CASE S T U D Y Dimensional Variation Reduction for Automotive Bodv Assemblv Dariusz Ceglarek and Jianjun Shi S l~orteningIauncli time, tlie period from no production to full ~roduction,nrl~ilesimultaS. 1 . I 1 1 u 1 c r i 1Ilcscar.cli neously satisfying cluality requirements, is a Cclrlel; Del)n~nre~r.! of~llccl~n~ricnl Elrgilrceri~rgn1rd /111/1licd priority of the inost advanced autolnobile inanufac~llcclronics,Tfic Unir.crsi/y tnrers. Sl~orte~li~lg lanncli tiine goes parallel with imofdliclrign~r,/11zlr A ~ C OJII~ 48109-212.5. c-nznil: ~ r o i i n gquality management, i.e., reducing the daalalek@e~rgi~z.rrcnriclr.c~lr~ opl)oiStunityfor work to be da~nagecla i d sliorte~ling THIS PRESENTS THE the time between defect occurreilce and defect detecMETHODOLOGY A N D F I N D I N G S O F tion [I]. Some studies done by Ayres [2] indicate that A S T U D Y O N T H E ROOTCAUSES 0F D I M E ~ S ~ OVARIATION ~ A ~ IN a sui-prisingly large Action of prodi~ctioncost is diA U T O M O T I V E BODY ASSEMBLY. rectly attributable to the prei7ention of avoidable deSOLVINGDIMENSIONAL FAILU R E ~DURING THE IB-MONTH fects (e.g., quality control), their detection (e.g., S T U D Y L E D TO T H E R E D U C T I O N inspection), or their eliiniilation after the fact (repair, O F DIMENSIONAL VARIATION OF THE AUTOMOTIVE BODY FROM AN I N I T I A L L E V E L O F 8.5 M M TO A BEST-IN-CLASS L E V E L O F 2 M M (6-SIGMA).THE S T U D Y F I N D I N G S I M P L Y T H A T V A R I A T I O N REDUCT I O N A C T I V I T I E S S H O U L D B E ESTABLISHED EARLY IN THE PRODUCT DEVELOPMENT PROC E S S S O T H A T PROBLEMSCAN BE IDENTIFIED AND DuRING PRE-PRODUCT1ON PHASES. rework). Informal esti~natesfroill 17arious sources suggest that lapses in quality control, i.e., design, inspection, ren~ork,repair, and warranty, may account for 40°/o or inore of total cost [2]. One aspect of wllicle quality is tlie dimn~siol~al integrity ol tlie ai~tolnotivebody (body-in-nrliite), nliicli llas great elfects on tlie quality and fimctionality 01 tlie veliicle. In automotive body assembly, geoinetrical accuracy is o i ~ eol the most important qnality factors. Variation in geometrical accuracy can 0 1995 AMERICAN SOCIETY OF MECHANICAL ENGINEERS I stclri fro111both tllc dcsign niitl tllc nssc~r~bly of nli nutoclnssificntion of tlic root cnilscs of tlic diliicr~siolinlfailurcs tllnt occ~~rrctl tl~rillgthe last 18 ~ ~ i o ~ iof t l ins ric\v ~iiotivcbody. In fact, tliinc~isiolinlvnrintio~lis iutroprotl~lctdcvclop~iic~it cyclc, \vl~icllwas cq~livnlc~it to tlic d~iccdinto virtunlly every tlcsig~~ \ ~ I I C I tlic I dcsig~iis period ~iccdcdto rcncl~t l ~ ctlcsig~l'si~ilicrc~it lcvcl of n~nnt~fncturctl [3]. Bccnl~scsoinc mnnc~fncturilignl~tl protll~cttli~~~c~isionnl vnrintio~i.Tllc pnpcr provides n tlcsig~ii~lducctlvnrintioii is incvitnllc, it is iniportnlit to l~cttcrundcrstnndirig of tlic nrca of nssc~~ibly ~norc 11nvca rr~ctllotlologyfor idclitifyilig root cnuscs of tli~nc~isiolinl \~nriatioli,as \vcll as tl~oroilgllly1111dcrstnnt1- prolic to clinicnsionnl failures, ant1 providcs rccolnmcndntiolls for fi~rtlicriniprovcrnclit. Adtlitiolinlly, it aims iilg t l ~ cSO~ICCCSof variation during tllc lnilncll of n ~ic\\. to liclp tllc rcscnrcl~co~nniuliityfocus 011 tlic critical product. nsl>cctsof cliliicl~siorinlvariation by bcttcr ~itilizi~ig CSCurrc~ltly,lliost COIIIIIIOI~ tccIiiiiq~~cs I I S C ~for ~ I I I isti~ig a~ialj-tical ~iictl~ods, or 011 dcvclopi~ig Iilorc cffccproving product quality arc 1):iscd oil Statistical Proccss tivc npl~roncllcs.Tlic rcsi~ltsof this pnpcr arc based on C o ~ ~ t r (SI'C) ol tcclu~iql~cs. SPC tccl~~~iqiics nrc used studies co~~ductcd in an nutoliloti\.c asscnibly facility during tlic ~ i i n ~ ~ ~ ~ f process n c t ~ rand i ~ garc bnscd on tlic \vlicrc sport utility vcl~iclcsarc riialiufact~ircd.Tllc statistical a~ialysisof ~i~cnsurcment dntn. SPC tccliannlysis of t l ~ c root carlscs of dil~~clisiol~nl faults ocII~~LIC nttc~ript S to ~ ~ l i d c r s t ntllc ~ ~critical d mrinblcs ill cirrrctl d ~ ~ r i t ~ l ~ i c g prc-productiol~ pllnsc (pilot procncll scqircl~ccof a ri~n~iufacturirl,nri~g proccss :uld csti~~iatc g m ~ n ) In~lncl~, , nlid full p r o d ~ ~ c t i 011 o l ~ olic nntl two IIO\\. tllcsc vorinblcs i~itcrrclatct l ~ r o i ~statistical gl~ snnlsllifts. Follo\viiig tliis introd~~ctioli, t l ~ c pnpcr is tli\.itlcd pli~ign ~ l dcspcl.il~lcrltntio~~ [f]. I-Io\vc\.cl; as stated in irito tllrcc scctio~~s. Tlic followirig scctiol~ pro\.itlcs bnck[f], SPC tlocs not csplain tllc actunl cnuscs of tlcfccts grour~d011 tlic nutoniotivc assc~nbl!- ~)roccss,n ~ illus~ d during tlic proccss. trntcs tlic product dcvclopmcnt pllascs. Tllcl~,\vc SOIIIC soorccs of asscrillly dcfccts arc givcii 1))prcsc~ltour ~ ~ ~ c t l ~ o d ofor l o grctlucing y tllc di~rlcl~sio~~nl [5]. Oric sourcc of nsscii~blydcfccts is iiitcrfcrcncc \~nriatiollof t l ~ cproduct. Follo\\.ing t l ~ crnetl~odology I)ct\vcc~~ rnating parts. Assc~l~bly tlcfccts can also be scction, \vc rcport on n ficltl study, summarizing root cnilscd by i ~ l s t a l l i ~n ~part g in arl irlcorrcct positio~~ or cnilscs of di~i~cxlsio~inl vnrintio~~ n~idprcscnting oilc iloricntntio~~ [5]. Solnc st~itlicsIlnvc bccn co~ldtictctlto lustmtivc cnsc sttitly. In tllc filial scctiol~,\vc discuss tllc n~inlyzct l ~ crclntionsliip bctwccn nssclriljly proccss mid i~riplicntiolisof t l ~ cwork nr~ds ~ ~ ~ n ~ r i ntlic r i zconductcd c product dcfccts in nssc~i~bly of clcctro~iicsdcviccs [GI. , stl~tlics. Currc~ltly,in n~~ton~otivc intliistrics quitc n fc\\. studics rclntctl to d i ~ ~ ~ c n s ifaults o ~ ~ nIlavc l bccli conductctl [7-IG]. Solllc of tlic~nn~inlpzctlic gcllcrnl cngi~iccri~~g I~UTOAIOTI\'E130DI': DESIGN, ASSEZIIR1,Y I'ROCESS, AND PRODUCT nntl ccorior~iicaspects of dilnc~isionnlfaults [7,8].,0t11DE\'ELOPI\IEI\I' I'IIASES crs n~inlyzcn spccific part of tlic product dcvclopmcnt h s s c ~ ~ ~of b ltllc y nutoliiotivc l)ody, sl~o\vnin cyclc, usunlly durii~gtlic full productio~~ pllnsc [9] or Figurc 1, is dol~~il~ntctl by gco~llctricnlrclntio~~s, so that ~ ~ C S C aI Igclicrnl ~ npproncll for r c d ~ ~ c tof i odi~licl~sio~~nl ~~ tlic tludity of tllc nssclrlbly proccss cnli be dctcnliincd vnrintiol~[lo-lG]. To t l ~ cbcst of our k~~o\vlctlgc, 110 I)y di~llcilsiol~al ilitcgrity of tlic product. This liicnlis col~~l)rcl~c~lsivc study clnssifyi~~g tllc root causes of t l ~ c tlint tllc lcvcl of product t l i ~ ~ i c ~ i s ivnrintiol~ o ~ ~ n l call bc t l i ~ ~ l c ~ ~ s fniltircs i o l ~ n l d ~ i r i-~tilg~ c~ v l ~ oni~to~~iotivc lc body n critical iridcs for final cvnluntio~~ of tllc nssc~~il)lctl dcvclol)~i~cl~t