Designs of Mixed Resolution for Process Robustness Studies

Abstract

A robust process is a process that is insensitive to changes in uncontrollable variables. In this article a class of designs that can be used to achieve a robust process is proposed. These new designs are similar in structure to classical central composite designs, but they are of mixed resolution. That is, the new designs are at least Resolution V among the signal factors and are at least Resolution III among the noise factors. A catalog of the new designs, known as composite mixed-resolution (CMR) designs, is included for the practitioner. A comparison of the sizes of robustness designs shows that many CMR designs are superior to or competitive with the corresponding Taguchi designs. The response surface models associated with these two classes of robustness designs are also compared. D efficiencies and G efficiencies of the CMR designs are included.