資源描述:
《基于SVM的復(fù)雜制造系統(tǒng)動態(tài)調(diào)度方法研究與應(yīng)用》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在工程資料-天天文庫。
1、碩士學位論文基于SVM的復(fù)雜制造系統(tǒng)動態(tài)調(diào)度方法研究與應(yīng)用摘要在制造業(yè)及自動化領(lǐng)域,復(fù)雜生產(chǎn)過程調(diào)度問題一直是研究的熱點。復(fù)雜制造系統(tǒng)具有大規(guī)模、多日標、工藝復(fù)朵、不確定性等特點,合理選擇調(diào)度策略對整個生產(chǎn)調(diào)度過程產(chǎn)生重耍影響。在給定的生產(chǎn)狀態(tài)下,為了能快速較優(yōu)地選擇調(diào)度策略(即動態(tài)調(diào)度),可以充分利用與調(diào)度相關(guān)的歷史生產(chǎn)數(shù)據(jù)來挖掘出相應(yīng)的調(diào)度策略。同時,為了提高調(diào)度的效率,應(yīng)當對生產(chǎn)數(shù)據(jù)進行去冗余處理,即對生產(chǎn)展性進行特征選擇。本文研究了復(fù)雜制造系統(tǒng)動態(tài)調(diào)度策略選擇方法,提出了一種基于生產(chǎn)屈性的動態(tài)調(diào)度策略選擇方法,該方法以歷史數(shù)據(jù)為基礎(chǔ),選取支持
2、向量機(SVM)作為數(shù)據(jù)挖掘工具,采用二進制粒子群優(yōu)化算法(BPSO)對生產(chǎn)屈性(特征)子集進行尋優(yōu),獲得基于SVM的動態(tài)調(diào)度策略分類模型。在此基礎(chǔ)上,對于任意給定的生產(chǎn)狀態(tài),通過該模型,能實時地獲取當前生產(chǎn)狀態(tài)下近似最優(yōu)的調(diào)度策略。另外,本文還對基于多目標的調(diào)度策略綜合評價方法進行了研究,并分別應(yīng)用功效函數(shù)法和信息爛法實現(xiàn)多目標決策評價和指標權(quán)重系數(shù)選擇,并應(yīng)用丁調(diào)度策略選擇方法的研究中。最后,本文針對某實際復(fù)雜制造系統(tǒng),對所提岀的動態(tài)調(diào)度方法進行實驗驗證。研究結(jié)果表明,該方法很好地提高了調(diào)度效率,保證了調(diào)度的實時性和準確性。關(guān)鍵字:動態(tài)調(diào)度,特征
3、選擇,SVM,參數(shù)優(yōu)化,多目標決策評價AbstractTheschedulingofcomplexproductionprocessispopularinthefieldofmanufacturingandautomation.Characterizedwithlarge-scale,multi-objectiveanduncertainenvironments,thecomplexmanufacturingsystemsshouldmakereasonablechoiceofschedulingstrategiesthathasgreateffec
4、tontheiroperationalperfonnance?Inordertochooseabetterschedulingstrategyquicklyunderagivenconditionofproduction(dynamicscheduling),itisagoodwaytomakefulluseofon-lineandoff-lineproductiondatarelatedwithscheduling?Meanwhile,inordertoimprovetheschedulingefficiency,someredundantdata
5、shouldberemoved,namelyaprocessoffeatureselectionofproductionattributes.Inthisthesis,adynamicschedulingstrategyselectionmethodforcomplexmanufacturingsystemsisstudiedandaschedulingstrategyselectionapproachbasedonproductionattributesisproposed.Thisapproachbasedonhistoricaldatauses
6、supportvectormachine(SVM)asadataminingtoolandbinaryparticleswarmoptimizationalgorithm(BPSO)tooptimizeproductionattributes(featuresubsets).Thus,thebestschedulingstrategycanbeachievedunderanygivenproductionstatusinreal-timethroughaSVMclassifierafteroptimization?Inaddition,themult
7、i-objectivecomprehensiveevaluationmethodsforschedulingstrategiesarestudiedinthisthesis.Inthestudyofthedispatchingstrategyselectionapproach,anefficacyfunctionmethodandaninformationentropymethodareappliedtomulti-objectivedecision-makingevaluationandindexweightcoefficientselection
8、respectively.Finally,thedynamicschedulingmethodpropose