A search method for the exploration of new regions in robust parameter design

An approach for the exploration of new operating regions of a process in which noise factors are present is proposed and studied. The search method considers the curvature in the variance function, following a dual-response approach to process optimization in which a second order model for the process variance is utilized. The method assumes a "Target is best" objective for the process mean response, and proposes to conduct a second order search that minimizes the variance function while keeping feasibility for the mean target constraint. The method is illustrated in detail using simulated experiments.