Limit to self replication

There are limits to the ability of even an advanced Reprap to self replicate. In particular, if the melting point of one of the components (such as the heating element) is higher than the extrusion temperature then that part will not be able to be made.

So we need to come up with a way of depositing high temperature materials at low temperature.

It would be wonderful if this led to a way of depositing metal at low temperature. It would be good to avoid laser sintering though this is the only method i can think of. Some sort of chemical process is likely to be the solution. Molecular beam epitaxy would just take too long.
any thoughts?

hey nathan, you bring up an excellent point that is something we definitely need to address. currently, we're only doing thermoplastic, and obviously we cant make the thermal barrier out of that. we wont be tackling this part with v1.0. however, for v2.0, i imagine we'll try a variety of methods.

the one that comes to mind for me, is using a syringe based extruder and ceramics. with the right consistency, we can extrude a stream of clay. once it dries, then we simply fire it in a kiln and have a fantastic, custom fit thermal barrier! of course there are a ton of other potential solutions, most of them just depend on having an extruder up and running =)

the problem in general that you're referring to is called closure (i think...), and it deals with how many external parts a system needs to be self sufficient. interesting side note: a very successful self-replicator (humans) doesnt even achieve closure. we need a variety of external materials to replicate (food, water, air, etc.)

***you bring up an excellent point that is something we definitely need to address. currently, we're only doing thermoplastic, and obviously we cant make the thermal barrier out of that.***

Well, the Mk 1 AEM extruder design definitely puts HPP within the realm of possibility as a thermal barrier ... and I can extrude HPP.

interesting. that would be AWESOME. the more external parts we can replace with reprapped parts, the better!

Same thing can be said of accuracy barrier (eg gears) and toughness barrier (load or tension carrying parts).

for the thermal barrier, apart from the clay method described above, we could also very easily use different types of plaster or cement. Some epoxys with metallic fillers are very heat resistant and can do the trick pretty well, as Zack showed with the jbweld. A range of polymers mixed with sand dust would provide an excellent material that have better thermal properties than the polymer alone. I think repraping extruders will not be stopped by the thermal barrier.

This is really what the polyceramic work is about. Taking a thermoplastic mold and using it to form a liquid catayst coldsetting thermoset composite. One of my future intentions is to look at then using the termoset material to mold INORGANIC (true ceramics) binders using 100C type heat to softset the inorganic then removing and firing it.

Your initial point is correct, there is a very real point of reduced returns from removing human input. But a little bit can go a long way.

Mike

The thoughts and ideas expressed in this post do not reflect those of my employer and are intended only as communications between individuals. Any attempts at implement are at your own risk

Edited 1 time(s). Last edit at 09/12/2007 09:42PM by ohiomike.

The way I see it, you'd have to do a paper-scissors-rock approach.

Any one material either won't withstand its own manufacturing process, or will "gunk up" with the material it is manipulating.
An alternate material, however, might both tolerate the conditions, and stay clean(er).

One idea for doing metals might be a wire-feed welder. The problem with this, and probably with your laser sintering, would be the metal itself wouldn't be organized in any meaningful manner. If you built up a long "structural support" by wrapping wire, much like a spring, then gave it a bend, it'd probably separate along the lines of the weld. Not necessarily at the weld.

As for ceramics...anyone ever hear of polymer clay? Something similar might work. Use a clay that will lay down firm, then bake it to cure it.

As for the energy of doing metal with heat, and the supposed energy cost. So? If you are going to melt the metal, aren't you going to need the same number of calories to melt one cc of metal regardless of whether or not it is one cc, or one cc out of one million cc's?

Personally...I think laser sintering in a additive system would be an excellent complement to a subtractive milling system. If you did it right, you could probably have a ready supply of dust from the milling, for the sintering. That, and you could save the milling for when you needed the metal to retain its crystalline structure, and use the sintering for everything else.

Long and rambling...I know.



