Page 385 - Six Sigma Demystified

Part 3 S i x S i g m a To o l S 365

= 1: Equal importance on target and bound (linear).
>1 (max = 10): High emphasis on target; must be very close to
target for desirability.

Response Surface Designs

Note: This example is a continuation of the example presented under “Factorial
Designs” above.

Collect data along the path of steepest ascent:

B E Response Conclusion
0 0 8,669.2 Centerpoint establishes initial response.
10 –8.8 –986.6 Response decreased; verify with replicate.
10 –8.8 –985.0 Verified response decreases in that
direction; try opposite direction.
–10 8.8 7,810 Response decreases in this direction;
confirm local maximum at origin.
0 0 8,668.8 Verified local maximum near origin.

A first-order experiment then is constructed and data collected, with the cen-
terpoint at a local maximum (0, 0), using the factor levels defined by the path
of steepest ascent.

Minitab

Use Stat\DOE\Response Surface\Create Response Surface Design. Depending on
number of factors, select either Central Composite or Box-Behnken. Generally,
response surface designs are used only after the number of factors has been
reduced using screening designs. One or more factors must have three or more
levels to generate a response surface design.
Select the “Designs” button to select a design. Usually the default values for
centerpoints and α are appropriate, as discussed earlier.
Select the “Factors” button to specify Numeric (i.e., quantitative) or Text (i.e.,
qualitative) and real experimental value for each factor level,

Use the “Options” button to randomize experimental trials.
(Continued)