qfd: validating robustness - university of...

Quality Engineering, 11(4), 593-611 (1999)

QFD: VALIDATING ROBUSTNESS

Kinnar K. GhiyaIntel Corporation

5000 W. Chandler Blvd., CH7-306Chandler, Arizona 85226

A. Terry BahillSystems and Industrial Engineering Department

University of ArizonaTucson, Arizona 85721-0020

William L, ChapmanRaytheon Systems Company

Tucson, Arizona 85734

Key Words

Quality function deployment (QFD); Concurrent engi-neering; Sensitivity analysis; Total quality management(TQM).

Introduction

Quality function deployment (QFD) started in Japan inthe late 1960s and is now used by half of Japan's majorcompanies. It was introduced in American automobilemanufacturing companies in the early 1980s and is beingused by numerous American corporations (1,2). QFD is ahandy tool for interdisciplinary teams. A typical QFD teamwill have members from marketing, sales, manufacturing,design, quality control, purchasing, and so on. "QFD en-hances communication levels within the core team" (3).

Quality function deployment strives to get the customer'sview of quality introduced in the early phases of the designcycle and considered throughout the product's entire life

cycle. "QFD therefore represents a change from manufac-turing-process quality control to product development qual-ity control" (1). In most implementations, QFD uses manymatrixlike charts to discover interrelationships among cus-tomer demands, product characteristics, and manufacturingprocesses, as shown in Figure 1. For example, the first QFDchart compares the customer's demands to quality charac-teristics. The second chart then investigates the relationshipbetween these quality characteristics and characteristics ofthe product. The third chart subsequently examines the re-lationships between these product characteristics andmanufacturing processes. Finally, these manufacturing pro-cesses are compared to the quality controls that will bemonitored during manufacturing.

ToothBrite Inc.: A Heuristic Case Study

In order to analyze QFD as a tool, we need an exampleto study; therefore, we will now present ToothBrite (4).Assume that you are the Chief Executive Officer of

593

Copyright © 1999 by Marcel Dekker, Inc. www.dekker.com

594 GHIYA, BAHILL, AND CHAPMAN

QualityCharacteristics

CustomerDemands

Phase I

QualityCharacteristics

ProductCharacteristics

Phase II

ManufacturingProcesses

ProductCharacteristics

Phase III

ManufacturingProcesses

QualityControls

Phase IV

Figure L Relationship of the four QFD charts.

ToothBrite Inc., a major toothpaste manufacturer, and yourmarket share has suddenly dropped, You suspect that thisis a result of your competitor's new innovation. Crest® hasdeveloped a new toothpaste container called the NeatSqueeze dispenser and has endowed it with a substantialadvertising budget, (To understand this example better, youmight cut open a Crest® Neat Squeeze dispenser or aColgate® Neat and Easy Stand Up dispenser and see howit is produced and what is inside.) To recapture your mar-ket share, you decide to redesign your product; therefore,you decide to use QFD as your analysis tool. We had ourmarketing department interview all people we thought couldprovide inputs for the system design. In the QFD literature,the aspects deemed important by the customer are calledcustomer demands. Our marketing department derived thefollowing customer demands:

NeatnessTidy tip: The tip stays neat and clean.

Retains shape: the container retains its original shape.Stays put: The container will not roll off the counter.

Hygienic: Toothpaste cannot touch the brush and then bedrawn back into the container.

Squeezable: People want to squeeze the container, notpump.

Easy open: The cap opens and closes easily.No waste: Almost all the toothpaste comes out.Small footprint: Container takes little counter space.Reasonable cost: It should cost about the same as present

containers.Attractive container: The sales department says that it

must look good.

After listing the demands, the customer assigns a weightindicating the relative importance of each demand. Usually,the weights are between 1 and 10, with 10 being most im-portant. Figure 2a shows the customer demands on the leftside and the associated weights in the right column.

§.CTQ

I

ISa

I

I

2

1

13

3

10

8

11

4

6

8

7

5

12

i

252

315

59

211

96

100

92

120

108

100

104

117

70

Alttractive Container

~

UJ

~~-so

~o

Reasonable Cost

U)

vO

^

SO

so

-

U)

SO

03

mall Footprint

_

so

~

OJ

Lh

•X

o Waste

U)

NO

so

SO

VO

-

VO

ON

Easy Open

U)

\

O-i

^

^

(Squeezable

SO

SO

so

-

U)

00

o'

so

sO

~U)

^

U)

sO

J^

Stays Put

^

SO

^

U)

ON

Retains Shape

SO

SO

so

-

-

u>

U)

•£>

| Tidy Tip

so

SO

OJ

u>

u>

0

?

g f |ill

«t

Suction Device

Dashpot & Air Chamber

Elastic Walls

Rigid Walls & Pump

Plastic Walls

Paste Viscosity Requirement

Fixed Amount Dispensed

New Cap Design

Flat Top & Bottom

Squeezable Top & Bottom

Inexpensive Materials

Simple Manufacturing Processes

Graphic Designer

Tamper Proof Package

Importance (1 to 10)

II11GO

I

3

9

6

10

8

7

5

4

1

2

i

137

72

94

36

76

82

110

126

184

144

(Attractive Container

U)

-

U)

SO

~sO

<l

ja

(fD*

i

«U)

NO

\

C0

T)

»'

\

-

Wi

iNo Waste

UJ

-U)

so

-

ON

S1^I

-

NO

w

^

00

?r

NO

«

Ui

-

oo

1o*

-vO

NO

^

Stays Put

oo

oo

ON

| Retains Shape

-

-

<jj

\

-

-

^>

1 Tidy Tip

SO

U>

-

t^i

SO

0

?o3 O

1 Is*-* f c

Amount of Mess

Amount of Fullback

Amount of Pressure

Amount of Effort

Amount of Waste

Counter Space

Amount of Deformation

Pleasing Appearance

Cost to Produce

Selling Price

Importance (1 to 10)

UiV4D


In this example, our customer is the person who brusheshis or her teeth with the toothpaste. However, "the cus-tomer" should also include all people who should provideinputs for the system design. This includes buyers, storemanagers, mothers, the manufacturer's sales force, the de-sign team, and the production facility. See Chapter 5 of Ref.5 for a fuller exposition of this matter.

Next, we asked our Systems Engineering Department toderive measures to assure that these customer demands arebeing satisfied. In the QFD literature, these measures arecalled figures of merit, quality characteristics, or, some-times, measures of effectiveness. Quality characteristicsshould be quantitative and measurable. The following arethe quality characteristics we used for the ToothBriteproject.

