A Primer On The Taguchi Method - 1

Copyript C 1910 by Van Nostrand ReiDhold
Ubruy of C ~ Catalog Card Number 89-14736

ISBN 0442-2l1.l9-4
copyright --1BaJ
All rights ~ No part of dlis work covered by the
be ~ Of: used in any foml Of:
by any ~. electronic. Of: meehanieal.including
photocopying. ~. taping. or bafomJation storage .and
mrieval ~ written permission of the
publisher.
Printed in the ~ States of America
VanNosb:mct ·1cinboId
IJ$Fifth Av. .
Nett Yark. HT 10003
VIII Noscrand ReInhoId International Company Limited
11 New Fetter lane
London EC4P 41$. England
Van Nostrand ReinhoId
480 La Trobe Sleet
Melbourne. YlCteJria 3000, Australia
Nelson Canada
1120 Birc~ Road
Scarborougb. 0DIari0 MIK 504. Canada
1615 14 13 12 U 1098765432 1
Library of C..- Cataloging-in-Pablication Data
Ray, Ranjit.
A primer _ the Tagucbi method I Ranjit Ray.
p.CIIt. -- (Competitive manufacturing series)
BibliographJ: p.
Includes index.
ISBN 0442-23729-4
1. Taguchi methods (Quality control) 2. Taguchi. Gen 'kbi, 1924-
I. Title. H. Series.
TSI56.R69. 1990
658.5'62--dcZO 89-14736
':IP
l
To my wife Krishna and

my daughters Purba and PtlIIla.
I
I
t
f
.
II
I
"-----'" ."-... -. -_.-.~-- . . /

People do,,~ t use their eyes. They never see a bird. they see v~
a ~ I (~_
"'r
sparrow. They never see a tree, they see a birch. They see concepts.
- /"',
.. ;;' ~;:;:-
T
UJ ?v -Joyce Cary
p:., bP
- VNR COMPEIIllVE MANUfACTURING SERIES-
,.. ..
PRACTICAL fJCPIUUMENT DESfGNbrWilfilmJ. Diamond
VALUE ANALYSIS IN DESIGNby·. . . . . C. fowlet
A·PRIMER 0" THE TAGUCHI UEIHODby Ranjit Ray
MANAGING NEW-PRODUCT DEVELOPMENTbJ Gcoft'Vir)cent
ART ANl>SQlNCE OftNVElft'lNG by' Gilbert .~er)S01l
RELlABlUlY ENGlNEERlNG INSYSTENS DESlGN-ANDOPERAnoN by
JW>ir S. DbiIoD . ,
m.wm.rrvANt> MAlNTAINABW'IY~ ~
·Batbir S. DbiUon ...... Reicbe
I
..............
APPU£D REUAl1U1Y by hulA. TohiIs and bawd C. Trindad
INDUSTRIAL ROBOT HANDBOOJt

CASE HISTORIES Of EffECTIVE ROBOT USEIN 1OlNOl1STRIES by·Rid1MI K MUIcr
ROBOTIC TECHNOLQGV:PRlNCIPLES AND ·fNCUCEbyWanerG. ~
MAOftNE VIStON by Ndo Zucch aadlic:hlm4 K ....
DESIGN Of AUJ'OMAnC .MACHlNllrbr ~W ..l.eldz."".
trwfSOUCERSPOR AlJ1'0MM'DI .. , HoIdedd
~IN JHl)lJS'tIr.,Horde$ki
DlSTRl8tnED CONTROL SVSTEM5 by Ilic:hatJP. 'Lutcas
BUlK MATERIALS HANDUNG HAtmBOOJ( by.Jacob f~
...........
MlCROCONPUTER SOfTWARE fO!t MEaWaCAl
WORI<JMG TOWARDSJUST.... :ma

by ...... fde
by Anthony Dear
GROUP TECHNOLOGY: fOUNDATION fOtl.COUPEIDWEMANUFACl"URING by
,
l
t
Charles S. Snead
FROM IDEA TO PROfIT: MANAGING ADVARteDlWtUfAC11JRlNG TECHN~lOGY
~]uleA. Iller
COMPEmlVE MANUFAC'IlJlUG by Stenky Mm.
STRATEGIC PlANNING fOR 1HE 1NDUSTRJAt. ~G fUNCTION by
Jack Byrd and L Ted IIoore
." ..
SUCCESSFUL COST REDUCTION PROGRAMS FOR ENGINEERS AND MANAGERS by
E. A Criner
MATERIAL REQUIREMENtS Of MANUfACTURING by Donald P. SmoDk
PRODUCTS LlABIUTV .by Warren Freeclrnan
lABORATORY MANAGEMENT: PRINCIPLES &: PRACTICE by
Homer Black. RonaId Hart. Ordn Peterson
................
TOTAL MATERIALS MANAGEMENT: THe FRONTIER FOR COSl"-CUl11NG
IN THE 19905 by Eugene L. Magad and John ArnQs
MATERIALS HANDUNG: PRlNCIPLES AND PRACTICE by Theodore H. AIlegri. Sr.
PRACTICAL STOCK AND INVENTORY TECHNIQUES THAT CUT COSTS AND
IMPROVE PROfITS by C. Louts HOMnstein
JAW 'Q
,
,
Contents
~face /xi
Abbreviations hdi
1. Qudty ......... Produet ......... ~ /1

1-1 BactarouBd It
1-2 Desip of Blpemnents-The Conventional Approach. It
Exetdses 15 ..
, .
2. The Tapdd Approadl toQudty . . Cost .' I [. . . . .
2·1B~ n
n
2-2 _Tapebi PbHosopby 18
2-3 The Concept of the ·Loss Function /10
2-4 ~Desip Strategy 114 f
2-5 Analysis QfR.esults 116
2-6 Areas ofApplieatioa 111
2-7 The New Approach--Its Appeal and Limi_ 117
I
~ III
,. M t 01 Quality /If
3--1 The QualilyOaaractedstic 119
3--2 Vadadoa as a Quality
3-3 Cost of·V.... 122
tJD Y....
3-4 QuaIit)' and Variation /12
3--S The Qalily We Ate After J22
3-6 The TapdU Quality Strateg 123
3-7 Selecdng Design P~for Reduced VadIrioa f1A.
3-8 Commoa T.-I., 127'
Exercises '1.8
4. Proc:echq. or file. Tapdli Method and Its Be,dts m
4-1 The New Discipline 129
.
I,
j
4-2 Up-Flont Thirdcing 131
4-3 Experimeutal Efticiency 131
4-4Effec:tive Use of Statistical Process CoJIroI 132
4-5 LoJw Tenn Benefits 132
4-6 ~ Cost Benefits--Taguchi Loss ftmdioa 133
4-7 Specifying Tolerance Levels 137
Exedses 139
'..... '.,- _. ~ . .- ... ".' ~. .. ~'. .. .....

"
YiD Contents
s. Working Mechanics of the TapddDesIp of Experbnents 140

5-1 F~ for Experiment Layout 140
5-2 BasE Methodology 140
5-3Desiping the ExperimeDl 144-
5-3-1 Order of Running the Experiments 146
5-3-2 Analysis of Re$ults 141
5-3-3 Quality Characteristics 149
5-3-4 ANOVA Terms and Notations ISO
5-3-S Variation Computation-Examples 151
5-3-6 Degrees of Freedom (DOF) 152
5-3-1 Projecting of the ~ Performance 155
5-4 Desiping with More than 'Ilne Variable$ 156
5-4-1 Designs with 2-Level V.nables IS6
5-4-2 Designs widl3-Level Variables 158
5-4-3 Designs with Mixed levels Using Standard Arrays 158
5-5 Desips with Interactions /SS
5-5-1 Steps in the Design and Analysis 162
5-5-2 Interaction Effects /68
5-5·) Key Observations 171
5-. More Designs with Interactions 172
5-6 DeIJps with Mixed Factor Levels ns
5-"'1 Preparation of a 4 Level Column m
542 Preparation of an 8 Level Column 179
S-7 DumtBJ Treatment (Column Degrading) 182
5-8 CombiBation Design 187 '
5-9 .Desiping Experiments to Investigate Noise Factors 190
5-10 Benefiting from Repetitions 191
5-11 De1iDibon of SIN Ratio 192
5-12 Repetitions under Controlled Noise Conditions 194
'I
, 5-13 DesiS- and Analysis Summary 196
I Exen:ises 198
·1
u
!;
~t
:: 6. Analysis of Variance (ANOVA) 1100
1\ 6-1 The Role 01 AHOVA 1100
1
1!
6-2 ANOY:t\ Terms, Notations and Development 1100
m 6-3 One Way AHOVA 1106
:1 6-4 One Factor Two Level Experiments (One Way ANOVA) 1112
:;1 6-4-1 Confidence Intervals 1116
:1 6-5 Two Way ~OVA 1117
J 6-6 Experiments with Replications 1121
11 6-6-1 Plocedures for Pooling 1124
I
I:
I
4i 4 '..: ' .44
6-7 StaDdanl ArJalysis wilhSie&Je and Multiple a.. /126

6-7-1 Analysis UsiaB SiDIk Run /126
6-7-2 Poet" 1134
6-7-3 ·Cod.deac:e·1JdervaI of Factor Bffeet 1131
6-7-4 &tint.cd Itesalt at Optimum Condi.. /139
6-7-5 Cod.deac:e latenaIof cbe Rault .at the Opti......
Conditka 1140
6-7-6 Analysis wilhMuItiple .bns /142
6-8 Application of the SfNRatio 1145
6-8-1 CoIlvenioaof. . . . iDto SIN RatioS 1145
6-8-2 AdvantaJe of SIN Ratio over Average 1146
6-8-3 Computadoa of cbe SIN Ratio /148 f
6-8-4 Effect of the SIN Ratio on the ADalylis 1150 ~
6-8..5 When to Usethe SIN Ratio for ADalysis 1154 . I
Exercises ISS
7. Loss FwtdIoa '156
7-1 Derivation of Ims F'UDCtion 1156
I
7-2 Average LosI·F'wlcIion for Product Population /160
7-3 Application of ..... Punedon Coneel* 1160
Bxetdses /172
8. BrainstonDIJla-Aa 1Rt", Part of the TagadIl PIdlosephy 1t73

. 8-1 Tbe~ofB"""'" /173
8-2 TheN. . . . . . . /173
8-3 Topks of the DistussioDs 1175 •
8-4 Typical ~ia cbe Session 1176
BurdseI 1119
I
9. Enmples of T...... ea. StacIIes /181
9-1 Applicad.oa ~ /tSI t
9-2 Application ~ lnda4ina Deaip and Malyais /182
Example 9-1 .~ ValYe-Ttain Noise Study 11~
Example 9-2 .• . , of e.tshaft Surface PiaWring Process It.
Bxanlple 9-3 Autemobile Oenetator Noise St8dy /190
Example 9-4 Engine Idle Stability Study /193
Example 9-SlnstmmentPaneI Structure De$igD 0ptimi1.af.i0D /195
Example 9-6 Study Lead. . to the Selection of the Worst·Qse
Bmier Test Vehicle 1198
Example 9-7 AiIbag Design Study 1203
Bxanlple 9-8 TtaIlSDJ_on Control Cable Adjustment
Param.eterI 1204
.• _ ...'" __ .'i. ...~ ....... _ .

,
J[ ConteDlS
Example 9-9 Front Stmcture Crush Characteristics /208
References 1210
Appendix A Orthogonal Arrays, Triangular Tables and

Linear Graphs /211 -
Appendix B F-Tables /231
Appendix C Glossary 1243
Index /245
\-
\
\.
•
Preface
Since the late 19408 Or, Genicbi Tagucbi has introduced several new
statistical concepts which have proven to be valuable tools. the su!'ject
of quality improvement, Many Japanese ~ .bave used his
approach for improving product and process qDIIky widlllllpRCCdeoted
success. The quality of Japanese automobiles is attributable largely to
the widespread application of the Taguchi Method.*11Iis boot iDtroduces
the basic concepts of the Taguchi methods tImJuP appIicaIioo examples
and speCial case studies.
Taguchi ~ s method is based upon an approKb which is completely
different from the conventional practice$ ofquality enp-rins. His meth-
odology emphasizes designing the quality into die produds and processes,
whereas the more usual practice relies upon iDspedioa. Iftbis quality
improvement practices Taguchi essentially uti)ias theCOGVentiouaI sta-
tistical tools" but be simplifies them by ideaIifyioga. of stringent
guid~lines for experiment layout and the anaIJIis of JaUIts. Taguchi's
approach has been extremely effective in impnwiDa die .qualily of Jap-
anese products. Recendy Western indus1ries Iave to ncognize
Taguchi"s method as a simple but highly effecthe· to improving
product and process quality,
Taguchi"s experimental design approaeh is b' • wide range
of applications. But his techniques bave 00fIlIIMW1y applied to what
be classifies as "off-line" qualitY control. He dRe phases
ofoff-line quality control: system design" paraIIICIer , IDd tolerance
design. Parameter design is primarily the subject of boot.
This book is intended for practicing engineas and managers who are
interested in developing a basic understanding and basic still in utilizing
the Taguchi concepts and methodologies. Common appIicaIioo methods
and simple analysis techniques are discussed ......'. More complex
analysis and application techniques are beyond die scope of tbis work.
$Jhe cum Tagldi Metlrot:f8 is a ttademart of dle AmeriaII

Micbipn. and is used by permission.
SiMJficr.'"' 7 . . . . DeatJcn.
.,
I ,
ACKNOWLEDGMENT
I would like to express my gratitude to Ron Gaizer, ElIen Menzies and
Doug Carr of Technicomp, Cleveland, Ohio, who approached me to
prepare a manual to be used to develop a video training program on the
subject. My thanks also to a group of reviewers whom Tecbnicomp
employed 10 review the text materials during the video production.
I am also grateful to the bundreds. of pnleticing engineers at General
Motors, my fonner place of employment, who have unknowingly served
as guinea pigs as I preached and' practiced the application methodologies
oudiRed in this book. I thank all the lead en.gineers who sooght my advice
in arrying out their Tagucbi case studies and asked me 10 facilitate the
braiDstorming sessions. These application exetcises provided much of the
material in this book.
I am also indebted to my erilleagues at Nutek, Inc., Girisb Bapu and
Rourme Lawrence for their help in puttiag together the original report
OIl the .subject. A large portion of the eady versions of this manuscript
we. distributed in summary form to my seminar attendees and returned
to me with suggestions for improvement. I thank AyenulHaque, Scott
TayIor, .Marcel Beauregard, Brad Hammer" Oary Engel, WiUiam HaweD,
Avi Shah, Phi! Ross and VlSWas Bowele for innumerable suggestions for
improvements. I would also like to express my thanks 10 Dr. R. Roy-
cboodhury and Dr. S. Khashnabis, Who have given me advice and sug-
gestions for this book that only closepel$Onal friends can give. Their
critieism h4lS~n invaluable. I am also deeply grateful to two members
ofVan Nosmutd Reinhold: Gene J?allaire for finding me and for providing
encouragement to undertake this endeavOl', and George M. Kaplan for _
editing the manuscript and for making a DUtnber of suggestions toward
improving the presentation of the material_
Finally, thanks to all members ofmy family for their kind understanding
of my frequent absence from family activities and for help preparing the
manuscript. -;
ABBREVIATIONS AND SYMBOLS
A,B J : • Variables used in ,the design of an experiment

Ai The sum of obserVations under condition A, (i = 1,2,3, etc.)
Ai The average of observations under conditioD Ai
C.l. The confidence interval
OOE Design of experiment
e Experimental error
f, n Degrees of freed6m
F Variance ratio
k A constant used in the expression for the loss function
L The Tagucbi loss function
Ls An ortbogonal array that has 8 experiments
MSD Mean .square deviation
N The number of experiments
OA Orthogonal array, L.., La, LI6t etc.
P The percent contribution of a variable
S The sum of squares
S' The net/pure sum of squares
SIN The signal to noise ratio
T The sum of all observations
V The variance (mean square, SIj)
Y Results measured in terms of quality characteristics.
Examples: cost, weight, length, surface fi_.
a Alpha
JL Population mean .
a Popuhltion standard deviation
er Population variance
1 . .
~
Quality Through Product and

Process Optimization
;..
1-1 BACKGROUND
Mankind has always had a fascination with quality. Today's tecMology
bears testimony to man's incessant desire to provide a higher level of
quality in products and services to increase market share and profits.
Sometimes, quality is essential. A pacemaker that controls beart action
must operate continuously and precisely. An erratic pacemaker is val-
ueless, useless, and dangerous.
Driven by the need to compete on price and performance audto main-
tain profitability,quality conscious manufacturers are increasingly aware
of the need to oPtimize products and processes. Quality achieved by
means of design optimization is found by many manufacturers to be cost
effective in gaining and maintaining a competitive position .m dle world
market.
1-2 DESIGN OF EXPERIMEN.TS-THE CONVENTIONAL

APPROACH
The technique of defining and investigating all possible conditions in an
experiment involving· multiple factors is known as the design of experi-
ments. In the literature this technique is also referred to as factorial design.
The concepts of design of experiments have been in use since Fisher's
work in agricultural experimentation, approximately half a century ago.
Fisher (Ref. 12) successfully designed experiments to determining opti-
mum treatments for land to achieve maximum yield. Numerous appli-
cations of this approach, especially in the chemical and pharmaceutical
.industries, are cited in the literature. A thorough coverage of this subject
is beyond the scope of this study, but the method and its advantages and
1
2 A Primer on the Tasdi Method
disadvantages from an engineering point of view, is illustrated by a simple .

example.
Consider a snack food company planning to introduce a new chocolate
chip cookie in the market. The product designers have standardized all
other ingredients acept the amount of sugar and chocolate chips. Two
levels of chips, Cl _ C2 and two levels of sugar, SI and 52 were selected.
(Subscripts I and 2.. respectively refer to the low and high levels of each
factor.) In order to select the best combin~tion of these ingredients that
appeal most to poIImtial customers, the market researcll group decided
10 conduct a marloeI survey of customer Preference.
This is olleof _simplest cases of design of experiments. It involves
. two fac;tors (chips" sugar) at two different levels (high and low) that
affect the taste of cookies. Such an experiment is described as a2x 2
factorial experiJ1;leaL There are four (22) possible treatments or combi-
nations~ The res. . .s to these factors, as obtained by a taste test, are
given in Table 1-1.
Examination of customer response shows a 10% (55-45) increase in
preference for the 8IPf level 52 at the low level (Cl) of chocolate chips
but the response iatreases to 15% (80-65) when Illore chips C2 are used.
These increases me called the simple effect of sugar. On the other hand,
lot the higher ~ of"chips, the taste preference increased from 45%
10 65% at sugar bel 51 and further increased to 80% with the higher
sugar level 52' ne mean response, Le., the difference between the av-
Table I-I: Taste Survey with Preference in Percent
SUGAR LEVEL
LEVEL OF (CHOCOLATE <:HIPS)

CHOCOLATE CHIPS SI Sz MEAN (C2 ~ Cl)
Cl 45 55 50.0
22.5
CI1 65 80 72.5
Grand mean
Mean 55.0 67.5 61.25
Mean response (sUP)

12.5
(S2 - SI)
erage effects at two levels of sugar (12.5%), is called the main effect of
sugar. Similarly, the main effect for chocolate chips is 22.5%. It is
important to note that in this example, oqly the main effects are analyzed;
no attempt is made to analyze the interactions between the factors. l'bese
interactions may or may not be present. The interactions between various
factors of an experiment can be quantitatively determined by using the
analysis of variance (ANOVA). This procedure is described in Chapter
6. ~
For the present, the degree of the interactions for a 2 x 2 experiment
can be determined from Figure 1-1 which graphs the response against
one factor (C) for two levels of the second factor (S). Since the lines
for the two levels, SI and S2are almost parallel, the factors ($ and C)
- are said to be independent and little or no interaction is assumed to exist.
NonparaUellines would indicate the presence of some interaction. Highly
skewed or intersecting lines would indicate strong interaction between
the two factors. Figure 1-1 indicates a slight interaction between the two
factors (sugar and chocolate chips).
In the.above example there wereoniy two factors, each at two different
levels. It would be rather easy to manufacture four types of cookies
reflecting all possible combinations of the factors under study and to
subject them to a market survey.
84
801-----------------------__.-4
76
72
I 68
&64
. ~ 601---~-----~------t
56
52
48
44
40 "'---'"-'-- -...&.---I
C1 ~
Level$of Chocolate Chips
Figure 1-1. Factor effects.

.
4 A Primer on the Taguchi Method
Table 1-2. Test Matrix with Three Factors at Two Levels

.4(
Cl
CeD
Average
For a fuU factorial design the number of possible designs N is
N = L'" (1.1)
I where L = Number of levels for each factor

and m = Number of factors
I J11us, if the qualities of a given product depends on three factors A,

S;' and C and each factor is to·be tested at two levels then, Eq.. (1-1)
m3icates 23 (8) possible design configurations. This three factor, two
%
level experiment is represented by Table 1-2.

The size of the test matrix is still easily managed and every combination
can be investigated. Each box in the table is called a cell. In order to
improve accuracy several observations are made per cell and the signif-
icance of the results are determined by statistical analysis (ANOVA).
Now let us consider the case where the cookies under consideration
'0
have'I5 different ingredients, at tWo levels of each. In this case, 2 15
! (32,768) possible varieties of cookies need to be investigated before the
i
most desirable recipe can be established. A market research program of
this magnitude would be exorbitant in .cost and time. Techniques such
as fractional factorial experiments are used to simplify the experiment.
Fractional factorial experiments investigate only a fraction of all the
possible combinations. This approach saves considerable time and money
but requires rigorous mathematical treatment, both in the design of the
experiment and' in the analysis of the results. Each experimenter may
design a different set of fractional factorial experiments. Herein lies Tag-
uchi's contribution to the science of the design of experiments. He sitp-
plified and standardized the fractional factorial designs in such a man-
ner that two engineers conducting tests thousands of miles apart, will
Quality Through Product and Process Opdmization 3
erage effects at·two levels of sugar (12.5%), is called the main effect of
sugar. Similarly, the main effect for chocolate chips is 22.5%. It is
important to note that in this example, only the main effects are analyzed;
no attempt is made to analyze the interactions between the factors. These
interactions mayor may not be present. The interactions between various
factors of an experiment can be quantitatively determined by using the
analysis of variance (ANOVA). This procedure is described in Chapter
6..
For the present, the degree of the interactions for a 2 x 2 experiment
can be determined from figure I-I which graphs the response against
one factor (C) for two levels of the second factor (S).Sinee the lines
for the two levels, SI and 82 are almost parallel, the factors ($ and C)
are said to be independent and linle Of no interactionJs assumed tQ exist. t
Nonparallellines would. indicate the presence of some interaction. Highly I.,''..,
skewed or intersecting lines would indicate strong interaction between .

the two factors. Figure 1-1 indicates a slight interaction between the two
factors (sugar and choco~ate chips).
In the above example there were only two factors, each at two different
levels. It would be rather easy to manufacture four types of cookies
reflecting all possible combinations of the factors under study and to
subject them to a market survey.
II
i
C, C2
Levels of Chocolate Chips
Figure 1-1. Factor effects.

, A Pritner on the Tagucti Metlwd
Taille 1·3. Two Factor ExperlmentData

A RESPONSE 1'0 B
At; Az AVERAGE <B. - Bz)
•
f6
70
45
8. =
82 =
Average All = A2 :::
Response
(A~ - AI)
1·3. A product involves three primary parameters at three different levels of eacb. In
order to optimizelk product. a full factorial design is plarmed for experimental
evaluaUoilS. How DlIliIlY possible design configurations need to be tested to achieve
. cbe objective?
1..... Draw a factor grapldDr the experiment shown in Table 1-3 and <tiscuss the results.
2
The Taguchi Approach to Quality
and Cost Improvement
;..
2-1 BACKGROUND
After the Second World War the allied forces found that the quality of
the Japanese telephone system..was extremely poor and totally unsuitable
for long term communication purposes. To improve the system the allied
conunand recommended that Japan establish research facilities similar to
the BeULaboratories in the United States in order to develop' a state--of-
the-art communication system. The Japanese founded the Electrical Com-
municationLaboratories (ECL) with Dr. Tagucbi in charge of improving
the R&D productivity and enhancing prodUd quality. He observed that
a great deal of time and money was expended in engineering experi-
mentation and testing. Little emphasis w~ given to theproceu of creative
brainstorming to minimize the expenditute of resources.
Dr. Tagucbi started to develop new methods to optimitethe process
of engineering experimentation. He developed techniques which are now
known as the Tagucbi Methods. His greatestcontributton lies not in the
matbematical formulation of the desip of experiments, but rather in the·
accompanying philosophy. His approacbis more than a method to lay
out experiments. His is a concept that basproduceda unique and powerful
quality improvement discipline that differs from traditional practices.
Two completely opposing points of views are commonly held about
Tagucbi~s contribution to the statistical design of experiments. One view
holds that Dr. Taguchi'scontribution to the field of quality control is one
of the most significant developments of the last few decades. The other
view maintains that many ofthe ideas proposed in his approach are neither
new nor were they developed by Or. Tagucbi. This text will not resolve
this controversy but will document successful case studies using Taguchi's
methods. These new techniques were transplanted to the United States
and created significant changes in quality engineering methods in this
country. The Taguchi approach has been successfully applied in several
7
8 A Primer 00 the Taguchi Method
industrial organizations and has completely changed their outlook on

quality control.
2-2 TAGUCW PWLOSOPHY

Taguchi espoused an excellent philosophy for quality control in the man-
, ,,,'" -.;
ufacturing industries. Indeed. his doctrine is creating an entirely different
breed of engineers who think, breathe and live quality. He has, in fact,
given birth loa new quality culture in this country. Ford Motor Company,
for example, has decreed that all Ford Motor and suppliers~ engineers be
trained in Taguchi methodology and that these principles be used to
resolve.tbeir quality issues. His philosopbyhas far reaching consequences,
yet it is founded on three very simple and fundamental concepts. The
whole of the technology and techniques arise entirely out of these three
ideas. These concepts are:
1. Quality should be designed into the product and not inspected into
it.
2. Quality is~st achieved by minimizing the deviation from a target.
'The prod~t should be so designed that it is immune to uncontrol-
lable envtftmmental factors.
3. The cost of quality should be measured as a function of deviation
from dlestandard and the losses should be measured system-wide.
Taguc.hi built upon W. E. Deming's observation that 85% of poor

quality is attributable to the manufacturing process and only 15% to the
worker. Hence he developed manufacturing systems that were "robust"
or insensitive to daily and seasonal variations of environment, machine
wear, and other external factors. The three principles were his guides in
developing these systems, testing the factors affecting quality production
and specifying product parameters.
Taguchi believed that the better way to improve quality was to design -:
and build it into the product. QU,ality improvement starts at the very
beginning, i.e., during the design stages of a product or a process, and
continues through the production phase. He proposed an "off-line" strat-
egy for developing quality improvement in place of an attempt to inspect
quality into a product on the production line. He observed that poor
quality cannot be improved by the process of inspection, screening and
salvaging. No amount of inspection can put quality back into the product;
"
"
it merely treats a symptom. Therefore, fIU8Iity coeccpts should be based

upon, and developed around, the philosophy ofpmention. The product
design must be so robust that it is immune to dle infI8ence otluDe01ltrolled
environmental factors on the manufaeturiDg processes. He emphasizes
that quality is what one desipsinto aprodlet.
His second concept deals with actual tnedlodS «effecting quality. He
conteaded that quality is directly relatedto.deviatioaefa desipparameter
from the target value, not to eonforma&ee to SOJICfixed .peeifications.
A product may be produced with propelties skewe4 towards one end of
an acceptance range yet show shorter life expectaneJ. However, by spec..
ifying a target value for the critical property...developing manufacturing
processes to meet the target value withlittledeviatUI. the life expectancy
may be much improved.
His third concept calls for measurina deN.tioJls from a given design
parameter in terms of the overall lifeqele ~.of the product. These
costs would include the cost of scrap, 1'eWOIk, iaspcetion, returns, war..
ranty service calls and/or product rep~L t'Jlesecosts provideguid..
ance regarding the major parameters to be GODtro8ed.
Tagucbi views quality improvement effort. He contin..
uaIly strives to reduce variation arolUld ·dle target A product under
investigation may exhibit a distribution dB has a mean val. different
from the target value. The first step ·towardsinlploving quality is to
achieve the population distribution as closetodle talJd value as possible.
To accomplish this, Taguchi designs expetimeDts -mg especially con-
structed tables known as 4'ortllolonal amtys"(OA). The use of these
tables makes the design of experiments very easy Md conSistent.
A second objective of manufacturing products to conform to an ideal
value is to reduce the variation or scatter arotmdd1e_get. To accomplish
this objective~ Or. Taguchi cleverly incorporates a unique way to treat
noise factors. Noise factors, according to his terminolop-, are factors
which influence the response of a process., buteaauot be economically
controlled. The noise factors such as weatherconditions, machinery wear,
etc., are usually the prime sources for variations.. Througbtbe use of
what lie calls the "outer arrays,'" Tagucbi devised an effective way to
study their inftuence with the least number of repetitions. The end result
is a "robusf' design affected minimally by noise.. i~e~, with a higb signal
to noise value~
To acbieve desirable product quality by design., Dr. Taguchi recom-
mends a tl'iree stage process.
10 A ...... on the Tagucbi Method
I. S}stems design
2. Penmeter design
3. T~ce design
The kus of the system design phase is on determining the suitable

workilwln'els of design factors. It includes designing and testing a system
based _die engineer's judgment ofselected materials, parts, and nominal
produ~ess parameters based on current technology. Most often it
involves _ovation and knowledge from the applicable fields of science
and tedlmlogy.
WbleSfstem design helps to identify the working levels of the design
factors" JlllaMeter design seeks to determine the factor levels that produce
the be$t pafonnance of the product/ProCess under study. The optimum
conm_is selected so that the influence of the uncontrolled factors (noise
factors)CEeS minimum variation of system performance. This text deals
solely will parameter design.
T91caa de~~p· is a step used to fine tune the results of parameter
design bJ IighteilJpg the tolerance of factors with significant influence on
the pl"Ollld. Sui' steps will normally lead to identifying the need for
better lSIferials;;buying newer eqUipment, spending more money for
.. .... . ,,'~;~,
mspec., etc. •
2-3 TB CONCEPT OF THE LOSS FUNCTION

The caEept of the 44totalloss function" employed by Dr. Taguchi has
forced aaPueers and cost accountants to take a serious look at the quality
control paetic.es of the past. The concept is simple but effective. He
defines .-Jity as uthe total loss imparted to the society from the time a
prodUct is shipped to the customer." The loss is measured in monetary
terms . . includes all costs in excess of the cost of a perfect product.
The deillilic>n can be expanded to include the development and manu-
facturm, plJases of a product.
A pocdJ conceived and designed product begins to impart losses to
society tiDm the embryonic stage and continues to do so until steps are
taken. improve its functional performance. There are two major cate-
gories ~ loss to society with respect to the product quality. The first
categolJrelates to the losses incurred as a result of harmful effects to
society(c.g., pollution) and the second relates to the losses arising because
of exc~e variation in functional performance. In this book, the total
loss fuDlttion refers essentially to the second category.
The Tagu.cbi Approach to Quality and Cost Improvement 11
The conventional method of computing the cost of quality is based on

the number of parts rejected and reworked. This method of quality eval-
uation is incapable of distinguishing between two samples, both widrin
the spec:ification linuts, but with different distributions of targeted prop-
erties. Figure 2-1 shows the conventional method and Taguchi."s view of
loss function. Thisgrapb depicts the loss function as a function of de-
viation from an ideal or the target value of a given design parameter.
Here T represents the target value or the most desirable value of;.tbe
parameter under consideration. This parameter may be a critical dimen-
sion') color of the product, surface finish or any other cbaracteristictbat
contributes to the customer'sconception of quality. How this ideal value
of the parameter was arrived at and bow significant this value is in
achieving our quality goals-, will be evident later.
UAL and LAL in Figure 2-1, represent upper and lower acceptable
limits of a design parameter, respectively. Normally the product is func-
tionally acceptable if the value of the specified parameter is within the
range between the UAL and LAL limits. No societal loss is assumed to
occur; the product is shipped to the consumer. However outside these
limits, as shown by the crosshatched region, 100% funetionaldeterio-
ration occurs and the product is either discarded or subjected to salvage
operations. Every .attempt is made to ~Qntrol the manufacturing process
to maintain the product within these limits.
However, according to Tagucbi, there is no sharp cutoff in the real
wodd. Performance begins to gradually deteriorate as the design param-
eter deviates from its optimum value. Therefore he proposed that the loss
function be measured by the deviation from the ideal value. This function
is continuous as shown by the dotted line in Figure .2-1. Product perfor-
/.
"" "- Old School
(No Loss RatlfP'1
/
/
'"
,/
100% Taguchi Conventional
777J"77/.7nmt~loss Function
_
Loss Function
........ .,/
....... ........ . /
..............
UAl Target Value (T) LAl
LAl = lower allowable limit

UAl = Upper allowable limit
Figure 2-1. Taguchiand conventional loss functions.

mance begins to suffer when the design parameters deviate from the ideal
or the target value. Taguchi's definition clearly puts more emphasis on
customer satisfaction, whereas previously all definitions were related to
the producer. Optimum customer satisfaction can be achieved by devel-
oping the products which meet the target value on a consistent basis. It
may be worthwhile to mention that Taguchi allows for more than 100%
loss imparted by a product. Such cases can occur when a subsystem
results in a failure of the entire system or when a system fails catastrophi-
cally. Thus the single most important aspect of Taguchi's quality control
philosophy is the minimization of variation around the target value.
A case study conducted by SONY makes it abundantly clear that these
two schools of thought are significantly different from each other,and
indeed affect customer satisfaction. SONY manufactures color TVs in
Japan and in the V.S. The TVs from both sources are intended for, the
U.S. market and·have identical design and system tolerances. Yet Amer-
ican consumers consistently preferredtbe color characteristics of the TV
setS manufacturei overseas.
A study wasjPnducted to detellllin.e a cause for the difference in
customer preferec.I. • '". P
. . . The resuItsindicated that the frequency distributions
for the sets man~ct\1red in V.S. and those manufactured in Japan were
significantly different as shown in Figure 2-2. Plants in both countries
produced TVs with color density within tolerance range. No TVs with
out-of-tolerance color characteristics were shipped to the consumer. How-
ever, the U. S. built sets followed a somewhat flat distribution consistent
with a go/nogo philosophy, while the product manufactured in Japan
followed a normal distribution with small deviation from the target value.
The large scatter. observed in lhe performance characteristics of the prod-
uct manufactured in the V.S., as is evident from Figure 2-2, was 're-
sponsible for significantly lower customer preference for these sets. Once
the process in the U. S. plant was brought under control and began to
produce the frequency distribution similar to the TVs produced in Japan,
customer satisfaction with the V. S. product achieved the level of satis-
faction seen with the imported sets. The Sony case demonstrated that
quality is more than just producing between upper and lower limits-
quality is achieving the target as much as possible and limiting deviations
from the target.
Consider another example which will further support this concept of
quality .~wo batches of main bearings for an internal combustion engine
were received from two different sources, A and B, for a new engine
development program. Under laboratory conditions, bearings from source
.. ~-_. --..'--~-----------------
1
t
1
Il
°LAL T
CoIorDensity
UA I
I
LAL - Lower _lIowable limit
UA - Upper allowJl)le limit
TAR - Target value
Figure 2-2. Color density distrib~
B wore much faster thantbose from souroe A. Topiapointthe cause of

unequal wear, selected performance cbaracteristicsafthe bearings were
measured and posted. Both batches of bearings 'WIle. within the design
specifications. However, the source B bearings co&tendymeasured a
mean diameter ·on.the larger side of the tolerance limits as depicted in
Figure 2-3. Although within the tolerance band die larger diameter re-
sulted in excessive clearance. Bearing analysis later .evealed that exces- I
sive clearance adversely effected the oil film thicbas, causing.~ poor
wear properties of this batch. The problem was d~ by adjusting the
manufacturing process to maintain bearing diameterJCaf the target value.
I
I
The loss function and its implications are diSC1i8rtd in detail in later I
sections. At present, it is important to note that: I
• The quality loss function is a continuous funetico and is a measure

of deviation from the target value. The confolmmce to specification
limits LAL and UAL is an inadequate measure 10 define the quality
loss function.
• Quality loss is related to product performance dmracteristics and can
14 A Primer on the Tagmhi Method
Nominal Diameter
41---------+-----------------1
Good Wear Character
Soun:eA
31--------I--+~r__--------------I
Poor Wear Ch8lllW!r

Source B
11----+-----+---+-+-----~--------I
UAL
+ Bearing Diameter
2-3. Bearing diameter distribution.
best be minirniJa by designing quality into the product. Prevention

of poor quality is less costly than rework, and yields far better returns.
• Quality loss results from customer dissatisfaction and should be mea-
sured system-wide rather than at a discrete point in the manufacturing
process.
• Quality loss is a financial and social loss.
• Minimization of quality loss is the only way-to be competitive and
survive in today's international business environment.
2-4 EXPElUMENT DESIGN STRATEGY

Taguchf constructed a special set of orthogonal arrays (OAs) to layout
his experiments. The use of latin squares orthogonal arrays for experiment
designs dates back to the time of World War H. By' combining the
I orthogonallatin squares in a unique manner, Taguchi prepared a new set
of standard OAs 10 be used for a number of experimental situations. A
comm~n OA for 2 level factors is shown in Table 2-1. This array, des-
ignated bytthe symbol L g , is used to design experiments involving up to
seven 2 level factors. The array has 8 rows and 7 columns. Each row
represents a trial condition with factor levels indicated by the numbers
TIle Taguchi Approach to Quality and Cost Improvement 15
Table 2-1. 0rth0g0DaI Array £,<2')
1'IJAL
NUMBER A B C D E F (J
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2 ;..
4 1 2 2 2 2 I l
S 2 1 2 1 2 1 2
6 2 1 2 2 1 2 I
7 2 2 1 1 2 2 I
8 2 2 1 2. I 1 2
in the row. The vertical columns correspond to the factors specified in

the study.
Each column contains four level 1 and four level 2 conditions for the
factor assigned to the column. Two 2 level faetorscom.binein fQur
possible ways, such as (1,1)., (1,2), (2,1), and (2,2). When two columns
of an array fonn these combinations the same number of times, the
columns are said to be balanced orotthogonal. Note that any two columns
of an La (2') have the same number of combinations of (l,I), (1,2),
(2,1), and (2,2). Thus, all seven oolumns of an L are otthogonal to each
other.
The OA facilitates the experiment design process. To design an ex-
periment is to select the most suitable orthogonal array, assign the factors
to the appropriate columns, and finally, describe the combinations of the
individual experiments called the trial conditions. Let us assume that
there are at most seven 2 level factors in the study. caU these factors A.
B. C, D. E. F and G, and assign them to columns I, 2, 3, 4, 5, 6, and
7 respectively of La. The table identifies the eight trials needed to complete
the experiment and the level of each factor for each trial run. The ex-
periment descriptions are determined by reading numerals I and 2 ap-
pearing in the rows of the trial runs. A factorial experiment would require
27 or 128 runs, but would not provide appreciably more information.
The array forces all experimenters to design almost identical experi-
ments. Experimenters may select different designations for the columns
but the eight trial runs will include all combinations independent of col-
umn definition. Thus the OA assures consistency of design by different
experimenters.
16 A Primer on die Taguchi Method
2-5 ANALYSIS OF RESULTS
In the Taguchi method the results of the experiments are analyzed to

achieve one or more of the following three objectives:
I. To establish the best or the optimum condition for a product or a

process
2. To estimate the contribution of individual factors
3. To estimate the response under the optimum conditions
The optimum condition is identified by studying the main effects of

each of the factors. The process involves minor arithmetic manipulation
of the numerical results and usually can be done with the help of a simple
calculator. The main effects indicate the general trend of the influence
of the factors. Knowing the characteristic, Le., whether a higher or lower
value prodl:lee8 the preferred result, the levels of the factors which are
expected to ~ce the best results can be predicted.
The knowledfe of the contribution of individual factors is a key to
deciding the dbte of the control lobe established on a production
process. Theatltysis of variance (ANOVA) is the statistical treatment
most commonly .apptied to the results of the experiment to determine the
percent contribution of each factor. Study of the ANOVA table for a
given analysis helps to determine which of the factors need control and
which do not.
Once the optimum condition is determined, it is USUally a good practice
to run a confinnation experiment. It. is, however, possible to estimate
performance at the optimum condition from the results of experiments
conducted at a non-optimum condition. It should be noted that the op-
timum condition may not necessarily be among the many experiments
already carried out, as the OA represents only a small fraction of all the
possibilities. .
Taguchi suggests two different routes to carry out the complete anal-
ysis. First, the standard approach, where the result of a single run, or
the. average of repetitive runs, are processed through main effect and
ANOVA analyses as identified above. The second approach, which he
strongly recommends for multiple runs, is to use signal to noise ratio.
(SIN) for the same steps in the analysis. SIN analysis determines the
«'{-
most. robust set of operating conditions from variations within the

results.
-_. ---_.._- -- "
The Tagucbi Approach to QQdty_ CosIImprovemcfa 17
2-6 AREAS OF APPUCATION

Analysis
In the design ofengineering products and processes, aaalytical simulation
plays an important role, transforming a conceptiato the fiDaI product
design. The Taguchi approach can be utilized to amve at the hestpa-
rameters for the optimum design configuration withdle least JUIQlber of
analytical investigations. Although there are seveal methods available
for optimization, using such simulations when tIle.fadors are coatiuuous,
the Taguchi method is the method that treats factomat discrete levels.
Frequently this approach significantly reduces com,.rtime.
Test ~nd Development
Testing with prototypes is an efficient way to see bow the concepts work
when they are put into a"design. Since experimentallaardwareis cosdy,
the need to accomplish the objectives with the least •. mber of ·testsis a
top priority. The Tagucbi approacb of laying out the experimental coo...
ditions significantly reduces the number of tests_ Cbe overaUtesda,
time.
ProeessDe'VeIopment
Manufacturing processes typically have a large IlIIlDber of factors lbat
influence the ·lk1aloutcorne. Identification oftheir inChidual eontribcdions
and their intricate interrelationships is essential in . . developMeBt of
such processes. The Taguchi concepts used in such projects ·bashelped
many U.S. and Japanese companies realize signifiamt cost savings in
recent times.
2-7 THE NEW APPROACH-ITS APPEAL

AND LIMITATIONS
The ApPeal
• Upfront improvement of quality by design and process development.

• Measurement of quality in tenns of deviation from the target (loss
function).
• Problem solution by team approach and brainstomling.
• Consistency in experimental design and analysis.
• Reduction of time and cost of experiments.
• Design of robustness into product/process.
18 A ftimer OB the Taguchi Method
• Reduction of variation without removing its causes.

• Reduction of product warranty and service costs by addressing them
wilh the loss function.
Tagudli's design methodology has wide ranging applications. Gen-

erally speaking, experimental design using OAs can be applied where
there aN a large number of design factors. Tagucbi's OAs for the design
of experiments, signal/noise analysis, and cost guidance based on the
loss fuattion has made his approach increasingly popular among prac-
ticing engineers. The items listed above are common features of the
Tagucbi approach. The team approach to problem solving resulted from
T~guchi"s extension of loss beyond the-Production line.
LimitadaDs
The m<8severe limitation of the Taguchi method is the need for timing
with respect to product/process development. The technique can only be
effective when applied early in the design of theproductlprocess system.
After tbe design 98riables are determined and their nominal values are
specified, experiJlintal design may not be cost effective. Also, though
the method has WIle ranging applications, there are situations in which
classical teclmiqw$ are better suited; in simulation studies involving fac..
tors that vary in a continuous manner, such as the torsional strength of
a shaft •• a function of its diameter, the Taguchi method may not be a
proper dJoice.
EXERCISES
. -
2-1. There are two types of losses that society incurs because of the poor quality of a
proded:. What are these losses?
2-2. Explain why the old definition of cost of quality is inadequate.
2-3. WhaI is the most important idea of Taguchi's concept of achieving higher produer
quafily? -
2-4. Nat1E three stages in the process of achieving desirable quality by design.
2-5. List mme areas in your field where the Taguchi approach can be used to improve
a procIu.ct or the efficiency of a manufacturing process.
2-6. A manufacturer of hi-fi speakers uses a gluing operation at the last s.tage in the
mauufacturing process. Recently because of a change in the bonding agent. the
quafily of the bond has been observed to be below specifications. Some engineers
rnaiDIain that the poor ~uality of the bond is attributable not to the change of glue
but mber to the accompanying application temperature. A 2 x 2 factorial experiment
inclldBg two different glues at two application temperatures is planned. List at least
u
three noise factors which may influence the outcome of the test.
3
Measurement of Quality
3-1 THE QUALITY CHARACTERISTIC

Every product is designed to perfonn some intended function. Some
measurable characteristic, generally referred to as the quality character-
istic iSllsed to express how well a product perfonns the function. Consider
a light bulb; its quality can be measured in tenus of its hours of life. FOl"
a machine automatically producing 2.00 inch diameter shafts, the devia-
tion from this target dimension may be a quality ehataderistic.1n a
majority of the cases, the quality characteristic may be a single measurable
quantity such as weight, length, hours, etc. For some produets subjective
measurements like "good," "bad," '1ow,"" and ~gk" maybe used. 10
other instances, .subjective and objective evaluations maybe .combined
into an Overall Evaluation Criteria (OBe).
No matter how the quality of the product is measured, by a single
aiterion, or by a combination of multiple criteria, the measure will pos-
sess one of the following three characteristics:
• the bigger the better

• the smaller the better
• the nominal the best
Suppose that we are investigating a pump la determine the best design

parameters which produce the maximum flow rate. In this case, the quality
of the design is judged by the flow rate measured in units of cubic feet
per minute and therefore will be of the characteristic uthe bigger the
better."
If on the other hand, the purpose of the study is to determine the least
noisy pump, the noise measured in units of say, decibels, will be of the
type described by "the smaller the better." When the object or process
under study has a target value, as for a battery of 9.0 volts or a process
19
20 A Primer on the Tapcbi Method
to machine a cylinder with a 3.00 inch inside diameter, the measure of

quality will possess "the nominal the best" characteristic.
3-2 VARIATION AS A QUALITY YARDSTICK

Variation is the la. of nature. In nature no two objects are absolutely
alike. They could be very similar, but hardly identical. No two people
are exactly alike. No two apples have precisely the same weight. Mother
Nature likes variety. Consider man/machine made items. Superficially
parts look and function alike. However, when examined closely, machine
made products also exhibit variation. Variations in nature are often ob-
vious to the human eye. This may not always be the case for manufactured
products. Two baD point pens of the same brand do not write in the same
way; two light bulbs _do not last the same length of time; two appliances
do not function in exactly the same manner; two engines of similar
specifications do not perform identically. This is because products made
for the same p will show factor influences.
Generally s "g, the quality characteristic of a product varies in
two ways. First~~lt differs from another of the same kind, and second, it
differs from the desired (target) value. Consider five 9 volt transistor
batteries. When their voltages are measured accurately with a voltmeter,
they may display a range of 8.90, 8.95, 8.99, 9.20, and 9.20 volts. All
of the batteries may work well for radios with a range of acceptance of
8.5 10 9.5 volts, exceeding the variation in these batteries. But for a
sophisticated instrument, only batteries which exhibit a voltage very close
to the target value, say 8.95 to 9.05 ,volts, will operate the instrument
properly. Batteries with excessive deviations from the target value may
produce unreliable readings or may even damage the instrument.
The first kind of variation can be displayed by comparing one item to
another. The maximmn voltage variation among the batteries is 0.3 (9.2
to 8.9) volts. Although all of the batteries are nominally rated at 9 volts,
most of them will deviate from this value. The deviation from this target
or nominal value, constitutes the other type of variation. These variations
are shown in Figure 3-1. In Figure 3-1(a), the average value of the
parameter deviates from the target value; the range of value (variation)
is also excessive. Figure 3-1(b) shows the average on target but the
variation is still excessive. Figure 3-1(c) illustrates the desired charac-
teristic-on target and with narrow variation.
Average value off target.
Too much variation around average value.
I
u.
Average Target Value
la)
Average value on target.

Too much variation around target value.
I
u.
Target Value
(b)
Average value on target.

Uttlt variation around target value.
>-
g.
I...
u.
Target Value
lc)
Figure 3-1. Typical quality distributions.

n A Primer on the Tagucbi Method
3-3 COST OF VARIATION

Early in his rese;u-ch, Taguchi observed that variation is common to all
manufacturing processes and that variation was the primary cause for
rejection of parts. Parts were rejected upon inspection when they did not
confonn to a predefined specification. Rejection increases the cost of
production. Often 100 percent inspection is excessively costly or im-
practical, thus a defective part may reach a customer and lead to warranty
costs and customer dissatisfaction~ -1ag~chi heldtbat variation is costly
even beyond the immediate fa<;toJY1lrodu.~tion cost and that excessive
. variation causes loss· ofquality( He contenCled that the cure of quality
I
loss is reduction of variation. Thus he reco~ended that effort should

be directed toward zero variation with less emph'asis placed on production
within fixed tolerance limits.
3-4 QUALITY AND V~TION

Taguchi view ·ation as a lack of consistency in the product giving
rise to poor q . With this view, Taguchi developed methodologies
aimed at reduc both the elements of variation--(a) deviation from the
target, and (b) ~ation with respect to others in the group. In name 3-
2, a typical quality measure of a product is compared to the desired state.
Note that the product mean value is off target and that the variation
around the mean is large though within upper and lower acceptance limits.
A much narrower distribution is desired with more frequentacmevement
of the target value and smaller variation around the target value.
How is this accomplished? What does it mean in tenns of cost savings?
The financial implications of variation will be covered in a later chapter.
In this chapter we will discuss, in detail., Taguchi's approach to variation
reduction.
3-5 THE QUALITY WE ARE AFfER

The quality of a product or a process may be difficult to define if! quan-
titative terms. Quality is what the customers perceive it to be, thusquality
varies froID product to product and from customer to customer. The
criteria customers use to judge .the quality of a product are related to the
satisfaction derived from the product and are numerous and often difficult
to quantify. Research has shown that a lack of product consistency is a
MeaSurement of Quality 23
4ro------------------......, Desired Condition

\ Mean == Target
/ STD. Dev.-low
I ,
3t-----_--+-_----------------I
I \
> I \
12t----+-----+---r---------~-....._f
u. I Current Status
. Mean <> Target
I STD. Dev. -High
It-----I----+-~~-t_---..JIi!c_------.....-f
Target Mean UAl

Quality Measure
Figure 3-2. Representation of the Taguchi approach.

J
major factor in the perception of poor quality. Consistency (reduced level
of variation) favorably affects most common elements of quality. To
t
customers~ quality may include service after delivery ~ ease of assembly ~
product perfonnance~ frequeneyof maintenance~ etc. Our focus is that
element of quality reflected by the performance of a product. The Tagucbi
approach for reducing
.
variation in performance is a two-step
. process.
1. Make the product/process perform in the best manner most of the

time (less deviation from the target).
2. Make all products perform as identically as possible (less variation
between the products).
3-6 THE TAGUCm QUALITY STRATEGY

Taguchi's approach to enhance quality in the design phase involves two
steps:
1. Optimizing the design of the product/process (system approach).

2. Making the design insensitive to the influence of uncontrollable
factors (robustness).
24 A Primer on the TapchiMethod
When a product is optimum, it performs best under the available op-

erating conditions. Depending upon the specified performance, the op-
timum will imply that the product has achieved the most, the least or the
target value of the quality measure. Optimizing the design of a product
means determining the right combination of ingredients or making the
proper adjustments to the machine so· that the best results are obtained.
Consider a baking process. Assume several bakers are given the same
ingredients to bake a pound cake, the object being to produce the best
tasting cake. Within limits, they can adjust the amount of ingredients,
but they can only use the ingredients provided. They are to make the
I
,
best cake within available design parameters. Taguchi's approach would
be to design an experiment considering all baking ingredients and other
inftuencing factors such as baking temperature, baking time, oven type
(if a variable) and so on.
3-7 SELEQ11NG DESIGN PARAMETERS FOR REDUCED

VARIATION
In the last 8ei~OD quality according to Taguchi's methodology was de-
tined. Tagucbistrives 10 attain quality by reducing the variation around
the target. In an effort to reduce variations, he searched for techniques
that allow variability to be reduced without necessarily eliminating the
causes of variation. Often in an industrial- setting, totally removing the
causes of variation can be expensive. A no cost or low cost solution may
be achieved by adjusting the level~ and controlling the variation of other
factors. This is what Taguchi tries to do through his Parameter Design
approach. There is no cost or Iow cost in reducing variability in parameter
design. Furthermore, the cost savings realized far exceed the cost of
additional experiments needed to reduce variations. -
The Taguchi method is most effective when applied to experiments
with multiple factors. But the concept of selecting the proper levels of
design factors, and reducing the variation of performance around the
i optimum/target value, can be easily illustrated through an example in-

volving only one factor. An electronic circuit which controlled-the color
characteristics of a television set, was significantly influenced by the line
I voltage. The experimenter, wishing to select the right voltage, investi-

gated the color quality at several input voltages. The influence of voltage
variation on color quality is shown in Figure 3-3. If the desirable range
of voltage for circuit design is between Vc and Va , then what voltage
Neuuremem of Quality 25
3 t - - - - - - - - - - - - # - - t - - - - t - - - -..........- - - - - I
l:
~ 2 .....-----...,...-----1----+------+----1
CS
8
1 ........--1------1----+---+----1
Voltage
i
Figure 3-3. Color quality response CURe..
should be specified for the circuit? Obviously. the choice should be a

point within working voltage range Vc and VD that p-ovides stable rolor
quality. In Tagucbi terminology, look for' the input voltage tbatreduces
variation of the cotor quality. The experimenter \\WkI initially select a
voltage at point B, so the variation around B, say to B' or B", would
minimally affect the output Le., the color quality.. 'foo1tage B is highly
attractive because small fluctuations in the line voIlage (B' to 8") will
not significantly affect the color quality of the TV as perceived by the
customer-user.
How much milk. butter. flour.

