Quality Engineering and Taguchi Methods: A Perspective

1 ~ ' ,orasit usedto becalled,qualitycontrol,isa fieldwitha ,qualitycontrolwasmostlyperceivedasa , ,managementis comingtothefundamentalconceptualun-derst andingthatqualitydoesnotnecessarilycomea ta ,qualityengineeringmethods,andqualityman agementprinciples,it is ,qualitymanagementisanintegratingcon-cep tforcost-effective,rational, ,wearehearinga ,engineeringstatis-tics,andqualityengine eringis properperspective,man-agersandengineersm ight commitPerhapsthemostimportantdevelopment isthatinsomecompanies,managementiscoming tothefundamentalconceptualunderstandingt hatqualitydoesnotnecessarilycomeata 'smethodsprovidedinthisarticleissomewhat critical,thepurposeis notto 'sengi-neeringideasareveryuseful,hisstat isticalinventionsareinefficient, ,wehavethefoundationfora ,it is 'Inmoremoderntimes, 'sefforts, ',' ,qualitycontrolwasseenasa wayoffendingoffthedepression, >13havebeenaroundsinceWorldWarII, ,foritiswidelyacknowl-edgedthata majorreasonforthesuccessofmanyJapaneseco mpa-niesistheuseofthequalityim-provement philosophytheyoriginallylearnedfromShewh art'sfriendandlongtimeassociate, BetterWayofDoingThingsWhatisqualityimpro vement?

2 Withouttryingtocorneupwithastringentdefi nition,qualityimprove-mentis operatingonthephilosophythatthereis alwaysa ~better,weusesimplescientif-icmethodsofd atacollection,dataanalysis, ;thefundamentalandper-hapsmostpowerfulto olsforqualityimprovementinmanufacturingi sIshikawa'sGuidetoQualityControl,origina llywrittenforfactoryfore-men'ThelateProf essorIshikawadescribeswhathasbecomeknown astheSevenTools:checksheets,Paretodiagra ms,causeandeffectdiagrams,histograms,str atification,scatterplots, ,whenaproductfailsitalsoproducesinfor-ma tion(thatitfaiiedandunderwhatcircumstanc es) ,thatinformationcanbeusedtogainunderstan dingof"why" "listen" ,weneed"hearingaids" ,manufacturingaswellasservice-related, is widelybelievedthatthesetools,despitethei rsimplicity(orperhapsbecauseofit),havesi g-nificantlycontributedtoJapan' "Taguchimethods"andQualityFunctionDeploy ment(QFD) "TheuseofQFDandTaguchiMethodshasbeeninst ru-mentalinhelpingJapanesecompa-niestoim provequality,reducecost,cutproductdevelo pmenttimeinhalf,andachievemajorcompetiti vemarketadvantages.

3 '"Taguchimethodsundoubtedlyareappliedext ensivelyinselectedindustries,butdatafrom JapanprovidedbyProfessorKusaba,MusashiIn stituteofTechnology,andtheJapaneseUniono fScientistsandEngineers(JUSE)seemtoindi- catethatTaguchi'spersonalcontri-butionst oqualityengineeringhavebeenappliedonlyon a ,Kusabawrites:Asregardstheso-calledTaguc himethod, ,itscomplicatedusagesrepresentedbythepse udo-factormethod, :1)Inspiteoftheiroriginality,thesemethod shavenotbeenprovidedwithsufficientlogica lexplanation;2)Theusageofthemethodsrequi resspecialskills;3)SN-ratioiseffectivein solvingalim-itedrangeofproblems,buttheus eoflogarithmsmakesit difficulttogetthephysicalmeaningoftheerr orscalculated,ordinaryanalytictech-nique sarefarsimpler,havehightestingpower, , ' "designofexperiments."I thinkJapan'ssecretweaponisthecompany-wid e,cooperativeuseofa ,thescientificattitudeascharacterizedbyt hePlan-Do-Check-Action(PDCA)cir-cle, ,ina spiritofgenuinecooperation, ,theseventoolsandmoreadvancedstatistical methods,suchasdesignofexperiments, , , , , >213747510777.

