Taguchi's Ina Seito Tile Experiment
Taguchi's Ina Seito Tile Experiment
Taguchi's Ina Seito Tile Experiment
In 1953, Ina Seito, a Japanese tile manufacturer, faced a serious quality problem: The size of too
many tiles did not conform to the tolerances. About one third of the tiles failed to meet the
requirements to be considered top grade. The production process includes five steps. In the first
step, materials are apportioned, pulverized and mixed to form clay. In the second stage, the clay
is molded into shape. The third stage is pre-firing in a tunnel kiln. The fourth stage is glazing,
and the last stage is the final firing. The quality problems occurred in the third stage, pre-firing,
and were the result of temperature differences at various locations in the kiln. Clay, when
subjected to different temperatures, shrinks at different rates, and thus some of the tiles went out
of tolerance. The first solution engineers came up with involved an investment of about $500,000
in an improved kiln with more consistent temperature across the section. This solution was too
expensive to be seriously considered. (In those years the dollar had a much higher buying power
than today, and the Japanese were short of foreign currency funds. For example, at the same
period Akio Morita of Sony acquired the rights to produce transistors for $25,000, and he had a
very tough time convincing MITI -- the Japanese ministry in charge of industrial development --
to allocate the scarce currency for this purpose!).
Because it was out of the question to solve the problem by reducing the temperature variation of
the kiln, Ina Seito decided to try Taguchi's recommendation and find out if it would be possible
to change the clay mixture or the operating conditions in such a way that the percent defective
would be reduced. To this end what's desired is not necessarily clay that does not shrink in the
kiln, but clay that shrinks consistently. Consistent shrinkage can easily be compensated for by
making the mold larger than the desired finished dimension, but inconsistent shrinkage will
invariably lead to out-of-tolerance tiles.
Ina Seito devised an experiment where they checked seven factors by eight experiments (a
questionable technical practice known as a highly saturated design). One factor addressed the
lime content of the clay mixture; five other factors addressed the type, texture, or quantity of
other additives; one factor measured the size of each batch, with a view to find out if it would be
possible to increase the production run without paying for it in terms of conformance. The
experiment suggested that by increasing the lime content, decreasing the percentage of recycled
material, and using a more expensive agalmotolite mixture (instead of replacing it by a cheap
alternative) the percent defective could be reduced by almost two thirds. In practice, the changes
suggested by the experiment proved themselves, and there was no need to invest a fortune in the
kiln. As an added benefit, lime, whose content had to be increased, is one of the cheapest
ingredients, so the new mixture was cheaper to produce. But the experiment suggested that the
production quantity should not be increased, in spite of the fact that it would save costs to do so
had it not caused more defects.
The Ina Seito experiment is a classic Taguchi achievement in three senses: (i) it is well
publicized; (ii) it achieved great benefits at low costs; (iii) the technical design of the experiment
was questionable (because of the highly saturated design without follow up with a less saturated
design), and it is highly likely that further improvement opportunities were missed or that some
results were not assessed correctly.
In generic terms, the temperature variation that caused the problems in the tile factory is referred
to as "noise," and the objective of robust process or product design is to be unsensitive (i.e.,
robust) to such noise.
35.3. Orthogonal Array L18 589
❒ Example [1]
The molded tiles are stacked in the carts and move slowly in a tunnel kiln as
burners fire the tiles. The newly constructed kiln did not produce tiles with uniform
dimensions. More than 40% of the outside tiles were out of specification. The inside
tiles barely met the specification. It was obvious to the engineers at the plant that
one of the causes of dimensional variation was the uneven temperature distribution.
The traditional way of improving quality is to remove the cause of variation. In
this case it would mean redesigning the tunnel kiln to make temperature distribution
more uniform, which was impossible because it was too costly. Instead of removing
the cause of variation, the engineers decided to perform an experiment to find the
formula for the tile materials that would produce consistently uniform tiles, regard-
less of their positions within the kiln.
