基于粒子群的供应链认知图预警方法研究

上海交通大学硕士学位论文基于粒子群的供应链认知图预警方法研究姓名：曹薇申请学位级别：硕士专业：软件工程指导教师：王东200902012009228200922820092285,55ABSTRACT6SupplychainriskanalysiswithCognitiveMapOptimizationAlgorithmbasedonParticalSwarmOptimizationABSTRACTWiththeglobalizationprocess,themarketcompetitionisnotonlythecompetitionbetweenoneenterpriseandanother,butalsothecompetitionbetweensupplychains.Supplychainisacomplexnetwork,includesthecollaborationamongmarket,sales,production,productdesign,purchase,logistics,financeandinformationtechology.Anyprobleminanylinkwillinfluencestheotherlinks,evencausethebreakdownofthewholesupplychain.Supplychainnetworkrisksoriginatefrommanyelements,andtheriskshaveapparentorinapparentrelationshipsbetweenthem.Supplychainisnotasteadynetworkbutischangingallthetime,allthesefactorsmakethestudyofsupplychainrisksabigchallenge.Thispaperdealswithsupplychainrisks,takesadvantagesofcognitivemapandparticleswarmoptimization.Icameupwiththeideathatsimulatethesupplychainnetworkwithcognitivemap,optimizethemapwithparticleswarmoptimizationmethod.Thismethodsimulatedtheruntimesituationofsupplychainnetwork,solvetheproblemofthepreviousmodelslackofself-studyability,andprovideusarealizablewaytostudythesupplychainpre-alert.Firstly,simulatethesupplychainnetworkwithcognitivemap.Cognitivemapcanmodeltheelementsandtherelationshipsbetweenthem,especiallyforthecomplex,uncertainandco-relationelements.HereIabstractedthenetworkintoaseriesofconceptswithcomplexrelationships,andscoredtheconceptsandrelationshipswiththehelpofexperts,definedtheprimarymodelofthesupplychain.Secondly,Itakethemethodofparticalswarmoptimization.Thisisoneoftherandomglocaloptimizationtechology,throughtheinteractionoftheparticalstofindthebestfieldofthespace,itsadvantageiseasytorealizeandhavegoodperformance.AccordingtotheABSTRACT7definitionofcognitivemap,Iprepared5groupstobeinthesteadystatusand5groupstobeinthedivergentstatus,totally10groupstotrainthecognitivemap,theresultmapisthepre-alertmapthatweneed.Atlast,wecanbaseontheoptimizedsupplychainnetwork,inputthesituationoftheconcepts,pre-alertthenetworkswiththecognitivemapmethod.Theresultwilltellusthecurrentsituationwillrunintoasteadystatusoradivergentstatus.Keywords:supplychain,pre-alert,cognitivemap,particalswarmoptimizaiton.101.121//11199319982000121.221[1],,,[2][3][4],[5][6]AHP[7]O.Feyzioglu1G.Buyukozkan1[8]131.31.3.112331.3.11,,:,,,1421.4151.11.42.42.32.22.11.31.23.43.33.23.13.63.54.14.34.25.25.11-1Fig.1-1structurediagramofthepaper162.12.1.1MarkDavid[9]JAYASHANKARM.SWAMINATHAN[10]200311[11]122000[12][13][14][15][16][17][18][19]3[20]17[21][22][16][23]24[24]25[25]26[26]2.1.218;,;,,2-1[27]2-1Fig.2-1classificationdiagramofsupplychainrisks2.1.31[28][29]Logit[30][31]191MDAMultivariateDiscriminantAnalysisMDAMDA2LogitLogitLogit2-1Logistic2-1Xi(Xip)iciiYLogistic220:;:()2-2Fig.2-2structureofneuralnetworks12[10000];[01000];[00100];[00010];[00001]553(22)nl=(n+m)/2+a(2-2)mna110:1()21BP:;:1.00.5O0.51:{}()[-ll]:pi=2*(IiImin)/(ImaxImin)1(2-3)piIiIminImax2,5ANN3SVMVapnik199522VCSVMSVMSVM2.2,,,,,,;,,,,231,,,,2,,,,3,,(),,,,4,,,,,,,,,,,,,,,,,,,,,,,,,,242.32.3.12-3Fig.2-3architecturofsupplychainpre-alertmanagementplatform(2-3)[32]ERPMRPCRMWMSAlertAnalyticsAdviceActionWorkflowWebServiceERPMRPCRMWMSETLREAL-TIMEADAPTOREDIWebServiceReal-timeAdaptorKPIKPI25webservice2.3.2():1CM,2CM3CM,4CMCM26NP-HardNP-Complete206070PSO2.42.4.1(CognitiveMap,CM)[33][34],[35].Tolman1948,[36].1976,AxelordStructureofDecision,CM[33].1986,KoskoAxelord,CM,[37][38].CM,CM.CM(),,.,CM,CM,,.272-4CMFig.2-4threemethodstoexpressCMCM,[33][33][34][36][37],,.2-4(a)(b)(c)CM.,N1,N2,N3,N4,N5,N6,N7,:()();,(+1-10):+1(),-1(),0;(+1-10).(2-4),f(x)xf(x)=1/(1+e)λ,f(x)=tan(x)i(k)i(k-1)ijj(k-1)A=f(A+wA)(2-4)3z28zzCM,CM(),CM.,CM,[48][49][50][51][52][53].,[48],();[49]:510,.,,,,,,;,,:,,().,Schncider,(),[50];Huerga,()[51];PapageorgiouHebbian,Hebbian,[52];Parsopoulos,[53],.:1.,();2..,CM.2.4.2SwarmIntelligenceKennedyEberhartPSO[54]203029(BeliefSpace)23KennedyEberhartReynoldsboids[55]PointsParticlePSO[56][57]PSO()pBestgBest1998ShiEberhartPSO[58]MNiXi=(xi1,xi2xiD)iPi=(pi1,pi2piD)pBestpi(i=1,2N)pggBestiVi=(vi1,vi2viD)vid(t+1)=w*vid(t)+c1*rand()*(pid(t)-xid(t))+c2*rand()*(pgd(t)-xid(t));xid(t+1)=xid(t)+vid(t+1);(1=i=N,1=d=M)(2-5)c1,c2rand()[01]wd(1dM)[-xd,max,xd,max][-vd,max,vd,max],xidvid(2-5)PSO2-5302-5Fig2-5Pseudo-codeofPSOPSO(Global)(Local)pipllBest2.5PgdPldvmaxvmaxvmaxvmaxPSOc1,c2c1=c2=2w1vmaxvmax=xmaxPSO313.1,12343-13-1Fig3-1architectureofSupplychainriskanalysiswithCognitiveMapOptimizationAlgorithmbasedonParticalSwarmOptimizationerp=/323.23.2.13-1,333-11.2.2301-11011-341.2.=/3.=/4.=/5.=/6.=/7.=/8.=/9.=/10.=/11.=/12.=/[90100]1001101/2/3/4/5/6/7/8/9/10/11/351++3.33.3.13-2Fig3-2cognitivemapofmanufacture3-2363-3Fig3-3cognitivemapoftransportation3-3373.3.21.2.3.4.3-43-53-4Fig3-4cognitivemapoftransportationwithrelationships383-5Fig3-5cognitivemapofmanufacturewithrelationshi3.3.33-2(a)bc3-2a.idintegernamevarcharisBestbool393-2b.idintegernamevarcharvaluefloatgraphIdintc.idintegernamevarcharfrominttointgraphId,fromto3-6ConceptGraphRelat