Examination of worker assignment and dispatching rules for managing vital customer priorities

内容需要下载文档才能查看 内容需要下载文档才能查看

Computers&OperationsResearch27(2000)525}537

Examinationofworkerassignmentanddispatchingrulesformanagingvitalcustomerprioritiesindualresourceconstrained

jobshopenvironments

HemantV.Kher*

DepartmentofBusinessAdministration,UniversityofDelaware,Newark,DE19716,USA

Received1March1998;receivedinrevisedform1January1999

Abstract

Thefocusinthisstudyisonevaluatingschedulingpoliciesthatmaybeusedtoo!eranear-perfectdeliveryperformanceforvitalcustomersindualresourceconstrained(DRC)jobshopenvironments.Priorstudiesinthisareahavemodeledshopswhereallordersareconsideredtobeofequalimportance,andshownthatanyreasonablecombinationofworkerassignmentanddispatchingrulesthataresensitivetotheshopcostfactorsmodeledworkswell.However,arecentsurveyindicatesthatshopsdoinfactdistinguishamongordersonthebasisofcustomeridentity,andthatcertaincustomershaveveryrigiddeliveryguidelineswhichmustbeadheredtoinordertomaintaintheirongoingbusiness.Thepresentstudymodelssuchenviron-ments,andshowsthatthechoiceofschedulingpoliciesisnotinconsequential,especiallyiftheshopaimsatprovidingnear-perfectdeliveryperformancetoalimitedsetofcustomers.Insuchinstances,policiesfordeployingworkerstodi!erentdepartments,aswellasdispatchingrulesthatusecustomeridentitybasedinformationoutperformtheircounterpartsthatarenotsensitivetoordersplacedbyvitalcustomers.Scopeandpurpose

Intoday'shighlycompetitivemarketplace,anincreasingnumberof"rmsarebeingrequiredtoprovideahighlevelofdeliveryperformancebytheircustomers.Certaintypesofvitalcustomers,suchasthoseoperatingundertheJITmanufacturingphilosophy,haveveryrigiddeliveryguidelineswhichmustbeadheredtoonaconsistentbasisinordertoretaintheirbusinessinthelongrun.Priorstudiesthathaveconsideredthisproblemfromaschedulingperspectivehavefocusedonlyondispatchingrulesinmachineconstrainedsettings.However,inmostrealshops,dispatchingrulesarenottheonlymeansavailableinmanagingordersfromvitalcustomers.Instead,managersrelyon#exibleworkersthatcanbemovedtodi!erentdepartmentsastheshoploadvaries.Theroleofworkers,anddecisionrulesthatgovernworker*Correspondingauthor.Tel.:#1803-775-6341Ext.3219;fax:#1803-775-2180.

E-mailaddress:hemantk@http://wendang.chazidian.comcsu.sc.edu(H.V.Kher)

0305-0548/00/$-seefrontmatter??2000ElsevierScienceLtd.Allrightsreserved.

PII:S0305-0548(99)00038-6

526H.V.Kher/Computers&OperationsResearch27(2000)525}537

assignmenttodi!erentdepartmentshasnotbeenconsideredinthecontextofmanagingvitalcustomerpriorities.Thepurposeofthisstudyistodeterminetherelativeimportanceofworkerassignmentanddispatchingrulesino!eringanear-perfectdeliveryperformancetovitalcustomersservedbythe"rm.??2000ElsevierScienceLtd.Allrightsreserved.

Keywords:Deliveryperformance;Customerservice;Multi-classjobshopsystems;Schedulingrules

1.Introduction

Inrecentyears,anincreasingnumberof"rmsarebeingrequiredbycustomerstocomplywiththeirstringentdeliveryrequirements.Asanexample,WisnerandSiferd[1]indicatethatover50%oftheUS-basedmake-to-order"rmsthatparticipatedintheirsurveyprocessedasigni"cantportionof`JITajobswithveryspeci"cdeliveryrequirements(i.e.deliveriesmustbereceivedbythecustomerswithintheirspeci"edtimewindow).Similarly,Malhotraetal.[2]alsoprovideexampleswhere"rmsdistinguishamongordersonthebasisofcustomeridentity,andtakethenecessarystepsinordertoensureanear-perfectdeliveryperformanceto`vitalacustomers.Non-compliancewiththedeliveryrequirementsmayhavepenaltiesassociatedincertaininstances[1],andcanadverselya!ectthe"rm'slikelihoodofsecuringfutureorlongtermcontractsfromvitalcustomers[2].Despitethisevidence,veryfewstudieshaveconsideredtheissueofmanagingduedateperfor-manceforvitalcustomerswithinthejobshopsetting.Wheretherehasbeensuchafocus,studieshavelimitedthemselvestomachineconstrainedsettingsonly,andfocusedprimarilyonevaluatingtheperformanceofvariousdispatchingrulesthatcanbeusedinsuchenvironments[2}4].Theconsiderationofworkeravailabilityinschedulinghasbeenignoredinsuchstudies.However,priorliterature[3]andindustrysurveys[1]indicatethatthechoiceofdispatchingrulesisnottheonlyconcernofshopmanagersinmanagingproductionschedules.E!ectiveschedulingalsorequirestheappropriatechoiceofworkerassignmentrulesthatcontrolworkeravailabilitytodi!erentdepartmentsthatmaycontainjobsfromvitalcustomers.

