基于神经网络的扣式电池自动化生产线故障诊断系统研究

河北工业大学硕士学位论文基于神经网络的扣式电池自动化生产线故障诊断系统研究姓名：闫少伟申请学位级别：硕士专业：机械制造及其自动化指导教师：林树忠20061101iiiRESEARCHONTHEFAULTDIAGNOSISSYSTEMOFLINEARAUTOMATICASSEMBLYLINEFORBUTTONBATTERIESBASEDONNEURALNETWORKABSTRACTNowadays,theprominentcharacteristicofproductionisautomaticandintelligent.Nowwiththegood-sized,high-speedofequipments,traditionaldiagnosetechnologyisnotenoughatall.Theexpensewouldbeverylargeoncethemodernproductionstoppedbecauseofthefaultofmachine.Sopeopleneedcomplex,exactandtimelyfaultdiagnosissystem.Alltheseforcedmentoinvestigatefaultdiagnosis,therisingresearchdomainisformed.Firstinthispaper,accordingtothesystemviewpointofassemblycraftandmechanicalautomationproductdesignofLi/MnO2buttonbatteries,anewlinearautomaticassemblylineforLi/MnO2buttonbatteriesisproposed.Wealsointroduceitsmechanicalandcontrolsystemdetailedly.Becausetheautomaticproductlineisacomplexsystemwithmoreinputsandoutputs,asthehard-core,thepaperproposedafaultinspectanddiagnosismethodbasedonNeuralNetwork.Theprincipleisasfollows:VC++isusedastheforegroundapplicationandSQLServer2000adthedatabaserunningbackstage.ThencollectedthenecessaryIn-Outsignals,basedontheanalysisanddisposalofthedata,throughdeeplyanalysisofmastersthinkingwhentheycheckedormaintainedasetofequipments,faultsofanfixturecanbelocatedbyviableknowledgeandheuristicconsequencemethod.Themethodscanusemaster’sexperienceadequately,soitcanreducetheworkoftheprocedureandenhancetheprecisionandtheveracityoftheresulttogreatextent.Bythecheckoutofpractice,itcansatisfytheclients.From80sinlastcentury,thediagnosetechnologyenterintointelligentstage,byhardlyexploreofvariouscountriesandexperts,theNeuralNetworktechnologyusedinfaultdiagnosishasobtainedplentifulproduction.Inthelast,thispapersummarizedandreviewedtheresearchesandapplicationsofNeuralNetwork,andgiveabriefpresentationabouttheapplicationforegroundofneuralnetworkusedinfaultdiagnosis.KEYWORDS:automaticproduct-line,neuralnetwork,faultdiagnosis,datacollection,SQLServerdatabase12SYsyggYS→ffSY→Fig.1.1Thepresentationoffaultdiagnosisfgfns1s2snsisfg3iY1iyimyy1ymysFig.1.2Theprocessoffaultdiagnosis456Fig.2.1AssembletechnologyofLi/MnO2buttonbatteries7Fig.2.2ThesamplelineforLi/MnO2buttonbatteries8Fig.2.3ThecontrolsystemdendrogramofproductlinePLC9×sµsµ1011Fig.2.4ControlflowchartofvibrationhopperFig.2.5TheconnectionmodeforPCandadapterYNNCD1SD3ER4SG5DR6CSCI89—1SD2RD3RS4CS5—677—8SG9RD2—RSIBMRS232C12Fig.2.6ThecableconfigureofserialportcommunicationbetweentouchscreenandPLCPWSCOM(25pinfemale)PLCCOM(9pinfemale)