Abstract
The ideas of robust parameter design focus on reducing product or process variability that is transmitted by noise variables. These variables are difficult to control in the process or are not constant across different levels of consumer usage. In this paper we develop and illustrate the use of response surface methods that are extended to cover modeling of the process mean and variance.
Considerable attention has been placed on the response surface for the process variance. A methodology is given that allows for a confidence region on the location on the control factors of minimum process variance. This is the location where the process variance is no larger than the experimental error variance. The mean and variance response surfaces can also be combined to produce prediction limits on a future response and one-sided tolerance intervals.
Additional information
Notes on contributors
Raymond H. Myers
Dr. Myers is Professor Emeritus in the Department of Statistics.
Yoon Kim
Dr. Kim is Assistant Professor in the Department of Mathematics and Statistics.
Kristi L. Griffiths
Dr. Griffiths is a Senior Statistician in Statistical and Mathematical Sciences.