Re: [CAD_CAM_EDM_DRO] Re: Soliciting Feedback: OS Design

I don't think that's an issue, after all we currently build machines of high
tolerance with machines built from lower tolerance. So how do we do it? With
the control system (humans and then computers). If the X slide has backlash,
we operate the machine against one of the stops to make parts of higher
tolerance than the mill's slides. Each generation should actually get more
accurate. That's actually one of the first objections I came up with.

In any case, the "self-replicating maching" is probably too ambitious for
today's technology, but I don't see a fundimental reason why a machine tool
shouldn't be able to make all parts it's made of, within reason -- I'm not
suggesting that it be able to make everything atom-for-atom. Yet, anyway.

It would need capability to:

* Accurately Positioning in at least 3 degrees of freedom, pref 4.
* Rotate a part against a fixed tool.
* Use rotary tool against fixed part.
* Accurately a measure part.
* Operate at wide range of speeds.
* Accomodate several different material hardnesses.

And in my opinion it would need to be able to support these operations
without complicated switcharound proceedures like a Shopsmith.

A tool changer would be nice too.

So what else?

On 7/24/06, Doug M <mccolld1@...> wrote:
>
> The biggest problem I see is cumulative error. Each generation is going
> to add accuracy degradation if it's copying itself. Even if building to a
> predefined spec., the accuracy is going to fade over time. If just copying
> itself, the accuracy of each "generation" will fade much quicker.
>
> Doug
>
>
>
> Mariss Freimanis <mariss92705@... <mariss92705%40yahoo.com>> wrote:
> This topic will quickly be ruled "OT" so I'll get in quick to beat
> the gun.:-)
>
> What you are looking to do almost fits in the "perpetual motion
> machine" round-bin. Almost.
>
> It violates some basic laws of logic and a whole bunch of economic
> ones.
>
> Logic law: A system can never be completely self-aware. The volume of
> information data to completely describe any system has to exceed the
> system itself. It's a never-ending and growing circle if you add that
> self-description data to the system. You do, the system gets bigger
> and the data to describe the added self-description grows as well.
> Around and around.
>
> Economic laws: So let's say you overcome the above. Some parts of the
> self-replicator are made of tool-steel and aluminum, other parts are
> made of silicon, copper foil and fiberglass-epoxy.
>
> Objection 1: You have to purchase the diverse raw materials your self-
> replicator needs to self-replicate. Some of these materials are of a
> special and rare nature. Will you pay significantly less for them
> than you you would for the fabricated replacement part?
>
> Objection 2: You pay for a specific function from a mechanism. A
> router does a truly crappy job milling tool-steel and a mill suitable
> for milling tool-steel makes for a really poor 4' by 8' wood router.
> Let's not even touch the pick-and-place function, board etching and
> drilling and reflow oven functions to self-replicate electronics. How
> much would this "Swiss-knife" machine cost versus one meant for the
> task at hand; route wood, mill steel or SMT-mount PCBs?
>
> Objection 3: You don't have a self-replicator turning out copies of
> itself when things are going well. You need a copy when the original
> breaks. When it's broken, how is it going to turn out a copy of
> itself?
>
> About ideas and names: Naming things is the easiest task. For that
> reason it should reserved for after you have done the hard work of
> actually building your idea and having a working version. Name it
> after the work is done, not before. That's good form.
>
> Mariss
>
> --- In CAD_CAM_EDM_DRO@yahoogroups.com <CAD_CAM_EDM_DRO%40yahoogroups.com>,
> "Dennis Schmitz"
> <denschmitz@...> wrote:
> >
> > A couple of years back, I saw a project being done out of a
> university
> > that was a first attempt at a self-replicating robot. The idea was
> > that it would use various plastics and metals to create circuits and
> > other parts. The idea was far to ambitious to succeed today, but it
> > got me thinking.
> >
> > A first step toward such a thing, and indeed toward a completely
> self
> > replicating factory might be a machine tool that would ship with
> > complete plans to make all of the parts of itself. Sure, this ruins
> > the business models of a company that might want to sell them, but
> > that's not the point, rather the point is, is it feasible?
> >
> > A complete machine would need lots of electronics and proprietary
> CAD
> > software to build, but that can all be bought off-the-shelf, making
> it
> > conceivable that once you have a machine, you can make more of them
> > with it.
> >
> > Sticking with mechanical stuff (because manufacturing multilayer
> PWBs
> > is still difficult), what would such a machine look like? How many
> of
> > the parts could be made with milling and lathe operations? Is free
> > motion control software mature enough? What CAD formats should be
> > used? How can we reduce the number of different materials required?
> > Could enough people get interested in it to create a
> > electro-mechanical equivalent of open-source software?
> >
> > I'm just interested in your thoughts about what it would take to
> build
> > a basic machine assuming you can buy some stuff like motors and
> screws
> > and single-board computers.
> >
> > What about licensing? Would people put up with software that
> > occasionally insisted on building a spare part?
> >
> > Imagine 20 or 100 years out, could this base design evolve into
> > something truly automatic, say garage sized, that had the capability
> > to create everything necessary to make itself, including
> > semiconductors. Assuming it could create all kinds of other stuff,
> > from watches to cars, how would this change the world? Do we really
> > want to give every tinkerer or psychotic teenager the ability to
> build
> > a fighter jet or nuke from scratch?
> >
> > What should it be called? (I'm partial to simple names
> like "Factory v0.001")
> >
> > Sorry if my random Sunday daydreams bore you -- feel free to ignore.
> >
>
>
> ---------------------------------
> Talk is cheap. Use Yahoo! Messenger to make PC-to-Phone calls. Great rates
> starting at 1¢/min.
>
>
> [Non-text portions of this message have been removed]
>
>
>


[Non-text portions of this message have been removed]