Constrained optimization: An effective alternative to Taguchi’s two steps to robust design

A parameter design procedure-applicable to robust design problem involving significant factor interactions, complex dependency between σ(x) and μ(x), and/or multiple response characteristics-is presented. The procedure is empirical as Taguchi’s two-step method is, but unlike Taguchi, it uses constrained optimization; it limits attention to only the truly relevant (feasible) designs that already meet the required target responses.