In the study, control factors were assigned to orthogonal arrays, and seven po-
sitions in the kiln were assigned to the outer array as noise factors. The interactions
between each control factor and the noise factors were studied to find the optimum
condition. Today, the SN ratio is used to substitute for the tedious calculation to
study the interactions between control and noise factors.
Following are the control factors and their levels:
D: type of agalmatolite
D1 ⫽ ⫺0.0%
D2 ⫽ 1.0% (current)
D3 ⫽
E: amount charged
E1 ⫽ smaller
E2 ⫽ current
E3 ⫽ larger
F: amount of returned material
F1 ⫽ less
F2 ⫽ medium (current)
F3 ⫽ more
G: amount of feldspar
G1 ⫽ 7%
G2 ⫽ 4% (current)
G3 ⫽ 0%
H: clay type
H1 ⫽ only K-type
H2 ⫽ half-and-half (current)
H3 ⫽ only G-type
These were assigned to an orthogonal array L18 as the inner array, and seven po-
sitions in the kiln were assigned to the outer array. Dimensions of tiles for each
combination between a control factor and the noise factor are shown in Table 35.2.
Data Analysis
As a quality characteristic, a nominal-the-best SN ratio is used for analysis. (For
the SN ratio, see later chapters.) For experiment 1:
Se 0.0454
Ve ⫽ ⫽ ⫽ 0.00757 (35.4)
7⫺1 6
(1/n)(Sm ⫺ Ve)
⫽ 10 log
Ve
7(714.8782 ⫺ 0.00757)
⫽ 10 log
0.00757
⫽ 41.31 dB (35.5)
The SN ratios of the other 17 runs are calculated similarly, as shown in Table 35.2.
35.3. Orthogonal Array L18 591
Table 35.2
Layout and results
A B C D E F G H
L18 1 2 3 4 5 6 7 8 P1 P2 P3 P4 P5 P6 P7 Mean SN
1 1 1 1 1 1 1 1 1 10.18 10.18 10.12 10.06 10.02 9.98 10.20 10.11 41.31
2 1 1 2 2 2 2 2 2 10.03 10.01 9.98 9.96 9.91 9.89 10.12 9.99 42.19
3 1 1 3 3 3 3 3 3 9.81 9.78 9.74 9.74 9.71 9.68 9.87 9.76 43.65
4 1 2 1 1 2 2 3 3 10.09 10.08 10.07 9.99 9.92 9.88 10.14 10.02 40.36
5 1 2 2 2 3 3 1 1 10.06 10.05 10.05 9.89 9.85 9.78 10.12 9.97 37.74
6 1 2 3 3 1 1 2 2 10.20 10.19 10.18 10.17 10.14 10.13 10.22 10.18 50.03
7 1 3 1 2 1 3 2 3 9.91 9.88 9.88 9.84 9.82 9.80 9.93 9.87 46.34
8 1 3 2 3 2 1 3 1 10.32 10.28 10.25 10.20 10.18 10.18 10.36 10.25 43.21
9 1 3 3 1 3 2 1 2 10.04 10.02 10.01 9.98 9.95 9.89 10.11 10.00 43.13
10 2 1 1 3 3 2 2 1 10.00 9.98 9.93 9.80 9.77 9.70 10.15 9.90 35.99
11 2 1 2 1 1 3 3 2 9.97 9.97 9.91 9.88 9.87 9.85 10.05 9.93 42.88
12 2 1 3 2 2 1 1 3 10.06 9.94 9.90 9.88 9.80 9.72 10.12 9.92 37.05
13 2 2 1 2 3 1 3 2 10.15 10.08 10.04 9.98 9.91 9.90 10.22 10.04 38.46
14 2 2 2 3 1 2 1 3 9.91 9.87 9.86 9.87 9.85 9.80 10.02 9.88 43.15
15 2 2 3 1 2 3 2 1 10.02 10.00 9.95 9.92 9.78 9.71 10.06 9.92 37.70
16 2 3 1 3 2 3 1 2 10.08 10.00 9.99 9.95 9.92 9.85 10.14 9.99 40.23
17 2 3 2 1 3 1 2 3 10.07 10.02 9.89 9.89 9.85 9.76 10.19 9.95 36.60
18 2 3 3 2 1 2 3 1 10.10 10.08 10.05 9.99 9.97 9.95 10.12 10.04 43.48
Tables 35.3 and 35.4 show the response tables for the SN ratio and average.
