简介:Laserblankweldingisbecomingmoreandmoreimportantintheautomotiveindustryandthequalityoftheweldiscriticalforasuccessfulapplication.Afullyautomatedsolutionisrequiredtoinspectthequalityoftheblanks.ThispaperpresentsavisioninspectionsystemwithaCMOScamerawhichusesART2networktoinspectthedefectson-linetoobtainthegeometryandthequalityoftheweldseam.TheneuralnetworkART2hasthecapabilityofself-learningfromtheenvironment.Itcandistinguishthedefectsthathavebeenlearnedbeforeandgivenewoutputsfornewdefects.SoART2networkissuitableforweldqualityinspectioninlaserblankwelding.Additionally,aCO_2laserisusedfortheblankbutt-welding.
简介:Basedonalargeamountofdissectionatweldedinterfaceandquantitativemicroscopicexaminationofweldedrate,thesuitablelimitgreyscalevaluewasdetermined,andtheweldedrateofsuperplasticsolid-stateweldinginterfaceofheterogeneoussteelwassystematicallystudiedbymeansofself-madeultrasonicC-scanningimaginginspectionsystem.Theexperimentalresultsshow:theweldedstateofsuperplasticsolid-stateweldinginterfaceofheterogeneoussteelcanbeconductedtobemoreaccurately,reliablyandquicklyinspectedbymeansofthissystem,andtheultrasonictestingresultsaregoodconsistentwithactualexaminationresultsoftheinterfacedefectivedistribution.Withintheextentofthesuitbleweldedrate,theweldedratein40Cr/T10Asuperplasticweldingprocesstestedbythissystemislinearwithitstensilestrengthofjoint.