RXD3TXD2GND7RTS4CTS53SD2RD5SG8CTS7RTS6DSR1CD4DTR13Fig.2.7ThesettingareaforsystemFig.2.8Thedebuggingmenuforthewholemachine1411104.1×5310~10~5310~1015Fig.3.1Thesketchmapofneuronmodeixjijw1xMixMnxjojojojof1xMixMnxjw1ijwnjw1xMixMnx1xMixMnxjw1jw1ijwijwnjwnjw16joix()ttjijo()tjjjo()t=()⎭⎬⎫⎩⎨⎧−⎥⎦⎤⎢⎣⎡−∑=jijiniijTtxwfτ1ijτjTjijwij()fjo()1+t=()⎭⎬⎫⎩⎨⎧−⎥⎦⎤⎢⎣⎡∑=jiniijTtxwf1ixinjojwijjTt'jnet()t=xwniij∑=1i()t'net()tj'net()tjT−jo()1+t()txiijw()t17jWX'jnet=XWjΤjWXjW=()Τnwww,...,,21()Τ=nxxxX,...,,2110−=xjTw=000wxTj=−XWxwnetTnetjiniijjjjΤ====−∑0'jWXjnet()()XWfnetfojjjΤ==()=xfx'jnetjT−'jnet≥jT'jnetjT()xexf−+=1110≥x00x18()xxxeeexf−−−+−=−+=11112()=xf()TxeP−+=111TFig.3.2Thenetworkstructureofmonolayer00≤xcxcxx≤01xxc19Fig.3.3ThelevelnetworkarchitecturewithconnectionbetweeninputlayerandoutputlayerFig.3.4Thelevelnetworkarchitecturewithinterlinkageinonelayer……Fig.3.5TheintervolvingnetworkstructureFig.3.6Thepartintervolvingnetworkstructure20W21jXijijwjjWjjwT0=0x1−()jjdXWrr,,=jWXjdjWt()tWj∆t()tXr()()()()[]()tXtdtXtWrtWjjj,,η=∆η()()()()()[]()tXtdtXtWrtWtWjjjj,,1η+=+∆()jjdXWr,,Fig.3.7Thecommoninstanceofweightadjustingjw0Xjw1ijwnjwjjO1−1xMixMnxjdηXjW∆()jjdXWr,,22ij()XWfrjΤ=()XXWfWjjΤ=∆η()ijijijxoxXWfwηη==∆Τni,,1,0L=()0=t()0jWjiodr−=jd()XWfojjΤ=()XWfojjΤ==()XWjΤsgn=()[]XXWdWjjjΤ−=∆sgnη()[]ijjijxXWdwΤ−=∆sgnηni,,1,0L=jd(){}1,1sgn−∈ΤXWjXWjη2±=∆1–10ΤXWj0≥ΤXWj()23δδδ()[]()()()jjjjjjnetfodXWfXWfdr''−=−=ΤΤδ()XWfjΤ'()jnetfδδ()()[]222121XWfdodEjjjjΤ−=−=EWj∇−=∆η()()XXWfodEjjjΤ−−=∇'()()XnetfodWjjjj'−=∆ηηXδjW∆()()ijjjijxnetfodw'−=∆ηni,,1,0L=δXWdrjΤ−=()XXWdWjjΤ−=∆ηjW∆()ijjijxXWdwΤ−=∆ηni,,1,0L=δ()XWXWfjjΤΤ=()1'=ΤXWfjδ24jdr=jW∆ijw∆XdWjjη=∆ijijxdwη=∆ni,,1,0L=jdixijw∆ijxdjjdo=Xp()XWXWipimΤ=Τ=,,2,1maxLmW()mmWXaW−=∆(]1,0∈amWXXXjWXjW25mdjW()jjWdW−=∆ηηajXjdFig.3.8InnerstarnodeandouterstarnodennXmmOjjjw1ijwLnjwLnixxxLLLLLL1jw1ijwLmjwLmiyyyLLLLLL126jWmW()Τ=nixxxxX,,,,,21LL()Τ=mjooooO,,,,,21LL()Τ=njijjjjwwwwW,,,,,21LLmj,,2,1L=jnetiniijjxwnet∑==1jo()()XWxwTnetojniiijjjjΤ==⎟⎠⎞⎜⎝⎛=−=∑sgnsgnsgn00ΤXWj0ΤXWj()Τ=21,xxX1ojomo……1x2xixnx……1WjWmWFig.3.9Monolayerapperceiveimplementjo1x2xixnx……Fig.3.10Singlecalaculatingnodeapperceiveimplementjw1jw2ijwnjw=jo27j02211−+jjjTxwxw02211−+jjjTxwxw02211=−+jjjTxwxw0jnet0jnet()Τ=321,,xxxX0332211−++jjjjTxwxwxw0332211−++jjjjTxwxwxw0332211=−++jjjjTxwxwxw0jnet0jnetn()Τ=nxxxX,,,21Lnn02211=−+++jnnjjjTxwxwxwLn=jo=jo28()Τ=nixxxxX,,,,,21LL10−=x()Τ=mjyyyyY,,,,,21LL10−=y()Τ=lkooooO,,,,,21LL()Τ=lkddddD,,,,,21LLV()mjVVVVV,,,,,21LL=jVjW()lkWWWWW,,,,,21LL=kWk()kknetfo=lk,,2,1L=jmjjkkywnet∑==0lk,,2,1L=()jine