Professional Documents
Culture Documents
org
doi:10.14355/srl.2016.05.001
OnInferencesinAcceptanceSamplingfor
GeneralizedRayleighDistributedLifeTime
MaroofA.Khan1,HasanMateenUlIslam2
DepartmentofBiostatistics,AllIndiaInstituteofMedicalSciences,Delhi,India
DepartmentofStatistics&OperationsResearch,AligarhMuslimUniversity,Aligarh,India
khanmaroofahmad@yahoo.co.in;2hasan_islam@rediffmail.com
Abstract
Forreliabilityinspection,aneffectiveacceptancesamplingplanisalwaysrequiredwhichcansatisfyboththeproducersand
the consumers quality and risk. This paper constructs the effective single sampling plan for reliability inspection when the
distribution of the failure items follows generalized Rayleigh distributed lifetime. For this purpose, we consider the ratio of
indexaveragelifetimeandtestingtimefortwovaluesofaveragelifetimeacceptableandnonacceptableonesandknownshape
parameterofthedistribution.Arelationshipbetweenindexandreliabilityfunctionisalsoderivedanditsuseisillustratedby
anexample.
Keywords
AcceptanceSampling;QualityandRisk;GeneralizedRayleighDistribution;TestingTime;AverageLifetimeandReliability
Introduction
Qualityconceptrevolvesaroundthemeetingorexceedingcustomerexpectationappliestotheproductandservice.
Managingqualityisalwayscrucialandachievinghighqualityisaneverchangingorcontinuousprocess;therefore,
qualitymanagementemphasizestheideasofworkingconstantlytowardsimprovingthequality.Itinvolvesevery
aspects of the company: processes, environment and their people. Quality product always helps to maintain
customersatisfactionandloyaltyandreducestheriskandcostofreplacingfaultygoods.Companiescanbuilda
reputationforqualitybygainingaccreditationwitharecognizedqualitystandard.Thatswhythedevelopingand
competitive world now has given great importance to statistical quality control techniques for ensuring the
trustworthiness of an item with regards to its lifetime by doing inspections. In this situation, it is necessary to
ascertainthevisibleoperatingcharacteristicvaluesoftheproposedplan.
Acceptance sampling plan ensures that the lifetime of the product is according to the specified/desired
level/standard of the consumer. It is often used to determine the disposition of incoming raw material or parts
when100percentinspectionisdestructive,timeconsuming,orexpensive.Aneffectiveacceptancesamplingplan
willallowtosetandmaintainanacceptablequalitystandard.Inacceptancesampling,inspectionisperformedona
sampletakenfromalotofincomingmaterials.Adecisionismadeconcerningthedispositionofthelotbasedon
theinformationobtainedfromthissample.Ingeneral,threeapproachedcouldbeconsideredwhenreceivingalot:
(1) no inspection, (2) 100% inspection, and (3) sampling inspection. If the life test of the items follows that the
averagelifeoftheproductisabovethespecified/desiredlevel/standard,thesubmittedlotisacceptedotherwisethe
sameisrejected.
The item failure is due to natural causes or due to some spurious ones i.e. environment, lack of maintenance
actions, mishandling, intensive operational tasks etc. If the item failure occurs after a certain specified period of
time,thenitwillbesatisfactoryoperations.Butwecannotexplaintheexacttimeofaspecifiedobjectwhenitwill
fail. We are in position to explain it in terms of probabilities and average expected time as its main parameters.
Now,thefailurebehaviourofthatspecificobjectistobemodeledandhencechosenasthemostsuitableclassof
life distributions describing this timetofailure phenomenon. The document MILSTD781 Reliability test:
exponential distribution used the ratio [ E (T )] / T0 , where E(T) is the average lifetime or durability of underlying
objectsand T0 isthetestingtimeintheexponentialcase(Ctuneanu,andMihalache,1989;Voda,andIsaicManiu,
1994).Later,someinferencesforWeibulldistributioncasehavebeendiscussed(IsaicManiuandVod,2009).Here,
www.srljournal.orgStatisticsResearchLetters(SRL),Volume5,2016
weareconsideringthegeneralizedRayleighdistributionwhichnowbecomesanimportantlifetimedistributionin
survivalanalysisandmanydistributionsareitsspecialcase.Thepresentstudydealswithsomenewresultsonthe
indexaveragelifetime/testingtimeintheconstructionofacceptancesamplingplansforreliabilityinspection,when
timetofailuredistributionisfollowinggeneralizedRayleighdistribution.