SUgar. eggs. etc.
make the best cake?
Figure 3-4. Cake baking experiment.

%6 A ~ on the Tagucbi Method
Figure 3-5. The desirable result of an optimized baking process.
The objective for products involving multiple factors is similar but

slightly lllOre~omplex. The idea is to combine the factors at appropriate
levels, each within the respective acceptable range, to produce the best
result and ye~xhibit minimum variation around the optimum result. To
I see bow tile Tagucbi technique is used for many factors, consider once
I again the process of determining the best recipe for a pound cake (Figure
3-4). Our objective is to determine the right proportions of the five major
I
I
i
ingredients, eggs, butter, milk, flour and sugar, so that the recipe will
Ingredients How much/many
A. Egg A,
00
B. Butter B, ( tJ
~
C. Milk C,
•
D. Flour 0,
,
~
E. Sugar E,
U
Figure 3-6. Factors and levels for a pound cake experiment.
Measurement of Quality 27
produce the best cake most of the time (Figure 3-5). Based upon past
experience the working ranges of these factors are established at the levels
as shown in Figure 3-6. At this point we face the fonowing questions.
How do we do detennine the right combination? How many experiments
do we need to run and in what combination?
3-8 COMMON TERMINOLOGY

;...
The technique for laying out the conditions of.experlments when multiple
factors are involved, has been known to statisticians for a long time. It
was first introduced by R. A. Fisher in England, in the 19208 and is
popularly known as the Factorial Design of Experiments. Themetbod
helps an experimenter determine the possible combinations of factors and
to identify the best combination. To detenninethe optimum combination
Taguchi prescribes carrying out a number of experiments under the con-
ditions defined by the rules he has developed. In experimental layout he
uses the same principles as that of factorial design, except that his methods
are much simplified and standardized.
In the case of the cake example, with five factors each at two levels
there are 32 (25 ) combinations of all possible factors and levels. If we
could bake 32 cakes, we would surely find the best tasting cake among
these 32 cakes and we would know the impact of each ingredient level
on taste. For most industrial situations, carrying out a large number of
experiments is not feasible.
Taguchi accomplishes the same objective with a smaller number of
tests. He selects a particular g trial fractional factorial design (orthogonal
array) that produces the most information regarding d1e best tasting cake.
The factorllevel combinations for these 8 experiments are defined by
using the LB (2') Ortbogonal Arrays. An OA is a table developed for two
level factOrs such that columns contain two sets of all possible combi-
n~tions of levels. OAs are' amenable to statistical analysis with a high
degree of confidence. Thus, an experiment planned to be consistent with
the array design, provides statistically meaningful results.
An OA experiment design leads to reduction of variation because of
controllable factors. Uncontrollable factors (noise, dust, etc.) can be
handled in two ways. First, the experiment trial can be repeated at different
noise conditions. Second, the noise factors can be included in a second
OA which is used in conjunction with the array of controllable factors.
This is clarified in Chapter 5.
21 A Primer on the Tapchi Method
Because OAs are used to define the unique experimental conditions as

well as the noise factors, Taguchi calls the former design Inner Array
and the latter Outer Array. When outer array experiments are performed,
the analysis includes the transfonnation of the results into a Signal to
Noise Ratio (SIN). This process identifies the optimum condition with
the least variability of the controllable as well as the uncontrollable fac-
tors.
The actual steps involved in designing the experiments using inner and
outer aaays will be discussed in Chapter 5.
EXERCISES
3-1.. How does Tagudti's view of quality differ (rom the conventional practice?,
3-2. How <k>es variation affect cost and quality?
3-3. What are the maiD causes of variation?
U. How is a product cfesign optimized?
3-5. How does Tagudti make the design less sensitive to the noise factors?
un.._.. !ibthAft_1 ?
~. "m;u. are!;Y~Q.OV6- anays.
':It JC.
.
3.7.. What is iri1;lied by the term parameter design and what is its significance in achieving
higher pt'd1Itet quality?
I
f
4
Procedures of the Taguchi Method and
Its Benefits
4-1 THE NEW DISCIPLINE

-The Taguchi method offers two new powerful elements. FU'St, it is a
disciplined way of developing a product or investipling complex prob-
lems. Second, it provides a means to cost effectively investigate the
available alternatives. Although Tagucm's method was built upon well
developed concepts of optimization through the design of experiments,
his philosophy regarding the value of quality and die procedure for car-
rying out experiments were new. The power and poplllarity of the method
lies in the discipline rather than the technique itself4 The power and the
resultant potential for cost savings will be reviewed in this chapter.
The technique is applied in four stqis.
1. Brainstorm the quality characteristics and design parameters im-

portant to the product/process.
2. Design and conduct the experiments.
3. Analyze the results to detel'l'Qine the optimUlB eonditioos.
4. Run a confinnatory test(s) using the optimum conditions.
These steps are contrasted with typical current practice. in Figure 4-1.
Brainstonning is a necessary and important step in the application

process. The nature and content oftbe bra.instorrni.a is dependent on the
type of project under study and thus no specific guidelines exist. Taguchi
recommends the participation of all relevant fimttional organizations,
including marketing. A suggested procedure for brainstorming for a Tag-
uchi experiment design, along with some general guidelines are described
in Chapter 8.
Taguchi experiments are designed according to some strict rules. A
.
·
1'.
30 A Primer on the Tagu<:hi Method
Curnmt Practice Taguchi Approach

(A Seritls Approach) (A Parallel PrOCf!$$)
let's Try This Brainstormingl
1. What Is the Quality Charaeteri~tic?

2. What Are the Design Parameters?
let's Try That
figu~~ I. Comparison of current practice and the Taguchi approach.

c<'>:1-.:,
set of OAs are used to design the experiments. A single OA may accom-
odate several eXperimental situations. Commonly used OAs are available
for 2, 3, and 4 level factors. Some standard arrays accomodate factors
of mixed levelS. In many situations, a standard OA is modified to suit a
particular experiment requiring factors of mixed levels. The process of
experiment design includes selecting the suitable OA, assigning the fac-
tors to the appropriate columns and determining the conditions for the
individual experiments. When noise factors are included in the experi-
ment, the condition of the noise factors for each individual experiment
is also determined.
The results of the Tagucbi experiments are analyzed in a standard series
of phases. First, the factorial effects (main effects) are evaluated and the
influence of the factors are determined in qualitative terms. The optimum
condition and' the performance at the optimum condition are also deter-
mined from the factorial effects. In the next phase, analysis of variance
(ANOVA) is performed on the result'! ANOVA study identifies the relative
influence of the factors in discrete terms. When the experiments include
multiple runs and the results are measured in quantitative terms, Taguchi
f I
ProceduIes oflbe Tagucbi Method and Its Beaefits 31

I
,
'I
. ,
recommends signal to noise ratio (SIN) analysis. In SIN analysis, the

multiple results of a' trial condition are first trans(onned into SIN ratios
and then analyzed. \
The optimum design identified in the analysis should be tested to
confinnthat perfonnance indeed is the best and that it closely matches
the performance predicted by analysis.
4-2 UP-FRONT THINKING :-

The value of brainstorming in product development, or in solving complex
problems is well known, yet it was rarely used in solving engineering
problems. Brainstonningprior to an experiment is a necessary require-
-mentin the Taguchi approach. However, Tagucbi does not Jive any
I
guidelilles to conduct brainstorming for an experiment. The content and
outcome of .a brainstorming session is latgely dependent on the oature of
a project, and as such, is a technique learned .primarily byex.perience.
Most application specialists consider brainstorming lObe the most im- t
portantelement in deriving benefits from the Taguchi method.
T~gucbi brings anew breadth to planning experimental studies. He t
.thinks through the whole process before starting the tests. This helps to
decide which factors are likely to be most important,how many exper- t
l
iments_ needed, and how the results would be measured and analyzed,
before .actuatly conducting any experiment. Figure 4-1 shows the typical
f
steps ofcummt praetice--some initial thillking, followed by some testing,
which, in turn, is followed by some more thinking and so on. In the
Tagucbi approach, the complete plan of bow to test, what to test, and
when to analyze the results, will all be decided beforehand. Ideally, a
brainstorming session will rely on the coUectiveexperience of the group
to determine the factors to be selected for testing in an appropriate design.
Practice of the Taguchi method fosters a team approach to design opti-
mization since participation of people from engineering~ manufacturing,
testing, and other activities may be necessary for complete variable iden-
tification.
4-3 EXPERIMENTAL EFFICIENCY

In most cases, the Tagucbi experiment design requires the least number
of test runs. A full factorial experiment with 15 factors at 2 levels each
requires a test matrix with 32768(2 15) test runs. A fractional factorial
experiment with an orthogonal array suitable for fifteen 2 level factors

consists of 16 test runs.
The experimental efficiency Taguchi offers can be described using the
following analogy. Assume that you are asked to catch a big fish from
a lake with a circular net. You are also told that the fish usually stays
around its hideout. But you have no knowledge of where this place is.
How do you go about catching this fish? Thinking analytically you may
first calculate the area of the net and the lake, then layout an elaborate
scheme to cover the entire lake. You may find, after all this planning,
that you need the whole day to locate the spot where the fish is. Wouldn't
it be nice to have a fish finder that could teU you the approximate locations
where to throw your net. The Taguchi approach in experimental studies,
to a great extent, works like a fish finder. It tens you which areas to try
first and then from the results of the trials, you determine, with a high
degree of certainty, the most probable location of the fish.
4-4 EFFE~TIVE USE OF STATISTICAL PROCESS CONTROL

After design}tnd development comes production. When we complete the
Taguchi experimental studies, it is time for statistical process control
(SPC). But where do we apply controls? Should we control an factors
across the board? If we knew which factors were most significant. it
would be wise to pay more attention to them. The information about the
relative merits of individual factors is obtained by analysis of variance
of the experimental results. This knowledge about the factor influence is
used to objectively determine the amount of manufacturing process control
necessary.
4-S LONG TERM BENEFITS

Most of the benefits of improved design quality, come after the product
is put in use. The reduced variation, a characteristic that is designed in
through the optimum combination of the factors, will yield consistent
performance of the product. This means that more of the products will
perform as designed. There will be happier customers, and.therefore, less
. warranty costs, and increased sales.
Procedures of the Taguchi MdJriod and Its Beaetits 33
4-6 QUANTIFYING COST BENEFlTS-TAGUCHI LOSS

FUNCTION
As indicated earlier, the major attraction of the Tapdliapproadl is the
discipline it introduces in the engineering practices rather than in the
direct benefits in time and cost savings. The value oftbe discipline is
extremely hard to quantify. The direct cost savings io terms of a better
Product or process, on the other hand, can be experielted in due course
of time after the product is sold. Can this potential costsaviog be estimated
before production begins? Taguchi suggests a way of quantifying such
cost savings. He uses his loss function concept to eatima.tethe potential
savings based on the improvement achievable, if_product were de- , f;
signed ~ the optimum condition prescribed by his approach. r
I
Suppose a manufacturer, XYZ Co., of a 9.00 volt transistor battery
applied the Taguchi method to improve the quality of ilsproduct. Before
the experiment, a measured sample of 10 batteries had the following
voltages:
BEFORE EXPERIMENT
VOLTAOQ
8.10 8.25 8.90 8.6$ 1.35

9.25 9.05 8.85 8.45 8.90
With the· target value of 9.00 volts, the above measured values produce
characteristics as shown below (also see Tables 4-1 and 4-2).
Average value == 8.67

Standard deviation == 0.37
Mean square deviation (MSD) == (8.1 - 9.0~ 4- (9.25 - 9.0)2
+ ... + (8.90 - 9.0f]/10
= 0.23
SIN ratio = - 10 Log lO (MSD)
= 6.36
34 A ftimer on the Tagu<;hi Method
Table 4-1. Standard Statistical Data Before

Experiment
Observation No. I ::= 8.100

Observation No. 2 = 8.900
Observation No. 3 = 8.450
Observation No. 4 ; : 9.250
Observation No. 5 ::= 8.860
Observation No. 6 == 8.350
Observation No. 9 ; : 8.900
TargettNominal value of result (Yo) == 9.00

Number of test results (HR) == 10
AVERAGE AND STANDARD DEVIATION:

Total of all test results == 86.79001
Aventl~ of test results == 8.679001
S~deviatiOD (SO) = 0.376252
V' = 0.141565
anaBI
LOSS FVNC110N PARAMETERS:
Mean square deviation (MSD) = 0.230449
$igoal to noise ratio (SIN) == 6.374235
Variance (modified fonn) - 0.127409
Square of mean value = 0.103041
VARIANCE DATA (ANOVA):

Target value of data/test result = 9.00
Mean of data/deviation from target - -0.321
Total variance (ST)
(ST = Variance • NR) 1.274089
Correction factor (CF)
fCF == (average of data)2 • number of data) 1.030404
Sums of squareslN 2.304499
DEFINITIONS:
Standard deviation (SD) = ~ ±
;= I
(Y; - y)2/(n - 1)
Variance................. = (SD)2
n
Mean square deviation (MSD) = ~ (Y; - Yo )2/d
;= I
Signal/noise ratio (SIN) .. == - IQ Log lO (MSD)
, '
, ,
Procedutes of the Tagucbi Method and Its Benefits 35
Table 4-2. StamIard StatIstkaI Data After Experiment

Observation No. 6 == 9.1SO
;..
TargetINominal value of result == 9.00

- Number of test results (HR) == 10

Total of all test results == 89.93999
Average of test results = 8..993999
Standard deviation (SO) == 0.159875
Variance == 0.02SS60
f
LOSS FUNCIlONPA.R.AME'l'ERS:
Mean square deviation (MSD) == 0.023040
I
Signal to llOise ratio (SIN) == 16.31511
Variance (modiied form) == 0.023040
Square of .1Ile8Il value ::: 3.6()()6E..()S
VARlANCE DATA (ANOVA):

Target value of dataitest result - 9.00
Mean of dataldeviation from target == -6.00004&03
Total variance<sn
(ST == Varianee • NIt) == 0.230040'
Correction factor (Cf)
(cP :;: (average of dataf • number of data) == 3.6006E-<l4
SUIIlS of squareslN == 0.230040
After the experiment a batch of 10 batteries showed the following char-

acteristics:
AFTER EXPERIMENT
9.10 8.93 8.69 8:92 9.08

8.08 9.02 8.91 9.15 9.25
Average value = 8.99
Standard deviation = 0.1598
Mean square deviation = 0.023
SIN ratio = 16.37
The signal to noise ratio (SIN) expresses the scatter around a target
value. The larger the ratio, the smallertbescatter. Taguchi's loss function
can be expressed in terms of MSD, aod thus SIN ratios. Knowing the
SIN ratios of the samples before and after the experiment, Taguchi's loss
function may be,used to estimate the potential cost savings from the
improved product.
Before estimates of savings can be made, some other pertinent infor-
mation needs to be gathered. Assuming the usual statistical distribution
of results, the two samples will exhibit the curve shown in Figure 4-2.
The producer, XYZ, Co., makes 100,000 units of the batteries per month
which ~ell for $1.25 each. For most customer applications, the battery
voltage should be within ± 1.00 volt, Le., between 8.00 and 10.00 volts.
If the voltage is beyond this range, customers request a refund ($1.25).
Taguchi's approach to the computation of cost savings is based on
determining the refund cost associated with variation of batteries, as
measured by the mean square deviation (MSD) from the target voltage.
Obviously the greater the variation the more likely that some batteries
will exceed the limits of customer acceptance. With the above infor-
mation, the loss is computed as $.288 per battery for the sample before
the experiment, and that for the sample after the experiment is $.028 per
unit. Since 100,000 units aremanufacwred per month, the total savings
per month is calculated to be $25,950.90 (Fig. 4-3).
Procedures of the Taguchi Method and Its Benefits 37
2.0 , . . - - - - - - - - - - - . - - - - - - - - - - - ,
1.5 ..--------------#--t-~~--------___t
I,·o............---------#-+-I--+-o\---------t
...
u..
0.0 "--"---t--&f--+"---""--........- ' 1 ! ' " - - i - - - -

Volts ......
I
COMPUTATION Of MANUFACTUREA&SUPPLIER TOLERANCE
Nominal value of the quality characteristic Y == 9 Volts
Tolerance of Y (range of deviation) == ± 1 Volt
Co$t to repair al'lonfunctioning unit by customer == $ 1.25
Co$t to repair a nonfunctioning unit by manufacturer = $ 0.5
Co$t to repair it nonfunctioning unit by supplier =: S 0.25
REQUIRED TOLERANCES
Manufacturer tolerance == 9 ± 0.63
Supplier tolerance == 9 :f: 0.40
NOTE: If thete Tolerantes were held. there will be no nonfunctional part in the customers
hands. For thesarne cost the Manufacturer will maintain satisfted eustOmet'$ and quality
products in the field.
Figure 4-2. Manufacturer and supplier tolerance.
4-7 SPECIFYING TOLERANCE LEVELS

Another application of Tagucm's loss function fQllllulation is in deter-
mining the levels of tolerances for various inspection points of the pro-
duction process. Suppose that the manufacturer XYZ is well aware of
the losses in his current production samples (before experiment). He
wishes to reduce the warranty costs and wishes to keep his customers
satisfied. But he does not want to disturb the current design and production
38 A Primer OB the TagudB Method
8efore Experiment After Experiment
A'If!f"age =8.67 Average = 8.99

Standard Deviation = 0.37 Standard Deviation = 0.1598
MSD =0.23 MSD = 0.023
SIN =6.36 SIN = 16.37
Observation 8.67 9.00
CALCULATION OF LOSS
PROBLEM DEFINITION
Target value of quality characteristic (m) = 9.00
Tolerance of quality characteristic = 1.00
Cost of rejection at production (per unit) := $1.25
Uniu produced per mQflth (total) = 100000
SIN ratio of current designlpart ::; 6.37
SIN ratio of new des9r/part = 16.37
COMPUTATION OF lOSS USING TAGUCHI LOSS FUNCTION
Loss function: L(y) ::; 1.25 x (MSOJ Also L (y) == K x (y-m)2
BEFORE EXPERIMENT:
Loss/unit due to daviation from target in current design :::;: $0.288
AFTER EXPERIMENT:
Losslunit due to deviation from target will be reduced from SO.288 to = SO.028
MONTHLY SAVINGS:
Ifproduetion were maintained at the improved
condition, then basEld on 100000 units/month = $25,950.90
Figure 4-3. Calculation of cost savings.
line. He is, however!J willing to explore ways to screen out the bad
products. The loss function offers some help here.
Let's say XYZ as the manufacturer of batteries, has a producer who
supplies the chemicals needed. Upon investigation, XYZ determines that
the chemical which is supplied by the company ABe, is substandard and
is the cause of voltage variation. The manufacturer has two options avail-
able. He can inspect his production with the hope to screen out all the
bad products. Or, he can ask the supplier to prescreen qte material so
inspection of the product becomes unnecessary. In either case he needs
to establish the limits to which the batteries have to conform. We have
Procedures of the Taguchi Method aDd Its Bene&$ 39
the customer tolerance, Le., + 1.00 volt. A manufacturer tolerance which

is what the company XYZ will use in the plant must be established, and
the supplier tolerance which the company ADC will be as~ed to use for
inspection must also be set. Taguchi determines these tolerances based
on the cost of rejection at the two places. Because the cost of rejecting
a finished part is probably greater than the cost of rejection of an ingre-
dient,the tolerances for the manufacturer and the supplier win differ from
that of the customers. ;.
In addition to what is already known about the product (battery), two
more pieces of information are needed for this calculation:
cost of rejection/replacement at the manufacturer, and

cost ofrejectionlreplacementat the supplier.
Suppose that a part costing 20 cents from a supplier can produce a loss
of 50 cents to the manufacturer if the part fails. The loss equations will
produce tolerances of +0.63 volts and±O.4 volts for the manufacturer
and the supplier,respectively, as shown in Figure 4-2. Either the supplier
or the manufacturer can screen the products. To assure defect free prod-
ucts, the supplier mayscreentbem before they are shipped to the man-
ufacturer, who in turn passes them to the customers. When a supplier
doesn't screen, the manufacturer must. (The calculations shown in Figures
4-2 and 4-3 are obtained by using the computer software in Reference
H.)
EXERCISES
4-1. The Tagucbi method is considered a technique that helps build quality into a product
or process. Explain what aspect of quality it inftuences and how.
4-2. Compare .theroles of the Tagucbi method with that of .statistical process control
(SPC) in a manufacturing process. Explain how the Taguchi me!hod can inftuence
decisions in the SPC activities.
5
Working Mechanics of the Taguchi Design
of Experiments
5-1 fORMULAS FOR EXPERIMENT LAYOUT

It should be quite clear by now that the Tagucbi method is designed to
improve the quality of products and processes wberetheperformance
depends on many factors. In laying out a test and development strategy,
simple logic will usually be sufficient to establish all possible combina-
..... of factors along with allowable ranges of each of the factors in-
volvecLBut for ,llgineering projects involving many factors, the number
of posslole comfiinations is prohibitively large. In addition higher order
interactions amqfig the influencing factors may be needed for specific
pojects. A custQinary method of redUcing the number of test combina-
dons is to use what are known as partial factorial experiments. Tagucbi
constructed a special set of general designs for factorial experiments that
cover may applications. The special set of designs consist of orthogonal
arrays. The use of these arrays helps determine the least number of
experiments needed for a given set offactors. The details of using standard
(no~ modified) OAs in designing 'experiments for a given set of factors
is the subject of this chapter.
The OAs provide a recipe for fractional factorial experiments which
satisfy a number of situations. When a fixed number of levels for all
factors is involved and the interactions are unimportant, standard OAs
will satisfy most experimental design needs. A modification of the OAs
becomes necessary when mixed levels and interactions are present. Simple
designs with smaller number of factors, at fixed levels, will be discussed
first.
5-2 BASIC METHODOLOGY

The technique of laying out the conditions (designs) of experiments in-
volving multiple factors was first proposed by the Englishman, Sir R. A.
Fisher. in the 1920s. The method is popularly known as the factorial
.1.(}
----. --- - ._ ~ - -
Working Mechanics ofdle TaguchfDesign of Experiments 41
design of experiments. A full factorial design will identify all possible

combinations for a given set of factors. Since most industrial experiments
usually involve a significant number of factors y a !fuU factorial design
results in a large number of experiments. For example yin an exPeriment
involving seven factors yeach at two levels y the total number of combi-
nations will be 128 (27 ). To reduce the number of experiments to a
practical level, only a small set .from aD thepossibilities yis selected. The
method of selecting a limited number of experiments wbich produces the
most information, is known as a partial factorial experiment. Although,
this shorteut method is well known, there are no general guidetines for
its. application or the analysis of the results obtained by perfonning the
experiments. TaguchiYs approach complements these two importatJt areas.
First, he-clearly defines a set of OAs y eacb of whicb can be used for
many experimental situations. Second, he devised a standard method for
analysis of the results. The combination of standard experimental design
techniques and analysis methods in the Taguchi approachyproduces con-
sistencyand reproducibility rarely found many other statistical method.
Before discussing how Taguchireduces the number of experiments,
an understanding will be belpful of how all possible combinations result
from a set of factors.
Suppose that we are concemedabout one factor A (say temperature).
If we were to study its effect on a product at two levels (say 400 and
SOOOf) Le.,
Factor A: Level 1 = Al (4O<fF)
then two e~riments become necessary, one at temperature AI ythe other

at A 2 •
Consider now 2 factors A and B. each at 2 levels (A., A 2 and B., B2 ).
This produces 4 combinations since, when A is held at AI, B can assume
values Bland B 2 and when A is held at A 2 , B can again be set at values
BI and B2 •
Symbolically these combinations are expressed as fonows:
With 3 factors, each at 2 levels, there are 23 (8) possible experiments

as described in the previous section. If A. B and C represent these factors~
the 8 experiments can be expressed as:
41 A Primer OD the Taguchi Method
A1B1C., A 1B 1C2 , A 1B 2C 1• AIB~2'
A 2B 1C h A 2B 1C2 , A-JJ2CJ, A 2B 2C].,
Using the above gener~l rule. the total number of experiments possible
for different numbers of factors at 2 or 3 levels and the corresponding
suggested Taguchi number of experiments is shown in Table S-l.
A factorial experiment of 7 factors, at 2 levels eacb, with 128 possible
combinations is represented by Figure 5-I(a). The factors are represented
by the letters A, B, C, D, E, F, and G. The subscripts land 2 (Fig. 5-
I(b» represent the value of a factor at level I and 2 respectively. Eacb
of the 128 cells corre.sponds toa unique combination of the factors. Cells
T-I through T-8 indicated the 8 trial numbers defined by Taguchi's partial
factorial OA (Fig. 5-1(b» for this experiment.
Tagucbi has establisbed OAs to describe a large number ofexperimental
situations. The symbolic designation for these arrays carries the key
informatioo oll.,tbe size of the experiment. The array of Figure 5-1(b) is
designated as.8 or La. The number 8 indicates 8 trials are required.
The next lower size of the OA is L4 • An L4 experiment requires 4 trial
runs. The arra,;,ttandles up to 3 factors at 2 levels eacb. To fit a situation
with factors between 4 to 7, all at 2 levels, an La will be used. For
situations demanding a larger number of factors, bigher levels as well as
mixed levels, a number of other OAs are available. (Ref. 4)
Experiment designs by OAs are attractive because of experimental
efficiency, but tbere are some potential tradeoffs. Generally speaking,
OA experiments work well when there is minimal interaction among
factors,- Le., tbe factor influences on the measured quality objectives
Table 5-1. Comparison of Factorial Design and

Taguchi Design
FAC'l'PIUAl. OESIGN TAGOOn

FACTORS LEVEl. TOTAL NUMBER OF EXPERIMENTS
2 2 4(2~ 4
3 2 8 (23) 4
4 2 16 (ZJ) 8
7 2 128 (21) 8
15 2 32,768 (215) 16
4 3 81 (34 ) 9
Working MecbaDics of the Tagucbi Design -of Experiments 43
At ~
Full
Factorial 8, ~ 8, ~
Experiments
C, C2 Ct C2 C, C2 C, C2
6 t T-1
Ft
62
Et
6t
F2
62 T-3 ;.
0,
6t
F,
62 T-5
E2
6t T-7
F2
62
6t
Ft
62 T-8
E,
6t T-6
F2
62
O2
Ft
6, . T-4
62
E2
6t
F2
6 2 T-2
la) Experiment Structure
~
1 2 3 4 5 6 7
Trial
Number . A 8 C D E F 6
T-1 1 1 1 1 1 1 1
T-2 1 1 1 2 2 2 2
T-3" 1 2 2 1 1 2 2
T-4
T-5
1
2
2
1
2
2
2
1
2
2
1
1
. 1
2
T-6 2 1 2 2 1 2 1
T-7 2 2 1 1 2 2 1
T-8 2 2 1 2 1 1 2
lb) Trial Runs and Conditions
Figure 5·1. Experiments using an La array.

44 A Primer on the Taaudri Method
are independent of each other and are linear. In other words when the
outcome is directly proportional to the linear combination of individual
factor main effects, QA design identifies the optimum condition and
estimates performance at this condition accurately. If however, the fac-
tors interact with each other and influence the outcome in a nonlinear
manner, there is still a good chance that the optimum condition will
be identified accurately, but the estimate of performance at the opti-
mum can be significantly off. The degree of inaccuracy in performance
estimates will depend on the degree of complexity of interactions among
an the factors.
5-3 DESIGNING THE EXPERIMENT

The word "design" in the expression design of experiments, is used in a
general sense to convey "a mental project or a scheme in which means
to an end are laid down." To design the experiment is to develop.a scheme
or layout of thelfferent conditions to be studied. In engineering, the
word takes on a special me.aning when used as "a design," ~'product
design," or a "prOcess design." In these expressions, it refers to some
form of an engin&rmg communication, such as a set of specifications,
drawings, or physkal models that describe a concept. Consider a state-
ment like "The Taguchi design of experiments can be used to optimize
many designs." The final "design" in this sentence, obviously refers to
some engineering design process.
An experiment design must satisfy two objectives. First, the number
of trials must be determined. Second, the conditions for each trial must
be specified. Tagudri's arrays are versatile recipes that apply to several
experimental conditions. For example, the design for experiments in-
volving 4, 5, 6, or 1 factors, may all be accomplished by using the same
orthogonalarray (Ls). The OAs contain information on both the number
as well as the configurations of the experiments.
Before designing an experiment, knowledge of the product/process
under investigation is of prime importance for identifying the factors
likely to influence the outcome. In order to compile a comprehensive list
of the factors, the input to the experiment is generally obtained from all
the people involved in the project. Taguchi found brainstorming to be a
necessary step for determining the full range of factors to be investigated.
Consider an example.
Working Mechanics of the Tagucbi Desip of Experiments 45
Table 5-1. MoIding Proeess Fador and LeftI-

Example 5-1
VARJABW OIl FACTOItS LEVEL I
A. Injection Pressure A, := 2SOpsi Ail = 350 psi ..

B. Mold Temperature B, := 150 OP lit = 100 OP
C. Set Tune Cl ::: 6 see. Ct =9 see.
;..
Example 5-1
An experimenter has identified three controllable factors for a plastic
moldingprocess. Each factor can be applied at two levels (Table 5..2).
The experimenter wants to determine the optimum combination of the
levels of these factors and to know the contribution of each to product
quality .
The Design
There are 3 factors, each at 2 levels, thus an L4 will be suitable. An L4
OA with spaces for the factors and their levels is shown in Figure 5-2.
This configuration is a convenient way to layout a design. Since an L 4
N ';; u..
S
Q..
I ~ i
0)
-! l u..
Jl
j
~ -e
~
::I
CD
ti
jl Je I
Ilii
>0
.1.1 :sI
£(L o.
21-
I-
i
Repetitions
~Experiment
Number
1
1
1
2
1
3
1
1
30
2 3 ...
2 1 2 2 25
3 2 1 2 34
4 2 2 1 21
Figure 5-2, An experiment layout using L may.

.
46 A PIimer on the Tagudti Method
has 3 columns, the 3 factors can be assigned to these columns in any

order. Having assigned the factors, their levels can also be indicated in
the corresponding column.
There are four independent experimental conditions in an L 4 • These
condiool1s are described by the numbers in the rows. For an experienced
user of the technique, an array with factors assigned as shown in Table
5-2 comains all the necessary informatiOn. But for others, a descriptive
arrangement of the factors constituting different conditions of the exper-
iment _y be helpful. In this case the fom conditions can be spelled out
as follows:
Experiment No. 1 Injection at 250 psi, I.e., AI

Pressure
Maid at 150 of, i.e., 8 1
Temperature
Set Time at 6 sec., I.e., Cl
E~mentNo.2 Injection at 250 psi, I.e. , AI
Pressure
Mold at 200 Of, I.e. , B2
Temperature
Set Time at 9 sec., I.e. , C2
Experiment No. 3 Injection at 350 PSI, I.e. , A2
Pressure
Mold at 150 OF, i.e., BI
Temperature
Set Time at 9 sec., I.e. , C2
Experiment No. 4 Injection at 350 pSI, I.e. , A2
Pressure
Mold at 200 ~, I.e. , B2
Temperature
Sel-Time at 6 sec., I.e., Cl
5-3-1 Order of Running the Experiments

Whenever possible, the trial conditions (the individual combinations in
a designed experiment) rhould be run in a random order to avoid the
inftueoce of experiment setup. If only one run for each of the above
conditions is planned, they could be run as experiment number 2, 4, 3,
and 1, or in any other random order. If on the other hand, multiple
Wmting Mechanics of the Tagucbi Design of Experiments 47
repetitions are planned, say 3 runs for each of the 4 conditions, then there
are two ways to .proceed.
Replication .
In this approach all the trial conditions will be run in a random order.
One way to decide the order is to randomly pull one trial number at a
time from a set of trial numbers including repetitions. Often a new setup
will be required for each run., This increases the cost Df the experiment.
Repetition
Each trial is repeated as planned before proceeding to the next trial run.
The trial rul! sequence is selected in a random order. For example, given
the trial sequence 2, 4, 3, and 1, three successive runs of trial number
2 are made followed by 3 runs of trial 4 and so on. This procedure reduces
setup costs for the experiment. However a setup error is unlikely to be
detected. Furthermore, the effect of external factors such as humidity,
tool wear, etc., may not be captured during the successive runs if the
runs are short in duration.
5-3-2 Analysis of the Results

Although, a detailed analysis of the results will be discussed in Chapter
6, a brief description of and objectives of such an analysis is introduced
here.
Following the specifications as prescribed above" the experimenter
conducted the four trials. The moldedproducts were then evaluated and J
the results, in tenns of a quality characteristic, Y. were measured as
shown below:
Y1 = 30, Y2 = 25, Y3 = 34 and Y4 = 27

I
These results are recorded in the right most column of the OA (Table
5-3). Since, there was only one test for each condition, the results are
recorded in one column. For each repetition of the experiment, there will
be another column of reSUlts.
To speed up analysis, Taguchi has provided some key procedures.
When these steps are strictly followed by different individuals performing
the analysis. they are likely to arrive at the same conclusions. The oh-
,
I 48 A Prittter on the Tagucbi Method
Table 5-3. L. with Test Data or MoIding

Process Experiment
RESULTS
TRIALS A B C eY)
I I I I 30
2 I 2 2 2S
3 2 I 2 34
4 2 2 I 2'}
I jective of the analysis of the Tagucbi experimental results is primarily to

seek answers to the following three questions:
1. What is the optimum condition?

2. Which factors contribute to the results and by how much?
3. What will.be the expected result at the optimum condition?
The predicted result should be verified by running an experiment.
Computation of Average Performance

To compute the average performance of the factor A at level 1, i.e., for
AI' add results (Table 5-3) for trials including factor A I, and then divide
by the number .of such trials.
For A I, we look in the column for A and find that level 1 occurs in
experiment numbers 1 and 2. -The average effect of A., is therefore
calculated by adding the results, Y. of these two trials as follows:
AI == (Y1 + Y2)/2 == (30 + 25)12 == 27.5
~ The average effects of other factors are computed in a similar manner.
A 2 == (Y3 + Y4 )/2 == (34 + 27)/2 == 30.5

BI ~ (Y1 + Y3)/2 == (30 + 34)/2 == 32.0
B2 == (Y2 + Y4 )/2 == (25 + 27)12 == 26.0
Cl == (Y1 + Y4)/2 == (30 + 27)/2 == 28.5
C 2 == (Y2 + Y3)/2 == C!5 + 34)/2= 29.5
Working Mechanics of the Tagucbi Desip of Experiments 49
33
32
31
i:
0:28
27
26
/
251.---£...-----.l.------a.-.-__+-_........_-~
A1
Factors •
Figure 5-3. Main effects.

,
The average effects can also be plotted for a visual inspection as appears
in Figure 5..3. Frequently the term ~~factorial effecf' or "main effect" is
loosely substituted for "average effect." Strictly speaking, the factorial
,
i
effect is the difference between the two average effecl$ of the factor at
the two levels. For instance, the factorial effect of factorC is the difference
between average effect of Cl and C2 • .
5-3-3 Quality Characteristics

In the previous chapter the quality characteristics weae described as:
• the bigger the better

• the smaller the better
• the nominal the best
For the molding process example, higher strength afthe molded plastic
part is desired, and thus, "the bigger the better." From Figure .5-3, the
A 2 B 1 C 2 will likely produce the best result and therefore represents the
optimum condition except for the possible effect of iDteractions between
the factors.
In terms of the actual design factors, the probable optimum condition
becomes:
A2 Le., Injection pressure at 350 psi
B. I.e., Mold temperature at 150 OF
C2 1. e... Set time at 9 sec.

58 A Primer on the Tagw;:bi Method
Relative Contribution of Variables

The relative contributions of the factors are determined by comparing
their variances. The technique, popularly known as the analysis of vari-
ance (ANOVA) is used for this purpose. ANOVA will be covered in
detail in Chapter 6. Here the procedure is briefly introduced to complete
the analysis.
5..3-4 ANOVA Terms and Notations
The analysis of variance computes quantities known as degrees of free-·

dom, sums ofsquares, mean squares, etc. and organizes them in a standard
tabular format. These quantities and their interrelationships are defined
as shown below using the following notation:
v = Mean squares P - Percent contribution

(variance) T - Total (of results)
s =. Sum of squares N - Number of experiments
s' - Pure sum of squares C.F. = Correction factor
f - Degrees of freedom - Total degrees of freedom
e - Error (experimental)
F = Variance ratio
Variance
The variance of ~h factor is determined by the sum of the square of
each trial sum result involving the factor, divided by the degrees of
freedom of the factor. Thus:
VA = SAlfA (for factor A)

VB = Solfo (for factor B)
Ve = Selfe (for factor C)
Ve = Sife (for error terms)
Variance Ratio
The variance ratio is the variance of the factor divided by the error
vanance.
·I
FA = VAlVe
Fa = ValVe
Working Mechanics of the Taguchi Drsign of Experiments 53
For factors
fit = Number of levels of A - I - 1

fs = Number of levels of B-1 - I
fe = Number of levels of C - 1 == I
andtbe DOF for error variance is
fe == IT - lA - Is - le
= 3- 1- 1- 1
=0
Variance
VA = SAlfA == 9/1 = 9
Vs = Sslls = 3611 = 36
Ve = Scllc = 111 == 1
Ve = SJle = 010 indeterminate
Note that if the experiment included repetitions, say 2, then:
IT - 4 x 2 - 1 == 7
fe=7-1-1-1=4
where Se need not equal zero, depending on test results, and Vg need not
be zero.
Variance Ratio
FA == VAlVe is indeterminate since Ve = o. Similarly Fb and Fe are
indeterminate (Table 5-4). However. Ve can be combined (pooled) with
another small variance (Ve) to calculate a new error Ve which can then
be used to produce meaningful results. The process of disregarding an
individual factor's contribution and then subsequently adjusting the con-
tributions of the other factors, is known as Pooling. The detailed pro-
54 A·Jtimer OD the Tagucbi MeIbod
Table 5-4. ANOVA Table for Molding Process Experiment

F~ i S V F s p
At I «) 9 - - 19.62
B 1 36 36 78.28
C 1 1 1 2.10
Error 0 0 0
Totah 3 46 100.00%
cedw:e for pooling and the criteria used to detennine its use will be
disCUS5ed in Chapter 6 along with ANOVA.
Ca.iderpooling effects of the factor C. Then .the new error variance
is computed as.:
+ Se)/(fc + le)
Ve :::: (Se
= (l.0 + 0)1(1.0 + 0) ::::; 1

With Ibis new Ve , all sums of squares, S. can be modified as
etc.
The result then can be shown in the ANOVA table with the effect of
the factor C pooled. The pooled effects are shown as the error term in
the AHOVA table (last row of Table 5-5).
11It last column of the ANOV A table indicates the percent contribution
of the individual factor. In the example, factor B contributes the most,
76.<ln>. The contribution of A is 17.39% and that of C is not significant. t
Table 5-5. Pooled ANOVAfor Molding Process Experiment

SD.IRCE f 5 V F s p
I
Ai 1 9 9 9 8 17.39
B 1 36 -36 36 35 76.08
C ~---- r--pooled ----- --------- -------- ----
I EmJF 1 1 1 6.74
~ TotIIs 3 46 100.00%
I
Working Mechanics pr the Taguclli Design pf ExperimelJts 55
Note that the difference in the percentage. inftlJellC'eS of factors before

(Table 5-4) and after pooling (Table 5-5) is not 1atJe. To increase the
statistical significance of important factors, those 'firdors withsmaJI van-
ances should be pooled.
5-3-7 Projection of the Optimum Perfor~
Recall that for a 4'bigger is better" quality charaetaisti£, the ,study of the
main effect shows that the optimum condition is.IJ III C2 • It.happens to
be the third trial run. This is just a coincidence. Most of the time, the
optimum condition will not be one of the trial I1JftSbecause the Taguchi
experiment represent only a small set of the fuB factorial experiment.
The optimum is one of the trials defiJ:led by the full faetorialexperiment.
As a general rule, the optimum performance will. calculated using the
following e x p r e s s i o n . ,
I
I
,
T - Grand total of all results
N== Total number of results
Yapt == Performance at optimum condition
For optimum combination A2 B 1 Cl (which ~ to be experiment

number 3)
Yapt == TIN + (A 2 - TIN) + (8 1 - TIN) + (C2 - TIN)

== average performance + contribution ofAl , B I, and Cl .
above average Performance
In this example:
T == 116, N = 4, A 2 == 30.5, 8 1 == 32. C2 - 29.5
therefore,
Yopt == 29 + (30.5 - 29) + (32 - 29) + (19.5 - 29)

= 34.0
which is the result obtained in the trial run 3.

56 A Primer on the Tapcbi Method
When the optimum is not one of the trial runs already completed, this
projection should be verified by running a confirmation test(s). Confir-
mation testing is a necessary and important step in the Taguchi method
as it is direct proof of the methodology. Generally speaking, the average
result from the confirmation tests should agree with the optimum per-
formance (Yopt) estimated by the analysis. The correlation can also be
established in statislical terms reflecting the level of confidence, influence
of number of confirmation tests, etc. The procedure for calculation of
confidence interval of the optimum performance is discussed in Section
6~. .
5-4 DESIGNING WITH l\'IORE THAN THREE VARJABLES

In the preceding seaion, the layout of a simple experiment involving
only 3 factors, was discussed. In this section, designs for a larger number
of factors will be treated.. The designs wiUuse the higher order OAs.
5.4-1 DESIGNS WITH 2·LEVEL VARIABLES

Example 5·2
Design an experiment to investigate
4 Factors all at 2 levels,

5 Factors all at 2 levels.
6 Factors all at 2 levels.
or 7 Factors all at 2 levels,
Let these factors be A, B. C. D. E, F. and G and their levels A.. A2 , etc.
Experiment Desigu
As seen in the last example, the smallest OA Le., L 4 , can handle up to
3 factors. What do we do when there are more than 3 factors?
A list of commonly used OAs is shown in Table A-I. Notice that L g
can be used for 4 to 7 factors. Therefore, L g is suitable for any of the
above situations.
An L g has 8 trial conditions and 7 columns. To design the experiment,
factors must be assigned to appropriate columns and 8 trial conditions
must be specified.
r
Working M~banics of the Taguchi Design of Experiments 57
Table 5-6.• La with Seven 1 Level Factors

FACTORS A B C D E F G
EX.PEJUMENT
NUMBER COLUMNS .... I 2 3 4 S 6 7 RESUL1'S
I I I I I 1 I
I
2 I I I 2 2 2 2
3 I 2 2 I I 2 2 ;..
4 1 2 2 2 2 I I
S 2 I 2 1 2 I 2
6 2 I 2 2 1 2 I
7 2 2 I 1 2 2 1
8 2- 2 I 2 I 1 2
Since all factors have the same number of levels, the factor can be
assigned to anyone column. Thus factor A can be assigned to any of the
columns 1 through 7. Then B can be assigned to any ORe of the remaining
columns. You can also assign them in the natural ascending order, like
A in column I, B in column 2 and so on. If there are only 4 factors,
ignore the 3 unused columns.
The experimental conditions are defined by reading across the row of
the OA. An Ls OA, as shown in Table A-2, has 8 rows. Thus it represents
8 unique combinations of factors and their levels. An La with the factors
assigned to its columns is shown in Table 5-6.
From La trial number 3, is defined as
Trial number 3 122 1 122
Example 5-3
Number of factors == 8 through II
Number of levels for each == 2
Use array L 12
Design
Use L12 shown in Table A-3 for this example. Assign factors 1 thru 11
in the 11 columns available, in any order. Express the 12 experimental
conditions by using the 12 rows of the OA. Note that L12 is a special
51 A Primer on the Tagucbi Method
if
I
array prepared for study of the main effects only. In this array the inter- f'
i
I
action effects of factors assigned to any two columns is mixed with all
other columns which renders the array unsuitable for interaction studies.
(Use L 16 , L32 and L64 shown in Appendix A to design experiments with
higher numbers of 2 level factors.)
5-4-2 Designs with 3 Level Variables

Example 54
Number of factors = S through 13
Number of levels for each = 3
Use Lv (Table A-g)
5-4-3 Desips with Mixed Levels .Using standard Arrays

Eumple 5·5
COJ,lSidei;the experiments where there are 8 factors to be investigated.
To study the nonlinear effect, 7 factors were set at 3 levels. The remaining
one (I) factor was examined at only 2 levels.
Design
This is an example of the mixed levels. The L l8 array shown in Table
A-7(b) is used. L I8 has 8 columns, of which column 1 has 21evels, and
the rest have 3 levels. Obviously, the factor with 2 levels will be assigned
to column 1. The remaining 7 factors can be assigned to columns 2
through 8 in any desired manner.
5-5 DESIGNS WITH INTERACTION

~ The term Interaction, expressed by inse~iJlg " x " mafk between the two
interacting factors, is used to describe--a: condition in which the influence
of one factor upon the result, is dependent on the condition of the other.
Two factors A and B are said to interact (written as A x B) when the
effect of changes in the level of A, determines the influence of B and
vice versa. For example, temperature and humidity appear to have strong
interaction with respect to human comfort. An increase in temperature
alone may cause slight discomfort but the discomfort increases as hu-
Working Mechanics of the Tagucbi Design of Experimeals 59
Table 5-7. Layout for Experiment with Two 2 Level Faetors

(a) The Case with Interaction (b) The Case without Interaetioa
Ta T2 10TAL
HI 62 80 142 75 J42
H2 75 73 148 78 148
Total 137 153 290 Total 137 153
;..
midity increaseS. Assume the comfort level is depeadent only upon two
factors T and H, and is measured in tenns of numbers rangm,g .fromO to
tOO. If T and H are each allowed to assume levdsas Tt, Tt. BI.and
8 2, asswne that two sets of experimental data (widltbe same grand·totaI
of all observations) were obtained and arerepreseotedby TablesS,..1a
and S,..7b. The data are plotted in Figures 5-4a ai S4b. FII_ S-4a
shows an interaction between the two factors since lhelines cross. figure
S4b shows no interaction because the lines are parallel. If the lines are
not parallel the factors may interact, albeit weakly..
The graphical method reveals if interactionexist$.ltis calculated from
the experimental data. But howean we know wIletber the factors will
interact before we design the eXperiment? The T~hi methods do not
specify any general guidelines for predicting interactions. One has to
determine that by some other means, perhaps frota experieneeora pre-
vious Taguchi experiment.
80 80
76
64 64
60 L..-....r..- - 60 ' - - - L - _ - -_ _- - ' - -
With Interaction Without Interaction
(a) (b\
Figure 5-4. Main effects of factors T and H.

