and more durable wood screws that won't torque off;
  4. Make the bending bar out of one wider piece of maple;
  5. Secure the bending bar with bolts tightened from the top, as with Dave Clay's brake (above).

Thursday, November 15, 2007

Of Recreational Vehicles and Army Navy Hardware

Last Sunday, I visited my friend Paul Eastham, builder of an RV-9A aircraft, at his hangar at South County airport. We chatted about riveted aluminum, and went on a short trip to Watsonville for lunch. He very generously let me take the controls and boy, I tell ya, that was a blast! He is building a camera mount, so he also took the opportunity to teach me how to drive solid rivets, and I learned about rivet smileys. :) All in all, I had a great time.

As a parting gift, he gave me some leftover hardware (mostly AN3) to experiment with for my own projects.

Now, meet my son, Aden. He is a nut (so to speak) for AN hardware. It was like showing a bag of diamonds to a jewel thief. He had to have an RV-9A. It had to be made of "real, lightweight" aluminum just like the real thing, and it had to be made with AN bolts. These were, so to speak, the design constraints. Here is the result:

You may notice that it ended up being an RV-9 instead of an RV-9A. That's life, I guess. You start out trying to build one airplane, and you end up building the other. It just happens.

Monday, October 15, 2007

Curta calculator

Yesterday, at the Harvest Festival at Ardenwood Historic Farm, I saw an exhibit of antique surveying instruments. Among those was a Curta mechanical calculator. Fascinating little thing. I subsequently found out that there's a Curta simulator in Flash; that these things go for about a kilobuck on eBay; and that someone out there had the chutzpah to disassemble theirs.

Back in the early 1960s, my dad was studying at the École Nationale Supérieure de l'Aéronautique in Paris. My mother told me stories about him doing his homework late into the night, while she listened to the clicking of one of these things.

Saturday, September 29, 2007

Fragment of bracket detail

I recently put together a fragment of this design just to get a feel for how things go together. This is also the first time I'm using the Tempo zinc oxide rattle can primer.

Notice that the bracket, made of 1/8" thick material, is not a complete "T" shape. This is because I just happened to have a thin strip of the stuff, so I cut whatever I could and worked with what I had. This Is Only A Test.

Notice also that I had trouble getting the primer to go on uniformly. It was scratch resistant on the sheet material, but seemed to easily de-bond from the 1/8" plate. I think the latter was because I didn't slap it on thick enough. Surface prep was to scuff with brown Scotch-Brite, wash with warm water and Dawn dish soap, dry, then apply the coating.

You might think these random pieces of stuff I make are useless. Not so! I'll have you know that this latest creation of mine was used as a scoop to rescue a crawfish from the neighborhood street. My wife tells me that the handle on the side was helpful.

Tuesday, August 28, 2007

On bondage

Folks have asked me about bonding (perhaps with backup rivets) versus just riveting. Elsewhere in this blog, I've mentioned that I am currently pursuing a "riveting only" strategy. The question is, "why"?

In order to be useful, a design must fulfill a purpose, or "market" niche. The purpose need not be monetary -- the market in question may be that of making folks happy, spreading Peace and Love, or winning a competition just for the sheer challenge of the thing. In my case, my purpose is this:
Hypothesis 1. Traditional aircraft monocoque aluminum construction, using thin sheetmetal and large cross-sections, occupies a useful niche between welded space frames and carbon fiber monocoques. It is competitively light weight and rigid while being easy for beginners to build.

There is also another claim, less easily made:

Hypothesis 2. The methods of Hypothesis 1 can be used as the basis for selling parts to homebuilders allowing them to design and build their own configurations.

With that in mind, the question is: should we just use plain rivets, or should we mess with bonding?