Mess: amount of toothpaste scraped off the tip when halfempty

Pull-back: amount of toothpaste pulled back when donedispensing

Pressure: amount of pressure needed to squeeze out thetoothpaste

Effort: number of turns, or time, or effort needed to re-move cap

Waste: amount of toothpaste left in the container at endof life cycle

Counter space: amount of counter space occupied bycontainer

Deformation: amount of change in shape of containerwhen half empty

Cost of materials: cost of raw materials used to make thecontainer

Pleasing appearance: based on results of customer survey

In general, QFD charts have a desire that needs to be satis-fied listed along the side, and measures or approaches forsatisfying the desires across the top, as shown in Figure 2a.The items listed on the left are called Whats and items listedalong the top are called Hows. To help determine the Howswe ask, "This is What the customer wants, now How canwe measure it?" We will often associate optimal or targetvalues with these measures. These measures become thedesires on the next chart,

The next step in QFD analysis is determining the strengthof the relationships (or the degree of correlation) betweenthe Whats and the Hows. This is done by filling in the cen-ter matrix on a column-by-column basis as shown in Fig-ure 2a. Each What is compared to each How. Four classifi-cations are given: If they are strongly related, a value of 9,or a black disk with a white dot inside, is recorded in the

appropriate cell. Moderate relationships are given a 3, or acircle; weak relationships are given a 1, or a triangle; andno relationship is given a 0, or the cell is left blank. Depend-ing on the customers, symbols and number can be mixed.

The next step is multiplying each cell's value by theweight of the customer demand and totaling the column foreach quality characteristic. This is shown in the row acrossthe bottom labeled Score in Figure 2a. The total score foreach column is an indication of the importance of that char-acteristic in measuring the customer's satisfaction. Typi-cally, measures with low scores receive little consideration.However, this does not necessarily mean that these measureswill not be used in the product design: They still may benecessary for contractual or other reasons. To satisfy thecustomer, we must pay strict attention to the measures withthe highest scores. Focusing attention on the customer is themain purpose of the QFD chart. The chart and its results arenot as important as the process of concentrating on the"Voice of the Customer" rather than the "Voice of theManufacturer." For the ToothBrite project, the Cost to Pro-duce (with a score of 184) and Selling Price (with a scoreof 144) were the most important measures,

Subsequent QFD Charts

To continue our QFD analysis, we will relate the qual-ity characteristics of Figure 2a to characteristics of the prod-uct. One of the purposes of a QFD analysis is to investigatemany alternative designs. However, as the analysisprogresses, we must limit the number of alternatives underconsideration. The characteristics of the product will bedifferent for each alternative design. Thus, if we wish tocontinue investigating alternative designs, then we mighthave to create a second QFD chart for each alternative. Thefollowing product characteristics, provided by the DesignEngineering Department, seem to imply a suction type ofcontainer:

Double lead threads on cap and tip—allowing cap re-moval with a half-turn

Size of dispensing hole in tipThickness of side wallsType of material for side wallsSize of dashpot (the portion of the tube containing air)Viscosity of dashpotTotal weight of the containerSize of the containerPrinting on label—must be colorful and easy to read

These product characteristics now become Hows in oursecond QFD chart shown in Figure 3a, The score of each

QFD: VALIDATING ROBUSTNESS 597

OriginalToothBrite

Phase II

Amount of MessAmount of Pull-backAmount of PressureAmount of EffortAmount of WasteCounter SpaceAmount of DeformationPleasing AppearanceCost to ProduceSelling Price

Score

Rank

03

J3

TiJ

<D

I

9

<NCO

o

HG

<u"3IT!

1133I3

1

Os£

0

«sO

Mat

eria

133

1

1

11

ON

*-•

vo

g.

'""'

i39913

3193

m

^r

-

103n

tMO

CO

33

11

*oOs

CO

o*%

•8>>

1399

1

31

P^r-<N

m

•|

60

00

I

3

1

mr

Os

83

G

o«+-(o8

M

391331

o(N

^f

*£

31oo.S

1

1

93

COvo

^m

c1«<-•0

S.1

19139

oo

(N

<N

1

I

1377294367682110126184144

(a)

ModifiedToothBrite

Phase II

Suction DeviceDashpot & Air ChamberElastic Walls

Rigid Walls & PumpPlastic WallsPaste Viscosity RequirementFixed Amount DispensedNew Cap DesignFlat Top & BottomSqueezable Top & BottomInexpensive MaterialsSimple Manufacturing ProcessesGraphic DesignerTamper-Proof Package

Score

Rank

Dou

ble

Lea

d T

hrea

ds

9

1

1<NOO

ON

Size

of H

ole

in th

e T

ip

33

331

1

«r>00••3-Ol

VO

(Tub

e) M

ater

ial T

hick

nes

333I

393

3rOsmro

TT-

O)PH

£

139991

3

393

3OxmCO

-

Size

of t

he D

ashp

ot

9

31

1

oVOr-<N

«0

Vis

cosi

ty o

f the

Das

hpot

9

9

39

00\m"!t

<N

Wei

ght o

f the

Con

tain

er

3

1

1

COm

0

Size

of t

he C

onta

iner

3

1

119

on

<N

r-

Prin

ting

on L

abel

1

3391

^o

5CO

Shap

e of

the

Con

tain

er

1

111913993

Sor-m

3s&o

I25231559

211961009212010810010411770

(b)

Figure 3. The original (a) and the modified (b) ToothBrite Phase II charts.


quality characteristic as determined in the first chart is usedas the weight in the second chart. The quality characteris-tics become the new Whats and the product characteristicsbecome the new Hows. The question becomes, "This is whatI am going to measure, now How will I build the productto make it optimum?" We fill out this chart using the sameprocess used for the first chart. Fill out each cell based onhow strongly each product characteristic is related to eachquality characteristic. Multiply the weights by the numeri-cal values for the relationships and sum the columns to givethe scores at the bottom of the chart, The column scores nowindicate how strongly each product characteristic is relatedto the customer's demands. For the ToothBrite project, thesescores indicate that the type of material used for the sidesof the container is the most important product characteris-tic. This is an important finding that was not obvious at theoutset.

