Abstract
Multi-stage processes are very common in both process and manufacturing industries. In this article we present a methodology for designing experiments for multi-stage processes. Typically in these situations the design is expected to involve many factors from different stages. To minimize the required number of experimental runs, we suggest using mirror image pairs of experiments at each stage following the first. As the design criterion, we consider their projectivity and mainly focus on projectivity P > 3 designs. We provide the methodology for generating these designs for processes with any number of stages and also show how to identify and estimate the effects. Both regular and non-regular designs are considered as base designs in generating the overall design.
Additional information
Notes on contributors
John Tyssedal
John Tyssedal is an Associate Professor at the Department of Mathematical Sciences at the Norwegian University of Science and Technology. He serves as an Associate Editor for QTQM. His research interests are in design of experiments and time series analysis.
Murat Kulahci
He is an Associate Professor in the Department of Applied Mathematics and Computer Science at the Technical University of Denmark and in the Department of Business Administration, Technology and Social Sciences at Lulea University of Technology in Sweden. His research focuses on design of physical and computer experiments, statistical process control, time series analysis and forecasting, and financial engineering. He has presented his work in international conferences, and published over 50 articles in archival journals. He is the co-author of two books on time series analysis and forecasting.