The aircraft industry has been using rivet bonded construction for many years. Thus it is instructive to follow their trajectory. The original F-18 A-D had a multi-part metal fin. The newer E-F models have a single piece, bladder molded, heat cured composite fin. Sound familiar? Bikes like the M5 Carbon High Racer and the Velokraft bikes all use this method, as do many upright carbon frames. Clearly, if you have the equipment, this is The Future. In addition, absent bladder molding, people like Garrie Hill, Jim Scozzafava, Tom Traylor and a whole host of others have shown that, if one is willing to do layups, carbon construction rules the roads. Hence, once again, the niche is to find something easier than carbon, but lighter than space frames.

There seem to be conflicting notions out there about how much one needs to prepare a metal surface for bonding. However, among people who rely on it for a living (or whose companies will fail dramatically if their structures come apart in service), the consensus seems to be that it's not easy. Check these out for starters:
In general, it seems to me that proper surface prep for bonding in order to achieve the rated strength of modern epoxies involves at least lots of care, and sometimes toxic and/or caustic chemicals.

Thus my current direction is to just rivet. Every Schmoe can build one and it will very likely stay together. It's easy to see a crappy rivet and, conversely, if a rivet looks pretty, it's probably adequately driven.

Thursday, August 23, 2007

Easy Racers Javelin clone

My current design is a clone of an Easy Racers Javelin: 700c rear, 451 front, LWB. I figure this is an easy way to get started -- later on, I can try more fancy designs with integrated seats. I'm trying not to "overmodel", so here is a sketch of just the parts I need to build a simple prototype of the rear wheel attachment. I would certainly not build such a long structure to the right, but I want to give an idea of how it would fit together:
The dropout is made from 1/4" 6061-T6 plate. The stays are 5/8" diameter, .035" wall 2024-T3 round drawn tubing, flattened at the ends and attached with two #8 MS27039 machine screws:
The stays are attached to the body by (roughly T-shaped) brackets made from 1/8" thick 6061-T6 material:

They are held together at the middle by a spreader piece made from 1/16" 6061-T6:
Removing the stay and spreader piece, we can peek in to see that the 1/8" bracket is built up on each side with a 1/16" thick spacer, such that the stay is flattened on both ends to an inside dimension of 1/4". This ensures that it has a nice radius without cracking:
The approach I show here, with a thicker bracket and a stay flattened onto the bracket, is a lot simpler and more symmetrical than either riveting the stay from the sides or inserting the stays into the structure. Note that I'm relying on the brackets themselves to resist side-to-side forces since the 1/8" plate has significant bending strength (though I haven't done the math on that part to know for sure, mainly because I am not sure what the design lateral loads should be).

Early monostay design

This early monostay design was far simpler than my subsequent ones. Here is the overall view:Removing the right side skins, we can see inside. Each monostay side is made up of a top and bottom channel piece; these come together in the middle and are attached to top and bottom plates. The skins hold the whole thing to the main body of the bike.
Looking back, this design is somewhat appealing. Perhaps I should return to it one day.

Triangulated stays, funky mounting

Another remark about the design discussed earlier in this post:
Mainly, I just want to record the treatment of the stay attachment. The stays are riveted into two straps, made of 1/16" 6061-T6:
A third component, made of the same material, acts as a brace in between the stays and is riveted to the back of the bike structure to cleanly transmit side-to-side forces:

An experiment with tubular monostays

This was a quick experiment with tubular monostays riveted into the structure. I didn't really do any strength calculations on this; I include it here just to record the idea for posterity. Here is an overall view:Zooming into the attachment area and removing the side skin, we have:
After we remove a couple more parts, we can see the tubes nestled in there:
This design assumes that each tube is attached by a line of rivets on each of 3 sides (top, bottom and outboard) to achieve the necessary strength.

Funky design with tubular stays

This is a rather funky-looking design, again in the dual 406 SWB configuration. Note the mid drive attachment, a 1/2" diameter tube that is riveted between two bulkheads in the main structure:Here is the seatstay attachment. All 3 brackets are made from 1/16" 6061-T6 plate:
And here is the chainstay attachment. The main bike structure is extended backwards a bit to avoid making the stays too long, on the theory that the large rectangular cross-section monocoque is more rigid and lighter than the stays, so we use it as far as we can till we get very close to the wheel:
Removing the right hand mounting bracket and skin, we get a clearer view into the structure:

Full monocoque monostay design

Perhaps the most complex of my riveted aluminum recumbent designs is this one, the culmination of a whole bunch of experimenting with parts and CAD work. The whole structure you see here (not including the wheels) weighs 5.02 pounds.

