Académique Documents
Professionnel Documents
Culture Documents
STATISTICALTOOLSINSIXSIGMADMAICPROCESSWITH
MINITABAPPLICATIONS
Chapter 6
Chapter6:ProcessCapabilityAnalysisforSixSigma
CHAPTEROUTLINE
ProcessCapability
ProcessCapabilityAnalysis
DeterminingProcessCapability
ImportantTermsandTheirDefinitions
ShorttermandLongtermVariations
ProcessCapabilityUsingHistograms
ProcessCapabilityUsingProbabilityPlot
EstimatingPercentageNonconformingforNonnormalData:Example1
EstimatingNonconformanceRateforNonnormalData:Example2
CapabilityIndexesforNormallyDistributedProcessData
DeterminingProcessCapabilityUsingNormalDistribution
FormulasfortheProcessCapabilityUsingNormalDistribution
RelationshipbetweenCpandCpk
ThePercentoftheSpecificationBandusedbytheProcess
OverallProcessCapabilityIndexes(orPerformanceIndexes)
Case1:ProcessCapabilityAnalysis(UsingNormalDistribution)
Case2:ProcessCapabilityofPipeDiameter(ProductionRun2)
Case3:ProcessCapabilityofPipeDiameter(ProductionRun3)
Case4:ProcessCapabilityAnalysisofPizzaDelivery
Case5:ProcessCapabilityAnalysis:DatainOneColumn(Subgroupsize=1)
(a) DataGeneratedinaSequence,(b)DataGeneratedRandomly
Case6:PerformingProcessCapabilityAnalysis:WhentheProcess
MeasurementsdonotfollowaNormalDistribution
ProcessCapabilityusingBoxCoxTransformation
ProcessCapabilityofNonnormalDataUsingBoxCoxTransformation
ProcessCapabilityofNonnormalDataUsingJohnsonsTransformation
ProcessCapabilityUsingDistributionFit
ProcessCapabilityUsingControlCharts
ProcessCapabilityUsingxbarandRChart
ProcessCapabilitySixPack
ProcessCapabilityAnalysisofMultipleVariablesUsingNormalDistribution
ProcessCapabilityAnalysisUsingAttributeCharts
ProcessCapabilityUsingapChart
ProcessCapabilityUsingauChart
NotesonImplementation
HandsonExercises
Chapter6:ProcessCapabilityAnalysisforSixSigma
Thisdocumentcontainsexplanationandexamplesonprocesscapabilityanalysisfrom
Chapter6ofourSixSigmaVolume1.Thebookcontainsnumerouscases,examplesand
stepwisecomputerinstructionwithdatafiles.
PROCESSCAPABILITY
Process Capability is the ability of the process to meet specifications. The
capability analysis determines how the product specifications compare with the
inherentvariabilityinaprocess.Theinherentvariabilityoftheprocessisthepartof
processvariationduetocommoncauses.Theothertypeofprocessvariabilityisdue
tothespecialcausesofvariation.
It is a common practice to take the sixsigma spread of a processs inherent
variation as a measure of process capability when the process is stable. Thus, the
processspreadistheprocesscapability,whichisequaltosixsigma.
PROCESSCAPABILITYANALYSIS:ANIMPORTANTPARTOFANOVERALL
QUALITYIMPROVEMENTPROGRAM
Thepurposeoftheprocesscapabilityanalysisinvolvesassessingandquantifying
variability before and after the product is released for production, analyzing the
variability relative to product specifications, and improving the product design and
manufacturing process to reduce the variability. Variation reduction is the key to
productimprovementandproductconsistency.
Theprocesscapabilityanalysisisusefulindetermininghowwelltheprocesswill
holdthetolerances(thedifferencebetweenspecifications).Theanalysiscanalsobe
usefulinselectingormodifyingtheprocessduringproductdesignanddevelopment,
selecting the process requirements for machines and equipment, and above all,
reducingthevariabilityinproductionprocesses.
DETERMININGPROCESSCAPABILITY
Thefollowingpointsshouldbenotedbeforeconductingaprocesscapability
analysis.
Chapter6:ProcessCapabilityAnalysisforSixSigma4
Process capability should be assessed once the process has attained statistical
control.Thismeansthatthespecialcausesofvariationhavebeenidentifiedand
eliminated.Oncetheprocessisstable,..
In calculating process capability, the specification limits are required in most
cases,Unrealisticorinaccuratespecificationlimitsmaynotprovidecorrect
processcapability.