P.S. I know it's a ways off, but I want to see a fabber that both adds and subtracts...in multiple materials, including semiconductors... Mk III?

yes! i like your thinking and enthusiasm. good call on the rock paper scissors approach... i have a feeling that is going to make it into the documentation somewhere. it certainly describes the situation well.

i've recently fallen in love with aluminum, its quite a sweet metal. i think it would be AWESOME to combine some of the backyard foundry stuff on the internet with a reprap machine. have reprap print out the molds for the furnace, add the refractory supplies, and you have a reprapped furnace. throw in your brick / propane + blower and you're in business.

of course the cool part (it gets better?!?!) is when you then use reprap to create lost wax casing molds by laying down thin streams of wax.

as you can see, i also want a versatile reprap system =)

I think we probably need to look at some sort of chemical process - use two part materials that form a higher temperature material when together. A dual syringe extruder should be enough. Self replication should be possible with smart design. (what would be amazing is to have one part as the regular material used so one one other material is required. But I don't know enough to say what would work.).

Maybe we should also be looking at non-metalic total replication for the near term. Which means removing all metallic rods, nuts etc and replacing with thermoplastic or two-part materials that can be easily printed. We also need a way to print screws (major challange i think) and also gears and belts.

I don't know that it matters, really.
Eventually, the goal would be for a replicator that can make objects from disparate materials. Even if the replicator could be made from a homogeneous material, not everything one would want from it could be. Consequently, the replicator will have to handle multiple materials, preferably without user manipulation. If after printing plastic, the thing stops until its user comes by to turn the item, and/or swap out the head or feedstocks, some complex items would take significantly longer to manufacture, simply due to the number of toolchanges. Complexity ceases to be free.

I suspect it'll be easier to make one fabber that manipulates multiple materials earlier than it would to discover or invent one material that fulfills all of the structural/thermal requirements for one fabber.

Count the heads.
Plastic, (possibly multiple types, which might require multiple heads, granted.)
metal, (again, possibly multiple types, although certainly something structural that is also a good conductor, and easy to work. And for these purposes, rubber counts as a plastic.)
ceramic,
semi-conductive.
Resistive, for heat.
paint.

Any consumer device you might name will have at least most of those materials in it. Some for structure, some for function.

As for what you mentioned. Isn't that what a binary epoxy is? Or cement?

Nathan Wrote:
-------------------------------------------------------
> I think we probably need to look at some sort of
> chemical process - use two part materials that
> form a higher temperature material when together.
> A dual syringe extruder should be enough. Self
> replication should be possible with smart design.
> (what would be amazing is to have one part as the
> regular material used so one one other material is
> required. But I don't know enough to say what
> would work.).
>
> Maybe we should also be looking at non-metalic
> total replication for the near term. Which means
> removing all metallic rods, nuts etc and replacing
> with thermoplastic or two-part materials that can
> be easily printed. We also need a way to print
> screws (major challange i think) and also gears
> and belts.

Whoops. I need to think more before posting. It'd save on posting.

Belts. I suspect it'd be easier to fabricate several dozen links to a motorcycle chain, although drawing a partially folded belt out of an elastic material would be possible.

Gears. I've about decided that eventually, any decent fabricator will have to have some machining ability to mill down the laid down material to tolerances tighter than the applicator head is capable of.

Additionally, a binary material that hardens into a contiguous material will probably tend to stick to itself.

belts and gears are something that we will probably (hopefully) be able to fabricate with darwin. also, once we get printing working, it would be fairly easy to design and print a bot that didnt use steel for structure, but rather printed parts.

granted, it would definitely take a while. and its never going to be 100% free. reprap simply reduces the formula to Time + Materials + Energy. of course, if you are printing something complicated or large, the 'cost' goes up. one way to get around this, is to go parallel. if you get multiple machines printing in tandem, you could very quickly increase your manufacturing capability.

luckily, you'll be able to print out more output capacity. ahhh... replicators =)

My point was that since you're going to need to manipulate metal anyway, and probably ceramic, why remove them from the design of the replicator?

If it takes equal effort to design the replicator to make things that combine all three materials, or to design the replicator to not need but one of those materials to duplicate itself, you benefit from the more complete replicator. The end result is still a machine that can reproduce itself, but a more capable machine. Granted, the bootstrapping of a more complex machine is itself more time consuming. Heck, the assembly of a more complex machine, even if you had a full kit, is itself more time consuming.

oh, i definitely agree. i would love to see a machine that combines plastic + ceramic + metal components that are all capable of being produced by said machine. that would be awesome!

I have to wonder if the 3 + 2 + 1 method, 3 axis controllers, plus 2 tool controllers, plus 1 interface, might be best replaced by 3 axis controllers, one interface controller, one tool controller, and some sort of "gripper" controller to select a tool.

