From: John Larkin
Subject: Re: Design of Experiments Workshop in NYC, SF, Miami
Date: Mon, 09 Dec 2002 16:54:18 -0800
Organization: Posted via Supernews, http://www.supernews.com
References: <3DF00F90.firstname.lastname@example.org> <email@example.com>
X-Newsreader: Forte Agent 1.91/32.564
On 9 Dec 2002 14:46:25 -0800, firstname.lastname@example.org (Brian Teasley)
>John Larkin wrote in message >
>> this is actually an interesting issue. I've read a bit of the DOE
>> stuff, and that seems, to me, to be nearly my natural enemy.
>> One cannot use DOE to design bridges, skyscrapers, or space shuttles.
>> They have to be *right* by design, with small margin for ignorance. If
>> electronics and software are designed with similar levels of theory
>> and discipline, there's no room or need for DOE or Taguchi techniques.
>Sure, I agree with you. I've written some software code myself, and
>have been involved in decent sized software development projects as
>well. During those times I have never thought of any reason to apply
>any DOE techniques.
>In reviewing the posts, there might be a little confusion about the
>term "test design" that I used. It refers to the way a test is set up,
>rather than a specific design of a "product" that is put together
>solely for a test.
>DOE methods are used when you do NOT know exactly how a process works.
>If your designed skyscraper is using some new type of glass, there are
>plenty of temperatures and times and chemicals, etc., that influence
>the thickness, strength, opaqueness (?), etc., of the glass. What is
>the optimal combination of chemicals levels, temperatures, etc. to
>make the glass?
>A "Full Factorial test design" would be an experiment in which you
>test all combinations of your chosen variables and their selected
>levels. You conduct your "test" or experiment, and carefully analyze
>the glass thickness (or opaqueness, or whatever) against all of the
>different settings of the different variables.
>As for the space shuttle, I know they DO run simulations to test
>designs, and they have some very serious computers to do the work. How
>do different angles of attack (when landing) and ambient temperatures,
>humidities, etc., effect the shuttle? (Including the nose cone and the
>famous "tiles"). The manufacturing of the tiles is certainly an area
>for DOE application. I haven't been involved in the shuttle
>simulations directly, so I don't know if they use DOE in any way or
>not, or even exactly what it is that they are simulating. I have been
>involved in the manufacturing of parts for the project, and DOE
>certainly is used in along the way.
I guess DOE would be more useful in a materials-development situation,
where the actual physics and chemistry is not fully understood. But
this is an electronics *design* newsgroup; a non-trivial electronic
design might have scores, maybe hundreds of tunable variables and
several, maybe dozens of optimizable outputs, some of which (ease of
maintenance, second-sourcing, sheer beauty) are not really
quantifiable. Things this complex can only be fine-tuned by simulation
or experiment, and that is most often to tweak the effect of one,
sometimes two continuous variables onto a measurable output... just
turn the pot until the noise figure is minimized.
Some solution spaces are just too big to wander around in.