Bailey, Steven P.; Chatto, Kenneth A.; Fellner, William H.; Pfeifer, Charles G. (1991, ASQC) Du Pont Company, Newark, DE
This paper demonstrates how the Du Pont Company has used statistical design of experiments (DOE) methodologies to minimize the effects of uncontrolled sources of variation during manufacture and include consumption (robustness) in product design.
The methods used in this process include classical methods, such as response surface analysis, which incorporates analysis techniques supplementary to standard regression output displays. The steps involved in exercising the response surface equation include: (1) Examine coefficients; (2) Plot results; (3) Specify D-Factor tolerances; (4) Calculate signal and noise; (5) Plot signal and noise; (6) Specify optimization criteria/constraints; (7) Choose a candidate product design; (8) Plot contours; and (9) Do sensitivity analysis at candidate.
To illustrate the approach, the authors use a circuit design example discussed by Taguchi (1988). After extracting information from the response surface equation, the practitioner can use this flexible, iterative process analysis to change assumptions and criteria until the desired levels of comfort and confidence are reached.
Du Pont Company,Product design,Response surfaces,Statistics,Taguchi, Genichi,Design of experiments (DOE)