Process capabilityanalysis using ahistogram or acontrolchart is based on the
assumption that the process characteristics follow a normal distribution. While
theassumptionofnormalityholdsinmanysituations,therearecaseswherethe
processesdonotfollowanormaldistribution.Extremecareshouldbeexercised
where normality does not hold. In cases where the data are not normal, it is
important to determine the appropriate distribution to perform process
capabilityanalysis.Incaseofnonnormaldata,appropriatedatatransformation
techniquesshouldbeusedtobringthedatatonormality.
:
SHORTTERMANDLONGTERMVARIATION
Thestandarddeviationthatdescribestheprocessvariationisanintegralpartof
processcapabilityanalysis.Ingeneral,thestandarddeviationisnotknownandmust
be estimated from the process data. There are differences of opinion on how to
estimate the standard deviation in different situations. The estimated standard
deviation used in process capability calculations may address "shortterm" or "long
term" variability. The variability due to common causes is described as "shortterm"
variability, while the variability due to special causes is considered "longterm"
variability.
Inprocesscapabilityanalysis,both"shortterm"and"longterm"indexes
arecalculatedandarenotconsideredseparatelyinassessingprocesscapability.The
indexes Cp and Cpk are "shortterm" capability indexes and are calculated using
"shortterm"standarddeviationwhereas,PpandPpkare"longterm"capabilityand
Chapter6:ProcessCapabilityAnalysisforSixSigma5
arecalculatedusing"longterm"standarddeviationestimate.Thesearediscussedin
moredetaillater.
DETERMININGPROCESSCAPABILITY
Followingaresomeofthemethodsusedtodeterminetheprocesscapability.
Thefirsttwoareverycommonandaredescribedbelow.
(1)Histogramsandprobabilityplots,
(2)Controlcharts,and
(3)Designofexperiments.
PROCESSCAPABILITYUSINGHISTOGRAMS:SPECIFICATIONLIMITSKNOWN
:
Supposethatthespecificationlimitsonthelengthis6.000.05.Wenowwantto
determine the percentage of the parts outside of the specification limits. Since the
measurements are very close to normal, we can use the normal distribution to
calculatethenonconformingpercentage.Figure6.2showsthehistogramofthelength
data with the target value and specifications limits. To do this plot, follow the
instructionsinTable6.2.
Table6.2
HISTOGRAMWITH
OpentheworksheetPCA1.MTW
SPECIFICATIONLIMITSFromthemainmenu,selectGraph& Histogram
ClickonWithFitthenclickOK
ForGraphvariables,..ClicktheScalethenclickthe
ReferenceLinestab
IntheShowreferencelinesatdatavaluestype5.956.06.05
ClickOKinalldialogboxes.
Histogram of Length
Normal
35
5.95
6.05
Mean
StDev
N
30
5.999
0.01990
150
Frequency
25
20
15
10
5
0
5.96
5.98
6.00
Length
6.02
6.04
Figure6.2:HistogramoftheLengthDatawithSpecificationLimitsandTarget
Chapter6:ProcessCapabilityAnalysisforSixSigma6
Histogram of Length
Normal
5.95
5.999
6.05
Mean
StDev
N
30
5.999
0.01990
150
Frequency
25
20
15
10
5
0
5.96
5.98
6.00
Length
6.02
6.04
Figure6.3:FittedNormalCurvewithReferenceLinefortheLengthData
Table6.4
CumulativeDistributionFunction
Normalwithmean=5.999andstandarddeviation=0.0199
xP(X<=x)
5.950.0069022
6.050.994809
Fromtheabovetable,thepercentconformingcanbecalculatedas0.994809
0.0069022=0.98790or,98.79%.Therefore,thepercentoutsideofthespecification
limitsis10.98790or,0.0121(1.21%).Thisvalueisclosetowhatweobtainedusing
manualcalculations.Thecalculationsusingthecomputeraremoreaccurate.
PROCESSCAPABILITYUSINGPROBABILITYPLOTS
Aprobabilityplotcanbeusedinplaceofahistogramtodeterminetheprocess
capability.Recallthataprobabilityplotcanbeusedtodeterminethedistributionand
shapeofthedata.Iftheprobabilityplotindicatesthatthedistributionisnormal,the
mean and standard deviation can be estimated from the plot. For the length data
discussed above, we know that the distribution is normal. We will construct a
probabilityplot(orperformaNormalitytest).
