Barker, Thomas B. (1986, ASQC) Rochester Institute of Technology, Rochester, NY
With the advent of modern, compact, high speed computers, Monte Carlo simulation methods have proven valuable in gaining experience with a complex process. However, a Monte Carlo simulation requires either a fundamental or empirical relationship in the form of an equation if we are to gain the experience that will lead to a properly specified product. Moreover, a Monte Carlo analysis by the nature of its random choice process is restricted in its ability to easily identify and quantify the major contributors to the variability we wish to control.Because of the impediments found in Monte Carlo, a process capability study often avoids any use of such pre-production analysis and must usually be satisifed with "best educated guesses" (BEG) for setting tolerances. If these tolerances are too lax, the end product exhibits wide variation and is of poor quality. If the tolerances are too tight, the unit manufacturing cowst (UMC) skyrockets and profits tumble. In either case, the BEG method causes a monetary loss to society. A quality oriented company will practice disciplines that avoid this loss by delivering on-target, low variation products with low UMC.To accomplish the two-fisted goal of quality at low cost, the product development/manufacturing team needs to know which omponents influence the variation in a process and how much control is necessary to produce a quality product. An efficient way to gain this knowledge is through experimental design methods.Like Monte Carlo (buth with far fewer trials) a designed experiment looks at the components' variations and rolls these variations into the overall process variation. Unlike the Monte Carlo method, a designed experiment can be dissected easily and made to reveal the quality sensitive components and the degree of this sensitivity. By having the qualitative and quantitative nature of the variation we possess a degree of understanding that is powerful enough to help the product development team set rational tolerances that deliver quality products.This direct application of experimental design to a tolerancing problem is an application of Statistical Experimental Design (SED) that has been developed and used in Japan and is one of the techniques that has given that country a quality/cost edge with its products in a world market. By combining this concept with the proven methods of response surface modeling as practiced in the Western World, we are able to bring East and West together utilizing the best parts of both design philosophies to produce quality products at affordable costs. By using the discipline of Statistical Experimental Design, we also perform our engineering tasks in a more professional and quality manner. A quality engineer is an engineer who produces a quality product in a quality manner.