Whileithasbeenacknowledgedthatworkerassignmentrulesusedinactualpracticeoftenincorporatethenotionofcustomerimportanceindeployingworkerstodepartments[1],therehasbeennoformalexaminationoftheimportanceassociatedwithsuchrulesinajobshopsetting.Consequently,thepresentstudywillconductanexaminationofworkerassignmentanddispatch-ingrulesthatcanbeusedtoo!eranear-perfectdeliveryperformanceforvitalcustomersindualresourceconstrained(DRC)jobshopsettings.ADRCenvironmentischoseninordertobeconsistentwiththe"ndingsinWisnerandSiferd[1]whichindicatethatmostrealjobshopsarelaborconstrained.

Section2discussesthe"ndingsfrompriorstudiesthathaveconsideredduedatemanagementwithrespecttovitalcustomersinmachineconstrainedjobshopenvironments.Italsodiscussesthe"ndingsondispatchingrulesandworkerassignmentrulesinDRCjobshopenvironments.Then,decisionrulesthatarelikelytoperformwellinDRCenvironmentsthatcatertovitalcustomersareproposedinSection3.ThisisfollowedbyadescriptionofthesimulatedDRCshop,theexperimentaldesign,andtheperformancemeasuresinSection4.Resultsofthesimulationexperiments,andconclusionsofthestudyarediscussedinSections5and6,respectively.

H.V.Kher/Computers&OperationsResearch27(2000)525}537527

2.Literaturereview

2.1.Incommensurableprioritysystems

Issuesrelatedtovitalcustomersinjobshopsettingshavebeenconsideredinthecontextofincommensurableprioritysystems,wherejobsarecategorizedonthebasisofcustomeridentity

[2].Typically,highpriorityisassignedtoordersreceivedfromvitalcustomers,whereasnormalpriorityisassignedtoordersplacedbyothercustomerscateredbytheshop.StudiesbyCobham

[4]andLawrence[5]haveconsideredincommensurablesystemswithoutaddressingtheissueofduedatemanagement.Thesestudiesrecommendtheuseofseparatequeuesforhighandnormalpriorityorders,withtherequirementthatordersfromvitalcustomersmustbeprocessedaheadofthoseplacedbyothercustomers.Thisprocedure,alsoknownasthetwo-queueruleinMalhotraetal.[2]helpsreducethroughputtimesforhighpriorityorders.Thesestudies[4,5]alsosuggestthatastheproportionofhighpriorityjobsprocessedbytheshopincreases,thethroughputtimesfornormalpriorityordersdeteriorates.Thisresultsuggeststhatwhenpossible,shopsmustjudiciouslylimitthenumberofhighpriorityorderstobeprocessed.

ThefocusinMalhotraetal.[2]wastoo!eranear-perfectdeliveryperformancetovitalcustomersunderdi!erentsettingssuchasduedatetightnesslevelsandtheproportionofhigh-priorityjobscateredbytheshop.Intheirstudy,severaldi!erentdispatchingruleswereproposedandtestedtoenhanceduedateperformanceforvitalcustomers.Resultsindicatedthatthechoiceofaspeci"cdispatchingruleinordertoo!eranear-perfectdeliveryperformancetovitalcustomersiscontingentupontheproportionofhighpriorityjobsintheshopworkload.Speci"-cally,theforcedpaced(FP)rulewasrecommendedforshopworkloadsthatcompriseof30%orfewerhighpriorityjobs.ThisruleprocessesnormalpriorityjobsusingEDD(earliestduedate),whereashighpriorityjobsareprocessedusingoperationduedatesforallbutthe"naloperation.Theduedateforthe"naloperationisidenticaltotheduedateatthepreviousoperationsoasto`forceathehighpriorityjobthroughthesystem.Forshopswhereworkloadscompriseofbetween30and70%ofhighpriorityorders,thestudyrecommendedtheuseofeitherthetwo-queue(2Q)rule,orthepreemptrule.Thepreemptrulesearchestheshopperiodicallyforhighpriorityjobsthathavebecomecritical.Ifacriticalhighpriorityjobislocated,thenitisscheduledforprocessingimmediatelyatthecurrentandsubsequentoperations.Implementingthisrulemayalsoinvolvepreemptingothernon-criticalhighpriorityornormalpriorityjobsthatarebeingprocessed.Thuspriorresearchonincommensurableprioritysystemsclearlysuggeststhattheappropriatechoiceofadispatchingruleisnecessarytoo!eranear-perfectdeliveryperformanceforvitalcustomers.

