半导体生产线工序参数的逻辑时序微粒群优化策略

第21卷第9期Vol.21No.9　控　制　与　决　策ControlandDecision　2006年9月Sep.2006:2005-07-11;:2005-09-13.:(70531020);(70271035,60104004);CNGI(CNGI-04-15-5A-2);973(2002CB312202).:(1980),,,,;(1947),,,,,.:1001-0920(2006)09-0969-05康　琦,汪　镭,吴启迪(,200092):提出一种具有逻辑时序特征的微粒群优化算法,并将其应用于半导体封装生产线的工序参数优化中.对实际的半导体封装生产线进行建模,并针对生产线中机器对不同产品的加工时间最优分配这一典型的工序参数优化问题,提出了以单位时间利润值和机器均衡度为评价指标的逻辑时序微粒群优化策略.最后进行了计算机仿真和结果分析.:半导体封装生产线;工序参数优化;微粒群优化算法:TP18:ALogicalTime-sequencedParticleSwarmOptimizationStrategyofProcedureParametersinSemiconductorProductLineKANGQi,WANGLei,WUQi-di(CollegeofElectronicsandInformationEngineering,TongjiUniversity,Shanghai200092,China.Correspondent:KANGQi,E-mail:[e-mail])Abstract:Akindofparticleswarmoptimizationmethodwiththecharacteristicoflogicaltime-sequencedisproposedandappliedtoprocedureparametersoptimizationofsemiconductorassemblyproductline.Theactualsemiconductorassemblyproductlineismodeled.Inallusiontosuchakindoftypicalprocedureparametersoptimizationasprocesstimedistributiontodifferentproductsoneachmachine,thelogicaltime-sequencedparticleswarmoptimizationstrategywiththeevaluateindexofaverageprofitinunittimeandequilibriumdegreeofmachineisputforwardandsimulated.Keywords:Semiconductorassemblyproductline;Procedureparametersoptimization;Particleswarmoptimizationmethod1引　　言,,.,,[13].,.Kennedy[4]1995,.,[5],,[68].,,,.2基本微粒群算法.,(),.,:N,i(i=1N)Dxi=(xi1,xi2,,xid,xiD),i,vi=(vi1,vi2,,vid,viD).id(d=1D)vid:vid=vid+c1rand1()(pid-xid)+c2rand2()(pgd-xid),(1)xid=xid+vid.(2):;c1c2;rand1()rand2()[0,1];pid,pgd;,rand1()rand2().,,.pbest,gbestpbest.,,;,pbestgbest;,(1)(2).,,,,.,,,.3逻辑时序微粒群优化描述,,,,.,,..,,,,.,,.,.:Step1:,,;Step2:,,;Step3:,,();Step4:((1)(2)),,,;Step5:.,,;,Step4.,,.,,.4半导体封装生产线建模,[9].:.,5:(Saw)(DA)(WB)(Molding)(Marking).,3,..970　　　控　　制　　与　　决　　策第21卷3A,BC,1.1/(/d)ASDIP5250BTSSOP4825CQFP8017.5,:1)Saw:.Saw-1Saw-2.2)DA:.5DA-1DA-5.3)WB:QC.20,5:WB-1WB-5,.4)Molding:,.3Md-1Md-3.5)Marking:.8,4:Mk-1(1,2),Mk-2(35),Mk-3(6,7)Mk-4(8).,2.,10%,.WBMarking,,,.2/dABCABCDA1303210Saw1607040230321029080603303210Molding16043032102605303210360WB1563Marking130402023102182512352.83253018445.5440503055.53.2:A,BC,.,(1d),(),().5逻辑时序微粒群优化策略与仿真5.1,,.:1),3,;2)3,;3).,.,:3,,,.,,,,.1.,1().Saw,,,,.Saw:Step1:M.,,XVD=,,D=23=6.,,(h),tijij,,971第9期康琦等:半导体生产线工序参数的逻辑时序微粒群优化策略　　　1X=(x1,x2,,x6)(t1A,t1B,t1C,t2A,t2B,t2C),():((t1A,t1B,t1C),(t2A,t2B,t2C)).Step2:,.,,,.SawSaw-2Saw-1,(t2A,t2B,t2C)(t1A,t1B,t1C).():Saw-2,(t2A,t2B,t2C)t2AT2A,t2BT2B,t2CT2C,t2A+t2B+t2C24.(3)T2A,T2BT2C,A,BCSaw-2,3.T2A24,T2A=24;T2BT2C.Saw-23,,24h.