Oh, and slots at one extreme dimension of the robot to hold tools.

This would allow for more flexibility in expansion. Need a new tool, or existing tool with a different material? Rack it in an available slot.

Of course, going back repeatedly for a different tool, just because you need two materials on pretty much every level, will almost double fabrication time.

Maybe when I get around to making one for myself, rather than commenting from the sidelines, I'll play with this.

There are Bismuth-Tin alloys with very low melting points in the range of 117 Degrees F (43 Deg. C), and others with higher temps. One interesting quality of some of these alloys is that they shrink when heated (not melted), which makes it easi to remove from molds in certain casting applications.

The alloys also exibit a very low thermal conductivity, which means depositing molten alloy on solid will not re-melt or deform multiple layers.

Roach_S Wrote:
-------------------------------------------------------
> Of course, going back repeatedly for a different
> tool, just because you need two materials on
> pretty much every level, will almost double
> fabrication time.

It seems to me that that would only be true if it took you as long to switch tools as it does to complete a level without a tool switch. Assuming you're making the same sized object either way, and both materials you're extruding can be fed at the same rate, the only extra time you would have to spend for a two material object layer is the time it would take to switch tools. That seems to me (from what I've read) to likely be a drop in the bucket, time-wise, compared to how long it takes to extrude even a single layer. This is especially true if you have stepper motors capable of significantly higher velocities than the extruder head can output at. This would mean you could zip over to the side of the project, grab the new head, and zip back relatively rapidly.


Getting closer to the point of this thread, perhaps the best solution would be to find a material thats physical properties change after being melted and re-hardened once. There are certainly materials out there (though I am no expert on them) that undergo this sort of change. Although this would be a case of choosing the materials for the machine, instead of designing the machine for the materials, this modified approach doesn't have the disadvantages pointed out earlier in the thread. Once you have an extruder made out of this custom material that is capable of high temperatures without a meltdown, it becomes possible to extrude more conventional high-melting-point materials as well.

also, keep in mind that not everything has to be printed in one go... it just has to be able to be printed. so, for example... we could print all the plastic parts, then print some ceramic parts, then some metal ones. if you really wanted to get printing in a hurry, you could have multiple machines running in tandem. then you wouldn't even need to switch heads.

that being said, being able to print multiple materials per layer is definitely a huge benefit. we'll definitely need the ability for doing support / thermoplastic, and it would be VERY nice if we could lay down conductors as well. then we could have complex geometric plastic parts with embedded electronics.

once we get that, who's going to start the open source roomba clone? =)

As the reprap can make it's own parts, it will be possible not only for it to replicate itself, but also upgrade itself. Things like automatic head switching will be able to be sent out in later updates. This has to be one of the greatest things about reprap, the ability to evolve

exactly... if someone can come up with an improvement that is capable of being printed within the current capabilities of the printer... then it can definitely upgrade itself. thats why i'm so excited about building my first one. its totally a chicken / egg type of dilemma.

I'm in the process of doing this right now. I'm printing a somewhat improved polymer pump for Tommelise.

I think there is going to be lots of diversity. That's a good thing. But it means there may be choosing in what your reprap can do and when. We will need lots of support from the original idea people and those that help make it better or more easily made. We will need to help develop hacking guides for adding software modules . We may need hardware development guides to help guide generation of additions that will be able to be self replicated and/or made with easily available parts. Steering committees to fold important additions or needed additions into a new release or branch. There will be families of lineage that may diverge and later converge as the differences arise or are encompassed. I'm looking forward to it all

Steering committees?

of course... to use the linux analogy again, it could work something like this: you have RepRap, which is the kernel of the whole system. then, you have various distros from that: debian, red hat, gentoo, slackware, etc. the equivalent for this would be things like tommeliese, biollante, and all the other various machines out there that approach the problem from various angles.

within each distro, there are committees and maintainers that decide which desktop to use, which librarys to include by default, basically how it will all be setup and what features they plan on supporting. it would be complete chaos to have it any other way. these can be just one person, or a group of people. regardless, there are decisions being made.

to bring it back to the world of reprap, we can see where it would be important to standardize on certain things like mounting brackets for toolheads, extruder barrel diameters, what toolheads are used, hardware interfaces, etc. if people dont agree to work within a standard environment, then there is chaos and the wheel is constantly being re-invented.

this isnt to say that one has to follow the plans exactly. just that for the sake of sanity, someone has to establish a nice, coherent, default base to work from.

committees!

yeah... committees have never accomplished anything.