Figure 35.2 shows the response graphs of SN the ratio and average. For example:
41.31 ⫹ 42.19 ⫹ 䡠䡠䡠 ⫹ 43.13
average SN ratio for A1 ⫽
9
⫽ 43.10 (35.6)
35.99 ⫹ 42.88 ⫹ 䡠䡠䡠 ⫹ 43.48
average SN ratio for A2 ⫽
9
⫽ 39.50 (35.7)
41.31 ⫹ 42.19 ⫹ 䡠䡠䡠 ⫹ 37.05
average SN ratio for B1 ⫽
6
⫽ 40.51 (35.8)
592 35. Introduction to Orthogonal Arrays
Table 35.3
Response table: SN ratio
Level A B C D E F G H
1 43.10 0.51 40.45 40.33 44.53 41.11 40.44 39.90
2 39.50 41.24 40.96 40.88 40.12 41.38 41.47 42.82
3 42.16 42.51 42.71 39.26 41.42 42.00 41.19
⌬ 3.60 1.65 2.06 2.38 5.27 0.31 1.57 2.92
Ranking 2 6 5 4 1 8 7 3
Optimization
The first thing in parameter design is to reduce variability or maximize the SN ratio.
Select the optimum combination from the SN ratio response table as
A1B3C3D3E1F3G3H2. Factors A, C, D, E, and H had a relatively strong impact on
variability, whereas B, F, and G had a relatively weak impact. Therefore, A1, C3,
E1, and H2 are definitely selected.
The second step is to adjust the mean. Adjusting the mean is easy by adjusting
the dimension of the mold. In general, look for a control factor with a large impact
on the mean and a minimum impact on variability.
Table 35.4
Response table: Mean
Level A B C D E F G H
1 10.02 9.93 9.99 9.99 10.00 10.07 9.98 10.03
2 9.95 10.00 10.00 9.97 10.02 9.97 9.97 10.02
3 10.02 9.97 9.99 9.94 9.91 10.01 9.90
⌬ 0.06 0.08 0.03 0.02 0.08 0.17 0.04 0.13
Ranking 5 3 7 8 3 1 6 2
35.3. Orthogonal Array L18 593
A1 A2 B1 B2 B3 C1 C2 C3 D1 D2 D3 E1 E2 E3 F1 F2 F3 G1 G2 G3 H1 H2 H3
Figure 35.2
Response graphs
A1 A2 B1 B2 B3 C1 C2 C3 D1 D2 D3 E1 E2 E3 F1 F2 F 3 G1 G2 G3 H1 H2 H3
F, and G are excluded from the predictions. The SN ratios under the optimum and
initial conditions, denoted by opt and initial, respectively, are predicted by
opt ⫽ T ⫹ (A1 ⫺ T) ⫹ (C3 ⫺ T) ⫹ (D3 ⫺ T) ⫹ (E1 ⫺ T) ⫹ (H2 ⫺ T)
⫽ A1 ⫹ C3 ⫹ D3 ⫹ E1 ⫹ H2 ⫺ 4T
⫽ A2 ⫹ C2 ⫹ D2 ⫹ E2 ⫹ H2 ⫺ 4T
Confirmatory experiments are conducted under the optimum and initial condi-
tions. From these results, two SN ratios are calculated. Their difference is the gain
confirmed. When the gain from prediction is close enough to the gain from the
confirmatory experiments, it indicates there is additivity and that the conclusions
are probably reproducible.