Page 295 - Six Sigma Advanced Tools for Black Belts and Master Black Belts

OTE/SPH
OTE/SPH
August 31, 2006
JWBK119-18
Taguchi Methods
280 3:6 Char Count= 0
or speed controller used for a process, or the resistors and capacitors in a circuit) can
be of a lower grade and hence lower cost, since the actual value of x can now be
subject to a variation as large as x around x 2 . (It could also mean that the product

containing this component can be subject to more severe environment stress with no
loss of performance.) In this way the design is optimized either for the most robust
performance for a given cost, or for the least cost with a given acceptable level of
robustness (or variability).
Overall, the problem formulation, experimental design, data analysis, and special
applications of Taguchi methods constitute a package that can fit the major objectives
of what has been termed robust design or quality engineering. In fact there is now
a proliferation of labels applied to studies involving the application of design of
experiments, for engineering objectives: Taguchi methods, off-line quality control,
parameter design, quality engineering, robust design, classical design of experiments,
and even robustization; 44 depending on the user, each label may or may not include
a particular idea or procedure. (As an illustration of the diversity of usage of terms,
it is noted that although ‘classical design of experiments’ is commonly taken to mean
designs after Fisher and Box, there are also some who have used it in formal writing
45
to label the ineffective ’one variable at a time’ designs. )

18.5.5 Illustrative examples
Some simple examples will illustrate how the popular version of Taguchi’s parame-
ter design procedure, as often found in commercial courses and computer software
packages, can easily be made appealing to the uninitiated. The frailty of the ‘power’of
parameter design, however, can be just as readily demonstrated; unfortunately that
is rarely seen in such courses and packages.
Suppose the black box P represents a bonding mechanism with response y, the
yield of manufactured parts with acceptable bond strength, and x 1 , x 2 , x 3 are design
parameters denoting bonding material hold time, and contact speed, respectively. The
objective of the study is to search for parameter settings to maximize the response.
3
According to Taguchi’s standard orthogonal array L 4 (2 ), an experiment can be carried
out as shown in Table 18.2, where the choice of material x 1 is between L and H, and
x 2 and x 3 are tried out at the extreme values L (for low) and H (for high) of their
respective acceptable operating ranges.
With experimental results as shown in last column of Table 18.2, an analysis can be
made as follows. The average of all y-values is 70 %. When x 1 is at L, the average y

Table 18.2 Illustration of a typical ‘parameter design’routine.
Material Time Speed Yield
x 1 x 2 x 3 y%
L L L 72
L H H 82
H L H 56
H H L 70