Search

The purpose of this visit was to get familiarised with the 3D weld pool surface monitoring technique developed at the University of Kentucky and extensively used during the past 15 years. Previous work focused on observation and measurement of a weld pool surface. Using the reflection law, it is possible to provide the base for the computation of the weld pool surface. Currently, at the University of Kentucky, their welding laboratory is interested in weld penetration identification topics. Using a deep learning approach, their model is capable of predicting the weld penetration status from top-side images during welding. Their current research, related to prediction of weld penetration by a neural network, does not utilise weld pool surface monitoring techniques. Instead, it is based on a passive vision system. As the model used to predict weld penetration proved to work well, the Welding Research Laboratory is seeking further improvements in the area of predicting the penetration and weld pool shape, by combining 3D weld pool surface monitoring techniques with currently developed deep learning methods. [ABSTRACT FROM AUTHOR]