Table6.5
NORMALITYTESTOpentheworksheetPCA1.MTW
USINGPROBABILITYPLOTFromthemainmenu,selectStat& BasicStatistics
&NormalityTest
ForVariable,SelectC1Length
UnderPercentileLine,..type5084inthebox
UnderTestforNormality,.AndersonDarling
ClickOK
Chapter6:ProcessCapabilityAnalysisforSixSigma7
:
:
Foranormaldistribution,themeanequalsmedian,whichis50thpercentile,and
the standard deviation is the difference between the 84th and 50th percentile. From
Figure 6.4, the estimated mean is 5.9985 or 5.999 and the estimated standard
deviationis
^
84thpercentile50thpercentile=6.01835.9985=0.0198
Note that the estimated standard deviation is very close to what we got from
earlier analysis. The process capability can now be determined as explained in the
previousexample.
Mean
StDev
N
AD
P-Value
99
Percent
95
90
5.999
0.01990
150
0.314
0.543
84
80
70
60
50
40
30
20
50
10
1
0.1
5.950
5.975
6.0183
5.9985
6.000
Length
6.025
6.050
6.075
Figure6.4:ProbabilityPlotfortheLengthData
ESTIMATINGPERCENTAGENONCONFORMINGFORNONNORMALDATA:
EXAMPLE1
Whenthedataarenotnormal,anappropriatedistributionshouldbefittedto
thedatabeforecalculatingthenonconformancerate.
Data file PCA1.MTW shows the life of a certain type of light bulb. The
histogramofthedataisshowninFigure6.5.Toconstructthishistogram,follow
thestepsinTable6.1.
Chapter6:ProcessCapabilityAnalysisforSixSigma8
Histogram of Life in Hrs
Normal
150
40
Mean
StDev
N
494.1
484.9
150
Frequency
30
20
10
-600
600
1200
Life in Hrs
1800
2400
Figure6.5:LifeoftheLightbulb
ForthedatainFigure6.5,itisrequiredtocalculatethecapabilitywithalower
limit because the company making the bulbs wants to know the minimum survival
rate.Theywanttodeterminethepercentageofthebulbssurviving150hoursorless.
TheplotinFigure6.5clearlyindicatesthatthedataarenotnormal.Therefore,if
weusethenormaldistributiontocalculatethenonconformancerate,itwillleadtoa
wrong conclusion. It seems reasonable to assume that the life data might follow an
exponential distribution. We will fit an exponential distribution to the data and
estimatetheparameterofthedistribution.Todothis,followtheinstructionsinTable
6.6.
Table6.6
FITTINGDISTRIBUTION
OpentheworksheetPCA1.MTW
(EXPONENTIAL)Fromthemainmenu,selectGraph& Histogram
ClickonWithFitthenclickOK
:
ClicktheDistributiontab,checktheboxnexttoFitDistribution
ClickthedownwardarrowandselectExponential
ClickOKinalldialogboxes.
The histogram with a fitted exponential curve shown in Figure 6.6 will be
displayed. The exponential distribution seems to provide a good fit to the data. The
parameter of the exponential distribution (mean =494.1 hrs) is also estimated and
shownontheplot.InFigure6.7,twoprobabilityplotsoftheLifedataareshown;one
fittedusingnormaldistributionandtheotherusingexponentialdistribution.
Chapter6:ProcessCapabilityAnalysisforSixSigma9
Histogram of Life in Hrs
Exponential
150
Mean 494.1
N
150
60
50
Frequency
40
30
20
10
0
400
800
1200
1600
Life in Hrs
2000
2400
Figure6.6:ExponentialDistributionFittedtotheLifeData
The plot using the exponential distribution clearly fits the data as most of the
plotted plots are along the straight line. To do the probability plots in Figure 6.7,
followtheinstructionsinTable6.7
Table 6.7
PROBABILITYPLOTOpentheworksheetPCA1.MTW
Fromthemainmenu,selectGraph& ProbabilityPlots
ClickonSinglethenclickOK
ForGraphvariables,typeC2orselectLifeinHrs
ClickonDistributionthenselectNormal.ClickOKinall
dialogboxes.
Normal
99.9
Mean
494.1
StDev
484.9
N
150
AD
7.566
P-Value <0.005
99
95
90
80
70
60
50
40
30
20
Percent
Percent
99.9
99
10
Mean
494.1
N
150
AD
0.201
P-Value 0.954
90
80
70
60
50
40
30
20
10
5
3
2
1
5
1
0.1
0.1
-1000
1000
Life in Hrs
(a)
2000
3000
10
100
Life in Hrs
(b)
Figure6.7:ProbabilityPlotsoftheLifeData
1000
10000
Chapter6:ProcessCapabilityAnalysisforSixSigma10
Table6.9
CumulativeDistributionFunction
Exponentialwithmean=494.1
xP(X<=x)
1500.261831
Thetableshowsthecalculatedprobability,P(x<=150)=0.2618.Thismeans
that26.18%oftheproductswillfailwithin150hoursorless.
