GTAW熔池圖像處理及熔池辨識模型研究.pdf

1、廣東工業(yè)大學(xué)碩士學(xué)位論文GTAW熔池圖像處理及熔池辨識模型研究姓名：劉明濤申請學(xué)位級別：碩士專業(yè)：機(jī)械電子工程指導(dǎo)教師：高向東20080501Abs仃actAsallimportantmanufacturingtechnologyGasTungstenArcWelding(GTAW)iswidelyusedinindustryRealtimeweldarccontrolisimportanttoweldingautomationandt

2、oimprovejointqualityBuildingallanalysismodelofweldpooldynamicsismainstepofweldingprocesscontrol，isveryimportanttojointqualityVisionsensorCancapturealotofweldpoolinformation，howtodetectthepoolinformationeffectively，pickup

3、characterparametersofweldingprocessandhowtobuildaveraciousidentificationmodelispreconditionforweldingautomationcontrolresearchThethesisisfocusedontheweldpoolidentificationmodelbasedonvisionsensewhichtheparametersarethewe

4、ldpoolwidthandpenetration，andithasbeenresearchedbyasetofweldarccontrolexperimentapparmuswhichincludetheCCD(Chargecoupleddevice)sensor，theworkingplatformandthestepdrivecontrolsystemandSOonFirstlytheCCDvisionsensorandother

5、assistantapparatusareusedforgatheringclearimageofmoltenpoolfromtheobverseweldinglinewithmanytechnicalexperimentsSecondlythepoolimageisfilteredbythetraditionalfilteringmethods，then，theedgeofmoltenpoolimageisenhancedbytop—

6、hatandbottom—hat(top—bot)ofGrayscaleMorphology，andtheimageisprocessedtobinaryimage，inthenextstep，themoltenpoolimageisprocessedbydilation、erosion、open、closeandcombinatorialmorphologybetweenthemThirdly，theedgeisdetectedbye

7、dgedetectionoperatorwhichmakeupofthebasicoperation，andthentheshapeinformationofmoltenpoolindifferentweldingprocessareextractedavailablyFinallybasedonthisinformation，theweldpoolimagecharacterparameters，forexample，thepoolw

8、idthandSOon，arecalculatedAstechnicalexperimentresearchprovidesthefoundationforcaptureofweldpoolandmodelingforweldpooldynamicsThispapermakesmanygroupsoftechnicalexperimentintheconditionofchangingparameters，andinvestigates

9、theinfluenceoftechnicalparametersonweldpoolwidthandpenetrationThemodeloftheweldingprocessisdesignedintwodifferentwaysbasedonexperimentparametersandthedataoftheweldpoolwidthandpenetrationwhicharemulti—memberregressionanal