Engineers at an aluminum-casting company were struggling to understand why a particular part came off the line filled with inclusions. Having conducted many one-factor-at-a-time tests to no avail, they turned to statistical software and a process called design of experiments. Optimizing based on this process let the engineers reduce the defect rate to zero.
See how to apply statistically based design of experiments (DOE) for mixtures - a proven method for making breakthrough improvements in cost and performance. Ultimately you may discover a sweet spot where all your customer specifications can be satisfied. To illustrate the method, this article lays out a case study on the formulation of rheology modifiers. (A somewhat different version of this article appeared in Modern Paint and Coatings.)
In many rubber and plastics processes, powerful interactions affect final performance. These remain undiscovered via traditional one-factor-at-a-time scientific methods. Multifactor design of experiments (DOE) reveals these interactions that lead to breakthrough improvements in process efficiency and product quality. The big gains come from a very simple form of DOE called two-level factorial design. This approach to experimentation has proven to be especially helpful for control of part shrinkage as demonstrated in a case study. However, it can be applied to any measurable response in rubber and plastics production. This primer provides the essential details on two-level factorial DOE from an engineering perspective with an emphasis on the practical aspects.
What would you do it confronted with an "opportunity" to make a major change, involving many factors, but you need to do it quickly? The traditional approach to experimentation requires you to change only one factor at a time (OFAT). However, the OFAT approach doesn’t provide data on interactions of factors, a likely occurrence with chemical processes. An alternative approach called “two-level factorial design” can uncover critical interactions. This statistically based method involves simultaneous adjustment of experimental factors at only two levels, offering a parallel testing scheme that’s much more efficient than the serial approach of OFAT.
This presentation details and demonstrates a procedure that, despite missing data, allows the use of user-friendly, normal-probability plots for two-level-factorial effect selection.
A look at augmenting the usual probability plot effects with points representing pure error.