The third QFD chart, shown in Figure 4a, compares theproduct characteristics to manufacturing processes providedby the Manufacturing Department. The manufacturing pro-cesses are as follows:

Molding process (cap, body, and bottom)—assume ablow molding process

Create moldBlow material—assume polycarbonate material will

be usedRemove container

Insert and bond liner—the bag that holds the toothpasteInsert toothpasteScrew on capWeld bottom—assume use of ultrasonic welding to at-

tach sides and bottomPaste or print label

These manufacturing processes are listed in the approximatetemporal order that we envision. From the scores and ranksat the bottom of this chart, we can see that blowing thematerial into the mold is the most important process.

Finally, our fourth QFD chart, shown in Figure 5a, com-pares the manufacturing processes to the quality controlsprovided by the Quality Control Department. These are thequality control items that will be monitored and controlledduring the manufacturing process:

Mold dimensionsMaterial controls (when material is being injected into

the mold)TemperaturePressureTime

Liner attachment inspectionToothpaste flow rateCap attachment torqueWelding controls (when attaching bottom to the sides)

IntensityDurationPressure

Labeling pressureCleanliness and hygiene controls

As we progressed through this ToothBrite project, the QFDcharts became more and more specific. The fourth QFDchart is very specific to particular alternative materials andmanufacturing process chosen.

This chart tells us that in order to satisfy the customer,during manufacturing we should pay very special attentionto the material temperature and the mold dimensions. Thismay not have been obvious to the manufacturing engineersbefore this QFD analysis.

Generalizations

This process of linking QFD charts together can continueuntil dozens of charts have been filled out (see Refs. 6-8),as suggested by the "Waterfall" chart of Figure 1. For ex-amples of using many QFD charts on one Heuristic ex-ample, we recommend Refs. 5 and 7. For many examplesderived from real manufacturing systems, see Ref. 8, whichis arguably the most definitive work on QFD in the Englishlanguage. QFD can also be applied to the top-level function,then to its subfunctions, and then to their subfunctions.Using QFD to design real systems will involve many, manyQFD charts. Managing such a large database will certainlyrequire computer assistance. Such programs are available(e.g., QFD/Capture by International TechneGroup Inc. andQFD plus by Ford Motor Co.). In this article, we have onlyscratched the surface in the use of QFD charts.

Alternative Approach

The QFD process can become complex and cumbersome,depending on the product complexity. Many times, the de-tailed four-chart analysis is not possible with available re-sources. Many QFD analyses in industry only use the firstQFD chart. In this case, the Hows can have a different defi-nition. Here, the Hows are listed as "how can we solve it"or "how can we meet a particular customer demand." There-fore, we will change the definition of the Hows in the firstQFD chart and compare the result with the original chart.


OriginalToothBritePhase III

Double Lead ThreadSize of Hole in TipMaterial ThicknessMaterial TypeSize of DashpotViscosity of DashpotWeight of ContainerSize of ContainerPrinting on LabelShape of Container

Score

Rank

1ffl

"S" «C3O

$sOS

'SI

IS

£s99

13939

9

§§

2

T3

1

S9999

9

9

I

1

£

'$

5

I33

1

3

3

16.

.S

?«

Inse

rt a

nc

31

11

3

oo

5

o

1I9

3

i7

n.

§

11

1

1

8

Ho$2O

ffl2£

P

39

3

i

3

ON

ON

3

•«

••p

£

O

If

3

399

3MS

4

1324100911494713955

2677430

215017682958

(a)

ModifiedToothBritePhase III

Double Lead ThreadsSize of Hole in the Tip(Tube) Material Thickness(Tube) Material TypeSize of the DashpotViscosity of the DashpotWeight of the ContainerSize of the ContainerPrinting on LabelShape of the Container

Score

Rank

o

IfU3

0

OT

0

Oflc53o£

21a>£j

99

13939

9

o00

s(N

rt

•j-j

1

g

(Y)

9999

9

9

•n

<N

rt

S3'«GOu

OS

^33

1

3

3

0§

^

n§S•gw

,£;

31

11

3

rr-

VO

3d,

3

•OT

,5

1

9

3

0sCN

t-

oH

S>

£!-H

Cfl

1

1

•*

00

sCQ

>

•ac8•£3P

3933

1

3

~

~m

42oo.b

'G

QO.aPI

3

399

VO

i-*

M<D

100224853597813927604368381

211319464034

(b)

Figure 4. The original (a) and the modified (b) ToothBrite Phase III charts.

I5a

og"a

?ak-

NJ

i—

ON

ON

VO

IZ!_

to

CO

1—1o

U)

4»

Ui

GOoo3

1474552

1913625

637875

637875

375957

267354

82269

F 400696

358923

1118542

761184

686380

s?5"

CJQ

s7

!S

1

SO

|

ooUl

S

GfaO

5'I£03O

I

Ui

Ui

so

t

—i

n5

Sc{'00

i£

-

so

U)

*

5"oJ

3uu

1p3

£

SO

so

M

1

5"bP3

P^

S

B.r*

1

-

so

_

_

^i

<B

I£J

w

>— i

UJ

?

£*

1

*OJ

UJ

to

1

gFfFb'a

SO

>— a

«

1

so&9

^

1^" s

I

i..

illi ? s*

??

Mold Dimensions

Material Control

Temperature

Pressure

Time

Liner Attachment Inspection

Toothpaste Flowrate

Cap Attachment Torque

Welding Controls

Intensity

Duration

Pressure

Labeling Pressure

Cleanliness and Hygene Controls

Weights

P"*

to

l

ON

ON

00

-

to

SO

~0

u>

*.

^

5?3

921769

1039230

346410

346410

238932

142695

45333

221305

194757

610819

568107

402395

JX)

I'35B£J00

go

£L

sO

~

atoOJ

G

yo'

froo3

Ui

Ui

so

-

2so

SO

5?

era

§o^•o

SO

Ui

R

a-.o3

1^*

1

SO

SO

O)« f\

w0)

?g,,rQ

-

SO

_

_

0^

oo

{a

0

y.|

Ui

,-K)

-0

cdisB.

so

Ui

Ui

~

0

nI

1a

SO

^,

sUi0

2g5'

OQ

%n

^Q

w*0g

f~ H

rt *"

