Page 200 - Design for Six Sigma a Roadmap for Product Development

Design for Six Sigma Project Algorithm 173

Data from the optimization experiment will be used to generate the
transfer function to be used for optimization and to improve design
robustness to the Six Sigma level. The validity of this function and the
resulting conclusions will be influenced by the experimental and sta-
tistical assumptions made by the DFSS team. What assumptions can
be made regarding the existence of interactions between design para-
meters? Is there an assumption that the variance of the response
remains constant for all levels within the transfer function, or is it
assumed that the variance in the response is due to the effect of the
noise factors? What assumptions can be made (if any) regarding the
underlying distribution of the experimental data? What assumptions
can be made regarding the effect of nuisance factors (other than the
experimental noise factors) on the variance of the response? Are the
optimum factor combinations predicted by the transfer function
optimal? What assumptions can be made about the transferability of
the results beyond the experimental environment, and what would sub-
stantiate these assumptions?
In the dynamic robustness formulation, the DFSS team should
decide on noise strategy and signal range, and develop a design para-
meter strategy. Design parameters are specified freely by the team. If
the experiment is exploratory, it is suggested to set levels at extreme
values of the feasible operating range. Two levels will be appropriate
for screening purposes, but more information on nonlinear effects will
require a three-level strategy. The quality metrics that will be used in
the analysis are FRs, loss functions of the FRs, and the signal-to-noise
ratio. Depending on the FR, there are two broad forms of ratio are
available. Static forms apply where the FR has a fixed value. Dynamic
forms apply where the FR operates over a range of signal input values.

5.10.5 Collect data and analyze results
(DFSS algorithm step 9)
The individual values of the appropriate metric are calculated using
the data from each experimental run. The purpose of determining the
metric is to characterize the ability of DPs to reduce the variability of
the FRs over a specified dynamic range. In a dynamic experiment, the
individual values for transfer function sensitivity coefficients are cal-
culated using the same data from each experimental run. The purpose
of determining the sensitivity values is to characterize the ability of
design parameters (DPs) to change the average value of the FRs across
a specified dynamic range. The resulting sensitivity performance of a
system is obtained by the best-fit curve.
DP level effects are calculated by averaging the metric to correspond
to the individual levels as depicted by the orthogonal array diagram.