Applications of Plackett and Burman designs to split plot designs

Split plot experiments are very common in industry for process and product improvement. Their applications range from robust product experimentation to experimentation in sequential processes. In this study, alternative split plot designs allowing large number of factors with relatively small number of runs are provided. This is mainly achieved by using Plackett and Burman (PB) designs.;The properties of PB designs and split plot designs with fractional factorials are reviewed first. The results such as switching signs in one column follow-up strategy for PB designs, the relationship between partial confounding and projective properties of PB designs and generalized alias matrix with its application to split plot designs are also introduced.;The applications of PB designs in split plot experiments are presented next. All possible combinations such as PB designs as the sub-plot array, as the whole-plot array and finally as both whole-plot and sub-plot arrays are explored. The strategies such as using half fractions of PB designs to reduce the number of runs and using a PB design and its fold-over (especially when it is used as the sub-Plot array) to clean out the confounding are also introduced.;In most cases using these strategies and with relatively easy to fulfil conditions, one can achieve designs where the main effects and two-factor interactions between whole-plot and sub-plot factors are free from confounding with any other two-factor interactions. This is valid for any n ran PB design. The competing fractional factorial split plot designs do not only lack the desired alias structure for robust product experimentation but they are also not provided in the literature for more than 15 factors. For these designs, increasing the number of factors requires an extensive computer search. On the other hand for the designs presented in this dissertation increasing the number of factors is simply a matter of choice and can be accomplished by fulfilling relatively simple conditions.