" A Primer OB die Tagucbi Method
Experimental design using Taguchi OA's is simple and straightforward

when there is no need to include interactions. It requires a little more
care to desip .an .experiment where interactions are to be included. In
Taguchi OAs the effect of interactions are mixed with the main effect of
a factor assigned to some other column. In the L 4 shown in Table 5-&
with factors A and B assigned to columns 1 and 2, interaction effects of
Ax B will be contained in column 3. If the interactions of A x B were
of no interest. a third factor C can be assigned to column 3 (Table 5..8).
II
The effect of mteraction A x B will then be mixed with the main effect I
of factor C. I
<
The interacting pair ofcolumns alona with the column where interaction
is shown, COMtitute an interactina triple. In case of an L 8 OA, the in-
teraction of ooIumns I and 2 goes to column 3• Which also means that
the interactioa between 2 and 3 is reflected in column 1 and that between
3 and 1 sho. up in 2. What about interactions between other columns
of an L8 or aay other 2 level orthogonal arrays? What about interaction
betweeniWo columns of 3 level orthogonal arrays? These will be difficult
tasks foripradicing engineers to keep track of. Tagucbi spent much of
his rese_b determining relationships for interacting columns. His find-
ings regardinl which columns interact with which other columns, are
presented in a table called the Triangular Table of Interaction. There are
tables to suit different levels of OAs. A large triangular table as shown
in Table A-6 made for 2 level columns will usually satisfy most commonly
used 2 levelorthogonal arrays.
Note that the first 1 columns of Table A-6, become the triangular table
for L 8 as shown in Table 5-9.
A triangular table contains inlormation about the interaction of the
various colunms ofan OA. The table should be interpreted in the following
way. The number in parentheses at the bottom of each column identifies
Table 5-8. L 4 with Two 2 Level Factors
EXPBUMENTSI AXB
COLUMNS A B C
I 1 1 I
2 1 2 2
3 2 1 2
4 2 2 I
Working Mechanics of the Taguchi Design of Experiments 61
Table 5-9. Interaction Between Two CoIl1IDD5 in - La Array

COLUMN: 2 3 4 S 6 7
(1) 3 2 S 4 7 6
(2) I 6 7 4 S
(3) 7 6 5 4
(4) I 2 3
(5) 3 2
(6) I
(7)
the column. To find in which column the interaction between columns 4

and 6 will appear, move horizontally across (4) and vertically from (6),
the intersection is "2'" in the tables. Thus the interaction effects between
columns 4 and 6 will appear at column 2. In a siInilar manner, other
interacting columns can be identified.
The triangular table facilitates laying out experiments with interactions.
. The table greatly reduces the time and increases the accuracy of assigning
propel' columns for interaction effects. To further enhance efficiency of
the experimental layout, Taguchi created line diagmms based on the
triangular tables. These diagrams represent standard experiment designs.
He calls such diagrams Linear Graphs. Linear graphs for Lt, ~ andotber ,.
2 levelorthogonal arrays are shown in Figures A-2 and A-3.
Linear Graphs .
Linear graphs are made up of numbers, dots and lines as shown in FIgure
5-4..1 where a dot and its assigned number identifies a factor, a connecting
line between two dotsindieates interaction and the number assigned to
the line indicates the column number in which interaction effects wilt be
compounded.
In designing experiments with interactions, the triangular tables are
essential; the linear graphs are complementary to the tables. For most
industrial experiments, interactions between factors are minor and the
1 2
0-----------0
3
Figure 5-4-1. Linear graphs for L4 •

Table 5-10. Factors for the Cake Baking Experiment

FACTORS. VAJUABl.E.S LEVEL 2
A. Egg 2 Eggs (AI) 3 Eggs (Arl

B. Butter 1 Stick (B I ) 1.5 Stick (B2l
C. Milk 2 Cups (Cl) 3 Cups (C2l
D. Flour I Extra scoop (D I ) 2 Extra scoops (~
E. Sugar I Extra scoop (El> 2 Extra scoops (E2l
triangular tables suffice. The following example shows how these two
tools are used for experimental design.
Example 5-6 IAf ~

1Y/~
In a baking experiment designed to determine the best reci~ for a pound
cake, five factq~ and their respective levels were identified as presented
in Table 5-10. ,i!~
Among the~Ffactors, milk (factor C) was considered to interact with
eggs (A) and titter (B). An experiment was designed to study tile in-
teractions A x~t: and B x Cin addition to the main effects of factors
A, B, C, D"j and E.
5-5-1 Steps in the Design and Analysis

Degrees of Freedom (DOF)
Each of the five factors (A, B, C, D, and E) are to be studied at two
levels, therefore, each factor has a OOF of I (OOF = Number of levels
- I). The DOF for the interaction is computed by multiplying the DOF
of each of the interacting factors. Thus, the DOF for A X C is 1. Likewise
the OOF of B xC:=: 1. The total OOF for the 5 factors and 2 interactions
in this case is 7. The apprOlfriate Tagucbi array cannot have a OOF less
than the total DOF of the experiment.
Selecting the Right OA

The experiment under consideration has 7 OOF and therefore requires an
OA with 7 DOF, hence an OA with at least 7 columns. Since an L 4 has
3 columns its DOF is 3. An Lp, has 7 columns and 7 DOF. It possibly
Woddng Mechanics of the Taguclti Design of ExperimeJlts 63
can work. An L 12 has eleven 2 level columns of which only 7 are needed.
L 12 certainly will work, but would require 12 trial runs in contrast to 8
for Ls. The smallest OA that will do the job should be selected iD order
to minimize experiment cost and time. How about an Ll(,? That is too
large for a 1 DOF experiment. In this case an Ls isa good match. Would
an Ls always work for an expeJjment with 7 OOF? Not necessarily. It
will depend mainly on how many interactions are expected to be inves·
tigated. It will work for Example 5·6.
Column Assignment'
In designing experiments with interactions~the columns for iDteIac-
tions must be identified fu:st. We have two interactions, A x C and B
x e. The trick is to select positions for A x C and B x e, such
that there are free columns for each of the factors A, B, and e as weD.
This can be done by using the triangular table for a 2 level OA or the
corresponding linear graphs. Let us examine the lineargrapb(a) ofAg-
i ure A·2. C is common to A x e and B x e. Assign eto column 2,
~. )~ a vertex with two connecting lines. Notice column 2 is a vertex of the
JV triangle with sides 2..3-1 and 2-64. With C at 2. assign A
toeilber
column 1 or column 4 ~ B to any remaining column. If Ais .as..
signed to 1, and B to 4, then A x C becomes column 3 and B x C
becomes column 6.
I
Five columns have been used by factors A, B, C ~ and interaerioas A
x C and B x C. The remaini~g two factors, D and E can be assiped
to columns 5 and 7 in any order. Let us assign D to column 5 and Eto
column 7. With factors and interactions successfully assigned to the
available columns, an La is obviously suitable for the design.
Having total OOF less than or equal to that for the OA, is oot always
a guarantee that a design can.be accomplished. Suppose instead of in·
teractions A x C and B xC, interactions A x C and B x D were to
be investigated. The total DOF will still be 7, the same as Ls.Dy "ex-
amining Figure A-2, notice that both linear graphs (a) and (b) have a
I
common factor such as I, 2 or 4. Since the interactions A x C and B
x D do not have a common factor, an Ls cannot be used. The next
higher order array should be tried.
The experiment designed for Example 5-6 uses the L g OA with column
assignments as shown in Table 5-11.
M A Primer on the TagdCbi Method
Table 5-11. La for the Cake Baking Experiment

E
7 RESULTS
1 1 I I I 1 1 I 66
2 1 1 I 2 2 2 2 75
3 1 2 2 1 1 2 2 54
4 1 2 2 2 2 1 1 62
5 2 1 2 1 2 1 2 52
6 2 1 2 2 1 2 1 82
7 2 2 1 I 2 2 1 52
8 2 2 1 2 1 1 2 78
Total == 521
DeseripU. 01 COIDbinatioos
The 8 trial oonditilns, contained in Table 5-11 can be described for each
individual trial as;ihown in Tables 5-12 and 5-13, for trial runs I and 2,
respectively. Thes,ther trial runs can be similarly described. Note that
the numbas inthllcolumns, where interactions are assigned (columns 3
and 6), lie not used in the description of trial run 2. Normally the
interaction column need not appear in the description and thus is deleted
from the description of the trial runs (trial run 2 Table 5-13). The complete
design information and analysis for this experhnentare shown in Table
5-14.
Taille 5-12. Deseription of Trial NUMBER 1 (Cake Baking)
VA1UA.8LE (FACTOR) LEVEL
I Egg 2 Eggs (A,)

2 Milk 2 Cups (C,)
J A x C (Egg x Milk)
4 Butter 1 Stick (B,)
j
6
Aour
8 x C (Butter x Milk)
1Extra scoop f'l
1 Sugar 1 Extra scoopAE,)
r
••
Working Mechanics of the Tagucbi Design of Experiments
Table 5-13. Description of Trial NUMBER 2 (Cake Baking)

VAJUABLE (fAC'IOR>
65
If
1 Egg 2 Eggs <AI)

2 Milk 2 Cups (Cl)
4 Butter l.S Stick (B,)
S Flour 2 Extra scoop (Dz)
7 Sugar 2 Extra scoop(Ez}
Running the Experiment

The order in which a specific combination of experiments is run is U8-
I
affected by the consideration of the interactions. Conditions 1 through 8
should be done in a random order.. A minimum of one trial run per
condition must be performed. Repetition of trial runs and the order of
repetitions are constrained by time and cost. I
Quality Characteristic (Results) t
Eight cakes were baked, one for each of the trial runs of Table 5-11.
i
The cakes were then examined by several experienced bakers. Before
cakes were baked, evaluation criteria were established. It was agreed that
me I
the cakes were to be evaluated not only for taste, but also for appearaace
and moistness. It was decided thattbe cakes were to be rated on a scale
of 0 to 100, using a scheme to reftect the weighting of each individual
attribu~ of the characteristic. For each condition, the average of the
evaluations by the bakers were recorded as shown in the column marked I
results (Table 5-11). Based on the definition, a higher value ofthe results
was considered favorable. For the purpose of analysis, this constituted
"the higher the better" type of quality characteristics. I
Analysis of Results
The analysis of data including interactions follows the same steps as
are taken· when there is no interaction. The objectives are the same:
(1) determine the optimum condition, (2) identify the individual inftu-
ence of each factor, and (3) estimate the performance at the optiImun
condition. The methods for objectives 2 and 3 are the same as before.
For the optimum condition, interactions introduce a minor change in
the manner in which the levels of factors are identified. To develop a
" A Primer on the Tagucbi Method
Table 5-14. Analysis of the Cake Baking Experiment

(a) Main Effects
LEVEL 1 LEVEL 2
COLUMNIFACTOR FACTOR AVERAGE AVERAGE
1 A Egg 64.25 66.00 1.15

2 C Milk 68.75 61.50 -1.25
3 AxC Ix 2 67.75 62.50 -5.25
4 B Butter 56.00 14.25 18.25
5 D Flour 10.00 60.25 -9.15
6 BxC 2 x 4 64. SO 65.15 1.25
7 E Sugar 65.50 64.15 -0.15
(1) ANOVA Table
COLtJMNSI SUM Of
fAC'l'OR FACTORS Of SQUARES VARlA1'fCE f PI!RC£NT
I A Egs 1 6.125 6.125 5.44 0.49

2 C Mill, I 105.125 105.125 93.44 10.13
3 AXC In.tionlx2 1 55.125 55.125 49.00 5.26
4 B Bup I 666.125 666.125 592.11 64.16
S D Flour I 190.125 190.125 169.00 18.41
6 B x C Interaction 2 x 4 I 3.125 3.125 2.77 0.19
7 E Sugar (I) (l.13) Pooled
All others/error 0 1.13 1.13 0.18
Total: 1 1026.880 100.00
(c) Estimate of Performance at the Optimum Condition of DesignlProCess:

Characteristic: Higher the Better
FAC'I'OR DESCRIPTION LEVEL DESCRIPIlON LEVEL NUMBER CONTRIBUTION
Egg 3 Eggs 2 0.815

Milk 2 Cups 1 3.625
Butter 1.5 Sticks 2 9.125
Flour I Extra scoop I 4.815
Sugar I Extra scoop 1 0.315
Contribution from all factors (total) ....................•.•.•.•....•.. 18.815

Current grand average of performance ........•..... '. 65.125
Expected result at optimum condition }84.000
Working Mechanics 01 the Tagucbi Design of Experiments 67
clear understanding of how the optimum condition is selected, the main

effects are discussed here in detail. (The details of ANOVA will be
covered in Chapter 6, only the results qf a computer analysis will be
.presented.)
The average effect of level I of the factor in column I (the effect of
2 eggs) is computed by adding the first 4 trial results of Table 5·11 and
dividing the sum by 4. Note that for trials 1 to 4, factor A (eggs) is
assigned level 1 (2 cups). Thus each of t!lese trial runs contains the effect
of factor A at level 1 (AI). The average effect of AI, therefore, is found
by averaging the results of the first four experiments. The notation Al
with a bar is used for this value. Thus, I
AI := (66 + 75 + 54 + 62)/4 := 64.25 j

.
Similarly the average effect of level 2, of A, is obtained by the last four
trial runs, since these were runs with factor A at level 2. Hence:
II
A2 + 82 + 52 +
:= (52 78)/4 := 66.00
I'I
f
Similarly
CI = 68.75
Cl := 61.50
and, I
(A X C)I = 67.75
(A x C)l = 62.50
The calculations for each factor and level are presented in Table 5-14(a).
The difference between the average value of each factor at levels 2
and 1 indicates the relative influence of the effect. The larger the differ-
ence, the stronger the influence. The sign of the difference, obviously
indicates whether the change from level 1 to 2 increases or decreases the
result. The main effects are shown visually in Figure 5-5. Figure 5-6
shows the interaction effects of A x C and B xC.
78
74
~ 70
c
j66 ~
62
58
Figure 5-5. Main effects.
Ignoring interaction effects for the moment, .notice that Table 5-14(a)
and Fig. 5-5 .show an improvemen~ at level 2 only for factors A and B
while level 2 effects for C, D and E cause a decrease in quality. Hence
the optinmtn levels for the factors based on the data are A 2 B 2 C. _. and
EI' Coincidentally, trial run number 6 tested these conditions and pr0-
duced the highes&;result (Table 5-11). Since interaction is ignored, the
average effects (;.(A X 4)1.2 and (AJ x C)I.2 shown in Table 5-14(a)
are not used in diermining the optimum.
~
5·5·2 lDteraction Effects

To determine whether the interaction is present, a proper interpretation
of the results is necessary. The general approach is to separate the influ-
ence of an interacting member from the influences of the others. In this
! !
lA x CIf 18 xCII
Interactions
Figure 5-6. Interaction effects.

-----~~~- ._--- -_•. - . ~ _ _--------------------~.
...
Working Mechanics of the T88Ucbi Design of ExperimenlS " ,

example A x C and B x C are the interactions with C common to both.
The infonnation about C can be extracted from the columns assigned to
factors A, B, and C, at the two levels of C. The steps involved are
described below.
I
The (AICIJ is first found from the results that contains both AI and Cl.
Note that (AIC I ) is not the same as the average value in levell of Table
5-14(a) for interaction (A X C) assigned to column 3 of Table 5-11.
I
I
I
This value is (A x C)I' In this analysis interaction columns, i.e. t columns I
3 and 6, are not used. Instead columns of Table 5-11 whicb represent I
the individual factors are used. Examination of column 1 shows A I is
contained in rows (trial runs) 1, 2, 3 and 4, but Cl is in trial runs 1, 2,
7
5, and 6. Comparing the two the rows that contain both A I and Clare
7
1 and 2. Theremre, AIC I comes fromtbe results of trial runs I and 2. _

The average effect of (AIC I ) = (66 + 75)/2 = 70.SO. The two
common trial nms for A I C2 are 3 and 4, and tbeaverage effect of (A I C21
== (54 + 62)/2 =58.00
In the calculations for AIC I and Aiel, factor level AI is common. The
difference between result 70.50 for AIC I and 58.00 for AtC2 is due only
to factor C.
Similarly A~., A~2' BIC.,B.Cz, BzC., and BzC2 are calculated as
shoWn below.
,
(Aiel) = 70.50 (A;c.> = 68.50
(A IC2 ) =58.00 (A 2C2 ) = 65.00
(BICI ) = 59.00 (B2C.) = 78.50

(B I C2 ) = 53.00 (JJ 2C2) = 70.00
These results can be easily visualized by inspection of Figure 5-7.
The intersecting lines on the left represents the interaetionbetween A
and C. The parallel lines on the right show that B and C probably do not
interact. Recall that in Table 5-14(a), the average influence of interaction
(A x C)I.2 assigned to column 3 was - 5.25. Further analysis for the
significance of this influence is made possible by the ANOVA table in
Table 5-14(b) which shows that the interaction A x C (column 3) is
5.26%, compared to the individual main effects of butter (B) 64.76%
and flour (D) 18.4%, etc.
To summarize, the suspected interaction between factors B and C was
.
. I
7. A Primer OD the Taguchi Method
82
78
8,
c2 Ct
. Fattors
Figure 5-7. Test of interactions.
found not to exist. The suspected interaction between A and C does exist;
its value is 5.26% t based on ANOVA.
Now to ~e the optimum condition determined. onI):. from the
factors A 2 , Cl' B~, D I , and El, we see from Figure 5-7 that A IC I has a
high~. value tbal?t~ I' Thus the optimum condition includes levels A 2
and Cl' The Jtew optimum conditions become A I B2 Cl D I El' However,
the performance at the new optimum should be compared with the original
optimum before the final determination of the interaction effects.
Consider the initial optimum which excluded the effects of interaction
(condition A 2 Cl B 2 D1 El). Using T = average result of 8 runs (Table
5-11) = 521/8 = 65.125. Compute the optimum performance using data
from Table 5-14(a):
fapt = T + (A 2 - T) + (Cl -:- T) + (B 2 - T)

. +(D I - T) + (El - T)
= r + (66.00 - T) + (68.75 - T) + (74.25 - T)

+ (70.00 - T) + (65.50 - T)
= 65.125 + 0.875 + 3.625 + 9.125 + 4.875 + 0.375

= 65.125 + 18.875
= 84.000
~-_..._~--_._- ~-..... - ~ -- ---- . --.--~.~ -'--~------------
Working Mechanics of d1e Taguchi Design of Experiments 71
Similarly for the revised optimum, considering interaction A I Cl B2 D I

El we compute:
\
Y. = f + (AIel - n + (.82- f) + (DI - n + (El - n
== T + (70.5 - + (74.25 - n
T)
+ (70 -- T) + (65.5 - T)
;. 65.128 + 5.375 + 9.125 + 4.875 + 0.375
== 84.875
Note that when the estimate of performance at the optimurnconditi<m

includes the interactions between A and C, the net result is obtained from
the combined effect of AIC I .
The second calculation yields a higher value, confirming that the re-
vised optimum is better.
I
,
Asa final check examine the interaction between B and c.. The sec-
ond pair of lines in Figure 5-7 representstbe effect of C at fixed levels
of B. The iines are almost parallel thus indicating little interaction. The
ANOVA calculations presented in Table 5-14(b) show a small inter-
action (0.19%). Observe that the highes~ value for the pair of lines
corresponds to C IB 2 • Comparing C IB 2 to the revised optimu.m condi-
tion; we find that C IB2 are included. Thus. the interaction B X C has
no intluence on the optimu.m. The optimum condition remains as re-
vised for interaction A x C and no further modification is needed.
Optimum condition == A I Cl B 2 DIEI·

E~ted performance at optimum condition == 84.875
I
5-5-3 Key Observations
• In designing experiments with interactions, Triangular Tables or Lin-
ear Graphs should be used for colunm assignments. To select the
appropriate 0 A, the types of interactions and their degrees offreedom
will have to be considered.
• For the purpose of analysis, interactions are treated as are any other
factors, however their presence is ignored for the preliminary deter-
72 A PIinJa' on the Tagucbi Method
miuarion of the optimum condition. The relative significance of in-

tm.ctions is obtained from an ANOVA study.
• "I1Ie determination of the effect of interactions requires a separate
study. Such study may suggest a change in the optimum condition.
• WIJen several interactions are included in an experiment, the level
selection may become extremely complex.
How sIlould interactions be handled? Should an extra variable be in-

cluded 10 study the interaction? Ifconstraints necessitate a choice between
includiBI an ex~ variable or studying an interaction, Taguchi recom-
mends. -. . wide, not down." When there is an extra column, study a
new VIIiaIie. not an interaction. On the second pass, if there are strong
feelinp ahoot the interactions, then they should be included.
5-5-4 Mere .Desips with Interactions

Example 5-7
Design .expe(inent with 5 factors at 2 levels each and two interactions.
VariaIJIa; A, B, e, D, E
IntemdioDs: A x B and e x D
Desiga
The 5 _tors and the 2 interactions each have one OOF. Thus the total
OOF is 7. In this case an Ls-will not work, because the triangular table
(Tables 5-9 and A-6) shows that there is only one independent triplet in
the first 1 columns. In other words, if A and B are assigned to columns
1 and 4 column 3 will be reserved for A x B. This means that columns
4, 5,' and 7 remain for factors C, D, E, and interaction C x D.
HowelU any combination of these columns 4, 5. 6, and 7 contains only
the values I, 2 or 3, hence their interaction would involve columns
previously assigned to A, B and A x B \ ,The next higher order OA is an
L 12 • DuI it is a special array where interaction effects are distributed and
thus cat:l be used to study interactions.. An L l6 OA is the next higher
order anay. Examination of the triangular table (Table A-6) will show
that it CID be used.
Using Table A-6, arbitrarily assigns:
------ ----- ,
Working Mechanie~ of the Taguclli Design of Experitnents 73
A to column I C to column 2
B to column 4 D to column 8
Then A x B is column 5~ and C x D is column 10
and E is column 3
There were many ways to achieve this column assignment. In this case
using Table A-6, A and B were arbitrarily assigned tocolunms I and 4,
with column 5 reserved for interaction A x B. C and D were then assigned
to two unused columns such that C x D becomes a column whicbis IS
or less and not previously assigned. Columns 2, 8, and 10 were the
interacting group of columns selected for factors C, D, and C x D,
respectively. The factor E was then assigned to one of the remaining 9
columns, column 3. The final design is sbown in Table 5·:15. Note that
columns 6, 7, 9, 11 to IS are unassigned and not used in the experiment.
Table 5-15. Lift Design with Five Factors and Two Interacticms
A C
x X
Ll6 A CE BB D D
EXPEIUMENTSI
COLUMN I 2 3 4 S 6 1 8 9 10 11 12 13 14 15
1 I 1 1 I 1 0 0 1 0 t 0 0 0 0 0
2 1 1 1 1 1 0 0 2 0 2 0 0 0 0 0
3 1 1 1 2 2 0 0 1 0 I 0 0 0 0 0
4 1 1 1 2 2 0 0 2 0 2 0 0 0 0 0
5 1 2 2 1 1 0 0 1 0 2 0 0 0 0 0
6 1 2 2 1 1 0 0 2 0 1 0 0 0 0 0
7 1 2 2 2 2 0 0 I 0 2 0 0 0 0 0
.., 0
8 1 2 2 2 2 0 0 - 1 0 0 0 0 0
9
10
2
2
1 2
I 2
1 2
1 2
0 0
0 0
.,I 0
0
1 0
2 0
0 0
0 0
0 0
0 0
1 2 0 0 0 0
11 2 2 1
.,I 0 1 0 0 0
12 2 1 2 2 1 0 0 - 0 2 0 0 0 0 0
2 1 2 0 0 0 2 0' 0 0
13
14
2
2 2
1
I I 2 0 0
.,1 0 I 0 0 0
0 0
0 0
15 2 2 1 2 I 0 0 0 2 0 0 0 0 0
.,1
16 2 2 1 2 1 0 0 - 0 1 0 0 0 0 0
.
Example 5-8
Design an experiment with 9 factors at 2 levels each and 5 interactions
as described below:
Variabks: A,B,e,D,E,F,G.H,J
Interactions: A x B, A x e, A x· E, A x F,and
BXD
Design
Tbe 9 factors and 5 interactions together have 1400F. An L16 OA has
15 OOF and is a good candidate. Because of the number of interactions,
a linear graph of L 16 is helpful. Since the factor A is common to 4 of the
5 interaetions,a linear graph with a hub will be used. In Figure A-3, the
lower left diagram can be adapted for the design by selecting columns
of interest as shown in Figure 5-8. Start by assigning A to column 1.
Then select end!" of four spokes for B, e, E, and F as shown in Table
5-16. If B is assigned to column IS, then D will be column 8 and B x
D column 7. Therefore when all the 5 interactions are assigned to the
appropriate columns, the remaining factors can be assigned to the avail-
able columns at random.
More examples ofexperiment designs are described in the later chapters
and specifically in Chapter 9.
B BxO 0
15 7 / 8
/
/
/
/
A/ Axe ~
C
11 10
E F
2 4
Figure 5-8. A linear graph for Example 5-8.

W",k,"1(. Mcchanic..""S of the Tl1guchi Desip of &periments 75
Tabtt' ~.t". 1".l",,11I1I wllh NIne 1 Level FaetlJ<S and Fm IIderadIoDS

A
,
A
;
F-
x
,! A
x
A
X
,. , ,- C} D D H CC I B B
L,...
A
"
EXPERtM' Nt ..1 J
, .. ~ 6 7 8 <) 10 tI 12 13 14 IS
-----
COlJIMN
I ,
I I
I
I
I
I
I
J
I
I
I
t
t
I
2 ...,
I
2 2
I I
2 2
I I I
2 2
2 2 2 2 t t I I 2 2 2 2
I 2
I 2 2 2 2 2 2 1 I 1 1
:l I' I 2 2
I ;..
4
I 2 2 t t 2 1 I I 2 2
2 2 I
I I 2 2 2 2 t I 2 2 I I
~ I
I 2 '2
tt I t I t 2 .1 2 2 I t
2 'l 2 2
I I t 2 2 I I I 1 2 2
7 ], 2 2 2
I
K I 2 I .1 I 2 I 2
, t J I 2 t 2 -
()
·.'• I .., I 2 I 2 2 I 2 I 2 I 2 I
to 2 I I 2 1 2 2 I 2 I
} I
') 2 t
",.
I 2 I 2
11 , I '1 2 I 2 I 2 I 2 I
12 · 2 I I ..., 2 I I 2 2 I
n .. ) 2 2
2 2 I 2 I t 2 2 I I 2
l '2
14 I 2 I 2 2 I 2 I I 2
....,
2 1 2
1-" .', J
...') 1 2 J 2 t I .1 I 2 2 I
Ht ·
5-6 .},,~o;.(;NS WITII MIXED FACTOR LEVW

D!:5ig • wUho,1I inlerpelion. and with all faclors al2 ~ are of the
n kim , They PR' lhe leasl clImbersoroe and DlOSloften can he
simplL'f l
d . '11(<1 hy nl<'lIl1-< .)f lhe slandard OAs. BlIllben: are mally occasions
;;~ mole IhplllWO levels will have 10 be iIlclu~d. hllbebaking process
experinW • IIlhird level,?f buller ~say 1.25 suck, a ~ between4 t and
he ."•..:iflc<l. • he expenroenl would c~ of factots al 2
e''"'"
nl
t.5)
level. 1111,1 onc III 3 level•. Now we ~ve 4 factors \hat are of 2 levels
and 0'1<' Ihlll hll' 3 levels. Whal expenment design is appropriate?
Wh\' nt" ,n,''''' Ih"n 21e vcls neces~? First•.3levels may be ilJlport8I1t
10 ,,")1<' h,d"'s. (',ms llkt "n expenroent desIgn of.8 molding process
which uses plastic pellet feed stocks from 4 different vendors; or 4 dif-
ferent specmcatioos may include a study of materials at 4 vendor supplied
specifications. The factor (material), in .such cases, will have 4 levels.
Another likely reatm for more than 2 levels is that the influence of a
factor on the result is suspected to vary nonlinearly. Considering only 2
levels will give a liDar output. Nonlinear behavior can only be determined
by a tlllrd level as sItown in Figure 5-9.
There are some standard OAs which treat mixed level factors. But,
they may not be ~ most economical or may not even suit your needs.
For most applica'" involving mixed levels, Tagucbi modifies the stan-
dard arrays to fit lie circumstances. By following his prescription, a 2
level column can. upgraded to a 4 or an 8 level column; a 4 level
column can be upttaded into an 8 level column. On the other hand, a
column can also be downgraded by lowering the number of levels, say,
from 4 to 3. The mr.tItod of reducing the levels is done by what is known
as Dummy Tre--..
Before consicleriulcolumn modifications, some additional words about
OOF are appropriate.. Recall that the OOF for a column iS'its number of
levels less I. Tb..". 2 level column has 1 OOF, a 3 level column has 2
OOF and a 4 level column has 3 OOF.
Thus to create a 4 level column, three 2 level columns are needed to
provide the sameOOF. To change one column of an L 8 to a 4 level
column, 3 columns are combined. Similarly, to change a column of an
L 16 into an 8 level column, seven of the fifteen 2 level columns are
combined.
Factors
Figure 5-9. Main effects of a factor with two and three levels.
".-------------------------------------------
S-6·1 Preparation of a 4 Level Column
A 4 level column is easily prepared from three 2 level columns. To

demonstrate, consider an La. The procedure will also apply for all 2 level
OAs.
Steps
1. From the linear graph for L s, select a set of three interacting columns
(Fig. 5-10). Example: columns I, 2, and 3.
2. Select any two columns. Suppose 1 and 2 are selected.
3. Combine the two columns row by row, by following the rules of
Table 5-17, to get a combined column such as is shown in Table
5-17. Replace the original columns 1, 2, and 3 by the new colu.nln
which has Just been prepared.
ExampleS-'
Design an experiment to accommodate one factor at 4 levels and four
others at 2 levels each.
Variables: A, 0, C, 0
Interactions: None
Levels: A == 4; B, e, D == 2
The Design
Factor A has 4 levels and 3 DOF. The other four 2 level factors each
have 1 DOE The total DOF is 7. Ants OA, sbov-'Il in Table 5-18, which
has 7 OOF appears suitable.
1 2
4 I
Figure 5-10. Groups of interacting columns for level upgrading.
78 A PrUner 00 the Taguchi Method
Table 5-17. Rules for Preparation of a

4 Level Column
LEVEL OF l£VEL OF LEVEL OF

FIRST COLUMN SECO!\lD COLUMN COMBf?Ioc TO FORM NEW cm.UMN
I I -+ I
I 2 -+ 2
2
2
I
2
-+
-+ .
3
BuildiDg Columns
The first dJree cofumns of an La can be combined 10 produce a 4 level
column following the procedure previously described. .
Step I. Start with an original Lt. and select a set of three interacting
columns, say l.~. . 2, and 3.
Step 2.lgnol column 3 (Table 5..19).
Step 3. Combine column 1 and 2 into a new colUlll1L Follow the
procedure as shown by Tables 5-20 and 5-21.
Step 4. Assign the 4 level factor to this new column and the others to
the remaining original columns as shown in Tables 5-22 and 5..23.
The experimental conditions and the subsequent analysis are handled
I in a manner similarly to the techniques described before.
Table 5-18. La Array

COLUMNS
1lUAL
NL"MBER 2 3 4 5 6 7
~
I I I I I I I I
2 I I I 2 2 2 2
3 I 2 2 I I 2 2
4 I 2 2 2 2 I I
5 2 I 2 I 2 J 2
6 2 I 2 2 I 2 1 \
7 2 2 I 1 2 2 I
8 2 2 I 2 I I 2
Working Mechanics of the Taguchi Desip of Experiments 79
Table 5-19. La Undergoing Column Upgrade

EXPEIUMENTS
COI..UMN 1 I 2 3 4 5 11 7
1 1 1 1 1 I 1
2 1 I 2 2 2 2
3 1 2 1 1 2 2
4 1 2 2 2 I 1
5 2;.. 1 1 2 I 2
6 2 1 2 1 2 1
7 2 2 1 2 2 1
8 2 2 2 1 I 2
5-6-2 Preparation of an8 Level Column

An 8 level column can be prepared by combining a set of seven 2 level
columns ofanL 16 OA. The procedure is similar to thecmeused in creating
a 4 level column. First we need to identify the seven collJmnsinvolved.
Then combine the columns using some establisbedguiddines.
Step 1. Select the set of seven columns.

One such set of seven columns consists of columns I. 2, 3, 4, 5, 6,
and 7. The seven columns are an interacting set amoag three factors, A,
B, and C. If A, B, and C are assigned to columns 1, 2, . .4,respectiveJy,
the triangular table or the linear graph corresponding to an L 16, will show
that, the interaction columns are 3 for A x B, 5 for A xC, 6 for B x
C and 7 for A x B xC. Here A x B x C represt8tS the interaction
between factor A (column l) and interaction B x C (column 6). 1be six
columns may be represented as a closed. triangle inlhe linear graph and
Table 5-20. Rules for 4 Level Colu. .

Preparation
0l.D COLUMNS NEW cou.l-iN
1 1 ~ 1
1 2 ~ 2
"'I I ~ 3
2 2 ~ 4
80 A Primer OD the Tapebi Method
Table 5-21. Preparing a 4 Level Column of an La Array

COU"MN-
EXPaUMENTI
r-.1:MBER 2 3 4 S 6 7
I -1-----1 > 1 I I I I 1
2 -1-1 > I 2 2 2 2 2
3 ;...1--2> 2 2 I I 2 2
4 -1--2> 2 2 2 2 I I
5 --2---1 >3 2 I 2 I 2
6 -2---1 >3 2 2 I 2 I
7 -2--2> 4 I I 2 2 I
8 -2----2> 4 I 2 I I 2
Table 5-21. Modified La Array with One 4 Level CoIuma

~
(."()lUMN IQiWCOLUMN 4 , 6 1
I I I I I I
2 I 2 2 2 2
3 2 I I 2 2
4 2 2 2 I I
5 3 1 2 I 2
6 3 2 I 2 I
7 4 I 2 2 I
8 4 2 I I 2
Table 5-23. ModIIed La with Factors Assigned (One 4 Level Column)

FACTORS-+
EXPERIMB\'TSiCOWMN
A
NEW COLUMN
B
4 ,
C D
6
E
7
I I ~ I I I I
2 I 2 2 2 2
3 2 I 1 2 2
4 2 2 2 I 1
5 3 1 2 1 2
6 3 2 1 2 I
7 4 1 2 2 1
8 4 2 1 1 2
2 ~---------4t------------... 4
6
Figure 5-11. Preparation of an 8 level column.
can be selected on this basis. A fourth line connecting the apex and the
base represents the interaction (A x B x C) as sbownin Figure 5-11 . I.
Step 2. Select the three columns to be used to form an Slevel column.

Select the three columns where the three factors, A, B, and C are
assigned. In general, select each apex of the triangle of the linear graph
for the set to represent the columns. These are columns I, 2, and 4 for
the set. The remaining four columns are eliminated, since the three col.. t
umns include the four interactive (A x B, A xC, B x C and A x B
x C),
Step 3. Combine three 2 level columns into an 8 level column.

Compare numbers in each row of the three columns and combine them
using the rules shown in Table 5-24. Note the rule isoot the previous
one for the 4 level array though it follows the same pattern.
Table 5-U.Rulesfof Preparation of an

8 Level ColUIDD fOf8ll,L t6 Array
NEW
FIRST SECOND TlUIID COLUMN
1 I 1 1
1 1 2 2
1 2 1 3
1 2 2 4
2 1 1 5
2 I 2 6
2 2 1 7
'2 2 2 8
Table 5-25. Converting LIt. to Include an 8 Level Column

Columns Combined to Form New Columns
LI6 NE\\' COLUMN
EXPERIMENTSI
,
COLUMN I 2 4 8 9 10 11 12 13 14 IS
I I ,zt I I 1 I 1 I I I I I
2 I 1 I I 2 2 2 2 2 . 2 2 2
3 I I 2 2 I I I . I 2 2 2 2
4 I I 2 2 2 2 2 2 I I I I
S I 2 3 I I 1 2 2 1 I 2 2
6 I 2 3 I 2 2 I I 2 2 I I
7 I 2 4 2 I I 2 2 2 2 1 I
8 I 2 4 2 2 2 I 1 I I 2 2
9 2 I S I I 2 1 2 I 2 1 2
10 2 I 5 I 2 I 2 I 2 I 2 I
11
12 , ;,~t i I
I
6 2
6 2
I
2
2
I
I
2
2
I
2
I
I
2
2
I
I
2
~;:J 2
13 2 7 I I 2 2 t t 2 2 I
14 2 2 7 I 2 I I 2 2 I I 2
15 2 2 8 2 I 2 2 I 2 I I 2
16 2 2 8 2 2 I 1 2 I 2 2 I
For the set of columns under consideration, the first, second, and third
are columns 1, 2, and 3, respectively (Fig. 5-11). The modified L l6 array
with its upgraded column is shown in Table 5-25. Note that the linear
graph (Fig. 5-11) represents seven columns consisting of three main
effects and four interactions. Thus combining the column representing
the three main effects includes the four interactions.
5-7 DUMMY TREATMENT (COLUMN DEGRADING)
Just as 2 level columns of QA can be combined to higher levels, so a

higher level column can be decomposed into lower level columns. The
method used is known as Dummy Treatment. I
Consider an experiment involving four factors A, B, C, and D. of
which A has only 2 levels. and all the others have 3 levels each. DOF
Working MecbaDics of the Tagucbi Design of Experiments 83
Table 5-26. Design with Degraded Column of L"

FACTORS- 8 C A D
I
EXPERlMENTSI
COLUMN 2 ) 4
1 1 1 1 1
2 1 2 2 1
3 1 3 l' 3
;.
4 2 1 2 3
5 2 2 l' 1
6 2 3 1 2
7 3 1 l' 2
8 3 2 1 3
9 3 3 2 1
(') indicates new lDOdified level
I' "" (level 3)
is 7. An 4 attay has four 3 level columns with 8 DOF.k could be used

if one column can be reduced to the 2 level for factor A and the three
remaining columns occupied by factors B. C• and D. In dummy treatment,
the third level of A == A) is formally treated as A), as if A) ex.ists. But
in reality A) is set to be either AI or A 2 •
The design with the modified column (3) of 4 is shown in Table
5-26. Factor A can be assigned to any column. Note that column 3
was. selected such that the modified level 3 = I' occurs once in each
group of three trial rons. This distribution enhances the experiment.
EX8IDple 5-10
In a casting process used to manufacture engine blocks for a passenger
car, nine factors and their levels were identified (Table 5-27). The op-
timum process parameters·for the casting operation are 10 be- determined
by experiment. Of the nine factors, two are of 3 levels each and another
of 4 levels. The remaining six factors are all of 2 levels each. The OOF
is at least 13 if no interactions are considered.
The Design
Since most factors are 2 level, a 2 level OA may be suitable. Each 3
level factor can be accommodated by 3 columns (modified) and the 4
level factor can also be described by 3 columns for a subtotal of 9 columns.
.t
84 A Prima' on the Taguchi Method
Table 5-27. Factors of the Casting Process ExperimeBt-Example S·10
VARIABLES l.EVEL I LEVEL 2 LEVEL 3 LEVEL 4
A: Sand compaction Plant X Plant Y Plant Z

B: Gating type Plant X Plant Y Plant Z
c: Metal bead Low High
D: Sand supplier Supplier 1 Supplier 2
E: Coating type Type I Type 2 Type 3 Type 4
F: Sand permeability 300 penn 400 penn
G: Metal temperature 1430F 1460 F
H: Quencbtype 450F 725 F
I: Gas level Absent High amount
The remaining six 2 level factors require one column each. Thus a min-
imum pf 15 columns is needed. L l6 satisfies this requirement. Nine col-
umns are to be converted to three 4 level columns, then 2 columns will
be reduced by dummy treatment to 3 level columns for this experiment.
Nonnally a 3 level column will have 2 DOF. But when it is prepared
by ,!ducmg a 4 level column, it must be counted as 3 DOF since it
inclttJes a dummy level. Thus, the total DOF fortbe experiment is:
6 Variables at 2 levels each 6 DOF

I Variable at 4 levels each 3 DOF
2 Variables at 3 levels each 6 DOF (Dummy Treated)
--------------
Total OOF ::: 15
L l6 has 15 DOF and therefore is suitable for the design. The three sets
of interacting columns used for column upgrading are I 23,48 12, and
7 9 14. The column preparation and assignment follows these steps.
t?~!,)
I. !!isc~~~~d use columns I and 2 to prepare a 4 level
00iWtiii: 'then d~ treat it to a 3 level column. Place it as column
1. Assign factor A (sand compaction) to this column.
2. Discard column 12 and use columns 4 and 8 to create a 4 level
column first, then dummy treat it to a 3 level column. Call the new
column, column 4. Assign factor B (gating type) to this column.
I 3. Discard column 14 and use columns 1 and 9 to create a 4 level

column for factor E (coating type). Call it column 7.
Table 5-28-1. Casting Process Optimization Design-Example 5-10

-,
Design Variables and Their .Levels
COWMN
NUMBS FACTOR NAME L£VEL I lEVEL 2 LEVEL 3
I Sand compaction Plant X Plant Y Plant Z

2 (Used with Col .I) MIU
3 (Used with Col I) MIU
4 Gating type Plant X PlantY Plant Z
5 Metal head Low High
6 Sand supplier Supplier 1 Supplier 2
7 Coating type Type t Type 2 Type 3 Type 4
8 (Used with Col 4) M/U
9 (Used with Col 7) WU
10 Sand perm 200 Perm 300 Perm
11 Metal temperature 1430 F 1460 F
12 (Used with Col 4) MIU
13 Quench type 450 F 725 F
14 (Used with Col 7) MIU
15 Gas level None Higb
Note: Modified Cols_ 1 ;! 3 4 8 12 and 7 9 14. No interaction objective: Detennine process parameter foe best
casting chl,lracleristic: 1k higher the better ISiq or od!er)
4. Assign the remaining seven 2 level factors to the rest of the 2 level
columns as shown in Table 5-28-1.
The detail array modified to produce two,3 level and one 4 level column
is shown in Table 5-28-2. Table 5-28-3 shows the modifications to create
~three 4 level columns and the dummy treatment two columns to 3 level.
Note that in new colun;tn 1, the four dununy levels l' occur together. In
this case, to avoid any undesirable bias due to level 1, the experimem:
should be carried out by selecting trial conditions in a random order.
Description of Experimental Conditions

Once the factors are assigned, the 16 trial runs are described by the rows
of the OA (modified). With experience, the run conditions are easily read
from the array. But for the inexperienced, and for large arrays. translating
the array notations into actual descriptions of the factor levels may be
subject to error. Computer software exists to reduce/eliminate chances
86 A Primer OD the Tagucbi Method
Table 5-28-2. Casting Process Optimization Design-Example 5-10

EXPERIMENrt
COLlJMN 2 3 4 S 6 7 8 9 10 II 12 13 1<4 IS
Expt 1 1 0 0 1 1 1 1 0 0 1 1 0 1 0 I
Expt 2 1 0 0 2 1 1 2 0 0 2 2 0 2 0 2
Expt 3 1 0 0 3 2 2 3 0 0 1 1 0 2 0 2
Expt 4 1 0 0 1 2 2 4 0 0 2 2 0 1 0 1
Expt 5 2 0 0 1 1 2 3 0 0 2 2 0 1 0 2
Expt 6 2 0 0 2 1 2 4 0 0 1 1 0 2 0 1
Expt 7 2 0 0 3 2 1 1 0 0 2 2 0 2 0 1
Expt 8 2 0 0 1 2 1 2 0 0 1 1 0 1 0 2
Expt 9 3 0 0 1 2 1 4 0 0 1 2 0 2 0 2
Expt 10 3 0 0 2 2 1 3 0 0 2 1 0 1 0 1
Expt 11 3 0 0 3 1 2 2 0 0 1 2 0 1 0 I
Expt 12 3 0 0 1 1 2 1 0 0 2 1 0 2 0 2
Expt 13 1 0 0 1 2 2 2 0 0 2 1 0 2 0 1
Expt 14 0 0 2 2 2 1 0 0 1 2 0 1 0 2
I
Expt IS 0 0 3 1 1 4 0 0 2 1 0 1 0 2
Expt 16 1 0 0 1 1 1 3 0 0 1 2 0 2 0 1
t
of such errors. A printout of the trial conditions for sample trial runs is
I
f
shown in Table 5-28-4.
Main Effect Plots for 3 Level and 4 Level Factors

The analysis of experimental data follows the same steps as before.
The results of a single test run at each of the 16 conditions are shown
in Table 5-28-5. The main effects of the factors are presented in Table
-; 5-28-6; the effects for the 3 and 4 level factors are displayed in Figure
5-12. The optimum combination is easily determined by plotting main
effects of all factors or from the data of Table 5-28-6 of the main ef-
fects, by selecting the higher values (since the quality characteristic is
"the bigger the better"). Note that for sand compaction the middle level
y produces the highest value. Such nonlinear behavior of the factor was
qV;~,r suse:cte~ from previous experience, hence three levels were selected
for the experiment.
Working Mechanics of the Tagucbi Design of Experiments 87'
Table 5-28-3. Casting Process OptinJization Design

(Column Upgrading Procedure).'
1 2 and 3 to form a- 4 8 and 12 to fQrm a 7 9 and 14 to form a

3 level col. "New I" 3 level col. "New 4~' 4 level col. "New T
2 3 NEW 1 4 8 12 NEW 4 7 9 14
1---1----1 1 1---1-1 I 1---1---1 I

• :t
1---1---1 I 2----2---1' ~.1t!Il 1---2-2 2
;.
1----1---1 1 2---1-2 3 2--1---2 3

1--1----1 1 2---2-1 4 = l' 2--2--1 4
1----2----2 2 1---1-1 1 2-1-2 3
1----2---2 2 1---2-2 2 2----2---1 4
1---2----2 2 2----1..--2 3 1---1----1 1
1---2----2 2 2----2--1 4 = l' 1---2--2 2
2----1---2 3 1----1---1 1 2---2----1 4
2---1--2 3 1----2----2 2 2---1---2 3
2---1--2 3 2----1---2 3 1----2---2 2
2----1---2 3 2---2---1 4 = l' 1----1-1 I
2----2---1 4 = l' 1----1----1 1 1--2-2 2
2----2----1 4 = 1/ I---J---l 2 1---1--1 I
2----2----1 4 = l' 2----1---2 3 2----2--1 4
2----2---1 4 = l' 2----2---1 4 = I' 2----1-2 3
Note: • indkales dummy treated levels.
5-8 COMBINATION DESIGN

Consider an experiment involving three 3 level factors and two 2 level
factors. An experiment design could consider L 16 OA with 3 columns for
each of the 3 level factors and 2 additional columns for the 2 level factors.
Such a design will utilize 11 of the available 15 columns and require 16
trial runs for the experiment. Alternatively consider the 4 OA.· Three
.columns satisfy the 3 level factors. If the fourth column can be used to
accommoWite two 2 level factors, then 4 with only 9 trial runs. could
be used.
:". ," Indeed, it is possible to combine two 2 level factors into a single 3
I.. level factor, with some loss of confidence in the results and loss of
./.'~ 0~' oPI!~ to study interactions. The procedure is given below.

- Define a new factor (XY) to be formed out of the combination of X
and Y and assign it to column 4. From the four possible combinations X
"
88 A PriIIl:r on the Taguchi Method
Table 5-28--4. Description of Individual Trial Conditions
11UAL NUMBER I
s.d compaction MIC ::::: Plant X · .. Level I

GDing type ::::: Plant X · .. Level 1
Nltalhead == Low · .. Level 1
"supplier ::::: Supplier I
Otadng type = Type I
·
·
..
..
Level
Level
1
I
.~
s-rperm = 300 perm · .. Level I I
.~'
\
.... temperature = 1430 F · .. Level I
Qlachtype = 450F · .. Level 1 t
GB level = Absent/none · .. Level I .I '
11UAL NUMBER 2 1
s.t compaetion MIC ::::: Plant X · .. Level I
GDing type = Plant Y · .. Level 2
llualhead = Low · .. Level 1
. .·supplier = Supplier I · .. Level 1
OJaIiRg type = Type 2 · .. Level 2
."penn.·~. = 400 perm · .. Level 2
. . . tempj'ature = 1460 F · .. Level 2
I
Qlachtype = 725 F · .. Level 2
Grra level = High · .. Level2
I
Table 5-28-5. Casting Process Optimization Data
TIUM. RI R2 R) 14 Rs Rt. AVG.
I 67.00 67.00
1 66.00 66.00
1 56.00 56.00
4 67.00 67.00
5- 18.00 18.00 I
6
7
90.00
68.00
90.00
68.00 I
(
,
8 78.00
89.00
78.00
89.00
10 18.00 18.00
11 69.00 69.00
12 16.00 76.00
13 18.00 78.00
14 66.00 66.00
15 17.00 17.00
16 87.00 87.00
Table 5-28-6. Casting Process Optimization Design Main Effects

COL
\
NUMBER FACTOR NAME LEVEL I LEVEL 2 (Lz - LI> LEVE;L 3 LEVEL 4
1 Sand compaction 70.50 78.50 8.00 78.00 00.00

4 Gating type 77.50 75.00 -2.50 67.50 00.00
5 Metal head 76.25 72.50 -3.75 00.00 00.00
;..
6 Sand supplier 76.25 72.50 -3.75 00.00 00.00
7 Coating type 69.25 72.75 3.50 74.75 BO.75
10 Sand perm 75.25 73.50 -1.75 00.00 00.00
11 Metal temperature 75.00 73.75 -1.25 00.00 00.00
13 Quencb type 72.50 76.25 3.75 00.00 00.00
15 Gas level 75.50 73.25 -2.25 00.00 00.00
and Y (XIY., XzY., Xlfz, and X2YZ). select any three and label them as
stated below.
call XIY. as (XY.r)1. Le., level 1 of new factor (XY)

XzY. as (XYh Le., level 2 of new factor (XY)
XIYz as (XYh Le., level 3 of new factor (XY).
Note that one combination, XzYz, is not included. With faetorXY assigned,
40A is shown in Table 5-29. From the array. the trial run conditions
defined for trial number I (row 1) are Al B I Cl (XY) I where (XY1I is
'XIY.. which was defined above.
82
80 78.50
78 17.5
76
o~ 74
in
a::
70
70.50
68 69.25
66 67.5
64'----'--_--J.._----JL...--_.L_._~.L_.~_L_ _ __I_._ _. . . I _ _ - - L . - _ . L . - -
L, L2 L3 L, L1 ~ L3
Coating Sand Compaction Gating
Figure 5-12. Plots of main effects-Example 5-10.

Table 5-29. L 9 with Five Factors

fACfoas A B C (X¥)
NUMBERlCOWMS I 2 3 4 Y- \f
I I I 1 I \
2 I 2 2 2 1.. 1
3 I 3 3 3 1 'Z.
'4 2 I 2 3 1. '1
5 2 2 3 I 1. 1
6 2 3 I 2 ~ 1
7 3 I 3 ~ "L 1.. 1
8 3 2 I 3 , ~
9 3 3 2 I , 1
The total data are analyzed with the two factors X and y treated as one
(XY). The analysis yields the main effect of (XY). The individual effect
of the constituents, X and Y is then obtained as follows: rI
i
I
Main effect of X :;:: (XY) 1 - (XY)2 and
Main effect of,Y :;:: (XY)I - (XYh
Note that the first equation has Y constant as Y. and the second equation
has X constant as XI'
After determining the main effects, the optimum condition, including
the levels of the two factors used in combination design, can be identified.
However, the interaction effects between factors X and Y cannot be ob-
tained from the data by this method. Should interaction be important, the
experiment design must be based on a larger array such as L 16-
5-9 DESIGNING EXPERIMENTS TO INVESTIGATE NOISE

FACTORS
Throughout this text the tenns factors. variables and parameters are
synonymously used to refer to factors which influence the outcome of
the product or process under investigation. Taguchi further categorized
the factors as controllable factors and noise factors. The factors identified
for the baking process experiment, namely sugar, butter, eggs, milk, and I'
i
flour, were easily controlled factors. Other factors which are less con-
trollable, such as oven temperature distribution, humidity, oven temper-
-- _ "'"_.----- ...
... ~.. '"'-' -
Working Mechanics Qf the Tagucbi Design of Experlments 91
ature cycle band width~ etc., may also influence the optimum product.
Since the goal is a ro'ffust optimum which is influenced minimally by
these less controllable variables, the study of the impact of noise factors
on the optimum parameters is desirable.
Taguchi fully recognizes the potential influence of uncontrQllable fac-
tors. No attempt is made to remove them from the experiment. Before
describing how the u~controUable factors are treated, additional defini-
tions are needed.
Controllable Factors-Factors whose levels can be specified and con-
trolled during the experiment and in the final design of the product or
process.
Noise Factors--These are factors which have influenoeon the product
or process results, but generally are not maintained at specific levels
during the production process or application period.
Inner Array--The OA of the controllable factors. All experiment designs
discussed to this point fail in this category.
Outer Array--Tbe OA of recognized noise factors. 1k tenn outer or
inner refers to the usage rather than the array itself, as will be clear soon.
Experimenl-The experiment refers to .the whole experimental process.
Trial Condition-The combination of factors!Jevels at which a trial run
.is conducted.
Conditions of Experiment-Unique combinations of factor levels de-
scribed by the inner array (orthogonal array).
Rep.etitions or Runs-These define the number of obsemWons under the
same conditions of an experiment.
The experiment requires a minimum of one run per condition of the
experiment. But one run does not represent the range of possible vari-
ability in the results. Repetition of runs enhances the available information
in the data. Taguchi suggests guidelines for repetitions.
5·10 BENEFITING FROM REPETITIONS

For some experiments, trial runs are easily and inexpensively repeated.
For others, repetitions of tests are expensive as well as time consuming.
Whenever possible trials should be repeated. particularly if strong noise
factors are present. Repetition offers several advantages. First, the ad-
ditionallrial data confirms the original data points. Second, if noise factors
vary during the day. then. repeating trials through the day may reveal
their influence. Third. additional data can be analyzed for variance around
a target value.
When the cost of repetitive trials is low, repetition is highly desirable.
When the cost is high, or interference with the operation is high, then
the number of repetitions should be determined by means of an expected
payoff for the added cost. The payoff can be the development of a more
robust production procedure or process, or by tbe introduction of a pl'(}o
duction process tbatgreatly reduces product variance.
Repetition pennhs determination of a· variance index called the Signal .
to. Noise ROOo (SIN). The greater this value, the smaller the product
variance around the target value. The signal to noise ratio concept has
been used in the fields of acoustics, electrical and mechanical vibrations,
and other engineeriJ'lg disciplines for many years. Its broader definition
and application will be covered in Chapter 6. The basic definition of the
SIN ratio is introduced here.
5-11 DIfINITION OF SIN RATIO

SIN;:::: -10 Log lO (MSD)
Where MSD =
Mean squared deviation from the target value of
the quality characteristic.
Consistent with its application in engineering and science, the value

of SIN is intended to be large. hence the value of MSD should be small.
Thus the mean squared deviation (MSD) is defined differently for each
of the three quality characteristics considered, smaller, nominal or larger.
For smaller is better:
MSD = (>1 + n + yj + .. .)/n

For nominal is the best
MSD = «Yl - m)2 + (Y2 - mf + .. .)/n
For bigger is better
MSD = (113'1 + l/y~ + lIy~ + .. .)/n

Where YI, Y2' etc. == Results of experiments, observations or

quality characteristics such as length,
weight, surface finish. etc.
m == Target value of results (above)
n== Number of repetitions (Yi)
Consi~er an experiment with three repetitions, using an L4 orthogonal

array as ~shown in Table 5-30.
In the table, trial number I is repeated three times with results 5, 6,
and 7. The average of these three is 6. The average is used for the study
of the main effects in a manner similar to that described for nonrepeated
trials. Slight differences in the analysis of variance for the repetitive case
is covered in Chapter 6. For experiments with repetitions,adalysis should
always use the SIN ratios computed as follows.
Assume that bigger is better is the quality characteristic sought by the
experimental data of Table 5-30. Then,
MSD == (l/yj + l/y~ + IIn + .. .)/n

Now, for row 1, YT == 5 x 5 - 25· .,~- == 6 x 6 == 36
rl ;:; 7 x 7 ;::: 49 n == 3
Therefore, MSD == (1/25 + 1/36 + 1149)/3

;::: (.04 + .02777 + .020408)/3
- .088l85/3
or MSD ;::: .029395, a small value.
Table 5-30. L 4 with Results and Averages

TRIAL NUMBER
COLUMN 2 3 RI R: R3 AVERAGE
I I I I 5 6 7 6
2 I 2 2 3 4 5 4
3 2 I 2 7 8 9 8
4 2 2 2 4 5 6 5
Table 5-31. L.. "ith Results. and SIN Ratios

TRIAL ~l.:MBER SIN
COLnfN ~ 3 Rl R2 R) RATIO
1 1 1 1 5 6 7 15.316
2 1 2 2 3 4 5 11.47
3 2 1 2 7 8 9 17.92
4 2 2 2 4 5 6 13.62
The SIN ratio is calculated as:
SIN = - 10 1.og 10 (MSD)

- - 10 1.og 10 (.029395)
- 15.31
SIN values for alt rows are shown in Table 5-31.

In. analysis~ the SIN ratio is treated as a single data point at each of
the test run conditions. Normal procedure for studies of the main effects
will follow. The only difference will be in the selection of the optimum
levels. In SIN analysis, the value of MSD or greatest value of SIN rep-
resents a more desirable condition.
5-12 REPETITIONS UNDER CONTROLLED NOISE

CONDITIONS
Repetitions show the variation of the product or process. The variation
occurs principally as a result of the uncontrollable factors (noise factors).
By expanding the design of the experiment to include noise factors in a
controlled manner ~ optimum conditions insensitive to the influence of the
noise factors can be found. These are Taguchi ~ s robust conditions that
control production close to the target value despite noise in the production
process. To incorporate noise factors into the design of the experiment
the factors and their levels are identified in a manner similar to those
used for other product and process factors (control factors). For example~
if humidity is considered noise, the low and high levels may be considered
a factor for the design. After determining the noise factors and their levels
for the test~ OAs are used to design the conditions of the noise factors
WOlting Mechanics of the Taguchi Design of Experiments 9S
which dictate the number of repetitions for the trial runs. The OA used
for designing the noise experiment is called an outer array.
Assume that three noise factors are identified for the cake baking
experiment (Tables 5-10 and 5-14) which utilized an Ls OA. The noise
factors are to be investigated at 2 levels each. There are four possible
combinations of these factors. To obtain complete data, each trial run of
La must be repeated .for each of the four noise combinations. The noise
array selected is an L4 OA. This outer array, with four combinations of
noise of the three noise factors, tests each of the 8 trial conditions four
times. The experiment design with inner and outer array is shown by
Figure 5-13. Note that for tile outer array, column 3 represents both the
third noise factor and the interaction of the first and second noise factor.
Note also the arrangement of each array, with the noise (outer) array
perpendicular to the inner tlITay. The complete design is .shown by Figure
5-14.
For most simple applications, the outer array describes the noise c0n-
ditions for the repetitions. This fonnal arrangement of the noise factors
and the sUhsequentanalysis influences the combination of the controllable
,.
Outer Array
~
M
N
...
.-
N
N
N
...
-
N
a
u..
i
fA
.-
-- N N
~ III
c:
Inner Array
cS
j - N M ...
Coatrolf=ac:tors Results
~ Number t 2 3 4 5 6 7 1 2 3 4
Experiment
Number
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 1 2 1 2
6 2 1 2 2 1 2 1
7 2 2 1 1 2 2 1
8 2 2 1 2 1 1 2
Figure 5-13. Inner and outer orthogonal arrays.