Mold Dimensions

Material Control

Temperature

Pressure

Time

Liner Attachment Inspection

Toothpaste Flowrate

Cap Attachment Torque

Welding Controls

Intensity

Duration

Pressure

Labeling Pressure

Cleanliness and Hygene Controls

Weights

Offlj?w

rr

i


We have used the ToothBrite example (4) to list the Howsin the first chart as "How we can achieve it" instead of"How we can measure it," The customer requirements arethe same for both charts (refer to Fig. 2). The Hows in themodified ToothBrite chart 1 of Figure 2b offer possiblesolutions for the customer requirements. For example, tomeet the customer requirement of Neatness, in the originalQFD chart, we have Amount of Mess, Amount of Pull-back,and Cost to Produce as significant measures (Hows). In themodified QFD chart, for Neatness, we are offering possiblealternatives—we could possibly have a good suction device,or plastic walls, or rigid wall and a pump arrangement. Forthe Attractive Container requirement, the important mea-sures in the original charts are Pleasing Appearance andSelling Price. For the same What, the modified QFD chartoffers a solution—one can employ a graphic designer. Now,one can look at the scores at the bottom of the chart. Elas-tic Walls is the most important How in this case and, so, thedesigner of the toothpaste dispenser should consider elasticwalls. Thus, this approach offers solutions after the firstchart itself and the scores at the bottom of the chart areuseful to determine which alternative should be consideredseriously. For small tasks, if one wants to limit the QFDusage to the first chart, this approach still brings good re-sults.

We have gone further to study how the effect of chang-ing these Hows propagates through the four charts. Figures3-5 compare the original ToothBrite charts with the modi-fied charts. The results after progressing through four phasesindicates that changing the Hows definition had almost noeffect on the final outcome! Compare the two final chartsfrom Figure 5. After progressing through the whole process,the ranks have hardly changed. Controlling the materialtemperature is still the most important task. This is an in-dication of the robustness of QFD analysis.

Bicknell (9) says the Hows ". . . should be controllablecharacteristics. They can be described in measurable termsto provide targets for specifications and desired productimprovements." Janet Fiero (through personal communica-tion) comments that the Hows of the first chart can be in-terpreted in different ways, depending on the problem inhand. Our example shows that it does not make a lot ofdifference.

Sensitivity Analysis

The next step in our research is to validate the robustnessof the QFD process, by conducting a sensitivity analysis. Wehave used examples from the literature and industry to carry

out the sensitivity analysis. Given the proprietary nature ofthese charts, some of them did not have complete informa-tion needed to carry out the sensitivity analysis. In suchcases, we modified the charts by filling out the missinginformation to the best of our knowledge and experiencewithout losing the original gist of the charts.

Figure 6 shows four charts of the Pencil example (7,10).These charts go through the design of a pencil as a newproduct. Figure 7 shows the four charts of a Door Systemexample (American Supplier Institute, 1992). These chartsare from the development of a car door system. Figure 8shows three charts of a Video Game example (11). We havealso used the modified ToothBrite example discussed ear-lier for this sensitivity analysis.

We have used two approaches to conduct this sensitiv-ity analysis. In the first approach, the correlation weightswere changed throughout all four charts and the effect ofthat on the final outcome was observed. In the second ap-proach, we used a two-level fractional factorial design tovary the customer importance weights in different combi-nations (e.g., for LI6, it will be 16 different combinations)and conducted an analysis to see the effect on the finaloutcome.

Change in Correlation Values

Deciding on the correlation values can be a very time-consuming process, especially when the chart has manyWhats and Hows. The relationships between the Whats andthe Hows are defined to be Strong, Medium, Weak, orNone. These traditionally have point values of 9, 3, 1, and0, respectively. It is important to analyze what effect thesevalues have in the final outcome for two reasons. First, asexplained earlier, the process gets tedious and there may bea tendency to fill up the rest of the boxes with minimumwork and understanding, due to a loss in concentration. Ifthis occurs, the QFD chart may not reflect the outcome asit should be. Second, there may be a conflict in deciding therelationship values. What one team member sees as a strongrelationship may look like a medium relation or weak rela-tion to the other team member.

The scale of 9-3-1 resulted from Weber's law. This is alogarithmic scale with a base of 3. Through discussions, wehave found that some people feel the need of using differ-ent scales and so we did some experiments. We have exam-ined the sensitivity of the QFD method using four differentchanges:

1. Replace the 9-3-1 scale with a 7-3-1 scale: Thisdeemphasizes the Strong relationship.

§lf *-8p'«" gf » Q S1 «S"T 3' £ s !? ~. .

aN>Lfi

5" 5"

CD • >-i £4-tr Z a B[•-• § 3 S1 I - a• ^ o ^

V*

D £*+",CD ' £

CD

CfQ ICD »

p

I

CD

5" to t* 8± p

B g - 8 .CL S \

CD M- ^ pr g.

I ^ i ^ "3 p «> §^ O 'P S 2 »

n 55

^<B SO CL

CD

s -8 .33 CD 7"£ B S•5' ^ 8»5* &> £2.S' «-* P"^ ±r E

CTQ CD g.

S" "cr ^D. g <

•B. | to ^jlJ L l cl

EL 3* r- ^T

& § 3^ CD

CL S O ^*S- BT sr a£ S I. g

"* i-> OOw (TO O.

aa

Q* OOs<j n

§?.>Tj CD

^ & cr

S a I

fB ! &. ^ ™ P5« CD ^ a g. w,

o ^ CL uo CL j^^ 8 >§ CD >§

CTQ &^ )S ^ i-3 PCD % g- 0 ft, 0oo CD ^

cr ^o8- S*CD °

CL f

1O "" <J ^ 00 H-

*-b oo "^ oo Pf oo. O X^ O o ^

^ cT _ cT crq cT5" * a < s ^1 ? I §•• g ?a- a= §H rv4^

§g. U.

8-^-g w^'o,W H- O 2 H-

sp,cT

sa rs

5?

1oo00

u>

o

u>Ji.

£

K>

?

f

O

0

o

o

so

*

S!

