Page 201 - Design for Six Sigma a Roadmap for Product Development

174 Chapter Five

The importance of DPs for decreasing sensitivity is determined by
comparing the gain in the metric ratio from level to level for each fac-
tor, comparing relative performance gains between each design para-
meter, and then selecting which ones produce the largest gains. The
level for each DP with the optimized metric ratio is selected as the
parameter’s best target value. All of these best levels will be selected
to produce the best parameter target combination. The same analysis
and selection process is used to determine DPs, which can be best used
to adjust the mean performance. These factors may be the same ones
that have been chosen on the basis of metric improvement, or they may
be factors that do not affect the optimization of the metric. DPs that
don’t contribute to improvements in the metric transfer function are
set to their most economical values.
The DFSS team needs to run confirmation tests of optimum design
combinations and verify assumptions, and perform a test with samples
configured at the combined optimum design level and calculate the
representative metric performance.
Compare the transfer function values to the predicted optimum values.
If the actual performance is within the interval of performance that was
predicted, then the predictive model validity is confirmed. There is a
good chance at this point that the optimum results experienced in the
confirmation run will translate to the usage environment. If the confir-
mation test values fall outside the interval, the team should reassess the
original assumptions for this experiment since, in all likelihood, other
conditions are operating which are not accounted for in the model.
A successful experiment will lead the team to clarify whether new
technical information has been uncovered that will greatly improve
the physical structure. The team will want to consider if other DP lev-
els should now form the basis of a revised experimental plan. If the
study failed to produce an improvement, the combination of noise fac-
tors that were in the original experiment may have overpowered the
ability of the DPs to generate improved performance. A redesign of the
DP strategy should be considered by the team. If improvement cannot
be realized and the team has exhausted all possible DPs, there may be
reason to conclude that the current concept being optimized will not be
able to support the performance requirements for the design under
development. This unfortunate situation usually happens because of
the violation of design axioms and would justify the consideration and
selection of a new concept or even a new physical structure.

5.11 Design for X (DFSS Algorithm Step 10)
The black belt should continually revise the DFSS team membership to
reflect a concurrent approach in which both design and process members