Factor
Description Level' Level 2
Egg 2 Eggs 3E.

Nla.)--NlNI-.... NI Milk 2 Cups 3 Cups
--NlNlNlNI-- W
Butter , Stick 1.5 Sticks

- N .... NlNI-Nl- U'I Flour 1 Extra ScooP 2 Extra .SCoops
........ -Nl-NlNI- ..... Sugar 1 Spoon 2Spooos
TYP80f Baking
Oven Time Humidity
1.. Gas 1.+5min. 1.80%
2. Electric 2.- 5 min. 2.60%
Columns 2 3
....
:D
1 , ,
:D 2
2 2
""
...:D 3 2 1 2
~
:D 4 2 2 ,
Figure 5-14. Cake baking experiment with noise factors.
factors for the optimum condition. The use of SIN ratio in analysis, is
strongly recommended.
5·13 DESIGN AND ANALYSIS SUMMARY

Application of the Taguchi technique is accomplished in two phases:
(1) design of the experiment, which includes determining controllable
and noise factors and the levels to be investigated, which determines
the number of repetitions, and (2) analysis of the results to determine
the best possible factor combination from individual factpr influences
and interactions. The two activities, experiment design and analysis of
Working Mechanics of the Tagucbi Design of Experiments 97
I Experiment
Design
I
I
I I
Simple Design Using Designs with Mixed levels
Standard Arrays and InteraCtions
I
Assign Factors to Columns • Modify Columns
as Appropriate • AssigrlFaetors Requiring
Levtl Modification
;.
• Assigrllnteraeting Factors
• Assign AD Other Factors
I
I
• Consider NolSIJ Factor
• Determine Noise Condition Using Outer Array
• Determine Number of Repetitions
I t
• Run Experiments In Random Order When Possible
I
Figure 5-15. Experiment design flow diagram.
• Nominal Is Best
• Smaller Is Better
• Bigger Is Better
Figure 5-16. Analysis flow diagram.
t.
test data are presented (flow charts) in Figures 5-15 and 5-16. The steps
involved are briefly described here.
Des. .
Depending on the factors and levels identified, follow one of the two
paths (Fig. 5-16). If all the factors are of the same level~ say 2, one of
the standard OAs can probably be used. In this case, the factors can be
assigned to the columns without much consideration about where they
should be placed. On the other hand, if the factors require many levels,
or one or more interactions are to be investigated, then carefully select
certain specific columns for factor assignments Of level changes. No
matter how simple the design, the applicable noise conditions should be
identified, and a second array (outer array) selected to include noise
ttffects. The number of repetitions wiU be dictated by the number of noise
factors. In the absence of a formal layout such as Figure 5-13 the number
of repetitions will be influenced by time and cost considerations.
Analysis
Analysis of results follows either paths (Fig. 5-16) of repetitions or no
repetition. Generally, for a single observation for each trial condition,
the standard analysis approach is followed. When there are repetitions
of the trial runs, whether by outer array designed noise condition, or
under random noise condition, SIN analysis should be performed. The
final analysis for the optimum condition is based on one of the three
characteristics greatest, smallest, or nominal value of quality.
~
EXERCISES
5-1. Identify each element (8, 2, 7, etc.) of the notation for an orthogonal array Ls(27 ).
5-2. Design an experiment to study four factors A. B. C. and D and three interactions
A x C. C x D and A x D. Select the orthogonal array and identify the columns
for the three interactions.
5.;3. An experiment with three 2 level factors yielded the following results. Determine
the average effect of factor C at levels Cl and C2 •
TRIALS!
COLt.:MSS A B C RESULTS
1 1 1 1 30
2 1 2 2 25
3 2 I 2 34
4 2 2 1 27
· ... _._-----------------------.
WCJrking Mechanics flf the Taguchi Design of Experiments 99
Table 5-31. Design Variables and Their Levels
,COLUMN
NUMBER LE\R-I LEVEL-2
1 Speed 2100 RPM 250 RPM

2 Oil viscosity At low TP At high t
3 Interaction 1 x 2 NlA NlA
;.
4 Clearance Low Higb
5 Pin straightness Perfect Bend
6 (Unused) MIU
7 (Unused) MIU
5-4. Describe the procedure yoo will follow to design an experiment to study one 3 level
factor and fo.Uf 2 level factors.
5-5. In an eXperiment involving piston bearings, an La OA was used in a manner shown
in Table 5-32. Determine the description of the trial number 7.
5-6. The average effects of the factors involved in Problem 5 are as shown in Table
5-33. Ifthe quality characteristic is '~e biggertbe better," determine (a) the optimum
condition of the design, (b) the grand average of performance, and (c) the perfor-
mance at the optimum condition. (Ans. (b) 35.01 (c)37.03)
Table 5-33. Average Factor Eft'eds
COLUMN
NUMBER VAlUABLE NAMES LEVEL.I LEVEL-2
1 Speed 34.39 35.63

2 Oil viscosity 35.50 34.52
3 Interaction.l x 2 33.60 36.42
4 Clearance 35.62 34.40
5 Pin straightness 35.31 34.70
·6
Analysis of Variance (ANOVA)
6-1 .THE ROLE OF ANOVA
Tagucbi replaces the full factorial experiment with a lean, less expensive,
faster, partial factorial experiment. Taguchi's design for the partial fac-
torial is based on specially developed OAs. Since the partial experiment
is only a sample of the full experiment, the analysis of the partial ex-
periment must include an analysis of the confidence that can be placed
in the results. Fortunately, there is a standard statistical technique called
Analysis of Vpee (ANOVA) which is routinely used to provide a
measure of con«~ence. The technique does not directly analyze the data,
but rather determines the variability (variance) of the data. Confidence
is measured from the variance.
Analysis provides the variance of controllable and noise factors. By
understanding the source and magnitude of variance, robust operating
conditions can be predicted. This is a second benefit of the methodology.
6-2 ANOVA TERMS, NOTATIONS AND DEVELOPMENT
In the analysis of variance many quantities such as degrees of freedom,

. sums of squares, mean squares, etc., are computed and organized in a
standard tabular fonnat. These quantities and their interrelationships are
defined below and their mathematical development is presented.
DEfINmONS:
C.F. = Correction factor n =Number of trials

e = Error (experimental) r = Number of repetitions
F= Variance ratio P = Percent contribution
f = Degrees of freedom T = Total (of results)
fe= Degrees of freedom of error S = Sum of squares
fr= Total degrees of freedom S' = Pure sum of squares
V = Mean squares (variance)
100
- -_.- -------------......-------------"I
I
Analysis of Variance (A.'liOVA) 101
Total Number of Trials

In an experiment designed to detennine the effect of factor A on response
\ Y. factor A is to be tested at L levels. Assume n 1 repetitions of each trial
that includes A 1. Similarly at level A 2 the trial is to be repeatedn2 times.
The total number of trials is the sum of the number of trials at each level,
Le. ,
Degrees of Freedom (DOF)

OOF is an important and useful concept that is difficult to define. It is a
measure of the amount of infonnation that can be uniquely determined
a
from a given set of data. OOF for data concerning factor equals one
less than the number of levels. For a factor A with four levels, A I data
can be compared with A 2 , A 3 and ~ data and not with itself. Tbus a 4
level factor has 3 OOF. Similarly an L 4 OA with three columns repre-
senting 2 level factors, has 3 OOF.
The concept of DOF can be extended to the experiment. An experiment
with n trial and r repetitions of each trial has n x r trial runs. The total
OOF becomes:
fT==nXr-1
Similarly, the DOF for a sum of squares tenn is equal to the number
of terms used to compute .the sum of squares and theDOF of the error
term le is given by:
ft! == fT - fA - fs - le
Sum of Squares
The sum of squares is a measure of the deviation of the experimental
data from the mean value of the data. Summing each squared deviation
emphasizes the total deviation. Thus
n
ST == :L (Y j - Y)2
i= 1
Where Y is the average value of Yi •

'-'
un A Prim&tr on the Taguebi Method
Similarly the sum of squares of deviations Sr, from a target value Yo, is
given by;
n
ST = L (Yi - y)2 + n(Y - Yo)2 (6-1-1)*
;= I
Varianc:e measures' the distribution of the data about the mean of the
data. Since the data is representative of only a part of an possible data,
OOF rad1er than the"number of observations is used in the calculation.
Sum of Squares
Varianee -
Degrees of Freedom
or V = Sri!
woo. die average sum of squares is calculated about the mean, it is
call~.aeral
:,
variance. The general variance 0-2 is defined as.
1 n
0-2 = - ~ (Y; - y)2 (6-1-2)
. n ;-1
•
* Sr:::::I (Y; - for
..
;,=1
:::: L ('j - Y+ Y- f O)2

i=J
..
:::: :I l(Yi - y)2 + 2(f; - Y)(Y - f o) + (Y - YO)2]
;==1
. , , " 11
:::: 2 ('; - f}2 + L 2(Y; - n(y - yo) + ~ (Y - Yo)2

isl i=1 ;=1
• ••
since :I (Y i - 1> :::: ~ Y; - L Y == nY - nY :::: 0
i.-I i_I i=1
•
and ~ cy - YoY :::: n(Y - f o)2
pi
The above c:cp1lltion becomes,

. \
Sr :::: L
i",,1
(Vi - n 2
+ n(Y - for
-- - -~~---------------~.
I Analysis of Variance (ANOVA) 103
1 Let m represent the deviation of the mean f from the target value f o,
I.e. ,
m::::: (f - f o) (6-1-3)
Substituting Eqs. (6-1-2) and (6-1-3) into Eq. (6-1-1),
(6-1-4)
Thus the total sum of squares of deviations (ST) from the target value f o
is the sum of the variance about the mean, and the square of .the deviation
of the mean from the target value multiplied by the total number of
observations made in the experiment.
ST of Eq. (6-1-4) also represents the expected statistical value of ST.
In this book, rigorous proofs are omitted unless necessary to clarify an
idea or concept. Further, the symbolST is used for both the expected
value and the computed value for a given sample.
The total sum of squares ST (Eq. 6-1-4) gives an estimate of the sum
of the variations of the individual observations about tbemean Y of the
experimental data and the variation of the mean about the target value
f o. This information is valuable for controlling manufacturing processes,
as the corrective actions to reduce the variations around the mean Y, Le.,
to reduce 0'2, are usually not identical to those actions which move the
mean toward the target value. When the total sum of squares ST, is
separated into its constituents, the variation can be understood and an
appropriate strategy to bring the process under control can be easily I
developed. Furthermore, the information thus acquired can be effectively I
utilized in Statistical Process Control (SPC).
Mean Sum (of Deviations) Squared

n
Let T::::: I (Yi - Yo) the sum of all deviations from the target value.
i= 1
Then, the mean sum of squares of the deviation is:

2 [~ (Y; - Y I]2/n
S,.:' T /n = O (6-2)
Eq. (6-2) can thus be written as:
It is important to note that even though from an over-simplistic deri-

vation of the value of Sm == nm2 , its statistical estimate or the expected
value, includes one part of the general variance. Therefore, representing
the statistically expected value by E(Sm):
(6-3)
The tenn (ST - Sm> is usually, referred to as the error sum of squares
and can be obtained from" Eqs. (6-14) and (6-3).
Therefore,
Se == ST - Sm == (n - 1)02
Rewriting th~equation, ST == Se + Sm' Thus tbe total effect of variance

ST can be d~mposed into the mean deviation Sm and the deviation Se
about the melt. Thus individual effects can be analyzed. Let
Y1 - f o == 3 f4 - f o == 4
Y2 - f o == 5 f s - f o == 6
f 3 - Yo == 7 Y6 - fo - 8
1
+ ... +
I
I
*5", == -- (fi
n
- f o)
Which can also be expressed as:

(f" - foW
1
5", == -- [(f. + f2 + ... + f .. - nYo)]2
n
or
1 -
s'" == -n [(nf - nfoW
or
Analysis of Variance (ANOVA, 105
where Yo is a target value, then
== 199
Sm == (3 +5 + 7 + 4 + 6 + 8)2/6
;.. . 33 2/6
== 181.5
Y == (3 + 5 + 7 + 4 + 6 + 8)/6
== 5.5
and
Se== [(3 - Yf + (5 - y)2 + ...]

== [(3 - 5.5)2 + (5 - 5.5f + .... (8 - 5.5)2)
== 17.5
Note that
Se == ST - Sm == 199 - 181.5 == 17.5
Also, since the standard deviation of the data 3, 5, 7, 4, 6, and 8, is

equal to 1.8708.
4
Se· == (n - 0')( 6
== (6 - 1) X (1.8708)2
== 17.5
Degrees of Freedom Sums

The DOF !e,fT,andfm of the sum of squares Se, ST, andSmare as follows:
fT'== n == number of data points.

fm == I (always for the mean)
le == IT - Im == (n - 1)
106 A Primer OD 6r: Taguchi Method
As pointed mt earlier, the DOF !T is equal to n because there are n

independent values of (Yj - YO)2. For investigating the effect of factors
at different levels, the OOF are usually one less than the number of
observations.
To summarize:
Sr = ner + nm2 (64)
Sm = er + nm2 (6-5)
Se = Sr - Sm = (n - 1)0-2 (6-6)
Also .as stated earlier, variance V, is

v= SI!
Therefore:
Vr = SrI!T er + m
2
(total variance)
Vm = S,,/I. a2 + nm2 (~ variance)
Ve = (Sr - Sm)/!e = er (error variance)
The example that follows should clarify the application of the concepts
developed a1:Jo'le. The data for this example are fictitious but suffice for
the purpose of Illustrating the principles.
6-3 ONE WAY ANOVA
One F8£tor ODe Level Experiment

When one dimensional experimental data are analyzed using ANOVA,
the procedure is termed a one way analysis of variance. The following
problem is aB example of one way ANOVA. Later ANOVA will be
extended to multi-dimensional problems.
Example 6.1
To obtain tile most desirable iron castings for an engine block, a design
engineer wants to maintain the material hardness at 200 BHN. To measure
the quality of die castings being supplied by the foundry the hardness of
Analysis of Variance (ANOVA) 107
Table 6-1. Hardness of Cylinder Block

Castinp-ExalDple 6-1
SAMPlE HARDNESS SAMPLE HARDNESS
I 240 6 180
2 190 7 195
3 210 8 205
4 230 9 215
5 220 10 215
10 castings chosen at random from a lot is measured, and displayed in

Table 6-1.
The analysis:
IT = Total number of results - 1

= 10' - 1 == 9
f o = Desired value = 200
the mean value is:
f == (240 + 190 + 210 + 230 + 220 + 180 + 195 + 205

+ 215 + 215)/10
= 210
then
ST = (240 - 200)2 + (190 - 2(0)2 + (210 - 200)2

+ (230 - 200)2 + (220 - 2(0)2 + (180 - 200)2
+ (195 - 200)2 + (205 - 200)2 + (215 - 200)2
+ (215 - 200)2
=4000
and Srn == n(Y - f o)! = 10(210 - 200)2 = 1000

Se = ST - Srn = 4000 - 1000 = 3000
1. A Primer on the Tagucbi Method
Table 6-2. Analysis of Variance (ANOVA) Table-Example 6- t
\ARIA."ICE VARIANCE PURE PERCENr

SOURCE OF SUM OF (MEA.'" SQUARE> RATIO SUM OF SQUARES CONTRIBUTION
VARIATION f SQUARES V F 5' p
Mean (m) 1 1000 1000.00

Error (e) 9 3000 333.33
Total 10 4000
And the variance is calculated as follows:

10 l.t"8o. D
VT = SrlfT = 4000/.9" = ~
Vrn = 1000/1 = 1000
V~ == (ST - Srn)l/e = (4000 - 1(00)/9 = 333.33
These results are summarized in Table 6-2. Table 6-3 represent a gen-
eralized formatt>f the ANOVA table.
The data cannot be analyzed further, but analysis of the variance of
the data can provide additional information about the data.
Let F be the ratio of total variance to the error variance. F coupled
with the degrees of freedom for VT and Ve provides a measure for the
confidence in the results.
To complete the analysis, the error variance V~ is removed from Srn
and added to S~. The new values are renamed as
S:" = pure sum of squares

S; = pure error.
Table 6-3. ANOVA Table for Randomized One Factor Designs-

~ Example 6- t
VARIANCE VARIANCE PURE PERCENr

SOURCE OF SUM OF (MEAN SQUARES) RATIO SUM OF SQUARES CONTRIBUTION
VAlUATION f SQUARES V F s' P
t:oanr (m) /m Sift Sml/m

(e) h S~ Sif~
Total fr_ Sr
This reformUlation allows calculation of the percent contribution, p.

for the mean, Pm' or for any individual factor (P A , Ps, etc.)
Table 6-3 presents the complete format for analysis F. Sand P. These
parameters are described below in greater detail.
Variance Ratio
The variance ratio, commonly called the F statistic, is the ratio of variance
due to the effect of'a factor and v~ance due to the error term. (The F
statistic is named after Sir Ronald A. Fisher.) This ratio is used to measure
the significance of the factor under investigation with respect to the vari-
ance of all the factors included in the error term. The F value obtained
in the analysis is compared with a value from standard F-tables for a
given statistical level of significance. The tables for various significance j
levels and different degrees of freedom are available in most handbooks
of statistics. Table C-l through C-5 in Appendix A provides a brief list
of F factors for several levels of significance.
I
To use the tables enter the DOF of the numerator to determine the
column and the DOF of the denominator for the row. The intersection is
the F value. For example, the value of F. l (5,30) from the table is 2.0492,
where 5 and 30 are the DOF of the numerator and t,he denominator, ..,
respectively. When the computed F value is less than the value determined
from the F tables at the selected level of significance, the factor does not
. :contribute to the sum of the squares within the confidence level. Computer
software, such as Reference 11. simplifies and speeds the determination
of the level of .significance of the computed F values.
The F values are calculated by:
(6-7)
and for a factor A it is given by:
(6-8)
Pure Sum of Squares

In Equations (6-4). (6-5). and (6-6) for each of the sum of squares, there
is a general variance (J"2 tenn expressed as <DOF) X (J":.
When this term is subtracted from the sum of squares expression. the
110 A I'bmer on the Taguchi Method
remainder is called the pure sum of squares. Since Sm has only one degree
of freedom, it therefore contains only one a.2 i.e., Ye' Thus the pure sum
of square for Sm is:
The portion of error variance subtracted from the sum of squares for
S. is added to the error term. Therefore,
S; = Se + Ve (6-9)
If factOD A, B, and C, havingOOF lA, IB' and le are included in an

experi.....t, their pure sum of squares are detennined by:
S~ = SA - lA X V,. (6-10)
Ss = Ss - Is x Ve
Se = Se - le X V,.
S; = Se + (fA + Is + Id X Ve (6-11)
Perceld Contribution
The pm:ent contribution for any factor is obtained by dividing the pure
sum of squares for that factor by SF and multiplying the result by 100.
The pen:ent contribution is denoted by P and can be calculated using the
followiDg equations.
Pm -S:" X lOO/Sr
PI( = S~ X lOO/Sr
Ps Ss X lOO/Sr
Pc = Se X lOO/Sr
Pe - S; X lOO/Sr (6-12)
The AHOVA Table 6-2 can now be completed as follows: Using Eqs.
(6-7) aDd (6-8) gives:
Fm VmlVe = 1000/333.33 = 3.00

F1 = V~, V 1 = 333. 33,333. 33 = 1. 00
'.,
Analysis of Variance (A-';OV A) III
Table 6-4. Analysis of VariaIlce (ANOVA) Table-Example 6-1

VARIANCE VAAIANCE R.lRE PERCENT
SOURCE OF SUM OF (MEAN SQUAl'tESl RATIO SUM OF SQUARES CON'l'IUBU11ON
VAlUATION f SQUAl'tES v F S' p
Mean (m) 1 1000 1000.00 3.00 666.67 16.67

Error (e) 9 3000 333.33 1.00 3333.33 83.33
Total 10 4000 100.00
The pure sum ofsquares obtained using Eqs. (6-9) and (6-10) is shown
below:
s:" == Sm - Ve == 1000 - 333.33 == 666.67

S; == Se + Ve == 3000 + 333.33 == 3333.33
And the percent contribution is calculated using Eqs. (6-11) and (6-12):
Pm == S:" X loo/Sr == 666.67/4000 == 16.67

Pe == S; X loo/Sr == 3333.33/4000 == 83.33
The completedANOVA tables are shown in Table 6-4.

A generalized ANOVA table for one factor randomized design is shown
in Table 6-5.
Returning to Table 6-3, the computed value for Fm, 3.00, is less than
the value from Table C-Ifor F. 1 (1,9) Le., 3.3603. Hence with 90%
confidence (10% risk) the castings appear to be similar. The apparent
data spread conlfoibutes only 16.67~ to the sample variability (sum of
Table 6-5. ANOVA Table for Randomized One Factor Design-

Example 6-1
SOURCE OF VARIANCE VAAIANCE f'lJRE PERCENT
VARIATION SUM OF (MEAN SQl'ARESl RATIO Sl"M OF SQUARES CO~IRIBIJTION
f SQUARES v F S' P'IOO
Mean (rn) Jm S", V", = S",IJm V".IV~ Sm - V~ S' ".1ST

Error (e) J~ S~ V~ = S"fJ~ Sm + V~ S'"fST
Total IT Sr
squares) whereas the remaining 83.33% variation is caused by other

factors.
6-4 ONE FACTOR TWO LEVEL EXPERIMENTS

(ONE "'AY A.'lOYA)
Example'.}
In Example 6-1 an experiment with one factor at one level was con-
sidered, the factor being the hardness ofthe cylinderblocks being supplied
by one source. Now consider the case with two different vendors suppling
the castings, These two sources are assumed to use similar casting pro-
cesses. Therefore, a new experiment is described with one factor, hard-
ness of castings, from two sourcesA. andA2 • The question to be resolved
is whether the castings being supplied by the two vendors are statistically
of the same quality. If not, which one is preferable. The target hardness,
200 SHN, is unchanged.
Ten sample~~ from each of the two castings sources were drawn at
random and thtir hardness was measured. The test yielded the results
shown in Table 6-6.
The analysis of this test proceeds as for the experiment of Example
6.1. Note that the error sum of squares term, S~, as given in Eq. (6.11),
contains the variation of the mean and that of the factor A. Therefore to
separate the effect of vendors, i.e., factor A. the sum of squares term SA
must be isolated from Se. The sum of squares for the factor A can be
tt calculated by:
(6-13)
n
-: Table 6-6. Measured HardDess-Example 6-2
H.UDNESS OF CASTINGS FROM HAIU»iESS OF CASTlNGS FROM

VENOORAI VENDOIt A2
240 180 190 191 202 198

195 210 205 205 203 192
230 215 220 208 199 195
215 201
Analysis of VlIriance (ANOVA) 113
Where,
L = nun,ber of levels
ni, nk = number of test samples at levels Ai and Ab respectively
T = sum total of all deviations from the target value
n = total number of observations = nl + n2 + ... + nj
~.
The Pin in Eq. (6-13) is a tenn similar to Sm and is called the correction
factor, C.F.
The expression for the total sum of squares can now be written as
(§-14)*
The OOF equation will be:
The Analysis:
f o = 200 (unchanged from Example 6.1)

Sr = (240 - 200)2 + (180 - 200)2 + ... + (215 - 200)2
+ (197 - 200)2 + ... + (195 - 200)2
+ (201 - 200)2 - C.F.
= 4206 - 500 = 3706
As T 2/n= C.F. = [(f1 + (f2 - f o) + ...
- f o) FIn
= [(40- 20 + . . . - 15 + 15)
+ (- 3 + 2 ... - 5 + 1)j2/20
=500
*Taguchi considers de"iation from the target more significant than that about the mean. The cost
of quality is measured as a function of the deviations from the target. Therefore. Taguchi eliminates
the variation about the mean from Eq. (6-14) by redefining 51 as follows:
ST = :2" (Yi - YO)2 - C.F. = Se + SA

1=1
or Se = ST - S...
Using Eq. (6-13), the value of SA, the square sum for the effect of factor
A (vendors) is obtained as:
SA = Squares of sum for vendor Al

+ Square of sum for vendor A 2
- Correction factor
or
SA =I(YiAI - y o)]2/nA1 + [(YiA2 - yo)]2/nA2 + T2/n
= 1'240 - 200) + ... + (215 - 200)]2

10
+ 1(197 - 200) + ..•. + (201 -.200)]2 _ 500
. 10
= 1000 +/() -
1~.
500 = 500
Also
S~ - 5r - SA
- 3706 - 500
= 3206
and the OOF will be:
IT = 20 - I = 19
!A-2-1=1
le = 19 - 1 = 18
Also,
and
i
Anal)'!ois of Variance (A.NOVA) 117
1
~
I
The number of experiments is 20 and there are two factors m and A I
involved in the estimates. Therefore,

\
20 20
ne = lA + Im - I + 1= 10
Since,
F. 1 (1,18) = 3.007
and
Ve = 178.11
m '" 205 + ~3,OO1 ':0 118, 11
= 205 + 7.32
= (197.68, 212.32) .
Therefore, it can be stated that there is a 9O'k probability that the true
value of the estimated mean will lie between 197.68 and 212.32.
The confid~nce interval can similarly be calculated for other statistics.
6-5 TWO WAY ANOVA

The one way ANOVA discussed above i~cluded' one factor with two
levels. This section extends ANOVA to experimental pata oftwo or more
factors with two or more levels. The following examples illustrate the
procedure.
Example 6-3
The wear characteristics of two brands of tires, Wearwell and Super-
wear are to be compared. Several factors such as load, speed, and air
temperature have significant effect on the useful life of the tires. The
problem will be limited to only one factor, i.e., the temperature. Let T.. .
and Ts represent winter (low) and summer thigh) temperatures, respec-
tively. Tire life is measured in hours of operation at constant speed and
load. The experiment design for this example is given in Table 6-8. This
Table 6-8. Tire Wear Experiment-Example 6-2

TE~fPERA n'RE
.'1
TIRE TYPE ·'11 = T.. -'2 = Tr SUM
Wea:rwell 70 65 135
BI Yl YJ
B
Supawear 75 60 135
B2 Y2 Y4
Sum 145 125 210 = Grand total
is called a two factor two level experiment. It has 4 possible trial runs
and the results of each run can be interpreted as follows:
With A at A I aDd B at B I the life is= 70 Hrs.

With A at Al ~ B at B2 the life is = 75 Hrs.
With A at A 2 8d B at BI the life is = 65 Hrs.
With A at A2 ad B at B 2 the life is == 60 Hrs.
The analysis of the data follows exactly the same procedures presented
in the previous example. In this case the total degrees of freedom,
fT = n - 1 = 3. The degrees of freedom and the ANOVA quantities
in this case become:
n· 22 = 4
fT == n - 1 =4 - 1 == 3
fA == number of levels - 1 =.2 -- 1 = 1
fB = types of tires - 1= 2 - 1 == 1
fAxB =1x I=I
le = fT .- fA - fB - fAx B
C.F. = Correction Factor- T 2/n = (270)2/4 = 18225.0
Yo = Target value = 0
ST = sum of square of all results - C. F.
Analysis of VariaDce (ANOVA) 117
The number of experiments is 20 and there are two factors m and A

involved in the estimates. Therefore,
20 20
ne = fA + fm -1 + 1 = 10
Since,
;.
F. 1 (1,18) - 3.007
and
Ve = 178.11
m -= 205 ± l-f101 70 118. 11
= 205 + 7.32
= (197.68, 212.32) .
Therefore, it can be stated that there is a 90% probahility that the true
value of the estimated mean will lie between 197.6Saad 212.32.
The confidence interval can similarly be calculated for other statistics.
6-5 TWO WAY ANOVA

The one way ANOVA discussed above included- one factor with two
levels. This section extends ANOVA to experimental. data of two or more
factors with two or more levels. The following elamples illustrate the
procedure.
Example 6-3
The wear characteristics of two brands of tires, Wearwell and Super-
wear are to be compared. Several factors such as load, speed, and air
temperature have significant effect on the useful life of the tires. The
problem will be limited to only one factor. i.e., the temperature. Let T",
and Ts represent winter (low) and summer (high) temperatures, respec-
tively. Tire life is measured in hours of operation at constant speed and
:C'ad. The experiment design for this example is given in Table 6-8. This
118 A Primer onlhe Taguchi Method
Table 6-8. Tire '''ear Experiment-Example 6-2

1E\4PERATt'RE
A
mE TYPE .4, = T.. A2 = T. SUM
Wearwell 70 65 135
B
BI Y, YJ
Saperwear 75 60 135
B2 Y:! Y4
Sum 145 125 270 :::: Grand total
is called a two factor two level experiment. It has 4 possible trial runs
and the results of eacb run can be interpreted as follows:
With A at AI and B at 8 1 the life is = 70 Hrs.

With A at AI and B at B 2 the life is = 75 Hrs.
With A at Az and B at B 1 the life is = 65 Hrs.
With A at Az ~d B
*.~:;
at 8 2 the life is = 60 Hrs.
1;<
The analysis d"the data follows exactly the same procedures presented
in the previous example. In this case the total degrees of freedom,
IT = n - 1= 3. The degrees of freedom and the ANOVA quantities
in this case become:
IT = n - 1 =4 - I = 3
lA = number of levels - 1 =.2 - 1 - 1
IB - types of tires - I = 2 - 1 - 1
IAXB = 1 x 1 = 1
le == IT - lA - IB - lAX B
C.F. = Correction Factor = T2/ n = (270f/4 = 18225.0
Y = Target value = 0
0
ST = sum of square of all results - C. F.

= (Yr + ... + Y~) - C.F.

- 702 + 652 + 75 2 + 6Ql- 18225.0
= 18350.0 - 18225.0 = 125.0
The total contribution of each factor is calculated as shown below:
AI = 70 + 75 ~ 145 .A 2 ::: 65 + 60 ::: 125

B I = 70 + 65 = 135 8 2 = 75 + 60 = 135
and,
SA ~. Ar/NA) + .A~/NA2 - C.F.
Ss = Bi/Ns ) + B~/Ns2 - C.F.

2 2
SAS = ~ 2 (A;8j)2/rij - C.f.
1=lj::)
SAXS = SAS - SA - Ss
• Sr ::: Se 4: SA + Ss + SA xIJ
where NA), N A2 etc., refer to the Dumber of trial runs included in the
sums A., Az etc . .All) is the lotalexperimental response for the factor A
at level fand the factor Bat levelj whereas r;j is the number ofreplicatioDS
(observations) for the cell ij. Thetenn SA x s represents the interaction
sum of squares.
For the above example,
SA = 14.52/2 + 1252/2- 18225.0 ::: 18325.0 - 18225.0 = 100.0

Ss ::: 1352/2 + 1352/2 + 18225.0 ::: 18225.0 - 18225.0 ::: 0.0
Since r = 1 (one observation per cell)
and
SJ.< 8 = 125.0 - 100. - 0 = 25.0

I2Q A Primer on me Taguchi Method
Therefore, using Eq. (6-11), the error variation becomes,
= 125 - 100 - 0 - 25 =0
Variance Calculations:
VA = SAl/A - 10011 = 100

VB :;:: SBIfB = 011 = 0
:;:: 0/0 indeterminate, hence not useful.
Variance Ratio
Once all the variances are computed, the results can be arranged, in a
tabular fonn as appears in Table 6-9. .
Observe that the OOF and Se of the error terms are zero, hence F, the
ratio of the variances, cannot be computed. Thus this experimental design
is not effective for studying the interaction of factors A and B. Additional
degrees of freedom are necessary for a complete analysis of the inter-
actions and main effects. This can be accomplished by repeating the
observations for each setup so that there will be an error term which will
have non-zero degrees of freedom and variance terms.
Table 6-9. ANOVA Table for Tire Wear Characteristics-Example 6-3

VARIANCE VARIANCE PURE PERCENT
SOt"RCEOf SUM OF (MEAN SQUARES) RATIO SUM Of SQUARES CON11UBl.l11ON
. VAlUATION ( SQUARES V F S' p
A 1 100.00 100.00
B 1 .0.00 0.00
AxB 1 25.00 25.00
Error (e) 0 0.00
Total 3 125.00
6·6 EXPERIMENTS WITH REPUCATIONS
Example 64
Example 6-3 is extended to two observations per cell as shown in Table
6-10. In Table 6-11, the data in each cell are replaced by a single value
obtained by adding f:he two data points. The total of the degrees of freedom
IT is increased:
SInce, n - r X 22 ::::: 2 x 4 ::::: 8
and IT : : : n - 1 ::::: 8 - 1 ::::: 7
The degrees of freedom for other factors are as follows:
lA : : : number of levels - 1 ::::: 2 - 1 ::::: 1
IB : : : types of tires - 1 ::::: 2 - 1::::: 1
IAXB : : : 1 xl::::: 1
le ::::: IT - lA - IB - lA xB ::::: 7 - 1 - 1 - 1 ::::: 4
and
C.F. ::::: Correction Factor ::::: T2/n = (542)2/8 ::::: 36720.5
Table 6-10. Tire Wear E~ts

with Repetitions--Example 6-4
TEMPERA11JRE
A
TIRE TYPE AI = T... A2 = Ts

Wea.rv.'ell 70.72 65.62
BI YI Y3
B
Superwear 75. 77 60, 61
B-;. Y-;. Y4
Table 6-11. Tire Wear Experiments with

Repetitions-Example 6-4
TEMPERATURE
A
tIRE TYPE .41 :::: Tw Az ::: Ts SUM
Wearwell 142 121 269

B
8. Y. Y3
Superwear 152 121 273
82 Yz Y4
Sum 294 248 542 = Grand total
Assuming
Yo == target value :=: 0

ST == sum of square of all 8 data points - C.F.
:=: (yi ....... + y~) - C.F.
== 7fil + 722 + 75 2 + 772 + 65 2 + 622 + 6fil + 61 2 - 36720.5

== 37028.0 - 36720.5 - 307.5
The contribution of each factor is shown below:
Al - 142 + 152 = 294 A2 == 127 + 121 == 248
B I == 142 + 127 - 269 B2 == 152 + 121 == 273
and
SA == 2942/4 + 248 2/4 - 36720.5 == 264.5
S8 == 2692 /4 + 273 2/4 - 36720.5 2.0

~ == {A I B t )2/2 + (A 1B2f/2 + .{A 2B 1)212 + (A 2B2 )2/2 - C.P.
AS
== [(142)2 + (152)2 + (127)2 + (121)2]12 - 36720.5
:=: 1.7019 '-' 1.6.. . ..,0 '
..- _. -. "~"
=. ")98 .: ;
·Pd .....
· I
Analysis of V&nce (ANOVA) 123
Also:
SAXB == 298.5 - 264.5 - 2 == 32

Se = ST - (SA + SB + SAXB)
;::: 307.5 - 264.5 - 2 - 32 = 9.0
Variance Calculations: ;.
VA = SA!fA = 264.5/1 ;::: 264.5

VB = SB!fB = 211 = 2
VAXB ~ SAxB'fAxB == 32/1 = 32
Ve == S/Ie = 9/4 == 2.25
Also
FA == VAlVe = 264.51.2.25 == 117.6

FB = VBIVe - 212.25 == 0.89
FAXB ;::: VAxBNe = 3212.25 = 14.22
Pure Sum of Squares:
S~ = SA - lA X Ve ;::: 264.5 - 1 x 2.25 ;::: 262.25

S~ == Ss - IB X Ve = 2 - 1 x 2.25 = -0.25
S~xB == SAXB - IAXB X VII' ;::: 32 .:- 1 x 2.25 = 29.75
S; = Se + (fA + IB + IAxB) X Ve = 9 +.3 X 2.25 = 15.75
and the percentage contributions will be:
PA == S~ X lOO/ST = 262.25 x 100/307.5 =15.28

PB = S8 X 100/ST -0.25 x 100/307.5 - -0.08
P.-\><B = s.~. B X 100lST 29.75 x 100/307.5 = 9.68
PI! == S;. X lOO,ST == 15.75 x 100/307.5 = 5.12
Table 6-12. ANOVA Table for Tire Wear with Repetitions-Example 6-4
SOLl'RCE VAIUANCE VARIANCE Pl."RE PERCENT
OF SUM OF fMEANSQUARES) RATIO SUM OF SQUARES CONllUBIJl1ON
VARIAnoN f SQUARES v F s' p
A 1 264.50 264.50 117.50 262.50 85.28

B 1 z.OO 2.00 0.80 -00.25 -0.08
A x B 1 32.00 32.00 14.20 29.75 9.68
Error (e) 4 9.00 2.25 1.00 1.00 5.12
Total 7 307.50 100.00
Since the number of the degrees of freedom for the numerator is 1 (see
Table 6-12) and that for the denominator is 4, from the F Tables at .05
level of significance (95% confidence), we obtain F.osO,4) :::: 7.7086.
The computed values of variance ratios F for factor A an~ interaction
A x S, are greater than the limiting values obtained from the table.
Therefore, there is a significant difference in the wear life of the tires
under summer and winter conditions. The interaction term, FAX B indicates
that the influence of temperature on the two brands of tires is also sig-
nificant. However, FB is less than the table F factor. Thus, there is no
difference between tire brands within the confidence level.
6-6-1 Procedures for Pooling
I, When the contribution of a factor is small, as for B in the above example,

the sum of squares for that factor is combined with the error S~. This
process of disregarding the contribution of a selected factof and subse-
I quently adjusting the contributions of the other factor, is known as Pool-

ing. Pooling is usually accomplished by starting with the smallest sum
of squares and continuing with the ones baving successively larger effects.
Pooling is recomm~nded when a factor is determined to be insignificant
by performing a test of significance against the error term at a desired
confidence level. A general guideline for when to pool is obtained by
comparing error DOF with the total factor OOF. Taguchi recommends
pooling factors until the error DOF is approximately half the total DOF
of the experiment (Ref. 10, pp. 293--295). Approaching the matter tech-
nically, one could Itest for significance and pool all factor influences below
the 90% confidence level. The procedure for significance testing will be
discussed later in this chapter. For now. we will arbitrarily select small
,
I
factor effects and pool. Consider the pooling effects of the factor B. If
the variance for Ibis factor is pooled with the error term, the new error
variance is computed as:
Ve = (SB + S,)/(fB + le)

, = (2.0 + 9.0)/(1 + 4)= 2.2
With a pooled V" all Si values must be modified to reflect pooling:
S~ = SA - (Ve x lA) = 264.50 - 2.20 =262.30

S~xB = SAXIl - (Ve X IAXB) = 32.0 - 2.2 x 1= 29.8
-
S; = Se + Ve(fA + IAxB) =: 11.0 + 2.2 x 2 = 15.4
The results of this procedure are summarized in Table 6-13, which makes
it apparent that pooling in this particular case does not appreciably change
the results. But in certain cases, the process may significantly affect the
results. No matter the effect on the results, insignificant factorssbould
always be pooled.
It is quite evident from these considerations, that the ANOVA pro-
cedure is cumbel'someandextremely time-consuming. The computations
needed increase tremendously as the size of the matrix increases. The
design of the experiment and the subsequent analysis of the test results
can be simplified using available software specifically made for analysis
of Taguchi experimental designs. Most of the computations shown in this
book have been carried out using QUALITEK-3 software (Ref. 11). To
Table 6-13. ANOVA Table for Tire Wear with Repetitions and PooliDg-
Example 6-4
SOURCE VARIANCE VARIA.~CE PURE PERCENT
OF SUM OF (MEAN SQCARES) RATIO SUM OF SQUARES CONTRIBUTION
VARIATION f SQUARES v F s' p
A 1 264.50 264.50 120.20 262.50 85.30

B Pooled
A x B 1 32.00 32.00 14.50 29.80 9.69
Error (e) 5 11.00 2.20 l.00 15.40 5.01
Total 7 307.50 100.00
further clarify the step by step procedure involved in analysis of variance,

the following numerical example is presented.
6-' STANDARD ANALYSIS \\lTH SINGLE AND

MULTIPLE RUNS
Example 6--5
A Taguchi experiment was designed (Table 6-14) to investigate five 2
level factors (A. B. e, D. and E) and two interactions (A x e and
B x C) of a certain manufacturing operation. The La ortbogo~ array
was used for the design and one sample that was produced under each
eXperimental configuration was examined. The results are shown in
Table 6-15.
6--'-1 Analy. Using SiDgIe Run

Level TotalslDd Their Averages
The factor ave!ages at each factor level are obtained by adding the results
of all trial conditions at the level considered, and then dividing by the
numbers of data points added.
Al = (YI + Y2 + Y3 + Y4)/4 = (42 + 50 + 36 + 45)/4

= 173/4 = 43.25
Table 6-14. Factors and Their Levels-Example 6-5

COLUMN FACI'OR NAMES LEVEL I LEVEL 2 LEVEL 3 LEVEL 4
1 Factor A Al
2 Factor C Cl
3 Interaction A x C N/A
4 Factor 8 81
6 Factor D DI
7 Interaction 8 x C N/A
8 Factor E El
Objective: Determine hest :fesign paralnetm.

:~J.rK~e:lst!c: '':'he ;[J1aller '.he '.letter
Table 6-15. Layout and Results-Example 6-5

FACTORS: A e Axe B D Bxe E RESULTS
l1UAl1COLl'MN I 2 3 4 S 6 7 I (y)
Trial 1 1 1 1 1 1 1 1 42.00
Trial 2 1 1 1 2 2 2 2 SO.OO
Trial 3 1 2 2 1 1 2 2 36.00
Trial 4 1 2 2 2 2 I 1 45.00
TrialS 2 I 2 I 2 1 2 ;.
35.00
Trial 6 2 I 2 2 1 2 1 55.00
Trial 7 2 2 1 1 2 2 1 30.00
TrailS 2 2 1 2 1 1 2 54.00
Total ::: 347.00
A 2 = (Ys + Y6 + Y7 + Y8)/4= (35 + 55 + 30 +54)/4

= 174/4 = 43.50
Cl = (Yl + Y2 + Y5 + Y6)/4 = (42 + 50 + 35 + 55)/4
= 18214 = 45.50
C2 = (Y3 + Y4 + Y7 + Y8)/4 = (36 + 45 + 30 + 54)/4
= 165/4 == 41.25
Similarly
B1 = 143 8 1 == 35.75
B2 - 204 B2 = 51.00
Dl = 187 Dl = 46.75
D 2 = 160 D2 = 40.00
El = 172 E I = 43.00
E2 = 175 E2 = 43.75
and
--- ,,- ,-- -,----,--,.
128 A PIimer on the Taguchi Method
(A x Ch ;: :;: (Y3 + Y4 + )'S + }'6)/4 ;:::;: 17l!4 = 42.75

(B x C)l ;:::;: (Yl + 'Y4 + )'5 + )'8)/4 ;:::;: 17614 ;:::;: 44.00
(B x Ch = (Y2 + Y3 + }'6 + }'7)/4 ;:::;: 17114 ;:::;: 42.75
The remits are shown in Table 6-16 and the main effects are plotted in
Figure 6-1.
Ignoring interaction effects and assuming the "smaller is be~ char-
acterislie is desired, the optimum condition becomes:·
Computation of Interaction
Interaction effects are always mixed with the main effects of the factors
assigned to the column designated for interaction. The relative signifi-
cance of the interaction effects is obtained by ANOVA just as are the
relative significltce of factor effects. To determine whether two factors
A and C interact; the following calculations are performed.
:li
Level Totals and Their Averages for A and C
AICl!;:::;: (YI + Y2)/2 == (42 + 50)/2 = 9212 46.0
AIC:! = (Y3 + Y4)12 - (36 + 45)/2 ;:::;: 8112 ;:::;: 40.5
~A2CIl ;:::;: (Y5 + Y6)/2 = (35 + 55)12 ;:::;: 9012 ;:::;: 45.0
A 2e:! ;: :;: (Y7 + Y8)12 - (30 + 54)/2 ;:::;: 8412 - 42.0
Table 6-16. Average Effects-Example 6-6

COLUMlil FAcroRS LEVEL 1 lEVEL 2 (L2 - I'd lEVEL 3 LEVEL 4
1 Factor A 43.25 43.50 -0.25

2 Factor C 45.50 41.25 -4.25
3 Interaction A x C 44.00 42.75 -1.25
4 Factor B 35.75 51.00 15.25
6 Factor D 46.75 40.00 -6J75
7 Interaction B x C 44.00 42.75 -1.25
8 . . . . . ._._Factor
____. .--.._._..._. .E 43.00
...__.. . . . __ ~_......-
43.75
__.. 0.75 . .--.. . . . . .__.. . . .
.·_w_._..-_. .. ~ . . .--'-
51
46.75
..
~
45.0
43.25
.. ~
43.5 •
43.0
....
43.75
41.25
40
35.75
C2 8, 82 0,
factors
Figure 6-1. Main effeas--Example 6-6,
Level Totals and Their Averages for B and C

BICI = (YI + Ys) = (42 + 35)/2 = 77/2 = 38.S
B I C2 = (Y3 + Y7) = (36 + 30)/2 = 66/2= 33.0
B~I = (Y2 + Y6) = (50 + 55)/2 = 105/2 = 52.5
B1.C2 = (Y4 + Ys) = (45 + 54)/2 = 99/2 = 49.5
These results are plotted in Figure 6-2. Note that lines B. and 8 2 appear
almost parallel. Hence, B and C interact slightly. Note also that A. and
A 2 intersect, thus, A and C interact.
52.50
38.50
c,
F:~ure 5-2. Ir.terac~lOn.-\ < C lr.J B ( C-Exam['ie 5-6.

130 A ~r 08 the Ta.guchi Method
The minor interaction of B x C is ignored but interaction A x C is

included in the optimum condition. For the "smaller is better" situation.
A I C 2 (40.5) produces the lowest average value. Therefore, Aiel is in-
cluded in the optimum condition. It so happens that A I C2 was already
included in the optimum condition, based on the main effects alone. (AI
Cl BI D2 El)
Analysis of Variance (AJ.~OVA)

ANOVA establishes the relative significance of the individual factors and
the interactioD effects. The steps are as follows.
Step 1.. Total of All Results:
Step 2. Correction Factor:
C.F. = PT", = 3471/8 = 15051.125

"~*~'"
,,'"
..
Note: n = To" number of experiments, 8.
Step 3. Total Sum of Squares:

8
ST = ~ i - C.F.
i= I
= (42 2 + 5()2 + 36: +

= 599.88
Step 4. Factor Sum of Squares:
SA = Ai/NAI + AYNAl - C.F.

