| The goal of any optimization experiment is to find the settings of the factors that can be controlled which results in optimal levels of the responses of interest. In robust parameter design, the two responses of interest are the mean and variance of a quality characteristic. In multiple response optimization, the responses of interest are the quality characteristics of the product. In both of these cases, a quantity that is a function of the estimates of the responses of interest is either maximized or minimized. A variety of quantities have been proposed for robust parameter design and multiple response optimization, but all of the proposed quantities are lacking in some respect-they may lack intuitive appeal, depend too heavily on the definition of subjective parameters, or fail altogether in certain situations. In addition, most of the quantities proposed for robust parameter design cannot be adapted easily to multiple response optimization. The probability that all of the responses are simultaneously within their upper and lower specification limits is a quantity which can be used for robust parameter design and multiple response optimization. The probability method also has an intuitive appeal that will make it easy to explain to people in fields outside of statistics. This method does not depend on the definition of subjective parameters, and it works in all of the situations that have been addressed. It may also be extended to multiple response robust parameter design, which none of the other methods has attempted. |