0

u>

o

0

K>

$

V!

so

SO

U)

U)

U)

gj

1g"

O

O

SO

O

0

2

E?«

0

0

SO

o

0

Ul

o

sf

0

_

o

0

1

offlHU)

Mold Body

Insert Graphite

Assemble Body

Add Clip

Apply Graphics

Weights

£>3

S

aUi

a

2

U>

a

B1w

0

3S

0

0

o

o

-

SO

VO

U,

g1£0

Tlb

0

o

o

sO

0

o

o

Ul

?wCfl

-

so

w.

o

sO

0

0

Ul

s>£ar/>

8

0

OJ

SO

0

so

-

0

.JSk

wS0

ffloD-

«

0

0

0

o

o

0

w

naj/dH1— i

Length of Pencil

Time Between Sharpening

Lead Dust Generated

Incline Angle Roll

Pages Per Pencil

Pressure Cycles to Erase

Eraser Dust Generated

Importance

o

1

£ s §c S P

3'

oJ3- CDP >-tS O

CTQ OCD l~t'

Op

"p p

s.^5U 5doo ^

II& °CD M

5?3

100

UJ

N)

KJ

Ul

Iu>

oo

g55

io

1p?132.«

o

o

0

o

so

SO

sO

Ul

N)

1n

"a

O

~

0

SO

0

O

0

u>

£i?rGd2

v - i

0

o

0

SO

O

o

0

U)

0

%a

£*•*eu

o

0

SO

o

o

0

0

i-0

1ffl

k<:

so

SO

o

o

_

0

_

SW

0ffi^*rH^A^

H^

Length of Mold

Diameter

Break Strength

Tension of Pocket Clip

Sharpness of Figures

Abraidability of Embossing

Depth of Embossing

Weights

00

S

sO

C?N

NJbO

M_ t

UJ

Mu>

10

S3

sifi1ri

O

iu>

SO

o

o

o

0

K)

?ic

O"<;£•

^

S1U)

SO

_

w

o

o

U)

"0rmSB

« •TJ3S:

sO

0

0

o

so

0

00

g1

i><&•§*

0

o

0

so

o

0

>0

5Tag

-*1so

0

o

o

so

0

Ul

3w&2&00

1sO

o

0

0

o

0

Ul

srs»'TS§

8:

—

o

o

w

o

U)

-

nffi

^{L^ j^wHK)

Graphite

Eraser

Eraser Holder

Body

Paint

Graphics

Weights

O

I

on

o*

00o3

K>

o

3

co

3oo

so

5

1

<£2.DfS'tj

t

O

O

0

o

SO

»

U)00ON

S1.£L

Sw,R'

o

o

0

o

o

«K)U)

s1w

**?oSh

1

o

o

o

0

-

i

00rr1

U)

so

SO

«o

0

Ul

S3*

§•2g&i!j

so

so

so

*

O

0

1

ow"£Ln>a§'3Q-fl

OJ

so

sO

SO

o

0

Ul

O

!"!>l ^5CJHoo

Melt Temperature

Extrude Temperature

Extrude Speed

Cure Temperature

Punch Diameter

Punch Sharpness

Weights

3

U)

00

-J

N)

aia

•n"£.sr

<=>

0

0

o

so

«

O

1S'•I

0

0

o

o

SO

SO

r""<w

$Ui

3O

1

O

O

so

SO

O

-

D

1

o

r>

^

0

o

0

so

o

K>

51w

O

ii,r>R

§.

u>

sO

—

O

o

»

m

n

>-hn

R

8-

SO

_

—

-

O

00

OffiJv^,Y*1^|"T*j

HH— i

Door Closing Effort

Pull Force (Inside)

Dynamic Hold Open Force

Static Hold Open Force

Water Leak Test

Importance

O

sw

GOGO

sn

11-j

sO

0

o\iU)UJ

K)

gsry

*3

{g

o

0

sO

SO

ISJ0

3

gsrni

o

0

0

SO

to

*!

JHB"81

0

sO

O

so

•5

00

ffe.00

nCL

SO

O

O

sO

300

fH,H

1

so

0

0

so

5

OO

1H3

"S

1

SO

O

0

so

NJ

U)

0

OPHTj>M

H

Quality Control Chart .

Preventative Maintenance

Mistake Proofing

Education & Training

Weights

8^

Ul

Ul

Ul

K)

1

U)

1Qir*g

tf

SO

sO

SO

SO

sO

-

to

1o'rna

fTlO

O

O

O

0

0

0

0

•-J

5s3o

CL

-a3

1o

0

0

0

0

o

u>00

2TjO

«

i.^

UJ

««

so

SO

sO

Ul

s>>-i

( J

Sn

CJQmS?a

U)

OJ

w

sO

sO

so

00•-J

o1"''

H"H

^H* .ro

Outside Diameter

Inside Diameter

Shape

Material Specification

Material Density

Vent Hole Diameter

Weights


Table L Results of Changing the Correlation Values for the ModifiedToothBrite Example

Results of changing correlation values for the first modified ToothBrite chartChart:!HOWSDashpot & Air ChamberElastic WallsRigid Walls and PumpPlastic WallsPaste Viscosity Req.Fixed Amount DispensedNew Cap DesignFlat Top or BottomSqueezable Top & BottomInexpensive MaterialsSimple Manuf, ProcessesGraphic DesignerTamper Proof PackageRoot Mean Square Change

Original93180

100

18

66

30

31

29

38

34

31

33

37

22

0.00

Replace931 with 731

79100

20

65

33

35

33

39

38

32

34

41

24

2.57

Replace931 with 531

79

100

22

63

38

40

40

40

44

34

36

47

28

6.50

Replace931 with 930

80

100

15

65

26

3020

38

28

29

29

34

20

3.69

Replace931 with 421

79

100

23

65

38

38

42

39

45

34

37

45

28

6.54

Results of changing the correlation values for the second modified ToothBrite chartChart: 2HOWSDouble Lead ThreadsSize of the Hole in the Tip(Tube) Material Thickness(Tube) Material TypeSize of the DashpotViscosity of the DashpotWeight of the ContainerSize of the ContainerPrinting on LabelShape of the ContainerRoot Mean Square Change

Original931

12

30

44

100

33

53

4

25

23

49

0.00

Replace931 with 731

13

36

50

100

33

51

5

28

26

51

3.16

Replace931 with 531

14

46

60

100

32

47

7

32

29

54

8.20

Replace931 with 930

7

29

42

100

32

54

1

22

19

43

3.19

Replace931 with 421

1342

54

100

34

49

7

31

28

52

5.84

Results of changing the correlation values for the third modified ToothBrite chartChart: 3HOWSCreate MoldBlow MaterialRemove ContainerInsert and Bond LinerInserting ToothpasteScrewing on TopUltrasonic Weld BottomPasting or Printing LabelRoot Mean Square Change

Original93167100201913

4

56

39

0.00

Replace931 with 731

701002624

16

5

60

43

3.74

Replace931 with 531

74

100

36

33

21

7

68

50

10.24

Replace931 with 930

6010016

17

13

0

56

37

3.34

Replace931 with 421

74

100

33

31

16

9

67

47

8.52

Results of changing the correlation values for the fourth modified ToothBrite chartChart: 4HOWSMold DimensionsTemperaturePressure (Matt. Control)TimeLinear Attachment InspectToothPaste FlowrateCap Attachment TorqueIntensityDurationPressure (Welding Control)Labeling PressureCleanliness & Hygiene CtrlRoot Mean Square Change

