|
|
|
 |
|
|
 |
Can't get the whole rep-rap stuff out of my head.
|
 |
 |
 |
|
|
|
|
|
Location: Blogs Dave Amphlett's Personal Blog |
 |
| Posted by: Dave Amphlett |
Tue, 15 Apr 2008 06:22:07 GMT |
I've got a big project that I've been working on for months, and I'm finding it tough to concentrate on because of this damned rep-rap stuff. More...
RepRap, for those who haven't heard of it yet, is an excellent project run by Dr Adrian Bowyer to produce an open-source designed Rapid Prototyper that is capable of replicating itself.
I've been following the project on and off for many months now, but
somehow this weekend it really got stuck into my head. I've never felt comfortable with the approach that Dr Bowyer it taking, although it's an established mech-eng route.
In the last few days my concerns have crystalised into an approach that I think is fundamentally better which I'm going to outline here just so that I have it written down and out of my head. That way maybe I can get back to completing the project that I'm already several months into, and come back to this later!
My approach would be a robotic arm (or arms) rather than translatable bed and tool head. The fundamental reason for this is that with one or more robot arms you at least have a fighting chance of performing assembly as well as parts production! A whole series of mental stages led me to the place where I believe that a basic single arm made from modular components has great bootstrap potential.
This weekend I saw for the first time the video of zcorp 310 printer producing a fully functional ball-bearing part, I love the simplicity and elegance of the technique which right now produces superb quality parts. For those not in a position to watch the video, the 310 is a very simple extension of an inkjet printer. it smooths a very thin layer of plastic powerder over an area, then prints 'binder' onto that 'paper-thin' surface. this binds not only the powder on that layer, but that layer to the layer beneath. drops the whole thing down by the depth of the layer, and smooths another layer over the top ready to perform the cycle again.
I want my self-replicating-rapid-prototyper (SRRP) to be capable of using techniques like the 310 print as it's such an efficient way of producing parts, but how is the compatible with a robotic arm as a fundamental approach? Well think about it like this - we use our hands to produce all sorts of marvelous goods, by using our hands to use TOOLS. Why can't a robotic arm use tools. Rather than the 310 printer being a self-contained printer, why can't it be a tool that a robotic arm employs.
What would a 310 printer NOT require if it was being operated by a robotic arm? well most of it's motors for a start. a robotic arm could draw the 'layer-smoothing' tool across the bed. It could press a button to lower the bed. A useful robotic arm would have to be able to do precision positioning, and so it could draw the print head across the surface. Hell, it can flatten the powder down, brush off any excess powder etc etc.
So the 310 printer designed for robotic arm operation becomes a relatively simple unit, most of the complexity being in the print-head. Let's call this a 3d powder printing tool. In fact I suggest that a robotic arm could produce 95% of, AND ASSEMBLE 100% of a 3d powder printing tool.
This gets to the heart of what I'm aiming for. What's the simplest robotic arm based prototyper that could build a 3d powder printing tool. Because once it can do that it can build so much more.
Ok, break here and let's attack from another angle. The other problem I have with reprap is the whole 'bed moving on a precision screw'. the precision screw shaft is so fundamental to how reprap works, and when can they expect a reprap to be able to produce one of those screws? Well I can't see it in the near future. It's a serious weakness in the whole approach.
So why is a robotic arm better than requiring precision screws? Well in and of itself it isn't. In fact the most 'likely' design people would choose for building one would probably involve a more complicated variant of a precision screw shaft. So I want to challenge this at a more fundamental level. Building precision things from precision things isn't a bootstrapable approach. So how about this. It's bad to need to use precision components to know that when we DO something (using outputs to motors), that we KNOW WHERE our tools are because they're all precision components. I keep coming back to humans as biology is the greatest bootstraping technology we've seen. You move your hand or fingers, and you don't KNOW exactly where they are just because you tried to move them by a certain amount. Instead you use your EYES!
So rather than doing certain things with precision components to magically know what the result will be, how about doing roughly what you think is needed using imprecise components, and using precision MEASUREMENT to establish whether you've achieved what you want. I need to work on this side of things more and more, but I'm sure that if we build some simple 'standard' (and very accurate) measures into the initial design, that it can boostrap more complicated but equaly accurate measuring tools into it's arsenal.
Optics is one way I suspect we could do all sorts of clever things in terms of measurements. 3 webcams in appropriate positions, with strategic led's on the robotic arm, might be enough to do some rather precise positioning. The true beauty of this approach is that it fundamentally provides for graceful degredation, which put another way is tollerance for imperfectly produce components - which is a mighty good thing in bootstraping!!!
So there you go - I can imagine a robot arm (two lateral extender joints, to rotator joints) capable say of holding and 'spining' a milling tool, and interchanging it with a drilling tool, all on a bed with a very basic clamping mechanism. I can imagine that being capable of machining wood or plastic to produce most of the components required to produce another arm.
Now I can imagine two of these arms being capable of producing ALL of the components required to make another arm. I'm serious here. Two robot arms could make stepper motors even!!! And if you take a proper carpentry approach you can construct much of this without the need for screws or bolts. Let's get back to proper 'joints'. That's bootstrapping!
I suspect that two robotic arms could produce virtually any tool they might need to produce pretty much anything else. Knocking out a 3d powder printing tool should be straight forward.
Making something that can make it's own tools which it uses to make more complex things - that's got to be the approach, and a translateable bed deposition printer just isn't on the right track. we need production and assembly, and what's going to be better at that than one or more robotic arms?
And bootstrapping is at it's most impressive when it's starting from a wobbly crappy easily achievable base, and how can that ever be achieved when precision components are a fundament part of a design. Instead let's work on precision measurement!
|
|
| Permalink |
Trackback |
|
|
 |
|
 |
|
|
|
|
|
|
|
|