The Model: Generalized Rayleigh Distribution
TheRayleighdistributionisusefulinlifetestingandreliabilitymodelingwhenagevarieswithtimeanditsfailure
rate is a linear function of time. The Rayleigh distribution was first derived for the problems in the field of
acoustics (Rayleigh, 1880). It was also applied in life testing of electrovacuum devices and in communication
engineering(Polovko,1968).Later,anacceptancesamplingplanassumingforageneralizedRayleighdistribution
has been developed (Dyer and Whisenand, 1973; Tsai and Wu, 2006). Rayleigh distribution has a linearly
increasingfailureratewhichmakesitasuitablemodelforthelifetimeofcomponents.Ageneralizedversionofthe
RayleighdistributioncalledthegeneralizedRayleighdistribution(GRD)isderived(Voda,V.Gh.1976).Thepdfof
thisGRDisgivenby
f (t )
2
2
1
t 2 k 1 e t / , t 0, 0, k 0.
( k 1) k 1
where () isincompletegammafunctionas
( z , a )
1
(a )
a 1e z dz
ThisGRDfamilyalsoincludesseveralimportantprobabilitydistributionsasspecialcase.
Forexample,if k 0 and 2 ,itreducestotheoneparameterRayleighdistributionwithdensityfunction
f ( x; )
t2
2
t exp
2
2
; t 0, 0.
t2
t 2 exp
22
3 (2)1/2
; t 0, 0.
For, if k [(a / 2) 1] and 2 , it becomes chisquare distribution with ' a ' degrees of freedom whose density
function
f ( x; , a)
t2
t a 1 exp
22
2( a /2) 1 a (a / 2)1/2
t 0, 0.
1
StatisticsResearchLetters(SRL),Volume5,2016www.srljournal.org
2
1 2k 1 t 2 /
, t 0, 0, k 0.
t
e
(k 1) k 1
andcumulativedensityfunctionis
F (t ) 1
j 0
2
(t 2 / ) j et /
.
j!
j 0
2
(t 2 / ) j et /
j!
andmeanvalueis E (T ) m 1/ 2 .
where m (k 3 / 2) / (k 1)
Hence,wehave
2
E (T )
andconsequently,weget
m t 2
k m t 2 m t 2 j
R (t )
exp
E (t )
j 0 E (t ) E (t )
j !
Therefore, for generalized Rayleigh distribution, the ratio E (T ) / t depends on its reliability function. If we fix
t T0 ,wehaveeithertoestimate R(T0 ) ortofixloweracceptableboundforit.Now
2
2j
2
k m T m T
m T0
0
0
R ( T0 )
exp
E ( T0 )
j 0 E ( T0 ) E ( T0 )
j !
www.srljournal.orgStatisticsResearchLetters(SRL),Volume5,2016
acceptance number which is to be determinedand T0 is the previouslyfixed testing time. Theprocedurefor the
constructionofasamplingplanhasthefollowingassumptions:
a) TheitemssubjectedtoinspectionarenonreparableanditsfailuretimefollowsGRD;
b) Itssizeistobedeterminedwithonlyonesamplenoreplacement;
c)
Foragivenrisk,thereisfixedanacceptableaveragelifetime [ E (T )] 1 ;
d) Foragivenrisk,thereisfixedannonacceptableaveragelifetime [ E (T )] 2 ;
e) Thetestingtime T0 smallerthantheactualoperatinglifeoftheunderlingitemsisfixed.
Then decision on the lot is taken as follows: submit to the specific reliability/durability test of a sample size n
drawnrandomlyfromalotofsizeN (n N ) ,duringaperiodofunitsof T0 ;recordthenthenumber (d ) offailed
elementsintheinterval [0, T0 ] ;if d c ,thenthelotisacceptedotherwise,if d c ,thenthelotisrejected.The
values of n and c are determined via the OC function (Operative Characteristic) of the plan which has the
function
L( p )
d 0
1
(np)d enp
d!
2
k (t 2 / ) j et /
where d ! 0,1, 2, , c. and p is the defective fraction in the lot given by p 1
, d is the
j!
j 0
numberoffailedelementsduringthetestingperiod T0 (Grant,andLeavenworth,1988).
approach avoids the knowledge of R(T0 ) since the input values are only T0 and [ E (T )] 1, 2 which are fixed
previouslytakingintoaccountoftheprespecificcaseathand.
