Premium
Design strategies for response surface models for the study of supersonic combustion
Author(s) -
Johnson Rachel T.,
Parker Peter A.,
Montgomery Douglas C.,
Cutler Andrew D.,
Danehy Paul M.,
Rhew Ray D.
Publication year - 2009
Publication title -
quality and reliability engineering international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.913
H-Index - 62
eISSN - 1099-1638
pISSN - 0748-8017
DOI - 10.1002/qre.983
Subject(s) - replication (statistics) , sampling (signal processing) , variance (accounting) , supersonic speed , computer science , response surface methodology , combustion , design of experiments , scope (computer science) , field (mathematics) , aerospace engineering , statistics , engineering , mathematics , machine learning , chemistry , accounting , organic chemistry , filter (signal processing) , pure mathematics , business , computer vision , programming language
An application of a classical design approach to an experiment involving the study of supersonic combustion is described in this paper. The case study described is that of an experiment whose objective is to create response surfaces of the mean and variance of several flow parameters as a function of location within a supersonic jet flow field. The approach demonstrated in this paper involves the use of a classic response surface methodology design in a unique manner. Additionally a unique application involving the sub‐sampling and replication strategies is developed in a similar manner to those of robust parameter design. The sub‐sampling and replication techniques allow for the ability to systematically account for the precision in mean and variance models of the output response variables. The final design prescribed met the experimental objectives of the project by creating the ability to fit response surfaces and allowing for the experimenters to understand the relative precision of their estimates based on the final sub‐sampling and replication techniques. Results from one section of the region of interest are used to illustrate two different modeling approaches. The performance of both modeling approaches in prediction of new data is illustrated. The conclusions also include a discussion of the future work that will follow. Copyright © 2008 John Wiley & Sons, Ltd.