4 ~'81828384.,8081(Fu)2916 191819 15263626 2040 40 445970885578UI6~1819~DY~~~~G~~ 710912210221721182023333634 344052 5 4 516121415111722282537 33367 4 I 8 5 5 8128IS1120232621232013373 412IS136 6 213 13IS2422I2 - I1412 7- 32 4----'---I- -(General)1881413IS141427121427294436594 9 45431059 4 3136 9 9101116 17102018 182 33 1 12 9 4 4 6 2 5 4 7510I 13 3 6 36215 I I I 3 3 9I1 IIII1121122I1- -221I 1(MultivariateAnalysis)3 2 3 3 5 2 4 3 512451712 14231381 I1 6 I 6 6 3 4 7 8 7 4 212-1-2211411342------45278---1---1(Misc ellaneous)36 630294218 14433240-11814101826 5 310- II129 4 43I13243211437233 5 1 4 5 3 5 7 1 42 5 35314 2 I4242231534 4 5 7 6 5 4IS7 62 2 4 2 1 11322131 31232 6 6 1105 1 I 4 2 1 12 4 2 I 2 4 -126 91266065941109795114107130138148 ""treeanalysis&QFD FfAFMEAW eibulldisttibutiondistributiontheorysens oI)'testrelationchaItPERT simulationcomputertechnic&BOPSoth"" ,eressionanalysisofvariancestatisticalte st&estimationoptimization& , ,theMassachusettsInstituteofTechnolo-gy( MIT) "fromBellLabs, , ,andlatergeneralandauthorofthebookAnEngi neer'sManualofStatisticalMethods'" ,"PrincetonUniversity, , ,sponsoredbyUnionofJapaneseSCientistsand Engmeers(JUSE).

5 ReproducedfromProfessorKusaba'sarticle," StatisticalMethodsm JapaneseQualityControl,"SocietasQualitat is, , , ,1 OProfessorof StatisticsatMIT, :1)"TheStatisticalPrincI-plesofExperimen tation,WithPartic-ularReferencetoExperim entationinTextileFactories,"and2)"Statis ticalAspectsoftheControlofQualityinTexti leManufacture."Theselec-tures,whichweres ubsequentlypub-lished, ,afterWorldWarII,Tippettwasagaininvitedb acktoMITtogivea 'smethodsforeffectiveexperi-mentation,de velopedforagriculture,werea fundamentalbreakwiththetraditionofonlyva ryingonefactorata ,herecom-mendedvaryingallfactorssimulta- neouslyinwhatisknownasa ,andthelaterdevelopmentsoffractionalfact orialsandorthogonalarraysbyBritish, Indian,andAmericanscien-tists, 'smethodalsoprovidesmoreinformationabout thesystemstudiedthantradi-tional, 'sex-perimentaldesignmethodsforim-provin gtheverydifficultprocessoftransformingra wcottonofgreatvar-iabilitytoa finalproduct(suchasdressshirts)

6 ,inadditiontocottonindustryexamples,elec tricai,me-chanical, visionarypaperwrittenin1936outlinesincle artermsTip-pett' ,hepointedoutthat"Theneedfortheapplicati onofstatisticalmethodstocontrolofquality inin-dustryliesinthefactthatallindus-tri alproductsaresubjecttosomedegreeofuncont rolledvariation.""Tippettwasincloseconta ctwiththeworkofShewhartandreofersoftento Shewhart' ' , ,whichinthefollowingdecadespublishedFall 198915lecturenotesarepreservedinanintere stingbookentitledTechnologi-calApplicati onofStatistics, devotedtocontrolchartsandotherprocesscon trol-orientedmethods,buttheotherhalfis , , ,FisherwasurgedtodothisbyhisfriendGos-se t,alsoknownunderthepennameStudent,21whow orkedforGuinness' ,fifties, wastooeasytomakea profitotherwise, contrary, theengineerGenichiTaguchiwhofora longtimeworkedfortheElectricalCommuni-ca tionLaboratoriesinJapan."'" ,unfortunately,toooftenrushproductsintot hemarket-placewithoutsufficientdevelopme ntandtesting, signifiedbyfrequentrecallsofdefectivepro ducts-thecostofwhichtheconsumer,ofcourse , 'smethodandlaterdevelopmentssuchasfracti onalfactorials,orthog-onalarrays,andresp onsesurfacemethodsconstituteda major,butlargelyignored, ,andeventhemostrudimentaryap-plicationof Fisher'sexperimentalde-signmethodsis 'smethodsarekeyingredientsinwhathavelate lybeendescribedintheUnitedStatesas"Taguc himeth-ods.

7 "Taguchihasinventedmethodssuchassignalto noiseratios,accu-mulationanalysis,andlin eargraphs,butit is 'sowninven-tionsarenotonlyinefficientine x-tractingtheavailableinformationfromoft enexpensiveandhardwondata, , ' , ' robustcar,forexampie,is a carthatnotonlystartsandrunswellunderidea lenviron-mentalconditions,butperformswel levenwhenthetemperatureis110 Fandthehumidityis90percent,orwhenthetemp eratureis-30 Fandthehumidityis 0 "robust" , assembiedproductscanbemaderobust(insensi tive)tocom-ponentvariation,productiontol er-ancescanbeopenedup, 'smethodformakingproductsmorerobustis ' ,theInaTileCompanyofJapanexperimen-tally foundthatbyaddinganinex-pensivelimeaddit ivetotheclay, ' 'susefulnotionofrobustproductdesign,thes olutiontotheproblemintheInaTileexample,a smanyotherexamplesof"Taguchimethods,"is a straightforwardappli-cationofFisher' 'sMethodsTocarryoutparameterdesignTaguch ioftenrecommendsexp~ri mentaldesignplans, , ,theaverageandthestandarddeviation,aswel lasagoodhomegrown"product,"calledExplora toryDataAnalysis,pio~, , a collectionofideasaboutplottingandgraphin gdatainwaysthatareinformativeandstimu-la tingfortheengineer' commonsenseapplicationoftheideathatapict ureis wortha 'scumbersomeandinefficientstatisticalmet hodsisbet-terthandoingnothingatall,butif wehaveacceptedthecompetitivechallenge,wh ynotusebetterandsimplermainstreamstatist icaltools?