CAPABILITYINDEXESFORNORMALLYDISTRIBUTEDPROCESSDATA
Process C apability (C p=1.0)
L SL
USL
L SL
USL
L SL
USL
MINITABprovidesseveraloptionsfordeterminingtheprocesscapability.Theoptions
canbeselectedbyusingthecommandsequenceStat&QualityTools&Capability
Analysis.Thisprovidesseveraloptionsforperformingprocesscapabilityanalysis
includingthefollowing:
Normal
Between/Within
Nonnormal
MultipleVariables(Normal)
MultipleVariables(Nonnormal)
Binomial
Poisson
10
Chapter6:ProcessCapabilityAnalysisforSixSigma11
DETERMININGPROCESSCAPABILITIESUSINGNORMALDISTRIBUTION
Thecapabilityindexesinthiscasearecalculatedbasedontheassumptionthat
theprocessdataarenormallydistributed,andtheprocessisstableandwithincontrol.
Twosetsofcapabilityindexesarecalculated:Potential(within)CapabilityandOverall
Capability.
PotentialCapability
Thepotentialorwithincapabilityindexesare:Cp,Cpl,Cpu,Cpk,andCcpk
variationwithineachsubgroup.Ifthedataareinonecolumnandthesubgroup
size is 1, this standard deviation is calculated based on the moving range (the
adjacentobservationsaretreatedassubgroups).Ifthesubgroupsizeisgreater
than 1, the within standard deviation is calculated using the range or standard
deviationcontrolchart(youcanspecifythemethodyouwant).
According to the MINITAB help screen, the potential capability of the process
tellswhattheprocesswouldbecapableofproducingiftheprocessdidnothave
shiftsanddrifts;or,howtheprocesscouldperformrelativetothespecification
limits(iftheshiftsintheprocessmeancouldbeeliminated).
OverallCapability
Theoverallcapabilityindexesare:Pp,Ppl,Ppu,Ppk,andCpm
Thesecapabilityindexesarecalculatedbasedontheestimateof
^
overall ortheoverallvariation,whichisthevariationoftheentiredata
inthestudy.
Ifthereisasubstantialdifferencebetweenwithinandoverallvariation,itmay
be an indication that the process is out of control, or that the other sources of
variationarenotestimatedbywithincapability[seeMINITABmanual].
11
Chapter6:ProcessCapabilityAnalysisforSixSigma12
Note:Accordingtosomeauthors,CpandCpkassessthepotentialshortterm
capability using a shortterm estimate of standard deviation, while Pp and Ppk
assess overall or longterm capability using the longterm or overall standard
deviation.Table6.17containstheformulasandtheirdescriptions.
FORMULASFORTHEPROCESSCAPABILITYUSINGNORMALDISTRIBUTION
Table6.17showstheformulasfordifferentprocesscapabilityindexes.
Table6.17
CapabilityIndexesforPotential(within)ProcessCapability
USL=upperspecificationlimit
Cp
U SL LSL
6 w ith in
CPL
LSL=lowerspecificationlimit
x LSL
^
3 within
within estimateofwithinsubgroupstandarddeviation
Ratioofthedifferencebetweenprocessmeanandlower
specificationlimittoonesidedprocessspread
x processmean
C PU
USL x
^
3 within
Ratioofthedifferencebetweenupperspecificationlimit
toonesidedprocessspread
C PK Min.C PU , C PL
Takesintoaccounttheshiftintheprocess.Themeasure
ofCPKrelativetoCPisameasureofhowoffcenterthe
processis.If C P C P K theprocessiscentered;if
C P K C P theprocessisoffcenter.
CCPK
USL
^
3 within
:
:
:
^
CCPK
3 within
:
:
12
Chapter6:ProcessCapabilityAnalysisforSixSigma13
Note:Inalltheabovecases,thestandarddeviationistheestimateofwithinsubgroup
standarddeviation.Asnotedabove,theformulasforestimatingstandarddeviation
differsfromcasetocase.Itisveryimportanttocalculatethecorrectstandard
deviation.Thestandarddeviationformulasarediscussedlater.