cyclc I~tivcbcc~lcol~ductcd.Currc~~tly, tllc vclliclc. kr~o\vlcdgcnbout rclntio~~s hct\vcc~~ tllc d i l l ~ c ~ ~ ~ i o l i n l vnrintio~~ of tllc vcliiclc and its f~u~ctio~inl ~)crforr~~:i~~cc, ns \vcll ns n s s c ~ ~ ~linc b l y fnilurcs during p r o d ~ ~ c tarc i o ~ ~ Proclucl n ~ l d1'1-occss i ~ o vc1-y t clcnrly u~~tlcrstood. This lack or cnitlcrstnntli~~g An nuto~~~otivc botl!. is I)uilt fr0111sllcct nlctnl results ill \vastcd cffort cluriug dcsigl~,I n ~ l l ~ cnntl l ~ , propacts tl~ntI~avctliffcrc~~t sllnpcs, sizcs, nricl tliick~~csscs, clirctio~~ of IIC\V vcl~iclc,I)ccn~~sc t l ~ c\vork tlocs 1101 clcpc~~tlil~g 011 their f i ~ ~ ~ c l i o Tllc i ~ sparts . arc cntcgorizctl tnrgct tlic ~llostcritical proccss/~)rotliictissllcs. ns structuml or 1io11str~1cti1ral. Strlicti~rnlparts sillq)ort Tllis pq1cr trics to fill tllis gap by l)rcsc~~ti~ig n tlic n~rto~liotivc body stri~ctiircns (1) 11lai11parts, S I I C ~ I MANUFACTURING REVIEW ' VOL. 8, NO. 2 J U N E 1995 Fig. 2 011rl;tteof ~ I I CnsscnrDIJ.pmcess: cri~ictrlslotiorrs of /Ire fig. 1. 7%e Docb. coonlirrnre sj.dern rcilh n~re.rnnrl,lc of serrsor locnlions. fnrtt~ir/glitre. as rails nlld ~)lcri~iiii, or as (2) rciiiforccliic~~t parts, for csal~iplc,door llil~gcrcil~forcclnc~~ts. Tlic otlicr parts arc callcd nonstr~ictirralparts, for csar~iplc,tloor outcr, co\vlsidc, roof, nrltl so 011. Bascd 011 tlrc rcscarcll coliductcd by [13, 171, it call 1)c co~iclildcdthat str~lctural parts, ~rinirrlyduc to tlicir grcatcr rigidity, Ila\-c 111ucl1 grcatcr ilripact 011 nutoriiotivc I)otly dil~icl~siolial accumc!- than ~iol~str~rcturnl parts. Tllc ai1tolnotivc botly asscr~iblyproccss il~cli~dcs tllrcc lllajor clc~l~cl~ts: ( I ) nsscliibly fisturcs, (2) \vcldillg robots or \\-cltliilg guris, nlitl (3) part Iiar~dlil~g 11iccI1arlisliis. Tlicsc tllrcc clc~iic~its co11t111cttlrc tllrcc major fi~l~ctiorrs of tllc nssclilbly proccss: part positioliil~g,part joilril~g,arid ~~art/subassc~r~l~ly tmlisfcr. 011n typical plant floor, tllcsc clcrr~c~rts arc lillkcd .irlto asscrrll)ly lilics tlo~iii~iatctl by pnrt positiorlilrg ant1 \vcltlillg stntiorrs. Eacli nssclnbly linc protluccs a lrrajor s~lbasscmblyor fil~alprotluct. Tllcrc arc usunll!- two typcs of lincs: (1) C O I I I I ) O I ~ C I Islrbnssc~nbly ~ lincs, nrltl (2) fralliirlg asscl~iblylilics. Tllc c o r i ~ p o ~ ~subassclnc~it 1)1!- lilics arc: lrlitlcrbotly lilic, and two al)crturc lillcsIcft- mid riglit-l~alitl.Tllc liiajor clc~~lcrit of tllc frall~il~g lint is all a~~torllotivc hotly fr:i~iiil~gstatioli, \vlicrc u11tlcrbotly, al)crturcs, nlitl roof Lo\\-s arc joil~tcdtogctllcr. Figure 2 sllo\vs nil outlillc of nil assclllbly lil~c\vitli dcscriptioils of tllc gcollictrical statior~sill tlic fralrrillg lilic. Tlrc frillilii~gstatioll is 111nrkct1as S3. 111gc11cra1,all stntiolis call l)c tlivitlctI illto gcoll~ctric;rl1u1t1rc-spot statiolls. 'l'lic gcorrrctrical statiol~sIlii~itllcIIarts \\'it11 p0sitio11-scttiligIIICCII~II~C;IIjigs dcsig~icdf01' cncli t y c of vclliclc. 'rllc part is sct ill ~)ositiori11y tlicsc tlctlicatctl jigs, illitinlly \vcltlctl, t11c11tral~s~crrcd to n rc-sl~otstntioil for fil~al\vcltli~~g. Durillg tllc rc-sl~otopcratiol~s, robots add \vcldil~gspots to tlic parts to iricrcasc tlicir strcllgtli. Rc-spot statiolls usr~allydo ~ i onffcct t tllc positioil or oricrltatiol~of tlic part or s11bassc1iibly.Bnscd oil tlic rcscarcli dolic by [13], tllc gcolrictricnl statiolis Ilavc a I I I ~ I C ~biggc; I illlpact 011tllc tli~iiclisior~al variatioii of tllc protluct than rc-spot statioris. Tllc total liurlrbcr of gcol~ictricalstntiolls varics fro111~ ~ O C C StoS ~ ~ O C C S S0.11avcmgc, at lcast 25-35 gcoliictrical stntiolls alid no lcss tliari 30-40 rc-spot stntior~sasscniblc all nl~tolnoti\'cbody iiintlc of 150250 slicct nictnl pnrts. T l ~ cuiijor fulictiori of tlic nsscliibly statioli is to corrcctly p ~ s i t i o pnrts l ~ bcforc nclding opcrntiol~sstart. Tllc usonl ~~ictlrotl of positiol~iligstalnpcd parts is bascd on tl1c3-2-1 Inyo~rtprilrciplc [7]. I.lo~vc\-cr,d11ri11g stutlics of actual asscl~~bly proccsscs, it was also l~otctl that nn n-2-1 layout prilrciplc, \vllcrc 11 > 3, is oftcli 11scd. Tllc sclcctio~iof tlic part locatilig prirlciplc dc1)critls OII tlrc sizc nlitl rigidity of tlrc pnrt [18-19], and tlilrctl!. affccts tlrc ilillcrcr~tIcvcl of protluct ~ ~ I I I C I I siolral variatiol~. Part tmllsport to and fro111statiolrs is ~ C C O I I I plislictl I>yirlnliy tlilfcrcli~~ncclianis~ns, s~lclias clcctrifictl ~l~or~or;lils, tip-1111tlrivcs, mid so 011. Dctailcd tlcscril~tiolisof otlicr a~rtoli~otivc body asscliibly proccsscs cilu I)c fourltl ill [12]. Alcns~~rct~~e~~t I\lotlcnr tli~~~c~lsioirnl dingl~osisof tllc a11t0111otivc I~otlyrcq~lircsacclrlatc tlntn to satisfy quality assllraircc rcq11ircl11clits.Currc~itly,tllrcc rlicasilrcrllcllt gngcs arc 111ostoftcli irsctl for cllcckil~gpnrt/s~rl~assc~iil~ly tlilncl~siolls: (1) Iinrtl p g c listurcs nlitl layout plntcs [8,20], (2) Coordi~~;~tc XIc;~silril~g hlncllil~cs(C;\IXIs), air11 (3) CEGLAREK AND S H I . D I M E N S I O N A L V A R I A T I O N REDUCTION Optical Coordiantc Alcns~rrilighlncl~i~ics (OCI\II\Is).T l ~ c ~)rcsc~~tctl stlidy rnninly usccl tlntn fro111CXlXI n~itl OCAIhI gages. . Procluct D c r c l o ~ ) ~ i i cI'linscs ~it Onc of the clinrncteristic fentilrcs of t l ~ cnutoiiiotivc ilidustry is tlic fretl~~c~icy of 111otlclc l ~ n ~ i and g s tlic vast nliiolllit of ti~iica ~ i t1nl)or l rctluiretl to ~iinkca c l ~ n ~ ~ g o v'fllc e r . a ~ ~ t o ~ ~ i oI)otly t i v cnssc11il)lyis rcgcnrdctl as tlic lcnst flcsil~lcprocess ill tlic ovcmll veliiclc nsscliiCoorcli~iatcAlcasl~ri~ig Alacl~i~ics 1)ly process [23]. Duri~iga ~iioclelc l i a ~ i gtooli~ig , nii~st Introducetl t111ri11gtlic '19605, CXlAls arc vc1-y ocI)c clin~igetlto ~iiatclitlic 11cn.lycstnblisl~ctlproccss a~itl cllratc and flcsil)lc. Tl~cy'allo\vfor Iiicasllrenicl~tsat product dcsig~i.Givc~itllc co~iiplcsityof tllis proccss, sclcctcd points 011tlic part \\.it11nli ilccllmcy of 0.01 tlic alito~iiotivcbody dcvclol)~i~c~it c!-clc rcquircs tlirce 11iIn. TI~eirflcsiLility is Lascd on tl~circapacity folto follr ycars of lead ti~iic. Eve11 tllc filial slngc of the quick r c p r o g r a ~ ~ i ~llieasllrc~iic~it ~~i~rg cycles \vl~cncl~cckal~to~iiotivc botly d c v c l o p ~ ~ i cycle, c ~ ~ t after vcl~iclca ~ ~ d i l ~ gdiffcrcrit parts or points. T l ~ cdra\vl)ack of CXIXls is lrard tooli~~g arc: dcsigl~cd, lasts OIIC ycnr. This final tlic ti~ilclag Lcforc rcccivi~~g llieasllrc~llcliti~~foniiation. stage i~lclr~tlcs t l ~ cfollo\vi~igpl~ascs:(1)pilot pmgra111, 1111 nl~to~notivc body 111l1stLC take11off-1i11cautl tmlls(2) ~ ) r c v o l i ~productio~i, ~~ic (3) Ialr~icli,and (4) 11111 fcrrcd to tllc ChII\l roo111 for ~~icasllrclliclit. 