Presented at 1996 Fall Technical Conference
Dr. Wayne A. Taylor, Taylor Enterprises, Inc. (www.variation.com)
Robustness is a key strategy for achieving high quality – low cost products and processes. Three different approaches to robust design are commonly used: the inner/outer array approach advocated by Taguchi, the dual response approach using response surfaces and the tolerance analysis approach which also uses response surfaces. Each of these approaches will be explained. The three approaches will then be contrasted.
The three approaches differ as to how the studies are run. Requirements for using the different approaches will be compared. In robust design, the objective is to estimate the effect that the targets of the input variables have on the variation of the output and to select the set of targets that minimize the variation while achieving the desired average. The three approaches differ as to the precision and accuracy of the resulting estimates. The accuracy and precision of the different approaches will be compared under a variety of circumstances including study conditions that are not representative of manufacturing conditions and when some sources of variation are not included in the study.
All three approaches have strengths and weaknesses. No approach can be said to be universally superior. However, these comparisons suggest that in most cases, the tolerance analysis approach is the best approach to use. The major weakness of the tolerance analysis approach is that it only estimates the variation caused by the inputs included in the study. It ignores other sources of variation. This weakness can be overcome using a combination of the tolerance analysis and dual-response approaches.