Original93177

100

33

33

19

13

4

20

18

58

39

35

0.00

Replace931 with 731

85

100

42

42

24

16

5

29

25

64

43

476.99

Replace931 with 531

100

9657

57

31

20

6

46

39

72

48

68

19.10

Replace931 with 930

6610033

33

17

13

0

18

18

56

37

13

7.29

Replace931 with 421

83100

50

50

19

13

4

33

28

61

39

62

11.62


o£§CT

CO

20.00 -r

18.00

16.00

14.00

12.00 -

10.00 --

8.00

6.004.00

2.000.00

-Q-931 with 531

-0-931 with 421

-»- 931 with 731

with 930

Chart: 1 Chart:2 Charfc3 Chart:4

Figure 9. Root mean square change due to global changes in correlation values of the modified ToothBrite charts.

heterogeneous groups (consisting of various engineering andmanagement disciplines), where the group members canhave different styles of thought process.

Change in Weights of Customer Demands

Some QFD professional facilitators state that determin-ing the importance of the customer demands is the most

time-consuming activity in constructing the first chart. Also,what one customer sees as an important requirement maynot be so important for another customer. So, there is aprobability of conflict generation while determining theimportance of the customer demands (Whats). In this situ-ation, it will be useful to discover what effect this can havein the final outcome. Typically, these importance values areassigned on a scale of 1 to 10, 10 being the most important.

Table 2. RMS Change in Scores Due to Changing the Correlation Values for the Other Examples

Pencil Example

CHART 1CHART 2CHART 3CHART 4

Original9-3-1

0000

Replace with7-3-13.11.53.13.6

Replace with5-3-17.83.83.610.4

Replace with9-3-02.60.83.63.4

Replace with4-2-15.72.96.883

ASI Door SystemExample

CHART 1CHART 2CHARTSCHART 4

Original9-3-1

0000

Replace with7-3-1

1.93.72.90

Replace with5-3-16.29.68.30.9

Replace with9-3-06.42.54.41.5

Replace with4-2-17.86.77

1.2

Video Game Example

CHART 1CHART 2CHART 3

Original9-3-1

000

Replace with7-3-1

1.44,43,3

Replace with5-3-13.611.28.9

Replace with9-3-0

1.83

4,4

Replace with4-2-13.39

8.1

QFD: VALIDATING ROBUSTNESS

Table 3. Change in Ranks Due to Change in Correlation Values

607

HOWS of Chart 4

Modified ToothBriteExample - Change in Ranks

at Chart 4

Original Rank at Chart 4Rank after 7-3-1 experimentRank after 5-3-1 experimentRank after 9-3-0 experimentRank after 4-2-1 experiment

HOWS of Chart 4

Pencil Example - Change inRanks at Chart 4


S

•s•sif

4

4

634

S

a33

3

33

fn

1ra

o.o

11

2

21

22

<u

E

11

5

54

55

Ia•s .

1<

7

7

757

Ofl

I

W

0rg

1

55

455

a(S.S3,

S***0

1.070.85

0

HOWS of Chart 4

ASI Door System Example -Change in Ranks at Chart 4


CJ

's,c

cS

Cf21

2

22

(O

"3

£1g3

33

3

3

eol§

|

1

en

£

1

1

32

.S

Q***

000

0

HOWS of Chart 3

Video Game Example -Change in Ranks at Chart 3

Original Rank at Chart 3Rank after 7-3-1 experiment

Rank after 9-3-0 experimentRank after 4-2-1 experiment

1(Ss?«0

55

54

oSo£:>S

22

2

2

Min

utes

.S

iSI•aQ.6'2E

44

4

5

I

1^o

8'?£

66

66

11

1

8

1£.a£

3

3

33

§

1

•S

1g

•J3

«

0*

1.aW)

«Q***0

0

0.58


Sometimes, the scale is restricted to 1-5 as in Figure 6(the pencil example) and Figure 8 (the video example).

In this section, the importance of the customer demandswere altered by (plus or minus) 1 and the impact was ob-served on all four charts. We used a design of experimentsapproach and a two-level fractional factorial design. We willuse the pencil example from Figure 6 to explain this analy-sis. Looking at Figure 6, we realized that the highest num-ber of parameters (either the number of Whats or the num-ber of Hows) in the pencil example is 7. Using a two-levelfractional factorial with eight trials (12), we can study up toseven factors (parameters); therefore, we will use a two-level fractional factorial with eight trials for the pencil ex-ample to conduct this analysis. The importance of the cus-tomer demands are the factors that are changed ± 1 from theoriginal value. For example, look at Chart 1 in Figure 6.Easy to Hold has an importance value of 3. So, for thiscustomer demand (What), the low and high values are 2 and4. Therefore, in four instances, we will have the importanceof the customer demands (for Easy to Hold) to be 2 and inthe other four instances, it will be 4. Table 4 shows the two-level fractional factorial with eight trials for the pencil ex-ample. Because the first chart has only five factors (Whats),the remaining rows represent noise. As we progress throughthis analysis, we will realize that the effect of the noise onall the responses is 0.

The importance column of chart 1 of Figure 6 is (3, 4,5, 3, 3)T. The superscript T represents the vector transposeoperation; that is, we are representing a column with a hori-zontal row. This column changes in the eight experiments.In the first experiment, this column will become (2, 3, 4, 2,2)T, which is column 2 of Table 4, the two-level fractionalfactorial with eight trials. When these numbers are used inChart 1 of Figure 6, the row labeled Score becomes (10,45,39, 18, 65, 21, 18), which becomes the top row of Table 5.Eight experiments are run. The last one uses an importancecolumn of (4, 5,4,4, 2)T, which is the ninth column of Table4 and produces a Score row of (16, 51, 57, 36, 83, 23, 18),which becomes the eighth row of Table 5. The effect Easy

to Hold can be computed by averaging the four values whenimportance was low (10,10, 12, 12) =» average 11, subtract-ing this from the average when the importance was high (18,18, 16, 16) => average 17, and dividing by 2, giving(17-11 )/2 = 3. This is the effect of the importance of Easyto Hold on the Length of Pencil. The other 48 cells of Table5 are computed in a similar manner. As a check on the arith-metic, we note that all entries in the noise rows must be zeroand the rest of the lower portion of Table 5 must be identi-cal to chart 1 of Figure 6.

