From: "Wayne Bengtsson"

An interesting solution to the scheduling problem of complex manufactured
parts has been found in something called "genitic algorithms".

Yes, I understand the difference between solving for the critical path using
exhaustive search and comparing 10,000 paths in the blink of an eye, selecting
the best 1000, and continuing for 1000 iterations. With the quantity matrix in
technological order, a finite problem not requiring exhausting search, you can
avoid exhaustive search and select candidate operations with a real good
initial population, adding the others with genetic selection.

The
result is NOT the ideal solution, but one that is plenty good enough, and
completed in a time frame quick enough to be useful.

Yes, that's all we need. I am not arguing for complete solution of the moment
to moment problem of which machine part to replicate next. Most frequently, you
just fix the one that's either broken, or the one that's worn out, and that
includes scraping ways. But what if the RS-232 level comparator on your CNC
goes out on Mars? Fabricate a new TTL gate? Stock the entire TTL inventory?

The nice people I work for manufacture,sell, and use machine tools that go
to the job.

I am familiar with running the work through the machine self-guided, dong the
work on the machine with controls or programs, and attaching the machine to the
work to produce a feature on the work which is too large.

For some items in metalworking, the only practical way to build even one off
is to build a jig or fixture.

Some. The process of replication starts with picking, from the existing
inventory of mills, lathes, drill presses, and various grinders, a pair of each
for study, determining which jigs and fixtures must be made to replicate each
and every part of those particular machines with themselves, and condensing the
resulting solution matrix into a specification for a pair of new machines,
building them, and knowing that each and every part on the new machine can be
reproduced by one of its interchangeable partners.

There's no CAD program that can solve that with a single click. And doing it
manually, or without a theoretical basis to guide progress and allocation of
resources, invites errors, budget overruns, and cancellation of the project,
never to be attempted again.



Yours,

Doug Goncz
Replikon Research (via aol.com)

Nuclear weapons are just Pu's way of ensuring that plenty of Pu will be
available for The Next Big Experiment, outlined in a post to
sci.physics.research at Google Groups under "supercritical"