Page 288 - Six Sigma Advanced Tools for Black Belts and Master Black Belts

OTE/SPH
OTE/SPH
Char Count= 0
August 31, 2006
3:6
JWBK119-18
Taguchi’s ‘Statistical Engineering’ 273
individual, isolated ‘customer’(which could range from an actual human being to ‘the
next process’ in a manufacturing system) at any one time, but is regarded as some-
thing that must be made so pervasive -- much like working telephones and electric
lighting in a modern office -- that its existence is felt only through its absence. Once
quality is not achieved, every party in an interlocking relationship suffers: suppliers,
manufacturers, customers and so on, an entire chain of elements in society. Indeed,
‘loss to society’reflects such a lofty ideal that it transcends all lesser visions of quality,
and remains the only major thought of Taguchi’s that has remained largely unscathed
in assaults on Taguchi methods through the years.
‘Loss to society’ is next measured through the deviation of actual product quality
from its target value. Use is made of a ‘loss function’, a mathematical expression that
can state, particularly for management purposes, the monetary value of the conse-
quence of any shortfall (or improvement) in quality. Although such a monetary value
more often than not represents only a virtual loss or gain, it is a convenient perfor-
mance index that can be readily appreciated by decision makers -- a very important
feature played up by Taguchi methods promoters to draw the interest of managers.
This is in sharp contrast to the presentations of statisticians which traditionally cannot
be divorced from the language of mathematical formalism and abstraction: hypothe-
ses, alpha and beta errors, white noise, OC curves and such like.
To management and technical personnel, the Taguchi’s approach to quality has an-
other obvious appeal that is not seen in the established SQC literature. This lies in the
way Taguchi admonishes them to leapfrog the earlier stages of statistical applications
shown in Table 18.1 and strive to improve quality at the out set during product and
process design and development, making use of design of experiments for what he
3
calls ‘off-line quality control’. Thus ‘Quality must be designed into the product’ is
not a mere slogan; Taguchi wants to have it actually carried out to provide a funda-
mental solution to possible quality problems. Basically this is to be achieved through
the understanding and exploitation of product and process behavior as expressed
in equation (18.1), which is a motivation shared by both Taguchi and mainstream
statisticians, but is also where Taguchi methods begin to depart from statisticians’
formulation of the physical system in question. The next section will highlight some
technical ideas that have helped shape the uniqueness of Taguchi methods, yet have
given rise to many of the controversies one has seen.

18.5 TAGUCHI’S ‘STATISTICAL ENGINEERING’

As Taguchi’s techniques for product and process optimization have already been well
9
documented and interpreted, an account of procedural matters will not be necessary
here.Itmaybefairlystatedthatthesetechniquesareinvariablymotivatedbyengineer-
ing considerations, using statistical tools to serve engineering purposes -- in contrast
to statisticians’ techniques that are primarily mathematically formulated, then subse-
quently exploited for their potential to solve engineering problems. Thus Taguchi’s ap-
proach may be regarded as ‘statistical engineering’rather than ‘engineering statistics’,
the latter being a term traditionally used by statisticians to cover mathematical tech-
niques extracted for learning and hopefully application by engineers. The main sta-
tistical contents of Taguchi methods may be appreciated from the following: problem