物流系统中心选址模型简介

Facilitylocationmodelsfordistributionsystemdesign物流系统设计的选址模型•Introduction•Typesofmodels•GeneralmethodsThedesignofthedistributionsystemisastrategicissueforalmosteverycompany.Theproblemoflocatingfacilitiesandallocatingcustomerscoversthecoretopicsofdistributionsystemdesign.IntroductionIndustrialfirmsmustlocatefabrication(制造厂)andassemblyplants(组装厂)aswellaswarehouses(仓库).Storeshavetobelocatedbyretailoutlets(零售网点).Theabilitytomanufactureandmarketitsproductsisdependentinpartonthelocationofthefacilities.Similarly,governmentagencieshavetodecideaboutthelocationofoffices,schools,hospitals,firestations,etc.Ineverycase,thequalityoftheservicesdependsonthelocationofthefacilitiesinrelationtootherfacilities.TypesofmodelsTheproblemoflocatingfacilitiesisnotnewtotheoperationsresearchcommunity(运筹学);thechallengeofwheretobestsitefacilitieshasinspiredarich,colorfulandevergrowingbodyofliterature.Tocopewiththemultitudeofapplications(众多应用)encounteredinthebusinessworldandinthepublicsector,aneverexpandingfamilyofmodelshasemerged.Facilitylocationmodelscanbebroadlyclassifiedasfollows:•Theshapeortopographyofthesetofpotentialplantsyieldsmodelsintheplane,networklocationmodels(网络选址模型),anddiscretelocation(离散选址)ormixed-integerprogrammingmodels(混合正数规划模型),respectively.•Objectives(目标函数)maybeeitheroftheminsumortheminmaxtype.Minsummodelsaredesignedtominimizeaveragedistanceswhileminmaxmodelshavetominimizemaximumdistances.Predominantly(此外),minsummodelsembracelocationproblemsofprivatecompanieswhileminmaxmodelsfocusonlocationproblemsarisinginthepublicsector.•Modelswithoutcapacityconstraintsdonotrestrict(限制)demandallocation.Ifcapacityconstraintsforthepotentialsiteshavetobeobeyeddemandhastobeallocatedcarefully.Inthelattercasewehavetoexaminewhethersingle-sourcing(单来源)ormultiple-sourcing(多来源)isessential.•Single-stagemodels(单阶段模型)focusondistributionsystemscoveringonlyonestageexplicitly.Inmulti-stagemodels(多阶段模型)theflowofgoodscomprisingseveralhierarchical(层次)stageshastobeexamined.•Single-productmodels(单产品模型)arecharacterizedbythefactthatdemand,costandcapacityforseveralproductscanbeaggregatedtoasinglehomogeneousproduct.Ifproductsareinhomogeneoustheireffectonthedesignofthedistributionsystemhastobeanalyzed,viz.multi-productmodels(多产品模型)havetobestudied.•Locationmodelsbaseontheassumptionthatdemandisinelastic(无弹性的),thatis,demandisindependentofspatialdecisions.Ifdemandiselastic(弹性的)therelationshipbetween,e.g.,distanceanddemandhastobetakenintoaccountexplicitly.Inthelattercasecostminimization(成本最小)hastobereplacedthrough,forexample,revenuemaximization(收益最大).•Staticmodels(静态模型)trytooptimizesystemperformance(性能)foronerepresentative(代表)period.Bycontrastdynamicmodels(动态模型)reflectdata(cost,demand,capacities,etc.)varyingovertimewithinagivenplanninghorizon.•Inpracticemodel(实践模型)inputisusuallynotknownwithcertainty.Dataarebasedonforecastsand,hence,arelikelytobeuncertain.Asaconsequence,wehaveeitherdeterministicmodels(确定模型)ifinputis(assumedtobe)knownwithcertaintyorprobabilisticmodels(概率模型)ifinputissubjecttouncertainty.•Inclassicalmodelsthequalityofdemandallocationismeasuredonisolationforeachpairofsupplyanddemandpoints.Unfortunately,ifdemandissatisfiedthroughdeliverytours(运输,投递)then,forinstance,deliverycostcannotbecalculatedforeachpairofsupplyanddemandpointsseparately.Combinedlocation/routingmodels(选址/路线模型)elaborateonthisinterrelationship.Generalmethods•AHP(AnalyticHierarchyProcess)层次分析法•FuzzyClustering模糊聚类法•Cross-medianmethod交叉中值法•gravitymethod重心法•P-medianmethodP-中值法•Systemicarithmetic系统模拟法•Geneticalgorithm(GA)遗传算法•Theshortestpathmethod最短路径法•SimulatedAnnealing(SA)模拟退火算法Thankyou!