Here are two SolidWorks renderings and one snapshot from the regular line view.

The way I build these assemblies, I generally construct a "part" that contains just pure geometry, containing most of the important driving dimensions and shapes for the rest of the work. The hope (rarely achieved in practice but true for some things) is that a change to the geometry causes the parts to automatically rearrange and resize themselves:
The design is a dual 406 SWB. The seat is a sheet of plywood, while the remainder is riveted .025" 2024-T3 Alclad sheet with fittings made of 6061-T6. Zoomed in on the front, with the side skin removed, we have:
The headtube mounting assembly is previously documented in this post so I won't belabor it further. The adjustable bottom bracket assembly is made of 2x2", 1/8" wall 6061-T6 tubing and 1/8" 6061-T6 plate. This was a trial design to see how far I could go without any welding at all; everything is held together using machine screws. Here is a detail of this area:
I suspect the middle part of the rail could be machined (or simply drilled with lightening holes) to save a bunch of weight. Here's an example from a different design to illustrate:

The main emphasis of this design was to investigate how I could "use the monocoque, Luke": my goal was to build as much of the primary structure as possible using the monocoque, eschewing tubular stays. First, a view of the overall structure with the monostay assembly and only the skeleton of the rest of the structure, to give some perspective:
When the rest of the skeleton is removed, we can see that parts of the monostay assembly extend into, and intersect, the main structure:
Removing one side skin and the two doublers on the dropout end, we can peek into the structure:
Removing the two long channel sections that come out towards the viewer, we have:
and removing one more part, we have this:
which should explain pretty well how things come together. The one part we just removed is what I call the "stay fork"; this one:
This proved to be a very difficult part to fabricate: in some parts near the "knee" of each leg, where it got pretty thin, the material kept cracking. Which led me to one important realization: parts are easier to draw than to make. One of the dangers of 3D parametric CAD such as SolidWorks is that one can go to town drawing complicated things that all auto-update whenever a dimension is changed, and have lovely geometries, but which are essentially unbuildable. :)

Sunday, June 24, 2007

Headtube assembly

This is my first attempt to record some old-ish (maybe 1.5 years ago) work. Here is a rendering of one of my riveted aluminum recumbent bike designs:

and here it is with the side skins removed for illustration:

This was back when I was thinking of using rivet bonded, rather than just riveted, construction. I planned to use 3M Scotch-Weld DP460 adhesive, and the seat was to be a sandwich of Rohacell structural plastic foam bonded between two very thin layers of aluminum. The only subassembly I actually constructed was the headtube mounting, shown here in CAD:

The parts are primed with Cortec 373 (thanks to Century Corrosion for arranging for a sample to be sent to me) prior to final assembly. The primer provides bonding support and corrosion resistance, and is non-toxic. However, the stuff is a bit hard to apply, and I was not able to get a uniform coat. Here is the final result:

As you might be able to see, the headtube has 4 flat surfaces providing solid surfaces for bonding and fastening into the sheetmetal. Lacking access to a machine shop, I improvised a jig from some hardware store metal and filed it down; it took several nights, and I would never, ever attempt this again, but I was finally done.