11f1111yopcrproductioli. Tlicsc pl~ascs nrc dcscriLccl ill thc follo\vi~ig atiolial Chlhl call Ilicasllrc up to eight al~to~~iotivc pamgmplls, foci~sing 011 t l ~ c di~iic~isio~ial ~SSIICS. Lodics d l ~ r i r ~ciglrt g 11o11rs.Tlic rcsl~ltsfro111tlic Chlhl arc 11scf111for asscssi~igt l ~ cfiual cll~alityof an alltornoPilot Progl-a111 tivc body or si~basscnillics. Tlic pilot progrnni is when prototype vcl~iclcsarc processes aftcr 100% of built to verify ~iin~ir~facti~rir~g t l ~ cvcliiclc d i ~ ~ i c ~ ~ and s i o ltolcm~iccs ~s arc npprovctl. Optical Coorclir~ateAIcnst~riilgAlechincs In rcccnt years, thc i~iiplc~iie~ltntio~i of tlic i ~ i - l i ~ ~ c Tlic pilot progralii is t l ~ cfirst pl~nsc,1)cgi111ii1ig011nvcragc Gvc lnol~tl~s allcad of t l ~ cI ~ I I I ~ C I 11 I . ~ I I C ~ I I ~ C(1) S: Optical Coortli~intchicns~~ring hlnclii~ic(OCXDI) in the tlicdcsigncd PLP (I'ri~~cil)nl Locnti~~g I'oinls) verifying nuto~iiotivci n d i ~ s t qlins provided ncnvopport~~nitics for scl~c~iics for 'tooling Gsturcs t111ri11gt l ~ c~ S S C I I I LproI~ nutoniotivc body nssc~iillydiag~~osis. OChII\Is arc illccss, mid (2) setting t l ~ c proccss cnpal~ility of t l ~ c clcstnllctl in-liric nt tlic c~iclof ~ilnjornsscmlly proccsscs, sigricd tooling. sl~clias framing, side frnmcs, untlcrl)otly, nl~tlso on. (Fig. 2). Each OCXII\I cor~sistsof lnnlly laser sensors Pre-Voluaic Productioo nllowing for nolicontnct Incnsllrcliic~itof a I~otlyor s111)Prc-volu~ilcp r o d ~ ~ c t i starts o ~ i OIIC to two ~ n o n t l ~ s nsscmLly rclativc to tlcsig~~ ~ ~ o ~ n i On ~ l nnvcmgc, l. tlic allcad of ln111ic11,after tlic tooling is sct up nt tlic nsscmIncasurcmcnt cycle takes a lc\v scconds, \\.it11 accllracy bly plant. The goal of prc-volunlc prod~lctio~i is to valiaro1111d0.25 1n111, and stntic nlid dy~in~uic rcpcatobility dntc process cnpnlility, n ~ i~iitially ~ d identify t11c sollrces \\.itliin 6-0 cqual to 0.14 and 0.25 111111, rcspcctivcly of v;~rintio~~. [21]. The OCAIXI gage ~~uensi~rcs fro111100 to 150 poiuts 011 cncli r~iajorassc~liblywit11 100% snnlplc mtc. As a rcsolt, tlic OCXIXI provitlcs ;i t r c l ~ ~ c ~ i cnlllolllit lo~~s La~liicll of d i ~ ~ l c l ~ s iio~~~~f aolr ~ ~ i \vl~icl~ n t i o ~call ~ , Lc used lor asAftcr ( l c t c r ~ ~ i itl~at ~ ~ i t~l ~ cgvcl~iclccall rcncl~ncsc11lLliproccss co~itrol.1\11 OCXlXIs ill tllc plnl~tuse t l ~ c ccptnllc c111nlityIcvcls, tlic lnluicl~pl~ascfollo\vs. 'l'ypisalilc coortli~~atc s y s t c ~callctl ~ ~ , t l ~ cLody coordi~~atc syscnlly, at I ~ I I I I C 1)11asc, ~I ;i 1iu111bcr of prol)lciiis I I I I I S ~still tc~ii,becansc it is msy to usc :uld co~iiparcstlntn fro111 I)c rcsolvccl bcforc a vcl~iclcof tlcsircd quality Icvcl cnli diffcrc~~t gages. Fig. 1 sI~o\\.st l ~ ccoordiiintc nscs a ~ ~ d I)c Luilt. Tlic 1~11gtli of t l ~ c1a1111cll I)liase tlcpc~~tls oil rcfcrc~lccpoiuts ill tlic body coordir~atcsystclli. T l ~ c Iio\v fnst ill1 tliniensio~~al ~)rol)lel~is call l ~ rcsolvccl c \vliilc pri~~ciplcs of t l ~ cllicnsllrclnclit sc~isorsare rlcscril)ctl Ly si~ilulta~~col~sly s p c c d i ~ ~upg the protlr~ctio~i rate. Itlcliti[22], nntl tllc olltli~lcof t l ~ csc~isorsetup for nl~to~iiotivc ficiltio~iof d i ~ ~ ~ c ~ ln111ts ~ s i o I)eco~iics ~ ~ a l O I I ~of t l ~ cLottlcbody assc~i~bly cnli I)c fo1111t1 in [lo]. ~ ~ c c ill k srcacl~i~rg t l ~ cf11l1 ~)rotluctio~l rate. M A N U F A C T U R I N G REVIEW VOL. 8, NO. 2 J U N E 1995 FnII I'mcl~~ctino T l ~ cfrdl protlr~ctionpl~nscI~cgins,nftcr tltc proclr~ctio~i mtc rcacl~cst l ~ nssigncd c Icvcl \vitl~ncccptnblc qridity. Typically, at this pl~asc,nssc~nblylinc maintcllallcc issrlcs al~clprotlllct quality assrlralicc rcqr~irc~ncntsbccol~~c clo~~ii~~a~it. t l ~ ca ~ ~ t o ~ i ~ oI~ody. t i v c Tl~cscpoi~itsnrc c11osc11for ~ncas ~ ~ r e i n cI~ccar~sc ~ ~ t s tl~cyarc t l ~ clilost critical points on tllc I)ody, cnllccl KI'C (Kcy I'rotlrlct Clinractcristics), or KCC (Kcy Control Cl~nmctcristics). T l ~ cKPC is n protl~lctclrnractcristic for \vl~icl~ di~nclisio~inl variatio~~ c o ~ ~ fsigliificantly tl affect n protlr~ct'ssnfcty or coinplia~~cc, as \vcIl'as cllstolilcr satisfactio~iwit11a product. Tlic KCC is n proccss pamlllc\'tlRMBILI'A1 COhTROL OF PRODUCT tcr for \vlticll tlic variation ~ i i r ~bc s t co~itrollcdto clisrlrc CEOAIETRY 1)URIKC ASSEfiIBLY t1i;it \.;rriatio~iill a KI'C is ~ i ~ a i ~ i t adi u~ r~ic~ g asscl~i- d~tllc I'ROCESS-I\IETIIODOLOGY bly proccss. For csa~i~l)lc, Illcnsurc~iic~it ~)oillts4, 17, Tlic tli~liclisionalquality of a product is directly 18, a~itl19, \vliicli tlefilic tlic frolit cloor opc~iing,arc rclntctl to t l ~ variability c of tlic proccss that crcatcd it. rccog~iizcdas KI'Cs. Tlicy colltrol tlic door-fitting proDi~i~ciisio~ial ~nriatioiicall IN usctl as nli i~itlcsto c \ . a l ~ ~ - ccss in tlic final vcliicIc. Tlic I I I C ~ S I I ~ C I ~ ~ CofI I II'L S I' loatc tllc di~iiclisio~inl quality of n prodrict. I h t a~ialysisof cators nrc t~suallyrccogiizctl as KCCs. ~ ~ O C C Svariability, S cspccially ill ii~tl~~strial practice a11t1 Tlic KPC and KCC arc sclcctctl during t l ~ cpilot rcscarcl~,can be a vcry co~iiplcsproblcni. 111protlucprogmm to dcscribc tlic d i ~ ~ i c ~ i s i ocl~arnctcristics ~ial of tioli, a ~ n c a s ~ ~\,ariati011 rcd of product colisists of a sct of tlic protluct and proccss, rcspcctivcly. r\ para1ilctcr dcsmallcr \.ariatio~iscauscd by ~ i i a ~diffcrclit iy factors scribing tlin~c~~sioi~al \.ariatioil of t l ~ a~~toniotivc c botly cnllcd root causes. Once t l ~ cparticl~larvarintion of n is tlcfi~icclns 6-sig~i~a sta~idartldcviatio~icalc~~latcd for protluct is clctcr~ninctlIly mcas~lrc~ilcllt data, it is inicncli liicnsllrcliicilt poilit \\-it11 n given sa~nplcsizc. Tlic portaat to n~ialyzcit aiid llrcak it dowri to tlic i~idivid6-signin for cncl~Iiicnsurcnlclit poilit call Ilc prcsc~itcd ual root causcs tllnt niny Iinvc occurrcd during tlic ill tlic for111 of n pcrcc~~tilc clinrt (Figurc 3), slio\ving tllc proccss. Tlic proccd~vcof tlctcr~liiningan individ~~al distribution of tlic cli~iic~isio~inl vnrintio~it l i r o i ~ g l ~ tol~~~ct root callsc of di~iiciisio~ial varintio~iis prcscntcd in tliis Incasrlrc1nclit:sclisors.Tlic f i ~ ~i~itlicator al of protluct papcr as a casc stutly. variatio~~ lcvcl can IIC prcsc~itctlas a C o ~ i t i ~ i 1111~~o~~s Tlic ~nctl~otlology applictl tluri~igtlic study call bc pro\.cnidiit I~ltlicntor(CII). Tlic Cil is sclcctctl as all r~pdivided into two parts. Tlic first part is usctl to c v a l ~ ~ a t c per liniit of 95% of nll 6-sign~nscalcl~latctlfor all or "tmck" t l ~ cIcvcl of cli~iic~isio~ial \.ariati011 r ~ s i ~ n~g ~ l i c a s ~ ~ r c ~points n c n t of tlic protlnct. For ccsnl~iplc,ill dcfincd variation indicator. Tlic scco~~tl part tlcscribcs Fig. 3, tlic CII i~ltlicntorslio\vs vnrintiol~\ ~ i t l i i 4.5 ~ i 111111 tlic casc study groccdurc riscd to identify nl~tllocalizc (6-signin) for 95% of olcnsllrclncllt points. root caliscs of dinicnsio~ial\.arintion bnscd on t l ~ cIncaTlic CII can bc ~isctlfor fnst tracking of product sorcnicnt data. Dot11parts arc dcscribcd in tlic followi~~g scctio~is. Trncki~igthe D i ~ i i e ~ i s i o ~\'nriniior~ inl Level Tllc tracki~igof di~~ic~isiolial varintio~~, L I S C ~as n q11nlityi ~ ~ t l c~~cccls s, to slio\v t l ~ cIcvcl of \.nrintion a ~ i d the ti~iicof occ~~rrcIicc. It is i111porta11~ to track varintio~l Icvcl tliro~rgl~ tliffcrc~~t ~)liascsto k n o \ ~[lot only tllc currc~itqunlity Icvcl of tlic prodnct, or acl~icvcdi~iipro\.c111c1itin ti~iic,b l ~ also t to IISC it as a 1)~11cl111iilrki1~g iridcs allo\viug for co~iq)nriso~i of tliffcrcnt protlucts or tllcir d c v c l o p r ~ ~cycles. c~~t Tlic di~i~c~isio~lal variatio~iIcvcl is asscssctl basctl 011 t l ~ ~licasurctl c p o i ~ ~oil t s t l ~ cprotl~rct.Fig. 1 slio\vs nil c s a ~ ~ i ~of) l~c~ i c n s ~ ~ r pc o~ ii~i loci~tio~~s c~ ~t ~ t 011 o~rcsitlc of CEGLAREKA N D S H I D I M E N S I O N A L VARIATION REDUCTION variation Ic\-el. Tlic prcsclltctl st~~dics' usctl tlic C1I indicator sllo~vriill tlic for111of a clinrt wit11 tlic liorizoiitnl alitl vcrticnl ascs rcprcsclitil~gtlic coliscc~~tivc \vccks of prodllctio~~ n~itl6-sigrna vnriatio~~ for 95% of IllcasllrcI I I C I I ~points rcspcctivcly (scc llppcr part of Figure 4). 0 e = u .s ;. 3 .8 B % y ' B z .g v, c .s & 2 .E ." -5 I U = a Cnsc Stutly ~ \ p l ~ r o ; l ctol ~Diagnose Root Cn11scs of I ) i ~ ~ ~ c ~ ~ s\'ariation ioanl TIic s t ~ ~ tof l yd i ~ i i c ~ i s i ofaults ~ ~ a l rcquiccs tlic usc of n ~~~ctliodolog!that will allo\\. for tlic itlc~itificatio~~ n ~ i dlocnlizntion of root cnrrscs of t l i n ~ ~ ~ svnrintior~. io~~a~ = -.$ .g .2n em Ez G -2 -2 sa f = u B g § $ v, 5 e e u -$ r5 : Fnture related faults 7Wo 40% MANUFACTURING REVIEW VOL. 8, NO. 2 J U N E 1995 100% c .2 .$ g 2 8 = 5 a 70% A nc\vl!- developed diagnostic ~nctliotlologywas preis basctl on a syssented in [13-161. Tlie ~netl~otlology tenintic, data-driven case study approacl~that e~iablcs qniek tletcction and locnlizatio~iof asscnil~l!- process f a ~ ~ lbased t s oil din~ciisionnl~neasllrell~c~~ts. T l ~ case c stutlics group and prioritize the currcllt variation level into tmccable antl solvable problc~nsin order to rcacli t l ~ ei~~tcndctl variation level. Therefore, n case study failure cl~aracdescribes a body-ill-\vl~itcdinic~~sional terized by a partict~lnrvariatio~~ pattern and caused by one (or several) root cause(s). Esaniplcs of dinlc~~sionnl \.ariatio~~ root causes are: asselilbly fixture-related (KC locator \\.car, inclusions OII the locating surface of locators, or cla~npsthat do not propcrl!. force tlie part against tlie locator), \vclding gun-rclatetl (niissi~~g \veldi~igspot, n~isnligucd \\.cldiiig gt111, \veltling tip \\-car), sta~llpcdpart-rclatccl (iiiconsistcnt press tonlinge, press stroke), arid rnatcrinl I~a~~dlir~g-related (interferences I)ct\vce~ipart I~andling Iocatow arid fixture locators). T l ~ problcm-solviag c rii~diot10Iogy~ I I C I I I ~ C S kno\\.lcdgc rcpresclltatio~iof protlact and nsscnl1)ly pro- cess, ant1 fa111troot calrsc itlcntificatio~ia~itllocalizatio~i npproaclies. Kno\vlcdgc Rcprcscnlnlion of Protluct nrld Proccss Tlie coniplesit!* of t l ~ ea i ~ ~ o ~ i ~ obody t i v cprcvcnts ~licasl~rc~nent data fro11111ei11gs~~fficient to localize root canses of dinici~sionalvariatioii. It is ncccssarp to inclr~tlekno\vlctlge a b o ~ the ~ t product a ~ ~t lt~l easscii~l)ly proccss 1,cfore the root causes can be detcrn~incd.Our kiio\vlcdgc rcprcsc~~tation is I~ased011t l ~ efinictional cliaracteristics of tlic protluct, tooling, process, and nlcastlrellicllts, \vliicli are ill turn tlefined as collections grotips [Id-151. The I~ierarcl~ical groups of I~ierarcl~icol s l ~ o ~tlie v esplicit rclationsliip bet\vee~~ the niost iniportant featu~.csof tlic antoniotivc body assen~bly,such as: (1) product relations (part-to-part and part-tosubassembly; locators positions of ~)arts/sul)nssc~~il)lies), (2) proccss relations (process layout; parts/s~~basse~nblies sequence; \velding spot locations n ~ scqucncc), ~ d and (3) nieasure~iientrelations (location of i~ispectctl poir~tsand type of mcnsnretl feature; allocation of encl~ mcasuretl point to ~ , a r t / s ~ ~ l ) a s s e ~ ~Figure ~ l ~ l y5 ) . sl~o\vs Aperture LAYER 1 LAYER 2 LAYER 3 I The hierarchical nrcuos of rneasurernenn The h i e m h t a l groups of lccatrxs CEGLAREK AND SHl . I D I M E N ~ I O N AVLA R I A T I O NR E D U C T I O N n11 csn~nplcof tlic I~icmrcl~icnl groups that i ~ i c l ~ ~ d c k~~o\vlc~Igc nbout product/proccss, Iiicnsurc~llcllts,nntl locators. \i 111ajo1.ndva~itngcof this k~io~vlctlgc rclmsc~i~atiori is t1111t a grcnt tlc;ll of k~io~vIc(Igc call IIC rcprcs c ~ i ~ c\vi~l~ili tl n 11nifictlfriinicn-orL, \vllicl~~ i r n ~ ~ l i f i c s ncccss to n11t1spcccls tllc cliagl~osticr c n s o ~ ~to i ~f~i ~gl t l root cnrlscs of f n ~ ~ l tTliis s . rcprcsc~itatio~~ is rlsctl t l ~ ~ r i ~ i g fnr~ltlocalizntio~~. Fi';lult Iclc~itificc~tio~~ As sllo\v~iill 1:igrlrc 6, fnrllt itlc~itificntio~i is cclrlivnlclit to tllc itlc~~tilicatio~~ of sorlrccs of tli~~iollsiol~al v:lriatio~~ ill tllc I~otly-ill-\vl~itc by gror~l)iligIiicasrlrc111c11ts11si1igvnrintio~rIcvcl n11tlcorrclntiol~cocllicic~it. Tlic procctlr~rcof liit~ltidc~ltilicatiolifor sr~stniiicddi~iic~isioiial vnrintio~~ sclccts alitl clnssifics tllc i~iforllinti011p c r t n i ~ i i ~to~ ~iicnstrrc~llclits g capttrrcd dr~rilign givc~ipcriod ol production ti111cand arc bnscd or1 two 1. Far~ltscvcrity cl.itcrio11-sclcctio~i ol iiicasurc111c1itpoilits wit11 vnriatio~icscccdilig vnriatioll t11rcslloltl T,. (Fig. 3). This critcrio~ic~isr~rcs that t l ~ cfault itlc~~tificntiol~ proccss will sclcct tlie Iiicnsr~rc~ncllt p o i ~ ~ttls~ n tlcscril)~ t tlic 111ostsevcrc di~ i i c ~ ~ s i ofarllts. ~ i n l \'arintiol~ tl~rcsl~old T,. was sct so t1111t 30% of all 1iic:lsllrcIilcllts \vcrc iriclrrtlctl for furtllcr n~inlysis. Fn~lltroot cnrlsc isolntioii critcrio~i-sclcctiol~ of Illcnsllrclllclit points \\-it11corrclntio~icscccdil~g corrclatio~ltl~rcsl~old T,. Tliis criterion sclccts ~llcirsrlrclllclit~ ) o i ~tlcscril)i~~g ~ts far~ltswit11 a sillglc root CIIIISC. It is I)il~cd011 t l ~ c;issr~~i~ptio~l tllnt tllc ~licasllrc~ilc~its wit11 1:lrgc vnrintio~~ will LC strollgly corrclntctl if n~itlo111yif tlicir vnriatio~is arc carrscd by tlic snmc root carlsc. For csn~iiplc, if lilcnsrlrclilc~itdntn include vnrintiol~cnr~scdby listrlrc locator fail~rrc,tlic pattcr~idcscribcd by t l ~ cdata will follo\v t l ~ cprc-dctcr~nincdpnttcr~lof tllc fnl~ltctllocator. Tl~crcforc,1)asctl OII t l ~ ccharacteristics of botl!. structure n~itlt o o l i ~ ~locators, g S O I ~ I C oilit its will c L ~ i ~ togctlicr ~ v ~ ' l during tllc nsscmbly proccss. 