>> "Greg House"
>> wrote in message
>> I have heard of most things in QA and CIP, but what is EVOP?
EVOP stands for "EVolutionary OPeration". Basically, it is a process (mainly manufacturing) improvement method developed by Box in de '50s. In this method, the process control variables (e.g. machine speed, temp, etc.) are continuously and systematically modified, the effect of these perturbations on the process output is monitored and the conditions when output is improved are noted. IMHO a combination of experimental design and SPC.
"LamarJL" wrote in message news:JoJz7.12110$B82.3776493699...
> Howdy Greg,
> EVOP stands for EVolutionary OPerations. The concept was also commonly
> referred to as the "string of pearls" approach to finding the sweet spot.
> Using EVOP you make a change to your process, if the change was good then
> you keep going in that direction. If the change was bad you backed up and
> went the other way. Like the myth of biological evolution - it would take
> forever to find the best spot for each process parameter and you would
> completely miss any interactions between parameters. EVOP gave way to DOE
> and Taguchi style experimentation years ago.
> Jim Lamar, CQE, CQA, QS-LA, MQM, ABC, Do-Re-Mi, 1-2-3, yada, yada, yada,
EVOP is a special application of DOE so it can't be said to replace it. It is often a 2 level, 2 factor factorial.
The whole idea is that it is done with a running process so you have to keep the changes to the independent variables small so the the output variable stays in spec.
You can't optimize a full-scale production process except by running it, unless in the unlikely case that you know in advance that a small scale or pilot run exactly represents full scale production.
It can certainly take a while to run enough replications to extract the significance of the variables and their optima, but what's the alternative? You can certainly find the interactions. I know there are at least 3 approaches to analyzing the data. One method uses ranking but I have never used it. Maybe that doesn't show the interactions.
Some old EVOP messages are Here
in case anyone is interested.
Anyone here doing any EVOP experiments?
From: Suzanne Fiorino
Subject: RE: EVOP - Any experience/Mullins/Bigelow/Fiorino
One can think of EVOP as 'baby steps" experimentation. Typically a small full factorial design is used to set up the EVOP test and the intent is to get meaningful information while still producing saleable product. QualPro in Knoxville, TN, has an excellent experimental design course, very practical, and a section of it covers EVOP. Also George Box, who is at the University of Wisconsin, may have a course on EVOP.
From: David Thibault
Subject: Re: EVOP - Any experience/Mullins/Thibault
One point not mentioned so far: Since the changes for EVOP are small relative to those for DOE, EVOP must rely on many replications to separate the signal from the noise.
Date: Sun, 23 May 1999 12:00:56 -0400
Subject: Electronic reprint of textbook about quality improvement using simplex EVOP
The textbook "Sequential Simplex Optimization: A technique for Improving Quality and Productivity in Research, Development, and Manufacturing" by F. H. Walters et al has been out of print for some time. CRC Press LLC has now licensed us t he right to make electronic reprints of this excellent text. We wish to continu e the work to popularize the simplex EVOP methodology, and make it an everyday tool for many more professionals and researchers. We have therefore decided to publish and distribute the electronic reprint through our web site. You may fre ely browse the whole text as HTML or download it as an Adobe Acrobat PDF-file from Grabitech.com