Morphy wrote:
> I have come across some abreviations describing
> improvement and trouble shooting tools.
> I basically need a 30 word description of these,
This is not as short as I would like, but it I haven't time to write
short.
> information
> on where I can find the "root" of the "idea" behind the tool,
> possible extensions that make it much stronger, and more such
> abreviations that are not listed below. Abreviations are
> language dependent, in this case English is the likely language
> used.
> FMEA
Failure Modes Effects Analysis. At each step of the product development
(part production), what _can_ go wrong, what impact might this have on
the final product, and what are you (the producer) measuring to ensure
that this failure is detected early. Finally, how do you prevent it
from happening.
Emphasis is on the 'worst case' scenario. Can be very cumbersome &
complex, but at least you have documentation & thought abut the
possible.
Good source: Automotive Industry Action Group, (AIAG). workbooks &
explanations for a small fee. As applied by them, very manufacturing,
product oriented.
>
> DOE
Design of Experiments. The most cost & technically effective way for
gaining information about a system. OR: How to make very few
measurements of an operation, yet gain exactly the knowledge of how that
operation works - how the inputs control the outputs.
About 1925 R. A. Fisher asked, "if I make this measurement, what will I
learn?" Then he answered the question generically. Think of your
introductory stat class type questions. Measurements are made on two
groups of product, and the data + equations tell us if there is a
statistical (sustainable) difference in the two product types. With
enough data, we learn if a difference is sustainable. Among other
items, Fisher worked out the math for comparing multiple groups, or
multiple characteristics in those groups, simultaneously. Each
measurement does extra duty. One measurement is used 4, 8 or more
times. That's effectiveness.
Say you are marketing a new product. How should you 'place' it, what
guarantee, what distribution outlets, should you use to maximize sales?
These options are inputs to the system which has customer demand as
output. I would use a designed experiment (via conjoint analysis in
this particular case), to survey potential customers, and find out what
they wanted in a product. This way, it takes far fewer potential
customers to find out the most popular product configuration. Works,
and works very well.
Works for any system, with inputs and outputs. We want the output
(consumer demand in the example above), but we can only change the
inputs. How are they related? Do a DoE and find out. Lots of results
available on the web, my own web site among them. Truth: DoE is the
_only_ way to discover certain major characteristics of a system.
Mathematically true. Accept no substitutes.
Source: Many books on DoE in engineering. Some software. Box, Hunter
& Hunter (1978) is still a contemporary foundation book. DeVor, Chang &
Sutherland is good too. Taguchi's methods are also based on DoE. Lots
of "we bring it down to you who don't do math well" texts. For any of
these, look at the pictures. Can they illustrate designs, and results,
with graphics you can get into? If so, maybe it will work for you.
Today, math should not be an issue. We have machines for that. How to
set up a design, collect useful data, and understand the results _is_ an
issue. If you can make the graphics visualizations your own, you can do
a DoE. I use tinker toy pieces and a simple hands on in-class design
problem. We do the arithmetic on the blackboard. You should see what
it does for Science Fair projects!
>
> SPC
Statistical Process Control - If we measure (the output of) a
continuing process at occasional intervals, we can detect when that
process drifts, or when it jumps to a new condition. We can separate
small, normal variation from serious variation. The word 'control' is a
misleading. Should be 'monitor.' All SPC charts will use time or
incident number (sequential number of sampling instead of time) on the x
axis of the chart.
SPC is fundamentally reactive - after the fact. You have to make a
deviant part before you can tell that you did. Why it happened is still
up to those who are at the point of production. By contrast, DoE is
fundamentally proactive - before the fact. You do the DoE to find out
what causes deviations, so you can avoid (or use) them. You do SPC to
find out when one of those nasty causes happens anyway.
Sources: Again, lots of these. Grant & Leavenworth (7th Ed.) is the
current foundation book, but it's not for the weak of heart. DeVor,
Chang & Sutherland has a section on it. The AIAG has a nice booklet on
the nuts & bolts, derived from a famous Ford Manual. Step by step, how
to set it up, how to interpret it, aimed at production supervisors.
Most good intro stat books today will have a chapter on SPC, and often a
chapter on DoE. If these answer your questions, have a good trip. All
3 items you mentioned are 'common' in Industrial Engineering circles.
The American Society for Quality (ASQ), has a library which covers these
and of course more.
http://www.asq.org/
Help any?,
Jay
--
Jay Warner
Principal Scientist
Warner Consulting, Inc.
4444 North Green Bay Road
Racine, WI 53404-1216
USA
Ph: (262) 634-9100
FAX: (262) 681-1133
email:
quality@a2q.com
web:
http://www.a2q.com
The A2Q Method (tm). What do you want to improve today?