Now, the results of these eight experiments provide val-ues to replace the Weights column in chart 2 of Figure 6.We will carry out eight different experiments for chart 2, aswas explained for chart 1. Refer to chart 2 of Figure 6. Thenominal weights column is (14, 57, 51, 27, 84, 31, 27)T. Forthe first experiment, this will be replaced by (10, 45, 39, 18,65, 21, 18)T, which is the first row of Table 5. This processwill continue for the remaining charts. Tables 6-8 show thefractional factorial analysis for the remaining charts of thepencil example.

So, how do we analyze these four tables? And what doesthis analysis tell us? Because all four tables are linked anda change in Table 5 will propagate through to Table 8, wewill concentrate on Table 8. Let us consider the factorLength of Mold from chart 4 of Figure 6. The effects ofWhats from chart 4 are the Effect rows of Table 8. Theaverage effect will be the average of these seven entries incolumn 1. So, if you average 729, 243,. . . , 2187, 0,0, thenyou will get the number 1423. Now, the original score fromchart 4 of Figure 6 for Length is 30,618. So the % effect is4.65% [(1423)(100/30,618)]. This means that changing theimportance of the customer demands in chart 1 by ±1 unitproduces an average change in the score for the How Lengthof Mold in chart 4 of only 4.65%. These effects are small.Similarly, the % effect for Break Strength is calculated tobe 3.36%. These are shown in Table 9.

A similar analysis was carried out for the modifiedToothBrite charts, the ASI Door System charts, and theVideo Game charts. These results are shown in Table 10.

Table 4. Two-Level Fractional Factorial with Eight Trials Array for Pencil Example

Experiment Number >»Easy to HoldDoes Not SmearPoint LastsDoes Not RollEasy to Erase

123422

223444

325622

425644

543624

643642

745424

845442


Table 5. Two-Level Fractional Factorial Analysis—FirstChart of Pencil Example

Table 7. Two-Level Fractional Factorial Analysis—ThirdChart of Pencil Example

Experiment Number1 10 45 39 18 65 21 182 10 45 39 36 67 39 363 12 69 63 18 101 23 184 12 69 63 36 103 41 365 18 63 45 18 85 39 366 18 63 45 36 83 21 187 16 51 57 18 85 41 368 16 51 57 36 83 23 18

Effect Easy to HoldEffect Does Not SmearEffect Point LastsEffect Does Not RollEffect Easy to EraseEffect Noise 1Effect Noise 2

3010000

1ew

c-J

0390000

1to

(DCQ

H

0930000

o

"83

0009000

"o

c<-.§~

0990100

1P-.

cu8

fi

0100900

1S&£9£

0000900

1

0

1Q

%

W

We used a two-level fractional factorial with eight trials forthe modified ToothBrite charts, a two-level fractional fac-torial for the ASI Door System charts, and a Plackett-Bur-man design for the Video Game charts.

Tables 9 and 10 indicate that changing importance of thecustomer demands by one point (in either direction) does nothave a large effect on the final outcome. This is an impor-tant conclusion, as the customer can have differences of

Table 6. Two-Level Fractional Factorial Analysis—SecondChart of Pencil Example

Experiment Number1 1468 351 21 255 936 302 1594 675 39 471 954 303 2232 369 23 267 1476 364 2358 693 41 483 1494 365 1980 675 39 333 1170 546 1854 351 21 441 1152 547 1984 693 41 333 1278 488 1858 369 23 441 1260 48

Effect Length of PencilEffect Time Between SharpeniEffect Lead Dust GeneratedEffect Incline Angle RollEffect Pages Per PencilEffect Pressure Cycles to EraseEffect Eraser Dust Generated

319219006300

O

0900

16200

I

0100900

11w

933

812700

tm

01621080900

1

9030000

S

1

Experiment Number

1 2295 14403 9919 765 1035

2 4239 15771 15121 1413 1683

3 2403 21831 12591 801 1125

4 4347 23199 17793 1449 1773

5 2997 19323 14913 999 1485

6 3969 18279 12627 1323 1809

7 2997 19467 15421 999 1431

8 3969 18423 13135 1323 1755

Effect Graphite

Effect Eraser

Effect Eraser Holder

Effect Body

Effect Paint

Effect Graphics

Effect Noise 1

81272772924300

$PQ

'o

361893182181

60300

Oj

ft?1

847955417291872

00

*OJ

'g

1

2799

2438100

0,

G

1

108936

2438100

o

?.&

opinion in what they see as the most important customerdemand. As long as these differences are not too large, theQFD process is robust enough to withstand the change.

In another experiment, which we will report in a futurearticle, we found that the most important decisions in com-pleting QFD charts were deciding whether each strong andeach medium should indeed be a strong or a medium.

Table 8. Two-Level Fractional Factorial Analysis—FourthChart of Pencil Example

Experiment Number H

1 20655 21420 129627 96156 11610 9315 116102 38151 39564 141939 148806 19386 15147 193863 21627 22428 196479 120528 12528 10125 125284 39123 40572 208791 173178 20304 15957 203045 26973 27972 173907 143208 16362 13365 163626 35721 37044 164511 125550 20250 16281 202507 26973 27972 175203 147780 15876 12879 158768 35721 37044 165807 130122 19764 15795 19764

Effect Mold BodyEffect Insert GraphiteEffect Assemble bodyEffect Add ClipEffect Apply GraphicsEffect Noise 1Effect Noise 2

7292432436561218700

1

'o

"S

7562522526804226800

,_

13

3241703716389729542700

fJa2

1

999

7236495087481757700

p.CJ

1b

1

H

1053108351291697200

<8

E0

co

8

1

97281324218772900

Ofl

1*gtf-1o

.0

T3"E3JO

1053108351291697200

00

1v,

Q


Table 9. Pencil Example—Results of Two-Level FractionalFactorial Analysis

Hows of Chart 4->(Refer to Figure 6 - Chart 4)

Ultimate Effect (In %)

"32?'o

§

4.65

11s

4.65

£j-f

00

1P3

3.36

0,

04>

1P-,

O

I§H

4.16

E•s1a

CO

4.54

I1'gW'o

.•s

114,51

00

•iJ=>s'o

t"

4.54

TYzft/e J$. Results of Two-Level Fractional Factorial Analysisfor the Remaining Examples