= 1132/4 + 174214 - 15051.125 = 0.125
SB = Bi/NBI + B~/NB2 - C.F.
= 1432/4 + 204:,4 -- 15051.125 = 465.125
Se = CrfNCl + C~i.vc: - C.F.
Analy~i~ of Variance (AHOVA) 131
So - DI!NDl + D~/No2 - C.F.
· 187 2/4"1 16(}2/4 - 15051.125 ::;:: 91.125

SE::;:: ET/NEt + ~/NE2 - C.F.
::;:: 1722/4 + 175 2/4 - 15051.125 ::;:: 91.125
,
SAXC ::;::(A X C)'i/N(AXC)I + (A X C)'}/N(AXCrz. - C.F.
- 1762/4 '+ 171 2/4 - 15051.125 ::;:: 3.l2S

S8xC ::;:: (B X C)'iIN(BXC)I + (B X C)'YNuJxcfJ. - C.F.
::;:: 1762/4 + 17P/4 - 15051.125 ::;:: 3.125
Alternative Formula for 2 Level Factors

In the case of 2 level factors, the sums of squares can be computed using
the following formulas:
SA::;:: (AI - A 2)2 ::;:: (173- 174)2 ::;:: 0.12S

(NAI + N A2 ) (4 + 4)
58 = (B I -B 2 )2 = (143 - 204)2 ::;:: 465.125

(NB I + N 82 ) (4 + 4)
Sc::;:: (Cl - C2)2 ::;:: (182 - 165)2 = 36.125

(NcI + NC2 ) (4 + 4)
SD= (D I - D 2)2 = (187 - 160)2= 91.125

(NDI + N02 ) (4 + 4)
SE::;:: (El - E 2)2= (172 - 175f= 1.115
(N£l -+. NE2 ) (4 + 4)
SAxC ::;:: «A X C)l - {A x Chf = 3.125

(N(A XCII + N(A XC)2)
132 A Nimer on the Taguchi Method
where
NA1 = Total number of experiments in which factor A. is present
NB I = Total n~mber of experiments in which factor 8 I is present
A. = Sum of results (Yi ) where factor A. is present

8.= Sum of results (Yi ) where factor 8 1 is present
Step 5. Total and Factor Degrees of Freedom (DOF):
DOF total = Number of test runs minus I

or IT=n-l =8-1 =7
DOF of each factor is 1 less than the number of levels:
lA = (Number;C)f levels of factor A) - I

=2 - 1 = 1
IB = fNumberOf levels of factor 8) - 1
=2- 1 =1
le = tNumber of levels of factor C) - 1
=2 - 1 - 1
ID - «Number of levels of factor D) - I
=2 - I = I
lE = (Number of levels of ,factor E) - 1
- 1 - 1 = 1
Also,
!<.Axn = lA X le
=I x I - 1
f'<8 XCI = 18 X Ic
.,'
DOF of the error term in this example:
le = IT - (fA + Is + le + ID + lE + f4. "c + fs xC>

=7-7 =0
With the error degrees of freedom equal to zero, t = O,infonnation

regarding the error sum of squares cannot be determined. In addition F
ratios for factors cannot be calculated since the calculationsiDvolve t.
To complete the calculations, smaller factorial effects are added together
(pooled) to form a new nonzeroestimate of thearor term. This is dis-
cussed in the following section.
Step 6. Mean Square "(Variance):
VA = SA!fA - 0.125/1 = 0.125

Vs = Ss!fs = 456.125/1= 465.125
Vc = Sc!fc == 36.125/1 == 36.125
VD = SolID == 91.125/1 = 91.125
VE = SE!fE - 1.125/1 = 1.125
0/0 = Indeterminate
As the variance of error term (Ve ) is zero. the variance ratio and pure
sum of squares (5') cannot be calculated. In this case the percentage
contributions are first calculated using sums of squares. Then, if there
are insignificant factors, pool them and recalculate percentages using the
pure sums of squares.
Step 7. Percentage Contribution:
0.125/599.88 = 0.02
PB = SB1ST = 465.125/599.88 = -7.54
Pc = SC/ST - 36.125/599.88 = 6.02

PD = SD ST - 91.125/599.88 = :5.20
134 A Primer OD die Taguchi Method
PAxe = SAxelSr = 3.125/599.88 = 0.52

PBxe = SBxc!ST = 3.125/599.88 = 0.52
and Pe cannot be calculated as Ve - 0
The results of the analysis of variance are summarized in Table 6-17.
6-7·2 PooUng
Note that in Step 7, the effects of factors A. E. and interactions B x C

and A x C are small, totalling slightly more than 1% (1.2%). These
factors are pooled to obtain new, nonzero estimates of S" and h.
Sum of Squares of ~rror:
Let: Se = SA + Ss + SAxe + SBxC

then Se = ST - (SB + Se + So) = 599.9 - (592.4) = 7.5
Degree of Freedom of Error Term:
h = IT - (fs + le + ID)
=7-3=4
Variance of Error Term:
Ve = Selle = 7.5/4 = 1.875
Table 6-17. ANOVA Table-Example 6-6

COLUMN FAcroRS f 5 V F p
I Factor A I 0.125 0.125 0.02

2 Factor C I 36.125 36.125 6.02
3 Interaction A xC 1 3.125 3.125 0.52
4 Factor B 1 465.125 465.125 77..54
5 Factor D 1 91.125 92.125 15.20
6 Interaction B XC 1 3.125 3.125 0.52
7 Factor E I 1.125 1.125 0.19
All other/error 0 0 0
Total 599,380 :00 ')Get-
';
Analysis of Variance (ANOV}\) 135
Factor F Ratios, for Significant Factors:
Fe = VC/Ye = 36.125/1.875 = 19.267

Fs = VB/Ye = 465.125/1.875= 248.067
Fo = VvlVe = 91.125/1.875 = 48.600
Pure Sum of Squares S', for' Significant Factors:
Se = Se - (Ve X le)
== 36.125 - (1.875 X 1) = 34.25
S~ = SB - (Ve X Is)
:::: 465.125 -' (1.875 X 1) = 463.25
So = So - (Ve X 10)
- 91.125 - (1.875 X 1) = 89.25
Note that the ANOVA in Table 6-18, the pure sum of squares, S' is not
shown.
Percentage Contribution:
Pc :::: SC/ST = 34.25/599.88 == 5.71%

, .
PB == SBlSr == 463.25/599.88 :::: 77.22%
Table 6-18. Pooled ANOVA-Example 6-6

COLUMN FACTORS { s V F P
1 FaetorA (1) (0.13) Pooled

2 Factor C 1 36.125 36.125 19.267
. 5.71
3 Interaction A X (I) (3.125)
Pooled
C
4 Factor B 1 465.125 265.125 248.067 77.22
5 Factor D 1 91.125 91.125 48.600 14.88
6 Interaction B x (I) (3.125)
Pooled
C
7 Factor E (I) (1.125) Pooled
All other'error .t 1.500 1.88 2.19
.....0(;\.· ~C}q 380 ·OO.00CC
..•._....•...• '_, '~'. __ .·_...._.• ·_'"~'.d .•.•• _. _,... " ',___ ..".• __.••• ' _ L . . . ". _
~ •• ,._ _ _ _ _ ...._ ......... ~ ................... _ •.••••_ . ~ •• ~ 4_~",,,
IJ6 A PriB1er on the Taguchi Method
Po = S!fSr - 89.25/599.88 = 14.88%

Pe 100% - (Pc + Ps + Po) - 2.19%
The AHOVA terms which are modified after pooling are shown in Table
6-18.
Tagudli~s guideline for pooling (Ref. 10, pp. 293-295), .requires a
start with dle smallest main effect and successively includes larger effects,
until the total pooled DOF equals approximately half of the total DOF.
The larger DOF for the error term, as a result of pooling, increases the
confideal:e level of the significant factors.
Note 6d as small factor effects are pooled, the Percentage contributions
and the coofidence level of the remaining factors decrease (Pc = 5.71
versus Pc = 6.02). By pooling, the error term is increased and in com-
parison . . other factors, appear less influential ~ The greater the number
of facton pooled the worse the unpooled factor effects look. Then we
must cotJSider why are column effects pooled?
Error Yariance represents the degree of inter-experiment error when
the DOF of the e~or term is sufficiently large. When the error OOF is
small or am, which is the case when all columns of the OA are occupied
and trials are not repeated, small column effects are successively pooled
to fonna larger error term (this is known as a pooling up strategy). The
factors . . interactions that are now significant in comparison with the
larger mapitude of the error term are now influential. Taguchi prefers
this strateIY as it tends to avoid the mistake (alpha mistake) of ignoring
helpful factors.
A large error DOF naturally results when trial conditions are repeated
and standard analysis is performed. When the error DOF is large, pooling
may not be necessary. Therefore, one could repeat the experiment and
avoid pooting. But, to repeat all trial conditions just'for information on
the error am may not be· practical.
A sure way to determine if a factor or inter~ction effect should be
pooled is to perform a test of significance (1- confidence level). But what
level ofamfidence do you work with? No clear guidelines are established.
Generally, a confidence level between 90 and 99% is recommended.
However, if the confidence level is below 90%, then it is a common
practice to pool the factor for Example 6.5, fac~or C is tested for signif-
icance at Cl 99% confidence level. Since factor C' contributes only 5.71 %,
test for significance and determine if this factor should be pooled. From
the :\~OV.-\ :able
..,. ------ -----------
Analysis of Variance 1A."lQVA) 137
Fe == 19.267
From the F Table 7 find the F value at
nl = DOF of factor C = 1
n2 =DOF of error term = 4
at a confidence level (say the 99% confidence level).
F == 21.198 (From Table C-4)
As Fe from experiment (19.267) is smaller than the F Table value (21.198),

factor C should be pooled.
SUMMARY RESULT
Description of the factor = Factor C

Column the factor is assigned to = 2
Variance ratio for this factor is - 19.267
DOF of the factor - 1
DOF of error term 7le = 4
Confidence level % = 99
Based on the level of confidence desired (99%),
the following is recommended.
"Pool this factor'7
The revised values are calculated as shown below:
Se = ST - (S8 + SD) = 599.9 - (556.25) = 43.625

le = IT - (f8 + ID) - 7 - 2 - 5
Ve = Selle ;:::: 43.625/5 ;:::: 8.725
Fe - VelVe = 465.125i8.725 = 53.309

138 A PJ:imer 011 the Taguchi Method
S~ = Ss - (Ve X Is) = 465.125 - (8.725 X 1) = 456.40

So - So - (Ve X 10) 91.125 - (8.725 X 1) = 82.40
Ps = S8IST = (456.40 X 100)/599.88= 16.08
Po = SDlST = (82.40 X 100)/599.88 = 13.74

PI: = 100% - (Ps + Po) 10.18
6-7·3 COIIfidenee Interval or Factor Effect

The confidence interval of estimates of the main effect is calculated using
the following expression:
C.I. = ± v'(F(l ,n~ X Ye/Ne)
Where
FO ,n2) = F value from the F Table at a required confidence level

and at DOF I and error DOF n2
Ve = Vanance of error term (from ANOVA)
Ne = Effective number of replications

Total number of results (or number of SIN) ratios)
- DOF of mean .( = 1, always) + DOF of all factors
included in the estimate of the mean.
Thus for the factor C at level Cl, the C.I. is calculated by first determining
the F factor:
Ne == 8/(1 + 1)= 4
F(l,4) = 7.7086 at 95% confidence level

then C.I. = ± 1.9034 at 95% confidence level.
SInce Cl = 45.50 (Table 6-16)
Analysis of Variant:e (ANOVA) 139
SUMMARY RESULTS
Based on: \
F value from the table (at a confidence level) = 7.7086
Error variance, Ve = 1.88
Number of effective replications =4
;..
The confidence interval C.l. is calculated as follows:
C.l. represents the variation of the estimated valueofdle main effect-

of a factor of the result at the optimum at .a confidence level used for
the F values.
Confidence interval (C. I.) = + 1.9034
6..74 Estimated Result at Optimum Condition

The perfonnance at the optimum condition is estimated only from the
significant factors. Therefore, the pooled factors are ootincluded in the
estimate.
Grand average of perfonnance: T = 347/8 = 43.375

As the factors B, C. and D are considered significant, die perfonnance
at the optimum condition will be estimated using only these three factors.
= T + (li 1 - T) + (D2 - T) + (C2 - T)
= 43.375 + (35.75 - 43.375) + (40 - 43.375)

+ (33 - 43.375)
= 22.000
Note that the optimum condition for the "smaller is better" quality char-
acteristic is B ~ C: D:.. The average values at these conditions were pre-
.,,, ,-:
,,\.C~,\ 'l~'
~a\. J,ak,.
I ,·'1
. rn
1::-- ,)l.Cl.!d3.nZc
I T,a bl.:>
'. "'J:.n. . I". {:.
)- I ,
'Q
140 A Pmuer on the Taguchi Method
Table 6-19. Estimate of Performanee at the Optimum

Condition-Example 6-6
LEVEl. DESCRIPllON
Fador C ;12 -10.3750

fadorB I - 7.6250
fauorD 2 - 3.3750
c.n"bution from all factors (total) -21.375
Cuamt grand average of performance 43.375
~ result at optimum condition 22.000 .
6-7-5 Ceddence Interval of tbe Result at the OptimUDl

Coadition
The eXplCSSion for computing the confidence interval, for perfonnance
at the opIimum condition, is calculated in the same way as are the factor
effects.
Where
F( l,llz} = The F value from the F Table ata required confidence

level at DOF 1 and error DOF n2
Ve = Variance of error tenn (from ANOVA)
!It = Effective number of replications
Total number of results (or number of SIN ratios)
-
DOF of mean ( = I, always) + DOF of all factors
included in the estimate of the mean.
Three factors, Bb C2 , and D2 are included in the making of the per-

fonnance estimate at the optimum condition. Therefore, the effective
number of replication, F value and the confidence intervals are calculated
as shown below. A confidence level of 90 to 99% is the non:nal range
of selection for common industrial experiments. A 90% confid~nce level
is arbitrarily selected for the following calculations.
_0 ~----------------l
Analysis of Variance (ANQVA) 141 ~
Ne ;::: 8/(1 + 3) ;:::: 2

I
a
t
F(l,4) ;:::: 4.5448 at the 90% confidence level
Ve ;:::: 1.88 ,
;
I
I
C.I. ;::: + 2.067 at the 90% confidence level. I
Therefore the result at the optimum condition is 22.0 + 2.067 at the

90% confidence level.
- SUMMARY RESULTS
where F( n1,»2) ;:::: Computed value of F with nI ;::: I, »2 ;::: error

DOF, ata desired confidence level,
Ve ;:::: Error variance
Ne ;:::: Number of effective replications
\
tI
.l, The confidence interval (C.I.) ;:::: + 2.067
,
Which is the variation of the estimated result at the optimum I.e., the
mean reSUlt, (m) lies between (m + C.I.) and (m - C.I.) at ~
confidence level.
The confidence interval formula assumes a sufficiently large number

of data points so that the sample approximates the population character-
istics. With a "small" sample, its characteristics may deviate from the
population. When the sample is small, Le., only a finite number of
continuation tests are planned, the C.I. of the expected result is expressed
as:
142 A Priltter on the Taguchi Method
Where
F( n I Jf21 = Computed value of F at a desired confidence level with

nl = 1 and error DOF nz
VI' = Error variance
and Ne = Effective number of replications.
Based on:
nl = 1 V,. = 1.88
NI' = 2 N, = 3
F.• O.,4) = 4.5448 at the'9O% confidence level.,

then C.l. == ± 2.668 at the 90% confidence level. This is a wider interval
than paviously adculated based on a large number of repetitions..
6-7-6 Analysis witb Multiple Runs
Assume die trial nlDS of the experiment were each repeated three times
and that die average result of each trial as shown Table 6-20., is the same
as that for a single trial in Table 6-15. The analysis takes a slightly
different fonn. Since the averages of these hypothetical results were kept
the same~ the main effects remain unchanged as shown in Tables 6-21
Table 6-28. Results with Three Repedtions-Example 6-6

TRLo\lJ
REPE'l1llONS RI R2 R3 Rs • AVERAGE
-:
I 38.00 42.00 46.00 42.00
2 45.00 50.00 55.00 50.00
3 38.00 36.00 34.00 36.00
4 55.00 45.00 35.00 45.00
5 30.00 35.00 40.00 35.00
6 65.00 55.00 \ 45.00 55.00
7 40.00 30.00 20.00 30.00
8 58.00 54.00 50.00 54.00
-----_ •.. --- -------------------------------------
Analysis of Variance (ANOVAl 143
Table 6-21. The Main Effeds-Example 6-6

COLUMNS FACTOR NAMES LEVEL I LEVELl (L2 - L.) LEVEL 3 LEVEL 4
1 Factor A 43.25 43.50 -0.25

2 Factor C 45.50 41.25 -4.25
3 Interaction A x C 44.00 42.75 -1.25
4 Factor B 35.75 51.00 15.25
6 Factor D 46.75 40.00 -6.75
;.
7 Interaction B x C 44.00 42.75 -1.25
8 Factor E 43.00 43.75 0.75
and 6-16. However, the results of ANOVA (Table 6-22) are significantly
different from- the corresponding result of the single run (Table 6-18).
,
I
! Computation for ANOVA:
, DOF = total number of results - 1

= number of trial x number of repetitions - 1
::: 8x 3 - 1 ::: 23
Sample Calculations Using Factor B:
B 1 = (38.0 + 42.0 + 46..0) + (38.0 + 36.0 + 34.0)

+ (30.0 + 35.0 + 40.0) + (40.0 + 30.0 + 20.0) == 429
Table 6-22. Pooled ANOVA
COLUMN FACTORS f S V F P
1 Factor A (I) (0.38) Pooled

2 Factor C 1 108.375 108.375 2.728 2.67
3 Interaction A xC (1) (9.380) Pooled
4 Factor B I 1395.375 1395.375 35.126 52.72
5 FactorD t 273.375 273.375 6.882 9.09
6 Interaction B x C 0) (9.380) Pooled
7 Factor E (I) (3.380) Pooled
All other/error 20 794.500 39.72 35.53
Total: ::!3 2571.630 lOO.OOC;C
Note that the trial condition for calculating B I is (1 3 5, and 7)

9 9
B2 == (45.0 + 50.0 + 55.0) + (55.0 + 45.0 + 35.0)

+ (65.0 + 55.0 + 45.0) + (58.0 + 54.0 + 50.0) = 612
Using the sum of squares fonnula for 2 level factor,
Sa = (B I --B2)'- (429-- 612)2

(Nal + Nai)
- 24
- 1395.375
Va = Sa/fa :::: 1395.375/1 - 1395.375

Fa == ValVe = 1395.315/39.72 = 35.126
S;' - Sa --Ve X fa:::: -1395.275 -- 39.72 = 1355.650
p. = 100 X S;,JST == 100 X 1355.65/2571.63 = 52.70%
The ANOVR and the performance at the optimum coadition are as

shown iD TabJ'16-22 and Table 6-23, respectively.
Wheu trial ru's are repeated, ANOVA produces different results with
larger error DOF and thus a higher level of confidence in the estimate
optimum perfonnance and the factor influences. ANOVA results from
multiple repetitions should always be preferred as this will yield the robust
design and reproducible perfonnance estimate. For this reason repetition
is highly desirable. But, since rePeating trial runs may be expensive, it
must be weighed against the need for robustness.
Table 6-23. Estimate of Performance at the Optim~m

Condition-Example 6--6
FACl'OR DESCRIPTION LEVEL CONl'RIBUTION
Factor C 2 -2.1250
Factor B 1 .-7.6250
Factor D 2 -3.3750
Contribution from an factors -13.375
O1rrent grand average of performance 43.375
Expected result at optimum condition 30.250
6-8 APPLICATION OF THE SIN RATIO

The change in the quality characteristics of a produc, under investigation,
in response to a factor introduced in the experimental design is the "signal"
of the desired effect. However, when an experiment is conducted. there
are numerous external factors not designed into the experiment which
influence the outcome. These external factors are called the noise factors
and their effect on the outcome of the quality characteristic under test is
termed ''the noise." The signal to noise ratio (SiN ratio) measures the
sensitivity of the quality characteristic being investigated in a controlled
manner, to those external influencing factors (noise factors) not under
control. The concept of SIN originated in the electrical engineering field.
Taguchi ~ffectively· applied this concept to establish the optimum con~
I dition from the experiments.

The aim of any experiment is always to determine the highest possible
SIN ratio for the result. A high value of SIN implies that the signal is
much higher than the random effects of the noise factors. Product design
or process operation consistent with highest SIN, always yields the ap-
timum quality with minimum variance.
From the quality point of view, there are three possible categories of
quality characteristics. They are:
I. Smaller is better, e.g., minimum shrinkage in .a cast iron cylinder

block casting.
2. Nominal is best, e.g., dimension of a part consistently achieved
with modest variance.
3. Bigger is better, e.g., maximum expected life of a component.
The SIN analysis is designed to measure quality characteristics.
6-8-1 Conversion of Results into SIN Ratios

The conversion of a set of observations into a single number. the SIN
ratio, is performed in two steps. First. the Mean Squared Deviation (MSD)
1
j
of the set is c.alculated. Second. the StN ratio is computed from the MSD
,
by the equation.
5 IV = - 10 Log w (~lSD) (6-16 )

146 A Primer OD the Tagnchi Method
Note that for the SIN to be large, the MSD must have a value that is
small.
The smaller is better quality characteristic:
MSD = {Yr + Yl + . . . + Y~)/N (6-17)
The nominal is the hest quality characteristic, Yo = n()minal or target

value:
MSD == «f. -- f~ + (Y2 - f o)2 + ... + .(YN -- Yo)~/N (6-18)
The bigger is better quality characteristic:
MSD = (l/ff + lIY~ + ... + lIY~)/N (6-19)
The Meaa ~ Deviation (MSD) isa statistical quantity that reflects

the deviation film the target value. The expressions for the MSD are
different for 4iJerent quality characteristics. For the nominal is best
characteristic, tile standard definition of MSD is used. For the other two
characteristics, the definition is slightly modified. For smaller is better,
) the unstated target value is zero. For larger is better, the inverse of each
large value becomes a small value and again, the unstated target is zero.
Thus for all three MSD expressions, the smallest magnitude of MSD is
being sought. In turn this yields the greatest discrimination between con-
trolled and uncontrolled factors. This is Tagucm's solution to robust
product or process Oesign.
Alternate forms of definitions of the SIN ratios exist (Ref. 7, pp.. 172~
173), particularly for the nominal is the best characteristic. The definition
in terms ofMSD is preferred as it is consistent with Taguchi's objective
of reducing variation around the target. Conversion to SIN ratio can be .
viewed as a scale transformation for convenience of better data manip-
ulation.
6-8-2 Advantage of SIN Ratio over Average

To analyze the results of experiments involving multiple runs, use of the
SIN ratio over standard analysis (use average of results) is preferred.
Analysis using the SIN ratio will offer the following two main advantages:
· --,
i
Analysis of Varianc;e (ANQVA) 147
I. It provides a guidance to a selection of the optimum level based on

least variation around the target and also on the average value closest
to the target. ,
2. It offers objective comparison of two sets of experimental data with
respect to variation around the target and the deviation of the average
from the target value.
To examine how the SIN ratio is used inanalY!is, consider the following
two sets of observations which have a target value of 75.
Observation A: 55 58 60 63 65 Mean == 60.2

Deviation of mean from target == (75 - 60.2) == 14.8
Observation B: 50 60 76 90 100 Average == 75.00

Deviation of mean from target == (75 - 75) == 0.0
These two sets of observations may have come from the two distri-
butions shown in Figure 6--3. Observe that the set B has an average value
Target Value
Figure 6-.\ Comparison of two j~-:;tributions.

which equals the target value. but has a wide spread around it. On the
other hand for the set A. the spread around its average is smaller, but
the average itself is quite far from the target. Which one of the two is
better? Based on average \"alue, the product shown by observation B.
appears to be better. Based on consistency, product A is better. How can
one credit A for less variation? How does one compare the distances of
the averages from the target? Surely, comparing the averages is one
method. Use of the SIN ratio offers an objective way to look at the two
characteristics together.
6-8·3 Computation of the SIN Ratio

Consider the first of the two sets of observations shown above. That is
set A: SS 58 60 63 65.
Case I. The nominal is the best.

Using Eq. 6-18, and with the target value of 75,
"e
MSD = «5~1l, 75)2 + (58 - 75)2 + (60 - 75)2 + (63 - 75)2

+ (65 - 75)l)I5
:::: (400 + 289 + 225 + 144 + 100)15
= 1158/5
= 231.6
therefore,
SIN = - 10 Log lO (MSD)

SIN = -10 x Log u) (231.6)
- -23.65
Case 2. The smaller is the better.

Using Eq. 6-17,
MSD = (55 2 + 582 + 6()l + 63 2 + 65 2 )/5

= (3025 + 3364 + 3600 + 3969 + 4425)15
- 18183/5
:".~ 16 -:'6 t)
Analysis of Variance (ANOVAJ 149
and
SIN = - 10 Log (3636.6)

- -35.607
Case 3. The bigger the better..

Using Eq. 6-19,
MSD = (1155 2 + 11582 + 1I6()2 + 11632 + 11652)/5

= (113025 + 113364 + 1/3600 + 113969 + 114425)/5
= (3.305 + 2.972 + 2.777 + 2.519 + 2.366) x lQ-"415
= (13.939) x IQ-"4/5
- .0002728
therefore,
SIN.. = - 10 Log (MSD) = - 10 Log (.0002787) = 35.548
The three SIN ratios computed for the data sets A and B under the three
different quality characteristics are shown in Table 6-24. The tJtreecol-
umns N, Sand B, under the heading "SIN ratios" are for nominal, smaller,
and the bigger the better characteristics, respectively.
Now select the best data set on the basis of minimum variation. By
definition lower deviation is indicated by a higher value of the SIN ratio
(regardless of the characteristics of quality). If the nominal the better
characteristic applies, then using column N, the SIN ratio for A is - 23.65
and is - 25.32 for B. Since - 23.65 is greater than - 25.32, set A has
less variation than st B, although set B has an average value equal to the
desired target value.
Table 6-24. SIN Ratios for Three Quality Characteristics

SiSRATIOS
OBSERVATIOSS AVERAGE 5 B
A: S5 58 60 63 65 - :3.65 - 35.6\ 35.5.1

3 ~() -ye -~ :'le .00
Similarly, set A is selected for the smaller is better characteristic and

B is selected for the bigger is better characteristic.
6-8-4 Effect of the SIN Ratio on the Analysis

Use of the SIN ratio of the results, instead of the average values, introduces
some minor changes in the analysis.
• Degrees of freedom of the entire experiment is reduced.

OOFwith SIN ratio = number of trial conditions - 1
(i.e., Number of repetitions is reduced to ".)
Recall that the OOF in the case of the standard analysis is
OOF = (Number of trial x Number of repetition) - 1
The SIN ratio calculation is based on data from all observations of
a trial condition. The set of SIN ratios can then be considered as trial
results without repetitions. Hence the DOF, in case of SIN is the
number" trials - 1. The rest of the analysis follows the standard
procedure.
• SIN must be converted back to meaningful terms. When the SIN ratio
is used, the results of the analysis, such as estimated performance
from the main effects or confidence interval are expressed in terms
of SIN. To express the analysis in terms of the experimental result,
\her \he
must be converted back l(l original units (If measurement.
To see the specific differences in the analysis using the SIN ratio, let
us compare the two analyses of the same observations for the Cam-Lifter
Noise Study shown in Table 6-25-1 (standard analysis) and in Table
6-25-2 (SIN ratio analysis). In this study the three factors (spring rate,
cam profile, and weight of the push rod) each at two levels., were inves-
tigated. The £4 OA definea the {pllr trial conditions: At e~ch of the four
trial con~tions, three observatio~ (in some noise~ate of 0 to 60) were
I
t
~
I
recorded. The results were then analyzed both ways as shown in these
two tables. "
A subtable "results" of the standard analysis (Table 6-25-1) presents
the average of the three repetitions for each trial run at the extreme right
1 hand column. The averages are used in calculating the main effects. The
~:
,
values shown in the subtabte titled "Main Effects" have the same units
as the original observations" Similarly, ~he e'(gec.ted "/alue It r.he )~)(lmUm
Table 6-25-1. Cam-Litter Noise Study Standard Analy.
COLt..'MN FACTORS lEVEL J l.E\U: LEVEL 3 LEVEL 4
I Spring rate Current Proposed

2 Cam profile Type 1 Type 2
3 Wt. of push rod Lighter He<nia:
Individual Results and Their Average
TIUALI
REPE11110NS R, Rz R] AVERAOE
1 23.00 30.00 31.00 30.00

2 35.00 40.00 45.00 40.00
3 50.00 30.00 40.00 40.00
4 45.00 48.00 51.00 48.50 -
The Main Effects
COLUMN FACI'ORS LEYEL J LEV£l : L;: - L,I l.EVELS l.EVEL 4
1 Spring rate 35.00 44.00 9.00 00.00 00.00

2 Cam profile 35.00 44.00 9.00 00.00 00.00
3 Wt. of push rod 39.00 40.00 1.00 oo.m 00.00
ANOVA Table
COLUMN FACTORS DOF SUM OF SQC:ARES vAAIAHCE F
I Spring rate 1 243.00 243:00 5.93 23.63

2 Cam profile 1 243.00 243.00 '5.93 23.63
3 Wt. of push rod (I) (3.00) POOLED
All others/error 9 369.00 41.00 52.16
Total: 11 855.00 100.00
Estimate of the Optimum Condition of Design!Process:

For Smaller the Better Characteristic
FACTOR DESCRIPTION LEVEL DESCRIPTIOS LEVEL (DfI"RIBUTION
Spring rate Current -4.5000

Cam profile Type 1 -45.000
Contriburion from all factors (total) -9.00
Current grand average of performance
E\pe:::ted result at optimum ~('Indition
_--_
......- .._.... ...._----._.•._---_._._._-........... _- 39.50
30.50
151 A I\imer on the Taguehi MetOOd
Table 6-25-2. Cam-Lifter Noise Study-SIN Analysis

Design factors and Their Levels
COLUMN FACTORS LEVEL I LEVEL 2 LEVEL) LEVEL 4
I Spring rate Current Proposed

2 Cam profile Type I Type 2
3 Wt. of push rod Lighter Heavier
Results . . the SIN Ratios

TIUAIJ
RE.PEt1TICR RI R2 Rl Rs SIN
I 23.00 30.00 37.00 -29.70

2 35.00 40.00 45.00 -32.09
3 50.00 30.00 40.00 -32.22
4 45.00 48.00 51.00 -33.64
The Maim Effects

COLUMN FACTORS LEVEL I LEVEL 2 (Ll - L.) LEVEL 3 LEVEL
1 Spring rate -30.9 -32.93 -2.04 00.00 00.00

2 ., cam profile -30.96 -32.86 -1.91 00.00 00.00
3 Wt. of push rod -31.67 -32.16 -0.49 00.00 00.00
ANOVATable
COLUMN FACTORS ooF SUM OF SQUARES VARIANCE F PERCENT
1 Spring rate 1 4.141 4.141 17.61 48.79

2 Cam profile 1 3.629 3.629 15.43 42.39
3 Wt. of push rod .(1) (3.00) Pooled
All othe«ettor I 0.24 0.24 8.82
Total: 3 8.01 100.00
Estimate of the Optimum Condition of DesignlProcess:

For Smder the Better Characteristic
FACTOR. mESOUPIlON LEVEL DESCRlPTION LEVEL CONTRIBlmON
Spring Ate Current 1 .1.0175

Cam pmile Type 1 1 0.9525
Contri_on from all factors (total) 1.%99
Current grand average of performance -31.9125
Expected result at optimum condition - 29.9425
---_.__ ._._----------
Analysis of Variarn:e (ANOVA) 153
condition, 30.5, has the same units as the original recorded data: The
degrees of freedom for the experiment (DOF column in ANOVA table)
is Il (4 x 3 - 1).
Comparing the standard analysis with the analysis using the SIN ratio
(Table 6-25-2), note that the average value of the results is replaced by
the SIN ratio. The SIN ratios are then used to compute the main effects
as well as the estimated performance at the optimum condition. Notice
also that the degrees of freedom for the experiment is 3. This difference
in OOF produces a big difference in the way the two analyses compute
ANOV A, i.e., the percentage contribution of the factors involved (for
spring rate the value is 23.6% from standard analysis as compared to
48.79% from SIN analysis). Likewise, the other factors will have different
magnitudes of contribution in the two methods.
In estimating the result at the optimum condition, only the factors that
will have significant contributions are included. In this case, both methods
selected level 1 of factors in columns 1 (spring rate) and 2 (cam profile).
This may not always be true.
When the SIN ratio is used, the estimated result can be converted back
to the scale of units of the original observations. For example, the expected
result in terms of SIN ratio' is -29.942S <Table 6-25-2, bottom line).
This is equivalent to an average performance, y', which is calculated as
.
follows.
Since
SIN == - 10 Log (MSD)
and
MSD = (Yi +. . . Y~)IN (For "smaller is better")

-- 2
= y expected
Therefore,
MSD = 10( -S/N)/IO == 10-( - 29.9425)110 - 986.8474
or
ye.. . pected :: (MSD)12 = (986.8474)12 = 31.41
\\ hich is con1Darable
, to 30.5 shown at the bottom of Table 6-25- i .
Table 6-26. Orthogonal Array and Test Data
TRIAL .4 B AXB C D
COLUMN I 2 J 4 S 6 7 RI Rz R3
Trial 1 1 1 1 1 I 0 0 45.00 56.00 64.00

Trial 2 1 1 I 2 2 0 0 34.00 45.00 53.00
Trial 3 I 2 2 1 I 0 0 67.00 65.00 60.00
Trial 4 I 2 2 2 2 0 0 45.00 56.00 64.00
Trial 5 2 I 2 I 2 0 0 87.00 81.00 69.00
Trial 6 2 1 2 2 1 0 0 78.00 73.00 68.00
Trial 1 2 2 1 1 2 0 0 45.00 56.00 52.00
Trial 8 2 2 I 2 1 0 0 42.00 54.00 47.00
6-8-5 When to Use the SIN Ratio for Analysis

Whenever an experiment involves repeated observations at each of the
trial eondition~.the SIN ratio has been found to provide a practical way
to ~easure an'~ontrol tbe combined in~~ of deviation of the pop-
ulation mean fljm tbe target and the vartafion around the mean. lnstan-
<lard analysis,Dle mean and the variation around the mean are treated
separately by a main effect study and ANOVA respectively. The analysis
of the Tagucbi experiments using SIN ratios for the observed results, can
be conveniently performed by using computer software such as the one
in Ref. 11.
Table 6..27. ANOVA
COLUMN FACTORS f S V F P
I Speed I 3.036 3.036 4.432 9.56

2 Oil viscosity (I) (1.91) Pooled
Interaction 1 15.820 15.820 23.093 61.57
4 Clearance 1 2.987 2.987 4.361 9.36
5 Pin straightness (1) (0.75) Pooled
All others/error 4 2.740 0.685 19.51
Total: 7 24.584 100.00%
Note: Insignificant factorial effects are pooled as shown ( ).

EXERCISES
6-1. In an experiment involving four factors (A. B. C. and D) and one interaction (A
x B). each trial condition as repeated three times and the observations recorded as
shown in Table 6-26. Detennine the total sum of squares and the sum of squares
for factor A.
6-2. Assuming "the bigger the better" quality characteristic. tnmsfonn the results to the
corresponding SIN ratios. (ADs. :t 0.962)
6-3. Table 6-27 shows the product of ANOVA performed on the observed results of an
experiment. Determine the following from the ANOVA table.
a.Pem:nt influence of the clearance factor.
b.Degrees of freedom of the speed factor.
c. Error degrees of freedom.
d. Intluence of noise factors and all other factors not included in the experiment.
e. Confidence interval (90%) of the perfonnanceat the optimum condition (use F
Table for 90% confidence level).
7
Loss Function
7-1 DERIVATION OF LOSS FUNCTION

In Chapter 2 Tagucm's philosophy regarding the cost of quality was
stressed. He believes that a poorly designed product causes society to
incur losses from the initial design stage through to product usage. There-
fore, he emphasizes good quality at the conceptual stage of a product
and onward, by optimizing the product design parameters as well as the
producti08 Qlftditions, to create a robust product. A question commonly
asked is how much effort should an organization expend on quality. How
can the point of diminishing returns be determined. In this section the
basic madtematical fonnulation of Taguchi's loss function is developed
and an oudine of the steps used to apply the loss function is presented.
The loss fimction has proven to be an excellent tool for determining the
magnitude of the process (manufacturer) and supplier tolerances, based
upon quatilyas perceived by the customer. The methodology for realign-
ment of tolerances is beyond the scope of this text.
Tagucbi defined the loss function as deviation as a quantity proportional
to the deviation from the target quality characteristic. At zero deviation,
the performance is on target, the loss is zero. If Y represents the deviation
from the target value, then the loss function L(Y) is:
L(¥) = key - YO)2 (7-1)
Where
Y = The quality characteristics~ as a dimension,

perfonnance, etc. .
Yo = The target value for the quality characteristic.
k = A constant which is dependent upon the cost structure of a
manufacturing process or an organization.
156
Loss Function lS7
It is important to note that:
1. The term (Y - f o) represents the deviation of the quality charac-

teristic f from the target value f o.
2. The equation for the loss function is of the second order in terms
of deviation of the quality characteristic.
The loss function represented by Eq. 7-1 is graphicallysbown in Figure

7-1; it possesses the following characteristics:
1. The loss must be zero when the quality characteristic of a product

meets its target value.
2. The magnitude of the loss increases rapidly as the quality charac-
teristic deviates from the target value.
3. The loss function must be a continuous (second order) function of
the deviation from the target value.
Target Va':..e
15 t---~r---------+---------f-.---I
le
..!! 10
0
0
.E
~
.s
5
-4 -2 o 2 4 6
y {Quality C"'a':::~risticsl
Figure 7-1. The Taguchi qua::~ loss function.

Taguchi determined the loss function from a Taylor Theorem expansion

about the target value Yo. Thus.
L(Y) = L(Yo) + L' (Yo)( Y - Yo) + J! L"(Yo)(Y - YO)2 (7-2)
where the terms with powers of (Y - Yo) higher than 2 are ignored as
being too small for consideration.
In Hq. 7-2, L(Y) is the minimum at Y := Yo hence its first derivative
L'(Yo) is zero. Therefore, Eq. 7-2 can be written as
L(Y) := L(Yo) + ~ L"(Yo)(Y - Yo)l (1-3)
The expression (lJ2)L"(Yo) in Eq. 7-3, is a constant and can be replaced

by a constant k. Also if Yo is the mean value of the product/process, then
Eq. 7-3 is interpreted as the loss about the product/process mean plus the
loss due to displacement of the process mean from the target. If the
process mean collcides with the target, the loss term L(Yo) is zero and
the loss function reduces to Eq. 7-1,
L(Y) = k(Y - Yo)2
The magnitude of the loss incurred because of the inability of a process

to meet the target value of a quality characteristic, as computed using
Eq. 7-1, is dependent on the target value, the manufacturing process, the
cost of rework, scrap, and warranty. For a given value of f o, the value
of k varies with the process and the organization. An organization seri-
ously committed to achieving higher standards of quality in an optimum
manner may develop families of loss curves for each process. One family
of curves with different values of k is shown in Figure 7-2. The value
of k for any application can be determined as outlined below.
Any mass produced product exhibits variation in its quality character-
istic. As long as the variation is small, the quality of the product is
acceptable to the customer. Customer acceptance determines the range
of variation. If the perceived quality falls outside the range, the customer
will not accept the product, and corrective actions at the design or process
level are required.
Loss Function 159
Target Value
20 t--~:If---------t-~~----p.-~
15
...
~
-0
0
.=
S
..J
10
51---------\---\+--+---++--1---------1
oL-_...L_-l.
-6 -4 -2
-=-~=--...l--....J.
o 2 4
__
-....J
6
Y (Quality Characteristics)
Figure 7-2. Family of loss functions.
Let this tolerance zone be + Ii, then the quality characteristic at the
extremes can be denoted as:
Yo + Ii and Yo - tJ.
Poor quality exceeding the extremes, necessitates corrective action such

as a warranty cost -of Lo.
Let Lo be the loss at Y = Yo + tJ.

Then from Eq. 7-1:
or
160 A ftimer on the Taguchi Method
Therefore:t Eq. 7-1 for this case becomes:
The above equation now completely defines the loss function in terms of
the deviation from the target value.
7-2 AVERAGE LOSS FUNCTION FOR PRODUCT

POPULATION
The lossfunction given in Eg. 7-1 represents the financial loss experienced
by a single product when the quality characteristic Y of the product
deviates from the target value Yo. In a mass production process. the
average loss per unit is expressed by:
L(Y)= [k(Y 1 - YO)2 + k(Y2 - Yo)2

+ k(Y3 - f O)2 + ... + le( Yn - Yo)2]/n
where 11 is the nttmber of units in a given sample.

In the above equation the factor k is common with every term. and
therefore~ it can also be written as: 1
L(Y) = k[(Y. - fo)l + (Y2 - YO)2
+ (Y3 - f ol 2 + ... + (Yn - Yo)2]/n
Note that the expression within the brackets is the mean squared deviation
(MSD)" the average of the squares of all deviations from the target value
f o. The average loss per unit can now simply be expressed by:
L( Y) = k(MSD) (7-4)
7-3 APPLICATION OF LOSS FUNCTION CONCEPTS

The loss function concept has two practical applications. The primary
application is for estimating the potential cost savings resulting from the
improvements achieved by optimizing a product or a process design. The
lO'iS Function 161
loss function can serve regardless of the method of the quality improve-
ment. As long as variation is reduced by corrective actions or design
improvements, the loss function presents a means for estimating the
savings in terms of dollars and cents. The second application is to de-
teonine manufacturer and supplier tolerances based on the customer's
perception of the quality range. In this case, the loss function provides
an objective way to set the limits for the inspection of products at man-
ufacturer or supplier location. The following examples illustrate the use
of tbe loss function. ;.
Example '-I
Machine Bracket Casting Process (~ost SavingS)
Engineers involved in casting a machine bracket designed an experi-
mental study to improve the process and reduce scrap rate. As a result
of the study, a number of improvements were incorporated. Data were
taken from 10 samples before and after the experiments. The foundry
had a production rate of 1500 castings per month. The .quality inspection
criterion was a length dimension of 12 + 0.35 inches. The parts which
did not fall within the limits were rejected. The average unit cost for
scrap or rework of the rejects was $20. The potential cost savings of the
optimized process was calculated by the Taguchi loss function.
Test Data before the Experiment:
11.80 12.30 12.20 12.4 12.1 12.2 11.9 11.8 11.85 12.15
Readings After the Experiment:
11.9 12.2 12.1 12.2 12.1 12.1 11.9 11.95 11.95 12.1
Other Data:
Target value - 12.00 in.