(t2A,t2B,t2C)(t2A,t2B,t2C),Saw-1Saw-2,(t2A,t2B,t2C).(t1A,t1B,t1C)t1AT1A,t1BT1B,t1CT1C,t1A+t1B+t1C24.(4)T1A,T1BT1C,A,BCSaw-1.Saw-1,3Saw-2,T1A=T2A-t2A,T1B=T2B-t2BT1C=T2C-t2C.Saw-13Saw-2,24h.Step3:.,.Step4:.,,A,BC,(:).Step5:(500).,;,.,,.,,,,,.972　　　控　　制　　与　　决　　策第21卷5.2,.:M=60,vmax=xmax,k1=0.4,k2=0.9.,3,.34.33A/hB/hC/h/%Saw17.913.875.0570.12528.064.093.6365.75DA17.235.927.5986.4227.186.975.5081.87536.233.4810.6284.7147.551.5110.3080.6758.940.927.9774.29WB113.9610.040100214.039.9710036.0117.99100423.970.03100523.520.4799.96Molding120.0083.3326.0325.12537.0029.17Marking111.722.174.9078.2928.561.943.9760.2939.981.404.2565.125411.103.696.5989.084SawDAWBMoldingMarking/%67.9481.5999.9945.87558.561.051.35809.3133.982,77.59%,3.406.,,,.,,..1d,,,.,.6结　　论.,,,,.,,.(References)[1]ShenYX,LeachmanRC.StochasticWaferFabricationScheduling[J].IEEETransonSemiconductorManufacturing,2003,16(1):2-14.[2],.B-T[J].,2005,17(4):993-996.(LvWY,DangYZ.SchedulingRe-entrantLinesBasedonGAandIntegratedRules[J].JofSystemSimulation,2005,17(4):993-996.)[3],.[J].,2003,9(8):641-644.(WuJW,XiaoYS.Multi-agentTechnologyinSchedulingofSemi-conductorProductionLine[J].ComputerIntegratedManufacturingSystems,2003,9(8):641-644.)[4]KennedyJ,EberhartRC.ParticleSwarmOptimization[A].ProcofIEEEIntConfonNeuralNetworks[C].Perth:IEEEPiscataway,1995:1942-1948.[5]EberhartRC,ShiY.ParticleSwarmOptimization:Developments,ApplicationsandResources[A].ProcofCongressonEvolutionaryComputation[C].Seoul:IEEE,2001:81-86.[6]EsminA,LambertTorresG,ZambronideSouzaAC.AHybridParticleSwarmOptimizationAppliedtoLossPowerMinimization[J].IEEETransonPowerSystems,2005,20(2):859-866.(下转第978页)973第9期康琦等:半导体生产线工序参数的逻辑时序微粒群优化策略　　　;2);3);4)[6,7].5.4DBSCANSUDBCDBSCAN,;SUDBC.43.4,SECDU,,.,SECDU.436结　　语SECDU.,.,SECDU.,.(References)[1]MacqueenJ.K-means:SomeMethodsforClassificationandAnalysisofMultivariateObservations[A].The5thBerkeleySymponMathematicalStatisticsandProbability[C].Berkeley,1976:56-68.[2]MarkusM,Breunig,Hans-PeterKriegel,etal.DataBubbles:QualityPreservingPerformanceBoostingforHierarchicalClustering[A].ACMSIGMOD[C].SantaBarbara,2001:99-112.[3]SamerNassar,JorgSander,CorrineCheng.IncrementalandEffectiveDataSummarizationforDynamicHierarchicalClustering[A].ACMSIGMOD[C].Paris,2004:13-18.[4]GuhaS,RastogiR,ShimK.CURE:AnEfficientClusteringAlgorithmforLargeDatabases[A].ACMSpecialInterestGrouponManagementofData[C].Washington,1998:73-84.[5]ZhangT,RamakrishnanR,LivnyM.BIRCH:AnEfficientDataClusteringMethodforVeryLargeDatabases[A].ACMSIGMODIntConfonManagemen