RelationshipbetweenCpandCpk
TheindexCpdeterminesonlythespreadoftheprocess.Itdoesnottakeinto
accounttheshiftintheprocess.Cpkdeterminesboththespreadandtheshiftinthe
process.
:
andtherelationshipbetweenCpandCpkisgivenby
pk
(1 k ) C
NotethatCpkneverexceedsCp.WhenCpk=Cp,theprocessiscenteredmidway
betweenthespecificationlimits.BoththeseindexesCpandCpktogetherprovide
informationabouthowtheprocessisperformingwithrespecttothespecification
limits.
OVERALLPROCESSCAPABILITYINDEXES(ORPERFORMANCEINDEXES)
MINITABalsocalculatestheoverallprocesscapabilityindexes.Theseindexesare
Pp, PPL, PPU, Ppk, and Cpm. The formulas for calculating these indexes are similar to
thoseofpotentialcapabilitiesexceptthattheestimateofthestandarddeviationisan
overallstandarddeviationandnotwithingroupstandarddeviation.Theformulasfor
theoverallcapabilityindexesrefertoTable6.18.
Table6.18
Capability Indexes for Overall Process Capability
Cp
USL=upperspecificationlimit
LSL=lowerspecificationlimit
USL LSL
6 within
overall estimateofoverallsubgroupstandard
deviation
Thisistheperformanceindexthatdoesnottakeinto
accounttheprocesscentering.Continued
13
Chapter6:ProcessCapabilityAnalysisforSixSigma14
PPL
3 overall
Ratioofthedifferencebetweenprocessmeanandlower
specificationlimittoonesidedprocessspread
x processmean
U SL x
PP U
x LSL
o v e r a ll
PP K M in . PP L , PP U
C pm
C pm
Thisindexistheratioof(USLLSL)tothesquareroot
ofmeansquareddeviationfromthetarget.Thisindex
isnotcalculatedifthetargetvalueisnotspecified.A
highervalueofthisindexisanindicationofabetter
process.Thisindexiscalculatedfortheknownvalues
ofUSL,LSL,andthetarget(T).
USL LSL
n
( tolerance ) *
(x
i 1
UsedwhenUSL,LSL,andtargetvalue,Tare
known
T )2
n 1
T=targetvalue=(USL+LSL)/2=m
tolerance=6(sigmatolerance)
C pm
M in .{ ( T L S L ), (U S L T )}
n
to le r a n c e
*
2
(x
i 1
T)
USL,LSL,andtarget,Tareknownbut
T (USL+LSL)/2
n 1
:
:
(Note:TheaboveformulasareverysimilartowhatMINITABusestocalculatethese
indexes.SeetheMINITABhelpscreenfordetails).
In this section, we present several cases involving process capability analysis
when the underlying process data are normally distributed. The process capability
reportandtheanalysesarepresentedfordifferentcases.
CASE1:PROCESSCAPABILITYANALYSIS(USINGNORMALDISTRIBUTION)
In this case, we want to assess the process capability for a production process
thatproducescertaintypeofpipe.Theinsidediameterofthepipeisofconcern.The
specification limits on the pipes are 7.000 0.025 cm. There has been a consistent
problem with meeting the specification limits, and the process produces a high
14
Chapter6:ProcessCapabilityAnalysisforSixSigma15
percentageofrejects.Thedataonthediameterofthepipeswerecollected.Arandom
sampleof150pipeswasselected.Themeasureddiametersareshowninthedatafile
PCA2.MTW(columnC1).
The process producing the pipes is stable. The histogram and the
probabilityplotofthedatashowthatthemeasurementsfollowanormaldistribution.
The variation from pipetopipe can be estimated using the within group standard
^
deviationor
within
.Sincetheprocessisstableandthemeasurementsarenormally
distributed,thenormaldistributionoptionofprocesscapabilityanalysiscanbeused.
PROCESSCAPABILITYOFPIPEDIAMETER(PRODUCTIONRUN1)
To assess the process capability for the first sample of 150 randomly selected
pipes,followthestepsinTable6.19.Notethatthedataareinonecolumn(columnC1
ofthedatafile)andthesubgroupsizeisone.