Tolerance - +0.35 in.
Cost of rejection = 520.00
Production rate = 1500 aer month.
Solution
For this application the expression of loss in tenns of the MSD will be
used.
for a single sample
and
L(Y) = k(MSD) for multiple samples.
Using Eg. 7~ 1, the constant k is determined as follows:
L = k(Y - YoY
i.e., L = k(Yo + Tolerance - Yo)2
But tbe loss L is equal to the cost of rejecting a part (520.00) and the-
tolerance is ..35
or 20 = k (.3t)2
~~~;
or k = 20/(.33)2 = 163.265
Therefore, from Eq. 7~4
L = 163.265 (MSD) (7~5)
Using the data from samples before the experiment,
MSD = [(11.8 ~ 12)2 + (12.3 - 12)2 + ... ]110.

= 0.0475
The MSD and other statistical parameters for this example, as shown in
Tables 7-1, 7-2 and 7-3, are obtained by using the software in Reference
11. The format of the design descriptions and the results are presented
in the manner displayed by the software.
From Eq. 7-5 the average loss per unit is calculated as
L == t63.265 X .0475::;: i.754 fin dollars)

Loss Ftmetion 163
Table 7-1. Machine Bracket Casting Process

(Before Experiment)
Observation I =11.800
Observation 2 - 12.300
Observation 3 -12.200
Observation 4 -12.400
Observation 5 = 12.100
;. Observation 6 = 12.200
Target/nomina) value of result = 12.00

Number of test results (NR) =: 10

Total of all test results =: 120.70000
Average of test results = 12.07000
Standard deviation (SO) =: 00.21756
Variance =: 00.04733
LOSS FUNCTION PARAMETERS:

Mean square de\"iation (MSD) =: 00.04749
Signal to noise ratio (SIN) =: 13.23301
Variance (modified form) = 00~04259
VARIANCE DATA (ANOVA):

Target value of dataltest result = 12.00
Mean of data/deviation from target = 00.069999
Total variance (ST) = 00.04899
(ST = Variance x NR)
Correction factor (CF) = 00.04B99
(CF = (average of data)2 x number of data)
Sums of squaresfN =00.47499
164 A ftimer OD the Tagucbi Method
Table 7-2. Machine Bracket Casting Process

(After Experiment)
Observation I = 11.900
Target/nominal value of result = 12.00

Number of test results (NR) 10

Total of an test results == 120.50000
Avemgeeltest results == 12.0s000
Standard.Viation (SO) == 00.11547
Variance,t;1t == 00.01333
LOSS dmON PARAMETERS:
Mea square deviation (MSO) == 00.01450
Signal to noise ratio (SIN) == 18.38631
Variance (modified fonn) - 00.01200
VARIANCE OATA (ANOVA):

Target ~ue of data/test result = 12.00
Mean of data/deviation from target = 00.05000
Total variance (ST) - 00.12000
(ST= Variance x NR)
Correction factor (<;F) - 00.02500

(CF = (average of data)2 x number of data)
Sums of squareslN - 00.14500

Loss Function 165
Table '-3. Calculation of Loss
PROBLEM DEFINITION
Target value of quality characteristic (m) = 12.00
Tolerance of quality characteristic = 0.35
Cost of rejection at production (per unit) = $20.00
Units produced per month (total) = 1500
SIN ratio of current design/part = 13.23307
SIN ratio of new design/part = 18.38631
;..
COMPUTATION OF LOSS USING TAGUCHI LOSS FUNCTION

Loss function: L(y) = 163.26 x (MSD) Also L(y) = Kx (y - m)2
BEFORE EXPERIMENT:
Losslunit due to deviation from target in CUtTent design = 57.754
AFTER EXPERIMENT:
Loss/unit due to deviation from target will be reduced from
57.754 to = 52.367
MONTHLY SAVINGS:
If production were maintained at the improved condition, then
based on 1500 units/month = 58081.34
Using the data from samples after the experiment,
MSD = [(11.9 - 12)2 + (12.2 - 12)2 + ... ]110

== 0.0145
From Eq. 7-6, the average loss per unit is calculated to be
L - 163.265 x .0145 = 2.367 (in dollars)
The average savings per unit is calculated by subtracting the loss after
the experiment ($2.367) from that before the experiment ($7.754). The
total savings is then obtained by multiplying the average savings by the
production rate as shown here.
Total saving per month = (7.754 - 2.367) x 1500

= $8080.50
.66 A Primer on the Tagucbi MeIhod
Example 7-2
Dryer Motor Belt (Manufacturer/supplier tolerance)
Alarmed by a high rate of warranty repairs of drive-belts for one of
its products. the distributor sought to reduce such defects. The field reports
suggested that the problem was mainly caused by the lack of adjustment
of tension in the drive belt. To correct the situation at the customer?s
location. the field repairmen had to adjust the tension to 100+ 15 lbs.
Field service cost is $40 per unit. Alternately, the adjustment of tensioll
could be made by the manufacturer at a unit cost of $15. The distributor
wants to ask the manufacturer to make such adjustments prior to shipment
in order to eliminate the field service and maintain satisfied customers.
What range of tolerance in belt tension should the distributor specify for
the manufacturer?
Solutio.
For this application, an understanding of the role of the three parties?
namely custa.r, manufacturer, and supplier will be helpfuL In the
context of tolefance specification, the three terms correspond to three
stages of product life. The supplier is the one who supplies a component
or part of the finished product to the manufacturer. The manufacturer is
the one who assembles the final product. The customer is the one who
uses the product and experiences its performance. In this example, the
distributor and the customers are considered to be the end users. The
customer and the manufacturer may have a supplier (not identified) for
the motor and belt assembly. The relationships among the three can be
represented in the following way.
Supplier ~ Manufacturer ~ Customer

(Belt + Motor) (Washer) (Washer in use)
Tolerance required:
(unknown) (unknown) + 15 Ibs.)
From Eq. 7-1 we have
L(Y) ;::: k(Y - YO)2 = k (Tolerance)2

Loss Function 167
where
Tolerance = Y(max. or min.) - Yo
Based on a repair cost of $40 at the customer's installation, the loss per
unit is $40.
Since
Yo = 100, Y = 100 + 15 (max.), and L(F} = 40

k = 40/(100 + 15 -- 1(0)2 = 0.17778
Therefore,
L(Y) = .17778 (Tolerance)2 (7-6)
Using the repair cost of $15 at the manufacturers facility as the loss, the
tolerance now can be determined by using the abOYe relation.
Tolerance = (LI.17778)1I2 = 9.18

(Manufacturer)
or
Tolerance Limits = 100 + 9.8
If the manufacturer determined the problem lobe caused by a specific

component, then correction at the supplier's facilities would be appro-
priate. A different set of specification may be retpJired. Assuming the
cost to make an adjustment at the supplier is $S per unit. What tolerance
should the manufacturer require of the supplier 10 assure the required
quality?
Using Eq. 7-4 with L(Y) = 5
5 - .178 (Tolerancef
and
Tolerance = (5/.17778)12 = 5.30

168 A Pmaer OD the Tapthi Method
2.0 ~--------'----r---------- ...
1.5
>
it 1.0
iL
0.5
...--$
I-----M------I
.......- - - - - C - - - - - - - - . (
"'"'$' = Supplier tol.ra~ 100 :t5.8

M ::::: Manufacturer tolerance 100 ± 9.18
C= Customer tolerance tOOt t 5 '
COMPUTATION Of MANUFACTURER SSUPPUER TOLERANCE

Nominal value of the quality characteristic Y ::::: 100 lbs.
Tolerance of Y (range of deviation) ::::: ± 15 lbs.
Cost to repair a nonfurlctioningunit by customer ::::: $40.0
Cost .. repair a Il()nfunetioning unit by manufacturer ::::: S15.0
Cost to repair a nonfunctioningunit by supplier ::::: $ 5.0
REQUIRED TOLERANCES
Manufacturer tolerance ::::: 100 ±9.18
Supplier tolerance ::::: 100 ±5.3
NOTE: If these Tolerances were held. there will be Il() Il()ofunetional part in the customers
hands. For the same cost the Manufacturer will maintain satisfied customers and quality
t
I
procb:ts in the field.
Figure 7-3. Manufacturer and supplier tolerances.
i
Loss ftmction 169
Therefore
Tolerance limits for supplier == 100 + 5.3
The solutions are presented in Figure 7-3.
Example 7·3
Fuel Pump Noise Study
In an experimental study of an automotive fuel pump noise, three 2
level factors were included as shown in Table 74-1. The Taguchi L4
orthogonal array was used to define the four trial conditions. Six samples
at each of the trial conditions were tested and the results were recorded
as shown in Table 74-2. The levels were selected so that the trial con-
dition I represents the current design of the fuel pump. If the decision
is made to change the design to the determined optimum configuration,
estimate the performance at the optimum design and the cost savings
when the new fuel pump is produced.
Table 7-4-1. Fuel Pump Noise Study-Example 7·3

Fuel Pump Noise Study
Design Factors and Their Levels
COLUMNS FACTORS LEVEL I LEVEL 2 LEVEl. 3 LEVEL 4
I Seal thickness Present Thicker

2 Rotor chuck type Present New design
3 Finger to drive Cl Present Increase
Note: Thn:e 2 level factors studied.
Objective: Design least noisy and best perfonning pump.
Characteristic: Nominal the best {SIQ = 60 Target).
This experiment will use ~
TRlAIJCOLUMN 2 3
Trial I I I 1
Trial 2 1 2 2
Trial 3 2 I 2
'1
Trial 4 2 1
170 A Primer OD the Taguc::hi Method
Table 7-4-2. Fuel Pump Noise Study (Result: Main EtTect and .ANOVA)
Original Observations and Their SIN Ratios
Quality Characteristic: Nominal is Best
TRIAL
REPETITlOSS RI Rl Rl R.4 Rs Il(, siN
1 61.00 85.00 87.00 65.00 59.00 16.00 -20.71

2 65.00 65.00 66.00 54.00 13.00 58.00 -18.99
3 54.00 45.00 56.00 45.00 63.00 46.00 -25.89
4 56.00 67.00 45.00 54.00 56.00 74.00 -23.36
The Main Effects .
COLUMN FACTOR NAME LEVEL I L.EVEL t (L. - Lit LEVEL 3 LEVEL 4
1 Seal thickness -19.85 -24.63 -4.18 00.00 00.00

2 Rotor chuck type -23.30 -21.18 2.12 00.00 00.00
3 Finger to drive -22.04 -22.44 -OAI 00.00 00.00
ANOVA Tabt'
COLUMN FACTOR OOF SUM OF SQUARES VARIANCE F PERCENT
1 Seal thickness 1 22.801 22.801 82.97

2 Rotor chuck type 1 4.512 4.512 16.43
3 Finger to drive 1 0.164 0.164 00.60
All others/error o
Total: 3 27.480 100.00
Assume that the following information is used.

,
~
Target value of quality characteristic = 70

Tolerance value of quality characteristic = + 20
Cost of replacement/rejection - $45.00

Production Rate - 20.000
I
I
Loss Fuaction 171
II
t Table 7-4-3. Fuel Pump Noise Study (Optimum and Cost Savings) I
l
Estimate of the Optimum Condition of Design/Process:
for Nominal the Best Olaraeteristic
Seal thickness Present design 1 2.3875

Rotor chuck: type New design 2 1.0625
Finger to driveclearanee Present design 1 0.2025
Contribution from all factors (total) 3.6524
Current gnmd average of perfonnance -22.2375
Expected result at optimum condition -18.5850
This estimate includes only" variables that have a silJ1ificant contribution. i.e.• pooled variabk$ are excluded
from the estimate. Estimates may also be made with variables of choice _.
1 CALCULATION OF LOSS
PROBLEM DEFINITION
Target va)u<: of quality characteristics (m) = 70..00
Tolerance of quality characteristic : 20.00
Cost of !ejection at production (per unit) ~ $45..0 0
Units produced per month (total) = 26000
SIN ratio of current design/part = -20.71
SIN ratio of new design/part = -18.585
COMPUTATION OF LOSS USING TAGUCHI LOSS FUNCTION

Loss function: L(y) = 0.11 x (MSD) Also L(y) = K x (y - m)2
BEFORE EXPERIMENT:
Loss/unit due to deviation from target in current design = S12.953
AFTER EXPERIMENT:
Loss/unit due to deviation from target will be reduced from S12.953
to = S7.941
MONTHLY SAVINGS:
If production were maintained at the improved condition, then
based on 20000 units/month = SI00246.9O
Solution
The .complete solution of this problem is shown in Tables 7-4-2 and
7-4-3. In calculating the cost savings, notice that the SIN ratios at the
trial condition 1 and the optimum condition, are taken directly from the
analysis of die Taguchi experimental results. (Solutions used software of
Ref. 11.)
EXERCISES
7-1. The mamlfacturer of a 10.5 volt smoke aIann battery employed the TaguclU method
to detelmine the better design parameters. The experimenters estimated the signal
to noise 'SIN) ratio for the proposed design to be 6.3. Based on a sample inspection
of the amentproduction process. the SIN ratio was calculated to be 5.2. The analysis
of wammty showed that when the battery voltage was beyond (to.50 z: 0.75) volts,
the smob alarm malfunctioned and customers returned the batteries for $6.50 each.
Determi.ae the monthly savings the proposed new design is expected to generate if
20.000 IDits are .manufaetured each month.
7-2. Suppo.sctbat the manufacturer in Problem 7-1 decides not to adopt the new design,
but ~ to screen all defective batteries in his manufacturing plant before they
are ~ to the customers. The cost for inspection in the plant is estimated ~ be
$3 per httery. If the same amount of warranty cost is incurred in the inspection
~ determine the tolerance limits for the inspection.
I
8
Brainstorming-An Integral Part of the
Taguchi Philosophy
1-1 THE NECESSITY OF BRAINSTORMING

Taguchiregards brainstonningas' an essential step· in the design of ef-
fective experiments. Taguchi uses brainstormfng to cross organizational
barriers. By including representatives of all departments, from design
Ibrough marketing, the quality demanded by the customer can be con-
sidered and those production factors which may contribute towards quality
can be identified and incorporated into the design of the experiments.
The benefits of brainstorming are obvious. The design does not belong
to anyone group, it belongs to all. Brainstorming identifies characteristic
effects and the environment known to the group as a whole. Measurement
techniques can draw on many disciplines. The OUteotne is better than
II would be the case if one department assumed all responsibilities.
Taguchi does not prescribe a standard method of brainstotming as
I applicable to all situations. The nature and content of the brainstorming
i session win vary widely on the problem. For most studies a formal
1 brainstorming session is highly effective. In many instances brainstorming
can be cOJllpleted in .a few hours. In s~meinstances the session may take
a good part of the day.
One method of conducting brainstorming, or what.is normally referred
to as a workshop for the Taguchi experimental design, is presented in
the text. The procedure has been found to be effective by the author. in
his experience with various client industries. The procedure is not standard
nor is it Taguchi's; it is the author's.
8-2 THE NATURE OF THE SESSION

The follO\\"ing guidelines determine the participants and the content of
the brainstonning session:
I
. Purpose of the BrainstormiBg Session
1. Identify factors, levels, and other pertinent information about the

experiment, collectively with all departments and personnel con-
cerned with a successful output from the experiment.
2. Develop team spirit and attitude to assure maximum participation
of the team members.
3. Develop a consensus on the selection and the'determination of those
items which are objective and those whicb are subjective in nature.
Team Leader
For the successful completion of a Taguchi case study, the appointment
of a team leader, from among the project team members, is necessary.
The team leader must recognize the need for a brainstorming session and
call for such asession. The leader should try to hold the session on neutral
ground. The leader should ensure the participation of all departments with
responsibilities fQ~ the productJprocess.
'l.
~'.:
Session Adviso~
The session sbqj1d have in attendance someone with· a good working
~&, ,
knowledge of the Tagucbi methodologies. Engineers or statisticians ded-

icated to helping others apply this tool often make better facilitators. They
should offer advice and counsel but not lead the discussion.
Who Sbould Attend?

All those who havefirsthand knowledge andlfJr a stake in the outcome
should be included. For an engineering design or a manufacturing process,
both the design and the manufacturing personnel should attend. If cost
or supplier latowledge are likely factors, then persons with experience in
these matters should be encouraged to attend (group size permitting).
Marketing personnel should attend to provide customer viewpoints.
How Many Should Attend?

The more the better. However. the time involved is proportional to the
number attending. The upper limit should be 15. It can be as low as 2.
No matter the number of people involved, brainstorming will immensely
benefit the whole process. A session with 15 people may take the whole
day. For 'Jp to g peor1e, half:1 jay :nay suffice. More important than
Brainstorming-An Integral Part of the Taguchi Philosophy 175
size is proper representation from all departments involved in design..

development, production, marketing, sales, and service.
What is the Agenda for the Session?

Any brainstorming session. leading to a Taguchi experimental design,
may use the topics presented in Section 8-3 for discussions. The emphasis
and lenth of the discussion, however, will differ significantly for differeat
;.. problems.
Is Tagucbi Training a Prerequisite?

No. Some exposure will help. Application experience On the part of some
participants will be a plus. A facilitator with application experience can
help the participants with brief overviews_when needed.
8-3 TOPICS OF THE DISCUSSIONS

The following topics should be included in the agenda for the brainstorm-
ing session.
1. Objective of the Study
A. Wbat is the characteristic of quality? How can the characteristic

be evaluated objectively?
B. How can the quality characteristic be measured? What are the units .
of measurement?
C. What are the criteria (attributes) for evaluation of the quality char-
acteristic?
D. When there are many criteria or several attributes of the quality
characteristic, how can they be combined into a single Overall
Evaluation Criterion (OEC)?
E. How are the different quality criteria weighted?
F. What is the sense of the quality characteristic? Lower is better,
nominal is the best. etc.
2. Design FactorslVariables and Their Levels
A. \\'hat are all the possible factors?

B. \Vhich ones are more important than others?
C. How many factors should be included in the study:
176 A Primer on the Tapchi Method
D. What level should be selected for the factors? How many levels?
E. What is the trade-off between levels and factors?
" F. How urgent are the results? Should the design be aimed at a fast
response with only a few factors, or is there time available to
investigate more factors?
3. "Noise" Factors
A. What factors are likely to influence the objective or outcome but

can· not be coatrolled in the experiment.
B. How can the product under study be made insensitive to these noise
factors, i.e., how can a robust solution be obtained?
C. How are these factors included in the study?
4. Interaction Stu6s
A. Which are the factors most likely to interact?

B. How many' ractions can be included?
C. Should an . . . tion be replaced by an additional factor?
D. Do we n~to study the interactions at all?
j"
5. Task ~ and Description
A. What steps are to be followed in combining all the quality criteria

into an OEe?
B. What about the factors not included in the study?
C. How well does the experiment simulate the customer/field appli-
cations?
D. How many repetitions will be necessary, and in what order will
the experimeDts be run?
E. Who will do what· and when? Who will analyze the data?
A brief outline of topics for discussion and their order is shown in Figure
8-1.
8-4 TYPICAL DISCUSSIONS IN THE SESSION
As indicated earlier the guidelines apply to most of the Taguc4i application

workshops but the specifics drawn will be unique to each situation. To
demonstrate a brainstorming session. ;;onsider the ~ake baking ;:>rocess
Brainstonning-An Integral Pan of the Taguchi Philosophy 177
Brainstorming
for Design of Experiment
• Assign Project Title and Defirnl Objectives

• Assign Personnel from Function Organization
• Identify QualitY Characteristics
• Determine How Each Attribute Is Measured
Determirnl:
;. • Control Factors
• Noise Factors
• Factor levels
Scopes of Projaet
• How Many Experimenu
• How Many Repetitions
I
I Assign Tasks:
• Who Does What
I Figure 8-1. Agenda for a brainstonning session.

II
Suppose there are five participants at a meeting called specifically to
discuss and plan a Taguchiexperiment to determine the best recipe for
the pound cake. The first question that comes up is, what are we after?
Ofcourse, everyone will agree thatthey are after the best cake. Obviously,
the taste ofthe cake will be a criterion. But how can the taste be measured?
How many will taste it? For a subjective evaluation like taste, evaluation
by more than one person is desired. If more than one individual is in-
volved, how can the net result be .evaluated? A possible solution is to
rate the cakes in terms of a numerical scale, say, on a scale of 0 to 10.
Is 10 "tor the best taste? This needs to be defined. It can be anything
agreed upon. But it is important that th~ maner be discussed, this rule as
will dictate the sense of the quality characteristic. If 10 represents the
best and 0 the worst, then the quality characteristic becomes "bigger is
better." If five people were to taste the .experimental cakes and the ratings
vary, whom do we believe? All of them. Record the average of all five
evaluations.
Is taste the only parameter to compare two cake samples'? If two cakes
have the same taste can a second parameter distinguish between them?
What about the moistness? One may prefer a moister over a drier cake.
How about shape. appearance. or shelf life'~ Perhaps moistness is some-
""'hat less impcnant than ~he taste, The ide1 of the brainstorming is to
~:l:se lit J'JcS~~,,'i;;;, jt'T<:: ,:c l!~ ~r.c :s:'uC'~. l.. :e:he; ~r.l;Jcr:aJ: :r 1'':-( ~~,::\\
178 A Primer on the Taguchi Mdf10d
Table 8-1. Overall Evaluation Criteria
CRfIBUA VALUE {;NITS WBGH'I1NG MAXIM{.'H VALUE
Taste 5 (YI) none 68 (WI) 10 b'tmaJ

Moistness 4.5 (yz) ounces 32 (w:z) 6(y~
would one measure moistness? By appearance? Perhaps one would decide

to take a fixed slice and weigh it, and note the weight in ounces. This is
an objective measurement. But what is a ustandard~' slice? One would
not necessarily need more than one measurement. Suppose that we agree
to consider just these two criteria, Le., taste and moistness. Taste is
subjective on a scale of 0 to 10; moistness is measured in. ounces. How
sould one make sense out of the two values? How would we compare a
cake rated 7 for taste and 4 ounces for moistness with another rated at 8
for taste and 3.5 ounces for moistness? What index can be devised to
distinguish between the two samples? How can dle data mix of subjective
and objective .ues be combined and analyzed?
Are these criferia to be given equal weight or unequal weight? How
important is mofitness compared to taste? Is the moistness 1/4 as importnat
as the taste? Is anassignmeDt such as taste is 80% and the moistness is
20% reasonably correct? In other words, if one were to split a dollar
among all the criteria, in accordance with the priority, how would one
disttibute the 100 cents? When such a question is asked of different
groups. the responses are never the same. One group may feel taste is
worth 80%, another group may say 60%. As a ge!leral rule, when con-
fronted with subjective areas such as these, let consensus prevail. Let
everyone participate. Let everyone offer their input. For eacb of the items,
determine tbe group average.
Suppose that the consensus of the group has been that the weighting
of the twQ quality criteria should be 68% for taste and 32% for moistness.
With this knowledge one. cad proceed to combine the evaluations and
produce a single quantified number (OEC). The evaluation parameters
for the problem can be summarized as in Table 8-1.
The Overall Evaluation Criterion (GEC) can be defined as:
OEC - (y',IYlmayJ X w! + (Y2JY2max) X W2 +

.- S,~O\ < 68, tOO .- ,;15;61 < 32/tOO
Brainstormin$-An Integral Part of the Tzpchi Phij~y 179
= 0.34 + 0.24
= 0.58
where Wi = Weight of ith component

Yi = Measurement of ith criterion
Yimax = Maximum value of itb criterion
Observe that the evaluation (y) in each case is divided by the maximum
value. This is done to get rid of its units (nonnalization). When multiplied
by the weigbting~ a dimensionless number., the resulting values for each
criterion are added to produce a net result in numerical dimensionless
terms.
Suppose that an La array is used to describe theeigbt trial conditions
I for the experiment. The eight cakes will have to be evaluated following
the scheme given above. The OEe calculated above (OEC= 0.58) will
represent the result for one trial. There will be seven other results like
1 this. The eight values (GECs) will then fOnn the result column in the
OA. The process will have to be repeated if there "-"ere repetitions in
each trial condition.
Discussions of factors and levels follow that of quality characteristics.
The nature of these issues are based on common sense and some under-
standing of the problem under investigation. A leader or facilitator will
often find it convenient to let their experience detemline the ftowof the
discussions. The remaining portion of the brainstorming session is left
up to the reader's imagination.
EXERCISES
8-1. In an experiment involving the study of an automobile dOOl" design. two criteria
were used for evaluation purposes. Deflection at a fixed point in the dQQr was
measured to indicate the stiffness. and the door closing effort was subjectively
recorded on a scale of 0 to 10.
a. Develop a scheme to define an overall evaluation criterion.
b. Explain why the overall evaluation may be u~ful.
8-2. During the brainstorming session for a Taguchi experiment. Cl t'Mge number of factors
were initially identified. Discuss the type of information you need to consider to
determine the number of factors TL1 r the experiment. and state nO\\ you will proceed
:0 select these factors.
8-3. A group of manufacturing engineers identified the following process parameters for
an experimental investigation:
• Founcen 2 level factors (not all are considered important)
• One interaction between two factors (consideted important)
• Three noise factors at t\\O levels each (considered important)
[f the total number of trial roDS (samples) is not to exceed 32. design the experiment
and indicate the sizes of the inner and outer arrays.
R
fj
-:
I,
"
I
~.
!
'!
· ---_
. ....,.;.---
9
Examples of Taguchi Case Studies
9-1 APPLICATION BENCHMARKS

Experiments designed and carried out according to the Taguchi meth-
odology. are generally referred to as case studies. Pedlapsthey are called
case studies to indicate that they are well planned experiments and not
simply a few tests to investigate the effects of varying one or more factors
at a time. The term case study may also be used to signify that such
planned experiments have been fruitfully carried out. that the results have
been analyzed 10 determine the optimum combinati08 of the factors under
study. and that tests to confirm the optimum conditions have been con-
ducted. But what does a case study look like? \Vhat are the steps to be
followed in completing a case study?
In Chapters 5 and 6, the mechanics of the Tagucbi design of experi-
ments and the procedure for the analysis of the experimental data were
discussed in detail. Those chapters included several application examples
(case studies). The examples in this chapter are representative of practical
problems the author has encountered during his associations with various
industries and clients.
A typical application of the method will include the following five
major steps (see also Figure B-1):
1. A brainstorming session
2. The designing of the experiment
3. The conducting of the experiments
4. Analyzing the results
5. The running of the confirmation test
Brainstorming for Taguchi experiments. is described in Chapter 8. It

is an essential element of a Taguchi case study. \\kn this step is com-
pleted. the planning is done, Each of the experimental situations may
Je:l13:ld :. 'Jnique quality )bjective. \\'hat lr~ ~he ::tt~ribute5 of :he quality
un A PSmer on the Taguchi Method
characteristics? In what manner should the results be monitored? How

many factors should be included in the study? These and many other
pertineDt questions are answered in the brainstorming session. Brain-
stonning was discussed in detail in Chapter 8. In this chapter, the re-
maining four steps of the Taguchi methodology wiUbe clarified by the
followiBg examples. In the solutions of these examples, extensive use is
made of Ihe software of Reference 11, which computes results following
procedures described in Chapters 5 and 6.
'-2 APPLICATWN EXAMPLES INCLUDING DESIGN AND

ANALYSIS
Example '·1
Engine Val'¥e-Train Noise Study
An experiment is to be designed to study the influence of six factors,
which were ideWjfied during brainstorming as influencing the noise emit-
ted by . . valv~$ainofa newly developed engine. Each factor is assigned
two levels. Thl~rainstonning session concluded that interaction effects
were I8:h lessJmportant than the main effects.
Soluti.
Since there are six 2 level factors, the smallest array is L8 • Since inter-
actions . . insignificant, the six factors can be assigned to the six of the
seven columns in any order desired. The factors involved and their levels
are shown in Table 9-l(a).
Assume that during the brainstorming session, the quality character-
istics aud the methods of measurement were also determined, in addition
to the &ctors and levels. Based on these criteria certain key elements of
the test plan are described in Table 9-1(a), using the principles of the
design of experiment. These are shown un~r the headings. '''Note'',
"Objective" and "Characteristic". For this exPeriment, the level of the
noise was to be measured in tenns. of some noise index on a scale of 0
to 100. The index was so defined that its smaller value was always
desirable.
The l.s array is shown in Table 9-1(b). Note that only six columns
define the test condition with the zeros in the 'unused column (column 7)
showing that 10 condition is implied. The 2 level array L s ' describes
--
..... _._- --_ .. _---
Eltample~ of Tapchi C~ Studies 183
Table 9-1. Engine Valve-Train Noise Study (Design)

(a) Design Factors and Their Levels
COLUMN FACTORS LEVEL I LEVB.2 LEVEL 3 LEVEL 4
I Valve guide clearance Low Higb

2 Upper guide length Smaller Larger
3 Valve geometJy Type I Type 2
4 Seat concentricity Quality I Quality 2
5 Lower guide length Location I LocatioB 2
6 Valve face runout Run out type I Run out type 2
7 (Unused)
Note: Six variables all at two levels studied.
Objective: Determine design configuration for least noise.
9taracteristic: Smaller the better (measured in tenns otnoise index.).
(b) The Experiment Uses the Following Array
lRlALICOLUMN 2 3 4 S 6 7
Trial 1 I I I 1 1 1 0
Trial 2 1 I 1 2 2 2 0
Trial 3 1 2 2 I I 2 0
Trial 4 1 2 2 2 2 1 0
Trial 5 2 I 2 I 2 1 0
Tria16 2 1 2 2 I 2 0
Trial 7 2 2 I I 2 2 0
Trial 8 2 2 1 2 I 1 0
eight trial conditions. The design may be cr~ated manually, but a computer
program will perform such computations in a matter of seconds and
without mathematical errors.
The results of the eight trial conditions, with one run per trial condition,
are shown in Table9-I-l(a). Examples in this chapter utilized the software
of Reference 11 which displays up to 6 repetitions and their averages.
These observed results are used to compute the main effects of the in-
dividual factors (Table 9-1-1(b». Since the factors have only two levels,
the main effects are shown under the two columns marked Level 1 and
Level 2. The third column labeled (L2-L 1) contains the difference between
the main effects at Level 1 and Level 2. A minus sign (in the difference
column) indicates a decrease in noise as the factor changes from Level
1 to Level 2. A positive value. on the other hand. indicates an increase
184 A Primer 00 the Taguchi Method
TaMe 9-1-1. Original Data and Their Averages (Results and ~ysis)
(a) Originai Observations and Their Averages

QaIty Characteristic: Smaller is Better
Results: (Up to 6 Repetitions Shown)
l'RIAIIi
REPE'l1'ml)NS RI
I 45.00 45.00
2 34.00 34.00
3 56.00 56.00
4 45.00 45.00
5 46.00 46.00
6 34.00 34.00
1 39.00 39.00
8 43.00 43.00
(b) 1k Main Effects
COLUID FACTOR NAME LEVEL I LEvEL 2 (L2 - LII LEVEL) LEVEL 4
,I Value guide
cleamace
45.00 40.50 -4.50 00.00 00.00
2 Upper guide length 39.15 45.15 6.00 00.00 00.00

3 Valve geometry 40.25 45.25 5.00 00.00 00.00
4 Seat concentricity 46.50 39.00 -7.50 00.00 00.00
5 Lower valve length 44.50 41.00 -3.50 00.00 00.00
6 Valve face run 44.75 40.15 -4.00 00.00 00.00
I
~
(c) ANOVA Table
COl.UJCi£ FACTORS OOF SUM OF SQUARES VARIANCE F
I Valve guide (I) (40.50) Pooled

2 Upper guide 1 12.00 12.00 2.011 9.96
3 Valve geometry (I) (50.00) Pooled
4 Seat I 112.50 112.50 3.142 21.10
concentricity
5 Lower guide (I) (24.50) Pooled
6 Valve face (I) (32.00) Pooled
All Of:f'rerslerror 5 179.00 35.80 68.94
Totll: 7 363.50 100.00
Examples of Tagm:hi Case Studies 185.
in noise. A quick inspection of the difference columnpennits selection

of the optimum combination, for example, the "smaller is bettern char-
acteristic. A negative sign in the column (L2-L1) indicates Level 2 of
the factor is desirable while a positive value indicates Level 1 is the
choice. This quick inspection is a sufficient test only when two levels
are involved and an factors are considered significant.
If the desired characteristic was~:''. ger is better" ·theft the level se...
... . ..-
..
." '"
-.~4:- ~,
Table '·1-1. Engine Valve-Train Noise SIaIJ

(Optimnm and Confidence Interval)
(a) Estimate of the Optimum Condition of Design/Process: For SmdIer the Better
Characteristics
I
FACfOR DESClUnJON LEVEL DESCRIP'TlON LEVEL CONTRIBL'TION
Up~guide length Smaller I -3.0000

Seat concentricity Quality 2 2 -3.7500
Contribution from ail factors (total) -6.15
Current grand average of perfonnance 42.75
This esti(lJate includes only those variables thal have a $ignificant contribution. i.e•.• ~ "arjables are ex.cluded
from the estimate. Estimates may • be made with variables of choice.
(b) Confidence Interval
Computing F function for I and 5 at 9O%coofidence leveL

Confidence Interval (c.i.) is expressed as:
CI."'~ ~F(l~:Z)Vfr
Where F(n" nz) = Computed value of F with nl = 1. nz = enor DOF
at a desired confidence level
Ve = Error variance
Ne = Effective number of replications
Based on: F = 3.2999999 nl = I. n;! =5 V.. = 35.8 N.. = 2.6667

The confidence interval C.l. = ± 6.65601 L which is the variatioo of the estimated result
at the optimum condition. Le .. the mean of the result. m lies between (m + C. I.) and
(m - C.Ll at 89.9 Yc confidence level.
186 A PriI1lel' OD the Taguchi Method
lectioD criteria will be the reverse of the scheme given above, positive
values indicate Level 2, and all negative values will indicate choice of
Levd I for the optimum condition. In this example with all factors, the
opti_m condition for (,(,smaIler is better" is levels 2, I, I, 2, 2, and 2
for factors in columns 1 through 6 respectively_The sign ( .... ) directs the
selection of levels, while the magnitude sugests the strength of the
inftuaJCe of the factor. The quantitive measure of the influence of indi-
vidual factors is obtained from ANOVA (Table 9-1-l(c».
ANOVA follows procedures outlined in Chapter 6. No new data or
decisions On the part of the experimenter a.rerequired. This is an ideal
situation for standard computer routines. Tberesults of ANOV A are
shown in Table 9-1-1(c). A review of the percent colUllUl shows that
Upper Guide (9.96%) and Seat Concentricity (21.10%) are significant..
The other insignificant factors are pooled (combined) with the error term.
Based on information from the ANOVA Table 9-1-l(c), the mean per-
as .wn
fOlll&lce at optimum condition and the confidence interval are calculated
in Table 9-1-2(a) and Table 9.. 1-2(b)~respectiveIy ..
11Ie last s~pin the analysis is to estimate the performance at the
optiomm condition. Normally,only the significant factors are used to
mate this esti.nlate. An examination of main effects indicates which levels
wilbeincluded in the optimum condition. In addition, ANOVA indicates
(by dle percentage column in Table 9-1-1(c»the relative influence of
each factor. Thus all the necessary information for the determination of
the optimum condition and the expected value of the response at this
condition is available. No new information is necessary to calculate the
performance at the optimum condition.
Example 9-2
Study of Crankshaft Sm1ace FiDishing Process
Recently, an engine was found to have an unusually high rate of
-: cnmbhaft bearing failures. Engineers identified the crankshaft surface
finishing as the root cause. A brainstorming session, with the engineers
and the technicians involved in design and manufactUring activities, re-
. which were considered to have a
sulted in the selection of six factors
major influence on the quality of the surface finish. The Taguchi approach
, of experimental design was considered an effective way to optimize the
process.
Examples of Taguchi Case Studies 187 I
The brainstonning also identified two levels for each factor and a likely
interaction between two of the factors. The group decided that the quality
characteristic of the surface finish should be measured in terms of du-
rability (life) under simulaled laboratory tests.
Solution
With six factors and one interaction involved in this study ~ La is suitable
for the experimental design. The first step is to decide where to assign
the interacting factors and which column to reserve for their interaction.
The table of interaction (Table A-6) for 2 level ortbogonaJ arrays shows
that columns I, 2, 3 form an interacting group. The two interacting factors
Table 9·2.. 1. Study of Crankshaft Surface-Finishing Process (Design)

COLUMNS FACfORS LEVEll LEVEl 2 LEVEl 3 LEVEl 4
1 Roundness (Jobing) 700 1600

2 Lay direction Least Most
(cross)
3 Interaction NJA
4 Tp (prQtess index) 65% 90%
5 Taper .00025 .0005
6 Wavines Lower limit Upper limit
7 Shape factor .0003 .0003
Note: Interaction between I'Ol.IlJdaess ad lay direction.
Objective: Detetmine best grimhg p;muneJers.
Characteristic: Bigger the better (bearing durability life).
(b) Orthogonal Array Used for the Experiment
TRIAliCOLUMN 2 3 4 5 6 7
Trial I 1 I 1 1 1 1 I
Trial 2 1 1 1 2 2 2 2
Trial 3 I 2 2 I I 2 2
Trial 4 I 2 2 2 2 I I
TrialS 2 I 2 1 2 I 2
Trial 6 2 1 2 2 I 2 1
Trial 7 2 2 1 1 2 2 1
Trial 8
., ")
I 2 I 1 2
.. A Primer on the Taguchi Method
Table '-2-2. Crankshaft Surface-Finishing Process

(Main Effect and A..'10VA)
(a) Original Observations and Their Averages
Quality Characteristic: Bigger is Better.
TBW.I
IW'EUiIONS R,
I 34.00 34.00
2 56.00 56.00
3 45.00 45.00
4 35.00 35.00
5 46.00 46.00
6 53.00 53.00
7 43.00 43.00
8 41.00 41.00
(la) The Main Effecl$
t't'.lUi"MN FActoRS LEVEL I l..E\B.. 2 (L2 - Lt) LEVEL 3 LEVB.4
1 Roundness 42.50 45.75 3.25 00.00 00.00

(lOOmg)
2 Lay direction 47.25 41.00 -6.25 00.00 00.00
3 Interaction 43.55 44.75 1.25 00.00 00.00
4 Tp (process index) 42.00 46.25 4.25 00.00 00.00
5 Taper 43.25 45.00 1.75 00.00 00.00
6 Waviness 39.00 49.25 10.25 00.00 00.00
7 Shape factor 41.25 47.00 5.75 00.00 00.00
(c) ANOVA Table
COl.UMN FACTORS DOF SUM OF SQUARES VARIANCE F PERCENT
1 Roundedness 1 21.13 21.13 5.02

2 Lay direction 1 78.13 78.13 18.56
3 Interaction 1 3.13 3.13 0.74
4 Tp (process index) 1 36.13 36.13 8..58
5 Taper I 6.13 6.13 1.46
6 Waviness 1 210.13 210.13 49.93
7 Shape factor 1 t>6.13 66.13 15.71
All others!error 0
Total:
., \00.00
420.91
Examples of Tapchi C~ Studies 189
Table 9-2-3. Crankshaft Surface-Finishing Process

(Pooled ANOVA and Optimwn)
(c) ANOVA table
COWMN FACTORS OOF SUM OF SQUARES VARIANCE F PEaCENT
I Roundedness 1 21.13 21.13 4.57 3.92

2 Lay direction I 78.13 18.13 16.89 11.46
3 lDteraetion (1) (3.13) Pooled
4 Tp (process I 36.13 36.13 7.81 7.48
index)
5
6
7
Taper
Waviness
Shape factor
(1 )
I
I
(6.13)
210.13
66.13
Pooled
210.13
66.13
45.43
14.30
48.83
14.61
I
~
.1\
t
All others/error
Total:
2
7
9.25
420.91
Nl.$I:: Insignificant factorial effects are pooled as shown ( ).

4.63 7.69
100.00 I
~) Estimate of the Optimum Condition of DesignlProcess: For 'Bigger the Better
Characteristics
FACTOR DESCRIPTION LEVEL CO!'-TRIBUTION
Roundness (Iobing) 1600 2 1.6250

Lay direction (crosshatch) Least 1 3.1250
r,. (process index) 90% 2 2.1250
Waviness Upper limit 2 5.1250
Shape factoJ: .0003 2 2.8750
Contribution from all factors (total) 14.875
Olrrent grand average of performance 44.125
&peeted result at optimum condition 59.00
This estimate includes only those \·ari;il:..les that have a signilicaiu contribution. i.e.• pooled variables are excluded
from the estimate. Estimates may also be made with variables of choice.
are therefore assigned to columns 1 and 2. Column 3 is kept aside for

their interaction. The remaining four factors are then assigned to any of
four remaining columns. The completed design, with descriptions of
factors. their levels and the ortho~onal.... array. are shown in Table 9-2-
Ha&b). Eight crankshafts were fabricated to the specifications described
1,)y the -.:ight ~rial conditions. Each sample was tested for durability llifel.
198 A Primer on the Taguchi Meth<xl
Since longer life was desirable. the quality characteristic applicable in

this case was Ubigger is better."
The observed durability, the main effects and the unpooled ANOVA
are shown in Table 9-2-2. The study of the main effects indicates some
interaction between the factors. This is shown by the magnitude 1.25 in
the cokmm labeled (LrL 1) in Table 9-2-2(b). This value is of the same
order of magnitude as the values 3.25~ - 6.25~ 4.25, etc. But is the
interaction significant? The answer to this question can be obtained from
the percentage column of the ANOV A table (Table 9-2-2(c». The inter-
action UDder column 3 is only 0.74%. Contributions below'5 percent are
generally not considered significant. The interaction and the factor in
COIUlDD 5 which has 1.46 in the percentage column, are pooled. The
pooled ANOVA is shown in Table 9-2-3(a). Observe that upon pooling
the pereentage values the significant factors are adjusted slightly.
In estimating the performance at the optimum, only the significant
faetOl'5 are used. As shown in Table 9-2-3(b)~ the expected improve-
me" in performance is 14.875 over the current average of perfOnDaDCe$
.,.,,*, .
(44.125). Since the interaction (Table 9-2-3(b» has little significance"

it irDOt considered in the selection of levels for the optimum condi-
tion; .
Example ' ...3

Automobile Generato.r Noise Study
Engineers identified one 4 level factor and four 2 level factors for
experimental investigation to reduce the operating noise of a newly re-
leasea- generator. Taguchi methodologies were followed to layout the
experiments and analyze the test results.
Solution
The factors in this example present a mixed level situation. Although,
experiment design is simplified if all factors have the same level, it is
not always possible to compromise the factor level. For instance, if a
factor influence is believed to be nonlinear, it should be assigned three
or more levels. The factor and its influence are assumed to be continuous
functions. If however, the factor assumes discrete levels such as design
type 1. design type 2, etc .. then the influence is a discr~te function and
each discrete step (level) must be incorporated in the design. The 4 level
~'ac~or ',n :he -;.'''<.lmDI~, :$ lis..::-et.e Since :he j :eve! ~actor has 3 DOF.
E'{amples of Taguchi Case Studies 1'1
and four 2 level fac~ors each have 1 DOF. the total DOF for the experiment
\ is 7. L 8 with seven 2 level columns and 7 OOF was selected for the
design. The first step provides for the 4 leyel factor. Columns 1, 2, and
3 of La are used to prepare a 4 level column. This new 4 level column
now replaces column 1 and is assigned to the 4 level factor. As columns
2 and 3 were used to prepare column 1 as a 4 level column, they cannot
be used for any other factor. Thus the four 2 level factors are assigned
to the remaining columns 4, 5, 6, and 7. The design and the modified