111n statistical scnsc, "~no\.ing" togctlicr call bc i~itclprctctlby corrclntio~ibct\\*cc~i tllc ~ n c n s u r c ~ ~poil~ts. ~ c n t Enc11 group ol fig. 6. Tlrejloei~ clircrl of tliri~rloslic rcnsorrir~g:Jirrrll tlelrcliorr. Selection of 111en~easrcren~en~s wi111 large wnblion forficrther anabsis; Settircg priorities ill diagtrosis: I Group all MLPs :vih large variation + Classification of /lie c ~ ~ e a s c t r e n ~with en~s large varialiorc illto groccps lvirh high correlalio~r: Enclt M P L group co~lslitulesone case; Ass~rrnptiorr: Single root calcse of the far111ntani/ets i~self~l~rncgl~ higlcl~comelaledda/a; 1 Group 1 Iftlre measc1re171ents are ~e~~correlaled, tlcey Ime DIFFERENTroot calcses. ... . . . . .. M A N U F A C T U R I N G R E V I E W * VOL. 8. NO. 2 J U N E 1995 corrclntcd lllcnsllrclilcllt poilits cnllctl cn~itlitl:itc Iiicnsllrclilciit poilits (CAILP) rcprcsciits n cnsc stutly. fcrciit types of root cnllscs. Follo~vil~g tllnt scctio~~, tllc nlinlpsis of all cnsc stlltlics is prcsc~itcd. Esa1111)lcof Cnsc Stucly: Co\\.lsiclc Rcililbrccli~c~it Cyclic \'nrintion This cnsc stlldy ill~~strntcs n di~ric~isio~inl fnillt pcrtni~liligto toolilig il~stnllntio~l, \vliicli wns dctcctcd Lp nu OCJIAl ~i~cnsurcmclit gngc ill t l ~ cfrnliiillg lillc. Tlic fn~llt occ~lrrctltll~rilig1:i1111clip11:isc. l'l~isf:iliIt 11:is direct illf111ciicc011 tllc tluiility of tlic fc~itlcrnlig~l~iicrlt. Corrcctilig this fn~lltinvolvctl rcprogm~ii~iii~ig tlic origi~inl progrn~ilof n \vcldilig robot. Fault 1,ocnlizntioli Tllc fnlllt locnlizntio~iprocctllrre locnlizcs tllc failthat ing part, as ~vcllas tltc nsscriibly stntioli cnl~si~ig fnil~lrcbnsctl 011 tllc sclcctctl CAILI's (Figllrc 7) [13-141. Tlic fnlllt is locnlizctl by idclitif!.i~ig tlic positi011of ChILPs OII tllc c o ~ i ~ l ) o ~sllbassc~iiblics ~c~lt of tlic Lod!.-ill-\vllitc usi~lgIiicmrcl~icolgrol~lxoofproduct/ process. 'l'llc c o ~ i i l ~ o ~ wit11 i c ~ i ttlic I~iggcst~lulllLcrof CA1LPs rcl)rcsc~~ts tlic lilost likcly failed co~iipo~ic~lt nud is cnllctl t l ~ ccnlitlidatc collipollellt. Additiolinllp, tlic Iiicrnrcliiciil g r o ~ ~ p s otllc f procl~~ct/proccss silo\\. tllc nssc~ilblpstntio~i\vlicrc tlic cnlididntc c o ~ i i ~ ) o ~is i c:is~it sc~liLlcd.'Tliis stntioli liiost likcly cnilscs tlic fnillt nlitl is cnllcd tlic colldidntc station. I;IEI,D STUDY: SPORT UTILITY \'EIIICI,E To tlctcr~lli~ic tllc root cnllscs of di~nc~isioiiol fnillts n ~ i dto clctcr~iii~ic Iio~vtllcsc faults \vollltl nffcct tllc di~iic~isio~inl clunlity, 1111 nnnlysis of solvctl cosc studics is contluctctl. Tlic objcctivc of tliis nlinlysis is to s111111ilnrizc tllc k~io\vlctlgcnlitl cspcric~iccsgainctl froin \.ariati011rctluctioi~ncti\,itics. It focllscs prillinrily 011 drn\vi~ig n guiclclinc for fr~trlrci~ii~)lc~iic~itntio~i tlurilig tlcsigl~as ~vcllas dllrilig ln~~iicli of n ~ie\vprotluct. Tlic foIlo\viilg scctioll prcscllts nn csnliiplc of olic cnsc stutlp wit11 tlif- I 1;:iuIt Iclc~~tification r\lcos~u.c~iic~it poil~ts3L and 5 L 011 tllc Lody-in\vliitc sl~o\vcdInrgc 6-sig~iio\,nriotioli in tllc l'axis (I~il)onrtl/Outl)onrtl)cqunl to 3.91 a11t1 3.16 111111, rcspcctivclp. Ilotli Iiicnslwcliicnt poilits slio\vctl strong COTrclntioli (0.94). Figilrc 8 slio\vs tlic locntiol~of tlic co\vlsitlc rci~iforcc~iic~it 1)nlicl alitl sclisors 2L nlid 5 L i l l tlic botly-in-~vliitc.Tlic otlicr liicnsilrcllicllt poiilts tlid not slio~vstrolig corrclotio~iwit11 poilits 2L illit1 5L. Falrlt 1,ocnlizntion Dccnllsc I)o~lisclcctcd Iiicns~~rc~lic~lt points 2L and panel, 5 L wcrc locntctl 011tlic co\vlsitlc rci~lforcc~licnt tllc gcoliictricnl stntioli nsscliibliiig tliis pnlicl ~vossllggcs~ctlas n cn1itlitl:itc stntioll. 'l'lic root cnusc of tllc 1. Fault identification: measurement points (4.6. 17) CEGLAREK AND S H l D I M E N S I O N A LV A R I A T I O NR E D U C T I O N I arrd rrrensrrrerrrerrl Cowlside reinforcement serrsors 2L IC. 51,. 2L 6-sigma = 2.9 1 m m Sample Size - 100 f n ~ ~was l t itlc~~tificd I)y 11singa n S-l)ar clinrt ant1 a tlctnilctl study of tllc nsscml~lyprocess in t l ~ ccniitlitlntc stntion. Tllc S-l)nr clinrts of points 2 L n~itl5 L ill Ynsis slio~vrcgulnr spikes o c c ~ l r r i ~c v~ cg ~ y2 6 1llcnsrlrciiiclits (Figure 9). Dctnilctl nlinlysis rcvcnlctl tlint ill tlic frniiii ~ i gli~ic,oiic ~vcltli~ig robot lnissctl two ~vcldingspots locntctl 011 t l ~ cco~vlsitlcpanel (Figure 10) cvcry 2 6 pailCIS nftcr n rcgulnr tip drcssi~igo p c m t i o ~ ~ . ' 6-sigma = 3.16 mm Sample Size - 100 so on; and (4) t l ~ cinfor~iiations o ~ ~ r c ~c s~ s to c ddctcct tl~cscprol)lcms. Root Cnrrscs of (Ire D i r ~ i c ~ r s i o r \~' n lr i n l i o ~ ~ Fifty_t~vo cnscs ~ v c r citlc~itifictlant1 solvctl nt tlic nssclnl~lyfacility to rcncli t l ~ cIlcst-iri-class Icvcl of tli~ I I C I I S ~ O I Ivariation, ~I ~ v l ~ i isc lcqual ~ to 2 111111 (6-sig111n stn~ltlirdtlcvintion). I ~ I I I O-I I~IICSC ~ C ~ S Cs t ~ ~ d i c118 s , root cnuscs 11-crc fo1111d.All f01111tlroot cnllscs call LC divitlctl Corrective Action T l ~ cproblcni was c a ~ ~ s cby t l tlic progrn~nmi~lg of illto f o ~ i~~ r~ n j cntcgorics: or product n ~ i dproccss design, thc rol~otsccluclicc, nliicl~wns corrcctcd. toolil~gi~~stnllntioii, tooling i i ~ n i ~ ~ t c i ~ nrind ~ i csc~, ~ p p l i c d ~ i ~ n t c r ivnrintio~~. nl Figurc 11 slio~vsthe d i s t r i b ~ ~ t iof o~i Evnluntior~ root cnllscs alllong tlicsc four cntcgorics. 