Table6.19
PROCESSCAPABILITY OpenthedatafilePCA2.MTW
ANALYSISUsethecommandsequenceStat&QualityTools&Capability
Analysis&Normal
IntheDataareArrangedassection,clickthecirclenexttoSingle
columnandselectortypeC1PipeDia:Run1inthebox
Type1intheSubgroupsizebox
IntheLowerspec.andUpperspec.boxes,type6.975and7.025
respectively
:
:
ClickOK
ClicktheOptionstabontheupperrightcorner
Type7.000intheTarget(addsCpmtotable)box
IntheCalculatestatisticsusingboxa6shouldshowbydefault
UnderPerformAnalysis,WithinsubgroupanalysisandOverall
analysisboxesshouldbechecked(youmayunchecktheanalysis
notdesired)
UnderDisplay,selecttheoptionsyoudesire(somearecheckedby
default)
Typeatitleifyouwantoradefaulttitlewillbeprovided
ClickOKinalldialogboxes.
TheprocesscapabilityreportasshowninFigure6.11willbedisplayed.
15
Chapter6:ProcessCapabilityAnalysisforSixSigma16
Target
USL
Within
Overall
P rocess Data
LSL
6.975
Target
7
U SL
7.025
Sample M ean
7.01038
Sample N
150
StDev (Within)
0.00971178
StDev (O v erall) 0.00946227
6.98 6.99
O bserv ed Performance
P P M < LSL
0.00
P P M > U SL 53333.33
P P M Total
53333.33
0.88
1.25
0.51
0.51
0.59
7.03 7.04
Figure6.11:ProcessCapabilityReportofPipeDiameter:Run1
INTERPRETINGTHERESULTS
1.
The upper left box reports the process data including the lower specification
limit, target, and the upper specification limit. These values were provided by
the program. The calculated values are the process sample mean and the
estimatesofwithinandoverallstandarddeviations.
ProcessData
LSL 6.975
Target 7
USL 7.025
SampleMean7.01038
SampleN
150
StDev(Within)0.00971178
StDev(Overall)0.00946227
2.
ThereportinFigure6.11showsthehistogramofthedataalongwithtwonormal
curvesoverlaidonthehistogram.Onenormalcurve(withasolidline)..
16
Chapter6:ProcessCapabilityAnalysisforSixSigma
3.
Thehistogramandthenormalcurvescanbeusedtocheckvisuallyiftheprocess
data are normally distributed. To interpret the process capability, the normality
assumptionmusthold.InFigure6.11,
4.
Thereisadeviationoftheprocessmean(7.010)fromthetargetvalueof7.000.
Since the process mean is greater than the target value, the pipes produced by this
processexceedtheupperspecificationlimit(USL).Asignificantpercentageofthepipes
areoutsideof
5.
Thepotentialorwithinprocesscapabilityandtheoverallcapabilityofthe
processisreportedontherighthandside.Forourexample,thevaluesare
Potential(Within)
Capability
Cp
0.86
CPL 1.21
CPU 0.50
Cpk 0.50
CCpk 0.86
OverallCapability
Pp
0.88
PPL 1.25
PPU 0.51
Ppk 0.51
Cpm 0.59
6.
ThevalueofCp=0.86indicatesthattheprocessisnotcapable(Cp<1).Also,Cpk=
0.50islessthanCp=0.86.Thismeansthattheprocessisoffcentered.Notethat
whenCpk=Cpthentheprocess
7.
17
Chapter6:ProcessCapabilityAnalysisforSixSigma
8.
Higher value of Cpk indicates that the process is meeting the target with
minimum process variation. If the process is offcentered, Cpk value is smaller
comparedtoCpeven
9.
TheoverallcapabilityindexesortheprocessperformanceindexesPp,PPL,PPU,
Ppk,andCpmarealsocalculatedandreported.Notethattheseindexesarebased
ontheestimateofoverallstandarddeviation.
10. PpandPpkhavesimilarinterpretationasCpandCpk.Forthisexample,notethat
CpandCpkvalues(0.86and0.50respectively)areveryclosetoPpandPpk(0.88
and0.51).WhenCpkequalsPpkthenthewithinsubgroupstandarddeviationis
.
11. The index Cpm is calculated for the specified target value. If no target value is
specified,Cpm..
12. .Forthisprocess,Pp=0.88,Ppk=0.51,andCpm=0.59.Acomparisonof
thesevaluesindicatesthattheprocessisoffcenter.
13.
Thismeansthatthenumberofpipesbelowthelowerspecificationlimit
(LSL)iszero;thatis,..
Theexpected"within"performanceisbasedontheestimateofwithinsubgroup
standarddeviation.Thesearetheaveragenumberofpartsbelowandabovethe
specificationlimitsinpartspermillion.Thevaluesarecalculatedusingthe
followingformulas
( LSL x
P z ^
*106
w ith in
fortheexpectednumberbelowtheLSL
18
Chapter6:ProcessCapabilityAnalysisforSixSigma
(U S L x
P z ^
*106
w ith in
fortheexpectednumberabovetheUSL
notethat
x istheprocessmean.
