Page 615 - Design for Six Sigma a Roadmap for Product Development

568 Chapter Fifteen

In many applications, high sensitivity is a desired property. If a tar-
get sensitivity is required, that is, if sensitivity should be neither too
high nor too low, the two-step optimization procedure can be used to
tune the sensitivity to target.

15.4.3 Why are interactions among control
factors not desirable?
Dr. Taguchi stated that the interactions among control factors are not
desirable. In a dynamic robust parameter design study, Dr. Taguchi
proposes the use of L 12 , L 18 , and L 36 , because in those arrays, the inter-
actions will be evenly confounded among all columns. Dr. Taguchi
treats those interaction effects as noises. This statement has drawn a
lot of criticism from statisticians.
However, if we study this issue from an axiomatic design perspec-
tive, we realize that interaction is again a form of complexity. The sys-
tem that has interactions is definitely more complex than one without
interactions. In other words, interaction is not a design choice. Because
Dr. Taguchi uses only an “additive model” among control factors and
S/N based only on linear ideal function, the S/N for the control factor
level combination with severe nonlinearity and nonadditivity will def-
initely be very low and so it will not be selected in parameter design.
Again, robust parameter design is equivalent to selecting a good
design by using dynamic S/N as a benchmark, and penalizing nonad-
ditivity will help in selecting a good design with less complexity.

15.5 Robust Technology Development

Robust technology development means building robustness into newly
developed generic technology, or new technology at its infancy. The
examples of such new generic technology include new memory chips,
new electronic bonding technology, and new materials. New technolo-
gies are usually developed at research laboratories under ideal condi-
tions, in small batches and on a small scale. After a generic new
technology is developed, product developers will try to integrate it into
new products. But usually there are a lot of hiccups in this integration
process; the new technology that works well in the lab may not work
well after integration, and its performance may not be stable and up
to people’s expectations. It usually takes many trials and errors to
“make it work right.”
Robust technology development is a strategy that tries to streamline
and expedite the integration of new technology with products and pro-
duction. Robust technology development proposes conducting robust
parameter design on the new technology when the new technology is