Analysis of errors in location of flaws in multipass welds using different clustering methods

Abstract

Titanium and duralumin inserts were introduced into a weld to imitate welding flaws in St3 steel samples. Analysis of the accuracy of locating the introduced flaws with different clustering methods (by digitized shape, main informative parameters, the rise rate of leading-edge envelope) has shown that these methods involve considerable time expenditures. The developed dynamic clustering has been combined with the clustering by main informative parameters and allowed in situ location of flaws during welding, while the weld has not been completed yet.