TABLE1:VALUESOFSINGLESAMPLINGPLAN [( n,
c )| T0 ] FORk=1ANDTHEINPUTRATIOS 100 T0 / [ E (T )] 1, 2
n
Valuesof 100 T0 / [ E (T )] 1 forwhich L( p2 ) 0.10
100
50
25
15
3
(11)
11
(6.3)
40
(2.8)
113
(1.9)
6
(25)
16
(10)
71
(7)
214
(3.8)
8
(33)
27
(15)
101
(7.2)
276
(5.2)
9
(35)
31
(18)
119
(9.1)
343
(6.1)
An Illustrative Example
Assumethatwehaveanacceptabledurability [ E (T )]1 4000 hoursandanonacceptableoneas [ E (T )] 2 800 hours.
Testingtimewasfixedatthevalue T0 200 hours.Theusualconsumerrisk 0.05 andproducersrisk 0.10
Therefore,tofindtheplan,wehave
100 T0
100 T0
100. 200
100. 200
25 and
5
[ E (T )]2
800
[ E (T )]1
4000
StatisticsResearchLetters(SRL),Volume5,2016www.srljournal.org
From the table, the nearest value of 100 T0 / [ E (T )] 1 for 100 T0 / [ E (T )] 2 25 is 7 and hence for the couple 50 (7). We
choose n 71 samplingunitswiththeacceptancenumber c 1 .Thesamplingplanisthen{(71,1)|200}andasa
consequence, we shall test n = 71 items on a period of 200 hours and record d the number of failed items. If
d 0 or1,weshallacceptthelototherwiseweshallrejectit.
Conclusions
Acceptance sampling plans are being widely used in industry to determine whether a specific batch of
manufactured or purchased items satisfy a prespecified quality to its customer. For determining economical
acceptance sampling plan, we have to fix the time and then estimate the reliability or we can fix the lower
acceptable bound. This method will help us in deciding the sample number for inspection and a small sample
numbercanbeselectedfordifferentcombinationoftheparametersfortheplan.
REFERENCES
[1]
Ctuneanu, V., M. and Mihalache, A., N. 1989: Reliability Fundamentals, Elsevier, Amsterdam (Fundamental Studies in
Engineering,No.10TranslatedfromRomanianbyA.N.Mihalache).
[2]
Voda, V., Gh., IsaicManiu, Al. 1994: The power distribution as a timetofailure model, Economic Computation and
EconomicCyberneticsStudiesandResearch,28,(14),4151.
[3]
IsaicManiu,AlVod,V.Gh.2009:Someinferencesonratioaveragelifetime/testingtimeinacceptancesamplingplansfor
reliabilityinspection.Reliability:Theory&Applications,4(3),5157.
[4]
Rayleigh, J. W. S. 1880: On the resultant of a large number of vibrations of the same pitch and of arbitrary phase,
PhilosophicalMagazine,5thseries,10,7378.
[5]
Polovko,A.M.1968:FundamentalsofReliabilityTheory:AcademicPress,NewYork.
[6]
Dyer,D.D.andWhisenand,C.W.1973:BestlinearunbiasedestimatoroftheparameteroftheRayleighdistributionPartI:
Smallsampletheoryforcensoredorderstatistics,IEEETransactiononReliability,22,2734.
[7]
Tsai, TzongRu and Wu, ShuoJye 2006: Acceptance sampling based on truncated life tests for generalized Rayleigh
distribution,JournalofAppliedStatistics,33(6),595600.
[8]
Voda,V.Gh.1976:InferentialproceduresonageneralizedRayleighvariateI,AplikaceMathematiky,21,395412.
[9]
Grant,E.L.andLeavenworth,R.S.1988:StatisticalQualityControl,6thEditionMc.GrawHillBook,Co.,NewYork.
Maroof A. Khan received his MSc and Ph.D. in Statistics from Aligarh Muslim University and an MBA in Operations
Management from IGNOU, India. Presently he is working as an Assistant Professor in the Department of Biostatistics at All
IndiaInstituteofMedicalSciences,NewDelhi,India.HismajorareaofresearchinterestisReliabilityManagement,Statistical
Quality Management and BioStatistics and he has authored and published a number of articles and papers in international
reputejournals.
HasanM.IslamisworkingasaProfessorofStatisticsintheDepartmentofStatistics&OperationsResearchatAligarhMuslim
University,India.HehasvastteachingandresearchexperienceinthefieldofAppliedStatistics,ProbabilityandBiostatistics.