基于系统动力学的供应链预警决策技术研究与实现

上海交通大学硕士学位论文基于系统动力学的供应链预警决策技术研究与实现姓名：费若愚申请学位级别：硕士专业：计算机软件与理论指导教师：王东20090101IIIIIIResearchandImplementationofEarly-warningandDecision-supportofSupplyChainbyUsingSystemDynamicsABSTRACTWiththerapiddevelopmentofinformationtechnologyandthewidespreaduseofInternettechnology,supplychainmanagementsolutionshaveincreasinglybecomeapopulartopicinbothindustrialandacademicfield.Effectivemanagementofasupplychainrequiresahighdegreeofsupplychainvisualization.Bymonitoringthesupplychainactivitiesinreal-time,boththeunexpectedvariationswhichhavebeenexposedandthosethathavenotbeenrevealedbecauseofthedelayphenomenonamongsupplychainnodesarerequiredtobedetectedandearly-warned.Theexistingresearchesmainlyfocusinthevariationsexposedatoperationallevelandtakeinterceptivedecisionstoremedytheexceptions.However,theexistingtechnologyisrelativelysimpleindetectingtheunexpectedvariationsatanearlystageandtakingshort-termtacticalpreventivedecisions.Thispaperfocusesontheresearchanddevelopmentofsupplychainearly-warninganddecision-supporttechnologyattacticallevel,detectingtheunwantedsituationsandpotentialrisksinadvance,whichbringsthepossibilityoftakingpreventivedecisionstomitigatethevariations.Accordingtothecurrentstatusinsupplychainmanagement,thispaperproposesamethodologyusedforsupplychainearly-warninganddecision-supportattacticallevelbycombingsystemdynamicsandneuralnetworks.Bymonitoringthedynamictrendsinsteadofstaticvaluesofsupplychainperformanceindicators,early-warningsforpotentialrisksIVcanbereported.Byfine-tuningthesupplychaintacticalstrategyinshort-term,theunwantedimpactcausedbyuncertaintycanbemitigated.Thispaperpresentsthesupplychainmonitoringmodelandadoptssystemdynamicssimulationmodelforsupplychainmodeling.Thesupplychainearly-warningtechnologypresentedbythispaperreportsearly-warningsbasedonthepredictionofthedynamictrendsofmonitoredindicators.Initially,asupplychainmodelisbuiltusingsystemdynamics.Theninordertomeetthereal-timemonitoringrequirementandintegrationneedofenterpriseapplications,aneuralnetworkthatisequivalenttothesystemdynamicsmodelisbuiltandactsasthekernelofmonitoringandearly-warningmodule.Thispaperalsodesignsthedecisionsupporttechnologybasedonthelearningabilityofneuralnetworks.Throughmappingtheweightadjustmentofneuralnetworksintodecision-makingofsupplychain,theshort-termtacticaldecisiontechnologycanbeputinpracticetomitigatetheunwantedsituationsreportedbyearly-warnings.Thispaperfirstlyanalyzestheexistingsupplychainearly-warninganddecision-supporttechnologyandpointsoutthatnotonlytheexposedexceptionsshouldbepaidattentionto,butalsothedynamictrendsofthemonitoredindicatorsisworthytobehighlighted.Then,awiderangeofsupplychainmodelingmethodsisanalyzed.Inordertomeettherequirementoftacticalmodeling,systemdynamicsisusedtomodelthesupplychain.Fortherequirementofreal-timemonitoring,systemdynamicsmodelisthentransformedintoequivalentneuralnetworksandtheinputsequenceisdesignedtoadapttothenecessaryoftime-varyinginput.Thearchitectureofmonitoringandearly-warningmoduleisproposed.Onthebasisofneuralnetworksusedbymonitoringandearly-warningmodule,thelearningandself-adjustabilityofneuralnetworksisusedtoimplementthetacticalshort-termdecision-supporttechnology.Finally,acasestudyofthemanufacturingVindustryanddetailsofsystemimplementationarepresentedtoillustratethemethodologyandarchitecture.KEYWORDSSupplyChain,Monitoring,Early-warning,Decision-making,SystemDynamics,NeuralNetworks20092220092220092211.1RightProductRightTimeRightQuantityRightQualityRightStatusRightPriceRightPlaceERPWMSCRP--21.2SAPI2CommerceOneAribaSCMKPISAPSAPEventManagement1.33(1)(2)(3)1.4452.12.1.1SCEM[1]SAPSCEMSAPEventManagement[2]web[2]KPI(KeyPerformanceIndicators)KPI[2]14KPIKPI2.1.26[3]SAPEventManagementSAPEventManagement[4][5][6][6]72.22.2.1(1)(2)[3][4]2.2.2[3]82.32.3.12-1ERPCRMWMS9MonitorEngineEarly-WarningEngineERPCRMWMSTMSOMS...DataIntegrationLayerPredictedBehaviorofMonitoredIndicatorsDesiredEarlywarningsDecisionSupportEngineNoYesSupplyChainModelNoActionTakenDecisionSettingSCMSolutions2-1Figure2-1ThearchitectureofSupplychainEarly-warningandDecision-supportplatform2.3.2102.3.3123112.42.4.1[7]12312:()[8]AminAitiokSIMAN[9](Multi-Agent)(MASMulti-Agentsystem)Multi-Agent[10]2.4.21Forrester1961[11,12]2070(1)13(2)(causalloopdiagram)(3)(stocksandflowsdiagram)21958ForresterBernard[13][14](1)AlnoosFrasier[15](JIT)BarlasAksogan[16](2)CakraeastiaDiawati[17]14(3)Hefeez[18]Hefeez2.4.3(neuralnetworks)1943McCullochPitts[19]Perceptron()[20]Grossberg[21,22]Kohonen[23]80Hopfield[24]BP[25]1989Hecht-NielsonHornikFunahashiBP12-2(a)TRpppp],...,,[21=],...,,[21Rwwww=R15b∑=+Riiibpw1))((bpwffoiiactout+=[26]2-2(b)2-3(a)1pRpip1wiwRw1pRpip1wiwRw(a)(b)2-2(a)(b)Figure2-2(a)neuronmodel(b)neuronmodelwithsite216Jordan[27]Elman[28]2-3(b)(c)JordanElmanJordanElman(a)(b)(c)2-3(a)(b)Jordan(c)ElmanFigure2-3(a)Feedforwardnetwork(b)Jordannetwork(c)Elmannetwork317Hebb(backprpogaatinoBP)1974Werbos1986RumelhartBPBP2.5183.13.1.1(1)3-112123-1Figure3-1Amanufacturingsupplychainmodel(2)19(3)(1)(2)203.1.2[29][30]3.1.3(1)(2)21(3)(4)(1)(2)(1)(2)