Design of Experiments
Design of experiments (DOE) structurally assesses the relationship of factors affecting a response. DOE can find cause-and-effect factor relationships impacting a process output.
Two-level DOE experimentation sets each factor input as a + or – value; e.g., temperature factor: high-temp=+1, low-temp=-1; operator factor: operator1=+1, operator2=-1.
Two-level DOE designs are efficient in assessing (in a minimal number of trials) the interaction of factors relative to impacting a response; e.g., high-temperature in conjunction with low-humidity yields an undesirable response. The resolution of a DOE describes how well two-factor interactions are aliased with other two-factor interactions and main effects. In a two-factor DOE, 15 two-level factors can be evaluated in only 16 trials; however, two-factor interactions are confounded with main effects – a resolution III experiment.
The DOE explanation in Integrated Enterprise Excellence Volume III provides the details for conducting effective DOEs, including how to assess interactions, even with 15 factors in 16 trials.
Design of Experiments (DOE) can be used within a Design for Six Sigma (DFSS) Define-Measure-Analyze-Design-Verify (DMADV) roadmap in both the Design (for optimization) and Verify (for design testing) phases. In this example, which is taken from a book, the application of a DOE is illustrated for verification test of the product design robustness to customer environmental conditions and future product configurations. The described technique is useful for capturing no trouble found issues before products are shipped.
In Design for Six Sigma (DFSS) and elsewhere, Design of Experiments (DOE) can be used in product design validation tests. In this example, Design of Experiment (DOE) techniques are used to validate a product design change in conjunction with an assessment of currently planned manufacturing tolerances. In this book excerpt, the reader is taken through a very robust sequence for the analysis of any DOE that provides a rapid and repeatable conclusion. This example includes a unique method to use probability plots to assess design limits.
An enhanced methodology is illustrated in the design of experiments transactional example that is available through the White Paper PDF link below.