I used 3/32" stainless steel POP rivets, and #2 stainless steel machine screws (yes, #2 -- these little guys are small -- but, in this design, all they do is stabilize the bonded joints to let the adhesive do its work). The screws are countersunk from the inside of the tube to provide clearance for the steerer (sorry for the blurry pic):

and this feat is accomplished by my handy reverse (aka back) countersink cutter and pilot (tools I didn't even know existed till I needed them). Here is one more view for completeness:

Lessons learned:
  1. Rivet bonding is tricky business. The adhesive is gooey and sticks to everything, and you have to quickly get all your parts assembled and riveted within the work time of the adhesive or else you have an ugly piece of scrap. It's all a bit stressful.
  2. Did I mention that the adhesive sticks to everything? This stuff is amazing. And it dries hard and tough as nails. I tried some test pieces, ripping them apart with my hands, to get an idea of how strong it is and, subjectively, that stuff is serious.
More broadly, this launched me into some navel-gazing regarding the use of bonding versus just riveting. In favor of bonding, I get more rigid joints and better strength. But the flip side is the stress of having to get the surface preparation just right or else the glue doesn't stick, and the stress of assembling everything on a strict time limit while the glue hardens. And my whole idea is to devise a construction technique that can be built without fuss (or else, why not just do a carbon fiber layup and be done with it?). As a result, my subsequent designs have moved progressively away from bonding towards pure riveting.

Saturday, June 23, 2007

Latest riveted aluminum recumbent bike part

This is the first of what I hope will be a series of posts describing my work to come up with a simple technique for building recumbent bicycles out of riveted aluminum. Some of these posts will work backwards in time since I have a backlog of un-blogged material (parts and CAD designs), but I'll start with the latest thing I've been working on.

Here is a hypothetical chunk of a riveted aluminum recumbent, held against my Volae Expedition to give you an idea of where it would fit. The boxy part is supposed to be the "body" of the recumbent, forming part of the seat. The tubes sticking out are the seatstays attached to the body:

This work started with a simple sketch on dead trees. Yes, I've done the CAD thing for a while (using SolidWorks), but I spent literally hundreds of hours futzing around with minor 3D modeling details, so my current strategy is, "don't touch the computer":

The first step is to build join the "seatstays" into a flat assembly with the proper spacing. Let's pretend that this is it:

The tubes are 1/2" diameter aluminum from the hardware store (I didn't want to wait for an online order). The brackets in between are .025" 2024-T3 Alclad. By assembling these on a flat jig, I can ensure that the distance between the tubes is exactly 2". I used 3/32" stainless steel POP rivets.

Next, I built the "box" with a slot for the seat tube assembly to be inserted:

This box is also made on a flat jig -- there are no fancy 3D fixtures needed to hold everything in alignment. The two channel sections that form the frame, and the skin doublers that bridge the gap between the channels, are made from .025" 2024-T3 Alclad, while the side skins are .016". (Picky rivet geeks will notice a couple of edge distance boo-boos.)

The next step is to mate these two parts. Presumably, this would happen when the builder completes the body and chainstay and seatstay assemblies. The three assemblies would be clamped to one another and to the dropouts, with an accurately dished wheel, and some 2x4s would be used to align everything like this example. The seatstay tubes can now be match drilled through the pilot holes in the side skins to fix the alignment:

The next step is to add some shear ties to the assembly, maintaining the continuity of the body around the slot and transmitting shear resulting from side-to-side forces to the body:

If you look at the following picture, you'll notice, as I did, that the structure is missing an extra shear tie between the stays to maintain proper continuity of the monocoque. Filed under "note to self":

The last step is to fabricate a cover for the other side. This will presumably be the surface to which the seat (perhaps made of plywood) would be attached. Here it is match drilled, primed and ready to rivet:

(Yes, a couple more edge distance boo-boos.) The finished product looks like this:

The craftsmanship on some of these parts leaves a little to be desired. You'll notice, in addition to the edge distance mistakes, some scratching where I accidentally pushed the drill too far through the structure, and a few parts that were made a bit skewed. I don't believe this is fundamental to the technique -- rather, I think that I, with a young family and a demanding job, am just being a bit hasty.

An inevitable question is, how strong is this stuff? Well, a 5.5" deep by 2" wide beam made from two pieces of .025" channel (top and bottom) and .016" side skins is stiffer and stronger, in the vertical direction, than a 2" diameter, .049" wall thickness CrMo round tube -- and half the weight. With no welding, gluing or composite layups. Which is why I would really like to see this construction technique scale up to a full vehicle.