I've had just under forty years experience with committees in both academia and industry. The only thing they're good for is wasting time and creating a captive audience for people with ego issues and misdirected yearnings for power and control. Good managers usually assign dsyfunctionional people to committees to keep them out of the way of people doing real work. Bad managers typically use committees to spread the guilt when it comes time to ask exactly what the organisation they're charged with leading accomplished.sarcasm whatsoever>

Edited 1 time(s). Last edit at 04/29/2007 10:21PM by Forrest Higgs.

Ok, forget I said committees, pretend instead I said interested hardworking individuals that are already doing lots of work and want to provide focus for those wanting to contribute.

Let me spotlight the obvious, to me at least. What's wrong with how it's happening now? Whether you realise it or not, Adrian has, I think, deliberately created an environment where technological progress is being made via a swarming strategy. It's been damned effective and has moved self-replicating printers along at a rate roughly a magnitude faster than what you'd ordinarily expect to see happen.

It just makes me crazy to hear somebody suggest that what's been happening is all very nice, but sooner or later we've got to slap a conventional organisational template on what we're doing. You want to see what a conventional organisational template accomplished in this field? Take a look at Stratasys' 10-K and it becomes very clear.

[library.corporate-ir.net]

They've got as good an conventional corporate organisation as you could want and they've had 15 years. Compare what they've accomplished with what this organisation has managed in about 12-18 months. We've got ambient FDM. We've got waterproof extrusions that don't delaminate. We've brought the costs down about 3 magnitudes.

What we've got is "swarm intelligence"

[en.wikipedia.org]

It's an artificial intelligence technique that's been around for 10-15 years except that we're using people instead of 'bots. We do rich interaction on the forums and blogs and we've got a bunch of people with varied and, relative to the staff at Stratasys (as an example), limited skills. All the same we're outperforming a cadre of highly skilled, experienced professionals with a lot more resources than we have.

[www.acfnewsource.org]

What we're doing technologically and organisationally, people, is so bleeding edge that it isn't even showing up in the literature.

For example, look what Zach just did with the aluminum bolts and acorn nuts. It's brilliant. Our strength is that even if Zach never comes up with another brilliant ideal like that, it doesn't really matter, because we have a couple of dozen people with very different experiences that, when confronted with a technology or diffusion, for that matter, obstacle are going to keep trying out different things till the problem goes away.

That's why I tend to freak out a bit when agents within our swarm try to lock things down with "standards" or "committees" that basically limit flexibility. That's why I find myself so nervous about the whole concept of Darwin and tend to be stand-offish about it. It's a reactionary impulse aimed at imposing an artificial order in what is otherwise an extremely revolutionary project.

Edited 1 time(s). Last edit at 04/30/2007 03:20PM by Forrest Higgs.

While I certainly understand the desire to use committee-equivalents to promote standardization, I agree with Forrest's assessment of their general usefulness. I know that with the wide range of ideas floating around for this project, things will get quite chaotic at times, but I feel that the best way to deal with the chaos is to let the standards evolve on their own.

Due to the self-replication aspect, this project can handle a less formal design process than other projects, since once someone comes up with a better idea, we can all simply print it out and try it, rather than having to change production processes, update kit parts, ship out new parts, etc. This makes it much easier for several designs to be tested by the community at large, and I think the more popular designs will get their own groups of followers, and the less popular ones will simply die off. New standards will be created simply by large numbers of people using the design that works best for them.

To determine the most popular designs and features, all you would have to do is look at the first page of the forums - the popular designs and hot topics keep getting bumped up, and the less interesting, less helpful, or less current ones get bumped to page 2 and beyond.

I agree that at times it will be necessary to have someone go through various systems and 'tidy them up' but I believe that a volunteer or two in times of need will be much more productive in the long run than a designated maintainer. Sure, it will be more chaotic, but we don't have to worry about design-on-a-deadline to keep the stockholders happy, and sometimes the most beautiful ideas can erupt from the twists and turns of utter chaos.