Page 319 - Six Sigma Advanced Tools for Black Belts and Master Black Belts

OTE/SPH
OTE/SPH
August 31, 2006
Char Count= 0
JWBK119-19
304 3:6 Economical Experimentation via ‘Lean Design’
2. It is foreseen that constraints in certain experimental resources, for example mate-
rials, equipment or manpower availability, are such that there will be long delays
if one has to await a complete set of data associated with a regular experimental
design.
3. Some quick answers concerning the subject of study are urgently needed, hence
one cannot afford the time for a full experimental data set.
4. Sometimes it may be deemed important to have information on experimental er-
rors; without increasing the total number of observations, replication can be made
in certain experimental runs in return for absence of response values in others, a
trade-off realizable via lean design.
5. It is envisaged that a certain condition specified in an experimental design matrix,
such as ‘high’(+1) settings for all the factors, may adversely affect a process, create
hazardous conditions, or cause undue inconvenience on the production floor, and
should be avoided: the response corresponding to such a condition may have to be
left out on purpose.
The above would mean that incomplete experimentation based on a lean design
is intentional. In situations where an experiment is originally meant to be run in
full, lean design may be invoked to substitute the original design, as illustrated by
the following.
6. The infeasibility of an experimental condition -- for a reason similar to that in case
5 -- is discovered only during the course of data collection, so arrangements for a
particular response have to be abandoned.
7 The experiment has to be aborted owing to unforeseen circumstances, such as
equipment breakdown, exhaustion of raw materials, or unavailability of personnel,
and the best has to be made of the data already collected.
8. Results of the study are suddenly called for before the completion of experiment,
to be used to help deliberations in an urgent meeting, for instance.
9. After an experiment is completed, it is found that certain data are associated
with faulty parts and materials, or have been wrongly measured or otherwise
corrupted and hence cannot be used, and it is not feasible to re-enact the spoilt
runs.

Regardless of the circumstances in which lean design is used, if there is subse-
quently an opportunity to make up for the absent response values, the additional
data can always be combined with those already available for a fresh analysis of the
experiment. This in fact also provides an opportunity to check on the validity of initial
assumptions concerning the non-existence of certain interaction effects. In this way
an incremental improvement of the experimental results is possible with the avail-
ability of each additional late response value. When this is intentionally done, for
example using L 5 2 4−1 as a start, then including one additional response value at a
time for a re-analysis of the experiment -- moving through L 6 2 4−1 , L 7 2 4−1 and L 8 2 4−1
designs until eight values are finally available -- is in fact execution of an incremental
experimentation scheme.
The use of lean design for incremental experimentation, depicted by the flow chart
in Figure 19.1, would prove useful in situations where every bit of early information
is crucial, for example in research and development for new advanced materials,
or urgent troubleshooting of high-cost processes. Not only does the approach incur
no more resources than called for in a regular experimental design, its self-correction