11ccordi1igto 11ftcr corrcctivc nctioii was co~~ll)lctcd, vnriatio~~ this classificntio~~, t l ~ c o ~ ~ t r i l , u t i oto~ ~t11c s vnrintio~i was rctlucctl fro1112.91 to 1.87 111111n~itlfro1113.16 Ivcrc: ~ ~ i n i ~ ~ t c ~procctlurcs i n ~ i c c (37%), product n~itlproto 2.01 1~1111ill t l ~ cY asis for sclisors 2 L n~itl5L, ccss tlcsig11 tliscrcpa~icics(27%), supplictl ~ ) n ~ ii11ncc11cl rcspcctivcly. rncics (23%),nlitl tooli~igi~istnllntio~i r~iistakcs(13%). This n~inlysisil~tlicntcstlint 111ost di~iic~isioiinl variatio~i Cnsc Strrdics Classification Bnscd or1 Proclucliorl problc~rlsof tlic botly-ill-~vliitcarc cnuscd by iilcorrcct Pllnscs i111d Di~licrlsio~lnl \'ilrintiorl niniiitc~in~icc procctl~~rcs n11tl dcsigli discrcl)a~~cics T l ~ IcX C S C I I ~ C ~n~ialysiscsnl~lilies5 2 case stl~tlics rntl~crtlia11s ~ ~ l q ~ lpn~icl i c d i ~ r n c c ~ ~ ~ xnlid c i c tooli~ig s illnccortli~~g to tllc iintllrc of tlic p r o b l c ~ i ~and s clnssifics stnllntio~it~iscrcpniicics.I lo~vcvcr,this clnssificntio~i tl1c11111si11gfour diffcrc~itcriteria: (1) t l ~ croot cnllsc of tlocs not n~lnlyzct l ~ cscvcrity of tlic occl~rrctlprol)lcnis. tllc ~ ~ r o l ) l ci.c., ~ ~ idcsigl~, , i~lstnllntio~i, ~ ~ i n i ~ ~ t c ~or~ n ~ i cnnsctl c , 011 cspcric~icc,tlic lilost scvcrc nntl ti~licco1is11111~lintcrinlvnrintio~~; (2) tlic p r o t l ~ ~ c t iplinsc o ~ i in ~vl~icli ilig prol)lc~nsarc citlicr si~pplicr-rcln~ctl o r tlcsig~i-rctlic root cnllsc occ~~rrctl, i.c., prc-lnui~cl~, Ini~nch,o r full lntctl prol)lciiis. 111t l ~ cnssc~~il)ly pln~it,tlic nvcrngc volu~iicp r o t l ~ ~ c t i o(3) i ~ ;t l ~ careas o r stngcs in 1)otly nsrcnctio~ltiulc for ~ i i n i ~ i t c ~ ~rclntctl n ~ i c c prol)lc~iiswns n s c ~ ~ i b ~lroccsscs, ly i.c., sol)nsscn~bly,I)otl!- framing, n11tl f c ~ vIio~irs( < 8 I i o ~ ~ r sIlo~vcvcr, ). corrccti~igtlcsig~i M A N U F A C T U R I N G REVIEW VOL. 8. NO. 2 J U N E 1995 tlg. % .T-bar chart i~i-oble~l~s or panels discrepancies took on avcragc ol lrorn 1 to 4 months. Prod~~c:tion Phases during which the Koot Cause of the Dimensional Variation Occurred Tn siinilar ways, tllc dirncr~siorralfa11lr.swere classified as a fuilction ol lirrie. This c:lassiIication provicles vduablc insigh1 illlo the strengths and weakuesses of the procl~c:~. and assembly process vers~isthe time of thcir dclcctior~. Figure 1 shows identifiecl fault root causcs ill cab11 w \ ~)rocl~~ction phase. The vertical axis indicates the 6-sigma variation of thc aulomolive bocly, 1)ased on the weekly average of daily balches of 100 hoclies with 100 measnremelirs takctl lor each automotive body. The d ~ t werc a captiucd during a 60-week periocl. The dimensional problerrls for each production phase are disc ~ ~ s s cwith d particular emphasis on those that allcci vuialion level. '1. Pre-productioll [~llasc. 'l'hc 6-signa variation of the 11ody-in-whi~cwas reducecl from 0.5 to 3.5 mm clurillg this phase. Tooling design and tooling iilstallalion as well as variat:ion of si~ppliecl Welding spots ......... Missed welding spots a L k ~ g related n problems lnstallatbn related problems jTm Maintenance related problems Cowlside Reinforcement panel a C E G L A R E K A N D SH1 Supplier related problems DIMENSIONAL VARIATION REDUCTION s t n ~ l l p i ~parts ~ g ~vcrct l ~ c . ~ ~ ~root n j o rcntlscs of varintio~~. Fig. 4 sl~o\vsthat 42% of t l ~ cfi1111tsat this stngc wcrc tooIi~!gdcsigu-rclntctl, 24% of t l ~ c f o ~ ~ lCOIIIC t s fr0111tooli~~g ilistnllntio~l,nntl 34% of root cntlscs C ~ I I I Cfro111s t n ~ ~ ~parts. p i ~ ~Arol111t1 g 40% of all f n ~ ~ lwcrc t s rclntctl to tllc fnilurc of nssc~~ihly fist~lrcs.Duc to tile lo\\- volu~ucof proc l ~ ~ c t itooliug o ~ ~ , ~ ~ i n i ~ ~ t c\vns ~ l nnot ~ ~nc111njor c issue c111ri1igprc-l)rod~~ctio~l. 2. I'roiluctio~~ la111ic11.111 t l ~ cln1111c1iplinsc, \.nrintio~i \vns rccluccd fro1113.5 to 2.5 I I I ~ S. i ~ ~ ~ itol ot lr~ c first ~)llnsc,dcsign o ~ l dtooli~igil~stnlli~tio~~ rclatcd faults \vcrc still t l ~ c~ ~ ~ nroot j o rcnuscs of v;lriatio~~, co~~tributi~lg nl)ol~t50% of t l ~ ctotal I I I I I I I ~ C I -of root causes. Root ciu~scsd r ~ cto t l ~ cst~pplicd s t n ~ ~ l p ipn~lcls l ~ g clccrcnsctl to 29%. I-lo\vcvcr, \,nrintio~lcauscd by tooli~~g ~linilltcrla;~cc-rclatcd fnlllts 1)cca111ci l ~ i p o r t n as ~ ~ttl ~ cproductiol~volu111ci~icrcascd.111gcncml, tllc f n ~ ~ lrclntcd ts to nssc111l)lyfisturcs mpidly i~~crcnsctl to 70% of tlic tocnl nun~bcrof faults. Figurcs 12 nntl 13 S I I I I I I I I ~ ~ ~t Zl ~C cclctcctio~~ of t l ~ c toolillg dcsigll tliscrc~)a~lcics, i~~stilIIntio~i ~l~istnkcs, as \vcll ns supplicrl pa~~cls' i~laccl~rncics tluri~lgt l ~ cI;roductlucctl fro1112.5 to 2.2 111111. .Alni~ltc~ln~lcc rcIncccI ti0111Il1nscs.T o o l i ~ ~dcsign g nntl il~stnllntio~~ tliscrc~)n~~problc~lis\vcrc the lnnjor root cnllscs in this cics bccml~cOIIC of tllc 111njorroot co~lscsof tllc pl~nsc.Rcfcrrillg to Fig. 4, it call bc seen tllnt olprol)lcnis oftcr tllc vnrintion \\.as rcduccd to 2.2 111111. ~liostall root cnuscs \vcrc cluc to ~ ~ ~ n i l l t c ~is~ a n c c This nllo\vs 11s to cstill~n~c tllot t l ~ c2.2-111111Icvcl \\.as sucs. Ib s ~ ~ ~ l ~ ~ toolilig ~ ~ n r itlcsigll z c , nl~cl tllc toolirlg tlcsign cnpnbility of t l ~ cproccss. Fr~rtl~cr rci~lstnllntio~~ tliscrcpn~icicsllntl b c c ~itlc~~tifictl ~ nntl cluctio~~ of variation to rcncl~t l ~ 2-nail c I c ~ c rcquircd l solvcd t l u r i ~ ~tllc g prc-protluctior~n~icl1n111lcl1 so111ctoolillg tlcsig~l~~lodificotio~~. pl~nscsnntl, nftcr illcrcnsi~~g t l ~ cprod~lctio~l volulrlc, tooli~~g I I I ~ ~ I I ~ C I I ~ proccdurcs IICC bccn~~lc lllorc clc~~ln~ldi~lg o~itlcritical for d i ~ ~ l c ~ ~ s i o ~ ~ o l vnri;~~io~~. 3. First sllift I1111 l)rooductio~l.Duriug t11c first full ~jroductio~i shift, the vnrintion \vns furtl~crrc- , t11ot t l ~ c11u111bcr of tlcsig~~, i~~stilllatio~~, and s111)plicd pa~lclsrclntctl prol)lc~nsi~~crcnsctl ngni11. This call bc cs~)lnincdI)!. rcalizi~~g tllnt I V ~ I C I It l ~ c rnriatiol~rcochcs its inl~crcntIcvcl, furtllcr vnrintion rcductioll rcquircs cl~n~lgcs in t l ~ cproduct n~idproccsscs. ' 4. Scco~~cl shift lnllllcl~al~tlI1111I)iocl~~ctio~~. Tl~c \ * n r i ; ~ ~Icvcl i o ~ sligl~tly ~ illcrcnsctl to 2.5 111111nt t l ~ c bcgil~uillgof tllc pcriod tl~rcto 1:1111lcl~il1g of tllc s c c o ~ ~shift. t l llftcr solvi~lgscvcml cast studics, t l ~ c vnriotio~ifi~inllyncllicvcd 2-111111IcvcI ill t l ~ cfifth I ~ ~ O I of I ~f111l ~ I ~)rod~~c 011 t itwo o ~ ~shifts. D11ri11g tl~is~~criotl, ~~~ni~itc~~n~~cc-rclntctl problcn~s\vcrc a 111njormot cnusc, nltl1011g11 tl~cy\vcrc rclnti\.cly less ( I O I I ~ ~ I I t11a11 ~ I I ~ c111ri11gt l ~ cprcviolls ~)l~nsc. Tllc ;1ssc1111)lyf i s t ~ ~ r fcns ~ ~ lc:~usctl ts 70% of t l ~ c total I I I I I I I of ~ ~ fnil~~rcs. ~ It is i~~tcl'csti~lg to 11otc M A N U F A C T U R I N G REVIEW VOL. 8, NO. 2 J U N E 1995 Locnlization of Dimensional Variation withi11 Assembly Process 'Thc. root cause based analysis preserlreil in the . prcvious sec~ion~illowsus to estirrlate the cnusrs of clinlerisiorlal variation. In addition to that, one of the most in~portantiss~lesis the loc:ilixatio~~ c~fthe variation within the assembly process: body-in-white suhasscmhlies (nnderhocly mid jide apertures), framing opcrations, ancl silppliecl panels. This classificalioll is sunlmarizecl in Figures 14 ancl 15. Based on Figs. 1 4 and 15, it coudtl be cor~c:h.idc:tl that thc variation rcdirction in the s1d)assernhlvlines has a major iirlpact or1 the variarion of ihe hocly-inwhile. 