硕士论文-基于神经网络PID的聚乙烯醇生产过程控制

华东交通大学硕士学位论文基于神经网络PID的聚乙烯醇生产过程控制姓名：鲁伟栋申请学位级别：硕士专业：交通信息工程及控制指导教师：杨辉20080418IPIDPIDPIDPIDPIDPIDPIDPIDPIDPIDPIDPIDBPABSTRACTIIPOLYVINYLALCOHOL(PVA)PRODUCTIONPROCESSCONTROLBASEDONNEURALNETWORKPIDABSTRACTTheMid-TemperatureControlofVinylacetate(VAC)CompoundReaction(MTCVCR)isanimportantpartofthePolyVinylAlcohol(PVA)productionprocess(PPP),andnowthePPPismainlyusednormalPIDcontroltechnology.ItisdifficultforthetraditionalPIDcontrolmethodtoreachanoptimumresultincertaincasesbecausetheMTCVCRisratheraclosedloopsystem,whichhasthecharacterofnon-linear,timevaryingandbighysteretic.Intelligentcontrolindependentofmodelofaplantandbasedonknowledgeoffersanewideaforimprovingtheprocesscontrolquality,ofwhichneuralnetwork,asoneofmoderninformationprocesstechnologies,havesomeadvantagesinmanyapplications.Neuralnetworkcontrolbecamearegardedresearchdirection.AndnerualPIDcontrolisoneofthemaintechniquesinresearchandapplicationduetothedominantplaceofPIDtypescontrolinindustrialprocesscontrolrecently.IfthecontrollerehichhasbetterperformanceandiswieldylikePIDcanbefound,therewillbesignificanceinboththeoryandpractice.ThisarticletakesJiangxiChemicalFibreandChemicalEngineeringLimitedLiabilityCompanyPVAextensionandreconstructionprojectasthebackground.AfteranalyzingtechnicalcharacteristicsofVACproductionprocessandbeingcombinedwiththeproblemswhichexistinpractice,theneuralnetworkcontrolmethodhasbeenimprovedwithPIDcontrolandmakesuseoftheiradvantagesincontrolling,andappliesittotheMTCVCR.Inthisway,wecangetagoodeffect,andsolvebigtimelag,strongjammingandsoon,thichtraditionalcontrolmethodscan'tsolve.Bythesimulationincomputerandanalyzingthetheory,thePIDcontrolmethodbasedonneuralnetworkhasbeenprovedtobeeffective.Atlast,amid-temperaturecontrolofVACCompoundReactioncontrolsystemisdesignedaccordingtothetheoryofneuralnetworkPIDcontroller,anditprovesthattheneuralnetworkPIDcontrolmethordsarefeasibleinthisproject.KeyWord:PolyVinylAlcohol(PVA),Mid-TemperatureofCompoundReaction,ProcessControl,NeuralNetworkPIDControl,BPArithmetic___________________________________________________________21.150[1],13607079%63%,30,10,,PIDPIDPIDPID1.21912F.Klatte60-1005%1921ConsortiumfurElectrochemischeIndustrie1925Shawinigan/1928Hoechst1.2/Wacker[2]350DuPont2/Celanese2.2719531970[3]60BordenBalawKnox30%Borden[4]RhonePoulene19601965ICI3/1969BayerHoechstUSI[5.6]1968Bayer6/;1970USI13.6/80%30%25%2004VAC550490HoechstAktiengesellschaftBPChemica1sVAC40%UnionCarbideMi1lenniumDuPontAcetexKurarayVAC1-11-1VAC%%15015010077.959.976.891823