Forthisprocess,theExpectedWithinPerformancemeasuresare
Exp.WithinPerformance
PPM<LSL
134.53
PPM>USL 66163.57
PPMTotal
66298.10
Theabovevaluesshowthe.
14.
TheExpectedOverallPerformanceiscalculatedusingsimilarformulasas
inwithinperformance,excepttheestimateofstandarddeviationisbased
onoveralldata.
Forthisprocess,theExpectedOverallPerformancemeasuresare
Exp.OverallPerformance
PPM<LSL
92.20
PPM>USL 61212.82
PPMTotal
61305.02
Thesevaluesarebasedon..continued
19
Chapter6:ProcessCapabilityAnalysisforSixSigma
CASE3:PROCESSCAPABILITYOFPIPEDIAMETER(PRODUCTIONRUN3)
Process Capability of PipeDia: Run 3
LSL
Target
USL
Within
Overall
P rocess Data
LSL
6.975
Target
7
USL
7.025
Sample M ean
7.00016
Sample N
150
StDev (Within)
0.0067527
StDev (O v erall) 0.00640138
1.30
1.31
1.29
1.29
1.30
O bserv ed P erformance
P PM < LSL 0.00
P PM > USL 0.00
P PM Total
0.00
CASE4:PROCESSCAPABILITYANALYSISOFPIZZADELIVERY
APizzachainfranchiseadvertisesthatanyorderplacedthroughaphoneorthe
internet will be delivered in 15 minutes or less. If the delivery takes more than 15
minutes, there is no charge and the delivery is free. This offer is available within a
radiusof3milesfromthedeliverylocation.
Inordertomeetthedeliverypromise,thePizzachainhassetatargetof122.5
minutes.
Usingthe100deliverytimes(showninColumn1ofdatafilePAC3.MTW),a
processcapabilityanalysiswasconducted.Toruntheprocesscapability,followthe
instructionsinTable6.20.TheprocesscapabilityreportisshowninFigure6.14.
20
Chapter6:ProcessCapabilityAnalysisforSixSigma
Process Capability of Delivery Time: 1
LSL
Target
U SL
W ith in
O v erall
P rocess D ata
LS L
9.5
T arget
12
USL
14.5
S am ple M ean
12.511
S am ple N
100
S tD ev (Within)
1.07198
S tD ev (O v erall) 0.986517
10
11
12
13
14
0.84
1.02
0.67
0.67
0.75
15
Figure6.14:ProcessCapabilityReportofPizzaDeliveryTime:1:
:
CASE6:PERFORMINGPROCESSCAPABILITYANALYSISWHENTHEPROCESS
MEASUREMENTSDONOTFOLLOWANORMALDISTRIBUTION(NONNORMAL
DATA)
TheprocesscapabilityreportisshowninFigure6.21.
Process Capability of Failure Time
Using Box-Cox Transformation With Lambda = 0
U S L*
transformed data
P rocess D ata
LS L
*
Target
*
USL
260
S ample M ean
107.115
S ample N
100
S tD ev (Within)
66.3463
S tD ev (O v erall) 74.8142
Within
O v erall
P otential (Within) C apability
*
Cp
C PL
*
C P U 0.60
C pk
0.60
C C pk 0.60
A fter Transformation
LS L*
Target*
U S L*
S ample M ean*
S tD ev (Within)*
S tD ev (O v erall)*
O v erall C apability
*
*
5.56068
4.46632
0.611239
0.647763
Pp
PPL
PPU
P pk
C pm
3.2
O bserv ed P erformance
P P M < LS L
*
P P M > U S L 60000.00
P P M Total 60000.00
3.6
4.0
4.4
4.8
5.2
5.6
*
*
0.56
0.56
*
6.0
Figure6.21:ProcessCapabilityReportofFailureTimeDatausingBoxCox
Transformation
21
Chapter6:ProcessCapabilityAnalysisforSixSigma
PROCESSCAPABILITYOFNONNORMALDATAUSINGJOHNSON
TRANSFORMATION
J o hns o n T r a ns f o r m a ti o n f o r F a i l ur e T i m e
99.9
99
90
Percent
S e le ct a T r a n s f o r m a tio n
N
AD
P-V alue
100
4.633
< 0.005
50
10
P r o b a b il it y P l o t f o r O r ig i n a l D a t a
0.77
0.8
0.6
0.4
0.2
R ef P
0.0
0.2
1
0.1
200
400
0.4
0.6
0.8
Z V a lue
1.0
1.2
P r o b a b i li t y P lo t f o r T r a n s f o r m e d D a t a
99.9
N
AD
P- V alue
99
Percent
90
100
0.249
0.743
P -V a lu e fo r B e st F it: 0 . 7 4 2 7 8 7
Z fo r B e st F it: 0 . 7 7
B e st T ra n sfo rm a tio n T y p e : S L
T ra n sfo rm a tio n fu n ctio n e q u a ls
-6 . 6 7 2 4 4 + 1 . 5 2 0 9 3 * Lo g ( X - 3 .5 8 4 4 6 )
50
10
1
0.1
-4
PROCESSCAPABILITYUSINGDISTRIBUTIONFIT
Theotherwayofdeterminingtheprocesscapabilityofnonnormaldataistouseadistribution
fit approach. In cases where data are not normal, fit an appropriate distribution and use that
distributionrather than a normal distributionto determine the process capability. We will
illustratethemethodusinganexample.
