Re: (CIM) Wooden Block CIM whitepaper

Hi Ted:

Neat!! I've been working on building a CNC workcell as well. I
like what you're envisioning; let me offer my own perspective and
ideas too....

The problem with CNC as I see it is in the programming, setup, and
babysitting it requires. We have "virtual paper tape" now, and it
runs faster and smoother, but it's all the same game as 40 years ago
pretty much.

Even with CAD data of a part, it's still a good hour or two of work
to set up the toolpaths and so forth in a CAM program. Setting up
the machine and edge finding, cut & measure, setting offsets, etc,
can take a half hour or more. Simulators are good, but you better
watch that first part just to be sure!

So I've been working on getting my mini-lathe, mill, and a robot
set up to service both with stock all on one table. Although
TurboCNC can handle driving all of that motion hardware now, it
easily triples the amount of programming and setup time. Since I'm
usually making only 10 of something, not 10k, it doesn't pay off that
much. A lot of people have this dilemna; all CNC does really is turn
the cranks - there's a lot of human input left over still.

I'd like to couple that to machine vision, but to actually handle
the machining ops, rather than to let a human watch. It should be
able to tell when an edge-finder jumps, whether the generated chips
are correct or not, know a crash when it sees one, check finishes,
and be able to measure and re-cut the part using a sort of flexible
optical comparator scheme and probes.

A second phase would be to encode decision heuristics into the
system, so that it can determine the optimal way to make a part in
the same way that machinists do, given nothing but a CAD
representation of the solid and the knowledge that drills make holes,
plastic has low modulus so it needs to be clamped differently than
steel, warpage is likely in such and such section types, so cut here
first, etc.... This should be modifiable over time based on feedback
from the workcell.

The end goal of all this is to be able to make machined parts as
easily as stereoliths now - send a .stl to the workcell, get parts
back. That'd be some real progress!

As far as I can tell from twiddling with possible algorithms, the
major concessions would be that stock would always have to be in
regular sections to start with, and that sections of the machine
would have to be painted some funny colors so that the computer can
figure out what's going on. Oh, and the cutting would be done dry
generally....

Easily a 5-10 year project to handle general cases, but I think
it'd be a landmark improvement. Faster drives, prettier interfaces,
more commands in the CAM - not so much... I think the EMC folks
working with STEP-NC are on the right track.

Dave Kowalczyk
Mercer Island WA
TurboCNC software --> http://www.dakeng.com

--- In CAD_CAM_EDM_DRO@yahoogroups.com, "tkosan" <tkosan@y...> wrote:
> Here is a link to the Home Shop CIM white paper that I promised
earlier:
>
>
http://tkosan.javadevices.org/jcim/files/home_shop_cim_white_paper.pdf
>
> My intent is to use this whitepaper as a guide for actually building
> the software and hardware needed to implement the vision it
describes.
>
{snip}