Results of Changing Importance of the Customer Demands for the Modified ToothBrite Charts

Hows of Chart 4«>(Refer to Figure 5)

Ultimate Effect (In %) 1.7 1.7 1.9 1.7 1.7 1.7 1.80

Results of Changing Importance of the Customer Demands for the ASI Door System Charts

Hows of Chart 4 ->(Refer to Figure 7)

Ultimate Effect (In %) 2.1 2.1 2.2

Results of Changing Importance of the Customer Demands for the Video Game Charts

Hows of Chart 3-->(Refer to Figure 8)

Ultimate Effect (In %) 2.6 2.4

Summary

These experiments suggest that the QFD process is ro-bust to small changes in the correlation values and the im-portance of the customer demands. The most significantchanges were caused by changing the correlation values

from 9-3-1 to 5-3-1: Eliminating the weak correlation orchanging the importance weights by ±1 unit had little effect.QFD brings together members from various departments ofthe company. There are members from marketing, applica-tion engineering, design engineering, manufacturing engi-neering, quality engineering, and so forth. Each member hasa different way of thinking and a different way of evaluat-ing given the diversity of his or her background. Thus, thereare going to be conflicts when it comes to deciding, say,correlation values or the scale for the correlation values inthe first chart. Now, conflict generation is a healthy part ofthe group dynamics, but it can sometimes diminish thepurpose and goals of the group. The above analysis indicatesthat QFD is robust to a certain extent to such conflicts. Fromthe surveys that marketing or application engineering groupscarry out as the first step of QFD process, we may see slightvariation in what customers think is the most importantneed. Our analysis indicated that the QFD process supportssuch variations to a certain extent. So, even with minor con-flicts, a group can utilize QFD outcomes effectively.

Acknowledgments

We would like to thank Jeff Cavanough, CurtHoffmeister, and Tim Stacy at American Supplier Institute(ASI) for providing us the Door System QFD charts. Wewould also like to thank Professor John Ramberg at theUniversity of Arizona for his helpful comments about thestatistical aspects of this article.

References

1. Sullivan, L. P., Quality Function Deployment, Qua!,19(6), 39-50(1986).

2. Hauser, J. R. and Clausing, D. P., The House ofQuality, Harvard Bus. Rev., 66(3), 63-73 (1988).

3. Griffin, A. and Hauser, J. R., Patterns of CommunicationAmong Marketing, Engineering, and Manufacturing — AComparison Between Two New Product Teams, Manag.Sci., 38(3), 360-373 (1992).

4. Bahill, A, T, and Chapman, W, L., A Tutorial on QualityFunction Deployment, Eng. Manag. /., 5(3), 24-35 (1993),

5. Chapman, W. L., Bahill, A. T., and Wymore, A. W., Engi-neering Modeling and Design, CRC Press, Boca Raton, FL,1992.

6. Re Velle, J. B., The New Quality Technology, An Introduc-tion to Quality Function Deployment (QFD) and TaguchiMethods, Hughes Aircraft Co., Los Angeles, 1990.

7. King, B., Better Designs in Half the Time — ImplementingQFD — Quality Function Deployment in America, GOAL/QPC, Methuen, MA, 1987.


CHART: 1

Easy to UseNot Break EasilyVersatile ControlsAffordableGood Visual EffectsEverybody Can Use ItSturdy ControlsInterruptable PlayWide Assortment of ActionNo Eye StrainDon't Get Hurt UsingScore

q00

0>

Bo

CO

9

0

9

9

3

0

0

9

9

9

0

186

0•"-3

I10

0

0

9

9

1

0

0

3

9

3

129

CS

S13Q

Ho

o

q

3

0

3

0

9

0

0

0

3

0

0

78

0>

H

U

13

S«q

'!

0

9

0

9

0

1

9

0

0

01

97

-oP.

1£

>

&

T30

9

0

9

9

3

1

31

1

1

1

125

qO

o"QQqI,8

11

31

3

3

0

90

0

9

9

138

<DO

|

4

3

2

3

5

34

2

5

3

4

CHART: 2

Software DesignScreen ResolutionInfo. Access to Technical Matt.Maintainance and Repair TimeProduct Development ProcessManufacturing OperationsScore

131

q

CO

0

9

0

3

3

3

2241

<D

1

M1S

9

0

0

3

1

0

2090

^

a

£5

0

0

9

3

1

0

1118

tn

^im

unic

atio

n

oQ

=»9

0

9

9

3

0

3624

q

OT3

for D

FM a

n

1•g

0

0

0

9

9

9

3240

0

§

Soq.2•1

&0

013

3

9

1986

J3

I186

129

78

97

125

138

CHART:3

Screen MaterialS/W Planning ModuleUser Friendly Contact# of Communication LinksTraining for DFM and CADDefinition of Mfg. OperationsScore

8o(XDOq

IO

9

0

0

0

0

3

26127

0g

Q

S1

0

9

0

0

0

9

36684

1.S

^q

1

13§0

(20

0

0

9

0

0

32616

1COS-,

O"cS

| •]

<uCO

0

0

9

0

0

3

16020

1

1

5 - V

alid

at

1Q

£

.5

'1

0

0

0

0

9

3

35118

z

1

t"03qoi&

9\

0

3

9

50971

£

£2241

2090

1118

3624

3240

1986

Figure 8. Video game example.


8. Akao, Y. (ecL), Quality Function Deployment: IntegratingCustomer Requirements into Product Design, ProductivityPress, Cambridge, MA, 1990.

9. Bicknell, B,, Total Quality: A Textbook of Strategic QualityLeadership and Planning, Air Academy Press, ColoradoSprings, CO, 1991, Chap. 8,

10. Shillito, M. L., ADVANCED QFD—Linking Technology toMarket and Company Needs, John Wiley & Sons, NewYork, 1994.

11. Bicknell, B. and Bicknell, K., The Road Map to RepeatableSuccess—Using QFD to Implement Change, CRC Press,Boca Raton, FL, 1995.

12. Montgomery, D., Design and Analysis of Experiments, 4thed., John Wiley & Sons, New York, 1997.

About the Authors: Kinnar Ghiya is a Senior MaterialsEngineer with Intel Corporation in Chandler, AZ.

A. Terry Bahill is a professor of systems engineering atthe University of Arizona in Tucson,

William L. Chapman is a Principal Systems Engineerwith Raytheon Systems Company in Tucson, AZ.

qfd: validating robustness - university of...

Documents