2.2.WorkerassignmentanddispatchingrulesinDRCjobshops

InaDRCsetting,thenumberofworkersistypicallylessthanthenumberofavailablemachines,andworkersarecross-trainedsothattheycanprocessjobsindi!erentdepartments.Consequently,theschedulingdecisionsinsuchenvironmentsmustconsiderbothmachinecapacityandworkeravailability[3].Thiscanbeaccomplishedbytheshopmanagerthroughtheappropriateselectionoftheaforementionedworkerassignmentrulesinadditiontothedispatchingrules.Workerassignmentrules,alsoknownasthe`whenaand`wherearulesintheexistingDRCliterature,are

528H.V.Kher/Computers&OperationsResearch27(2000)525}537

usedtodictatethefrequencyofworkertransfers,andtoselectthedepartmenttowhichtheworkerwillbetransferred,respectively.

PriorstudiesthathaveinvestigatedworkerassignmentrulesinDRCenvironmentssuggestthattheappropriatechoiceoftheserulesissituational,anddependstoalargeextentuponthetypeofconditionsthataremodeledinthestudy.Theconditionsthatareshowntohaveabearingontheselectionofworkerassignmentrulesincludethetimerequiredtoexecutetransfersbetweenworkcentersordepartment[6}9],andworkere$cienciesinprocessingdi!erenttasksintheshop

[10}15].Speci"cally,studiesthathavemodeledDRCenvironmentswithtransferdelays[6,7,9,15]recommendtheuseofdecentralized`whenarulewhichallowsworkerstotransferonlyafterthequeueattheircurrentdepartmenthasbeenfullyexhausted.Inshopswhereworkersprocessjobswithvaryinge$ciencylevels,studiesrecommendtheuseofcentralized`whenarulewhichconsidersworkertransfersafterthecompletionofeveryjob[10,11,15].Finally,inshopsthatmodelbothtransferdelaysandvaryingworkere$ciencylevels,theuseof`moste$cienta`wherearulehasbeenrecommendedinpriorstudies[15].

Thechoiceofworkerassignmentrulesisshowntobelessimportantwhentransferdelaysarenegligibleandwhereworkerscanprocessjobsintheshopwithcompletee$ciency(i.e.workershavehomogenouse$ciencies).Asanexample,WeeksandFryer[16]showthattheuseofcentralized`whenaruleallowsforamore#exibleuseofworkersincomparisontothedecentralizedrule,however,they"ndthatthereductionsinmeanandvarianceof#owtime`duetothisaddedyexibilityaresmalla(p.1368).Similarly,TrelevenandElvers[17]testelevendi!erenttypesof`wherearulesandconcludethatthechoiceofthisruleisnotimportant,andthat`amanagercanmakethelaborassignmentdecisionbasedonothercriteriasuchaseaseorcostofapplicationoftherulesa(p.51).However,acommonassumptionthroughouttheabovementionedstudiesintheDRCliteraturehasbeenthatallordersintheshopareofequalimportance.Thus,itisnotclearwhethertheabovementioned"ndings(e.g.WeeksandFryer[16]andTrelevenandElvers[17])regardingthechoiceofworkerassignmentrulescanalsobeextendedintoDRCjobshopswithincommensurableprioritysystems.

