Quality Optimization by Design: Global Synergy


Barker, Anne M.   (1990, ASQC)   Rochester Institute of Technology, Rochester, NY

Annual Quality Congress, San Francisco, CA    Vol. 44    No. 0
QICID: 9590    May 1990    pp. 1047-1050
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

The products of today and tomorrow require a degree of quality and reliability that often seems unattainable. The engineer is assaulted with many solutions, sometimes conflicting, which promise a means to achieve that goal of high quality, and do it quickly.

The concepts and practice of statistically designed experiments can appear to the engineer to be concentrated in two areas: the traditional Western techniques and the methods of Genichi Taguchi, both with the same roots and both with the same goal of a quality product. On one hand is a series of experiments yielding a functional understanding of a process, concentrating on location, while on the other hand is the concept of robustification against outside noise with low variation, as well as achieving the target.

There are disadvantages, as well as advantages, to each approach. The traditional methods can seem overly complicated, especially for the study of several variables, along with little attention paid to variability, while the Taguchi approach is criticized for such things as ignoring the detection of interactions, lack of randomization, and the use of signal to noise ratios., Aside from the advantages of an understanding of a process on the traditional side and robustification with low variation on the other, the fact that experimentation is being done by either approach is a laudable achievement.

The system described in this paper is a blend of the two approaches, taking the best of each by using two-level fractional factorial designs for screening a large number of variables, followed by (or as part of) a central composite design to detect curvilinearity, with all of the experimental results used to generate' an empirical equation through regression. Then the equation is exercised through a parameter design by computer to find the optimal points for a robust product that is on target with low variation. The simulation is completed with a tolerance design by computer .to find and reduce the variation of the quality sensitive components.

This is a technique which produces much faster development of quality products, while attaining an understanding of the underlying system. An example will illustrate the concept.



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.