Robust Factorial Designs for Manufacturing Quality


Kenney, James M.; Cabana, Louis R.   (1987, ASQC)   Polaroid Corporation, New Bedford, MA

41st Annual Quality Congress, May 1987, Minneapolis, MN    Vol. 41    No. 0
QICID: 3299    May 1987    pp. 227-235
List $10.00
Member $5.00

This article is not available online. Contact us to receive a scan of the archive, in PDF format.

Article Abstract

A comprehensive, engineer-friendly approach to the important task of planning and executing manufacturing experiments is presented. The objective is to develop an appreciation for te concept of "filling out experimental space" when running experiments. Such an approach has been shown to be invaluable in producing robust products and processes which can move more rapidly through scaleup and stand up better in the customer environment.Emphasis is placed on the "up-front" considerations of the design and set-up of an experiment or series of experiments. The designs themselves are based on novel arrangements of two-level orthogonal tables. These tables - in a manner similar to the Taguchi approach - are then used during analysis to pool insignificant effects, providing more precise estimates of real factor effects. Additionally, through the use of unique standardized experimental layouts and a menu-driven LOTUS 1-2-3 program, data analysis is presented in a very practical and compact tabular format.Many hurdles which have traditionally created an unfriendly environment for industrial experimentation are addressed and put aside. Among some of these issues are:1. How can the proper number of trials (sample size) be computed to ensure that only the important effects are detected?2. If a fractional factorial test is called for, how well will the design be able to resolve main effects separately from interactions?3. What if there is an untested factor (e.g. day-to-day) present in the experiment - can the design and analysis easily accommodate this problem?4. Will the designs allow the engineer to avoid unwanted confounding of 2-way interaction?5. Can the designs easily handle difficult-to-change factors which almost always come up when trying to execute a test?Through the use of illustrations, worksheets and examples, engieners - even with minimal familiarity with statistics - are made to feel comfortable with laying out, running and analyzing factorial experiments. A Lotus 1-2-3 template program is used to take all the drudgery out of the repetitive ANOVA type calculations, and is both friendly and flexible. It has been developed to include most of the advanced thinking in today's experimental design philosophy:1. Leads the experimenter through the design and set-up phases - culminating in the printout of a comprehensive "Experimental Set-Up Matrix".2. Allows for flexible and iteractive use of error to aid in judgements of significance.3. Lays out results in an expanded, mcuh more engineer-friendly ANOVA table which contains both statistical and engineering data.4. Allows the experimenter to "pool" insigificant effects into error to obtain more precise estimates of the impact of the tested factors.



Browse QIC Articles Chronologically:     Previous Article     Next Article

New Search

Featured advertisers

ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.