Withregardstothechoiceofdispatchingrulesseveraldi!erentstudiesareinagreementthatdispatchingruleswhichperformwellinmachinelimitedenvironmentsalsoperformwellinDRCshopenvironments[3].Asanexample,Park[18]showsthatdispatchingrulessuchasmodi"edoperationduedate(MOD)andshortestprocessingtime(SPT)workwellwithrespecttocustomerservicemeasuressuchasmeantardinessinDRCenvironmentsintheabsenceofworker#exibility.However,thisstudyalsoshowedthat`therelativeimportanceofsophisticateddispatchingrules(suchasMOD)decreaseswithcross-trainedworkforcesbecausetheworkermovementcanreducetheoccurrencesoftemporarybottleneckoperationsa(p.298).However,theissueofo!eringanear-perfectdeliveryperformanceforvitalcustomerswasnotaddressedinPark[18].Assuch,itisnotclearwhetherhisassertionabouttheimportanceofdispatchingruleswouldholdtrueinDRCjobshopenvironmentsthatmustdealwithcustomerswhohaveveryrigiddeliveryrequirements.

2.3.Focusofthisstudy

Althoughpriorstudiessuggestthatworkerassignmentrulesmustbeselectedinconjunctionwiththeexistingshopconditions,theyhavenotconsideredsituationswheretheshopmustensureanear-perfectdeliveryperformancewithrespecttovitalcustomers.Thisisdespitetheevidencethat

H.V.Kher/Computers&OperationsResearch27(2000)525}537529

manyactualshopsareservingcustomerswithrigiddeliveryrequirements(e.g.JITvsnon-JITcustomers[1]).Consequently,theimportanceassociatedwiththeappropriatechoiceof`whena,`whereaanddispatchingrulesinsuchenvironmentsisnotclearfromtheexistingresearchinthisarea.Basedontheresultsofpriorstudiesthathaveconsideredincommensurableprioritysystemsinmachine-limitedenvironments,itcanbeconjecturedthatruleswhichfavorordersfromvitalcustomersarelikelytohelpachievethegoalofattaininganear-perfectdeliveryperformancewithrespecttohighpriorityorders.However,whatistheroleplayedby`whenaor`wherearulesinsuchenvironments?Willworkerassignmentrulesthathavebeenproposedinpriorliteratureworkinsuchenvironments,oristhereaneedtousenewerassignmentrulesthatbaseworkertransferdecisionsbyusinginformationabouthighpriorityjobsintheshop?

Thepresentstudyattemptstoaddresstheseissuesbytestingavarietyof`whena,`whereaanddispatchingrulesinDRCjobshopenvironmentswithincommensurableprioritysystems.Thespeci"cdecisionrulesthatwillbetestedinsimulatedDRCsettingswillbedescribednext.

3.Workerassignmentanddispatchingrules

3.1.`Whenarules

Threedi!erent`whenarulesaretestedtodeterminetheire!ectivenessinachievinganear-perfectdeliveryperformanceforvitalcustomersintheDRCshop.Theseincludethedecentralized(DEC)rule,thecentralized(CEN)rule,andthenewlyproposedmodi"edversionofthedecentralized(DECP)rule.TheDECservesasthebasecaserule,andisleast#exibleofthethreerulesinthatitconsidersworkersfortransfertoanotherdepartmentonlywhenthequeueinthecurrentdepartmentisempty.TheCENruleismost#exiblesinceitconsidersaworkerfortransfertoanotherdepartmentaftercompletionofeveryjob.Finally,theDECPruleliesinbetweentheDECandCENintermsof#exibility.TheDECPruleincorporatesinformationabouthighpriorityjobsatthecurrentdepartment,anddoesnotallowaworkertotransferuntilallhighpriorityjobsinthecurrentdepartmenthavebeenprocessed.

3.2.`Wherearule

The`wherearuleswhichselectthelocationtowhichaworkerwillbetransferredincludethelongestqueue(LNQ),longestqueueofhighpriorityjobs(LNQP),andthehighpriorityjobwiththeearliestduedate(EDDP).TheLNQruleassignstheworkertothedepartmentwiththelongestqueue,andisselectedasthebasecaserule.Thisrulehasperformedwellwithrespecttoduedaterelatedperformancemeasureinpriorstudies(e.g.MalhotraandKher[15]).TheLNQPisasimplemodi"cationoftheLNQrule,andattemptstodeployworkerstodepartmentsthatcontainmaximumnumberofhighpriorityjobs.Finally,theEDDPruletransfersworkersbyconsideringtheduedateinformationofhighpriorityjobsintheshop.

Thechoiceof`whenaand`wherearulesthatincorporatecustomeridentity-basedinformationindeployingworkerstodi!erentdepartmentsisbasedontheempiricalevidenceprovidedbyWisnerandSiferd[1].Accordingtotheirsurvey,thethreetoprankedworkerassignmentruleswere`whencurrentjobdone,sendtomostimportantjoba,`whenalljobsdone,sendtomostimportantjoba,