E xponential - 95% C I
99.9
99
90
90
50
P er cent
P er cent
N orm al - 95% C I
99.9
50
10
N orm al
A D = 3.359
P -V alue < 0.005
E xponential
A D = 0.303
P -V alue = 0.818
10
1
1
0.1
-2000
0.1
0
2000
4000
L ife o f T V T ube ( Da y s)
10
100
1000
L if e o f T V T ube ( Da y s)
Weibull - 95% C I
50
50
P er cent
P er cent
90
99.9
99
90
10
10
0.1
0.1
0.1
1.0
10.0
100.0
1000.0
10000.0
Weibull
A D = 0.323
P -V alue > 0.250
G am m a
A D = 0.309
P -V alue > 0.250
G am m a - 95% C I
99.9
10000
0.1
L ife o f T V T ube ( Da y s)
1.0
10.0
100.0
1000.0
10000.0
L if e o f T V T ube ( Da y s)
Figure6.26:ProbabilityPlotsforSelectedDistribution
22
Chapter6:ProcessCapabilityAnalysisforSixSigma
PROCESSCAPABILITYSIXPACK
Anotheroptionavailableforprocesscapabilityanalysisisprocesscapabilitysix
pack.Theprocesscapabilityusingthisoptiondisplays
x
Achart(orindividualchartforsubgroupmeans),
AnRchart(Schartforasubgroupsizegreaterthan8),
Arunchartormovingrangechart,
Ahistogram,
Anormalprobabilityplotoftheprocessdatatocheckthenormality,and
Thebetween/withinstatisticsandoverallcapabilityindexes.
Between/Within Capability Sixpack of Shaft Diameter
I ndiv id ua ls C ha r t o f Sub gr o up M e a ns
C a p a b ility H isto g r a m
Individual Value
UCL=75.01672
75.01
_
X =75.00095
75.00
74.99
LCL=74.98518
1
11
13
15
17
19
21
23
25
74.98
M o v ing R a ng e C ha r t o f S ub gr o up M e a ns
Moving Range
0.02
75.02
LCL=0
1
11
13
15
17
19
21
23
25
74.96
UCL=0.04854
0.04
_
R=0.02296
0.02
0.00
LCL=0
1
11
13
15
17
19
75.00
75.04
C a pa bility P lo t
R a nge C ha r t o f A ll Da ta
Sample Range
75.01
__
MR=0.00593
0.00
75.00
No r m a l P r o b P lo t
A D : 0.593, P : 0.119
UCL=0.01938
0.01
74.99
21
23
S tD ev
B etw een 0.0028568
Within
0.00987
B /W
0.0102751
O v erall
0.0105927
B /W
O v erall
S pecs
25
C apa
Cp
C pk
C C pk
Pp
P pk
C pm
S tats
1.62
1.59
1.62
1.57
1.54
1.57
Figure6.34:ProcessCapabilitySixpackofShaftDiameter
INTERPRETINGTHERESULTS
ThereportshowsthecontrolchartsforXbarandR.Thetestsforspecialcausesareconducted
andreportedonthesessionscreen.Nospecialcauseswerefoundindicatingthattheprocessis
stableandincontrol.Thecapabilityhistogramshowsthatthe..
Chapter 6 of Six Sigma Volume 1 contains detailed analysis and
interpretation of process capability analysis with data files and step-wise
computer instructions for both normal and non-normal data.
To buy chapter 6 or Volume I of Six Sigma Quality Book, please click on
our products on the home page.
23