|
After World War II, W. Edwards
Deming, the father of the American quality movement, was sent to
Japan by the United States government to help rebuild Japan’s
destroyed industrial base. While there, Deming befriended Genichi
Taguchi, a young Japanese electrical engineer. Deming trained Taguchi
in statistical process control and factorial design, which are primary
topics of Statistical Experiments. The first industrial
test of the statistical process known as Design of Experiments (DOE)
soon followed.
INA Tile, a Japanese manufacturer of bathroom tile, had built a
brand new kiln at a cost of $500,000. Unfortunately the device had
an uneven heating pattern, and all fired tiles broke and crumbled
before they could be shipped to market. Taguchi suggested that instead
of rebuilding the kiln—an economically unfeasible proposition—they
should make an improved slurry mixture (tile paste) that could withstand
uneven heating without breaking.
Taguchi identified eight active ingredients used in the slurry mixture.
Each ingredient could be added to the mixture like salt and pepper,
in varying strengths. Using manufacturing records, he chose two
representative strengths for each ingredient that he felt would
keep the fired tiles within overall performance specifications.
The goal was to find the optimal combination of all eight strength
levels leading to durable tile. The INA officials panicked, believing
they could not afford to test all possible combinations of 256 batches
of slurry. Taguchi conducted a successful test program, however,
using a fractional factorial design consisting of 16 individual
trials. INA was saved, and a new era of industrial dominance began
in Japan, initiated by Deming, and taken to its full potential by
Taguchi.
In the 1970s, American industries were surprised when Japanese manufactured
goods were suddenly found superior to their American counterparts.
The superiority
of the Japanese goods was a result—at least in part—of
Deming’s passion for statistical experiments. Now, these experimental
design techniques are used worldwide wherever quality goods are
made, including the United States. In fact, anyone who expects to
have a place in modern American industry must be proficient in the
statistical methodologies first put forth by
W. Edwards Deming.
Statistical Experiments provides activities and real problems
to be solved by analyzing data and making appropriate changes in
a process to improve it. You learn to extract information using
minimal data to solve employers’ problems with maximum savings
of cost
and time.
Design of Experiments (DOE) is a very extensive and broad field
that has split into several different schools of application and
interpretation. Among them are classical DOE concepts, ideas of
Shannin, of Taguchi, and hybrid factorials. Many of the techniques
from the various schools overlap in some areas. The schools tend
to emphasize or de-emphasize specific DOE areas. Some focus on highly
customized DOE techniques that expedite the test process for select
cases.
Statistical Experiments is a foundational module that neither
emphasizes nor de-emphasizes any particular school of thought in
Design of Experiments (DOE). It does, however, provide enough foundation
to make it possible for you to further pursue more specialized DOE
training in any one of the predominant schools today.
After completing this module, you should be able to demonstrate
the following competencies:
- Use sound sampling techniques and data analysis to make repeatable,
defensible inferences (Comp. 1).
- Solve real world problems by designing and conducting factorial
and fractional factorial experiments (Comp. 2).
- Use analysis of variance (ANOVA) to interpret experimental results
(Comp. 3).
- Graphically display outcomes using regression analysis to communicate
the recommended action based on the experiment (Comp. 4).
Table of Contents
|
