Professional Documents
Culture Documents
6
mandagden24.oktober2016 11.13
Exam Question: Inventory planning over multiple time periods linear-cost case. The single-unit, single-customer
approach: single-location case finding optimal stock levels. Capacity limited systems. (Chapter A6)
InventoryPlanningoverMultipletime.Periods:Linearcostcase.
Preliminaries
Adecisionatonepointintimeaffectsdecisionsinsubsequenttimeperiods.
Forexample,inventoryleftattheendofaperiodcanbeusedinthenextperiod.Thiscanbeproblematic,becauseit
hadsomescrabworth,wasthrougingoutorwasnothingworth.
PreliminariesTerminology:
Inventoryposition=onhandinventory+inventoryonorderbackorderlevel.
Basestock(ororderupto)polity:
Inanygivenperiod,iftheinventorypositionisbelowacertainbasestocklevelSn,thenitistriedtoraisetheinventory
positionuptolevelSn.Ifthecapacityinthisperiodisinsufficienttodoso,thenitisallusedupsothattheinventory
positiongetsasclosetoSnaspossible.
Basestockpolicyisoptimalinanumberofdifferentenvironments.
Onorderinventory(pipelineinventory)=thenumberofunitsthathavebeenorderedbuthavenotbeenreceived.So
weorderedbuttheproductsdidn'tarrive.
Onhandinventory=thenumberofunitsyettobesatisfied.
Inventorylevel=Onhandinventorybackorders.
Orderuptolevel,Snisthemaximuminventorypositionortargetinventorylevelorbasestocklevel.
Assumptions:
Singleproduct.
Randomandpossiblynonstationarydemandindependentfromperiodtoperiod.
Finiteplanninghorizon.
Orderplacementdecisionsaremadeatdiscreteintervalsoftimeatthebeginningoftheperiod.
Allexcessdemandisbackordered.
Leadtimearedeterministic.
Nofixedorderingcosts,nocapacitylimits.
Notations:
SequenceofEvents:
1. Weobservethenetinventory(whichequalstheinventorypositionsincetheleadtimeiszero).
2. Anorderisplacedonanexternalsupplier,whichraisestheinventorypositiontoSn.
3. Theinventoryarrives,whichraisesthenetinventoryleveltoSn.
4. Next,weobservethedemandthatarisesfortheitem.
5. Ifthisdemandislessthanthesupply,aholdingcostofhischargedperunsoldunit;ifdemandexceedssupply,a
backordercostofbdollarsisassessedperunitshort.
Basestockpolicyisoptimalforthissystem.Outobjectiveistoobtaintheoptimalbasestocklevelthatminimizesthe
longtermexpectedaveragecostperperiod.
OverageInventory(inventoryHolding):
Oritcanlooklikethis:
order placement inventory position. Sn
Yn
UnderageInventory(Backlogging):
HereorderuptolevelS.
Eachperiod'sorderquantity=Sinvposition.
Letustakeanexample:
S=4
Inv.position=1,order41=3.
Inv.Position3,order4(3)=7.
Orwecanhave:
D1=10,fort=2.
Inv.Pos:110+3=6
Thus:Order4(6)=10.
ThisisakindofthePullsystem.
FinitePlanningHorizonAnalysis
Soherewewanttominimizetheoverallcosts.
Let assume that inventory remaining on hand at the end of the horizon is
returned to the supplier, and a credit of kr. c per unit is given. Assume also
that if shortages exist at the end of period N, we satisfy the excess demand
through purchases from the supplier at a cost of kr. c per unit. Then gN+1y
cy and the values of sn can be found recursively by applying dynamic
programming algorithm.
Nowwecansettheprogrammingalgorithm.
Wenowdefinethedifferentnotationsandequations:
S=onorderinventory+inventorylevelatthebeginningofperiodbackorderinglevel.
Inventorylevelofendofperiod +1=Sdemandovertherecentperiod +1periods+exp.Backordering.
( =leadtime)
Sonowinthemultipleperiodmodel,wetakeanexample:
S=6>demandattheendofperiod4=6D1D2D3D4.
Keyperformanceindicators:
1. Stockoutprobability=Probability[Dover +1periods>S].
>1Probability[Dover +1periodsS].
2. InStockprobability=1StockoutProbability.
1. >Probability[Dover +1periodsS].
3. Fillrate=1exp.Backorders/Exp.Demandinoneperiod.
Nowwewilltakeanexamplewiththis:
Wenowassumewehave:
Period:1week.
Leadtime :3.
Demand:Normallydistributed,sowehaveaMean:80.6andSt.Dev:58.8.
Sowewilldiscusshowthesingleunit,singlecostumerapproachcanbeextendedtoconsideramoregeneraldemand
modelandstatedescription:
6.1.1.5GeneralizedDemandModel
NowsupposethatdemandsaregovernedbyaMarkovmodulatedprocess.Thatis,demandineachperiodisgoverned
byanexogenous,stationaryergodicMarkovchain,sn.
Thisstateisobservedinthebeginningofeachperiodn.
6.1.1.6capacityLimitations
Itcouldbethatwehaveexternalsupplier,whichcanonlyshipmaximumCunitsineachperiod.
ThismeansthattheoptimizationproblemcanbedecomposedintoCproblems,eachofwhichrepresentsasubsystem
consistingoftheserialsystemwithasingleunitofcapacityateachstage.
Nowwehavefoundtheformofanoptimalpolicywhentherearenofixedorderingcosts,letusnowfindtheoptimal
stocklevels.
6.2Findingoptimalstocklevels
Thisisbasedoncreatingadynamicprogrammingrepresentationofthedecisionproblem.