Visual information acquirement and real-time control methodologies for weld pool dynamics during pulsed GTAW

This paper presents a sequential research work on visual information acquirement and real time control methodologies for weld pool dynamics during pulsed GTAW. The visual information acquirement methods are focused in computer vision sensing, image processing and characteristic extraction of the weld pool surface from the single-item pool images by particular algorithms; Based on acquired visual characteristics of weld pool and established neural network, fuzzy and knowledge models for predicting weld pool dynamics during pulsed GTAW, corresponding control methods, such as neural network, self-learning and other composite intelligent control strategies are developed to realize real time control of weld pool dynamics during pulsed GTAW.