Page 264 - Six Sigma Demystified

244 Six SigMa DemystifieD

• The change in the mean, also known as curvature, refers to the difference
between the edge points of the 2 experiment and their center point. If
k
the process is centered on the current optimum, there should be a sta-
tistically significant difference between the center point and the edge
points.
• Juran suggests choosing two or three important factors to keep the EVOP
manageable. Choose levels for each factor as “small steps” to avoid large
changes in quality or operating conditions, and center the first experiment
at the current “best operating condition” for the process.

3
2
• Run a 2 (for two factors) or 2 (for three factors) experiment with a
center point. Repeat the experiment, and after the second cycle, begin to
estimate the error and the significance of effects.
• Continue with this experiment for a third cycle (i.e., third replicate), and
if a factor is significant after this third cycle, then begin phase 2 with a
new experiment centered on the new “best condition.”
• When factors are not calculated as statistically significant, consider in-
creasing the range of the levels for these factors because it is possible that
the levels were too similar to detect a statistical difference. Alternatively,
consider replacing the insignificant factors with new factors that currently
may be contributing to error.
• If no factor is determined to be significant after eight cycles, then either
change the factor-level ranges or select new factors.
• When the optimal condition has been reached, run additional experiments
with new factors or new factor-level ranges to verify the optimal condition.

Consider this example chemical process to maximize yield using two process
factors—temperature and reaction time. The current process setting for tempera-
ture is 150ºC. Levels are chosen at 145 and 155ºC. The current process setting
for reaction time is 30 minutes. Levels are chosen at 28 and 32 minutes.
The measured responses are shown in Figure F.10 for cycle 1 and Figure F.11
for cycle 2 at each experimental condition. The number in parentheses refers
to the order of the trial. For example, the first data point of cycle 1 was run at
a temperature of 150ºC and reaction time of 30 minutes, with a resulting yield
of 74 percent.
Note: See “Factorial Designs” for information on defining and analyzing the
experimental runs in Minitab and MS Excel.