Streamlining robust parameter design efforts

Abstract

This paper proposes and evaluates techniques to streamline robust parameter design, that is, to make products robust using fewer experimental runs. The evaluations are conducted using a combination of hierarchical probability models and six case studies. A procedure is presented for choosing among methods to balance the needs for speed and accuracy. Simulations of robust parameter design being deployed in this way suggest that the proposed approach gives good results and reduces sensitivity of the design process to the fidelity of a priori judgments of which subsystems need robustness improvement.