OA are shown in Tables 9-3-l(a) and (b).
One run at each trial condition was tested in the laboratory and the
performance was measured in terms of a noise index. The index ranged
between 0 (low noise) and 100 (loud noise). The lower value of this
index was desirable. The data and calculated main effects are shown ill
, Table 9·3·1. AutomobUe Generator Noise (Design)

COLUMNS FACTORS LEVEL 1 LEVEL 1 LEVEL 3 LEVEL 4
I Casement structure Present Tex.tured

2 (Unused)
3 (Unused)
4 Air gap Present Increase
5 Impregnation Present type Harder type
6 Contact brush Type I Type 2
7 Stator structure Present design Epoxy coated
Note: One 4 level and four 2 level factors are studied.
Objective: Detennine generator design paralneters for least noise.
Characteristic: Smaller is better (measuted in .predefined indexl.
(b) An Ls Orthogonal Array Used in the Experiment
TRIAlJC01.UMN 2 3 4 5 6 7
Trial I I 0 0 I 1 I I
Trial 2 I 0 0 2 2 2 2
Trial 3 2 0 0 I 1 2 2
Trial 4 2 0 0 ")
2 I I
Trial 5 3 0 0 2 I 2
Trial 6 3 0 0 "} "l
.:. I
Trial "7 -+ 0 0 ') ")
I
f •
Trial 3 -+ 0 0 -"' -"\
~
Table 9-3-2. Automobile Generator Noise (Main Effects)

Ca) Original Observations and Their Averages
Quality Characteristic: Smaller is Better.
TRIAU
1W'ETI110N$ RI Rs AVERAGE
1 50.00 50.00
'2 62.00 62.00
3 70.00 10.00
4 75.00 15.00
'5 68.00 68.00
6 65.00 65.00
7 65.00 65.00
8 74.00 74.00
(b) Main Effects
LEVEl I LEVEl 2 (L} - LI) LEVEL 3 LEVEl 4
I Casement structure 56.00 72.50 16.50 66.50 69.50

4 Air gap 63.25 69.00 5.75 00.00 00.00
5 Impregnation 64.75 67.50 2.75 00.00 00.00
6 Contact brush 66.75 65.50 -1.25 00.00 00.00
7 Stator structure 63.75 68.50 4.75 00.00 00.00
Tables 9-3-2(a) and (b), respectively. Note that the 4 level factor in
column I (Casement Structure) has its main effects at the four levels.
This factor has 3 DOF as noted in the ANOVA table (Table 9-3-3(a»
under the column marked DOF.
From the ANOVA table it is clear that the factor assigned to column
6 (Contact Brushes> has the smallest sum of squares and hence the least
influence. This factor is pooled and the new ANOVA is shown in Table
9-3-3(a). Using the significant contributors, the estimated performance
at the optimum condition was calculated to be 49.375. In this case the
optimum condition is trial I (Levels I I 1 1 I). The result for trial I was
50 (Table 9-3-2(a». The difference between the trial result and the es-
timated optimum performance (49.375) resulted from the dropping of the
minor effect of the contact brush factor from the estimate.
,
p"'
" t
..
Example~ of Ta!uchi C~ Studies 193 I!
Table 9-3-3. Automobile Generator Noise (A~OVA and Optimum)

(a) ANOVA Table
I
CQLlJMN FACTORS OOF SUM OF SQUARES VARIA....a: F
1 Casement Structure 3 309.38 103.13 33.00 70.49

4 Air gap 1 66.13 66.13 21.16 15.07
5 Impregnation 1 15.13 15.13 4.84 3.45
6 Contact brushes (1) ;. (3.13) Pooled
7 Stator structure 1 45.13 45.13 14.44 9.51
Total: 7 438.88 100.00
Note: Insignificant factorial effects are pooled a1J shown ( l.
(b) Estimate of the Optimum Condition of DesignlProcess: For Smaller the Better
Characteristic
FACTOR DESCRIYl10N LEVEL DESCRIPTION LEVEL CONTRlBtJTlOS
Casement structure Present design I -10.1250

Air gap Present gap 1 -2.8750
Impregnation Present type 1 -1.3750
Statot structure Present design 1 -2.3750
l
;
Contribution from all factors (total) -16.750 ~
Current grand average of performance 66.125
This estimate includes only lhose variables that have significant contnbutions. i.e.• pooled "ariables are excluded
from the estimate. Estimates may also be made with variables of chOice. .
II
!
I
Eutnple 9-4
EDgine Idle Stability Study
An engine development engineer identified three adjustment parameters
controlling tbe idle performance or an engine. Each of the factors is to
be studied at three levels to determine the best setting for the engine. A
Taguchi experiment design is to be utilized.
Solution
The smallest three level OA, ~, has four 3 level columns. With three 3
level factors in this study, the L 9 is appropriate for the design. The factors
are placed in the first three columns. leaving the fourth column unused.
Table 9-4-1. Engine Idle Stability Study (Design)

(a) Desiga Factors and Their Levels
COLUMNS FACTORS LEVEL 1 LEVELZ LEVEL 3 LEVEL 4
1 Indexing -5 Deg ODeg +5 Deg

2 Overlap area. 0% 30% 60%
3 Spark advance 25 Deg 30Deg 35 Deg
4 (Unused)
Noce: 1'Ilrec ] level factors studied.
()bjective: Daermine best e11Jine setting.
~ SmaIJer is beUer (Speed deviation).
(b) The Elperiment Used an LJ Orthogonal Array

~ Z 3 4
Trial I 1 1 1 0
TriaI.l I 2 2 0
Trial) I 3 3 0
Trial 4 ""'" 2 1 3 0
Trial , 2 2 1 0
Trial 6 .- 2 3 2 0
Trial 1 3 I 2 0
Trial a 3 2 3 0
Trial' 3 3 I 0
The fadors, their levels, and the modified OA are shown in Tables
9-4-1 (a) and (b).
The performance of the engine tested under .various trial conditions
was measured in terms of the deviation of the speed from a nominal idle
speed. A smaller deviation represented a more stable condition. Three
separate observations were recorded for each trial condition as shown in
Table 9-4-2(a). The signal to noise ratio (SIN) was used for the analysis
of the results. The main effects, optimum condition, and ANOVA table
are shown in Tables 9-4-2 and 9-4-3. Based on the error DOF and vari-
ance, the confidence interval of the estimated performance at optimum
is also computed as shown in Table 9-4-3(b). The confidence interval
(C.L) value of + .3341 will mean that the estimated optimum perfor-
mance (SIN ratio) will be - 25.878 + .3341 at 90% confidence level
(89.77% as a result of numerical solution by the computer software of
Reference 11).
Examples of Taguchi Case Studies 195
Table 9-4-2. Engine Idle Stability Studies (Main Effect and A.'10VA)
(a) Original Observations and Their SIN Ratios
1 20.00 25.00 26.00 -21.54

2 34.00 36.00 26.00 - 30.19
3 45.00 34.00 26.00 -31.10
4 13.00 23.00 22.00 -25.%
5 36.00 45.00 35.00 -31.81
6 23.00 25.00 34.00 -28.87
7 ~5.00 45.00 53.00 -33.06
8 56.00 46.00 75.00 -35.60
9 35.00 46.00 53.00 -33.12
(b) The Main Effects
COLUMN FACTORS LEVEL I LEVEl. 2 (Lz - Lo LEVEL 3 LE\U4
1 Indexing -29.61 -28..88 0.73 -33.93 00.00

2 Overlap -28.85 -32.53 -3.69 -31.03 00.00
3 Spark. advance -30.67 -29.76 0.91 - 31.99 00.00
(c) ANOVA Table
COLUMN FACTORS OOF SUM OF SQUARES VARIANCE F PEBca.-r
1 Indexing 2 44.636 22.318 932.80 -61.26

2 Overlap 2 20.541 10.271 429.27 28.15
3 Spark direction 2 7.565 3.783 158.10 10.33
Total: 8 72.79 100.00
Example 9-5
Instrument Panel Structure Design Optimization
A group of analytical engineers undertaking the design of an instru-
ment panel structure. are to study the influence of five critical struc-
tural modifications on the system. A finite element model of the total
"
structure was available for a static stiffness analysis. The objective is
Table 9-4-3. Engine Idle Stability Studies

(Optimum and Confidence Interval)
(a) Estimate of the Optimum Condition of DesignlProcess: For Smaller the Better
Characteristic
FACTOR DESCRJPIlON LEVEL DESCRlPTIOH
Indexing 0 Deg 2 1.9256

Overlap 0% I 1.9522
Spark direction 30 Deg 2 1.0489
Contribution from.all factors (total) 4.92667
Current grand avaage of performance -30.80SS6
Expected result at optimum condition -25.81889
Computing F function for I and 2 at 9O'JJ confidence level.

Confidence Interval (C.I.) is expressed as:
C.1. =
Where £(nl. nv = Computed value of F with nl = I, n2= eITOf DOF

Vc: = Error variance
Vc: == Effective number of replications
Based on: F = 5.999996 nl = I, n2 = 2 V.. = 2.392578E-2

N.. = 1.285714
The confidence interval C.I. == ±O.3341461

This is the variation of the estimated result at the optimum condition, Le., the mean of
the result, m lies between (m + c.1.) and (m - C.I.) at 89.77% confidence level.
to determine the best combination of design alternatives. To reduce the

number of computer runs, a Taguchi experiment design was selected
to determine the number and conditions of the computer runs necessary
if each factor is~ be studied at 2 levels. Two interactions are believed
to be important.
Examples of Ta;pchi Case Studies 197
Solution
This investigation is an analytical simulation rather than a hard-
ware experiment. The factors' and levels shown in Table 9-5-1 (a) Me
used in an 40A to set up the simulation. Only one om per trial cond-
ition is necessary since the computer results should not change with
repetition. The observation (stiffness values), the main effects aDd
the optimum condition are shown in Tables 9-5-1(1), 9-5-2(a), and
9-5-2(b).
Table '-5-1. Instrument Panel Structure OpImdzation

(Design and Data)
COLUMN LEVEL I LEVEL 2 LEVEl. 3 l.£\'El. 4
I Front dash beam Solid Hollow

2 Under structure With Without
3 Interaction I x 2 N/A
4 Forward panel Current New desip
5 Interaction 1 x 4 N/A
6 Plenum structure Steel Plastic
7 Surface structure Baseline New
Note: Interactions I x 2 and I x 4 studied.
Objective; Detenninesuuetural paramelel'li for maximum sm=ngtb.
Characteristic; Bigger the better (measured in terms of stiffne$s).
(b) Original Observations and Their Averages

Quality Characteristic: Bigger is Better.
TRIAL!
REI'ETITIONS RI R2 R3 Rs A\-'BAGE
1 13.50 13.5'0
2 14.00 14.00
3 14.30 14.30
4 13.10 13.10
5 22.00 22.00
6 18.00 18.00
7 29.90 29.90
8 16.00 16.00
198 A PIimer on the Taguchi Method
Table 9-5-2. Instrument Panel Structure Optimization

(l\fain Effect and Optimum)
(al The ~bin Effects
COl..\.'"\1'S fACTORS LEVEl ! LEVEl 2 IL2 - Ld LEVEl 3 LEVEl 4
I Froot dash beam 13.73 21.48 7.75 00.00 00.00

2 Cnder structure 16.88 18.33 1.45 00.00 00.00
3 Interaction I x 2 18.35 16.85 -1.50 00.00 00.00
4 forward panel 19.92 15.28 -4.65 00.00 00.00
5 Intel3Ction I x 4 15.45 19.75 4.30 . 00.00 00.00
6 Plenum structure 16.15 19.05 2.89 00.00 00.00
7 Surface structure 18.63 16.58 -2.05 00.00 00.00
(b) EstirnIR of the Optimum Condition of Design/Process: For Bigger the Better
Charat1rmstic
fACTOR DE'9Om'TION LEVEl DESCRIPTION LEVEL CONTlUBUTION
Front das8 i:leam Hollow 2 3.8750

FOlWard pmel Current design I 2.3250
Plenum stmeture Plastic 2 1.4500
Contribun- from all factors (total) 7.64999
Current gatd average of performance 17.64000
Expected aultat optimum condition 25.25000
Th:i e£:::r.ae:m::!udes only those variables that make a significant contribution. i.e•• pooled variables are excluded
from the e5ti!mte. Estimates may also be made with variables of choice.
Example 9-6
Study Leading to the Selection of the
Worst Case Barrier Test Vehicle
To asswe that the design of a new vehicle complies with all the ap-
plicable Federal Motor Vehicle Safety Standards (FM'V.SS) requirements,
engineers involved in the crashwortbiness development of a new vehicle
design, want to determine the worst combination of vehicle body style
and optioos. This vehicle is to·be used as the test specimen for laboratory
validation tests instead of subjecting several prototype vehicles with all
available options and body styles to tests under all compliance conditions.
Four :! level factors and one 4 level factor were considered to have major
Examples of Tagucbi Case Studies 199 .
Table 9-6-1. Seledion of Worst Case Barrier Vebicle (Design)

(a) Study leading to Selection of Worst Case Barrier Vehicle Design Factors and
Their Levels
COLUMNS FACTORS LEVEL I LEVEL 2 LEVEL 3 LE\'EL4
1 Test type o Deg-F 30 Deg-R 30 Deg-L NeAP

2 (Unused)
3 (Unused)
;.
4 Type of vehicle Style I Style 2
5 Power train Light duty Heavy duty
6 Roof structure Hard top Sun-roof
7 Seat structure Standard Reinforced
NOIe: One 4 level and four 2 level fllctOr'S studied.
Objective: Detemtine the worst vehicle/option combination.
Cbaqctetistic: Smaller is better (one or more injury criteria). t
(b) The Experiment Used an La Onhogonal Array
TRIALtCOLUMN 2 3 4 S 6 7
Trial I I 0 0 I 1 1 1
Trial 2 1 0 0 2 2 2 2
Trial 3 2 0 0 1 I 2 2
Trial 4 2 0 0 2 2 1 1
Trial 5 3 0 0 1 2 1 2
Trial 6 3 0 0 2 I 2 I
Trial 7 4 0 0 1 2 2 1
Trial 8 4 0 0 2 1 1 2
I
influence on the perfonnance. A Taguchi experimental design approach
was followed. I
I
Solution t
The design involved modifying a 2 level column of an 4 into a 4 level
one. The process is similar to that described in Example 7-3. The factors,
their levels, and the modified OA are shown in Tables 9-6-1(a) and (b).
The description of the trial conditions derived from the designed exper-
iment served as the specifications for the test vehicle. For barrier tests.
the specimens are prototype vehicles built either in the production line,
or hand made. one of a kind. test vehicles. In either case the cC'st for
_ A Primer on the Taguchi Method
Table 9-6-2. Selection of Worst Case Barrier Vehicle

(Main Effect aDd ANOVA)
(a) Original Observations and Their Averages
Rz AVERA<iE
I 45.00 . 45.00
2 65.00 65.00
3 38.00 38.00
4 48.00 48.00
5 59.00 59.00
6 ·32.00 32.00
7 36.00 36.00
8 38.00 38.00
(b)'Ibe Main Effects
COLl'MN FACTORS LEVEL I LEVEL 2 (Lz - Ll) U!VEL3 LEVEL 4
I Test type 55.00 43.00 -12.00 00.00 00.00

2 Type of vehicle 44.50 45.75 1.25 00.00 00.00
5 Powertrain 38.25 52.00 -13.75 00.00 00.00
6 Roof stnlcture 47.50 42.75 -4.75 00.00 00.00
7 Seat stnlcture 40.25 50.00 9.75 00.00 00.00
(c) ANOVA Table
COLl"MN FACTORS OOF SUM OF SQUARES VARIANCE F PERCENT
1 Test type 3 336.38 336.38 4.648 27.71

4 Type of vehicle (I) (3.13) Pooled
5 Powertrain 1 378.13 378.13 15.674 37.15
6 Roof stnleture (I) (45.13) Pooled
7 Seat stnlcture 1 190.13 190.13 7.881 17.42
All others/error 2 48.25 24.13 17.72
Total: 7 952.88 100.00
--~ _._-------~ '"
Examples of Tapchi Case Studies 261
Table 9-6-3. Selection of Worst Case Barrier Vehicle

(Optimum and Contidence Interval)
(a) Estimate of\the Optimum Condition of DesignlProcess: For Smaller the Better
Characteristic
FACTOR DESCR1J7IlON LEVEL DESCRIPTION
Test type NeAP 4 -8.1250

Powertrain ;.. Light duty 1 -6.8750
Seat stlUcture Standard 1 -4.8750
Contribution from all factors (total) -19.8750
"This estimate includes only Ibose variables that mak.e a significant contribution. Le.• pooled vllti$lcs are ellduded
Computing F function for I and 2 at 90% confidence level.

Confidence Interval (C.I.) is expressed as:
C.l. :::
Where F(nJ, nz) ::: Computed value of F. withn\ ::: I, nz ::: error DOF
Vc ::: Error variance .
Ne ::: Effective number of replications
Based on; F == 5.999996 n\ == 1, n2::: 2 V.. = 24.125 N.. = 1.333333

The confidence interval C.I. == ± 10.4
preparing the test vehicles could easily run in the hundreds of thousands
of dollars. 'Proper specification, in a timely manner" is crucial to the cost
effectiveness of the total vehicle development program. For the purposes
of the tests, eight vehicles were built in the production line following the
specifications which correspond to the eight trial conditions. These ve-
hicles were barrier tested and the results recorded in tenns of a predefined
occupant injury index. The results and the analyses are shown in Tables
9-6-2 and 9-6-3. By using eight test vehicles. the engineers were able to
202 It. Primer on the TaguchiMethod
Table 9-7-1. Air Bag Design Study (Design)

(a) Desip Factors and Their Levels
FACTOR NAME LEVEL I LEVEL 2 LEVEL 3 LEVEL 4
1 (Unused)
2 Steering colUmIl rotation Rot I Rot 2 Rot 3 Rot 4
3 Steering column crash 300 350 400 SOO
stiffness
4 Knee bolster stiffness SI S2 S3 S4
5 Knee bolster location 100 mm 115 mm 150 mm 175 mm
6 laftation rate Rate 1 Rate 2 Rate 3 Rate 4
7 Development time 14 ms 18 ms 24ms 32ms
8 Vent site 650 mm 970 mm 1300 mm 1625 mm
9 Bag size (E-7 mm) 5.0 5.8 6.5 7.2
10 Maximum bag pressure PI P2 P3 P4
(b) 0nIl0pnaI Array
TRJAU
COLL"MN 1 2 3 4 5 6 7 8 9 10 RI
Trial I 0 1 I 1 1 1 1 1 I I 8.00
Trial 2 0 1 2 2 2 2 2 2 2 2 5.50
3 3
Trial 3
Trial
Trial 5
. 0
0
0
I
I
2
3
4
I
3
4
1
3
4
2
3
4
2
4
3
4
3
3
4
4
3
4
4
5.00
7.00
8.00
Trial 6 0 2 2 2 1 1 4 4 3 3 4.00
Trial 1 0 2 3 3 4 4 1 1 2 2 5.00
Trial 8 0 2 4 4 3 3 2 2 1 1 7.00
Trial 9 0 3 1 2 3 4 1 2 3 4 8.00
Trial 10 0 3 2 1 4 - 3 2 1 4 3 3.00
Trial B 0 3 3 4 1 2 3 4 1 2 5.00
Trial 12 0 3 4 3 2 1 4 3 2 1 4.00
Trial 13 0 4 1 2 4 3 3 4 2 1 6.00
Trial 14 0 4 2 1 3 4 4 3 1 2 4.00
Trial 15 0 4 3 4 2 1 1 2 4 3 7.00
Triall6 0 4 4 3 1 2 2 1 3 4 5.00
Trial 11 0 1 1 4 1 4 2 3 2 4 3.00
Trial 18 0 1 2 3 2 3 1 4 1 4 5.00
Trial 19 0 1 3 2 3 2 4 1 4 1 7.00
Trial 20 0 I 4 I 4 1 3 2 3 2 9.00
Trial 21 0 2 I 4 2 3 4 I 3 2 5.00
Trial 22 I 0 2 2 3 I 4 3 2 4 I 6.00
T
~
I
1 Examples of Taguchi Case Studies 203
Table 9-7-1. Continued I I

~
I I
TRlAlJ \ f
I
COLUMN I 2 3 4 5 6 7 8 9 10 RI I'
I I
Trial 23 0 2 3 2 4 I 2 3 I 4 7.00 ! II
I
Trial 24 0 2 4 I 3 2 I 4 2 3 7.00 . I
Trial 25 0 3 I 3 3 I 2 4 4 2 6.00
Trial 26 0 3 2 4 4 2 I 3 3 I 5.00
Trial 27 0 3 3 I I j 4 2 2 4 4.00
Trial 28 0 3 4 2 2 4 3 I 1 3 5.00
Trial 29 0 4 I 3 4 2 4 2 I 3 6.00
Trial 30 0 4 2 4 3 I 3 I . 2 4 7.00
Trial 31
Trial 32
0
0
4
4
3
4
I
..,
2
I
4
3
2
I
4
3
3
4 2
I 8.00
4.50 i
j
learn the worst vehicle configuration. This information was then used to
adhere to several of the compliance regulations.
Example 9-7
Airbag Design Study
t
I
Engineers invoJved in the development of an impact sensitiveinftatable
\
II airbag for automobiles, identified nine 4 level factors as the major inftu~
! ences on performance. Using this information the Taguchi experimental
design approach was used to determine the optimum design.
Solution
Since the experiment involves n~e 4 level factors, L 32 with the nine 4
level columns and one 2 level column was ~elected for the design. Since
there is no 2 level factor in this design, the 2 level column (column 1)
of the OA shown in Table 9~7-l(b) is set to zero. The factors, their levels.
and the analyses are shown in Tables 9-7-1(a) through 9-7-3(b). The
study was done using a theoretical simulation of the system. The trial
conditions were used to set up the input conditions for the computer runs.
The results of the computer runs at each of the trial conditions were
recorded on a scale of I to 10 and are as shown with the OA (right mo:-~ '-
column) in Table 9-7-Ubl. The :ruin effects and :\~OV.-\ are DreSe:1te:
Table 9-7-2. Air Bag Design Study (MainEft'ed and ANOVA)

(a) The Main Effects
COLUMN FACTOR NAME I..EVEL I LEVa.2 (L2 - Lt> I..EVEL :3 LEVa.4
2 SceeriBg column rotation 6.19 6.13 -0.07 5.00 5.94

3 ~ ooIwnn crush stiffDess 6.2S 4.94 -1.32 6.00 6.06
4 Knee bolster stiffness 6.38 5.88 -0.50 S.2S 5.15
5 Knee "-location 4.94 5.94 1.00 6.38 6.00
6 IdacioB rate 6.so 6.06 -0.44 4.94 5.15
7 Dewlopmenl time 6.19 5.56 -0.63 6.38 5.13
8 V_ _
5.63 6.56 0.93 5.06 6.00
9 B., size (B-1 mm) 5.88 5.19 -0.69 6.13 6.06
10 Muiumm bag pteSSWe 6.63 5.SO -0.88 5.00 6.38
(c) ANOVATaWe
fACTORS DOF SUM Of SQUARES VAJUANCE F PERCENT
2 SteaiBg column rotation (3) (7.313) Pooled

3 SteaiDg C()Iumn crush (3) (8.438)
Pooled
stiffness
4 Knee bolster stiffness (3) (5.125) Pooled
5 Knee bolster location 3 9.063 3.021 1.631 4.65
6 Inftatioo rate 3 10.438 3.479 1.879 6.48
7 Development time (3) (7.938) Pooled
8 V_size 3 9.563 3.188 1.721 5.32
9 Bag size (B-7 mm) (3) (4.438) Pooled
10 Muimum bag pressure 3 11.125 3.708 2.002 7.39
All others/error 19 35.19 1.85 76.17
Total: 31 75.38 100.00
in Tables 9-1-2(a) and Cb). The optimum vehicle option combination and
confidence level of the design appear in Table 9-7-3(a) and (b).
Example 9-8
Tnmsmission Control Cable Adjustment Parameters
A Tagucbi experiment was conducted to detennine the best parameters
for the design of a transmission control cable. The engineers identified
one 4 level factor and five 2 level factors as well as three interactions
Imong ~hree of the five ~ :evel factors.
-,.
I
"
j
I
Examples of TapdU Case Studies 205
J !
Table '-7-3. Air Bag Design Study (OptimUBi and Confidence) . !
(a) Estimate of the Optimum Condition of DesignlProcess: For S8Iler the Better
Characteristic
FACTOR. DESCIUP1lON
I
Knee bolster location
Inflation rate
Vent size
100
Rate 4
1300 mm
I
4
3
-0.8750
-0.062S
-0.7500
I;
i
I
i
Max bag pressure P3 3 -0.8125

Contribution from all factors (total) -2.SOOO
I
- 1bis estimate inclucJes only dtose variables that have a significant contributioa. i.e.,·JIlIi*d variables are excluded
Computing F-functions for I and 19 at 90% confidence level.

Confidence Interval (C.L) is expressed as:
C.l. " r l
• ~~ · V,
Where F(nh nz)= Computed value of F with nl == I. "l == aror DOF,
at a desired conf1dence level
Ve ::: Error variance
Ne == Effective number of replications.
Based on: F ::: 2.600 nl == I. n2 == 19 V.. == 1.85IfH Ne. == 2.461539
The confidence interval C.I.;;::!: 1.3
Solution
This experiment required both level of modification aml interaction study.
The total DOF for the experiment was 11 [(4 - 1) + 5 x (2 - 1) +
3 x (l x 1)]. L 12 has 11 DOF. But it requires a special OA which
cannot be used for interaction studies. L I6 is selected for the design. The
factors and their levels are described in Table 9-8- Ha). For a 4 level
column and for the three interaction. four g:-,,",ups of natural interaction
:olumns :ire first selected. The sets selected :ue t. 2. 3: j . 8. 15: 11 . ..1.,
_ A P.timer on the Taguchi Method
TaMe '·8·1. Study of Transmission CtlDtrol Cable Adjustment

(Design and Modified OA)
(a) Study of Transmission Conttol Cable Adjustment Design Factors and Their Levels
COL~ FA.C'TOR NA.l,fE LEVEL I LEVEL 2 LEVEL 3 LEVEL 4
1 Adjuster typeIsource Type 1 Type 2 Type 3 Type 4

2 (Used with column 1)
3 (Used with column I)
4 Bracket deflection Low High
S Adjuster load Low High
6 Cable elasticity 1 x 19 7 x 7
7 Interaction 8 x IS NlA
8 Speed of adjuster Low High
9 (Unused)
to (Unused)
11 Interaction 4 x IS NlA
12 Interaction 4 x 15 NlA
13 (Unused)
14 (Un~~
IS Adjustilg torque Low High
NGf,e: A four IweIs factor and three ~.
()bjeetiw: DIIPmine bcstscttingl~
~.NooDnal die best (deviaIionl'rom oomina1 measured).
(b) Orth.twmal Array Used for me Experiment
TRIALl
(.'()l.1J)Di 2 3 4 S 6 7 8 9 10 11 12 13 14 IS
Trial 1 1 0 0 1 1 1 1 1 0 0 1 1 0 0 I
Trial 2 1 0 0 1 1 1 1 2 0 0 2 2 0 0 2
Trial 3 1 0 0 2 2 2 2 1 0 0 I 2 0 -0 2
Trial 4 I 1 0 0 2 2 2 2 2 0 0 2 1 0 0 1
Trial S 2 0 0 1 1 2 2 1 0 0 2 1 0 0 2
Trial 6 2 0 0 1 1 2 2 2 0 0 I 2 0 0 1
Trial 7 2 0 0 2 2 I I I 0 0 2 2 0 0 I
Trial 8 2 0 0 2 2 I I 2 0 0 1 1 0 0 2
Trial 9 3 0 0 1 2 I 2 1 0 0 2 1 0 0 2
Trial 10 3 0 0 1 2 I 2 2 0 0 1 2 0 0 I
Trialll 1 3 0 0 2 1 2 1 1 0 0 2 2 0 0 1
Trial 12 ;3 0 0 2 1 2 I 2 0 0 1 I 0 0 2
Trial 13 4 0 0 1 2 2 1 1 0 0 1 1 0 0 1
Trial 14 , 4 0 0 1 2 2 1 2 0 0 2 2 0 0 2
Trial 15 4 0 0 2 1 I 2 1 0 0 2 2 0 0 2
Trial [6 ~ 4 0 0 2 I I 2 2 0 0 I I 0 0 I
T Examples of Taguchi Case Studies 207
Table 9-9-1. Front Structure Crush Characteristics (Design and Data)

COLUMN FACTOR NAME LEVEL J LEVEL 2 LEVEL 3 LEVEL 4
1 Lower rail section Current Proposed

2 Upper rail geometry Open section Closed section
3 Cross member Present ReiJJforced
;..
Note: Thtee 2 level factors studied.
Objective: Determine best design for barrier crush.
Characteristic: NomiDal the best (impact defonnation).
(b) This Experiment Uses an ~
1'RIAUCOLUMNS 2 3
Trial 1 1 1 1
Trial 2 1 2 2
Trial 3 2 1 2
Trial 4 2 2 1
(c) Original Observations and Their SIN Ratios

Quality Characteristic: Nominal is Best
TRIAL!
REPETITIONS RI R2 R) Rs S!~
1 12.00 14.00 11.00 -6.37

2 18.00 16.00 15.00 -8.46
3 14.00 15.00 15.00 .... 1.76
4 19.00 18.00 15.00 .-11.47
15; and 12, 4, 8. Columns 1, 2, 3 are used to upgrade column 1 into a

4 level column. The other three sets are reserved for the interactions
among the three factors assigned to columns 4, 8, and 15 such that the
interaction between 4 x 15 is shown in column 11, the interaction
between 4 x 8 in column 12, and the interaction between 8 x 15 in
column 7. The factor with 4 levels is assilmed
.... to column 1 which is now
a 4 level column. The remaining two 2 level columns are assigned to the
columns 5 and 6 . The columns 9. 10. 13. and 14 remain unused. The
modified L 1(, and the factors assign~d to the appropriate columns are
;hO\\n :n Table 9-8-11a \.
108 A Primer on the Taguchi Medwd
Table '-9-2. Front Structure Crush Charaderistics

(Main Effect~ ANOVA and Optima)
(a) The Main Effects
COUIMS FACToa~AME LEYE.L I LEVEL ~ IL2 - Ltl LEVEL 3 IDU4
I Lower rail section -7.42 -4.86 2.56 00.00 00.00

..I 2 Upper rail section -2.30 -9.97 -7.66 00.00 00.00
3 Cross member -8.92 -3.35 5.57 00.00 00.00
(b) ANOVA Table
COLUMN FACTOItS DOF SUM OF SQUARES VARIANCE F
I Lower rail section I 6.554 6.554 6.81

2 UpPer rail section 1 58.676 58.676 60.96
3 Cross member 1 31.025 31.025 32.23
All others/error o
Total: 3 96.250 100.00
(c) Estimate of.&&: Optimum Condition of DesignlProcess:
LEVEL CONTRIJJUTION
Lower rail section Proposed design 2 1.2800

Upper rail sectioe Open section 1 3.8300
Cross member Reinforced design 2 2.7850
Contribution from all (actors (total) 7.8950
Current grand a~wage
of performance -6.1350
This estimate includes only those variables that have a significant contribution. i.e.• pooled variables are ~luded
from the estimate. &nm.ates may also be made with variables of clroice.
Example , ..,
Front Strnctu.re Cl'1Ish Characteristics
The Taguchi design of experiment methodology was used to oprimize
the design of the basic load carrying members of an automobile front
structure.iTbe development engineers were interested in determining the
best combination of designs. with three factors. each of which had two
.l:te;nanves. The 'Jerformance ·)f the 5tructure was measured in terms of
'f
,
!
I
Examples of Taguchi Case Studies lO9
I
the deformation under a drop silo test. For the test variability, three I
samples at each configuration were tested.
I
Solution
The factor descriptions and the analyses are shown in Tables 9-9-1 and
9-9-2. The design and the analysis is straightforward. The analysis utilizes
a SIN ratio with "nominal is best" quality characteristic.
References
1. Patrict M. Burgman. "Design of Experiments-The Tapcbi Way.» Manufacturing

Engil'leering. May 1985. pp. 44-46.
2. YuiD Wu and W'lllie Hobbs Moore. 1986. Quality Engmeering-Product and Process
Optimization. Dearbom, Michigan: American Supplier Institute.
"3. Rou.Id L. Iman and W. J. Conover. 1983. A Modem Approach to Statistics. New
York John WUey &. Sons•
....4. YuiDWu. 1986. OrthogonalAmrysandLinearGraphs. Dearbom, Michigan: Amer-
I icaa Supplier Institute.

5. Burton Gunter. 1987. A Perspective on the Taguehi Methods. Quality Progress.
" AnIaiean Society for Quality Control, Milwaukee, WI.
6. Lawtenee P. SuUivan. June 1987. The Power of the Taguchi Methods. Quality
Pmpess. American Society for Quality Control. Milwaukee, WI.
7..PbiIip J. Ross. 1988. Taguchi Techniques for Quality Engineering. New York:
Mc<bw Hill Book Company.
8. Thomas B. Baker and Don P. Causing. Quality Engineering by Design-The Taguchi
Method. 40thlnnual ASQC Conference, March 1984.
9. limQuinlan.lt85. Product Improvement By Application ofTaguchi Methods. Mid-
vale. 000: PIft Products, Inc.
10. Oeaiehi Taguchi. 1987. System ofExperimental Design. New York: UNIPUB, Kraus
Imauational Publications.
It. Ranjit K. Roy; QUALITEK-3:ffiM PC Software for Design and Analysis of Taguchi
Experiments. NUTEK Inc. 30400 Telegraph, Suite 380, Birmingham, MI 48010.
(PIl. 313-642-45(0).
12. R. A. Ftsher. 1951. Design of Experiments. Edinburgh: Oliver &. Boyd.
13. W. G. Cochran and G. M. Cox. 1957. Experimental Designs. Second edition. New
York: John Wiley &. Sons.
14. Diane M. Byme and Shin Taguchi. 1981. "The Taguchi Approach to Parameter
Design.» 40th Annual Quality Congress Transactions. MUwaukee, Wisconsin: Amer-
ican Society for Quality.Control.
15. G. Taguchi and S. Konishi. 1981. Orthogonal Arrays and Linear Graphs-Tools
for Quality Engineering. Dearbom, Michigan: American Supplier Institute, Inc.
16. 1biId Supplier Symposium on Taguchi Method. American Supplier Institute, Inc.
Dearbom, Michigan. October 8, 1985.
J I
1
I
l,
- - --'-- --~ --_.- ..--
- .. -
APPENDIX A
.•!'*
Orthogonal arrays, triangular tables and linear graphs

I
Table A-I. Common Orthogonal Arrays·· _

»
NUMBER OF NUMBER OF
ARRAY FACTORS LEVELS
~(t~) 3 2
7
4(2 ) 7 2
Ld2") 11 2
L I6(2 15 ) 15 2
Ld231 ) 31 2
~(3-') 4 3
1 7
*L 1S(2 .3 ) I 2
and 7 3
L27 (3 13) 13 3
L I6(4s) 5 4
*L3~21.49) 1 2
aod9 4
~(4:!1) 21 4
*Mixed level a1TaYS

**Orthogonal a1TaYS from G. Tagucbi and S. Konisbi. Onhog()nal Arrays and Linear Graphs. DeaIbom. Michigan:
American Supplier Institute. Inc. 1987.
2U A Primer on the Taguchi Method
Table A-2. Orthogonal Arrays L 4 and La (2-Level)*

(a)
L..(23)
I 2 3
I I I I
2 I 2 2
3 2 I 2
4 2 2 I
(b)
I Ls(27 )
1
2 3 4 S 6 7
I I 1 I 1 I 1 1
2 1 I 1 2 2 2 2
3 I 2 2 1 I 2 2
4 I 2 2 2 2 I I
5 2 I 2 I 2 I 2
6 2 I 2 2 I 2 1
7 2 2 I I 2 2 I
8 2 2 I 2 I I 2
·Reprinted with pennissWn of the American Supplier Institute. Inc.
-------__e.
J
3
1•• 2
Figure A-I. Linear graph for L 4 •
1
2
4!E--------e4
2411--------1114 7
6
ial lbl
Figure .-\.~. Linear graphs for L~.

--'
Appendix A 213
Table A-3. Orthogonal Arrays L 12 and L a6 (2 Level)·

LI212". -
NO. 2 3 4 5 6 7 8 9 10 It
I I I I I I I I I I I 1
2 I I I I I 2 2 2 2 2 2
3 I I 2 2 2 I I I 2 2 2
4 I 2 I 2 2 I 2 2 I I
2
5 I 2 2 I 2 2 I 2 I 2 I
6 I 2 2 2 I 2 2 I 2 I I
7 2 I 2 2 I I 2 2 I 2 I
8 2 I 2 I 2 2 2 I I I 2
9 2 I I 2 2 2 I 2 2 I I
10 2 2 2 I I 1 Z
I 2 I 2
II 2 2 I 2 I 2 I I 1 2 2
12 2 2 -I I 2 I I 2 2 2 I
The l..u(2 11 ) is a ~pecially designed Ilmt). in that interactions are disributed trn)re er k$s unifonnly to allcofumns.
There is 110 linear graph fer this attay. If $hould not be used to anai}B interactions. The advantage of this design
is its ~apability to investigate 11 main effects. making it a highly recommended array.
(5
l 11 )
NO. 2 3 4 5 6 7 8 9 10 U 12 13 14 15
I I I 1 I I I I J 1 I I I I I
2 I I I I I 1 2 -2 -2 2 2 2 2 2
3 I I 1 2 2 2 2 I I I 1 2 2 2 2
4 I I 1 2 ...
., -2 2 2 2 2 2 1 1 I I
5 I 2 2 I 2 2 I 1 2 2 I I 2 2
6 I 2 2 I 2 2 2 2 I I 2 2 I I
7 1 2 2 2 2 I I 1 2 2 2 2 I I
8 I -2 2 2 2 I 2 2 I I I I 2 2
9 2 1 2 2 2 1 2 I 2 I 2 I 2
10 2 I 2 ")
2 2 I 2 I 2 I 2 I
II 2 I 2 2 I 2 I 2 I 2 2 I 2 I
t2 2 I 2 2 I 2 2 I 2 I I 2 I 2
1 ...
., ....,
., - 2
"\
13 2 2 I I 2
14 2 2 I - ... ")
1 2 2 I
.,
., ...
15 -...
")
-... -... - ..,
"I "I
-
")
2
...
")
16 - - - - -
"I
-
"I ")
-
'~::'i'r.t~j "·l:1 X'·-"!,:,';\..'C .'/ 'r':..: ....r~'·- __ - ~ _~':-1~' ;r,: r.;'~ -.
1
L4 ,. 3
-2
La
1
2 4 2
6 15
2 L,a
La
t 1 4
7 9
15 8
10 7
L,a
12 10 L'6
2.-------------..4
Figure A-J. Linear graphs for 2 level orthogonal arrays.
: as it:: Jl.& ! ; ; :, -
.
Table A-4. Orthoaronal Ar....vs L 32 (2 LtveI, 31 Factors)*
"COLUMN I L,2 (231 )
~
('ONI>ITION , I 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2J 22 23 24 25 26 27 28 29 30 31
I I I I I 1 I I 1 1 J 1 I J 1 I 1 I I I J I I I 1 1 I I J 1
2 I I I 1 I I 1 I I 1 I I 1 I I 222 2 2 2 2 2 2 2 2 2 2 222
.3 I I 1 I I 1 I 2 2 2 2 2 222 I I 1 I I I J I 2 222 2 2 2 2
4 I J J J J J J 2 222 2 2 2 222 2 2 2 222 1 I J J J J J J
,,
--," 5 I 1 I 2 2 2 2 I 1 I I 222 2 1 I J I 2 222 1 I I 1 2 2 2 2
6 I I J 2 2 2 2 I I 1 I 2 2 2 2 2 222 I I I J 2 2 2 2 J I I I
7 1 1 I 2 2 2 2 2 2 2 2 I I I I 1 I I I 222 2 2 222 1 I I J
H 1 1 2 2 2 2 2 2 2 2 I 1 1 I 222 2 I 1 I I 1 I I I 2 2 2 2
I) 2 2 1 1 2 2 I 1 2 2 I 1 2 2 1 I 2 2 1 1 2 2 I 1 2 2 1 I 2 2
10 2 2 1 1 2 2 1 I 2 2 J 1 2 2 22 I I 2 2 1 I 2 2 1 I 2 2 1 I
11 2 2 1 I 2 2 2 2 1 I 2 2 I I 1 I 2 2 1 1 2 222 I I 2 2 I I
12 2 2 I 1 2 222 I I 2 2 1 I 2 2 I I 2 2 I I I 1 2 2 1 I 2 2
13 J 2 2 2 2 I I I J 222 2 I 1 I I 222 2 1 I I J 2 2 2 :2 I I
14 I 2 2 2 2 I J J I 222 2 1 I 2 2 1 I I I 222 2 I I I I 2 2
15 1 2 2 2 2 I I 2 2 I I I I 2 2 I I 222 2 I 1 2 2 I I I I 2 2
(rolftinutd)
........ ,"'. > ~ ~.~ ');
~- , Mlr t4f~ - . . - . , ._ _ ... _~ $"""" 1

Table A-4. OrthOS!onal Arravs L:n (2 Level, 31 Factors) (Continued)
31
"COLUMN I L
; 32,(2 )
'~~:~,*
I 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 3\
16 I 2 2 2 2 I 2 2 1 I
I I 1 222 2. I I 1 1 2 2 1 1 2 2 2 2 I 1
17- 2 I 2,(
I 2 I 2 I 2 1 2 I 2 1 2 I 2 1 2 I 2 I 2 1 2 I 2 I 2 \ 2
I~ 2 I 2 I 2 12 I 2 1 2 I 2 I 2 2 I 2 1 2 I 2 I 2' I 2 I 2 I 2 1
19 2 I 2 I 2 1 2 2 I 2 I 2 1 2 I 1 2 1 2 1 2 1 2 1 2 2 1 2 I 2 \
20 2 I 2 1 2 I 2 2 I 2 I 2 I 2 I 2 I 2 1 2 I 2 I I 2 I 2 I 2 I 2
,
-"
-. 21 2 1 2 2 I 2 I \ 2 I 2 2 I 2 I I 2 1 2 2 1 2 1 I 2 1 2 2 I 2 \
22 2 I 2 2 I 2 I 1 2 I 2 2 1 2 I 2 1 2 I I 2 1 2 2 I 2 I \ 2 I 2
23 2 1 2 2 I 2 I 2 I 2 I 1 2 I 2 I 2 1 2 2 1 2 1 2 1 2 I I 2 I 2
24 2 I 2 2 I 2 I 2 I 2 I 1 2 I 2 2 I 2 I I 2 I 2 I 2 1 2 2 I 2 I
25 2 2 I I 2 2 1 I 2 2 1 I 2 2 I 1 2 2 1 1 2 2 I 1 2 2 1 I 2 2 I
26 2 2 I 1 2 2 1 I 2 2 I I 2 2 I 2 I I 2 2 I I 2 2 1 1 2 2 I I 2
27 2 2 I I 2 2 I 2 I 1 2 2 I I 2 I 2 2 I 1 2 2 I 2 I I 2 2 I 1 I
28 2 2 I I 2 2 I I 2 2 1 I 2 2 I I 2 2 1 I 2 2 I I 2 2 I I 2 2 I
29 2 2 I 2 I 1 2 I 2 2 I 2 1 I 2 I 2 2 I 2 1 I 2 I 2 2 I 2 I I 2
30 2 2 1 2 J I 2 1 2 2 I 2 1 I 2 2 I I 2 I 2 2 I 2 I I 2 I 2 2 I
31 2 2 I 2 I I '2 2 I I 2 I 2 2 I I 2 2 I 2 1 I 2 2 I I 2 I 2 2 I
32 2 2 I 2 I 1 2 2 I 1 2 I 2 2 I 2 I I 2 1 2 '2 I I 2 2 1 2 I I 2
...Reprinted with pennission of the American Supplier Institute. lne,

I
--------
$ J 2 .- u £ t .,
\
I
I
Table A-!·l. Ortboaonal Array! 4. (2 Level)·
I
I
63
44 (2 )
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
No. I
.
1 1 I I 1 1 1 1 I 1 1 1 1 1 I 1 1 1 1 1 1 1 t 1 1 1 1 1 1 1 1 1
2 1 1 1 t t 1 1 t 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 t 1 t 1 1 1 1 1
3 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 222 2 2 222
4 1 I 1 1 1 1 1 1 1 1 1 1 1 1 1 222 2 222 2 2 2 2 2 2 2 2 2
5 1 1 1
1 1 1 1 222 2 2 222 1 1 t 1 1 1 1 t 222 2 2 2 2 2
6 1 1 I 1 1 I 1 222 2 2 222 1 t 1 1 1 1 1 1 222 2 2 2 2 2
7 1 1 1 1 1 I t 2 2 2 2 2 2 2 2 2 2 2 2 2 222 I t t 1 1 1 1 I
K 1 1 I 1 I 1 I 2 2 2 2 2 222 2 2 2 2 2 2 2 2 I 1 1 1 1 I 1 I
9 I I 1 2 2 2 2 1 1 1 I 2 2 2 2 1 I I I 222 2 1 I 1 1 2 2 2 2
, >
-; 10 I 1 1 2 2 2 2 t 1 t 1 2 2 2 2 t 1 1 I 2 222 1 1 1 t 222 2
I1 1 1 I 2 2 2 2 t 1 1 t 2 2 2 2 2 2 2 2 1 1 1 1 222 2 1 1 1 1
12 I I I 2 222 1 1 1 1 2 2 2 2 2 222 I I t 1 2 2 2 2 1 1 1 1
13 I 1 1 222 2 2 2 2 2 1 .I 1 1 1 1 I 1 222 2 222 2 1 1 I 1
14 I I 1 Q 2 2 222 2 2 1 I I I 1 I I t 2 2 2 2 2 2,. 2 2 1 1 t I
15 1 1 1 2 222 2 2 2 2 1 1 1 1 2 22 2 1 1 1 1 1 1 1 1 2 2 r 2
16 1 1 1 2 2 2 2 2 2 2 2 1 1 1 1 2 2 2 2 I 1 1 1 1 1 1 1 2 2 2 2
17 I 2 2 I 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2
18 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2
19 1 2 2 1 1 2 2 1 1 7 2 1 1 2 2 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 1
20 1 2 2 1 1 2 2 1 1 221 1 2 2 2 2 t 1 2 2 1 t 2 2 1 t 2 2 1 I
21 1 2 2 1 I 2 2 2 2 I 1 2 2 1 t t 1 2 2 1 t 2 2 2 2 1 1. 2 2 1 1
. 22 1 2 2 1 1 222 2 1 1 2 2 I I I I 2 2 I I 2 2 2 2 I I 2 2 I I
2:' I 2 2 I I 2 2 2 2 I 1 2 2 1 1 2 2 I 1 2 2 I I I I 2 2 I I 2 2
(contimu!d)
_.", .
- ........ ~,~ ~~,,/ ~, le _1 . [ I g $ U P t I &t _.. _-..
Table A-S-l. Orthogonal ArraysL... (2 Le!ttI)JColllinued)
,
W1~
"tL ,("\63)
.... .,'
No. I 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
24 122 1 122 22112211 2 2 1 122 I 1 I J221122
25 I 2 22 2 1 1 11222211 1 ° 1 22221 I I 1222211
26 1222211 11222211 l 1 2 2 221 I I 1222211
27 122221'1 11222211 2 2 I 1 1 1 2 2 2 2111122
28 12222 I 1 11222211 2 2 1 1 11 2 2 2 2111122
.. 29 1222211 221 1 1 122 1 1 22221 1 2 211 1 122
00
30 1222211 22111122 1 1 22221 1 2 211 1 122
31 1222211 22111122 2 2 1 1 1 1 2 2 . 1 1222211
32 1222211 221 1 1 122 2 2 1 1 1 J 2 2 1 1222211
33 2 1 2 1 212 121 2 1 2 1 2 1 2 1 2 1 2 1 2 1 212 1 212
34 221 212 1 2 1 2 121 2 1 2 1 21212 1 2 1212112
35 2 1 212 1 2 I 212 1 212 2 1 2 J 2 1 2 1 2 1 2 1 2 1 2 1
36 2 I 2 121 2 1 2 I 2 I 2 1 2 2 1 2 1 2 1 2 1 2 1212121
37 2 1 2 I 212 2 1 2 121 2 I 1 2 1 2 1 ,2 1 2 2 121 2 121
38 2 1 2 1 2 1 2 212 1 2 121 I 2 1 2 1 2 2 1 2 1212121
39 2 1 2 1 2 1 2 2 1 2 I 2 121 2 J 2 I 2 1 2 1 1 2 I 2 1 212
40 2 1 2 1 212 212 1 2 1 2 1 2 I 2 I 2 I 2 1 1 212 1 212
41 2 1 221 2 1 1212212 1 1 2 12212 1 1 2 1 221 2 1
42 2 1 2 2 121 1 2 1 2 2 121 1 2 J 2 2 1 2 1 1 2 1 2 2 1 2 1
43 2 1 2 2 1 2 1 1 2 122 121 2 1 2 1 I 2 I 2 2 1211212
. ~"""""._ +o,.""."!..,,_.~',,,,_~_W'" ._"_.~;,- ., _ .:.1< "" $ JX .... """'*""" 4 .. "
-------_...--
.'Ill I " • • • Bb
I
44 2 J 2 2 J 2 J J 2 J 2 2 J 2 I 2 J 2 I I 2 I 2 2 J 2 J I 2 J 2
45 2 I 2 2 1 2 1 2 1 2 1 1 2 1 2 I 2 1 2 2 1 2 1 2 1 2 1 1 2 1 2
46 2 1 2 2 J 2 J 2 1 2 t 1 2 1 2
2
1 2
2 1
1 2 2
2 1 1
. J 2
·2 J
t 2 1
2 1 2
2 1 1 2
1 2 2 1
1
2
2
1
47 2 I 2 2 I 2 I 2 I 2 t I 2 1
48 2 1 2 2 1 2 t 2 1 2 t 1 2 1 2 2 1 2 1 1 2 1 2 1 2 1 2 2 1 2 I
49 2 2 I I 2 2 1 1 2 2 1 1 2 2 1 1 2 2 1 t 2 2 1 t 2 2 t t 2 2 1
50 2 2 t 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 I 1 2 2 1 1 2 2 t I 2 2 I
5\ 2 2 1 1 2 2 1 t 2 2 t t 2 2 t 2 t t .2 2 1 t 2 2 t 1 2 2 1 1 2
52 2 2 1 1 2 2 t 1 2 2 1 1 2 2 1 2 1 1 2 2 t I 2 2 1 1 2 2 J t 2
53 2 2 t t 2 2 1 2 1 1 2 2 1 1 2 t 2 2 1 1 2 2 I 2 1 1 2 2 1 1 2
54 2 2 1 1 2 2 t 2 t I 2 2 1 1 2 1 2 2 1 I 2 2 I 2 t 1 2 2 1 1 2
55 2 2 1 1 2 2 1 2 1 I 2 2 I t 2 2 t t 2 2 1 1 2 t 2 2 1 1 2 2 1
~6 2 2 J I 2 2 J 2 J I 2 2 1 I 2 2 I I 2 2 1 J 2 1 2 2 1 1 2 2 t
2 2 I '2 I 1 2 1 2 2 I 2 I I 2 I 2 2 I 2 I 1 2 1 2 2 I 2 1 I 2
,.
57
.-
~
58 2 2 I 2 J 1 2 1 2 2 t 2 t 1 2 1 2 2 1 2 1 I 2 I 2 2 J 2 J I 2
59 2 2 J 2 1 1 2 1 2 2 t 2 I I 2 2 I I 2 I 2 2 t 2 1 1 2 I 2 2 t
60 2 2 J 2 I 1 2 1 2 2 I 2 I 1 2 2 1 I 2 J 2 2 I 2 1 I 2 1 2 2 I
61 2 2 1 2 I I 2 2 1 1 2 I 2 2 I 1 2 2 J 2 t 1 2 2 1 1 2 I 2 2 1
62 2 2 I 2 I 1 2 2 1 1 2 I 2 2 I I 2 2 I 2 I J 2 2 1 t 2 J 2 2 1
63 2 2 J 2 I 1 2 2 1 I 2 t 2 2 I 2 1 1 2 I 2 2 I I 2 2 J 2 J I 2
M 2 2 1 2 1 I 2 2 J t 2 I 2 2 I 2 I I 2 1 2 2 J 1 2 2 1 2 I J 2
"'Reprinted with permission (If the American Supplier Institute. Inc.
• 4
Table A-S-2. Orthogooal Array L 64 (Contioues Table A-5-1) (2 Level)
I
32 H 34 35 36 37 38 39 40 4\ 42 43 44 45 46 47 4lJ 49 so 51 52 53 54 ss 56 57 ss 59 60 61 62 63
~ k ):tW-" ,S __' .. ' -,.-o~- _tiWSXi;:mh.___ ~ ___ ",_" N-'-_,· __ -~.tk.M.Jm:.o ,_ ,lZ.U:e:u,;;@ J 4:4, ~ .. AJdlik.:Q&. ..W0",-" -.",)$"
" g ."S),t),-,·
$
I I I I I I I I I I I I I I I 1 1 I I I I I I I I I I I I I I I
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2. 2 2 2 2 2
I I I I I I I I I I I 2 2 2 2 2 2 2 222 2 2 2 2 2 2 2 2 2 2 2
22222 2 2 2 2 2 2 I I I I I I 1 1 I I I I I I I I I I I I I
I I I I I I I I 22222222'111*" I 1 I I I 2 2 2 2 2 2 2 2
2 2 2 2 2 222 I I I I I I I I 2 2 2 2 2' 2 2 2 I I 1 1 I I I I
I I I I I I I I 2 2 2 2 2 2 2 2 I 1 I 1 1 I I 1 2 2 2 2 2 2 2 2
2 2 2 2 2 2· 2 2 I I I I I I I I I I 1 I I 1 1 I 2 2 2 2 2 2 2 2
I I I I 2 2 2 2 I I I I 2 222 I I I I 2 2 2 2 I I 1 I 2 2 2 2
2 2 2 2 I I I I 2 222 I I I I 2 2 2 2 I I I I 2 2 2 2 I I 1 1
I I 1 I 2 2 2 2 I I 1 I 2 2 2 2 2 2 2 2 1 1 1 I 222 2 I I 1 I
2 2 2 2 I I I I 2 2 2 2 I I I I I I I I 2 2 2 2 I I I I 222 2
I I 1 I 222 2 2 2 2 2 I I I I I I 1 1 2 2 2 2 2 2 2 2 1 1 1 I
.,
~
2 222 1 I I I I I I I 222 2 2 222 1 I 1 I I I I I 222 2
I I I I 222 2 2 2 2 2 I I I I 2 2 2 2 I I I I I I I I 2 2 2 2
2 2 2 2 I I I I I I I I 2 2 2 2 I I I I 22222 2 2 2 I I I I
I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 1 I 2 2 I I 2 2 I I 2 2
2 2 I 1 2 2 I I 2 2 I I 2 2 1 I 2 2 1 I 2 2 I I 2 2 I 1 2 2 I I
I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2
2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 1 I 2 2 1 I 2 2 I I 2 2 I I
I I 2 2 I I 222 2 I I 2 2 I I I I 2 2 I 1 2 2 22 I I 2 2 I I
2 2 1 1 2 2 I I I I 2 2 I I 2 222 I I 2 2 I I I I 2 2 I I 2 2
I I 2 2 I I 222 2 I I 2 2 I 1 2 2 I I 2 2 1 I I I 2 2 I I 2 2
2 2 1 I 2 2 I 1 I I 2 2 1 I 2 2 I I 2 2 I I 222 2 1 1 2 2 1 1
I I 2 2 2 2 I I 1 I 2 2 2 2 1 I I 1 2 2 2 2 I 1 I I 2 2 22 I 1
2 2 I I 1 1 222 2 1 1 I I 222 2 I I 1 I 2 2 2 2 1· I I I 2 2
I I 222 2 I 1 I I 2 2 2 2 I I 2 2 I I I I 2 2 2 2 I I I I 2 2
2 2 I I I I 2 2 2 2 I I I I 2 2 I I 2 2 2 2 I 1 I 1 222 2 1 I
I I 2 2 2 2 I I 2 2 I I I J 2 2 I I 2 2 2 2 1 1 2 2 1. 1 I 1 2 2
2 2 I I 1 I 2 2 1 I 2 2 2 2 1 1 2 2 I 1 1 I 2 2 I I 2 2 2 2 I I
I I 2 2 2 2 I I 2 2 I I I I 2 2 2 2 I I I I 2 2 I 1 2 2 2 2 I I
_ _.
~_~ ~ _ ... 4"""".
: : e 21t wc:sit SJ 2 a ~
~~ .... ;;:: I .a2 2 ,U".,SV";.2J. t
- £
.....i ; 1
.- ."
$ $
,
2 2 I I 2 2
I I I I 2 2 2 2 I I I I 2 2 2 2 I 2 2 I I I
I I 2 2
I 2 I 2 2 I 2
I I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2
2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 1 2 I 2 I 2 I 2 1
I 2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 2 I 2 I 2 I 2 I 2 I 2 I 2 I2 I
2 I 2 I 2 I 2 I 2 I 2 I 2 I 2 I I 2 I 2 1 2 I 2 1 2 I 2 1 2 I 2
I 2 1 2 I 2 1 2 2 I 2 1 2 I 2 1 I 2 I 2 1 2 1 2 2 1 2 I 2 1 2 1
2 I 2 I 2 I 2 1 I 2 1 2 I 2 1 2 2 1 2 1 2 1 2 1 1 2 1 2 1 2 I 1
I 2 I 2 I 2 I 2 2 I 2 I 2 I 2 I 2 1 2 I 2 I 2 I I 2 1 2 I 2 I 2
2 I 2 I 2 I 2 1 I 2 I 2 I 2 1 2 I 2 1 2 1 2 1 2 2 1 2 1 2 1 2 1
I 2 I 2 2 1 2 I 1 2 I 2 2 1 2 I 1 2 I 2 2 I 2 1 1 2 1 2 2 1 2 1
2 I 2 I 1 2 I 2 2 I 2 1 1 2 I 2 2 I 2 I I 2 I 2 2 I 2 I I 2 1 2
I 2 I 2 2 I 2 I I 2 1 2 2 I 2 I 2 I 2 I I 2 I 2 2 1 2 1 1 2 I 2
2 I 2 I 1 2 I 2 2 I 2 1 I 2 I 2 I 2 I 2 2 I 2 I I 2 1 2 2 I 2 I
I 2 I 2 2 I 2 I 2 I 2 I I 2 I 2 I 2 I 2 2 I 2 I 2 1 2 I 1 2 I 2
2 I 2 I 2 I 2 I 2 1 2 2 I 2 I 2 I 2 I I I 2 I 2 I 2 1 2 2 I 2 I
I 2 I 2 2 1 2 1 2 1 2 I 1 2 I 2 2 I 2 1 I 2 I 2 1 2 I 2 2 I 2 I
2 I 2 1 I 2 I 2 I 2 1 2 2 I 2 1 I 2 I 2 2 1 2 I 2 1 2 I I 2 1 2
..
.. I 2 2 1 I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I
2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 I I 2 2 1'1 I 2
I 2 2 I I 2 2 I I 2 2 I I 2 2 I 2 I I 2 2 I I 2 2 1 I 2 2 I I 2
2 I I 2 2 I I 2 2 I I 2 2 I I 2 I 2 2 I I 2 2 I I 2 2 I I 2 2 I
I 2 2 I I 2 2 I 2 I I 2 2 I I 2 I 2 2 I I 2 2 I 2 I I 2 2 I I 2
2 I I 2 2 I I 2 I 2 2 I I 2 2 I 2 I I 2 2 1 I 2 I 2 2 I I 2 2 I
I 2 2 I 1 2 2 I 2 1 I 2 2 I I 2 2 I I 2 2 I I 2 I 2 2 I I 2 2 I
2 I I 2 2 I I 2 I 2 2 I I 2 2 I I 2 2 I I 2 2 I 2 I 1 2 2 I I 2
I 2 2 I 2 I I 2 I 2 2 I 2 1 I 2 I 2 2 I 2 I I 2 I 2 2 I 2 I I 2
2 I I 2 I 2 2 I 2 I I 2 I 2 2 I 2 I I 2 I 2 2 I 2 I I 2 1 2 2 I
I 2 2 I 2 I I 2 I 2 2 I 2 I 1 2 2 I I 2 1 2 2 I 2 I I 2 I 2 2 I
2 I I 2 I 2 2 I 2 I I 2 I 2 2 I I 2 2 I 2 I I 2 1 2 2 I 2 I I 2
J 2 2 I 2 I I 2 2 1 I 2 I 2 2 I I 2 2 I 2 I I 2 2 I 1 2 I 2 ' 2 1
2 I I 2 I 2 2 I I 2 2 I 2 I I 2 2 I I 2 I 2 2 I I 2 2 I 2 I I 2
I 2 2 I 2 I I 2 2 I I 2 I 2 2 I 2 I I 2 I 2 2 I I 2 2 I 2 I I 2
2 I I' 2 I 2 2 I I 2 2 I 2 I I 2 I 2 2 I 2 I I 2 2 I I 2 I 2 2 I
·Rcprilltcd with penni1\I\ltm of the American Supplier Institute. 1nl.',
.
\ '.j
•
Table A-6 TrlMlllar Tabl.fQf 2 IAVeI 011hOSOlla. Atra)'"
.
,,.;,
, 1 2 :I 4 .5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 2S 26 27 28 2~ )0 31
(I) :I 2 .5 4 7 6 4.) 8 11 10 13 12 15 14 17 16 19 18 21 20 23 22 2S 24 27 26 29 2K 31 30
(2) I 6 7 4 5 10 1I 8 9 14 15 12 13 18 19 16 17 22 23 20 21 26 27 24 25 30 31 28 29
(3) 7 6 5 4 11 10 9 8 15 14 13 12 19 18 17 16 23 22 21 2027 26 25 24 31 30 29 28
(4) I 2 3 12 13 14 15 8 4.) 10 11 20 21 22 23 16 11 18 19 28 29 30 31 24 2.'i 26 27
(5) 3. 1 2 13
(6) I 14
12
15
IS
12
14
13
9
10
8
11
11
8
10
9
21
22
20
23
23
20
22 17
21 18
16- 19
19 16
18 29
17 30
28
31
.
31
28
30
29
25
26
24 27
27 24
26
25
(7) 15 14 13 12 11 10 9 8 23 22 21 20 19 18 17 16 31 30 29 28 27 26 2S 24
(8) I 2 3 4 5 6 7 24 25 26 27 28 29 30 31 16 17 18 19 20 21 22 23
(9) 3 2 5 4 7 6 25 24 27 26 29 28 31 30 17 16 19 18 21 20 23 22
(10) I 8 7 4 S 26 27 24 25 30 31 28 29 18 19 16 17 22 23 20 21
(11) 7 6 .5 4 27 26 25 24 31 30 29 28 19 18 17 16 23 22 21 20
(12) I 2 3 28 29 30 31 24 25 26 27 20 21 22 23 16 17 18 19
(13) 3 2 29 28 31 30 25 24 27 26 21 20 23 22 17 16 19 18
~~ Al Si! _.~ -
- - _ ''I4iJliLL~W';A"J4it_~"Z2V!UJ!it;;U.JJ al ilL 2 _ 12; i I IQM 2 2 2 22.1." PI ~.M"l
(14)
(IS)
I 30
31
31
JO
28
29
29
28
26
27
27
26
24
25
25 22
24 23
23
22
20
21
21
20
18
19
19
18
16 17
17 16
I
(16) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
(7) 3 2 5 4 7 6 9 8 11 10 13 12 15 14
(l8) 1 6 7 4 .5 10 11 8 9 14 15 12 13
(19) 7 6 5 4 1I 10 9 8 15 14 13 12
(20) I 2 3 12 13 14 15 8 9 10 I I
(21) 3 2 13 12 15 14 9 8 II 10
(22) I 14 15 12 13 10 II tI 9
(23) 15 14 13 12 11 10 9 8
(24) I 2 3 4 5 6 7
(25) 3 2 S 4 7 6
(26) I 6 7 4 5
(27) 7 6 5 4
,,
(28) I 2 3
(29) 3 2
(J()) I
+lh'llIllItnl with pcnllih~iun u(thc American Supplier Institute, Inc,
--
t
" \
224 A Primer on the Taguchi Melhod
Table A-7. Orthogonal Arrays (3

Level, L., and L ls )*
(a)
2 .l 4
I I I I I
