Using Augment Design to Resolve a DOE Blocking Mistake
DOE plays an important role in AMAT End-End continuous process, design improvement and optimization. Unfortunately, most DOE practitioners may randomly pick any DOE design with JMP that they are familiar with while lacking a DOE statistics foundation, resulting in poor predictive models. As one of the most powerful Resolution IV algorithms, DSD allows us to study both main and interaction effects of a large number of predictors in a relatively small DOE run size. However, DSD cannot tolerate any Orthogonality Violations such as GRR Noise, SPC Time Noise, Design Constraint or Recursive Stepwise Algorithm.
We studied a special case regarding DSD Blocking Design. Instead of assigning the two operation systems as Predictor factor, Blocking factor has been assigned. The DOE experimental operators did not pay full attention and changed the run order, not following the Blocking plan. After completing the first half of nine runs (1st Blocking), the design process owner found Blocking mistakes and stopped the DOE runs immediately, which induced poor design evaluation. After discussion, we found five alternative resolutions to improve the Orthogonal design structure. We conducted detailed Design Diagnostics on each alternative DOE proposal considering the DOE schedule and cost constraints. Among the five proposals, Augment DOE, adding nine new Augment points (all at corners) is the best to recover the most Orthogonal risks at the highest Return of Investment Ratio. Through this DOE Blocking case study, we have further upgraded our JMP DOE knowledge with effective communication through this crisis.