8 Formoredetails,seeBox,BlsgaardandFung."I ndustrialstatisticiansintheUnitedStatesa ndGreatBritainhavesincethe1950s developedmanyusefuI andpractical ' setoftechniqueswherebytheexperimenter,in a step-by-steplearningprocess, ,moresurelyleadstooptimalcondi-tions,and mostimportantly, ,responsesurfacemethodsarenomorecomplica tedthan"Taguchimethods."Partoftheproblem withthepromotionofTaguchimethodsisthatth eyaresupposedtoconstitute"aseparatetechn ology.""Thatim-pliesthattheuseris askedtomakea ,andadvisingtheusertochooseTaguchiMeth-o ds,leadstothedismissalofgoodpracticaltoo lsinventedintheWestsuchasCuthbertDaniel' sNormalPlottechnique."Daniel'sNormalPlot isa (ANOVA)favoredbyTaguchi,itis alsoextremelysimpletoapplyandspeaksdirec tlytotheright(intu-itive) ,Sulli-van" (SPC).However,traditionalDesignofExperim ents(DOE)hashadnoimpactatallun- ,"Letustalkwithdata."27 Infact,the"27thQualityControlConferencef orForemen"heldrecentlyinJapanhadasitsmot to"Makejudgmentaccordingtofactsandactont hefacts.

9 " , ,AppliedStatistics,SuppiementtotheJourna loftheRoyalStatisticaiSociety,Proceed-in gsoftheAmericanSocietyforQualityControl, ElectricalManufac-turing,Technometrics,J ournaloftheAmericanStatisticalAssociatio n,FoodTechnology,ToolingandPro-duction,B iometrics,JournaloftheAmericanSocietyofN avalEng/-neersJournaloftheInstituteofPe- troleu'm,TheIndustrialChemist,AmericanSt atistician,ChemicalEn-gineering,Journalo ftheSocietyofPlasticEngineering,Journalo fAmericanLeatherChemistsAssoci-ationAmer icanSocietyforTestingMat~ria/sBulletin,E lectricalEngi-neering,AnalyticChemistry, RubberAge,Tappi,BellSystemsTechnicalJour nal,BlastFurnaceandSteelPlant,Industrial andEngineeringChemistry,AmericanSocietyo fMe-chanicaiEngineersTransactions,Journa lofQualityTechnology, ,ofcourse, seemshardtobelievethatpeopleovera periodofmorethan40yearswouldcontinuetopu blishexamplesthathad"noim-pactatall."..S owhydidWesternstatisti-ciansinthepastnot succeedinget-I>17tinga largenumberofengineerstousedesignofexper iments?

10 Andwhywas Taguchimoresuccessful?"Quitesimply,theJa panesehaveasecretingredient:theydoit andwedon't."7 Thatis , hectic(oftenbecauseequip-mentandemployee timearetiedupwiththeproductionofdefectsa ndrework). "worksmarter-notharder," 'srevenueandthereforeisnotreceptivetoide asthatmighthavea ,theeconomicbenefitsofquali-tyimprovemen tcanbehardtoquantify,orevenbeunknownandu nknowableasDemingsays.'Thus1believethere asonwhyexperimen-taldesignmethodswerenot usedsomuchintheWestinthepastisunrelatedt omethodology,butin "discovered"intheWestuntiltheJapanesesho wedusthebenefitofqualitymanagement, oftentakesa 'sfactori-als,fractionalfactorials, theprimaryrea-sonforthesuccessof" Taguchi methods."Infact,mostofthecasestudiesofTa guchimethodspub-lishedoverthepastyearsca nbeanalyzedwithsimplermainstreamstatisti calmethodsprovidinga ,seetheanalysisprovidedbyBox"ofQuinlan's " ' timeofseriouscompetitiveproblemsforAmeri canindustryit seemsunfortunatetofol-lowTaguchi' 'sbooksandarticles,forex-ample, ,ontheotherhand, recentarticle,Taguchiwrites:Thereis "proves"thatahorseflycan'tfly, , ,wherea de-tailedscientificexplanationis completescientificcauseandeffectexplanat ionwouldsometimesevenbea ,engineerswillfindthemselvesstuck,unable tomakefurtherprogressif theydon'tknow"why.