An Experimental Strategy for Fractionating 33 and 34 Factorial Experiments

Abstract

In the design of statistical experiments, situations may arise when resource constraints hinder the use of factorial designs for process improvement. This paper explores how 9, 18 and 27-run orthogonal arrays compare against each other and against a proposed experimental plan referred to as a 'Segmented Fractional Plan' when used to fractionate 3 3 and 3 4 factorial experiments. Based on the analysis of 8 responses from 6 factorial experiments, it was observed that to identify the process setting that produces the desired product quality, with a reduced number of experimental runs, the segmented fractional plan can perform as well or better than some orthogonal arrays thus, providing an option for fractionating 3 3 and 3 4 factorial experiments.

A new family of quality loss functions

“…This type of factorial design was discussed extensively by Ozoemena et al [29]. Box-Behnken design (BBD) [30] or Taguchi's orthogonal array L 16 [31] are also very useful for this type and size of experiment since they provide a good balance between the cost of the experiment (including time) and the required accuracy of the results.…”