2 1 2 2 2
3 I 3 3 3.
4 2 I 2 3
5 2 2 3 1
6 2 3 1 2
7 3 I 3 2
8 3 2 I 3
9 3 3 '2 I
(b)
Ll~21 X )',
NO, 2 ) 4 S 6 7 8
I 1 I 1 1 I I
2 2 2 2 2 2 2
3 3 3 3 3 3 3
4 I 2 1 I 2 2 3 3
5 I 2 2 2 3 3 I 1
6 1 2 3 3 1 1 2 2
7 1 3 I ')
I 3 2 3
8 1 3 2 3 2 1 3 1
9 I 3 3 I 3 2 1 2
10 2 I 1 3 3 2 2 1
11 2 1 2 I I 3 3 2
12 2 1 3 2 2 1 I 3
13 2 2 I
., 3 I 3 2
14 2 2 2 3 I 2 1 3
15 2 2 3 1 2 3 2 I
16 2 3 1 3 2 3 1 2
17 2 3 2 I 3 1 2 3
18 2 3 3 'l
1 2 3 1
i
Note: Like the LI~ ell\. this isa speciall~ designed array. An intaaCuon is builit in between the first two columns.
This interaction information can be obtain.ed without sacnficing an~ other column. Interactions between three·
!e\'el columns are distributed more or le~s uniformly 10 .111 the ~ three-level columns. which permits investigation
)f l1alll effects. Thus. :t ;S l !lighl~ ~ecotr.mended arra.. for ~~[)l;:n~nls
'l"t~crT'eJ ~ 'll1 X,TI'-;;rf _\( "r.r: ·\;r.C~(JJ': SU~Ct~e" ·.n~l~('Jte r:<:
- _.__ .-.
.,
r-- .
_/ I
•
I
Appendix A 225
Table A-8. Orthogonal Arrays (3 Level, Lr,)*
2 3 4 S 6
L27 1313 ,
7 8 9 10 11 12 13
Il
1
2
3
I
I
1
1
1
1
1
1
1
1
1
1
I
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
I
2
3
1
2
3
IJ;
'*,
4 I 2 2 2 1 1 1 2 2 2 3 3 3
;..
5 I 2 2 2 2 2 2 3 3 3 I I 1
6 1 2 2 2 3 3 3 I 1 I 2 2 2
7 I 3 3 3 I 1 I 3 3 3 2 2 2
8 I 3 3 3 2 2 2 I I I 3 3 3
9 I 3 3 3 3 3 3 2 2 2 1 I I
2 1 2 3 I 2 3 1 2 3 I
10
1I 2 1 2 3 2 3 I 2 3 I 2
-32 3
1
12 2 I 2 3 3 1 2 3 I 2 3 I 2
13 2 2 3 I I 2 3 2 3 I 3 I 2
14 2 2 3 I 2 3 I 3 I 2 I 2 3
15 2 2 3 I 3 I 2 I 2 3 2 3 I
16 2 3 I 2 I 2 3 3 I 2 2 3 1
17 2 3 I 2 2 3 I I 2 3 3 1 2
18 2 3 I 2 3 1 2 2 3 1 I 2 3
19 3 I 3 2 I 3 2 1 3 2 I 3 2
20 3 I 3 2 2 I 3 2 I 3 2 I 3
21 3 I 3 2 3 2 1 3 2 1 3 2 I
22 3 .2 1 3 I 3 2 2 1 3 3 2 I
23 3 2 1 3 2 I 3 3 2 I I 3 2
24 3 2 I 3 3 2 I I 3 2 2 I 3
25 3 3 2 I I 3 2 3 2 1 2 I 3
26 3 3 2 I 2 I 3 1. 3 2 3 2 1
27 3 3 2 I j 2 I 2 I 3 I 3 2
*Reprinled .... ith pennissioe of the American Supplier Institute, Inc.
It
I
, I
I I
Appendix. A 227
~I
Table A-IO. Ortbogonal Arrays (2 Level" 4 Level)·
L32 12 1 X ill
No, I 2 3 4 S 6 7 I 9 10
I I 1 I I I I 1 I 1 I
1 1 1 2 2 2 2· 2 2 2 2
3 1 1 3 3 3 3' 3 3 3 3
4 1 1 4 4 4 4 4 4 4 4
5 1 2 1 1 2 2 3 3 4 4
6 1 2 2 2 1 1 4 4 3 3
7 1 2 3 3 4 4 1 I 2 2
8 1 2 4 4 3 3 2 2 1 1
9 1 3 1 2 3 4 1 1- 3 4
o. 3
10 1 3 2 I 4 3 2 I 4
11 1 3 3 4 I 2 3 4 1 2
12 1 3 4 3 2 I 4 .3 2 1
13 1 4 I 2 4 3 3 -4 2 1
14 1 4 2 I 3 4 4 3 1 2
15 1 4 3 4 2 1 1 2 4 3
16 1 4 4 3 I 2 . 2 I 3 4
17 2 1 I 4 1 4 2 3 2 3
18 2 I 2 3 2 3 1 4 1 4
19 2 I 3 2 3
., 4 I 4 I
20 2 1 4 I 4 I 3 2 3 2
21 2 2 1 4 2 3 4 I 3 2
22 2 2 2 3 1 4 3 2 4 1
23 2 2 3 2 4 I 2 J I 4
24 2 2 4 1 3 2 1 4 2 '3
25 2 3 1 3 3 I 2 4 4 2
26 2 3 2 4 4 2 1 3 3 1
27 2 3 3 1 1 3 4 2 2 4
28 2 3 4 2 2 4 3 I 1 3
29 2 4 1 3 4 2 4 2 1 3
30 2 4 2 4 3 1 3 I 2 4
31 2 4 3 1 2 4 2 4 3 1
32 2 4 4 2 1 3 1 3 4 2
*Reprimed with p.:rmissil'n of the AlTIerican Surrher Institute. Inc,
~,-<,,, ~i~"~
~
Table A-II. Triangular Table for 4 Level Orthogonal Arrays*
>
(1)
2
I 3 2
3
"
2
( 'W,\
~
2
,-$
6
7
7
6
M
6
9
6
IQ
11
11
10
12
10
,,-
10
14
15
1$
14
• /(1
14
,,,
14 19
III '11
18
~tl
19
JI
18
r
0
=
4 4 3 3 8 8 7 7 12 12 1I 11 16 16 15 15 20 20 19 19 f
5 5 5 4 9 9 9 9 13 13 13 12 • 17 17 17 16 21 21 21 20
I I I 10 11 12 13 6 7 8 9 6 7 8 9 6 7 8 9 ~
OQ
(2) I 4 3 3 14 15
19
19
20
17 14 IS 19 17 10 11
19
12 13 10 11 12 IJ _.
c
I')
:::s-
5 5 4 18 21 18 19 20 21 18 20 21 14 15 16 17
I
() I 2
5
I
2
4
I
11
16
21
12
10
17
20
13
13
14
19
10
12
15
18
1I
7
17
20
8
6
18
21
9
9
15
18
6
8
14
19
7
8
13
19
9
9
12
18
8
6
11
21
7
7
10
20
6
9
12
15
7
8
13
14
6
7
10
17
9
6
11
16
R
I
(4) I 2 17 16 15 14 15 14 17 16 11 10 13 12 13 12 11 10
3 19 18 21 20 21 20 19 18 20 21 18 19 16 17 14 15
13 12 11 10 9 8 7 6 7 6 9 8 8 9 6 7
(5) I 15 14 17 16 16 17 14 15 12 13 10 11 11 10 13 12
20 21 18 19 19 18 21 20 21 20 19 18 17 16 15 14
I I I 2 3 4 5 2 5 3 4 2 4 5 3
(6) I 8 7 7 14 16 17 15 10 13 11 12 10 12 13 12
9 9 8 18 21 19 20 18 20 .21 19 14 17 15 16
.1 I 3 2 5 4 5 2 4 3 4 2 3 5
(7) I 6 6 17 15 14 16 12 II 13 10 13 11 10 12
9 8 20 19 21 18 21 19 18 20 18 15 17 14
I 4 5 2 3 3 4 2 5 5 3 2 4
(8) I 6 15 17 15 14 13 10 12 1I 11 13 12 J()
7 21 18 20 19 19 21 20 18 17 14 16 15
5 4 3 2 4 3 5 2 3 5 4 2
.1 (9) I 16 14 15 17 IJ 12 10 13 . 12 10 11 D
19 20 18 19 20 18 19 21 15 16 14 17
I I 1 2 4 5 3 2 5 3 4
(10)1 12 lJ 11 6 8 9 7 6 9 7 8
13 12 18 21 19 20 14 15 18 15
."';':'~;,.,,,,,,~.
_ . M ~e",;...
.. iiJjt t t £ 4t £J 2 t 51 re• t J _ ' ....", '" ,A,'NV''''''' . :-
L
••
. t ~.
••• • _
$;; .; g :" • 'MO'" . M" J
"
l'
I I 4 Z 3 , :l Z 4 3
1
(11 ) 10 10 9 7 6 8 9 7 9 6
13 12 20 19 21 18 17 15 14 16
1 5 3 2 4 3 4 2 5
(12) 10 7 9 9 6 9 6 8 7
11 21 18 20 19 15 17 16 14
3 5 4 2 4 3 5 2
(13) 9 6 7 9 7 8 6 9
19 20 18 21 16 14 15 17
1 1 1 2 3 4 5
(14) 16 15 IS 6 8 9 7
17 17 16 10 13 11 12
I I 3 2 5 4
(IS) 14 14 9 17 8 8
17 18 12 11 13 10
I 4 5 2 3
(6) 14 7 9 9 8
15 13 10 12 11
5 4 3 2
(17) 8 8 7 9
lJ I 12 10 13
1 I 1
i (18) 20 IQ 19
21 21 20
I I
(19) 18 18
21 20
.,
1
(20) 18
19 .
+1<cplilltc\1 with pcmlission of the American Supplier lnlltitute, Inc. f
~
>
..
~
.L
APPENDIX B
F-Tables
Brainstol"ming
1. What are we after?
A 2.
3.
4.
5.
I
B I Design Experiments
I
I I I I
c Experiment 1 I I Experiment 21 I I IExperiment X
I I I I
I
D I Analysis of Test Results
I
I
E Run Conformation Test with Optimum Condition
Figure B-1. A Taguchi experiment flow diagram.

Table 8-t. F Table F.•.<f.,(2), 9D~ Conftdence
jj = Number of degrees of freedom of numerator

h == Number of degrees of freedom of denominator
.',..>....
~
I.
2 3 4 5 6 7 8 9 10 12 15 20 24 30 40 611 120 co
12
1 39.864 49.500 53.593 55.833 57.241 5K.204 58.906 59.439 59.8$8 60.19$ 60.705 61.220 61.740 61.002 62.265 62.529 62.7 114 6.U.l61 63.32tc
l. 8.5263 9.0000 9.1618 9.2434 9.2926 9.3255 9.3491 9.36689.3805 9.3916 9.4081 9.4247 9.4413 ,9.44% 9.4579 9.4663 9.4746 9.48211 9.4913
.~ 5.5383 5.46Z4 5.3908 S,34~7 5.3092 S"~847 S,2662 ',2SI' "ZtiOO S,lJ04 ',21~ 5.2003 ',11145 5.1'64 5,1(;11 5.1597 5.1512 5.142~ 5.1337
4 4.5448 4.3246 4.1908 4.1073 ".0506 4.0098 3.9790 3.9.549 3.9357 3.9199 3.8955 3.86&9 3.8443 3.8310 3.8114 3.8036 3.7986 3.7753 3.7607
:; 4.0604 3.7797 3.6195 3.5202 3.4530 3.4045 3.3679 3.3393 3.3163 3.29743.2682 3.2380 3.2067 3.1905 3.1741 3.1573 3.1402 3.122K :1.1050
Cl 3.7700 3.4633 3.2888 3.1808 3.1075 3'()546 3.0145 2.9830 2.9571 2.9369 2.9047 2.8712 2.8363 2.8183 2.8000 2.7812 2.7620 2.742.1 2.7222
7 3.5894 3.2574 3.0741 2.9005 2.8833 2.8274 2;7849 2.7516 2.7247 2.7025 2.6681 2.6322 2.5947 2.5723 2.5555 2.5351 2.5142 2.49211 2.4708
8 3.4579 3.1131 2.9238 2.8064 2.7265 2.6683 2.6241 25893 2.561,2 2.5380 2.5020 2.4642 2.4246 2.4041 2.3830 2.3614 2.3391 2.3162 2.2926
I} 3.3603 3.0065 2.8129 2.6927 2.6106 2.5509 2.5053 2.4594 2.4403 2.4163 2.3789 2.3396 2.2983 2.2768 2.2547 2.2320 2.2085 2.1843 2.1592
10 3.2850 2.9245 2.7277 2.6053 2.5216 2.4606 2.4140 2.3m 2.3473 2.3226 2.2841 2.2435 2.2007 2.1784 2.1554 2.1317 2.1072 2.081K 2.0554
11 3.2252 2.8595 2.6002 2.5362 2.4512 2.3981 2.3416 2.3040 2.2735 2.2482 2.2087 2.1671 2.1230 2.1000 2.0762 2.0516 2.0261 1.9W7 1.9721
12 3.1765 2.8068 2.6055 2.4801 2.3940 2.3310 2.2828 2.2446 2.2135 2.1878 2.1474 2.1049 2.0597 2.0360 2.0115 1.9861 1.9597 1.93:23 1.9036
13 3.1362 2.7632 2.5603 2,4337 2.3467 2.2830 2.2341 2.1953 2.1638 2.1376 2.0966 2.0S32 2.0070 1.9827 1.9576 1.9315 1.9C143 I.H7511 I.H4b2
""":.---====~~~
~. .. • 4 $ U J. [caUl Ui ild i ' 8 £ , ) " ld .. uas t L .. 221 .. L &$ 2 £ tu $ :; tZi , '1
H .\.1022 2.7265 2.5222 2.3947 2.3059 2.2426 2.1931 2.1539 2.1220 2.0954 2.0537 2..0095 1.9625 1.9377 1.9\19 1.8852 1.8572 1.8280 1.7973
1'\ .'.1m2 ~.6952 2.4898 2.3614 2.2730 2.2081 2.1582 2.1185 2.0862 2.0593 2.0171 1.9722 1.9243 1.8890 1.8728 1.8454 1.8168 1.7867 1.7551
III .H)4KI 2.6b82 2.4618 2.3327 2.2438 2.1783 2.1280 2.0880 2.0553 2.0281 1.9854 1.9399 1.8913 1.8656 1.8388 1.8108 1.7816 1.7507 1.7182
1'1 ,H)262 2.6446 2.4374 2.3077 2.2183 2.1524 2.1017 2.06·13 2.0284 2.0009 J.9S77 1.9117 1.8624 1.8362 1.8090 1.7805 1.7506 1.7191 1.6856
IX .H1070 2.(l239 2.4160 2.2858 2.1958 1.1296 2.078S 2,0379 1.0041 1.9110 1.9333 1.8868 1.8368 1.8103 1.7827 1.7537 1.7232 1.6910 1.6567
I') 2.9891} 2.6056 2.3910 2.2663 2.1760 2.1094 2.0580 2.0111 1.9836 1.9557 1.9117 1.8647 1.8142 1.7873 1.7592 1.7298 1.6988 1.6659 1.6308
20 2.9747 2.5S'I3 2.3801 2,2489 2.1582 2.0913 2.0397 1.9985 1.9649 1.9367 1.8924 1.8449 1.7938 1.7667 1.7382 1.7083 1.6769 1.6433 1.6074
21 2.96()1,) 2.5746 2.3549 2.2333 2.1423 1.0151 2.0232 1.9819 1.9480 1.9197 1.8150 1.8272 1.7756 1.7481 1.7193 1.6890 1.6569 1.6228 1.5862
22 2.9486 2.5613 2.3512 2.2193 2.1279 2.0605 2.0084 1.9668 1.9327 1.9043 1.8503 1.811l I. 7590 1.7312 1.7021 1.6714 1.6389 1.6042 1.5668
.'1 2,\).'74 2.5493 2.3387 2.2065 2.1149 2.0472 1.9949 1.9531 1.9189 1.8903 1.8450 1.7964 1.7439 1.7159 1.6864 1.6554 1.6224 1.5871 1.5490
24 2.9271 2.5310 2.3274 2.1949 2.1030 2.0351 1.9826 1.9407 1.9063 1.8775 1.8319 1.7831 1.7302 1.7019 1.6721 1.6407 1.6073 1.5715 1.5327
.''\ 2.9177 2.5283 2.31702.1843 2.0922 2.0241 1.9714 1.9292 1.8947 1.8548 1.8200 1.7708 1.7175 1..6890 1.6539 1.6272 1.5934 l.S51O 1.5176
.'h 2.9091 2.5191 2.3015 2.1745 2.0822 2.0139 1.9610 1.9188 1.8841 1.8550 1.8090 1.7596 1.7059 1.6771 1.6468 1.6147 1.5805 1.5437 1.5036
:7 2,l1012 2.5106 2.2987 2.1655 2.0730 2.0045 1.9515 1.9091 1.8743 1.8451 1.7989 1.7492 1..6951 1.6662 1.6356 1.6032 1.5687 1.5313 1.4906
2X 2.8939 2.5028 2.2906 2.1571 2.0645 1.9959 1.9427 1.9001 1.8652 1.8359 1.7895 1.7395 1.6852 1.6560 1.6252 1.5925 1.5575 1.5198 1.4784
21) 2.KK71 2.4955 2.2831 2.1494 2.0566 1.9678 1.9345 1.8918 l,856O 1.8274 1.7808 1.'7306 1.6759 1.6465 1.6155 1.5825 l.S472 I.~ 1.4670
,.
,- 10 2.HII07 2.4887 2.2161 2.1422 2.0492 1.9803 l,9269 1.8841 1.8498 1.8195 1.7727 1.7223 1.6673 1.6377 1.6065.. 1.5732 l.S376 t.49K9 1.4564
·m 2.11354 ~.44()4 2.2261 2.0909 1.9968 1.9269 1.8725 1.8289 1.7929 1.7627 1.7146 1.6624 1.6052 1.5741 1.5411 1.5056 1.4572 1.4'!'4K 1.376Q
1111 2.7lJ14 2..\lJJ2 2.1774 2.0410 1.9457 l,K747 1.8194 1.7748 1.1380 1.7070 l.M74 1.6034 1.S435 1.5107 1.4755 1.4373 1.3952 1.3476 1.2915
I '11 :' 7·17x .' 1·17' 2 1:\(1() Il)fm I.KlJ5Q 1.14238 1.7675 1.7220 1.6843 I.M24 I.WI2 1.5450 1.4821 1.4472 1.4094 1.3676 1.3203 1.2646 1.19:!6
'.', .' Ill"''' .' 10.'11 1.IItIIK 1
•• ' . . '.'~"" n . . . ·~·~ _ ...... _
I"'' ', I 847\ 17741 1.1Hl7 I MfIl I (I \ I'i
_.- I '1 1'K 1. 1 ""'iK
--.-. --_._-_._
I .JK71 I ·1~J(1
•...--.,..,
1.,\1"2 1..\410 I. 2't51 \.24(11) 1.1 hKtl I (KIIIII
"'"""'._-~:_, ... ......
~, ....•~ ~
I
\ .
_. ...1
Table B-2. F Table F.05,(rhr~, 95~ Confideace
I. = Number of degrees of freedom of numerator

12 = Number of degrees of freedom of denominator
...
(.,J
J- /,
2 3 4 S 6 7 8 9 10 12 IS 20 24 30 40 60 120 00
Il
I 161.45 199.50 215.71 224.58 230.16 233.99 236.77238.88 240.54 24Ul8 243.91 245.95248.01 249.05 250.09 251.14 2S2.20 253.25 25·U2
2 111.513 19.000 19.614 19.247 19.296 19.330 19.353 19.371 19.385 19.396 19.413 19.429 19.446 19.454 19.462 19.471 19.4711 19.487 111.4%
j 10.1211 9.5521 9.2766 9.1172 9.0135 8.9406 8.8868 8.8452 8.8123 &.7855 8.7446 8.1029 8.6602 IU385 8.6166 8.5944 8.5720 8.5494 11.5265
4 7.70116 6.9443 6.5914 6.3883 6.2560 6.1631 6.0942 6.0410 5.9988 $.9644 5.9117 5.8578 5.802S ".7744 5.7459 5.7170 5.61478 5.6581 5.6281
5 6.6079 5.7861 5.4095 5.1922 5.0503 4.950) 4.8759 4.8183 4.7725 4.7351 4.6777 4.6188 4.5581 4.5272 4.4957 4.4638 4.4314 4.3984 4.J650
6 5.:.~1l74 5.1433 4.7571 4.5337 4.3874 4.28)0 4.2066 4.1468 4.0990 4.0600 3.9999 3.93~U 3.8742 3.8415 3.8082 3.7143 3.73911 3.7047 3.6688
7 5.5914 4.7374 4.3468. 1 4.1203 3.9715 3.8660 3.7870 3.7257 3.6767 3.6365 3.5747 3.5108 3.4445 3.4105 3.3758 3.3404 3.3043 3.2674 3.2298
11 5.3177 4.4590 4.0662 3.8378 3.6875 3,5806 3.SOO5 3.4381 3.3881 3.3472 3.2840 3.2184 3.1503 3.1152 3.0794 3.0428 3.0053 2.9669 2.9276
9 5.1174 4.2565 3.8626 3.6331 3.4817 3.3738 3.2927 3.2296 3.1789 3.1373 3.0729 3.0061 2.936$2.900S 2.8637 2.8259 2.7872 2.7475 2.7067
10 4.11646 4.1028 3.7083 3.4780 3.3258 3.2172 3.1355 3.0717 3.02042.9782 2.9130 2.8450 2.7740 2.7372 2.6996 2.6609 2.6211 2.5801 ;UJ79
11 4.1l443 3.9823 3.5874 3.3567 3.2039 3.0946 3.0123 2.9480 2.8962 2.8536 2.78762.7186 2.6464 %,1,090 2.5705 2.5309 2,490! 2.4480 2.4045
12 4.7472 3.8853 3.4903 3.2592 3.1059 2.9961 2.9134 2.8486 2.7964 :US34 2.6866 2.6169 2.5436 2.JO$S 2.4663 2.42S9 2.3842 2.34IU 2.2%2
IJ 4.6672 3.110S6 3.4105 3.1791 3,0254 2.9153 2.8321 :2.'1669 2.7144 umo 2,6031 2,8331 2.4$19 3.42022.3103 2.3392 2.2!Jfl6 ~UU4 ;1.;1(164
""?"!£~C'~~~U,~fjHII I1 illi· [ ~ u;t. If U!!!j!l!~j_ I .

•.•• _ed; I oh M£&.;; i t ! i1",1 ; ; I. . JUS ; u:; J & • i nit 2 ~ 2 2 2 a t ! :: t 1
1·\ .l.(,O()I J. Utl9 3.3439 3.1122 2.9582 2.8477 2.7642 2.69872.6458 2.6021 25342 2.4630 2.3879 2.3487 2.3082 2.2664 2.2230 2.1178 2.1307
1'1 4.54.\1 3.6H23 3.2874 3.0556 2.9013 2.7905 2.7066 2.6408 2.5816 2.5437 2.4753 2..40352.3215 2.2878 2.2468 2.2043 2.1601 2.1141 2.0658
If> 4.49411 .U:U7 3.2389 3.0069 2.8524 2.7413 2.6572 2.59H 2.5377 2.4935 2.4247 2.3522 2.2756 2.2354 2.1938 2.1507 2.1058 2.0589 2.0096
i1 .l.4SD :U915 3.1968 2.9647 2.8100 2.6987 2.6143 2.5480 2.4943 2.4499 2.3807 2.3077 2.2304 2.1898 2.1477 2.1040 2.0584 2.0107 L9604
IX 4.4139 3.5546 3.1599 2.9277 2.7729 2.6613 2.5767 2.5102 2.4563 2.4117 2.3421 2.2686 2.1906 2.1497 2.1071 2.0629 2.0166 1.9681 1.9168
19 ·UXOH 35219 3.1274 2.8951 2.7401 2.6283 2.5435 2.4768 2.4227 2.3779 2.3080 2.2341 2.155S 2.1141 2.0712 2.0264 1.9796 1.9302 1.8780
10 4.3SI.l J.4928 3.0984 2.8661 2.7109 2.5990 2.5140 2.4471 2.3928 2.34792.2776 2.2033 2.1242 2.0825 2.0391 1.9938 1.9464 1.8963 1.8432
21 .t ..124H 3.4MB 3,0725 2.8401 2.6848 2.5727 2.48.76 2.4205 2.3661 2.3210 2.2S04 2.1757 2.0960 2.l)S40 2.0102 1.9645 1.9165 1.8657 1.8117
22 4.3001}' 3.4434 3.0491 2.8167 2.6613 2.5491 2.4638 2.3965 2.3419 2.2967 2.2258 2.1508 2.0707 2.0283 1.9842 1.9380 1.8895 1.8380 1.7831
2.\ 4.2793 3.4221 3.0280 2.7955 2.64002.5277 2.4422 2.3748 2.3201 2.2747 2.2036 2.1282 2.0476 2.0050 1.960S 1.9139 1.8649 1.8128 1.7570
24 4.2597 3.4026 3.0088 2.7763 2.6207 2.5082 2.4226 2.3551 2.3002 2.2547 2.1834 2.1077 2.0267 1.9838 1.9390 1.8920 1.8424 1.7897 1.7331
1'\ 4.2417 3.3852 2.9912 2.7587 2.6030 2.4904 2.4047 2.3371 2.2821 2.2365 2.1649 2.0889 2.0015 1.9643 1.9192 1.8718 1.8217 1.7684 1.7110
2(, 4.2252 3.3tt90 2.9751 2.7426 2.5868 2.4741 2.3883 2.32052.2655 2.2197 2.1479 2.0716 1.9898 1.9464 1.9010 1.8533 1.8027 1.7488 1.6906
27 4.2100 .'..1541 2.9604 2.7278 2.57192.4591 2.3732 2.3053 2.2501 2.2043 2.1323 2.0558 1.9136 1.9299 1.8842 1.8361 1.7851 1.7307 1.6717
11\ 4.1960 3.3404 2.9467 2.7141 2.5581 2.4453 2.3593 2.2913 2.2360 2.1900 2.1179 2.0411 1.9586 1.9147 1.8687 1.8203 1.7689 1.7138 1.6541
29 4.llBO 3.)277 2.9340 2.7014 2.5'454 2.4324 2.3463 2.2782 2.22292.1768 2.1045 2.0275 1.9446 1.9005 1.8543 1.8055 1.7537 1.6981 1.6377
:. i
,0 4.170t) J ..H58 2.9223 2.6896 2.5336 2.4205 2.3343 2.2662 2.2107 2.1646 2.0921 2.0148 1.9317 1.8874 1.8409 1.7918 1.7396 1.6835 1.6223
..
',..,
(12
.\0 4.0H41\ .':nI7 2.8387 2.6060 2.4495 2.3359 2.2490 2.1802 1.1240 2.0772 2.0035 1.9245 1.8389 1.7929 1.7444 1.6928 1.6373 1.5766 I.SOKQ
1.11 ·1 {K)J~ '1'i()4 2.7581 2.5252 2.3683 2.2540 2.1665 1.0970 2.0401 1.9926 1..9174 1.8364 1.7480 1.700t 1.6491 1.5943 1.5343 1,4673 1.38Q.,
I '11 \'I!O; \ 071H 2.hH02 2.4472 2.2tX10 2.17~ 2.C1867 2.()IM 1.95M3 1. 11105 l.ttll7 1.7~IS J.{,5K7 1.60tC4 U543 1.4952 1.42'X1 1..\1114 US.'"
,. \.H415 l.WS7 2.6t149 2..\71" 2.2141 2.(1')86 2.()()l)6 1.9384 I.K7W 1.11307 1.7522 1.6M4 1.S70S I.S 173 1.4591 UCI4$I 1.31 KO I. n 14 J.(KIIllI
,.
Table B..3. F Table F.ou<'.,fa), 97.S~ Confidence
\
11 == Number of degrees of freedom of numerator
12 == Number of degrees of freedom of denominator
: ..
f....
0'\
'. :I 3 .. s 1 u 9 10 IZ IS JO 24 JO 40 12U
6 flU co
I~
I 647.79 799,50 K64.16 899.58 921.85 937.11 948.22 956.66 963.28 968.63 976.71 984.87 993. JO 997.2S 1001.4 IOOS.6 1009.ll 1014.0 10111.3
2 31<1.506 39.000 39.165 39.248 39.298 39.331 39.355 39.373 39.387 39.398 39.415 39.43J 39.448 39.456 39.465 39.473 39.4lH 39.490 .19.498
.i 17.443 16,()44 15.439 15,101 14.885 14,7lS 14.624 14.540 14.473 14.419 14,337 14.253 14.167 14.124 14.081 14.037 13.99~ 13,947 13.902
4 12.218 10.649 9.9792 9.6045 9.3645 9.1973 9.0741 8.9796 8.9041 8.84398.1512 8.6565 8.'s599 8.5109 8.4613 8.4111 8.3604 8.3092 8.2573
5 10.1107 It4336 7.7636 7.3879 7.1464 6.9777 6.K53! 6.7572 6.6810 6.6192 6.5246 6.4277 6.3215 6.2780 6.2269 6.1751 6.1225 6.0693 6.0153
b K.1l131 7.2598 6.5988 6.2272 5.C>876 5.8197 5.6955 5.5996 S.5~ 5.4613 5.3662 5.2617 5.1614 5.1172 5.0652 5.0125 S.9:'i8tJ 4.'KM!! 4.841J1
7 lt0727 6..5415 5,8898 .5..5226 5.2852 S.IIK6 4.9949 4.8994 4.8232 4.7611 4.6658 4.5678 4.4667 4.41SO 4.3624 4.3089 4.2.S44 4.19K9 4.142.\
II 7.5709 6.0595 5.4160 5.0526 4.8173 4.6517 4.5286 4.4332 4.3572 4.2951 4.1997 4.1012 3.9995 3.9472 3.8940 3.8398 3.7344 3.7279 .\.6702
9 7.2093 5.7147 4.0781 4.7l81 4.4844 4.3107 4.1971 4.1020 4.0260 3.9639 3,8682 3.7694 3.6669 3.6142 3.5604 3.5055 3.4493 3.3918 3.3329
10 6.9367 5.4564 4.8256 4.4683 4.2361 4.0721 3.9498 3.8549 3.7790 3.7168 3.6209 3.5217 3.4186 3.3654 3.3110 3.2554 3.1984 3.1399 3.0798
11 6.7241 5.25S9 4.6300 4.2751 4.0440 3.8807 3.7S86 3.6638 3.5879 3.S2S7 3.4Z96 3.3299 3.2261 3.lns 3.1176 3.0613 3.0035 2.9441 2.88211
12 6.5538 5.0959 4.4742 4.1212 3.8911 3.1293 3.6065 3.5118 :,'-4358 3.3736 3,2713 3.1772 3.0721 3.0187 2.9633 2.9063 2,8478 2.7874 2.7249
13 6.4143 4.9653 4.3472 3.9959 3.7667 3.6043 3.4827 3.3880 3.3120 3.2497 3.1532 3.0527 2.9477 2.8932 2.8313 2.77972.1204 2.6590 2.5955
.1
H"""''''''''''',~'*'''_4>$,~4I1ll··IIU._.·-(*j''~.·'''_rnm.tJI
~"""""''1'"<'"''
_""'}_-' __:~i:"';:;:':c;,_f¥,*<>,';-"'O"'" {-".t;::~'_~;(*"'_' .-,,,, _. _ , , '. .,"1:',,$1:"''-' ,,'''','''{''_-';li{--C-. __, : _ ' ' ; 7,,::;,,<~?__ ," , , , PI
' _ . 7 ' 72 •
__.-.w. __ , , • i : L I.,. J dj!;t 1 t.. £ I LUS L ! M $; ... ; :: : . . , .. -~~~~
i·; II '1 1)79 4.H~67 4.2417 3.8919 3.6634 3.5014 3.3799 3.2853 3.2003 3.1469 3.0501 2.9493 2.8437 2.7888 2.7324 2.6742 2.6142 :1.5.519 2.4117:!
I~ Cl.I!)95 4.7650 4.1528 3.8043 3.5764 3.4147 3.2934 3.1987 3.1227 3.0602 2.%33 2.8621 2.7559 2.7006 2.64,37 2.5856 2.5242 2.4611 2.3953
Ih 6.1151 4.6!!67 4.0768 3.7294 3.5021 3.3406 3.2194 3.1248 3.0488 2.9862 2.8890 2,1875 2.6808 2.6252 2.5678 2.5085 2.4471 2.3831 2.3163
I1 Cl.1l420 4.6189 4.0112 HMS' 3.4379 3.2767 3.1556 3.0610 2.9849 2.9222 2.8249 2.7230 2.6158 2.55982.5621 2.4422 2.3801 2.3153 2.2474
III ~.97!!1 4.~597 3.9539 3.6083 3.3820 3.2209 3.0999 3.00s3 2.9291 2.8664 2.7689 2.6667 2."90 2.~7 2.4445 2.3842 2.3214 2.2558 2.1869
1'1 5.9216 4.5075 3.9034 3.5587 3.3327 3.1718 3.0509 2.9563 2.88(l() 2.8173 2.7196 2.6171 2.5089 2.4523 2.3937 2.3329 2.2695 2.2032 2.1333
l() 5.117IS 4.4613 3.8587 3.5147 3.2891 3.1283 3.0074 2.9128 2.8365 2.7737 2.6758 2.5731 2.4645 2.4076 2.3486 2.2873 2.2234 2.1562 2.0853
.?l 5.8266 4.4199 3.8188 3.4754 3.2501 3.0895 2.9686 2.9740 2.7977 2.7348 2.6368 2.5338 2.4247 2.3675 2.3082 2.2465 2.1819 2.1141 2.0422
!.l 5.71163 4.31128 3.7829 3.4401 3.2151 3.0546 2.9338 2.8392 2.7628 2.6998 2.6017 2.4984 2.3890 2.3315 2.2718 2.2097 2.1446 2.0760 2.0032
2~ 5.74911 4.3492 3.7505 3.4083 3.1835 3.0232 2.9024 2.8077 2.7313 2.6682 2.5699 2.4665 2.3567 2.2989 2.2389 2.1763 2.1107 2.0415 1.9677
24 5.7167 4.3187 3.7211 3.3794 3.1548 2.9946 2.8738 2.7791 2.7027 2.6396 2.5412 2.4374 2.3273 2.2693 2.2090 2.1460 2.0799 2.0099 1.9353
2'; 5.611M 4.2909 3.6843 3.3530 3.1287 2.9685 2.8478 2.7531 2.6766 2.6135 2.5149 2.4110 2.3005 2.2422 2.1816 2.1183 2.0517 1.9811 1.9055
2h 5.65R6 4.2655 3.6697 3.3289 3.1048 2.9447 2.8240 2.7293 2.6528 2.5895 2.4909 2.3867 2.2759 2.2174 2.1565 2.0928 2.0257 1.9545 1.87S1
27 :'i.6.HI 4.2421 3.6472 3.3067 3.0628 2.9228 2.8021 2.7074 2.6309 2.5675 2.4688 2.3644 2.2533 2.1946 2.1334 2.0693 2.0018 1.9299 1.8527
211 5.6()9(} 4.2205 3.6264 3.2863 3.0625 2.9027 2.7820 2.6872 2.6106 2.5473 2.4484 2.3438 2.2324 2.1735 2.1121 2.0471 1.9796 1.9072 1.8291
."1 :'i.~H7H 4.2006 3.6072 3.2674 3.0438 2.8840 2.7633 2.6686 2.5919 2.52862.4295 2.3248 2.2131 2.1540 2.0923 2.0276 1.9591 1.8861 1.8072
,,
,~
, lO <;5(175 ·US21 3.S8t )4 3,2499 )'()265 2.8667 2.7460 2.6513 2.5746 2.5112 2.4120 2.3072 2.1952 2.1359 2.0739 2.00fl9 1.9400 1.8664 -1.7867
.1\} :'i.42.W 4.0510 3.4633 3.1261 2.9037 2.7444 2.6238 2.5289 2.4519 2.3882 2.2882 2.1819 2.0677 2.0069 1.9429 1.8752 1.8028 1.7242 \.b.\'7I
I.ll '\ .'H.';7 I l l:!5,1 :U425 .1.0077 2.7863 2.6274 2.5068 2.4117 2.3344 2.27022.1692 2.0613 1.9445 1.8817 1.8152 1.7440 1.666tI UlllO 1.4H22
I.'ll ·U'i2·1 II\II-$l .1.2270 2.K1)43 2.6740 2.5154 2.3948 2..2CJ94 2.2217 2.t57tl 2.()S48 I. I J4S0 l.K24lJ 1.7597 1.6HtJ9 I. hi 41 U2Qt) 1.4.l27 !..\1Il4
.,tj
'i 02.W \.fIR!!$ .U 161 2.711SS 2.5665 2.4082 2.2875 2.191S 2.1 '-'6 2.()493 1.9447 I.R.U6 I. 70tlS 1.6402 1.5660 1.4RJS URtI.' UMI4 IIMKKI
, . , ,_ _
!
r
, ..
,.'
./
Table 8-4. F Table F.01(f. ,f;a), 99CJ, Conlid,nee

f. = Number of degrees of freedom of numerator
12 = Number of degrees of freedom of denominator
12~~~~~
.~ I1
-j~
2 3 4 5 6 7 8 9 J() 12 15 20 24 30 40 80 120 QC
I 4052.2 4999.5 5403.3 5624.6 5763.7 5859.0 5928.3 5981.6 6022.5 6OS5.8 6106.3 6157.3 6208.7 6234.6 6260.7 62&6.8 6313.0 6339.4 h366.0
2 98.503 99.000 99.166 99.249 99.299 99.332 99.356 99.374 99.388 99.39999.415 99.432 99.449 99.458 99.466 9!U74 99.483 99.491 99.S01
3 34.116 30.1:H1 29.4" 28.710 211.231 21.91127.6'12 2'1.489 27.345 tt.229 tt.• 26.1'12 26. • M.SII 26.501 26.411 26.31626.221 26. US
4 21.198 18.000 16.694 15.971 15.522 15.207 14.986 14.79914.659 14.546 14.374 14.198 14.020 13.929 13.838 13.745 13.652 13.558 13.463
5 16.258 13.274 12.060 11.392 10.967 10.672 10.456 10.289 10.158 10.051 9.8883 9.7222 9.5527 9.466$ 9.3793 9.2912 9.2020 9.1118 9.0204
6 13.745 10.925 9.7795 9.1483 8.7459 8.4661 8.2600 8.1016 7.9761 7.8741 7.7183 1.5590 7.3958 7.3127 7.2285 7.1432 7.0S6!i 6.9690 6.81«11
7 12.246 9.5466 8.4513 7.8467 7.4604 7.1914 6.9928 6.8401 6.7188 6.6201 6.4691 6.3143 6.1554 6.0743 5.99215.9084 5.8236 5.7372 ~.64CJ5
8 11.259 8.6491 7.5910 7.0060 6.6318 6.3707 6.1776 6.0289 5.9106 5.8143 5.6668 S.SISl 5.3591 5.2793 5.1981 5.1156 5.0316 4.9460 4.8588
\) 10.561 8.0215 6.9919 6.4221 6.0569 5.8018 5.61295.4671 5.351t 5.2S65 5.1114 4.9621 4.8080 4.7290 4.6486 4.5667 4.4831 4.3978 4.3105
III 10.044 7.5584 6.5523 5.9943 5.6363 5.3858 5.2001 5.0561 4.9424 4.8402 4.70S9 4.5582 4.40S4 4.3269 4.2469 4.1653 4.0619 3.9965 ,1.9090
11 9.6460 7.2057 6.2167 5.6683 5.3160 5.0692 4.8861 4.7445 4.6315 4.5393 4.3974 4.2509 4.0990 4.0109 3.9411 3.8596 3.7761 3.6904 3.6025
12 9.3302 6.9266 5.9526 5AIl9 5.0643 4.8206 4.6395 4.4994 4.3875 4.2961 4.1553 4.0096 3.8584 3.7805 3.7008 3.6192 3.5355 3.4494 3.3608
13 9.0738 6.7010 4.7394 5.2053 4.8616 4.6204 4.4410 4.3021 4.1911 4.1003 3.9603 3.8154 3.6646 3.5868 3.5070 3.4253 3.3413 3.2548 3.1654
Kt I : tit! § 1 ;£.JJ49,,! h a $ • !2;: : it $ .• _-~
14 IU~616 6,5149 5.5639 5.0354 4.6950 4.4558 4.2779 4.1399 4.0297 3.9394 3.8001 3.6557 3.5052 3.4274 3.3476 3.2656 3.1813 3.t>942 3.0040
15 1I.6l:131 6.35119 5.4170 4.8932 4.5556 4.3183 4.1415 4.0045 3.8948 3.$049 3.6662 3.5222 3.3719 3.2940 3.2141 3.1319 3.0471 2.9595 2.Sb84
16 H.531O 6.2262 4.2922 4.7726 4.4374 4.2016 4.0259 3.8896 3.7$04 3.6909 3.5527 3.4089 3.2588 3.1808 3.1007 3.0182 2.9330 2.8447 2.7528
17 fI,3997 (1.1121 5.1850 4,5590 4.3359 4.101$ 3.9267 :1.1910 3.6822 3.5931 3.4552 3.3f17 3.1615 3.0835 3.0032 2.9205 2.8348 2.7459 2.tl530
18 8.2854 6,IH2t) 4.0919 4.5790 4.24~ 4.0146 3.8406 3.7054 3.5971 3.5082 3.3706 3.2273 3.0771 2.9990 2.9185 2.8354 2.7493 2.6597 2.5660
19 l:U850 5,9259 5.0103 4.S003 4.1708. 3.9386 3.7653 3.6305 3.5225 3.4338 3.2965 3.1533 3.0031 2.9249 2.8442 2.7608 2.6742 2.5839 2.4893
20 H.0960 5.H489 4.9382 4.4307 4.1027 3.8714 3.6987 3.5644 3.4567 3.3682 3.2311 3.0880 2.9377 2.8594 2.7785 2.6947 2.6077 2.5168 2.4212
21 H.Oltm 5,7804 4.8740 4..:\688 4.0421 3.8117 3.6396 3.5056 3.3981 3.3098 3.1729 3.0299 2.11976 2.8011 2.7200 2.6359 2.S4114 2.4568 L\60.\
22 7.9454 5.711)() 4.8166 4.3134 3.9880 3.7583 3.5867 3.4530 3.3458 3.2516 3.1209 2.971lO 2.M274 2.7488 2.6675 2.5831 2.4951 2.4029 2..\055
23 7.8811 5.6637 4.7649 4.2635 3.9392 3.7102 3.5390 3.4051 3.2986 3.2106 3.0740 2.9311 2.7805 2.1017 2.6202 2.5355 2.4471 2.3542 2.2559
24 7.8229 5,6136 4.7181 4.2184 3.8951 3.6661 3.4959 3.3629 3.2S60 3,1681 H)316 2.8887 2.7380 2.6591 2.5713 2.4923 2.4035 2.3099 2.2107
25 7.76911 55.6l:l0 4.6755 4.1774 3.8550 3.6272 3.4568 3.3239 3.2172 3.1294 2.9931 2.8502 2.6993 2.6203 2.5383 2.4530 2.3637 2.2695 2.16tJ4
26 7.721.' 5,5263 4.6366 4,1400 3.8183 3.5911 3.4210 3.2884 3.1818 3.t>941 2.9579 2.8150 2.6tl4O 2.5848 2.5026 '2.4170 2.3273 2.2.1lS 2.\315
27 7,117f17 5AHIlI 4.6009 4.1056 3.71'148 3.5580 3.3882 3.255'8 3.t494 3.0618 2.9256 2.7827 2.6316 2.5522 2.4699 2..:t84O 2.293R 2.1984 2.09M
2K 7.6356 S·t529 4.5681 4.074() 3.1539 3.5216 3.3581 3.2259 3.1195 3,()J20 2.8959 2.7530 2.NU7 2.52232.4397 2.3535 2.2529 2.1670 2.0642
2l.J 7.5976 5.4205 4.5378 4.0449 3.7254 3.4995 3.3302 3.1982 3.0920 3.0045 2.8685 2.7256 2,5742 2.4946 2.4118 2.3253 2.2344 2.1378 2.()J42
..... 30 7.5625 5.3l.J04 4,5097 4.0179 3.6990 3.4735 3.3045 3.1726 3.0665 2.9791 2.8431 2.7002 2.5487 2.4589 2.3860 2.2992 2.2079 2.1 1tl7 2.0062
~
40 7.3141 5.17M5 4.3126 3.8283 3.5138 3.2910 3.1238 2.9930 2.8876 2.8005 2.6649 2.5216 2.3689 2.28llO 2.2034 2.1142 2.0194 1.9172 1.8047
fiO 7.0771 4.lJ774 4.1259 3.6591 3.33S9 3.1187 2.9530 2.8233 2.7185 2.6318 2.4961 2.3523 2.1978 2.1154 2.0285 1.9360 1.8363 1.726.\ 1.6006
120 6.11510 4.7K65 3.9493 3.4706 3.1735 2.9559 2.7918 2.6629 2.5586 2.4721 2.3363 2.1915 2.0.\46 1.9500 1.8600 1.7629 1.6557 1.5330 I.JM05
:le 6.hWJ ~()(.52 .\,7MI6 .U192 .1.0173 2.8020 2.6393 2.51132.4073 2.3209 2.I1'I411 2.<HS5 1.8783 1.7908 1.6964 1.5(tD 1.47.'0 J..'\246 1.lltl0U
. -~'l
Table 8-5. F Table F.005(f.,f1 ), 99.5% Confidence
/1 == Number of degrees of freedom of numerator

12 == Number of degrees of freedo·m of denominator
S 11 I
2 3 4 S
,6 7 8 9 10 12 IS 20 24 JO 40 110 120 cc
1:·1
16211 20000 21615 22500 23056 23437 23715 23925 24091 24224 24426 24630 24836 24940 2S044 25148 252S.l 253'9 254M
2 II,JK.50 199.00 199.17 199.25 199,30 199.33 199.36 199,37 199.39 199,40 199.42199.43 199.45 199.46 199.47 199.47 I99AIl 199.49 199.5 I
~ 55..552 49.799 47.467 46.195 45.392 44.838 44.434 44.126 43.882 43.686 43.387 43.085 42.778 42.622 42.466 42.308 4~1.149 41.9ft9 41.K29
.H.33.1 26.284.24.259 23.155 22.456 21.9,75 21.622 21.352 21.13920.967 20.705 20.438 20.161 20.030 19.892 19.752 19.611 19.468 19.-'25
'"
"l 22.7H5 IH.314 16.530 15.556 14.940 14.513 14.200 1:\.961 13.772 13.618 13.384 13.146 12.90312.780 12.656 12.530 12.402 12.274 12.144
Cl 1!l.635 14.544 12.917 12.028 11.464 11.073 10.786 10.566 10.391 10.250 10.034 9.8140 9.5888 9.4741 9.3583 9.2408 9.1211J 9.0015 K.1I793
16.236 12.404 10.882 10.050 9.5221 9.1554 8.8854 8.6781 8.5138 8.3803 8.1764 7.9578 7.7540 7.64$0 7.5345 7.4225 7.JO!lK 7.1933 7.0760
H 14.6!l1l 11.042 9.5965 8.8061 8.3018 7.9520 7.6952 7.4960 7.3386 7.2107 7.0149 6.8143 6.6082 6.5029 6.3961 6.2875 6.1772 6.{)649 5.Q505
'l 13.614 10.107 8.7171 7.9559 7.4711 7.1338 6.8849 6.6933 6.5411 6.41716.2274 6.0325 5.8318 5.7292 5.6248 5.Sl86 5Alt14 S.•lOOl 5.lIm
l(J 12.826 9.42708.0807 7.3428 6.8723 6,5446 6.3025 6.1159 5.9676 .5.8467 5.6613 5.4707 5.2740 5.1732 .5.0705 5.9659 4.H5tJ.'! 4.1$01 4.h.\K5
Ii 12.226 K.1J122 7.6004 6.8H(}t) 6.4217 6.1015 5.8648 5.6821 5.5368 5.4182 5.2363 5.0489 4.8552 4.7557 4.6543 4.5$08 4.44511 4.3367 4. 225tl
L! 11.754 8.5096 7.2258 6.5211 6.071l 5.7570 5.5245 5.3451 5.2021 5.0855 4.9063 4.7214 4.5299 4.4315 4.3309 4.2282 4.1:!.N 4.tH49 .'.\}()JIJ
H 11.374 8.1865 6.9257 6.2335 5.7910 5.4819 5.2529 5.0761 4.9351 4.8199 4.6429 4.4600 4.2703 4.1726 4.0727 3.9704 3.HM5 3.7577 3.6465
. ~• •tl::_';';"Ir9jrrnC[1 SPiUSST 7 q I'DIXIT'

...•.•••..._-------_ ....-----', 40"'; £t '"
.
14 11.<)0() 7.9217 6.6803 5.9984 5.5623 5.2574 5.0313 4.8566 4.7173 4.6034 4.4281 4.2468 4.0585 3.9614 3.8619 3.7600 3.6553 3.5473 3.4359
1
1'1 1O.7911 7.7(lOIl 6.4760 5.8029 5.3721 5.0708 4.8473 4.6743 4.5364 4.4236 4.2498 4.0698 3.8826 3.7859 3.6867 3.5850 3.4803 3.3722 3.2602
If> 10.575 7.51~!I 6.3034 5.6378 5.2117 4.9134 4.6920 4.5207 4.3838 4.2719 4.0994 3.9205 3.7342 3.6378 3.5388 3.4372 3.3324 3.2240 3.1115
11 IlUll4 7.3536 6.1556 5.4967 5.0746 4.7789 4.5594 4.3893 4.2535 4.1423 3.9709 3.7929 3.6073 3.5112 3.4124 3.3107 3.2058 3.0971 2.9839
IX 1O.:!lll 7.2148 6.0277 5.3746 4.9560 4.6627 4.4448 4.2759 4.1410 4.0305 3.8599 3.6827 3.4977 3.4017 3.3030 3.2014 3.0962 2.9871 2.8732
1'1 J(um 7.0935 5.9161 5.2681 4.8526 4.5614 4.3448 4.1770 4.0428 3.9329 3.7631 3.5866 3.4020 3.3062 3.2075 3.1058 3.0004 2.8906 2.7762'
:!() l).1)4:~9 6.91165 5,8177 5.1743 4.7616 4.4721 4.2569 4.0900 3.9564 3.8470 3.6779 3.5020 3.3178 3.2220 3.1234 3.0215 2.9159 2.8058 2.6904
21 9.1129.'i 6.11914 ,5,7304 5.0911 4.6808 4.3931 4.1789 4.0128 3.8799 3.7709 3.6024 3.4270 3.2431 3.1474 3.0488 2.9467 2.8408 2.7302 2.6140
l!. '1.7271 6,Il064 5.6524 5.0168 4.6088 4.3225 4.1094 3.9440 3.8116 3.7030 3.5350 3.3600 3.1764 3.0807 2.9821 2.8799 2.7736 2.6625 2.5455
21 'l.ClJ411 (,,7.'00 5.51123 4.9,500 4.5441 4.2591 4.0469 3.8822 3.7.502 3.6420 3.4745 :,\,2999 3.1165 3.0208 2.9221 2.8198 2.7132 2.6016 2.4107
.'·1 "'I'll' 6.6610 5,SIIJO 4,lIll98 4.4857 4.2019 3.9905 3.8264 3.694t) 3.5870 3.4199 3.2456 3.0624 2.9%7 2.8679 2.7654 2.6585 2.5463 2.4276
1'1 '1.4753 (,59112 S.4615 4.8351 4.4327 4.1500 3.9394 3.7758 3.6447 3.5370 3..l104 3.1953 3.0133 2.9176 2.8187 2.7160 2.6099 2.4%0 2.3765
2(, "'.·,10.'i9 6.5409 4.4091 4.7852 4.3844 4.1027 3.8929 3.7297 3.5989 3.4916 3.3252 3.1515 2.9685 2.8728 2.7738 2.6709 2.5633 2.4501 2.3297
.'1 I) .\423 6.4111l.'i 5.361.1 4.7396 5.3402 4.0,594 3.8501 3.6875 3.5,571 3.4499 3.2839 3.1104 2.9275 2.8318 2.7327 2.6296 2.5217 2.407112.2867
.!X ".211311 h.440.' :U170 4.6977 4.2996 4.0197 3.8110 3.6487 3.5186 3.4117 3.2460 3.0727 2.8891) 2.7941 2.6949 2.5916 2.48342.3689 2.2469
.!,t} .
IJ.2297 6,3lJ511 5.2764 4.6591 4.2622 3.9830 3.7749 3.6130 3.4832 3.3765 3.2111 3.0379 2.8551 2.7594 2.6601 2.5565 2.4479 2.3330 2.2102
,,
.... 10 9.1797 (,,3547 5,2388 4.62334.2276 3.9492 3.7416 3.5801 3.4505 3.3440 3.1787 H)(),57 2.8230 2.7272 2.6278 2.,5241 2.4151 2.2997 2.1760
·11) H.K2711 6.0664 4.97~9 4.37311 3.9860 3.7129 3.5088 3.3498 3.2220 3.1167 2.9531 2.7811 :!.59114 2.5020 2.401S 2.2958 2.18311 2.0635 1.9.'114
1>0 1l.4 I '4() 5.711~() 4,7290 4.1399 3.7600 3.4918 3.2911 3.1344 3.0083 2.9042 2.7419 2.S705 2.3872 2.2989 2.1874 2.0789 1.9622 1.1i)41 1.61485
121) H.17IJO 5.~W.' 4.4973 3.9207.3.5482 3.2849 3.0874 2.9330 2.8083 2.7052 2.5439 2.3727 2.1881 2.0890 1.9839 1.11.709 1.7459 1.6055 ..·BII
..,
. 7 K7«I.1 'I.:!'lll' 4,2794 3.7151 .U4Q(1 Hl91.l 2.89614 2.7444 2.62Hl2.51118 2.35874 2.1868 I. t /Cl911 1.8983 1.7891 1.M91 1.S32~ 1.3637 -r.(Kl(lO
"
.,.,.
·-4
APPENDIXC
.,•
Glossary
ANOVA: An analysis of variance is a table of infofl11dion tbat displays the
contributions of each factof.
Controllable fador: A design variable that is considered to infiueftcethe response and
is indudedin the experiment. Its level can be controlled by the experimenters.
DesigB of experiment: A systematic procedure to la)' out the factors and conditions of
an experiment. Taguchi employs specific partial factorial arrangements
(orthogonal arrays) to detennine the optimum design.
Factorial Experiment: A systematic procedure in which all COOlroUable factors except
one are held constant as the variable factor iseered discretely or continuously.
Error: The amount of variation in the response caused by factDf5 other than
controllable factors included in the experiment.
Interaction: Two factors are said to have interaction with each other if theinftuence
of one on the response function is dependent. the vaIR of the other.
Linear Graph: A graphical representation of relative column locations of factors and
their interactions. These were development hyDr. TagudD to assist in
assigning different factors to columns of the OtdIDgonal ana)'..
Loss Function: A mathematical expression proposed by Dr. Tapchi to quantitatively
detennine the additional cost tosociet}' caused by the lack of quality in a
product. This additional cost is viewed as a loss to society and is expressed as
a direct function of the mean square deviation from the target value.
Noise Factors: Noise faetorsare those factors that have an influence over a response
but cannot be controlled in actual applications:. They are of three kinds.
Outer Noise: Consists of environmental cooditions such as humidity
temperature. operators. etc.
InDer Noise: The deterioration of machiRes. tools and parts.
Between Product NoiSe: The variation from piece 10 piece.
Off-Line Quality Control: Refers to the quality enbaocement efforts in activities
before production. These are activities such as upstream planning. R & D.
systems design. parametric design. tolerance design and loss function. etc.
Orthogonal Array: A set of tables used to determine the least DUmber of experiments
and their conditions. The word orthogonal means balanced.
Outer Arra~': An orthogonal array used to define the conditions for the repetitions of
the inner array design to measure th~ effects C'f variou:,- ocise factors.. An
experiment with outer arrays will reduce the product variability and sensitivity
to the ini1uence of noise factors.
Parameter Design: Parameter design i~ u::-~j h' dc:,if= a prod:.:.:'! by selecting che
Jpimll;;J ,:C'nJition ;)( !hc parametc: !e". ;:~~ 5(1 j':Jt the pn.'\iuct is least sensitive
Quality Cltaracteristic: The yardstick \10 hich measures the performance of a product
or a process under study. For a plastic molding process this could be the
stragth of the molded piece. For a cake. this could be a combination of taste.
s . and moistness.
Response: A quantitative value of the measured quality characteristic. e.g. stiffness.
wright. flatness.
VariatioDlleductioo: The variation in me output of a process produces nonunifonnity
in dle product and is perceived as quality. Reduced variation increases
C1r1'5IQ·rner satisfaction and reduces warranty cost arising from variation. To
acIIieve better quality. a product must perform optimally and should have less
Vlriation around the optimum performance.
RobllStDal:Describes a condition in \iOhicba product or process is least influenced by
the -wariation of individual factors. To become robust is to become less sensitive
to ~ariations.
SIN Ratic Stands for the signal to noise ratio. i.e.• the ratio of the power of a signal
to die power of the noise (error). A high SIN ratio will mean that there is high
seEitivity with the least error of measurement. In Taguchi analysis using SIN
ratios. a higher value is alwa)"s desirable regardless of the quality characteristic.
Signal Fader: A factor that influences me average value. but not the Variability in
taplDse.
. System Daip: Th~:design of a product or a process using special Taguchi
tedmiques./:
Taguchi Design: Afttethodology to increase quality by optimizing system design.
panmeter deSign. and tolerance design. This text deals with system design.
Target V~: A value that a product is expected to posses. Most often this value is
c:liftUent from what a single unit actually does eXhibit. For a 9 vo~t transistor
baIrery the target value is 9 \'0115.
ToleratU Design: This is a sophisticated version of parametric design used to
OJI'imize tolerance. reduce costs. and increase customer satisfaction.
Variables" factors. or Parameters: These words are used synonymously to indicate
tBe controllable factors in an experiment. In the case of a plastic molding
e:qleriment. molding temperature. injection pressure. seftime. etc., are factors.
Index
Accuracy. 4 Confirmation test. 56

AgriculturalCQndition, 1 Conformance to speci6c&ms., 13
Air bag #.bigo, 202. 203 Consistency in experiroemal_go. 11
Allied forces, 7 Controllable factors. 27. 45. 90
Analysis of results, 16, 65 Couections faetors.SO, 113
signal/noise, 18 Cost of rejectiens, 39
Analysis of variance (ANOVA), 3, 16,32. variation. 22
100, 115. 117. 130. 143. 144. 152, 153, CI'lUlk shaft surface. 186
170. 188, 189. 191, 193. 194, 199, 200. Critical dimension. I I
204,208 Customer satisfaction. 12. 14
Analytical simulation, 17 Customer preference, 2
Application e~ample, 182
Array notation. 85. 184 Degrees of freedom (00Fl. SO. SI, 52.62.
Automobile generator, 190 101, 10S. US. 120. 132.144, ISO
Average effects. 2. 3. 48. 49, 67. 69 Degrees of interaction. 3
Average factor effects. 99 Deming. w. E.• 8
Average performance. 48 Design of experiment. 2
Average value. 33 Design factors. 175, 181. 191. 194. 197. 206.
207. See also Factot
Baking process, 24 Dryer motor belt. 166
experitnel1t. 62 Dummy treatment. 76. 82.83. 84
Barrier test vehicle. 198 Dummy level. 84. 85
Ben Laboratories, 7 Effective number of reptications. 140
Best combinations. 2 Electrical communicatioB . . . .ories (ECU. 7
Brainstorming. 17.29.31.44. 1.73 Engine valve train, 182~ 183. 185
major steps of. 181 Engine idle stability.193
session. 173. 175 Engineering. 2
communication, 44
Casting process. 83. 161 Evaluation criteria. 175
Characteristic effect. 173 Error sum of squares. 1&4-
Chemical and Phannaceutical industries. 1 DOF.136
Chocolate chip cookie. 2 variance. 136
Classical technique. 18 Estimated results. 139
Colunm modification. 76 Examination of customer response. 2
Colunm preparation. 79 Experimental efficiency. 31. 32. 42
Combination design. 90 data. 59
Combinations. 2 errors. 5
Competitive position. 1
Conclusions. 47 Factorial design of expe:1rnent. 1. 2i. 30. 42
Contiden..:-e int~f\al. 116. 138. 139. IJO. 141.
~ 8:', ~ 96. :(i!, :'):,
· ...;-'
FlSb finder. 32 f~.9.3O,90,92,94.95. 176

Fisher. R. A.• 1.27,40 experiment. 9S
Fractional factorial design, 27 NonparaIlel.3
Frequency distribution. 12
Front structure crusb, 207. 208 Objective evaluation. 19
Fuel pump noise. 169 Off-line strategy. 8
Full factorial experiment. 31. 55 quality improvemeat. 8
design, 41 One way ANOVA. 106. 112
Fundamental concepts. 9 Operating conditioa, 16
Optimilation. I, 29
General variance, 102 Optimum combination, SS, 86
Grand total. SS condition, 10. 16, 29, 33, 44. 48. 49, 94,
Graphical method, S9 96
performance, SS
Ingredients, 2 treatment. I
Inner amy. 28, 91. 9S Ortbogonal arrays (OA), 14,93. 187. 202,
Instrument panel structure, 195 206
Interactingpais'. 60 Outer anay. 28. 91. 9S
Interaction, 3.40. S8. 59. 60. 81, 128 Out-of-toIerance. 12
effects, 70. 90 Overall ev.,auation aitetia (OEC>. 19. 178,
pair, 69 179
stQdies. 176
Intersecting lines. 69j\jh
,;;i~, Pace maker. 1
Parallel. 3
Parameter design, 10. 24
Partial faaorial experiment. 40, 41
Latin squares, 14
Percent~, 13
Levels. 2
Percent contributioa, SO, SI, 110. liS. 123
Life expectancy, 9
Percent inftuenee. SS
Linear combination. 44
Pooled effects, 54
Linear graph. 61. 63. 71. 74. 77.82
Pooling. 124, 134
Loss function, 10, 11. 13. 18.• 33. 1.56, 157.
Pound cake, 26
160, 163. 164, 171
Price and performance, I
Problem solving. 18
Main effects. 3. 30. 44. 49. 82. 86. 93, 94.
Product parameter. 8
143, 152. 176. 184. 188. 192. 195. 198,
Product Watranty. 18
200,204.208
Profitability. I
Manufacturing process control. 32
Pure sum of squares. SO. SI. 109, 111. 123.
Manufacturing tolerance, 39, 161. 167. 168
133
Market survey, 2, 3
Mean squared deviation (MSP>. 33. 36. 92.
146. 149. 162 QUALITEK-3 software, I2S
Mean squares. 50 Quality characteristic, 19, 20, 156, 158
Muedleveh.40, 58 Quality strategy. 23
Molding process. 45. 75
Monthly savings. 165 Random order. 46, 47
Multiple runs. 142 Relative influence. 30
contribution. 50
Noise condition. 94 significance. 72
UN
/1
Index 147
Ikpetition. 47. 91 Tagucbi pbilosopby."8

Ieplication.47. 121 Target value. 9. 158
Ieproducibility. 41 properties. 11
Response Tmn approach. 17. 18
customer. 2 Team leader. 174 I
mean. 2 Technique. I
Robust condition. 9. 17, 91 Total ~ of freedom. SO
Tolenmce
tj
design. 10
Second World War, 7. 14
levels. 37
Session advisor. 174
limits. 167. 168
Sipal to noise ratio (SIN). 28. 31. 33. 36.
Traditiooal practices. 7
92,96. 145. 148, ISO. 153. 1S4
Transmission control able. 204
Simulation studies. 18
Sack food. 2
TreatmeDt.2 11
Trial condition. 14. 89. 98
SocietaIloss, 11. 14
Trial run. 89
Sony. 12
Triangular table, 60. 61. 71. 72
Specification llinit. 11
Two way A.."'lOVA. 117
Standard analysis. 126
approach. 16
arrays. 30. 58 Uncontrollable factors. 27
deviation. 33. 105. 163. 164 Up-front thinking. 31
Statistical process control (SPC). 32. 103
Subjective evaluation. 19 Variability. 100
Sugar. 2. 3 Variance. 102
Sums of squares. SO. 115 ratio. SO. 53. 109. 115. 120. 133
~I~r~~ce. 161, 168 data. 163. 164
System design. 10 Variation. 9. 20. lIS
\..
I
I
II
f . ." .
1·.·.1·>
!
t'
~.'"
,~
j
'~i;
.j
I

A Primer On The Taguchi Method - 1

Uploaded by

Copyright:

Available Formats

A Primer On The Taguchi Method - 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Primer On The Taguchi Method - 1

Uploaded by

Copyright:

Available Formats

Copyript C 1910 by Van Nostrand ReiDhold

Ubruy of C ~ Catalog Card Number 89-14736

Printed in the ~ States of America

Library of C..- Cataloging-in-Pablication Data

To my wife Krishna and

"-----'" ."-... -. -_.-.~-- . . /

- VNR COMPEIIllVE MANUfACTURING SERIES-

INDUSTRIAL ROBOT HANDBOOJt

WORI<JMG TOWARDSJUST.... :ma

1. Qudty ......... Produet ......... ~ /1

'..... '.,- _. ~ . .- ... ".' ~. .. ~'. .. .....

s. Working Mechanics of the TapddDesIp of Experbnents 140

6-7 StaDdanl ArJalysis wilhSie&Je and Multiple a.. /126

8. BrainstonDIJla-Aa 1Rt", Part of the TagadIl PIdlosephy 1t73

.• _ ...'" __ .'i. ...~ ....... _ .

Example 9-9 Front Stmcture Crush Characteristics /208

Appendix A Orthogonal Arrays, Triangular Tables and

$Jhe cum Tagldi Metlrot:f8 is a ttademart of dle AmeriaII

A,B J : • Variables used in ,the design of an experiment

Quality Through Product and

1-2 DESIGN OF EXPERIMEN.TS-THE CONVENTIONAL

disadvantages from an engineering point of view, is illustrated by a simple .

Table I-I: Taste Survey with Preference in Percent

LEVEL OF (CHOCOLATE <:HIPS)

Mean response (sUP)

Figure 1-1. Factor effects.

4 A Primer on the Taguchi Method

Table 1-2. Test Matrix with Three Factors at Two Levels

For a fuU factorial design the number of possible designs N is

I where L = Number of levels for each factor

I J11us, if the qualities of a given product depends on three factors A,

level experiment is represented by Table 1-2.

skewed or intersecting lines would indicate strong interaction between .

Figure 1-1. Factor effects.

Taille 1·3. Two Factor ExperlmentData

industrial organizations and has completely changed their outlook on

2-2 TAGUCW PWLOSOPHY

Taguc.hi built upon W. E. Deming's observation that 85% of poor

it merely treats a symptom. Therefore, fIU8Iity coeccpts should be based

The kus of the system design phase is on determining the suitable

2-3 TB CONCEPT OF THE LOSS FUNCTION

The conventional method of computing the cost of quality is based on

LAl = lower allowable limit

Figure 2-1. Taguchiand conventional loss functions.

Figure 2-2. Color density distrib~

B wore much faster thantbose from souroe A. Topiapointthe cause of

• The quality loss function is a continuous funetico and is a measure

Poor Wear Ch8lllW!r

2-3. Bearing diameter distribution.

best be minirniJa by designing quality into the product. Prevention

2-4 EXPElUMENT DESIGN STRATEGY

Table 2-1. 0rth0g0DaI Array £,<2')

in the row. The vertical columns correspond to the factors specified in

2-5 ANALYSIS OF RESULTS

In the Taguchi method the results of the experiments are analyzed to

I. To establish the best or the optimum condition for a product or a

The optimum condition is identified by studying the main effects of

most. robust set of operating conditions from variations within the

The Tagucbi Approach to QQdty_ CosIImprovemcfa 17