A comparison of the response model/combined array method and the traditional methods for robust design

Abstract

Genichi Taguchi alerted statisticians to the importance of considering variation as well as target when designing experiments. The Box-Taguchi method ( a variation of his classical method) has placed little emphasis on C-N interactions. In this study, an attempt is made to understand the C-N interactions. To achieve this, simulated examples are used to demonstrate a modern day technique called Response Model/Combined Array. Also, Taguchi's designs call for a complete cross between a (possibly) fractionated so-called control array and a (possibly) fractionated so-called noise array. It is sought to explore the possibility of cutting down on the total number of experimental runs by fractionating a design consisting of control and noise factors put together in one design. The conclusion drawn is that the Response Model/Combined Array method offers a better possibility of saving in the total number of experimental runs than the Box-Taguchi method.