Page 287 - Design for Six Sigma a Roadmap for Product Development

Axiomatic Design 257

4. After fixing the values of DP 1 and DP 3 , use DP 5 to adjust and control
FR 5 .
5. Use DP 2 only to adjust and control FR 2 .
6. After fixing DP 2 , use DP 4 to adjust FR 4 .

{ } [ A 11 0 0 0 0 0 ][ ] (8.11)
FR 1
DP 1
A 22 0
0
0
0
0
DP 2
FR 2

A 33 0
A 31 0
0
DP 3
FR 3
A 36
A 44 0
A 42 0
0
0
FR 4
DP 4
A 51 0
FR 5
0 0 A 53 0 A 55 0 DP 5
0
0
0
FR 6 6
1 A 66 6
6 DP 6 6
1
where A is a nonzero sensitivity entry. Note that steps 1, 2, and 5 above
can be performed simultaneously. This is not the case for the remain-
ing rules. The greatest potential here to reduce the work at the opti-
mization step is to utilize this step as a first perspective of the ideal
functions definitions. Definition of the ideal function for each parti-
tioned function is central to selecting and understanding good struc-
tural (solution entity) choices, as well as to setting the foundation for
the rest of the DFSS algorithm. As activities progress through the
development process, the definition of the ideal function will be clari-
fied, refined, and finalized. The activities of analysis and synthesis
require iterating rapidly with multiple forward and backward motion
between design domains and within these domains. Iterative looping
forward and backward is good practice in structure creation, as it is in
most engineering activities.
A suggested sequence to be followed in this example with respect to the
rest of the FR optimization studies is to use DP 1 only to optimize FR 1 , use
DP 3 only to optimize FR 3 , use DP 4 only to optimize FR 4 , and use DP 5 only
to optimize FR 5 . Any lower-level design parameters of DP 1 should not
appear in the optimization of FR 3 , for example. The selection of the DPs
for optimization to avoid coupling, eliminate the interaction between con-
trol factors, and enforce an additive transfer function model.
8.5.3 Decoupling of coupled design
(DFSS algorithm step 6)
There are many ways to decouple a design depending on the situation:
1. Make the size of array y equal to the size of array x: m p.
According to Theorem 2 in Suh (1990, p. 68), when a design is cou-
pled because the number of FRs is greater than the number of the
design parameters, it may be decoupled by the addition of new