Robustness and Prediction Error in Multi Response Parameter Design Optimization


Özateş Gürbüz M., Köksal G., Köksalan M. M.

2018 INFORMS Annual Meeting, Arizona, United States Of America, 4 - 07 November 2018, pp.241

  • Publication Type: Conference Paper / Summary Text
  • City: Arizona
  • Country: United States Of America
  • Page Numbers: pp.241
  • TED University Affiliated: No

Abstract

Parameter design optimization that involves two or more responses of products or processes is a well-known research problem. This problem involves multiple objectives typically formulated as response surface models using regression. These models are constructed with some prediction error even though the model structure is appropriate. In comparing alternative solutions of the model, the decision maker (DM) has difficulty to comprehend closeness of mean response is to its target, variance of true response (robustness) and magnitude of prediction error. In this study, these components are analyzed under certain conditions and guidance is provided for the DM to facilitate the comparison.