111general, prol)leirls in s~ihasse~nbly are twice as fre.q~ierltas those in the framing station. This inclicatcs thal for this pnrticl~lardesign of the hoclp-in-white, ~ h c s~lhnsseml~ly is much more seilsitive to variation tllc~n the final framing of subassemblies. 111 additioll, the coopcration and support fro111 stamping operations playcd n major role in r.ecluc:ing dilne~lsionalvariation. Ahout one-fourth of the root ca uses wrcrccai~scdby the part suppliers. Information Sources for Fauli Delection T l ~ eallalysis of the inforlnntion sources is focused on the rneasi.lrr.ment and verificarion. of gages such as: (1)Optical Coordinate Measuring Ylachule (OCMM), (2) C~oordinateYleasuriilg )lachine (CYIY), (3) visi~al obse-l~ation,and (4) Theodolite gagc. Figwe 1 6 shows the clistribi~tionol the infonr~atioil sources usccl lor dclcclion and ar~alvsisof the sol\:cd casc s~uclics.'l'hc rlleasurelileIlt gag: llsecl rnost oltcll LO idcn~ifyvariatioi~prollelns was the 0C:;LIhI (56%). C:Y14ls artcl Theoclolires wcrc cac11 usccl in 19% of all mse stuciies. mostly d.tlrillg analpis allcl vrrific.aL ~ ~ofI faults. I Sumnlarizi~lp~11ccxpcricl~ccgained while solving tlicsc casc stildics. \\:c C O I I C ~ ~ I Cthat ~ ~ Cinformation ~ sourccs arc i~scd~r~i-lirlly for t h e e purposes: 1. Dctcc~iont u ~ diclcn tificatior~of thr clirr~erlsionnl variation-OCM-\I. visud ol)servnrinn, mlrl Feedback horn Pirlal Assernhly. 2. .4nal!.sis of already icle11tifiet:l fa1~lts-OC\li'~,l mid C3iNl. 3. Qiian~i~ativc verification of the root sn~:~se of diuiellsio~lalvariatioi~-(:bINI and theodolite. Thc, classificatiorl basecl or1 irlfonrl~~tion source has been corrclatccl largely with proc11lc:tion phases. Figire 17 allocates tlifferc:nt information solirces to different proctuctioi~pl~ases.This classification call he surrllr~arizetl as ~ollons: 1. In-line ineasurement cloinina~csas all informatioil sotlrce during detection and ~~nalysis of h e problems causing din~cnsionalvariation. 2. Mosl lrcy ucnt applicatior~of (:MI gage (35%) was during pre-productior~phase, *.hen the voltune of produced cars was small (small salllple size). 3. Theodolite was used during lannch phasc (34%) for velific,ation of toolirig installatior1 problcrils.. and second shifr phase (35%)for vcrificatiori of tooling clesigii modifications ncccssaly ~oreacl~ 2-rrm~variation lcvcl ancl not preclicted earlier hy designers. CONC1,USIONS AhD SUMMARY El Panel dated problem a Subassembly related problems Framing related problems Find Asembly rclated pmblems There is a move ro blend control of dirliansional variario~~ into the antomotivc ioclus~ry.\Vii:hir~the ut~rolnotive industl?, it is viewed as assets able to c:reotr high qualit! product. This papcr srnrirrl:~rizc\s:I devcloped ~nethoclologyfor rcdt~ci~lg cli~nensionalvariation :is well as n study or] ~ h root c causes of variation. This s~lrily.cottductecl in oue of t11t facilities of the leaclil~p ;it110 I I I I I I I I I ~ I L C\\-;IS ~ L It'o(:~~s(!d ~ I W . 011 oljtainirlp tllc 1jcs1ill-class cli~r~c:~~sio~lcll i~~uro~~lorivc bocly bldlcl basccl on clti~~liiy O I : I I C ~ I I I I ;~I ~s~ ~~ i ~ l ~asl i2s IlI iI I~~( ~6 -~S~~ ~ I IAI ~ I ) . CEGLAREK AND SH1 . D I M E N S I O N A L VARIATION REDUCTION F(q. 15. Ilnnt Pre-Production Launch One shift Two shifts 100 cc~rrsrr / i s s ~ f i r f ~ / i n n Lnsed on 111e 90 loculizcr/inn of /he 80 clinrc~r~.siorrcr,,/r~~~I~s 70 u>i&lrin lfrc -'2 ns.nerr~b(~.prucrss Q) 60 a n d ocrilrlmc.ta Q) 2 so $1 during procirrc/tr,rr 40 phar~r. 30 20 10 0 -- 3 $E 2'z E a M ;5 .z5 q rt: 2 vl 0, m data-tirive~i,casc-based approach was used in solving s. dimensional l a ~ ~ l tAn of one case st~lclyis presented in die paper. All of the case st~idiesIleccssary to achieve hesT-in-class climensionnl ~ariaTi011were analyzed and classified according to the iintnre of the fa~ilts using four. different critexia: (1) root causes of h e laults, such as tooling design, u~stallation.mainterrance, or ir~corningmaterial v;uiation; (2) root causes in the c:liffercni productioii ph:ises, i.e., prc-lauu~ch,launch: u ~ full d vol~uneproduction; (3) areas ol body assenihly process, LC.,subassembly, body fia~l~ing, clc.; and (4) the infor~rle~tion sources used to detect din~cl~sio~lal faults. Wlost of the \:arialion problems di~rir~g the firsi 1.5 A 5 a L E M .5 B ?&i 3m E 2 3 M .E E E P z ?i '2" -2 I0 ~iionthsol the conclucted test werc cat~scclby prohleins in ~ ) r ~ d uand c l process (tooling) design, tooling installntion. tooling nlaintenallcc, and inc:ornirig lrlaterial variation. Their relativc contributions :ire classificcl accorrling I O the nt~rlherof problem> ill each calcgoy: Tooling ~naintcnanccrelatecl prohlerns: 37%. Product and process clcsign related prohlnris: 27%. Sta111pingpalel related prohlerns: 23%. Tooling instnllation relatcd problems: 13%. For a11 ol the cases icleiitified, spumdic problelns or special cause prohlenls (such as spikes) arc as frecyuent a s chro~licp~'ohlcmsor common cause problc~ns (35% of total case s ~ ~ r d i c sIIowever: ). the imderlying root causes for the ~rlajorityof the sporadic: ~)roblerrls arc chronic iii nature. 'l'hcsc problems are indllc:etl by in~crferellce,either pi-~rr/parlinterference or tooling1 part i11~crferenc:e. As rar as the i.oot causes or variation during rlilfereiit Ij~~oclac-tiori p1ias1:s. finclings related to llie loiincl~ stage arc as I'ollows: Prol)lrrns during p1.c-proclnctio~~ ancl p r o c l ~ ~ r ~ I t i ~ ~ ~:lr(' i ~~ ~~ l il i ~ irclatecl ~ l l ! to cle.sigrl arlcl i~ls~allalion. :i11(1are 1)nnc-1-rclniecl. MANUFACTURING REVlEW VOL. 8, NO. 2 J U N E 1995 fig. 17. IZoof Pre-Production Launch One shift 100 Two shifts cnrrsc clns$i/;cntiorr 1 Lnsecl orr tlre irforrrrntion sorrrce rritcrior~nrrrl orcrrrmrcc clrrrirrg pmdrrrtiorr plrases. D ~ ~ r i n gIICW vclliclc program, prc-l~rotluction nritl protll~ctln~llicllarc tllc i~uportnntplinscs in \\,llicll to corrcct all di~ncnsiolinlp r o l ~ l c ~nlid ~is ncllicvc tllc proccss cnpnl~ility. lnost significantly to tllc varintio~~ of tllc body-ill-\vllitc. In tllc co~iductcdstudy, 56% of tllc problclrls arc subnsscmblY-riclntcd. 