基于流型识别的气液两相流相含率估计方法研究

《基于流型识别的气液两相流相含率估计方法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、天津大学硕士学位论文基于流型识别的气/液两相流相含率估计方法研究姓名：吴萌萌申请学位级别：硕士专业：检测技术与自动化装置指导教师：董峰20070101AbstractThetwo-phaseflowsystemsareverycommoninmanyindustryfields，suchaspetroleum，chemistryengineering，metallurgy,energypower,pharmacyandSOon．Becauseofitscomplexityandvariability,thereal-timeparametermeasurementoftw

2、o—phaseflowbecomesadifficultissueinthefieldofengineeringandscience．Therefore，theadvancedsensorandmodeminformationprocessingtechniquesarecombinedtoprovidethenewmeasurementmethodoftwo-phaseflow．Asanon-linemeasurementtechniquewiththeadvantagesofvisualization，non—invasion，lowcosLandnon—radiat

3、ion，ElectricalResistanceTomography(ERT)basedontheprincipleofelectricalsensitivityisplayinganincreasingly-importantroleinthefieldoftwo—phaseflow．Inthispaper,bycombiningtheERTtechniquewiththemodeminformationprocessingtechniques，theidentificationofflowregimesandtheestimationofvoidfractionwer

4、eachieved．Themainachievementsinthispaperareasfollows：1．Thereal—timeexperimentswereconductedusingERTsysteminoil／gas／liquidthree—phaseflowlaboratoryofTianjinUniversity,andthedatawithdifferentflowregimesofgas／liquidtwo-phaseflowinverticalupwardpipewereobtained．2．Accordingtothereal-timeexperi

5、ments，thesoft—sensormodelofgas／liquidtwo—phaseflowinverticalpipehasbeenestablished．Threefeaturesinamplitudedomainareextractedbytimeseriesstatisticalanalysis，andthefeaturesofmultiplescaleswaveletenergypercentarederivedbydiscretewaveletstransform．Allthefeaturesconstitutedthefeaturequantitie

6、sdatabaseofsoft-sensormodel．TheSupportVectorMachine(SVM)theoryisadoptedastheclassificationmodeltoidentifythethreetypicalflowregimes。Bycomparingthetwomethods，theresultsindicatethatthemethodofmulti．featureinformationfusionwhichcombinedthetwofeatureextractionmethodscouldbetterexpresstheessen

7、tialinformationofgas／liquidtwo·phaseflow,andachieveahigheridentificationrateofflowregimes．3．Basedontheaccurateflowregimesidentificationofgas／liquidtwo·phaseflow,anewmethodforestimatingthevoidfractionofgas／liquidtwo·phaseflowhasbeenbroughtforward．ByusingANSYSsimulati