硕士论文-基于群集智能的产品公差优化设计方法研究

华中科技大学硕士学位论文基于群集智能的产品公差优化设计方法研究姓名：邹洪富申请学位级别：硕士专业：机械设计及理论指导教师：肖人彬20060429IParetoParetoParetoPareto(TSP)(GA)(SA)IIAbstractManufacturingisaveryimportantindustryinnationaleconomy;itsconditionshowsnationalindustrializationlevel.Manufacturingisfacingfiercemarketcompetition,manufacturingcost,productqualityandtimethesereflectthemanufacturingabilityofanenterprise.Someadvanceddesignmethodwaspresentedbysomeresearcherstomeethigherconsumers’requests.DesignforQuality(DFQ)isoneofadvanceddesignmethod,whichbelievesthatdesignstagehaseffecttoproductqualityintensively.ToleranceoptimaldesignisamainpointinDFQfortoleranceisrelatedtoproductdesignandmanufacturingatthesametime.Atolerancedesignmulti-objectivemodelandahybridswarmintelligencealgorithmweregiveninthepaperbyresearchingtoleranceoptimaldesign.Tosolvetheproblemthatproductqualityisignoredincost-tolerancemodel,amulti-objectivemodeloftolerancedesignispresented,whichisbasedonTaguchi’squalityviewandtheconceptionofParetooptimumset.Manufacturingcostandqualitylossaretakenasdesignobjectiveatthesametime,theobjectivesaresubjecttoassemblysuccessrateofstatisticaltolerance,andthetolerancezoneobtainedislooserthantheworsttolerancemethod.Thetraditionalparticleswarmoptimizationalgorithmisimproved,theparticleisredefinedaccordingtotheconceptionofParetooptimum,andthenthefastnon-dominantsortingtechnologyisadoptedtosequencetheparticlesbytheirfitnessvalues,somulti-objectivemodeloftolerancedesigncanbesolvedintheimprovedalgorithm.Usedtoengineeringexample,onlyforonerunninggoodParetooptimumsetwasobtained.Solutionscanbeselectedaccordingtomanufacturingrealityandmarketdemand.BytheanalysisofParetofront,thetolerancedesigncharacteristicofthiskindofpartcanbegot;thegenerallawsoftolerancedesignarealsovalidatedbytheresults.Traditionaltolerancedesignincludestwosequencedstepswhicharedesigntoleranceandmanufacturingtolerance,Serialdesignmodeoftraditionaltoleranceischangedbyconcurrenttoleranceoptimaldesign,whichisvirtuallyhybridvariablecombinationoptimizationIIIproblem;itwasdefinedasaspecialkindofTSP,sosolvingbecomessimpler.Makingthebestofantcolonyoptimizationandparticleswarmoptimizationinsolvingdiscreteproblemandcontinuousproblem,ahybridswarmintelligencealgorithmofthemwaspresented,whichwasappliedtoanexampleofconcurrenttoleranceoptimizationdesignandsatisfiedresultsweregained.Thealgorithmishighefficientandshowsstronglysearchingabilitycomparedwithgeneticalgorithmandsimulationannealingalgorithm,It'sbetterwhenmulti-workingprocedurecombinationexplosionishappened,forit'sconcurrentcharacteristicandinformationsocialshare,andanewwayisalsogivenforsolvinghybridvariableoptimization.Keywords:ToleranceoptimaldesignMulti-objectiveoptimizationConcurrenttolerancedesignHybridvariableoptimizationSwarmintelligence11604740771.1V.Hubka[1]M.Morup[2]DFQ(DesignforQuality,DFQ)CADCAPPCAMCAD/CAMCIMS(ComputerAidedTolerancingCAT)CADCAPPCAMCAD/CAM/CAPP2Pareto[3]011.23[4]1978C.Hillyard[5]O.Bjorke[6]Computer-AidedTolerancing1983A.A.G.Requicha[7]1988R.Wei[8]TolerancingforFunction”[9][10][11]Ngoi[12][13][14][15][16]Langrage[17][18][19]MonteCarlo[20]()[21][22][23]4CATCATCAT[4][24]1.3(SwarmIntelligence,SI)[25,26]Bonabeau[27]AgentAgentAgentAgentAgentAgent[28]AgentAgentAgent[29]()()51.3.12090Dorigo,Mahiezzo,Colorni(AS)[30,31](TSP)()TSP[32,33]job-shop[30,33][30,34]()1996GambardellaDorigo(AAS)[35,36]ASStutzle(MMAS)[37][τmin,τmax]MMASTSP,QAPACO(AntColonyOptimization,ACO)1998ACO(ANT'98)[38][39]G.BilchevI.C.ParmeeACS[40]1.3.2(PSO)1995KenndyEberhart[41,42]PSOPSOPSO(CEC)PSO6ClercVandenBergh[43,44,45]PSOPSOPSOPSOPSO(HPSO)PSOPSO[46]ShiEberhart[47,48]wwPSOwwPSOwPSO(HPSO)[49]PSOPSOPSOPSOKenndyEberhartPSO[50]PSO(TSP)PSOPSOPSOPSO[51][52]PSO[53]PSOPSOPSOPSO(1)PSOPSOPSOwvidwvidc1rand()(pidxid)c2Rand()(pgdxid)(1.1)xidxidvid(1.2)Shi(1998)w=0.90.47(2)PSOKennedyEberhart[50]PSOClerc[54]PSO(TSP)PSOPSOPSOvidwvidc1rand()(pidxid)c2Rand()(pgdxid)(1.3)1111()1;0kkkkididididifsigvthenxelsexr++++==(1.4)111()1exp()kidkidsigvv++=+-sigmoid1[0,1]kidr+∈1kidr+n(0,1)PSOPSO(3)PSO(HPSO)PSO(HybridPSO,HPSO)PSOAngeline[55]1998AngelineHPSOPSOBenchmarkPSOLovbjerg,RasmuwsenKrink2000PSOHPSO112()*()(1.0)*()childXpparentXPparentX=+-(1.5)221()*()(1.0)*()childXpparentXPparentX=+-(1.6)XDchildk(X)parentk(X)k=1,2PDP[0,1]121112()()()()()()parentVparentVchildVparentVparentVparentV+=+(1.7)122212()()()()()()parentVparentVchildVparentVparentVparentV+=+(1.8)8PSOPSOPSOPSOPSO(4)ClercClercPSOvidk[vidc1rand()(pidxid)c2Rand()(pgdxid)](1.9)1222,424kccjjjjj==+---(1.10)xidxidvid(1.11)j4.1k(1.10)0.729VmaxVmax100000Xmax()1.3.3[56][57]1.491.4.1ParetoPareto[58]()ParetoGA1985Schafer[59]ParetoHajela[60]ParetoFonseca[61]ParetoParetoParetoHorn[62](Niche)ParetoParetoParetoZitzler[63]Pareto4Pareto[64]Pareto[65]Pareto[66]()101.4.2Eberhartkennedy(1995)(ParticleSwarmOptimizationPSO)()PSO()PSO[67][68][69]PSOParsopoulosVrahatis[70]PSOPSOXiaohuiEberhart[71]PSOCoelloLechuga[72]PSO(repository)PSOPSOPSOPSO1.5(1)11Pareto(2)01(3)ParetoPareto121.11.1132.12.32.2[73]2.2.1(DimensionalChain)(Tolerancechain)(1)(2)(Link)(closing)2.1142.1x0x1x2x3