产能约束下半导体芯片测试生产线调度优化-研究

《产能约束下半导体芯片测试生产线调度优化-研究》由会员上传分享，免费在线阅读，更多相关内容在应用文档-天天文库。

1、西南交通大学硕士研究生学位论文第JI页AbstractTheperformancecheckingofpackagedintegratedcircuitsisnamedsemiconductorfinaltesting．Nowadays，thecostoffinaltestinghasarisentothirtypercentofthetotalcostinintegratedcircuitsmanufacturingduetotheincreaseofintegrationandcomplexity．Thesemiconductorfinaltestinghasbecomeabot

2、tleneckinthedevelopmentofChina'ssemiconductorindustryandattractedtheattentionsofresearchersinmanagementscience，industrialengineeringandautomation．Semiconductormanufactureschedulingisallimportantcombinatorialoptimizationproblemcharacterizedbylarge—scale，reentrant,mixedprocessingmodeandmulti·reso

3、urceconstraints．Foralongtime，researchersconsideredresourceconstraintonlyasmachinecapacityconstraint,assumingotherresourceswereinfinite．Moreover,jobsequencedependentsetuptimeWasrarelyconcernedinresearchesonsemiconductormanufacturescheduling．Therefore，hardlyCaneveryresolutionsappliedtothesemicond

4、uctorfacility．Inthisthesis，theschedulingproblemencounteredinsemiconductorfinaltestingoperationWasresearched，consideringthemultipleresourcesconstraintsandjobsequencedependentsetuptimes。Wesimplifiedtheproblemasaflexibleflowshopschedulingproblemwithtwostages，testandbum-in，toprocessandusedadivide-a

5、nd-conquerstrategytosolveit．Atfirsttwostageswereoptimizedseparately,thentheywereintegratedtoatwo-stageflexibleFlowshopschedulingproblemandacombinatorialparticleswarmoptimizationalgorithmWasdesignedtosolveit。Theschedulingproblemonteststagewasrepresentedbyanidenticalparallelmachineschedulingprobl

6、emwithmultipleresourcesconstraintandjobsequencedependentsetuptimes．Amixedintegerprogrammodelandtwoalgorithms，includingaGeneticAlgorithmandaVariableNeighborhoodSearchalgorithm，wereproposedforthisproblem．Wedescribedtheschedulingproblemonbum—instageasasinglemachineschedulingproblemwithjobsequenced

7、ependentsetuptimesandusedcommercialoptimizationsoftwareandsixdifferentalgorithmstosolveit．Experimentalresultsshowthattheheuristicperformedmuchbetterthanotheralgorithmsonsolvinglarge—scaleproblems．Aslast，wedesignedacombinatorialpar