Response Surface Design for Correlated Noise Variables

Abstract

Response surface designs for process robustness studies and robust parameter design have historically been chosen and performed under the assumption that noise variables are uncorrelated, but this may not always be a valid assumption. This paper presents a framework for quantifying the effects of correlation among noise variables on the estimates of model parameters. Both numeric and visual assessments of design optimality are employed. Nonstandard designs with superior performance compared to traditional designs are suggested for some example cases in which noise variables are correlated and have a practically significant interaction with control variables.

Department(s)

Mathematics and Statistics

Keywords and Phrases

design of experiments; fraction of design space graph; prediction error variance; slope estimation variance; variance dispersion graph

Document Type

Article - Journal

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2005 Fu Jen Catholic University, All rights reserved.

Publication Date

01 Jan 2005

Link to Full Text

Share

 
COinS