Blocked Fractional Factorial Split-Plot Experiments for Robust Parameter Design

Abstract

Fractional factorial experiments are commonly used for robust parameter design and, for ease of use, such experiments are often run as split-plot designs. If the control factors are at the subplot level and the noise factors are at the whole-plot level, this also results in gains in efficiency. If all runs of the fractional factorial split-plot design cannot be run under homogeneous conditions, such designs are frequently blocked. In this paper, we explore the choice of blocked fractional factorial split-plot designs for use in robust parameter design. A ranking scheme for such designs is developed and, using a search algorithm, a catalog of 32-run optimal designs is provided. Two situations are considered, one in which the control factors are at the subplot level and one in which the control factors are at the whole-plot level. An example from the aerospace sector is used to illustrate the concepts.

• Blocking
• Control factors
• Noise factors
• Optimal designs
• Restricted randomization

Additional information

Notes on contributors

Robert G. McLeod

Dr. McLeod is an Assistant Professor in the Department of Mathematics and Statistics. His email address is [email protected].

John F. Brewster

Dr. Brewster is a Professor in the Department of Statistics. His email address is [email protected].