RESPONSE-SURFACE METHODS AND THE USE OF NOISE VARIABLES

The ideas of robust parameter design focus on reducing product or proc ess variability that is transmitted by noise variables. These variable s are difficult to control in the process or are not constant across d ifferent levels of consumer usage. In this paper we develop and illust rate the use of response surface methods that are extended to cover mo deling of the process mean and variance.Considerable attention has bee n placed on the response surface for the process variance. A methodolo gy is given that allows for a confidence region on the location on the control factors of minimum process variance. This is the location whe re the process variance is no larger than the experimental error varia nce. The mean and variance response surfaces can also be combined to p roduce prediction limits on a future response and one-sided tolerance intervals.