111c\-nluntilig tllc iliiportnncc of ~iicasurc~ncrit systclils ili. fault idcl~tilicntiolio l d localizntion, tlic li~idiligs nrc: :. Findings rclntcd to tllc f11Il prodr~ctio~i stngc nrc ns follows: To sllcccssf~~lly i(lc11tifytli~iic~isio~inl problems, sllfficicl~tulcnsllrcllicrlt is rlcccssn1-y. For fast tlctcctioli of tllc t l i ~ ~ ~ c ~ ~prol)lclrls, s i o ~ ~ n11sc l of in-linc lricnsllrclrlcllts is cssclltinl. For fault clingr~osis, flcsiblc ~ ~ l c a s u r c ~ ~ systcriis l c ~ i t nrc i~liportarlt. Each vcliiclc Ibody Ilns, as n rcslllt of product nlid ~ ~ O C C Stlcsig~i, S an irilicrcrit Ic\.cl of tlinlc~lsionnl \ynrintio~~. Oricc tllc proccss rcncllcs its irlllcrcllt Icvcl of vnrintio~l,hlrtllcr rcductioll of tllc tlilncllsionnl vnriatio~iOII tllc botly call LC ncllicvctl. I.Io\vcvcr, tllis cn11 LC dollc olily tllr011gl1costly nlltl tiliic co~lsr~~iii~ig product n ~ i dproccss ~iiodificntioll. I11 tllc cnsc of tlic nsscliibly plnlit \vllcrc study \vns co~ldllctcd,tllc tli~i~c~isio~inl cnpnlility is nt 2.2 To sl~cccssfullyvcrify tliiilc~isio~lnl problc~nsill tooling, tlicodolitcs, or portnblc CABls arc ncccssnry. 111111. h1:11iy pr0bIc111sill ~)rotI11ctioll st:~gcarc ~linilltclinlicc rclntctl. For csnlllplc, drlrirlg O I ~ Csllift productio~l,nll of tllc cnsc stutlics itlclitifictl wcrc ~~ini~~tc~~n~icc-rclntctl. During two sllift procluctio~l, 52% of tllc cosc st~ltlics\vcrc ~~~ni~itc~la~lcc-rcl:~tctl. for root causes of \.nrintioll ill tliffcrcllt arcits of tllc nsscllil)ly proccss, s~~l)nssc~i~l)ly wrintiol~co11tri1)lltcs [\s F ~ l s i aof l ~ nsscllll)ly process kllo\vlcdgc wit11 t1uti1tlrivcll nllnlysis is tlic kcy to frlst itlc~~tificotio~~ of tllc root cnllscs. CEGLAREK AND S H 1 DIMENSIONAL VARIATION R E D U C T I O N REFERENCES 1. Hopp, W.J., Spearman, M.L., Woodruff, D.L., "Practical Strategies for Lead Time Reduction," Manufacturing Review, 3(2): 78-84, 1990. 2. Ayres, R.U., "Complexity, Reliability, and Design: Manufacturing Implications," Manufacturing Review, 1(1): 26-35, 1988 3. Parkinson, A., Sorensen, C., Pourhassan, N., "A General Approach for Robust Optimal Design," Trans. of ASME, Journal of Mechanical Design, 115(1): 74-80, 1993 4. Gupta, Y.P., Kumar, S., "Controlling the Production Process Through Statistical Process Control," Manufacturing Review, 4(1):18-32, 1991 5. Barkan, P., Hinckley, C.M., "The Benefits and Limitations of Structured Design Methodologies," Manufacturing Review, 6(3): 211-220, 1993 6. Brannan, B., "Six Sigma Quality and DFA-DFMA Case Study/Motorola Inc., Boothroyd & Dewhurst DFM Insight, 2:1-3, 1991. 7. Plonka, F.E., "A Methodology for Tolerancing, Process Evaluation and Control of Automobile Body Subassembly Designs," Ph.D. Dissertation, Univ. of Michigan, 1974 8. Baron, J., "Dimensional Analysis and Process Control of Body-In-White Processes," Ph. D. Dissertation, University of Michigan, Ann Arbor, 1992 9. Roan, C., Hu, S.J., Wu, S.M., "Computer Aided Identification of Root Causes of Variation in Automotive Body Assembly," ASME Winter Annual Meeting, 391-411, 1993 10. Hu, S.J., "Impact of 100% Measurement Data on Statistical Process Control (SPC) in Automobile Body Assembly," Ph. D. Dissertation, University of Michigan, Ann Arbor, 1990 11. Hu, S., Wu, S.M., "Identifying Root Causes of Variation in Automobile Body Assembly Using Principal Component Analysis," Trans. of NAMRI, XX: 311 -316, 1992 12. Roan, C., "Identification, Monitoring, and Diagnosis for Dimensional Control of Automobile Body Assembly," Ph. D. Dissertation, University of Michigan, Ann Arbor, 1993 13. Ceglarek, D., "Knowledge-Based Diagnosis for Automotive Body Assembly: Methodology and Implementation," Ph.D. Dissertation, University of Michigan, Ann Arbor, 1994 14. Ceglarek, D., Shi, J., Wu, S.M., "A Knowledge-based Diagnosis Approach for the Launch of the Auto-body Assembly Process," Trans. of ASME, Journal of Engineering for Industry, 116(4):491-499, 1994. 15. Shi, J., Hu, S.J., Ceglarek, D., "Process Navigator for the Automobile Body Assembly Process," Proceedings of the First S.M. Wu Symposium on Manufacturing Science, Evanston Il: 325-332, 1994. 16. Ceglarek, D., Shi, J., "Fixture Failure Diagnosis for Autobody Assembly Using Pattern Recognition," accepted for publication in the Trans. of ASME, Journal of Engineering for Industry simultaneously in ASME Winter Annual Meeting, PED- 68: 263-275, Chicago, Ill., November 6-11, 1994. 17 Takezawa, N., "An Improved Method for Establishing the Process-Wise Quality Standard" Rep. Stat. Appl. Res., JUSE, 27(3):63 -75, 1980. 18. Menassa, R.J., DeVries, W.R., "Locating Point Synthesis in Fixture Design," Annals of CIRP, 38(1): 165 -169, 1989. MANUFACTURING REVIEW VOL. 8. NO. 2 J U N E 1995 19. Rearick, M.R., Hu, S.J., Wu, S.M., "Optimal Fixture Design for Deformable Sheet Metal Workpieces," Trans. of NAMRI, XXI:407 -412, 1993. 20. Datamyte Handbook, "A Practical Guide to Computerized Data Collection for Statistical Process Control," AllenBradley Co. Inc, 1993 21. Perceptron 1000, ,“System Manual,” Perceptron Inc, 1991. 22. Greer, D., "On-line Machine Vision Sensor Measurements in a Coordinate System," SME Paper #IQ88-289, 1988. 23. Sekine, Y., Koyama, S., Imazu, H., "Nissan's New Production System: Intelligent Body Assembly System," SAE Technical Paper Series, No. 910816, pp. 1-12, 1991. Dariusz Ceglarek is a research investigator at Department of Mechanical Engineering and Applied Mechanics at the University of Michigan. His research interests include quality control and assurance methodology with implementation; intelligent manufacturing, knowledge-based diagnosis using online fault detection and isolation with emphasis on automobile industry, application of AI in manufacturing. His current research is being sponsored by General Motors Corp., Chrysler Corp., National Institute Standard and Technology - Advanced Technology Program, and the National Science Foundation. He is a associate member of Society of Manufacturing Engineers, and American Society of Mechanical Engineers. Jianjun Shi is the associate director of the S. M. Wu Manufacturing Research Center and a faculty member at Department of Mechanical Engineering and Applied Mechanics at the University of Michigan. His teaching and research interests include quality control and assurance methodology, systems, and its implementation; dynamic system modeling and control; intelligent manufacturing using on-line fault detection, isolation, and diagnosis with emphasis on automobile industry. His current research is being sponsored by General Motors Corp., Chrysler Corp., Auto Body Consortium, National Institute Standard and Technology - Advanced Technology Program, and the National Science Foundation. He is a member of Society of Manufacturing Engineers, and American Society of Mechanical Engineers.