.11666610057.657.610041.741.710020201004151510012.112.110011.211.210011.511.51005510052.5502.550551002.62.61002210022100221002210022100489.989.418.25400.581.751.3(VAC)50(PVAC)(PVA)(PVA)(VAE)(EVA)(PVAC)(EVC)(EVOH)(vc/vAc)[7](PVB);(VC/VAC)VC/VAC[8]5AVC-(EVOH)(VAE)(EVA)EVOH2000200410-15%VPBPVB2004VAC4400kt(PVAC)44%(PVA)41%-(EVA)9%6%1995-20032.5%2003-20082008-20133.4%2.1%PVA80%PVACEVA14%3%3%6.9%[9][10]80%PID95%[11]1.3.1↑+→+COCaCCCaO231-1CO2CO1kg10~20kg6↓+↑→+22212)(2OHCaHCOHCaC1-21.3.2232322)()(OCOCHZnCHHCOCOCHZnHC•≡→+1-3)()(3323OCOCHZnCHHCOCOCHOCOCHZnCHHC•=−→•≡1-42323333)()(OCOCHZnCHHCOCOCHCOOHCHOCOCHZnCHHCOCOCH+=−→+•=−1-5CHOCHOHHC3222→+1-6CHOCHOCOCHOCOCHCHCH323233)()(+→1-7OHCHOCHCHCHCHOCH22232+→1-8CHOCHCHCHCHOCHHC22322→+1-966223HCHC→1-10OHCOCOCHCOOHCH22233)(2++→1-11ZnOCOCOCHACZn++→2232)()(1-12:7:1.3.31.4CS1000DCSPIDCS1000CS1000PIDPIDPIDPID81.5PIDPIDPVAVAC1PIDPIDPIDPID2PIDPID(SNPC)BPPIDBPPIDPIDPID3matlabMC++VC++BPPIDPIDPIDPID0.5PIDBP9PIDBP2.1PIDPIDPIDPID(P)(I)(D)PID2.1.1PIDPIDPID2-1PID)(tr+−)(te+++++)(1tu)(tu)(tv)(tc2-1PIDFig.2-1BlockDiagramofSimulationPIDControlSystemPID)(tr)(tc)()()(tctrte−=2-1++=∫10)()(1)()(dttdeTdtteTteKtuDIp2-2)11()()()(sTsTKsEsUsGDIp++==2-3PIDBP10PK--IT--DT--PID(1))(te(2)ITIT(3)()2.1.2PID(2-2)PID[12][])2()1()()1()1(10−−−++−=−∑−=kekeKjeKkeKkuDkjIP2-4)(ku)1(−ku[][][])1()()()()2()1(2)()()1()()(−∆−∆++∆=−+−−++−−=∆kekeKkeKkeKkekekeKkeKkekeKkuDIPDIP2-5)1()()(−−=∆kekeke2-5PID2-2PIDPIDr+−c∫eu∆u2-2PIDFig.2-2BlockDiagramofIncrementalPIDControlSystem2-5PIDBP11)2()1()()(−+−−=∆kCekBekAeku2-6)1(TTTTKADIP++=)21(TTKBDP+=TTKCDP=ABCe(k)c(k)r(k)e(k-2)=e(k-1)e(k-1)=e(k)u(k)2-3PIDFig.2-3BlockDiagramofIncrementalPIDControlAlgorithm2.1.3PID(IPIPID)(PPD)PIDBP12PIDPID[13-15]aPIDPIDPKITDT(DT)12IT(0.8)3T:PK1.6~5IT3~10(min)DT0.5~3(min)PIDBP13PIDbPIDPIDPIDPID1rKrT-(Ziegle-Nichols)(2-1)2-1PKITDTP0.5rKPI0.45rK0.85rTPID0.6rK0.5rT0.12rTPID)()(0202∫∫∞∞=dtedte1.05PIDBP142-120urKuT(2-4)2-22-2TPKITDTPI0.03rT0.53rK0.88rT1.05PID0.014rT0.63rK0.49rT0.14rTPI0.05rT0.49rK0.91rT1.2PID0.043rT0.47rK0.47rT0.16rTPI0.14rT0.42rK0.99rT1.5PID0.09rT0.34rK0.43rT0.2rTPI0.22rT0.36rK1.05rT2.0PID0.16rT0.27rK0.4rT0.22rTPIDBP15yt∞y0uKKir=uTgTiguKTTy=∞2-4Fig.2-4StepCurveMethodtoDetermineBenchmarks2-3PKITDTPrK1PIrK8.03uTPIDrK2.12uT0.42uT:1/41/4PIDPIDPIDBP161.2.3.idTTδdiTTδδ0idTTδ→∞=0idTTδ→∞=δ4:11.70.50.125idkkkTTTTδδ===0.80.30.1idsssTTTTδδ===12NY3NNYY2-5Fig.2-5RegulatorDiagramofParameterSettingProcedurePIDBP17δ-dT-kδ-kT-sδ-41sT-41iT-2.2BP(ANNArtificialNeuralNetworks)()()()[16](MFNNMultilayerFeedforwardNeuralN